Taxonomic revision of genus Ablattaria Reitter (Coleoptera, Silphidae) using geometric morphometrics

Abstract The genus Ablattaria Reitter, 1884 (Coleoptera: Silphidae: Silphinae) is revised. Four taxa are recognized as valid species: Ablattaria arenaria (Kraatz, 1876), Ablattaria cribrata (Ménétries, 1832), Ablattaria laevigata (Fabricius, 1775) and Ablattaria subtriangula Reitter, 1905. Ablattaria laevigata var. meridionalis Ganglbauer, 1899 is newly treated as a junior subjective synonym of Ablattaria laevigata. Lectotypes are designated for Phosphuga arenaria Kraatz, 1876, Ablattaria arenaria var. punctigera Reitter, 1884, Ablattaria arenaria var. alleoni Portevin, 1926, Silpha cribrata Ménétries, 1832, Silpha laevigata Fabricius, 1775, Silpha gibba Brullé, 1832, Ablattaria gibba var. costulata Portevin, 1926, Ablattaria gibba var. distinguenda Portevin, 1926, Ablattaria gibba var. punctata Portevin, 1926 and Ablattaria subtriangula Reitter, 1905. The distribution of all taxa is mapped, based on material examined. Geometric morphometric methods were used to evaluate shape variability in Ablattaria. Results indicated sexual dimorphism in all species. Shape inconsistency was found between the sexes of all taxa when tested independently. The first two relative warp axes indicated 65.17% shape variation in males and 65.72% in females. Canonical variate analysis separated the taxa studied. There was minimal overlap between some groups in both sexes. Differences in body shape between populations of Ablattaria laevigata from Central Europe, Italy and Greece + Turkey were also examined. Relative warps implied 58.01% shape variability on both axes in males and 64.78% in females. CVA revealed noticeable overlaps between the groups, although the Italian population demonstrated a higher separation in both sexes.


Introduction
The genus Ablattaria Reitter, 1884 (Silphidae: Silphinae) is a specialized group of gastropod predators. Distributed in the Western Palaearctic Region, these beetles inhabit forests, gardens, scrubland and generally damp localities (Portevin 1926, Heymons andLengerken 1932). Reitter (1884) erected Ablattaria as a separate genus to accommodate five taxa of carrion beetles: the widely distributed European Silpha laevigata Fabricius, 1775; Silpha gibba Brullé, 1832 from Greece: Arcadia (originally described as separate species, but treated by Reitter as a variety of A. laevigata); Silpha cribrata Ménétries, 1832 from southern Russia; Phosphuga arenaria Kraatz, 1876 from Asia Minor; as well as the newly described A. arenaria var. punctigera Reitter, 1884 from Haifa. Later, Ganglbauer (1899) described A. laevigata var. meridionalis (merely as a geographic variety occurring in a large area ranging from southern Hungary to Greece) and Reitter (1905) added A. subtriangula from Spain. Portevin (1926) in his world revision of carrion beetles treated A. gibba once more as a separate species, and added several new varieties: Ablattaria arenaria var. alleoni (no distribution provided, but the type specimen is labelled as coming from Turkey: Adana), A. gibba var. costulata (type specimen from Turkey: Istanbul), A. gibba var. distinguenda and A. gibba var. punctata (type locality not specified for either taxon). Silpha laevigata Fabricius, 1775 is the type species of Ablattaria by subsequent designation by Hatch (1928), who treated Ablattaria as a subgenus of Silpha Linnaeus, 1758. Probably the broadest review of this genus was published by Schawaller (1979), who distinguished four species: A. arenaria, A. cribrata, A. laevigata, and A. subtriangula, and formally ranked two additional taxa as subspecies of A. laevigata: A. laevigata gibba (Brullé, 1832), and A. laevigata meridionalis Ganglbauer, 1899. Schawaller provided re-descriptions of all taxa, a key to adults and a brief summary of their distributions. Recently, Nikolaev and Kozminykh (2002) treated only two taxa as full species. They regarded A. cribrata as a subspecies of A. laevigata, conditionally stated that A. arenaria should be considered also as a subspecies of A. laevigata, and formally treated A. gibba as a junior subjective synonym of A. laevigata. Most of these changes were followed in the Palaearctic catalogue by Růžička and Schneider (2004). The main morphological characters used to delimit separate species are differences in shape and surface punctation of pronotum and elytra. No consistent differences were found in the shape of male genitalia (Schawaller 1979).
Historically, most controversies have concerned the delimitations and distributions of A. laevigata, A. gibba and A. laevigata var. meridionalis (also treated at different ranks, see above). A. laevigata is a widely distributed European species (e.g., Portevin 1926, Schawaller 1979. Its distribution in Central Europe was given in detail by Horion (1949) for Germany and Austria and mentioned by Mroczkowski (1955) from southern Poland. A. gibba was originally described from southern Greece (Peloponnese Peninsula: Arcadia region) (Brullé 1832), and A. laevigata var. meridionalis was delimited as coming from "Illiria, Dalmatia, southern Hungary and Greece" (Ganglbauer 1899). However, later authors confused the distributions of the two taxa: Porta (1926) treated A. l. var. gibba from "Lombardia, Veneto, Toscana, Lazio, southern Italy" and A. l. var. meridionalis from "Corsica". Portevin (1926) reported A. gibba from Romania, Greece and Anatolia and A. l. var. meridionalis from "southern Europe". Hatch (1928) repeated Portevin's distribution data for A. gibba as "Rumania [sic], Greece, Anatolia" and added a record for A. l. var. meridionalis from "Eastern Europe". Schawaller (1979) reported A. laevigata laevigata from the south of Central Europe and from France and Spain, A. l. gibba from the Balkan Peninsula to central Anatolia, and A. l. meridionalis from Italy, including the surrounding islands.
The ecology and detailed adult and larval morphology of A. laevigata were described in detail by Heymons and Lengerken (1932). Colkesen and Sekeroglu (1989) examined the development and biology of A. arenaria adults and larvae. Further, Sekeroglu and Colkesen (1989) studied the feeding and prey preferences of A. arenaria larvae.
In this study, we revise the taxonomy of the genus. We provide new lectotype designations and synonymies based on morphological characters and using the valuable technique of geometric morphometrics on the adult beetle's body shape. These methods helped us to distinguish taxa and understand variation within and between populations. Based on the material examined, we further summarize information about the precise distribution of the taxa.

Materials and methods
Overall, 2729 specimens were examined from various European museums and collections with acronyms as follow: Types of most taxa were located and examined. Lectotypes for most taxa are designated below to fix the concept of the taxon in question and to ensure its universal and consistent application and interpretation.

Morphological analyses
Photographs of habitus and morphological details were taken using a Canon MP-E 65 mm or EF-S 60 mm macro photo lens on a Canon 550D, and several layers of focus combined in Zerene Stacker 1.04 software (Zerene Systems 2014; http://www. zerenesystems.com/cms/stacker). Exact label data of primary types were cited verbatim. Separate lines on labels are indicated by a slash "/", separate labels by double slash "//". Author's remarks and comments are enclosed in square brackets. The following abbreviations are used: p -preceding data are printed; hw -preceding data are hand-written. Interpreted label data of non-type material examined is summarized in Appendix 1. Data are available from the Dryad Digital Repository (http://doi. org/10.5061/dryad.7dn7m). To determine the coordinates of the localities, Google Earth (2014; http://earth.google.com) was used along with maplandia (http://www. maplandia.com). Distributional maps were created in ESRI ArcMap 10.2 of ArcGIS Desktop 10.2 suite. For map layers, free level 0 data from Global Administrative Areas (http://www.gadm.org) and World Shaded Relief (http://www.arcgis.com/home/item. html?id=9c5370d0b54f4de1b48a3792d7377ff2) were used.

Geometric morphometrics
Four species of the genus Ablattaria were examined: A. laevigata (145 males, 174 females), A. arenaria (85 males, 87 females), A. cribrata (49 males, 33 females) and A. subtriangula (5 males, 8 females). Moreover, three groups representing populations of A. laevigata were tested: one population from Greece and Turkey (26 males, 37 females), a population from Italy (39 males, 33 females), and one from Central Europe (Austria, Hungary and one specimen from the Czech Republic) (35 males, 33 females). Images were captured using an Olympus digital reflex camera (model E-330) connected to an Olympus stereoscopic microscope (model SZX7) and combined body length of pronotum and elytra was measured.
The geometric morphometric analysis was performed using the thin-plate spline (TPS) package; available free at http://life.bio.sunysb.edu/morph/index.html (Rohlf 2014). This technique utilizes coordinates of specific locations called landmarks that are precise points on each specimen describing the overall shape and representing the specimen's morphology (Bookstein 1982(Bookstein , 1986(Bookstein , 1989(Bookstein and 1991. In TpsDig 2.10 (Rohlf 2006) the "draw background curves" tool was employed to digitize a curve that outlined only the left half of the pronotum and the left elytron formed from 55 points. The homology of these points on all samples and their reliability in demonstrating the highest shape variability was considered (Bookstein 1991, Slice 2007. The curve points were converted into landmarks using TpsUtil 1.44 (Rohlf 2009) for further analysis.
Landmarks were then superimposed by generalized Procrustes analysis, which allows calculating variability between the taxa after aligning their landmark configurations in a specific process that ensures homology (Rohlf 1990, Rohlf and Slice 1990, Rohlf and Marcus 1993, Zelditch et al. 1995. This was conducted in TpsRelw 1.53 (Rohlf 2013). Relative warp analysis was also performed, wherein the relative warps (RWs) are transformations that express the patterns of shape variation among the specimens and visualize it using D'Arcy Thompson's transformation grids. The deformations in the grids represent the shape changes (Rohlf 1993, Richtsmeier et al. 2002, Adams et al. 2004, Zelditch et al. 2012. Multivariate analysis of variance (MANOVA) and discriminant analysis (DA) were applied on the relative warp scores matrix to test the significance of the variations between groups (taxa/sexes), and canonical variate analysis (CVA) was performed to illustrate these differences (Zelditch et al. 2004(Zelditch et al. , 2012. Graphical visualization of the CVA results was also made. All of the preceding analyses were executed in PAST ver. 2.11; freeware available for download at http://folk.uio.no/ohammer/past/ (Hammer et al. 2001).
Geometric morphometrics employs centroid size rather than linear size in calculations associated with allometry (which is the influence size has on shape) (Bookstein 1991;Klingenberg 2010, Zelditch et al. 2004. The natural logarithm of centroid size was used here, as it increases the statistical power (Viscoci and Cardini 2011). The taxon groups were first tested independently. Furthermore, multivariate analysis of covariance (MANCOVA) was used in the size correction when comparing groups to test its effect on body shape. In this analysis, the log of centroid size was used as the covariate. TpsRegr 1.38 (Rohlf 2011) was applied to calculate this influence and run permutation tests (Rohlf 1998, Viscoci and Cardini 2011, Zelditch et al. 2012.

Type species. Silpha laevigata
Diagnostic description. Body, in general, dull-black (brown to dark brown in subteneral specimens), total body length 9-19 mm.
Legs strong with fine spines, femur of hind legs broad, tibia ends with an apical spine stretching out (Figs 5,6). Tarsi with robust tarsal claws. Males with laterally expanding tarsomeres, females with cylindrical and more slender tarsomeres (e.g., as show in Figs. 4 and 5).
Phylogenetic position. Ablattaria is classified preliminarily as a sister lineage to Phosphuga Leach, 1817 and Silpha Linnaeus, 1758, based on 2.1 kB sequence of cytochrome oxidase subunits I and II (Dobler andMüller 2000, Sikes et al. 2005), sometimes treated also as a subgenus of Silpha (Sikes et al. 2005 Additional material examined. 391 specimens, see Appendix 1. Diagnostic description. Total body length 11-15 mm, body matt. Pronotum semi-elliptical, with only very superficial, very fine punctures medially on disc (which looks impunctate under lower magnification), much larger punctures more peripherally ( Fig. 12). Elytra with fine punctures that are finer in size and less close together than in A. cribrata and A. laevigata, with few intermixed larger punctures dispersed mostly toward the inner elytral margin (Fig. 16).
Remarks. Subtle differences in punctation of the scutellar shield and elytra, which differentiate the two varieties described by Reitter (1884) and Portevin (1926), fall within the intraspecific variability of A. arenaria. We confirm their status as junior subjective synonyms, as already proposed by Schawaller (1979).  (Fig. 13). Elytra more flattened; with middle-sized, more densely arranged larger punctures, very dense toward the inner elytral margin and here sometimes subquadrate in shape (Fig. 17).
Elytra more flattened. Large punctures are dispersed over the entire elytra with a higher concentration towards the inner elytral margin, which makes the elytra appear coarse. Smaller punctures are also present on both elytra, scutellum and pronotum, although they appear to be larger than those of A. laevigata but less frequent.
Remarks. Both Fabricius (1775) and Sulzer (1776) refer in their descriptions of S. laevigata and S. polita to Geoffroy (1762: 122, species #8). However, the book of Geoffroy is not consistently binominal and Opinion 1754 (ICZN 1994) placed it on the Official List of Works in Zoological Nomenclature with only some generic names available. Accordingly, the author of S. laevigata is Fabricius and the author of S. polita is Sulzer. In the syntype series of Silpha laevigata from ZMUC and BMNH, consistent with current understanding of Ablattaria laevigata, we also found intermixed a single specimen of Silpha tyrolensis Laicharting, 1781 (in ZMUC, "Kiel collection"; see above for details). This syntype specimen is here considered a paralectotype. We have designated a female from ZMUC, the "Copenhagen collection", as the lectotype to fix this name as currently used. Ablattaria laevigata is a widely distributed species with regional variation in size and shape between populations (see Geometric morphometrics section below), and also with some variability in punctation of elytra, sometimes with intermixed larger punctures or an impunctate pair of longitudinal lines present on elytra.
There are no distinct differences in the description of Silpha polita to separate it from A. laevigata, and we believe that this taxon is correctly considered as a junior subjective synonym of A. laevigata by Reitter (1884) and Schawaller (1979). In our opinion, the variation in body size, proportions and surface sculpturation which led to the description of Silpha gibba and several varieties of Ganglbauer (1899) and Portevin (1926) fall within the infrasubspecific variation of A. laevigata. We agree with Schawaller (1979), who considered Ablattaria gibba var. costulata, Ablattaria gibba var. distinguenda and Ablattaria gibba var. punctata as junior subjective synonyms of A. laevigata. Further, we consider Ablattaria laevigata var. meridionalis of Ganglbauer (1899) as a junior subjective synonym of A. laevigata.

Ablattaria subtriangula
Remarks. Additional male specimen (MNHN, coll. Marmottan), pinned, labelled: "Soto [hw] // TYPE [p, red modern label] // S. subtriangula / Reitt. / Co-type [hw, same handwriting as on identification label of lectotype specimen]" is not considered here as paralectotype, because its locality is not consistent with precise information provided in the original description by Reitter (1905). "Soto" is vague, as there seem to be more than 10 localities with this name across Spain (http://en.wikipedia. org/wiki/Soto), none of which are in either Cáceres Province or elsewhere in the Extramadura autonomous community.

Biology.
Regarding the seasonal activity of A. subtriangula, in the limited adult material examined, most specimens were collected between April and June.

Geometric morphometrics
Relative warps (RWs) of both males and females of the four Ablattaria taxa were calculated and plotted on an axis system. The first RW (RW1) axis represented 44.25% of shape variability and the second axis (RW2) accounted for 20.22%. Subsequently, discriminant analysis (DA) was applied between the sexes on the first 30 axes representing 99.94% of variability. The results indicated shape sexual dimorphism (Hotelling's test: 444.2, F: 14.071, p < 0.0001). Specimens correctly classified to their means showed a percentage of 82.59. Male groups of the four taxa were tested independently and RW1 accounted for 44.74% of the total variance whereas RW2 accounted for 20.43%. A higher 46.98% of variability was explained by the RW1 axis in females and 18.73% by RW2. Both scatter plots of the two first RWs for male and female Ablattaria displayed a high overlap between the groups of the different taxa. The thin-plate spline (TPS) transformation grids (not included in the article) indicated some shape differences between the taxa especially in A. arenaria; less rounded or curved pronotal margins posteriorly and more parallel elytra. In A. laevigata the posterior pronotal margins appeared more rounded (semielliptical) and the elytra were more robust than the other taxa particularly in the females, whereas the pronotal shape of A. subtriangula was more narrowed to the front (conical).
Multivariate analysis of variance (MANOVA) was performed on the four groups. The results indicated significant shape variations, but the separations between the groups were weak, given that the number of A. subtriangula specimens was very low compared to those of other groups. Hence, the analysis was repeated without the A. subtriangula samples to obtain a clearer separation.
Two individual canonical variate analyses (CVA) for males and females (separately) were performed to obtain separation of the four groups on the first 20 axes of the RW scores matrix. These axes covered 99.81% of the shape variation between male groups and 99.82% between female groups. Results indicated no overlap between A. arenaria and either A. laevigata or A. cribrata in males and only with one specimen in females (Fig. 34). The overlap between A. laevigata and A. cribrata was minimal and more evident in males than in females.
The jackknifed (or leave-one-out) values of the confusion matrix in A. laevigata males illustrated a correct mean classification of 131 from 144 specimens (13 showed means closer to that of A. cribrata). In A. arenaria, this was the case for 84 of 85 (1 was closer in its mean value to that of A. cribrata). A. cribrata had 42 accurate classifications of 49 in total (7 specimens were closer to A. laevigata). In the females, 150 specimens of 175 in A. laevigata were correctly classified (24 were closer to A. cribrata and 1 to A. arenaria). In A. arenaria, 82 of 87 were correctly classified (4 were closer to A. laevigata and 1 to A. cribrata). In A. cribrata, 25 of 33 were clearly classified (8 were closer to A. laevigata). These findings strongly indicate the shape variations of these taxa, and thus support the hypothesis that all three taxa constitute separate species.  Nevertheless, both males and females (independently) of A. subtriangula were tested and compared with one group formed by the three other taxa to ensure its independence by discriminant analysis (DA). Results indicated significant shape variability in males (Hotelling's test: 20.598, F: 5.0946, p < 0.001) with 86.93% correct classification of specimens to their means. For females (Hotelling's test: 40.282, F: 10.465, p < 0.0001), specimens correctly classified were 85.48%. As a result, A. subtriangula indicates its division from the other taxa and therefore may also be considered as a separate species.
To examine allometry effects, the influence of size on body shape was tested first on the four taxa by separating them into groups based on taxon and sex. The multivariate regressions of shape onto size were performed one group at a time. Results showed significant relationship in both sexes of A. laevigata, males of A. arenaria and A. subtriangula, and females of A. cribrata. The results were insignificant for female A. arenaria, A. subtriangula and male A. cribrata (Table 2).
Since allometry was significant in most taxa groups, size correction was provided by multivariate analysis of covariance (MANCOVA). This tool indicates if variation in shape is a result of size difference alone. MANCOVA was applied on male and female groups of the four taxa. Results suggested a significant interaction between body shape and body size (Table 3). Permutation tests with 1000 random permutations demonstrated a p-value of 0.00021 in males and 0.00087 in females. Considering that the percentage explained by size was 16.09% in males and 11.14% in females, some effect on the body shape variability between the taxa can be observed.
Given that A. laevigata has such a wide ranging geographical distribution, it was interesting to examine the species' body changes in various populations. Three different populations were studied: one from Greece and Turkey (Gr. & Tr.), a population from Italy (It.), and a population from Central Europe (CE) (geographic origins of examined specimens are summarized in Fig. 33). Relative warps were calculated in male and female populations separately and plotted on an axis system. The RW1 axis of males corresponded to 39.02% and RW2 to 18.99% of shape variability. In females, RW1 indicated 43.87% of shape variation and RW2 indicated 20.91%. TPS transformation grids (not included in the article) showed little shape variability; the elytra appeared, in general, more parallel in the Greek and Turkish populations. Populations from Italy had a more arched elytra and the pronotum was slightly broader (Figs 7 and 8).
MANOVA was performed subsequently. Male populations revealed significant shape dissimilarity (F = 10.35; Wilk's lambda = 0.121; DF = 30/166; p < 0.00001). Shape variability was found to be also significant in the female populations (F = 8.337; Wilk's lambda = 0.166; DF = 30/172; p < 0.00001). Canonical variate analysis on the first 15 axes was performed and represented 99.54% of the shape variation in males and 99.68% in females. Results indicated overlap between all groups (Fig. 35). The jackknifed values of the confusion matrix for both sexes are presented in Tables 4 and 5. The most obvious separation was seen in the Italian population, which showed incorrect classification of only 7 specimens in the two sexes taken together of a total 72 specimens. The Central European population showed higher variation in the male than was that in the female populations from Greece and Turkey.
In order to determine whether allometry played a role in this categorization, and even though the sample was too small, regression results indicated significant relationship between size and shape in both sexes (Table 6). Despite the fact that size explained a low percentage of the body shape (11.49% in males and 9.91% in females), its effect cannot be denied and the influence of allometry can be noted.
The linear size (body length of pronotum and elytra) of both sexes was measured and plotted in a simple boxplot (Fig. 36). In general, females were larger than males. Even though body length differences did not appear to be very marked, the smallest of the three groups was the population from Central Europe, particularly the males, whereas the females were only slightly smaller than the Greek and Turkish females. The largest specimens measured here were those from Italy.

Discussion
Geometric morphometric techniques were applied to this genus for the first time. Based on the results we obtained and the observed morphological traits, A. cribrata can be considered a separate species. This agrees with Schawaller (1979) but not with Nikolaev and Kozminykh (2002), who treated A. cribrata as a subspecies of A. laevigata.
A. laevigata has a vast distribution over the Western Palearctic region. Hence, both its shape and size vary greatly across its range. Applying the geometric morphometric techniques to various populations indeed confirmed these shape variations (see above). These results, together with further morphological examination of other populations, revealed that beetles from the Mediterranean region (specimens from  Croatia, Italy, Greece and Turkey) tend to be larger in body size than those from Central Europe. The geographical separation of the identified species can be observed most particularly in A. subtriangula, which is endemic to Spain and is syntopic only with A. laevigata. A. subtriangula is absent from the Balearic Islands, where only A. laevigata is found. This is in agreement with Piloña et al. (2002). A. cribrata is widespread throughout Iran (in contrast to data in Portevin (1926) and Schawaller (1979)) and is separated from A. arenaria at the eastern Iraqi -western Iranian borders, where only one specimen of A. arenaria has been noted (although two specimens of A. arenaria were recorded from Iran: Huzestan prov. by Růžička and Schneider (2002)). A. laevigata overlaps in its occurrence with A. arenaria mainly in eastern and more sparsely in northern Turkey. A. arenaria does not appear to cross into continental Europe. A. laevigata and A. cribrata partially overlap in Georgia, Armenia and southern Russia (specifically in Chechnya). The genus does not seem to extend beyond southern Russia and Iran to Kazakhstan or Afghanistan. It is rather scarce in Turkmenistan, where it is known only in the southwestern part (as also mentioned by Nikolaev and Kozminykh (2002)). Moreover, there are no known records of Ablattaria from northern Africa or from the Arabian Peninsula. Schawaller (1979) stated that the genus only rarely occurs at higher elevations. We examined about 52 specimens from localities above 2000 m (140 specimens from localities above 1000 m). Some of the records of A. laevigata from high elevations were from Italy: Molise Reg., Majella Mt. at 2793 m. Also, A. arenaria was cited from Syria: Nur Mts. (Amanus) at 2240 m, and Israel: Mount Hermon at 2000 m. Schawaller (1979) also speculated whether the genus could be found in the higher Pyrenees and Alps. Two specimens of A. laevigata were recorded from the French side of the Pyrenees (Languedoc-Roussillon Reg.: Lac d'Estom and Arles sur Tech), and two specimens from the Spanish side (Catalonia: La Jonquera and Espot). Six specimens of A. laevigata were recorded from localities higher than 1000 m in the Alps, the highest being from Provence-Alpes-Côte d'Azur Reg.: Vaucluse Dept., Mont Ventoux, at an altitude of ca. 1900 m.