The millipede family Ammodesmidae (Diplopoda, Polydesmida) in western Africa

Abstract Ammodesmidae are represented in western Africa by two species of a single genus, Ammodesmus Cook, 1896 (= Cenchrodesmus Cook, 1896, syn. n.). The type-species Ammodesmus granum Cook, 1896 (= Cenchrodesmus volutus Cook, 1896, syn. n.) is redescribed, based on neotype selection, as well as on additional samples, often containing numerous specimens, from Liberia, Guinea and the Ivory Coast. A new species is described from Mount Nimba, Guinea: Ammodesmus nimba sp. n.


Introduction
The small Afrotropical family Ammodesmidae has hitherto been known to comprise only three genera. One of these is Elassystremma Hoffman & Howell, 1981, a recently reviewed oligotypic genus currently comprising four species from Kenya, Tanzania and Malawi Howell 1981, VandenSpiegel andGolovatch 2004). Two fur-ther genera, both monobasic and both described from western Africa, have remained enigmatic ever since their proposals.
The taxonomic history of the family Ammodesmidae is rather confusing (Jeekel 1970). It was originally introduced invalidly, because it contained two genera, Ammodesmus, a nomen nudum, and Doratodesmus Cook, 1895, the latter name proposed to replace the preoccupied Doratonotus Pocock, 1894 (Cook 1895). Cook (1896a) validated Ammodesmidae only through selecting and diagnosing Ammodesmus granum Cook, 1896 as type species. A little later, Cook (1896b) slightly rediagnosed Ammodesmus and also described the new genus Cenchrodesmus for the sole species C. volutus Cook, 1896, he had collected syntopically in Liberia together with A. granum. This latter species was said to have been based upon the holotype while C. volutus upon three syntypes. Both were first mentioned to have been taken from "the western part of the country", but the exact locality, Mt Coffee, was soon provided elsewhere (Cook 1896c).
The diagnoses of the two genera and species were rather anecdotal and provided little useful information (Cook 1896b, page 414): "Both genera have the habit of coiling into a sphere. The second segment is enormously enlarged so as to completely conceal the head and the first segment when viewed from the side as well as to cover the space left between the decurved carinae of the other segments when the creatures are coiled. Ammodesmus has the dorsum roughened by a transverse row of large papilliform tubercles rising from the posterior part of each segment, while Cenchrodesmus has the segments nearly smooth. When disturbed it coils up and lies motionless, and then is perfectly concealed, having exactly the appearance of a grain of sand". In summary, Cook (1895Cook ( , 1896a distinguished the family Ammodesmidae, as well as both Ammodesmus and Cenchrodesmus, by their extremely small size, claimed to be the smallest in Polydesmida (only about 2 mm in length), and their ability to volvate.
As no genital structures had been mentioned, VandenSpiegel and Golovatch (2004) suggested that all type specimens from Liberia might have been females. Furthermore, since the types could not be relocated for revision, Ammodesmus and Cenchrodesmus have ever since remained "nomina dubia" (Hoffman 1980). So even when Hoffman and Howell (1981) described the new genus Elassystremma and its sole, and type, species E. pongwe Hoffman & Howell, 1981 from Tanzania, the need was again emphasized in revising the two western African genera before unequivocally assigning E. pongwe to the family Ammodesmidae. The same uncertain situation has also remained after the latest review of Elassystremma which added three further congeners from eastern tropical Africa (VandenSpiegel and Golovatch 2004).
Between 2008 and 2011, a rich material of diplopods was taken or amassed by the first author from Liberia, Guinea and the Ivory Coast. This large collection appears to contain a proportion of Ammodesmidae, fortunately also with males from each locality becoming available for study. Three different morphotypes could be distinguished at once, including the two forms described by Cook (1896a-c), as well as a new one with a very peculiar colour pattern. Moreover, quite unexpectedly, both of Cook's species happen to be the most common and widespread, in larger samples always coexisting, with all specimens showing the papilliform metatergal tubercles typical of Ammodesmus granum being males and the samples with nearly smooth tergites representing juveniles or females, i.e. just like the situation described for Cenchrodesmus volutus. This striking sexual dimorphism (which somewhat resembles that observed in some volvatory species of Eutrichodesmus Silvestri, 1910, a genus of the Australasian family Haplodesmidae -see Golovatch et al. 2009aGolovatch et al. , b, 2010 allows for the following unequivocal identifications and synonymy to be proposed: Ammodesmus granum Cook, 1896 must have been based on a male holotype, while Cenchrodesmus volutus Cook, 1896 is its junior subjective synonym which seems to have been based on three female syntypes. The third form appears to be a new species of Ammodesmus, taken from a single locality (Mt Nimba) and described below.
The present paper provides a review of Ammodesmus, the sole ammodesmid genus that appears to populate western Africa. Its gonopod and numerous other characters are documented here for the first time, and compared to those of Elassystremma, the sole eastern Afrotropical counterpart ammodesmid. The type species Ammodesmus granum is redescribed, based on neotype designation, and a new species is added to this genus. A distribution map of and a key to the Ammodesmus species are also given.

Material and methods
The bulk of material belongs to the collection of the Royal Museum for Central Africa (MRAC), Tervuren, Belgium, with only a few duplicates shared with the collections of the Zoological Museum, State University of Moscow (ZMUM), Russia and the Muséum national d'Histoire naturelle, Paris (MNHN), France, as indicated hereafter. All samples are stored in 70% ethanol. Photographs were made with a Leica digital camera Leica DFC 500 mounted on a Leica MZ16A stereomicroscope. Images were processed with a Leica Application Suite program. Specimens for scanning electron microscopy (SEM) were air-dried, mounted on aluminium stubs, coated with gold and studied in a JEOL JSM-6480LV scanning electron microscope.
Gonopods mostly complex; coxae globose, usually but not always strongly enlarged and deeply excavate in the middle (= gonocoel), cannulae very evident. Telopodites basically unipartite, slender or stout, sometimes with a small lateroparabasal outgrowth, only seldom strongly exposed (Ammodesmus granum), more usually deeply sunken into gonocoel, leaving only tips exposed. Seminal groove mostly running on mesal face, turning laterad due to telopodite twisting only distally to subapically, with a very evident (Ammodesmus granum) or small solenomere either devoid of or supplied with a hairy pulvillus; accessory seminal chamber absent.
Distribution. Liberia, Guinea and Ivory Coast (western Africa), as well as Tanzania, Kenya and Malawi (the African eastern Arc Mountains). Head small, epicranium and interantennal region finely and densely granulose, lower half setose. Three labral teeth equal in size and length. Antennae short; antennomere 5 longest and largest, strongly enlarged, about as high as long; antennomeres 5 and 6 each with a distodorsal group of 20 to 30 bacilliform sensilla; antennomeres 4, 5 and 6 each with a single macroseta on dorsal side near apical third; terminal antennomere with usual four apical cones. Collum rather large and moderately convex, nearly not covering the head from above, surface finely and densely granulose. Tergum 2 with particularly strongly enlarged, spatuliform paraterga, latter of following segments not enlarged; lateral end subtruncate, but rounded; overlap pattern typical from paratergum 4 onward. Ozopore formula: 5, 7, 9, 10, 12, 13, 15-17(18); ozopores opening flush on tergal surface about anterior third of paraterga, the opening sometimes being concealed by preceding paratergum. Limbus smooth. Telson relatively small, its posterior edge with a row of macrosetae, ventrolaterally with 2 macrosetae borne on small knobs; epiproct very short and stout, surmounted by four conspicuous macrosetae in pits (= apparently, a spinning apparatus); hypoproct subtrapeziform with a paramedian pair of macrosetae borne on small knobs. Sterna very narrow, most coxae subcontiguous medially. Legs moderately robust, rather short and setose. First ♂ tarsus with modified setae, each last ♂ tibia showing an elongated process bearing a long apical seta, tarsus reduced, claw vestigial; other legs not modified.

Key to recognized genera of Ammodesmidae
Distribution. Western Africa (Liberia, Guinea and Ivory Coast). Species included. Ammodesmus granum and A. nimba sp. n.

Key to Ammodesmus species
1 Coloration pinkish to brownish with darker metaterga (Fig. 1   Gonopod with a small globular coxa reaching in length only about one-third of telopodite; the latter slender, flattened and twisted mesad in distal part, with a small sacshaped outgrowth laterally at base. Redescription. ♂ ca 1.9 mm long; maximum width, 0.9 mm. Entire dorsal surface covered with a thin layer of secretions (= cerategument), under which the body integument is light brown to pinkish with metaterga of each segment brownish (Fig. 1). Body with 18 or 19 body rings (16 or 17+1+T), shape as in Figs 1, 2 & 3, with caudal body end tapering towards a relatively small telson not concealed by paraterga (Fig. 6).
Distribution. Known from Liberia, Guinea and the Ivory Coast. It is noteworthy that at Mt Nimba this species occurs parapatrically with the new congener described below.

Remarks.
A. granum is striking in being perhaps the only species in Polydesmida in which both sexes vary in the number (18 or 19) of body rings. Infraspecific variations in the number of body segments in this order are usually quite rare, always being stable per sex. Thus, in such cases males always have fewer body rings (18 or 19) than females (19 or 20), a situation not too uncommon, e.g., in Haplodesmidae (Golovatch et al. 2009a(Golovatch et al. , 2009b(Golovatch et al. , 2010 and, especially, Opisotretidae (Golovatch 1988).
Neotype designations for both A. granum and C. volutus are necessary, because the original types can be presumed as being lost. A special search undertaken among Cook's diplopod collections, currently housed in the Smithsonian Institution, National Museum of Natural History, Washington, D.C., had failed already before the description of Elassystremma pongwe by Hoffman and Howell (1981). Diagnosis. Minute polydesmidans with a characteristic, spotty colour pattern of the caudal edge of each segment. Gonopods with extremely large coxae concealing the telopodites inside a deep gonocoel.
Relationships. Superficially, both species of Ammodesmus might look sufficiently different to consider them as representing different genera, especially as regards the absence of sexual dimorphism in metatergal structure and the presence of a deep gonocoel in A. nimba as opposed to A. granum. The main distinctions can also be summarized in a tabular form (Table). However, based on all evidence, we are rather inclined to recognize only two valid genera in Ammodesmidae, both quite disjunct also geographically (Fig. 47). Name. Referring to the type locality, a noun in apposition. Distribution. Known only from the type locality and probably endemic to Mt Nimba.

Conclusion
Despite extensive efforts applying the same collecting techniques in similar habitats in many places, A. nimba appears to occur, and to be apparently common, only at the single locality whence it has been taken, whereas A. granum has a surprisingly wide distribution. The vast range of A. granum, currently reported from the western part of Liberia, at Mt Nimba in Guinea and in the Taï forest in the western part of Ivory Coast, is rather unusual for such a tiny and hygrophilous animal. Certainly being likewise rather poorly vagile, this species can be assumed to represent a relict which must have been widely distributed in the past when woodlands were continuous in western tropical Africa (Couvreur et al. 2008). Geographically, the family Ammodesmidae seems to be purely Afrotropical, Ammodesmus being obviously confined to western Africa while Elassystremma to eastern Africa (Fig. 47). All four Elassystremma species (E. pongwe Hoffman & Howell, 1981, E. michielsi VandenSpiegel & Golovatch, 2004, E. leave VandenSpiegel & Golovatch, 2004and E. prolaeve VandenSpiegel & Golovatch, 2004 are slightly larger than Ammodesmus (up to 5 mm long), and their gonopods are invariably complex, sunken inside a deep gonocoel (VandenSpiegel and Golovatch 2004). Likewise, only one species, E. prolaeve, is widespread, occurring not only in Kenya and Malawi, but obviously also in-between in Tanzania (Fig. 47).
The use of Winkler-Mocsarski apparatuses, or Winkler apparatuses for short, appears to be the most appropriate technique in sampling particularly cryptic soil/litter fauna. This technique has allowed for material to be collected in almost any East and West African tropical forest prospected by the first author and it is most likely that new species will be revealed in the central parts of the continent, if the collecting efforts use appropriate techniques. At least the wide geographical gap between both genera certainly invites further studies (Hoffman 1993). More refinements to the distribution of already known species are also most plausible.