Taxonomy of the Proisotoma complex. VI. Rediscovery of the genus Bagnallella Salmon, 1951 and epitoky in Bagnallelladavidi (Barra, 2001), comb. nov. from South Africa

Abstract The genus Bagnallella Salmon is restored and given a diagnosis. It takes an intermediate position between Proisotoma Börner and Cryptopygus Willem complexes and is characterized by the separation of the two last abdominal segments (like in Proisotoma) and 3 and 5 s-chaetae on the fourth and fifth abdominal segments (like in Cryptopygus and its allies). A list of and key to species belonging to Bagnallella is given. Bagnallellabisetacomb. nov., B.dubiacomb. nov., B.sedecimoculatacomb. nov., and B.tenellacomb. nov. are commented and redescribed. Morphology of Bagnallelladavidi (Barra), comb. nov. is described from the specimens from South Africa. So far B.davidi appears to be a complex of forms differing in size of the furca and macrochaetae. Two types of strongly modified males were found and described. Antennae, ventral side of abdomen, posterior edge of abdominal tergites, and mandibles are affected with epitoky. The nature of the discovered strong polymorphism is unclear.


Introduction
Knowledge on the Collembola of South Africa has increased significantly over the last decade, with most new species described from the Western Cape Province where the majority sampling has been made (Janion-Scheepers et al. 2015). From these collections, a rich diversity of Isotomidae has been discovered, including Parisotoma (Potapov et al. 2011) and Cryptopygus . From these collections, we also recorded three known species (P. davidi, P. tenella, P. sedecimoculata), which resemble the genus Cryptopygus but have Abd. V and VI separated. Thus, these species cannot be attributed to any genus of the Cryptopygus complex but rather belong to the Proisotoma complex. This paper determines the position of these three species by recovering a genus erected in the past by Salmon (1951). Also, several other species mostly distributed in the Southern Hemisphere belong to this taxon. In addition, we describe an unusual polymorphism in Proisotoma davidi (Barra, 2001) which remains unsolved.
Position of the genus in the subfamily Anurophorinae. To date an appropriate generic name did not exist for the small group species related to Proisotoma Börner, 1901 sensu lato which were discussed in the monograph of Potapov et al. (2006). This group, so-called "Proisotoma tenella, ripicola, biseta", consists of forms sharing characters such as: the three last abdominal segments separated, manubrium with anterior chaetae, four prelabral chaetae, and presence of three and five s-chaetae on Abd.IV and V, respectively.
Recently, one more species, Proisotoma sedecimoculata (Salmon, 1943), became a probable candidate to belong to this group (Potapov and Janion 2017). This species was described by Salmon (1943) as Folsomia sedecimoculata and was afterwards proposed as a generotype for the new genus Bagnallella Salmon, 1951. Bagnallella was erected based on three last abdominal segments fused, bidentate mucro and eight ocelli. Later, Bagnallella was lost in the taxonomy of the subfamily and was mostly treated as a junior synonym of either Folsomia or Proisotoma. After the examination of the type specimen, it was discovered that the three last abdominal segments were actually separated (Potapov and Janion-Scheepers 2017). Here, we suggest restoring Bagnallella for the group of species mentioned above, rather than erecting a new generic name. Several other forms described under different generic names also fit to Bagnallella at lesser or larger degree of accuracy. For these species, the two Bagnallella key characters were mentioned in the associated descriptions or were seen by us, apart from three forms with unknown sensillar chaetotaxy. Nevertheless, we suppose the last ones (notated with question marks in the list of species of Bagnallella below) belong to the genus. Among these species, Bagnallella sedecimoculata is poorly described and so is not the best to be a generotype, but we prefer to keep a generic name already created by John Salmon.
The incertae sedis genus Bagnallella combines the characters of two large generic groups by the separation of its two last abdominal segments and the presence of three and five s-chaetae on Abd.IV and V respectively in characteristic position. The former character is a diagnostic feature of the Proisotoma Börner, 1901 complex, the latter indicates basic set of s-chaetae in Cryptopygus Willem, 1902 and related genera belonging to Cryptopygus complex (Potapov et al. , 2013. The latter complex is characterized by the fusion of the two last abdominal segments. The genus Bagnallella takes a neatly intermediate position between Cryptopygus (Cryptopygus complex, Southern Hemisphere) and Scutisotoma (Proisotoma complex, Northern Hemisphere). The three genera share, apart from the characters of the subfamily, the presence of a furca, a mid-tergal position of macro s-chaetae on body tergites, and the absence of any specific apomorphy. The combinations of the two key characters mentioned above are shown in Figure 1. After the separation of the two last abdominal segments, we suggest treating Bagnallella in the Proisotoma complex.
Distribution and ecology of Bagnallella. The genus is distributed worldwide. More local species (B. davidi, B. douglasi, B. mishai, B. biseta, B. koepckei) are distributed in the Southern Hemisphere that indicates further relation to the "austral" genus Cryptopygus. Due to apparent ruderal B. tenella and pioneer B. ripicola, the genus also occurs in the Northern Hemisphere. The ecology of the former species is especially remarkable, as B. tenella is often recorded in mass abundances (Agrell 1939;Arle 1970;Neves and Mendonça 2016). We speculate that B. davidi has an unusual biology and ecology due to its morphological modifications (described below).
Key to known species of Bagnallella * 1 Mucro tridentate (Fig. 10)  Descriptions and remarks to species of the genus. Below we give the diagnosis, description, remarks, and distribution, with, if possible, ecological remarks of all species of Bagnallella. Some sections may be omitted if the associated species have good morphological descriptions in the literature. (Rapoport, 1963), comb. nov. Rapoport, 1963 Material examined. Adult male from Argentina under label "Tucuman, 19/4/59, entre maderas ...", deposited in the Museum national d'Histoire naturelle (MNHN), Paris, France. This individual was obviously among the material the original description was based on by E.H. Rapoport.
Discussion. Our redescription is based on one individual of E. Rapoport, and more material is needed to complete the understanding of B. biseta. The species obviously belongs to the genus Bagnallella by separation of two last abdominal segments and s-chaetotaxy of Abd.IV and V. After chaetotaxy of tibiotarsi, ventral tube, and dens, B. biseta appears to be a more polychaetotic species than its congeners. Description. Colour grey. Cuticle, ocelli, outer mouth parts, and antennae as in B. sedecimoculata. PAO as long as 0.8-0.9 Ant. I and as 1.4-1.5 as long as Claw III. Ventral side of head with 4+4 postlabial chaetae. Th.III with 2+2 ventral axial chaetae.
Discussion. This species was named after its dubious position in generic system of Proisotoma/Cryptopygus (Deharveng 1981). It resembles B. sedecimoculata (see the Discussion below) and apparently belongs to the genus Bagnallella by separation of two last abdominal segments and s-chaetotaxy of Abd.IV and V. Our specimens from New Zealand match the first description.
Distribution. Bagnallella dubia was described from Marion Island and recorded in Macquarie Island (Greenslade and Wise 1986) and alpine sites of New Zealand (Babenko and Minor 2015). The species is possibly widely distributed in cold sites of high altitudes of the Southern Hemisphere. Its occurrence in Australia (Greenslade 2006) needs to be verified. Discussion. The species can be attributed to Bagnallella by the two key characters of the genus. It is characterized by 33/22235 s, 11/111 ms, bifurcate maxillary palp, long polychaetotic dens (34-35/14), and bidentate mucro. The presence of seven ocelli and 2+2 chaetae on anterior side of manubrium are two unique characters among members of the genus.
Distribution. This species is currently only known from SE Brazil.
Discussion. The full redescription is given by Fjellberg (2007). Distribution and ecology. Europe. It prefers sandy places along the edge of water.
Discussion. Bagnallella sedecimoculata was described from New Zealand and was designated as type species for the genus Bagnallella (see the discussion to the genus above). The type specimen from New Zealand was studied (Potapov and Janion 2017), although only one generic character (separation of two abdominal segments) was proven. The redescription given above is based mostly on the South African material which looks conspecific to one individual in hand from New Zealand (L.D. and A.B. leg.). Bagnallella sedecimoculata resembles B. dubia and differs by ventral chaetae on Th.III (absent vs present) and a few more anterior chaetae on dens (16-20 vs 12-16). The latter character is not stable enough to separate the two species.

Bagnallella tenella
Number of posterior chaetae on dens, clavate tibiotarsal chaetae, and chaetae on retinaculum vary. We treat all this variation within one species, but further study is needed. The independence of Proisotoma nigromaculosa (Hawaiian Islands) is doubtful.
Although we have no material from Europe, Stach's (1947) concept of P. tenella based on the specimens from Poland is accepted by us (for details see Potapov 2001;Fjellberg 2007). Our tropical material fits Stach's (1947) descriptions.
Distribution and ecology. Widely distributed cosmopolitan species. Common in tropics. In higher latitudes only in protected soils.
Discussion. Bagnallella davidi is a rather peculiar species of the genus due to few chaetae on dens (vs many more chaetae both on anterior and posterior sides), tridentate mucro (shared with B. mishai only), and 3+3 postlabial chaetae (vs 4+4 or more in other species). The first description is almost complete and, therefore, we made very few additions concerning mouth parts. The species exhibits high variation in length of macrochaetae and show different modifications of males. All the forms (described below) can indicate either high plasticity of a single species or the complex of separate although closely related species, calling for further morphological, biological, ecological, and molecular investigations.
Distribution. Eastern Cape, Amatola Mountains (type location) and widely in the Western Cape and Free State (our material) provinces of South Africa.
2. "Short-haired" form (Figs 12-14). Macrochaetae are short, shorter than common chaetae on most abdominal segments. Ratios: Mac : Abd.V Mac on Abd.V as long 0.2-0.3 of tergal midline; Mac : mucro = 0.9-1.4; Mac : dens 0.2-0.3. In spite of their small size, macrochaetae are erect and stiff and so well recognized indicating their possible ecomorphic nature although the integument and mouthparts are not modified. Head and furca appear to be relatively larger than in typical form. Ratio manubrium : dens : mucro as 5.1-6.6 : 4.1-5.8 : 1; dens : Claw as 3.3-4.5. All other significant characters (s-chaetotaxy, mouth parts, chaetotaxy of extremities) are as in typical form. All instars and both sexes can belong to this form.
We also found individuals with middle-sized macrochaetae (as in Fig. 15), which is on Abd.V twice as long as mucro (vs 3.3-4.1 in typical form and 0.9-1.4 in short-haired form) and half as long as dens (vs subequal to dens in typical form and 0.2-0.3 as dens in short-haired form).
Each studied population consists of only one of the forms, and we have not found a continuous range of macrochaetae variability, apart from short-haired clasping adult supermales occurring in "normal" populations.
3. "Clasping supermales" (Figs 15-20). Ant.I-III expanded and partly fused. Antennae joints probably lost mobility. Inner side of Ant.II and III is armed with thickened, flame-shaped, and bifurcate chaetae which probably form a clasping organ. Front of head have chitinized tubercles. Anal valves are armed with spines. Mandibles without apical teeth. Macrochaetae short. Subadult clasping supermales, i.e., males without fully developed genital plate and without developed ejaculatory duct, have also expanded antennae although without modified chaetae on inner their side. They show normal (longer than on adult supermales) macrochaetae and normal mandibles and have no spines on anal valves (Fig 15-20). The females of the same population belong to the form with middle-sized macrochaetae. 4. "Spiny supermales" (Figs 23-25). One of the males has a row of spiny p-chaetae on Abd.IV and strong thickened macrochaetae on lateral parts of Abd.VI. Other macrochaetae on the body are weakly developed. Common chaetae on dorsum of Abd. IV-VI curved at apex (Fig. 25). Mandibles without apical teeth and molar plate (Fig.  24). Outer mouth parts (labrum, maxillary outer lobe, and labium) not fully developed. Unmodified males are much more frequent than the two male forms described above. In most populations, only unmodified males are known. They can show all possible length of macrochaetae and belong to associated forms (Figs 21, 22).

Discussion
If present, the clasping organ of Collembola is formed by two symmetrical complexes associated with, respectively, left and right antennae, or, more rarely, other limbs. Paired "clasps" are evolutionary formed in different taxa, for example, in the family Sminthurididae, Vertagopus reuteri (Schött, 1893), Rhodanella minos Denis, 1928, Seira raptora (Zeppelini & Bellini, 2006) (Delamare Deboutteville et al. 1969Betsch 1980;Fjellberg 1982;Bellini et al. 2009). In "clasping supermales" of Bagnallella davidi, all modified chaetae (curved spines, flame-shaped, flattened, and bifurcate) are found on the inner side of the antennae, forming a unique type of clasping organ. This allows males to grasp females between the two antennae at the axial region. The only possible similar case was described in Vertagopus pseudocinereus Fjellberg, 1975, which clasps the female with right and left antennae. This species has curved and serrated chaetae on antennae, without any strong modifications (Goloschapova et al. 2006: fig. 2). We can assume that males of B. davidi uses this unpaired "clasp" (Fig. 20) in a similar manner, for clasping onto the female.
The function of the spiny row in "spiny supermales" of B. davidi is more difficult to explain. The armature of the posterior row on Abd.IV somewhat resembles backward-shifting macrochaetae in males of Scutisotoma acorrelata Potapov, Babenko & Fjellberg, 2006, while strong lateral macrochaetae on Abd. VI indicate some similarity to Ephemerotoma huadongensis (Chen, 1985), which shows similar armature on both Abd. V and VI (described by Potapov et al. 2015.