Hidden species within the genus Ocys Stephens: the widespread species O. harpaloides (Audinet-Serville) and O. tachysoides (Antoine) (Coleoptera, Carabidae, Bembidiini)

Beetles previously considered to be Ocys harpaloides (Audinet-Serville) from northern Africa, Spain, France, the United Kingdom, France, and Belgium belong to two species. These species can be distinguished using DNA sequences of 28S rDNA, 18S rDNA, COI, CAD, and topoisomerase I. A key, diagnoses, and images are provided to enable identification of specimens based upon characteristics of male and female genitalia, as well as microsculpture and other external structures. Through examination of the holotype of Bembidium harpaloides v. tachysoides Antoine, as well as designation of lectotypes of Bembidion harpaloides Audinet-Serville and Ocys melanocephalus Stephens, and designation of a neotype for Tachis rufescens Guérin-Ménéville, the valid names of the two species were determined to be Ocys harpaloides and Ocys tachysoides (Antoine).


Introduction
Among European carabid beetles of the tribe Bembidiini Stephens, Ocys harpaloides (Audinet-Serville, 1821) is one of the more distinctive, looking more similar to members of other tribes (for example, the trechine Trechus Clairville, 1806 or the tachyine Porotachys Netolitzky, 1914) than to other bembidiines. As currently circumscribed, Ocys harpaloides is a common and widespread Mediterranean-Atlantic species, ranging from the Azores, Algeria and Morocco north to southern Norway, Britain and Ireland, and east to Italy, the Balkans and Greece (Lindroth 1992).
The only congener with which O. harpaloides is documented to coexist throughout much of this range is Ocys quinquestriatus (Gyllenhal), a species from which it can easily be distinguished by numerous morphological characters (Lindroth 1974;see below). There are more than 20 additional species of Ocys Stephens; almost all of these are rare, and restricted to small geographic areas (e.g., Giachino and Vailati 2012;Magrini and Degiovanni 2009;Neri et al. 2011;Toledano and Wrase 2016;Vigna Taglianti 1994). Many of these distinctive, localized endemics occur at higher elevations in mountains in Spain, northern Africa, or Italy (Magrini and Degiovanni 2009;Neri, et al. 2011;Netolitzky 1942), or on islands (Magrini et al. 1998). However, one species similar to O. harpaloides, Ocys tachysoides (Antoine, 1933), has been reported from low elevation habitats in Morocco (Antoine 1933;1955), Spain (Toribio 2013), and a montane region of Portugal (Sciaky 1998).
In Northern Ireland, two morphologically distinct forms of "Ocys harpaloides", correlated with habitat, have been reported (Anderson et al. 2000): specimens found among coastal rocks (Fig. 1A) are paler and larger ( Fig. 2A), and those found in woodlands (Fig. 1B) are darker and smaller ( Fig. 2B) (Anderson et al. 2000). Although this pattern might be a result of ecophenotypic variation, the observation suggested the possibility that two species could be hiding within "Ocys harpaloides". However, initial morphological investigations did not reveal additional, definitive differences (Anderson et al. 2000).
In 2011 we began an examination of the DNA sequences of these forms, and more detailed study of morphological variation. This revealed that two species were indeed present in Northern Ireland, and elsewhere in the range of "Ocys harpaloides". As this paper was being prepared for publication, Raupach et al. (2016) reported the existence of two forms of "Ocys harpaloides" in Germany and France based upon the cytochrome oxidase I gene; these two forms correspond to the two species detailed here. In this paper, we document the differences between the two species, and through examination of types of the available names, show that the valid name of the second species is Ocys tachysoides (Antoine), which is much more widespread than previously believed.

Methods
Members of Ocys were examined from the collections listed below. Each collection's listing begins with the coden used in the text. GCR: gonocoxite ratio = gonocoxite 2 length / gonocoxite 1 length (Fig. 4) Measurements were made either using Microvision's Cartograph software processing images from a JVC KY-F75U camera on a Leica Z6 lens, or on a Leica M3Z stereomicroscope with a Leitz graticule.
The density of the microsculpture lines on the left elytron was measured by counting the number of microsculpture lines that cross a 0.1 mm longitudinal line placed in the centre of the third elytral interval approximately 1/3 of the distance between the front and back of the elytron.
Taxon sampling for DNA studies. We obtained DNA sequence data for 15 Ocys specimens (Table 1), and combined these with results from the two Ocys specimens sequenced in Maddison (2012). For outgroups, we chose two or more species representing different lineages within each of the five largest genera of Bembidiina (Bembidion Latreille, Asaphidion Chaudoir, Amerizus des Gozis, Sinechostictus Motschulsky, and Lionepha Casey (Maddison, 2012)). The outgroup species are listed in Table 2.
For all except specimen number 4606, DNA was extracted from specimens preserved in 95% ethanol using a Qiagen DNeasy Blood and Tissue Kit. Fragments for the seven genes were amplified using the Polymerase Chain Reaction on an Eppendorf Mastercycler Pro Thermal Cycler, using TaKaRa Ex Taq and the basic protocols recommended by the manufacturers. Primers and details of the cycling reactions used are given in Maddison (2012). The amplified products were then purified, quantified, and sequenced at the University of Arizona's Genomic  Table 2. Sampling of Bembidiina species. Four-digit numbers in the "#" column entries for Ocys are D.R. Maddison voucher numbers for sequenced specimens. Data for all non-Ocys specimens, as well as to specimens 1077 and 0569, were obtained from Gen-Bank from previous studies (Maddison 2008;2012;Maddison et al. 1999;Maddison and Ober 2011;Maddison and Swanson 2010;Wild and Maddison 2008), and more data about the specimens are presented in those papers. Locality details for Ocys specimens are given in Table 1 and in Maddison (2012 and Technology Core Facility using a 3730 XL Applied Biosystems automatic sequencer. Assembly of multiple chromatograms for each gene fragment and initial base calls were made with Phred (Green and Ewing 2002) and Phrap (Green 1999) as orchestrated by Mesquite's Chromaseq package (Maddison and Maddison 2014;Maddison and Maddison 2015b) with subsequent modifications by Chromaseq and manual inspection. Multiple peaks at a single position in multiple reads were coded using IUPAC ambiguity codes. For specimen 4606, RNA was extracted from an RNA-Later preserved specimen using a Qiagen RNeasy Kit and Trizol (Life Technologies). Isolation of mRNA was done with an NEBNext® Poly(A) mRNA Magnetic Isolation Module (New England BioLabs). The library was prepared using an NEBNext Ultra RNA (New England BioLabs) kit using 1000 ng of input RNA and NEB-Next Multiplex Oligos for Illumina (New England Bio-Labs). The finished library was quantified using a Qubit Florometer (Life Technologies) and fragment distribution characterized on a 2100 Bioanalyzer (Agilent Technologies). The library was then sequenced on an Illumina 2500 at the Oregon Health Sciences University's Massively Parallel Sequencing Shared Resource, with reads demultiplexed using CASAVA v1.8. The 100-base, paired-end run produced 36 million reads, which were assembled using CLC Genomics Workbench version 8.5, using the de novo assembly feature with default parameter values. The resulting assembly was converted to a blastable database using NCBI's makeblastdb tool. Using the Asaphidion yukonense sequences from Maddison (2012) as query sequences, the specimen 4606 transcriptome was searched using nBLAST, yielding single contig sequences for 28S, 18S, and Topo. For CAD, three overlapping contigs were found; they were identical in the overlapping regions, and were merged into a single sequence. For COI, three contigs were found, one containing full-length COI; the other two short fragments could be eliminated from consideration because they BLASTed to non-bembidiines in Gen-Bank or contained stop codons.
Alignment and phylogenetic analysis. The appropriate alignment was obvious for COI, CAD, and Topo, as there were no insertion or deletions (indels) evident in the sampled sequences. The two ribosomal genes were aligned by MAFFT version 7.130b (Katoh and Standley 2013), using the L-INS-i search option and otherwise default parameter values. No data were excluded from analyses, other than trimming of ends of sequences that were notably longer than others.
Models of nucleotide evolution where chosen with the aid of jModelTest version 2.1.7 (Darriba et al. 2012;Guindon and Gascuel 2003). Among the models supported by GARLI, the models chosen by the Bayesian Information Criterion were 28S: TVMef+G, 18S: SYM+I+G, COI: TIM1+I+G, CAD: TVMef+I+G, and Topo: TrN+G. Likelihood analyses of nucleotide data were conducted using GARLI version 2.0 (Zwickl 2006), as orchestrated by the Zephyr package of Mesquite (Maddison and Maddison 2015a). Twenty-five search replicates were employed for maximum likelihood trees, and 200 replicates for bootstrap analyses.

Results
There are two distinct clades in each of the five gene trees ( Fig. 5): one clade contains all of the larger, paler specimens, and another contains all of the smaller, darker specimens. As outlined below in the taxonomic section, the valid name for the larger, paler specimens is Ocys harpaloides (Audinet-Serville), and the valid name for the smaller, darker specimens is Ocys tachysoides (Antoine). Based upon the arguments presented below in the Discussion, we consider these different species, and for simplicity's sake we will speak of them as different species in this section, and use their valid names. Likelihood bootstrap support for these two species is strong (Table 3), being above 90% in at least two of the well-sampled genes.
In COI, there are 11 sites at which three to six of the O. tachysoides (all specimens except 2899) show a secondary peak in chromatograms; there are three different sites at which 1 to 3 O. harpaloides (2759,2760,2853) have a secondary peak. The secondary peaks suggest an alternative version of COI, which is presumably a nuclear copy, or "numt" (Thalmann et al. 2004), thus causing some uncertainty about the source of our sequenced COI. However, in spite of the uncertainty, the gene tree clearly supports two distinct species (Fig. 5).
A detailed examination of differences between and within each of these two species also reveals many sites at which they differ (Table 4). There are six fixed amino-acid differences between the two species across the three protein-coding genes.
Morphological characteristics were in general correlated with DNA sequences. All five males in the Ocys harpaloides clade in Fig. 5 had genitalia matching those shown in Fig. 6A, B, and all seven males in the Ocys tachysoides clade had genitalia matching Figs 6C, D (Table 5). Irish populations also had the colour of the elytral disc correlated with the gene tree clades, but the correlation was not as clear in Belgium (Table 5).
On the island of Ireland, all three O. harpaloides clade males were found on the ocean shore, at two localities, whereas all five O. tachysoides clade males were found in woodland habitats, at three localities (Table 5). That distinction was not as evident on the mainland of Europe, as two of the O. harpaloides specimens from Belgium were found in a broadleaf forest, inland (Table 5).   Table 5. Morphological features and habitats of the sequenced specimens. "Disc colour" refers to the colour of the elytral disc relative to the apex and lateral margins. Male genitalic type A is that shown in Fig. 6A, B, and genitalic type C is that shown in Fig. 5C

Discussion
Our data show that specimens in northern Europe referred to "Ocys harpaloides" belong to two species. The DNA data alone strongly suggests that two species live in Northern Ireland and Belgium: the gene trees for two certainly unlinked genes (nuclear 28S and mitochondrial COI) show the same clades for specimens that broadly overlap in geographic range, with specimens of both forms sampled from each region. This pattern, combined with identical patterns for 18S, CAD, and Topo, indicates that the two forms are not currently exchanging genes. This is confirmed by the correlated morphological traits, in particular the distinctive male genitalia. Toribio (2013) recognized these two species in populations in southern Spain and Morocco, but did not recognize that O. tachysoides was as widespread as O. harpaloides.
By comparing our sequences with those presented by Raupach et al. (2016), we have confirmed that the two COI forms they documented correspond to O. harpaloides (their French specimens) and O. tachysoides (their German specimens).

Taxonomic section
The following key allows one to identify Ocys specimens from much of Europe, outside of those regions containing localized endemics (that is, outside of higher elevations in mountains in Spain, northern Africa, or Italy (Magrini and Degiovanni 2009;Neri et al. 2011;Netolitzky 1942), on islands (Magrini et al. 1998). Elytra rufous or brown, in most specimens paler centrally with darker apical and lateral margins. Microsculpture of elytra consisting of more closely spaced, transverse lines with less of a tendency to form meshes (Fig. 7A, B). Aedeagus with ventral margin in most specimens bent downward toward the thinner apex; anterior sclerites of the internal sac more rounded; with a brush sclerite of normal size (Fig. 6A, B). Spermatheca more curved (Fig. 9A, B)  -Elytra dark brown to black in colour, in some specimens paler along the suture, but entire discal region not paler than margins. Microsculpture of elytra consisting of less closely spaced, transverse lines with more of a tendency to form meshes (Fig. 7C, D). Aedeagus with ventral margin straighter, and apex wider, with a blunter tip; anterior sclerites of the internal sac more angulate; with a very small brush sclerite (Fig. 6C, D). Spermatheca straighter (Fig. 9C, D)   Dufour's collection. The four pins contain a total of eight specimens: the first two pins contain one specimen each, the third pin houses five specimens, each on a separate card, and the last pin houses a single specimen. The last specimen can be excluded from consideration, as its label contains a locality ("Carcassonne", in southern France) that differs from the locality stated by Audinet-Serville. The other three pins each have one or two labels: the first specimen has a label that apparently contains the number 249; the second the number 5244, and the remaining pin has two labels, one with a small amount of indecipherable text, the other with the number 124. All of the Ocys in Dufour's collection belong to this species, as determined by colour pattern and microsculpture. In addition, the first two are males, and dissection revealed genitalia that match genitalic type A (i.e., as in Fig. 6A, B). The labels on the three pins are in different handwriting, and the pins themselves differ in structure. Thus, it appears as if the three came from different sources, and likely only one of them represents authentic Audinet-Serville material. According to Antoine Mantilleri (Thierry Deuve, pers. comm.), the handwriting on the first specimen is likely that of Dufour. This suggests that the first specimen is not an authentic Audinet-Serville specimen, but rather an original Dufour specimen; however, it is possible that Dufour added a label to the specimen after having acquired Audinet-Serville's collection. The writer of the other labels is not known. We could discover no information about Audinet-Serville's handwriting, other than a single example of his signature, and thus we do not know if any of the labels might have been written by him, or whether they may have been added by later researchers. Thus, it is not clear which specimen on the three candidate pins is Audinet-Serville's, and it likely will never be determined with certainty. We consider the value in stabilizing the name more important than the uncertainty of the choice. As all of these specimens belong to the same species, as the specimen on the second pin is a male in sufficient condition and with undamaged genitalia, and as there is hint that the specimen on the first pin is not an Audinet-Serville specimen, we have chosen the second specimen as the lectotype of Bembidion harpaloides Audinet-Serville. The collection of Félix Édouard Guérin-Méneville was incorporated into the Chaudoir collection at the MNHN (Cambefort 2006) in the Chaudoir magasin. David Kavanaugh and Azadeh Taghavian searched for types of Tachis rufescens Guérin-Ménéville in the Chaudoir magasin. They found numerous Ocys, but every one could be eliminated as Guérin-Ménéville specimens through consideration of the labels, as all specimens had a mechanically printed label (against other Guérin-Ménéville specimens), and the labels specified a locality, collector, or collection that did not coincide  Figure 5. Phylogenetic trees of the five studied genes. Outgroups not shown. Branch length is shown proportional to relative divergence, as estimated by GARLI. The larger, paler specimens with genitalia as in Fig. 6A and B are shown in orange, and the smaller, darker specimens with genitalia as in Fig. 6C and D are shown in brown.
with those of the Ocys examined by Guérin-Ménéville (Kavanaugh, pers. comm.). The Guérin-Ménéville specimens of Ocys are thus lost. In the interests of stabilizing the nomenclature, we sought a neotype for Tachis rufescens. Within the MNHN, there are specimens from three localities around Paris (Vaires, Chelles, and Chalifert). All of these specimens belong to the following species, Ocys tachysoides (Thierry Deuve, pers. comm.); we have confirmed this for the specimens from Vaires by examination of male genitalia. However, the lack of Ocys harpaloides in the twentieth-century specimens in the MNHN does not necessarily mean that the species was absent from Paris in 1823: in the 123 years that passed between the description of Tachis rufescens and the collection of the Viares specimens in 1946, the habitats around Paris may have changed. The possibility that Ocys harpaloides lived in Paris in the 1820s is made more likely by the presence, in the collection of Mr. Dominique Echaroux, of two series of specimens of Ocys harpaloides from just outside of Paris: one specimen from Bouray-sur-Juine and seven from Juvisy-sur-Orge. The latter is only 18km SE of the type locality of Tachis rufescens. Thus, both species have lived in the general Paris area, and a specimen of either species could be designated as a neotype. However, designation of the Vaires specimen, for example, would lead to Ocys rufescens as the name of the following species, with O. tachysoides as a junior synonym. This would change the name used for the following species in the Iberian Peninsula and Africa, where it is known as O. tachysoides (per Toribio 2013). In the interests of sta-bility, we choose as neotype of Tachis rufescens one of the Juvisy-sur-Orge males; with this designation, Tachis rufescens is maintained as synonym of O. harpaloides.
In the BMNH, under "melanocephalum", are seven specimens (four males, three females) in the J. Stephens collection, all with the label "British Isles J. Stephens Coll. BM. 1853 -46". We consider this to be the type series of Ocys melanocephalus Stephens. Three of the males in this series were dissected, and genitalia examined; they all match genitalic type A (i.e., as in Fig. 6A, D). We have chosen one of these specimens as the lectotype.
In the Wollaston collection of the BMNH, in Insecta Maderensia, Drawer 2, is a single teneral female, without a determination label. This specimen is presumed to be the holotype of Bembidium dubium Wollaston, 1854. Its colour suggests that it belongs to this species. However, the species membership of the type is of no nomenclatural importance, as the name is a junior primary homonym of Bembidium dubium Heer, 1838, itself a junior synonym of Bemidion assimile (Gyllenhal, 1810).
Diagnosis. Body length 4.2-6.1 mm (Toribio 2013 and our observations; average length of males 5.0 mm (n=5), of females 5.51 mm (n=5)). Head and pronotum a clear orange-brown; anterior and central part of elytra of the same colour, but sides and posterior region of elytra dark brown (Fig. 3), and with the epipleural gutter and suture yellow-brown. Microsculpture of elytral disc more transverse, with more close-set parallel lines and fewer meshes (Fig. 7A, B); density of microsculpture lines 25-26 per 0.1 mm (n=4 males). Hind margin of pronotum in most specimens straighter than in O. tachysoides, such that the hind margin is directed more or less laterally near the hind angles (Fig. 10A), occasionally with a slight emargination laterally (Fig.  10B). Elytra more parallel-sided, greatest width behind middle. Basal margin at shoulder slightly arcuate with a forward-directed concavity (Fig. 11A, B). Elytral striae 2 through 4 less marked in the apical third. Aedeagus with ventral margin bend slightly downward toward apex (Fig. 6A, B); apex more rounded. Anterior edge of central sclerite complex of internal sac more rounded; brush sclerite larger; dorsal membranes of internal sac darker. Gonocoxite relatively long (Fig. 8A), GCR=0.64-0.69 (average 0.67, n=5); spermathecal margin (opposite the efferent duct of the spermathecal gland) curved (Fig. 8A, B; n=5), with tip pointed toward duct of gland.
Geographic distribution (Fig. 12). In Africa, from Morocco, Algeria, and Tunisia. In Europe from Spain, France, Belgium, Italy, Ireland, and the United Kingdom. Examination of specimens in additional collections will likely show it to be more widely distributed.   Specimens examined. In addition to type specimens, and those listed in Table 1, (n=7)). Head and pronotum a deep redbrown contrasting with the dark brown to blackish elytra (Fig. 3), only the epipleural gutters and sometimes sutures paler, light reddish-brown. Occasionally with a bluish lustre. Microsculpture of elytral disc less transverse, with more of tendency to form sculpticells (Fig. 7C, D); density of microsculpture lines 20-21 per 0.1 mm (n=4 males). Hind margin of prothorax with a slight emargination laterally, such that the hind margin at the hind angles is directed posteriorly (Fig. 10C, D). Elytra less parallelsided, greatest width around middle. Basal margin at shoulder in most specimens more or less straight (Fig. 11C), relatively few specimens slightly arcuate with a forwarddirected concavity (Fig. 11D). Elytral striae 2 through 4 more marked in the apical third. Aedeagus with ventral margin bend more or less straight (Fig. 6C, D); apex less rounded, blunt. Anterior edge of central sclerite complex of internal sac angulate; brush sclerite smaller; dorsal membranes of internal sac paler, less evident. Gonocoxite relatively short (Fig. 8B), GCR=0.55-0.60 (average 0.58, n=5); spermathecal margin opposite the efferent duct of the spermathecal gland more or less straight (Fig. 8C,D; n=5).
Geographic distribution (Fig. 12). In Africa, known from Morocco (Antoine 1955). In Europe, from Portugal (Sciaky 1998), Spain (Toribio 2013), France, Belgium, Germany, and the United Kingdom. Examination of specimens in additional collections will likely show it to be more widely distributed.
Specimens examined. In addition to the type specimen, and those listed in Table 1, we examined specimens from France: Vaires (MNHN). Only records of which we are confident are included. In addition to specimens we have examined, and those from Toribio (2013), Antoine (1956), and Sciaky (2009), we have included three publically available records from the Barcode of Life Database (http://www.boldsystems.org), corresponding to BOLD records GBCOU8213-14, GBCOU8667-14, and FBCOH116-12, as well as specimens identified by Luca Toledano and Paolo Bonavita in their collections (CTVR and PB, respectively) based upon images and diagnoses we provided to them. we puzzled through the best approach to resolving the nomenclatural issues, were invaluable. Azadeh Taghavian's help in this regard can also not go unmentioned. In addition, David Kavanaugh spent many valuable hours while in Paris personally looking for Ocys types, and we could have not completed the project without his help; we are also thankful for his advice as we considered the best path, and for hand-carrying the Audinet-Serville material back from Paris. Several others aided these efforts as well, and we wish to thank them: Antoine Mantilleri (for his advice about Audinet-Serville material), and Paul Johnson (for providing copies of pages from Horn & Kahle). Finally, we wish to thank Mr. Dominique Echaroux, from Etréchy (Essonne, France), for providing specimens of Ocys harpaloides from his collection, and for allowing us to designate one of them as a neotype of Tachis rufescens.
We are also very thankful for Beulah Garner's stewardship of the specimens at the Natural History Museum in London, and for sending a portion of the type series of Ocys melanocephalus Stephens so that we might designate a lectotype.
For various helpful discussions we thank Marcos Toribio, Luca Toledano, Paolo Bonavita, and Paolo Neri. For their thoughtful reviews of the manuscript, we thank Michael Raupach, Pier Mauro Giachino, and Joachim Schmidt. We also thank Luca Toledano and Paolo Bonavita for examining the specimens in their collections and providing additional localities for Fig. 12.
We thank all those who provided material preserved for DNA studies, including Joachim Schmidt, Frederik Hendrickx, and Wayne Maddison, and in particular the late Konjev Desender. We also thank John S. Sproul for extracting the RNA of specimen DNA4606, James M. Pflug for preparing the sequencing library, and Illumina Inc. for helping resolve some sequencing issues.
This work was funded in part by the Harold E. and Leona M. Rice Endowment Fund at Oregon State University.