We describe a new stiletto snake, Atractaspis, from western Liberia and southeastern Guinea. The new species shares with morphologically similar western African Atractaspis species, A. reticulata and A. corpulenta, the fusion of the 2^nd infralabial with the inframaxillary. From A. corpulenta the new species differs by a more slender body (276–288 ventrals and 19 or 20 dorsal scale rows versus 178–208 ventrals with 23–29 dorsal scale rows), a divided anal plate and divided subcaudal scales (both non-divided in A. corpulenta). The new species differs from most A. reticulata by having 19 or 20 dorsal scale rows at midbody (versus 21–23, rarely 19), and a lower ventral count (276–288 versus 304–370). The new species thus has a relatively longer tail: snout-vent-length / tail-length in the female holotype (15.7) and paratype (21.5) versus a mean of 23.6 in seven female A. reticulata. The new Atractaspis likely is endemic to the western part of the Upper Guinea forest zone and thus adds to the uniqueness of this diverse and threatened biogeographic region.

Biodiversity hotspot, biogeography, rainforest, Reptilia , Squamata , species delimitation, Upper Guinea forest

The stiletto snakes or burrowing asps, genus Atractaspis Smith, 1849, currently comprise 22 (Wallach et al. 2014), or 21 (Uetz et al. 2018) valid species. Most species are restricted to sub-Saharan Africa where they occur in a wide range of habitats from semi-deserts to rainforests (Spawls and Branch 1995); only two occur in the Middle East and Arabia (Wallach et al. 2014; Grossmann et al. 2018). These fossorial and venomous snakes are famous for their unique skull anatomy and venom delivery system, enabling them to use a single fang to bite, with closed mouth, in a lateral stabbing movement (Broadley 1990; Greene 1997; Cundall and Irish 2008; McDowell 2008). Various herpetologists have already been suffering from this behavior, making it impossible to hold the snakes in the usual way behind the head without being bitten (see Wagner et al. 2009). Their systematic position was matter of a constant debate and the snakes have been placed for instance within Viperidae, within Aparallactinae as subfamily of Colubridae, as a separate family Atractaspididae, and most recently as a subfamily Atractaspidinae within the Lamprophiidae (Broadley 1990; Vidal et al. 2007; Wallach et al. 2014; Uetz et al. 2018; Portillo et al. 2018).

From West and western Central Africa 11 Atractaspis species are known: A. aterrima Günther, 1863, A. boulengeri Mocquard, 1897, A. coalescens Perret, 1960, A. congica Peters, 1877, A. corpulenta (Hallowell, 1854), A. dahomeyensis Bocage, 1887, A. irregularis (Reinhardt, 1843), A. microlepidota Günther, 1866, A. micropholis Günther, 1872, A. reticulata Sjöstedt, 1896, and A. watsoni Boulenger, 1908 (Chippaux 2001; Chirio and LeBreton 2007). On recent surveys in north-western Liberia and south-eastern Guinea, we collected stiletto snakes deviating from all other known Atractaspis species. These snakes are described here as a species new to science. In addition, we redescribe the holotype of A. reticulata, the species which is morphologically most similar to our new species, and summarize its known distribution.

The holotype was euthanized by smearing a benzocaine cream into its mouth. The paratype and one additional specimen were found dead. From the holotype and paratype we collected tissue samples, which were preserved in 97% ethanol. The snakes were preserved in 75% ethanol and are inventoried in the herpetological collections of the Museum für Naturkunde, Berlin, Germany (ZMB; holotype), the Museo di Storia naturale del Salento, Calimera, Italy (MSNS; paratype), or will be inventoried at the Institut Royal des Sciences Naturelles de Belgique, Brussels, Belgium (IRSNB; additional specimen, see below). Measurements and assessment of morphological characters of the new species and comparative material, including its pholidosis, has been done by one person (CK). Ct-scanning and measurements of skull bones have been compiled by KM. Dorsal scale rows were counted at three points along the trunk, i.e. at one head length posterior to the end of the head, at midbody (at half of the snout-vent length), and at one head length anterior to the anal scale. Dorsal scale row reduction formulae were based on Dowling (1951a), the formulae for the supracaudal scales are analogous to this procedure. We added a ‘x’ to the formulae if no fusion or reduction takes place. Ventral counts are according to Dowling (1951b). Preventrals are the scales anterior to the ventrals, characterized as being broader than long. The terminal scale was not included in the subcaudal count and is given as ‘+1’. Values for symmetric characters are given as left/right. Measurements of snout-vent length and tail length were rounded to the nearest millimeter, all other to the nearest 0.1 millimeter. Sex was determined by a small incision at the base of the tail. Comparative measures have been compiled from the literature, material housed at ZMB and the holotype of A. reticulata from the Naturhistoriska Rijkmuseet, Section for Vertebrate Zoology, Stockholm, Sweden (NRM 1796 collected in “Kamerun, Ekundu” by Sjöstedt; Tables 1, 2). Wallach et al. (2014) and Uetz et al. (2018) erroneously cited the Zoological Museum of the University of Uppsåla, Sweden (ZMUU), as the repository of the holotype of A. reticulata.

The heads of the holotype and paratype of the new species, the A. reticulata holotype and further comparative material were subjected to micro-tomographic analysis at the Museum für Naturkunde Berlin, using a Phoenix nanotom X-ray|s tube at 80–100 kV and 100–230 μA, generating 1000–1440 projections with 750 ms per scan. The different kV-settings depended on the respective specimen size. Effective voxel size, i.e. resolution in three-dimensional space, ranged from 5.71–15.67 μm. The cone beam reconstruction was performed using the phoenix|x-ray datos|x version 2.2 software (GE Sensing and Inspection Technologies GMBH) and the data were visualized in VG Studio Max, version 3.1.

We sequenced 509 bp of the 16S ribosomal RNA of the types of the new species, following the procedures and using the primers as described in Portillo et al. (2018). We compared the two sequences to each other and to the sequences of the other Atractaspis species from which 16S have been published (Portillo et al. 2018): Atractaspis boulengeri (IPMB J355; GenBank AY611833), A. corpulenta (IPMB J369; GenBank AY611837), A. irregularis (UTEP 21655; GenBank MG746901) and A. micropholis (IPMB J283; GenBank AY611823). We also received an unpublished 16S sequence from A. reticulata heterochilus (UTEP 21664; Democratic Republic of the Congo, Tshopo Province: road between Nia Nia and Kisangani; Portillo et al. submitted) for comparison. The sequences of the new species have been deposited at GenBank.

The phylogenetic relationships within the genus Atractaspis are still unclear. Two recent contributions included only some species of Atractapis (Underwood and Kochva 1993; Moyer and Jackson 2011). The study by Underwood and Kochva (1993) deviates in its phylogenetic conclusions partly from Laurent’s (1950) species grouping. However, for morphological comparisons between species Laurent’s (1950) paper is still the most useful. Our new species morphologically falls within Laurent’s (1950) section “D” (reticulata group) of the genus. Together with A. reticulata and A. corpulenta it shares the fusion of the 2^nd infralabial with the inframaxillary. This character likewise distinguishes these two species and A. branchi sp. n. from all other congeners.

Atractaspis corpulenta is a comparatively very robust, heavily built snake (only 178–208 ventrals with 23–29 dorsal scale rows at midbody vs 276–288 ventrals and 19 or 20 dorsal scale rows in A. branchi) and has a non-divided anal plate and subcaudal scales (both divided in A. branchi). Its West African rainforest subspecies, A. corpulenta leucura, has a white tail tip (see fig. 3 in Rödel and Mahsberg 2000), which is lacking in A. branchi. Atractaspis branchi thus can be only confused with A. reticulata, from which it differs by a lower ventral count while having simultaneously a higher number of subcaudals and, thus, a relatively longer tail: snout-vent length / tail length in A. branchi: 15.7 and 21.5; in A. reticulata: mean = 22.3, minimum = 16.2, maximum = 29.2 (N = 15). The two A. reticulata with the comparatively longest tails (16.2 and 17.6) originate from Gabon and Angola, respectively, and are both males. The seven A. reticulata females, for which we could calculate a snout-vent length / tail length ratio, had a mean of 23.6 (Tables 1, 2).

The new species has 19 dorsal scales rows at midbody (the paratype has mostly 19 rows but rarely 20 around midbody) and thus differs from almost all A. reticulata vouchers investigated by us and reported in the literature by having 21–23 rows (Table 1). However, the holotype of A. reticulata reticulata has only 19 scale rows and mid-body scale rows. thus numbers of scales at mid-body is not diagnostic for these two taxa. Apart from ventral and subcaudal scale counts, differences in skull anatomy between the new species and A. reticulata are a shorter frontale and a greater number of palatine teeth in the new species. An additional diagnostic character might be larger eyes in A. branchi compared to A. reticulata (Table 2). However, the holotype is a young female and relative eye size seems to diminish with body size, as indicated by the two types and our measurements of other differently sized Atractaspis, including variously sized A. reticulata (Table 2). The types of the new species indicate other potential ontogenetic changes in characters, i.e. body colour changing from brown to grey, fang length increasing with body length, and rows of the dorsal body scales changing from perfectly straight in the smaller holotype to oblique in the larger, and presumably older, paratype.

However, confusion of A. branchi with A. reticulata taxa seems unlikely, as they most probably do not overlap geographically. From West Africa sensu stricto (Senegal to the Nigerian Cross River; see Penner et al. 2011 for discussion), A. reticulata has been reported from Nigeria (Eniang and Ijeomah 2011: Oban Division, Cross River National Park, Cross River Province; Butler and Reid 1990 and Hughes 1983: without further details) and Ghana (Hughes 1983). The Ghanaian record is of particular interest, as it might be conspecific, based on biogeographic arguments, with the new species. Unfortunately, this snake could not be investigated.

The Cross River area is zoogeographically part of the Lower Guinea forest zone. It is not yet clear where the zoogeographic barrier between the West and Central African fauna exactly runs: at the Cross River, the Niger Delta, or the Dahomey Gap. Nor has it been clarified what exact processes are responsible for the separation of fauna and flora, and what is the time scale during which taxa in both areas evolved (see Penner et al. 2011; Jongsma et al. 2018 and studies cited therein). The geographic scale, processes and time most likely varies between taxa (see Bell et al. 2017 for Central African examples). However, it is becoming more evident that there is only very little to almost no overlap in occurrences between forest species of the Upper and Lower Guinea realms. Indeed, many studies, mostly recent, either have discovered related but distinct species in both forest blocs or have shown that “widespread” taxa actually comprise cryptic species complexes, including species that either occur in West or Central African rainforests. Examples of recent discoveries of species pairs within amphibians are Sclerophrys taiensis and S. tuberosus (Bufonidae; Rödel and Ernst 2000); Acanthixalus sonjae and A. spinosus (Hyperoliidae; Rödel et al. 2003); Cardioglossa occidentalis and C. leucomystax (Arthroleptidae; Blackburn et al. 2008); and Amnirana “albolabris-West” and Amnirana albolabris (Ranidae; Jongsma et al. 2018). Among reptiles different species in West and Central Africa have been discovered in the crocodile genera Mecistops and Osteolaemus (Crocodylidae; Shirley et al. 2014); the turtles Pelusios cupulatta and P. gabonensis (Pelomedusidae; Rhodin et al. 2017); the cobras Naja guineensis and N. melanoleuca (Elapidae; Wüster et al. 2018); and the vipers Atheris hirsuta and A. squamigera (Viperidae; Ernst and Rödel 2002). Among birds there are for instance Picathartes gymnocephalus and P. oreas (Picathartidae; Treplin 2006); and among mammals there are different West and Central African species within the antelope genera (Neotragus spp., Tragelaphus spp.; Philantomba spp.; Hernández Fernández and Vrba 2005; Moodley and Bruford 2007; Colyn et al. 2010), within small carnivores (Genetta johnstoni and G. piscivora; Gaubert et al. 2004), otter shrews (Micropotamogale spp.), or within various bats such as the genus Rhinolophus (Fahr et al. 2002). Many more examples have been published (Huntley et al. 2019) and several more examples, still unpublished, are known to us.

Apart from a species-level uniqueness of Upper Guinean forest assemblages, it is also evident, that this region is an important area for old endemic lineages. Prominent herpetological examples comprise the frog family Odontobatrachidae (Barej et al. 2014), and the frog genera Pseudhymenochirus (Pipidae; Evans et al. 2004) and Morerella (Hyperoliidae; Rödel et al. 2009). Therefore, the discovery of a new and presumably endemic species of fossorial snake from the western Upper Guinea forests is not very surprising. However, further surveys are needed to determine the geographic range of the new snake species and to gather more information about its ecological needs and biological properties.

We thank the Forestry Development Authority (FDA) of the Republic of Liberia for permission to undertake our research, and for the collection and export permit (ref: MD/037/2018/11). We especially thank the Liberia Forest Sector Project of the FDA for financing our surveys. We are indebted to the Society for the Conservation of Nature of Liberia and the Wild Chimpanzee Foundation and their respective employees in general and Jerry C. Garteh and Annika Hillers in particular for the invitation to conduct this survey and for logistic support. Without the work of our guide, Junior Kerkpel, our driver Prince, and the support of the people of Dulcor Gbondo Town, the survey in Foya would not have been possible. For technical support with the ct-data we are indebted to Anna Sophie Welter (MfN) and Jonathan Brecko (RBINS). Robert Schreiber and Isabelle Waurick (MfN) supported our work by barcoding the type specimens. Frank Portillo and Eli Greenbaum provided an unpublished 16S sequence of an Atractaspis reticulata heterochilus. Frank Tillack (MfN) photographed the type specimens. Göran Nilson (GNM, Göteborg, Sweden), and Erica Mejlon (ZMUU, Uppsåla, Sweden) searched the holotype of A. reticulata for us. However, the type proved to be in the Stockholm collection. Sven Kullander and Andrea Hennyey (Swedish Museum of Natural History, NRM) made possible the loan of this snake and permitted ct-scanning and sex determination. Jakob Hallermann (ZMH, Hamburg, Germany) clarified the status of ZMH material, and Frank Glaw (ZSM, Munich, Germany) made possible the examination of material under his care. Alexander Kupfer (SMNS, Stuttgart, Germany), Dwight P. Lawson (Oklahoma Zoo, USA), Joyce Lowe (Newtown Linford, UK), and Tomáš Mazuch (Dříteč, Czech Republic) provided to us some difficult to get literature. All this support is gratefully acknowledged!