Careful selection of geographical settings facilitates assessment of the spatial and temporal components of the organic evolution. Oceanic islands such as the Galápagos Islands or the Sunda Arc are in this respect the absolute favourites since the time of Charles Robert Darwin and Alfred Russel Wallace. Their nearest inland alternatives are the “sky islands”, such as the Eastern Arc Mountains (EAM) of Eastern Africa. They consisting of 10 main forested blocks broadly scattered through Kenya (Taita Hills) and Tanzania (North Pare, South Pare, West Usambara, East Usambara, Uluguru, Nguru, Kaguru, Rubeho and Udzungwa; Fig. 1A). A few other adjacent forested areas, such as geologically young forested volcanoes (i.e. Mt. Kilimanjaro, Mt. Meru, Mt. Hanang, Fig. 1A) or lowland forests (i.e. Kimboza, Pugu Hills, Fig. 1A) do not belong to EAM, even though their Biota might have been variously connected with that of EAM in the geological past. The forests of EAM are remarkable for their archipelago-type habitats supporting disproportionately high biodiversity with many narrow endemics (Lovett & Wasser 1993). The predominant hypothesis behind this phenomenon is the high biotic and abiotic stability of EAM during the last 5–10 Ma, when most of the Afrotropical forest repeatedly shrank to a few small and widely separated core survival zones (Hamilton & Taylor 1991; Maley 1996). Indeed, the global glacial cycles of the Plio- and Pleistocene manifested themselves in Africa by repetitive droughts reducing the forest cover to a few small and widely separate refugia (deMenocal 2004). The EAM, however, owning to their height and relative proximity to the Indian Ocean, are believed to continuously precipitate atmospheric moisture sufficient to support rainforest and its associates animal life ever since the Miocene, when the African forest had it last maximum (Hamilton & Taylor 1991). This hypothesis is directly supported by the pollen, charcoal and carbon isotope evidence obtained from two deep soil probes taken in Udzungwa (Mumbi et al. 2008) and Uluguru (Finch et al. 2009) and reveling stable forest composition during the past 48,000 yr, that is through a period greatly exceeding the Last Glacial Maximum with its peak at about 25,000 ya.
The pivotal book edited by Lovett and Wasser (1993) placed EAM in the spotlight of modern phylogeographical studies. The EAM were thought to act as “species pumps” (Fjeldså & Lovett 1997; Murienne et al. 2013) accumulating and maintaining diverse and endemic forest-dependant clades of different age. Since then, a number of EAM organisms had their phylogeny assessed at the fine scale and involving the spatial and temporal aspect of evolution. The most suitable organisms for such analyses should meet criteria as: (1.) form a clade with the age of diversification not exceeding that of EAM forests; (2.) be common enough to permit predictable and efficient sampling; (3.) be biologically dependant on forests for survival; (4.) have relatively low dispersal capacity across intervening dry and hot savannah and (5.) have their diversity satisfactory known and Linnaean taxonomy stable. Different clades variously meeting these criteria have been recently employed to shed light on the phylogeographic past of EAM. In plants, widely cultivated African violets (Saintpaulia H. Wendl., Gesneriaceae) nearly endemic to EAM attracted much phylogeographical attention (Möller and Cronk 1997; Lindqvist & Albert 1999; Dimitrov et al. 2012). Low dispersing and moisture dependant Vertebrata were frequently targeted, such as frogs (Loader et al. 2014), caecilians (Loader et al. 2011), chameleons (Measey & Tolley 2011; Tolley et al. 2011; Ceccarelli et al. 2014), snakes (Menegon et al. 2014) and small mammals (rodents by Bryja et al. 2014; shrews by Stanley et al. 2015). Non-vertebrate animal life of EAM, although undoubtedly highly diverse and with a number of forest-dependant low-dispersal clades, is still too fragmentary known taxonomically to permit their reliable usage for phylogeographical purposes. Among the latter, attempts were focussed on katydids (Hemp et al. 2016), flat (Heiss & Grebennikov 2016) and assassin (Weirauch et al. 2017) bugs, ground beetles (Grebennikov et al. 2017) and weevils (Grebennikov 2015a). Some of these studies, however, relied on the non-recombinant mitochondrial COI gene, which when used alone (i.e. without nuclear markers) is prone to various shortcomings obscuring past evolutionary events (Funk & Omland 2003).
This paper highlights the discovery of a new weevil genus with three new species each narrowly endemic to a single EAM block, as well as finding its sister group and assessing it phylogeographically. The clade’s representatives (Fig. 1B,C) were first detected in 2002 by sifting wet forest litter (Fig. 1F–I) near Amani village in the East Usambara Mountains, Tanzania (Fig. 1A) and for about a decade could not be assigned to any known genus. Based on current systematic practice, they belong to the presumably polyphyletic subfamily Molytinae comprising dozens of dubiously defined tribes and hundreds of genera, some of them recently discovered (Grebennikov 2014b) or inadequately known (Grebennikov 2016a). Externally, the East Usambara specimens resembled those of the Afrotropical genus Typoderus Marshall, 1953, with 11 poorly known species, some of which (Fig. 1D,E) were found in the same litter samples. Lack of adequate comparative material coupled with the absence of DNA data delayed phylogenetic assessment of the East Usambara specimens. During this time sampling in EAM (Fig. 1) and in some nearby forested areas was conducted. It revealed similarly shaped, although morphologically distinct beetles in Uluguru and Udzungwa, but nowhere else (Fig. 1A).
Discovery of these novel, morphologically similar, and phylogenetically puzzling weevils, seemingly restricted to three EAM blocks, evoked a number of evolutionary questions. First, do they form a clade, and if so, what might be its sister-group? Second, do specimens from each of the three localities form a reciprocally monophyletic group? Third, how did these flightless and presumably low-dispersing, habitat-dependent organisms come to populate three blocks of wet forest widely separated by seemingly highly unsuitable dry and hot savannah? More specifically, can the observed distribution be explained through normal ecological dispersal (Heads 2014) with subsequent subdivision into three allopatric lineages by means of climatically induced wet forest fragmentation (= vicariance) or, alternatively, may a hypothesis of long range dispersal be justifiably evoked? The latter, although infrequent, has been convincingly demonstrated for a number of animals crossing large saltwater barriers, such as at least some among 40 species of eyeless arthropods of the Galapagos Islands (Peck 1990), non-human primates of the New World (de Queiroz 2014), and minute blind and wingless Orthotyphlus Zaballos & Mateu, 1998 ground beetles colonizing New Caledonia (Andújar et al. 2016). Last but not least, how old are the evolutionary events leading to the present-day diversity and distribution of these novel beetles and do they coincide with the Pliocene-Pleistocene climatic fluctuations? This paper attempts to shed light on all these evolutionary questions by doing a series of DNA-based phylogenetic analyses and interpreting the results taxonomically, phylogenetically and phylogeographically.
Material and methods
Specimen sampling. A total of 130 individual litter samples were taken in 2010–14 in 14 discrete Tanzanian forests of different age and genesis, representing those on EAM (9), volcanic highlands (3) and lowland forests (2; Fig. 1A, Table 1). Fine fraction of the litter (Fig. 1H) was physically separated by using a large sifter (35 cm in diameter, Fig. 1G). The same litter was sifted twice: first through a larger mesh (square side: 10 mm) firmly fixed in the sifter, following by sifting through a wire insert (square side: 5 mm; Fig. 1H). Such two-step sifting was judged more efficient in processing a larger volume of litter without clogging the finer mesh, particularly in wet conditions. Taking a sifting sample lasted for about 2–4 hours and the wet mass of the final fraction (< 5 mm) was on average 7.7 kg (Fig. 1H, varying between 1.4 kg and 36.1 kg per sample with the total of 996 kg of fine litter fraction sampled, Table 1). Living organisms were then extracted (typically the following night) by placing approximately one handful of litter in a mesh bag suspended inside Winkler non-electric funnels and operated in a shelter (Fig. 1I). Funnels were suspended for 2–8 hours and then re-loaded with the same litter at least once with the aim to shake and excite organisms and thus to stimulate their active movement and, therefore, extraction. Six funnels each containing nine mesh bags were simultaneously in operation. A container at the bottom of each funnel containing extracted live organisms was emptied into a Whirl-Pak sealable plastic bag with 96% ethanol. Ethanol was drained and replaced at least three times in intervals of 1–3 days, to effectively remove water from specimens. All specimens from the same sifting sample received identical geographical labels (given verbatim in Table 1). During fieldwork, specimens were stored for up to two months at room temperature and then brought into a lab and placed in a freezer at –9°C. All herein reported specimens (including all outgroup taxa) are deposited in the Canadian National Collection of Insects, Arachnids and Nematodes in Ottawa, Canada (CNC). Each specimen can be traced through a unique identifier label pinned under a beetle and bearing the code CNCCOLVG0000XXXX; the last four X’s correspond to a unique number referred to on the topologies (Figs 3–5).
Tissue submission and DNA laboratory procedures. Within two months of their capture, specimens were sorted from the samples and processed for DNA barcoding using the standard animal COI-5’ fragment (Hebert et al. 2003; Ratnasingham & Hebert 2007). One leg per specimen (normally a right femur cut open on both ends to expose muscle tissue, and additionally partly crushed with forceps) was placed in two drops of 95% ethanol in a sealed well on a standard 96-well microplate used for tissue submission. Three sets of primers were used (Table 2) to amplify one mitochondrial and two nuclear DNA markers (Table 3). All laboratory work related to DNA extraction, purification and sequencing was performed in a commercial laboratory “Canadian Center for DNA Barcode” (CCDB, http://www.ccdb.ca/) at the University of Guelph, Ontario, Canada, following the standard laboratory protocol (Ivanova et al. 2006, Ivanova et al. no date). Resulting sequences and additional relevant information such as gel images and trace files were uploaded to the “Barcode of Life Database” (= BOLD, http://www.boldsystems.org/).
Alignment and dataset concatenation. Each of three DNA markers (Table 3) was aligned using a different strategy. Alignment of the COI fragment was trivial and did not result in introduction of insertions or deletions (= indels). Alignments were checked for stop codons and frame shifts. Alignment of ITS2 and 28S was done using the online version of MAFFT 7 (Katoh et al. 2002; Katoh & Toh 2008a), with the Q-INS-i algorithm (Katoh & Toh 2008b) utilising the secondary structure information and resulted in introduction of 959 and 70 indels, respectively (Table 3). No parts of the alignment were excluded from the analysis. Three aligned single-fragment datasets were concatenated using Mesquite 3.11 (Maddison & Maddison 2011). The concatenated matrix contained 42% of gaps (mainly from indels in ITS2).
Analytical strategy, matrix design and phylogenetic analyses. Three analyses were designed and implemented: Analysis 1 (A1, phylogenetic) was designed with multiple goals to (a.) test monophyly of the herein hypothesised new clade of novel beetles from East Usambara, Uluguru and Udzungwa; (b.) if found to be monophyletic, then assess its relationships with the genus Typoderus, itself a taxonomic unit of questionable monophyly and (c.) if both form a clade, place it in the practically non-existing phylogenetic framework of Molytinae weevils. For these purposes, a matrix was created containing 46 terminals sequenced for three markers (COI, ITS2 and 28S; Table 4) and containing three terminals representing novel beetles from East Usambara, Uluguru and Udzungwa, three terminals representing three named Typoderus species, 30 other various Molytinae, nine non-Molytinae Curculionidae, and a member of the closely related family Dryophthoridae to root obtained topologies. Phylogenetic analysis was conducted using the CIPRES Science Gateway (Miller et al. 2010) using the maximum likelihood (ML) method. ML trees were obtained using RAxML 7.2.7 (Stamatakis 2006), with default parameters unless otherwise stated. The concatenated matrix was partitioned into three fragments and an independent GTR+G model was applied to each data partition. This evolutionary model is the most complex for nucleotide transitions, since it gives a different rate for each of them and accounts for rate heterogeneity (G). It is also the only model implemented in RAxML. The best scoring ML tree was selected among 100 searches on the original alignment with different randomized parsimony starting trees. Support values were obtained with 1000 bootstrap (Felsenstein 1985a) replicates as strong (> 75%), moderate (40%–75%) and low (< 40%). GenBank accession numbers for all 46 specimens are given in Table 4, while their locality data, specimen images, electropherograms and sequences can be found online in a public BOLD dataset dx.doi.org/10.5883/DS-LUPANG2.
Analysis 2 (A2, phylogeographic) was designed based on the results of A1 (corroborated sister-group relationships between the herein reported new clade and monophyletic Typoderus) and was aimed to assess the interrelationships within the new clade, including testing the hypothesis of reciprocal monophyly of its three geographical groups (from East Usambara, Uluguru and Udzungwa, respectively). For this purpose a matrix was created containing 31 ingroup terminals sequenced for the mitochondrial marker COI-5' and representing populations from East Usambara (12), Uluguru (7) and Udzungwa (12), plus a single representative of its sister group (Typoderus) to root obtained topologies. Phylogenetic analysis was conducted using the ML method implemented in MEGA7 (Kumar et al. 2016) with a GTR+G model determined in MEGA7 as having the best fit. Support values were obtained with 1000 bootstrap replicates. GenBank accession numbers for all 32 specimens are on the topology (Fig. 4), while locality data, specimen images, electropherograms and sequences can be found online in a public BOLD dataset dx.doi. org/10.5883/DS-LUPANG1.
Analysis 3 (A3, temporal) was designed based on the results of A2 (corroborated reciprocal monophyly of the three geographical groups) and was aimed to estimate relative and absolute time of the main evolutionary events leading to the present day diversity and distribution of the new clade. For this purpose the A2 matrix was reduced in size to only 11 terminals best representing the two most basal levels of branching inside each geographical clade, as detected in A2. Bayesian phylogenetic analyses in BEAST 1.8 (Drummond et al. 2012) was used to simultaneously estimate an ultrametric phylogenetic tree and ages of diversification. Lacking fossils and unambiguous biogeographical events to calibrate the phylogeny, a uniform a priori substitution rate was implemented. Two calibration schemes were used, each utilizing a different rate. The first calibration scheme was based on the rate of 0.018 nucleotide substitutions per site per million years per lineage (subs/s/Myr/l), in agreement with results obtained for COI-5’ in other beetles (Papadopoulou et al. 2010; Andújar et al. 2012), other insects (Brower 1994) and other arthropods (Bauzà-Ribot et al. 2012). The second scheme was based on the unusually high rate of 0.0793 subs/s/Myr/l estimated for the biologically most similar Trigonopterus Fauvel, 1862 weevils inhabiting forest litter of the Oriental region (analysis 2 in Tänzler et al. 2016), in agreement with the hypothesis that molecular evolution in flightless beetles, especially groups inhabiting stable habitats, might be highly accelerated (Mitterboeck & Adamowicz 2013). Monophyly of the East Usambara + Uluguru clade was enforced, following the topology obtained in A2. The GTR+G evolutionary model was used and the MCMC chains ran for 10 million generations. Consensus trees were estimated with TreeAnnotator (Drummond et al. 2012) discarding the 25% initial trees as a burn-in fraction, after checking ESS of likelihood, evolutionary rates and root age values, and ensuring that the tree likelihood values had reached a plateau. Posterior probabilities were considered as a measure of node support.
Taxonomic procedures. The ingroup organisms dealt with in this paper are new to science and it is therefore necessary to perform their formal taxonomic description. Since higher taxonomic categories do not exist objectively and need to be decided by the first reviser based on the best available evidence (Ward 2011), the following logic was implemented. The herein reported new clade, since morphologically easily diagnosable from its sister group genus Typoderus, should be ranked as a new genus. Its evolutionary lineages from East Usambara, Uluguru and Udzungwa, since reciprocally monophyletic and morphologically distinct, should be ranked as either one broadly defined, or three narrowly defined new species. Both approaches are logically equally valid by fully meeting all three primary taxon-naming criteria (Vences et al. 2013): monophyly of the taxon in an inferred species tree, clade stability and phenotypic diagnosability. The choice between one versus three species scheme should, therefore, be based on practical need to have names when referring to these organisms (Ward 2011). It should be explicitly stressed that logically sound species naming, like any assertion about a biological object, requires explicit knowledge of branching phylogeny (Felsenstein 1985b) and cannot be argued in oversimplified terms of “intraspecific genetic distance”. Since specimens representing three geographically determined lineages are easily distinguished morphologically, each lineage was given a rank of a Linnaean species. All three new species are mutually allopatric and it is, therefore, impossible to assess how effectively they can preserve their genetic and morphological distinctness in case of possible introgression with other herein described congeners. The situation of putative allopatric “species” repeatedly diversifying and then succumbing to introgression was recently termed “Sisyphean evolution” and illustrated on the iconic Darwin’s finches (McKay & Zink 2015). This Sisyphean scenario is likely widespread in nature and might, perhaps, be applicable to the new species described below. It is, therefore, important to fully embrace the understanding that any taxonomic arrangement represents the practical situation of today (= status quo) and should, therefore, be revised if and when new conflicting evidence came to light.
To expedite the formal descriptive taxonomic process, suggestions by Riedel et al. (2013) are followed. Each new species is illustrated by the standardized images of the holotype and its genitalia, and DNA barcoding data are provided. Species-level diagnostic descriptions are given by means of Table 5, which consistently lists all easily observed morphological differences. No absolute measurements are reported (which are variously and inconsistently measured in weevils, i.e. with or without rostrum and/or the head capsule visible from outside), which should be instead calculated from scale bars provided for the Holotype images (and from those provided for all herein reported specimens and accessible online on public BOLD dataset dx.doi.org/10.5883/DS-LUPANG1). The subfamily Molytinae is taxonomically defined following mainly Alonso-Zarazaga & Lyal (1999), and without the recent addition of Cryptorhynchinae (Oberprieler et al. 2007; Lyal 2014; for reasons see Riedel et al. 2016).
Lupangus gen. n.
Type species: Lupangus asterius sp. n., by present designation.
Diagnosis. Adult specimens of Lupangus can be immediately recognized among Molytinae weevils (including those of the genus Typoderus, its sister clade) by the combination of at least two easily observed characters: markedly vertical eyes about 4–5 x as high as wide (Fig. 4; not more than 2.5 x in Typoderus) and a deep transverse groove extending dorsally between the dorsal edge of eyes (Fig. 4; absent in Typoderus).
Description. Adult body robust and heavily sclerotized, dark coloured and medium-sized (about 5–7 mm between anterior edge of pronotum to elytral apices); head, body and legs with numerous thick and short yellowish to orange setae; body with numerous large punctures; pronotum and elytra between striae 2–3, 4–5 and 6–7 with longitudinal ridges bearing separate rounded elevations or sharp peaks. Head with eyes markedly vertical (about 4–5 x as high as wide); cornea of ommatidia markedly globular; rostrum delimited posteriorly by deep and narrow dorsal transverse groove extending between dorsal eye corners; rostrum with variously developed longitudinal grooves; antennae with scapus, funicle with 7 antennomeres and club with three antennomeres. Prothorax without postocular lobes or prosternal channel; procoxae moderately separated by prosternal process and closed posteriorly; mesocoxae moderately and metacoxae markedly separated. Hind wings absent. Elytra interlocked with meso- and metathorax and among themselves; each elytron with ten rows of punctures. Aedeagus short, bent and cylindrical, with ventrally-directed hair in apical part.
Species composition and distribution. The genus Lupangus consists of three new allopatric species restricted to the Eastern Arc Mountains in Tanzania. Elevation: 501–1921 m.
Biology. All known specimens of Lupangus were detected by sifting floor litter in wet primary Afromontane forests. Host plants, immature stages, parasites or any other biological aspects remain unknown.
Etymology. Toponymic, after Lupanga, one of the principal peaks of Uluguru; gender masculine.
Lupangus asterius sp. n.
Distribution. This species is known only from East Usambara, Tanzania. Elevation: 501–1020 m.
Etymology. The species epithet is a Latinized Greek mythical name of Asterius, an Argonaut from Thessalia; noun in apposition.
Material examined. Holotype, male (CNC), specimen #3060, Tanzania: “TANZANIA, E Usambara Mts., Amani NR, 5°10'34''S 38°36'01''E, 15.xii.2011, 1004m, sift.05, V.Grebennikov”. Paratypes (CNC): 11, as in Fig. 4.
Lupangus jason sp. n.
Distribution. This species is known only from Uluguru, Tanzania. Elevation: 1569–1921 m.
Etymology. The species epithet is a Latinized Greek mythical name of Jason, the leading Argonaut, husband of Me deia; noun in apposition.
Material examined. Holotype, male (CNC), specimen #3636, Tanzania: “TANZANIA, Uluguru Mts., Lupanga Peak, 6°51'54''S 37°42'28''E, 10.i.2012, 1921m, sift.27, V.Grebennikov”. Paratypes (CNC): 6, as in Fig. 4.
Lupangus orpheus sp. n.
Material examined. Holotype, male (CNC), specimen #7714, Tanzania: “TANZANIA, Udzungwa Mts., –7.8419 36.8546, 1083m, 7.x.2014, sift03, V.Grebennikov”. Paratypes (CNC): 11, as in Fig. 4.
Distribution. This species is known only from Udzungwa, Tanzania. Elevation: 1083–1693 m.
Etymology. The species epithet is a Latinized Greek mythical name of Orpheus, an Argonaut, a magically talented musician; noun in apposition.
Results of three DNA analyses
Analysis A1 resulted in a topology (Fig. 3) with monophyletic Lupangus a sister to monophyletic Typoderus; all three clades are strongly supported. The rest of the topology is poorly resolved, with only a few clades showing strong support. The majority of these clades are formed by a few presumably most closely related terminals, such as genera each represented by more than one terminal or by pairs of genera such as Trachodes Germar, 1824 and Acicnemis Fairmaire, 1849, or by both Cryptorhynchinae genera. Cossoninae genera do not form a clade, two of them (Trichopentathrum Osella, 1976 and Caenopentarthrum Voss, 1965) are strongly linked to the molytine genus Otibazo Morimoto, 1961. Lepyrus Germar, 1817 is strongly supported as the sister of Plinthus Germar, 1817, and both of them to Adexius Schoenherr, 1834. Entiminae and the rest of Curculionidae are reciprocally monophyletic, except that the Molytinae Prothrombosternus Voss, 1965 is nested within Entiminae.
Analysis A2 resulted in a topology (Fig. 4) with three clades of Lupangus corresponding to three newly described allopatric species, and with L. orpheus from Udzungwa being the sister to the rest. While specimens of L. asterius (from East Usambara) exhibit no phylogeographic structure (note that the specimens of sample EU08 define the basal-most dichotomy), both other species have such a structure. That of L. orpheus (from Udzungwa) is moderately pronounced with specimens from sample UD03 forming a shallow sister clade to the rest of the species sampled from the opposite side of Mt. Mwanihana some 400–600 m higher and about 5 km away. Phylogeographic structure of L. jason is the strongest, with the single specimen #3636 (Holotype) being the deeply divergent sister to the rest.
Analysis A3 resulted in two topologically identical trees (Fig. 5) with different timescales. The slow evolutionary scheme (0.018 subs/s/Myr/l) suggested separation of mitochondrial lineages representing all three allopatric Lupangus species taking place at about 7–8 Ma, while the fast scheme (0.0793 subs/s/Myr/l) placed these events at about 1.9–2.1 Ma. The basalmost splits inside each of three species are dated between about 3 Ma (in L. jason under slow scheme) and about 0.25 Ma (in L. orpheus under fast scheme).
Clade of Lupangus and Typoderus in unresolved Molytinae. The main result of analysis A1 is the recovery of a strongly supported clade formed by reciprocally monophyletic Lupangus and Typoderus. This is the second pair of reciprocally monophyletic Afrotropical Molytinae genera convincingly shown to form a clade using phylogenetic analysis of DNA data; the other pair is Amorphocerus Schoenherr, 1826 and Porthetes Schoenherr, 1838 constituting the tribe Amorphocerini and known to develop exclusively on cycads of the genus Encephalartos (Zamiaceae) (Downie et al. 2008). The rest of Afrotropical Molytinae, consisting of many dozens, if not hundreds of genera (Alonso-Zarazaga & Lyal 1999), remains in the painful state of not only phylogenetic, but basic taxonomic neglect and obscurity (Grebennikov 2015a, 2016a). The herein analysed genus Typoderus (Fig. 1D, E) might be a good example of neglected genera, with all of its 11 nominal species known only from the original description published in the short period between Marshall (1953) and Voss (1965). Specimens of this genus were, however, exceedingly common and diverse in the majority of the 130 litter samples (Table 1), suggesting that the real species diversity (and its phylogeographic potential) is much higher. Both morphological characters of Lupangus stressed in the diagnosis (Fig. 4) seem autapomorphic to this genus, leaving Typoderus without known morphological support. From the presently released data one might suspect that Lupangus is an unusually shaped Typoderus sister to the three species represented in the analysis, but subordinate in Typoderus if the remaining species of this genus are considered. This, however, is unlikely, since analyses of a significant amount of unpublished data (about 400 Typoderus specimens sequenced for COI-5’ and about 200 of them sequenced also for ITS2 and 28S, data not shown) consistently resulted in a monophyletic Typoderus excluding Lupangus.
Molytinae weevils lack a comprehensive molecular phylogeny compared with those proposed for some other comparably large weevils subfamilies such as Platypodinae (Jordal 2015), Cryptorhynchinae (Riedel et al. 2016) or arguably the best studied economically important Scolytinae (reviewed in Kirkendall et al. 2015). Until now, either a few Molytinae representatives were included in broader multi-marker analyses (i.e. McKenna et al. 2009), or a larger subset of Molytinae genera was analysed using a single marker (Grebennikov 2014a,b). Each approach is limited in either coverage, or rigour, or perhaps in both and, therefore, developing of a robust Molytinae phylogeny (if indeed monophyletic, see Riedel et al. 2016) is still a pending task. The herein reported tree of Molytinae (Fig. 3) is an extended version of the barcode-only topologies from Grebennikov (2014a,b) with addition of a few terminals and of two nuclear ribosomal markers: ITS2 and 28S. The tree, even though with low support values, appears plausible, since with the exception of two Leiosoma Stephens, 1829, all a priori most closely related terminals (congeners, or closely related genera such as Trachodes and Acicnemis) predictably form strongly supported clades. The backbone resolution of the tree is, however, not much better than those obtained earlier using a single mitochondrial marker. Only a few non-anticipated clades are recovered, and some of them with low statistical support, for example Mediterranean Aparopion Hampe, 1861 + Anchonidium Bedel, 1884. Afrotropical Aparopionella Hustache 1939, once thought to be sister to Typoderus (Marshall 1953), does not group with other Typoderina, thus rejecting monophyly of the subtribe (sensu Alonso-Zarazaga & Lyal 1999). The recently described genus Morimotodes Grebennikov, 2014 is again (Grebennikov 2014b) recovered forming a clade with one of two included Leiosoma, while the latter genus is surprisingly not monophyletic. Novel is the strongly supported clade Adexius + (Lepyrus + Plinthus). Unexplainable is the paraphyly of three members of Entiminae with respect to Prothrombosterus (Molytinae). The strongly supported sister-group relationship between Otibazo and two poorly known Cossoninae genera suggests that the latter might be taxonomically misplaced. Overall, results of analysis A1, besides resolving the Lupangus + Typoderus clade, provide relatively little novel information. This can be attributed to the inadequate set of three markers (some of them, like COI and ITS2 perhaps too fast evolving) unable to resolve relatively deep phylogeny of the selected terminals.
Pre-Pleistocene vicariance best explains Lupangus speciation and distribution. The topology obtained in analysis A2 with all three Lupangus species reciprocally monophyletic and strictly allopatric strongly suggests the simplest phylogeographical scenario of speciation through the normal ecological dispersal (Heads 2014) with subsequent subdivision into three lineages by wet forest fragmentation (= vicariance). Such interpretations are most commonly inferred for other EAM clades analysed in sufficient detail (Kinyongia chameleons by Tolley et al. 2011; Trioceros chameleons by Ceccarelli et al. 2014; Praomys rodents by Bryja et al. 2013; Parepistaurus flightless grasshoppers by Hemp et al. 2015). Timing and possible causes of this process pertaining to Lupangus remain, however, highly elusive due to the lack of a reliable time calibration. The herein implemented two flat evolutionary rates differ about four times (slow versus fast; see Methods and Fig. 5) and currently no information is available to allow choosing one of them. Adding to this the unavoidable uncertainty of the 95% confidence intervals (Fig. 5) deprives the herein implemented dating of most of its precision. Even though widely varying, the timing results consistently suggest that with 95% probability Lupangus allopatric speciation, even if estimated with the staggeringly fast evolutionary rate of 0.0793 subs/s/Myr/l, took place not later than about 1.3 Ma, and perhaps not later than 1.8 Ma (Fig. 5). These time points are in mid- and early-Pleistocene, respectively, and predate the most dramatic shrinkages and expansions of the African wet forest during the last million years (Hamilton & Taylor 1991). Application of the slow evolutionary rate (0.018 subs/s/Myr/l) pushes Lupangus speciation about four times deeper in the past and before the onset of the Pliocene-Pliocene climatic cycles. So widely interpreted, the results consistently suggest that the hypothesised ecological dispersal of the most recent common ancestor of Lupangus and its subsequent speciation through forest fragmentation took place only once and before the most pronounced dry/wet cycles of the last million years. This, in turn, suggests that during at least the last million years the local climate was not wet enough to permit forest expansion sufficiently pronounced to re-connect EAM forests into a single forested block. Such timing agrees with that estimated, for example, by Tolley et al. (2011) and much predates that of Hemp et al. (2015). Observed dating disagreements are, however, fully expected, since evolutionary history of each clade, even if most similar in dispersal capacities to many others occurring sympatrically, is expected to be fully unique through a combination of numerous stochastic evolutionary events.
Basal splits in two among three Lupangus species (Figs 4, 5) are geographically structured, rejecting panmixia. This can be fully expected for low-dispersing organisms, so the specimens of L. orpheus from the sample UD03 might form the sister group to the rest of the species sampled on the other side of Mt. Mwanihana only about 5 km away. Phylogeographic structure inside L. jason from Uluguru is even more pronounced, with the Holotype from Lupanga peak forming the deeply divergent sister to the analysed rest of the species sampled some 18 km southwards at the Bunduki village. The timing of the latter split using fast evolutionary rates suggests that with 95% probability it occurred not later than 0.5 Ma, and likely as early as 0.8 Ma (Fig. 5), which seems relatively old. It should, however, be remembered, that only the 5P fragment of the mitochondrial COI gene was used for such estimations, therefore the reported results might at least partly be linked to the phenomenon of maternal inheritance with all its known analytical advantages and shortcomings (Funk & Omland 2003).
Surprisingly and in spite of dense sampling, Lupangus beetles were documented in only three among nine studied EAM blocks, and in none among three volcanic and two lowland forests (Fig. 1A). Part of these absence data might be the sampling artefact of randomly failing to detect the beetles because of their highly fragmented distribution through seemingly uniform forests. This hypothesis gains support in an observation that no specimens of L. jason endemic to Uluguru were discovered in the densely sampled Uluguru forest adjacent to Lukwangule Plateau (samples UL01–12, see Table 1). An alternative explanation to the absence of Lupangus in other Tanzanian forests is that the suitable habitat was perhaps irreversibly lost due to anthropogenic changes (= human-driven extinction). Both assumptions, although not explicitly tested herein, appear unlikely, since the sampling in each forest was geographically diversified, while all sampled forests (with the exception of that on Pugu Hills, Fig. 1A) appeared large and healthy enough to support these beetles. Absence of Lupangus in forests of all three volcanic highlands (Fig. 1) might perhaps be linked to the relatively young age of these forests not pre-dating volcanic activities responsible for forming these highlands and commencing about 2–3 Ma (Nonnotte et al. 2008). Thus, unless otherwise demonstrated, the observed seemingly non-random distribution of Lupangus through 14 sampled Tanzanian forests (Fig. 1A) can most plausibly be attributed to two main factors: exceedingly low dispersal capacity coupled with highly stochastic nature of the colonizing/surviving events.
Inadequate taxonomy impedes unlocking phylogeographic potential of flightless weevils. Similarly to perhaps all presumably low-dispersing organisms, flightless weevils such as Lupangus are highly suitable but underutilized model organisms for testing competing phylogeographical hypotheses, particularly those pertaining to the dispersal versus vicariance dilemma. Their usefulness is compromised, however, by often acutely inadequate taxonomic knowledge (Riedel et al. 2010; Tänzler et al. 2012), making their use unpractical due to the lack of, or confusion in, their Linnaean names. In the present study, for example, the phylogeographical hypothesis is derived from a lineage entirely new to science, which, therefore, has to be first formally named and described according to the rules of the International Code of Zoological Nomenclature (ICZN 1999). This additional taxonomic burden seems, however, a much lesser hazard when compared to more numerous situations in which inadequately attributed and often synonymous historical taxonomic names making reference to organisms practically impossible (“clogging taxonomy”; Grebennikov 2016b).
Critical dependency of all biological assertions on having explicit knowledge of organisms’ branching phylogeny is a logically unavoidable requirement (Felsenstein 1985b). The taxonomic impediment might perhaps be obeyed and eventually solved though the practice of classical taxonomic Holotype-based revisions (Riedel 2011; Riedel & Tänzler 2016). This task is time consuming, cannot be significantly automated, requires high-end skills and, therefore, is becoming more expensive than the more and more democratized discovery and documentation of the inner branches on the Tree of Life (Maddison 2016). An alternative solution might be adoption, at least temporary, of a non-Linnaean DNA-based nomenclature (Ratnasingham & Hebert 2013) free of the historical taxonomic burden. Which way the biological science will choose to develop will soon be decided empirically.
Most of pioneering phylogeographical work utilizing flightless weevils (mainly to address the dispersal vs. vicariance dilemma) was done within the last decade focussing, predictably, on oceanic islands, such as those in Australasia near the Wallace line (Tänzler et al. 2014, 2016; Toussaint et al. 2015), Pacific Islands (Claridge et al. 2017), Macaronesia (Stüben & Astrin 2010) including the Canary Islands (Emerson et al. 2006; Faria et al. 2016; Machado et al. 2017), the Caribbean archipelago (Zhang et al. 2017), subantarctic islands (Grobler et al. 2011), Galapagos Islands (Sequeira et al. 2000, 2008) and Mauritius (Kitson et al. 2013). Attempts to extend this approach to the “sky island” faunas of the continental landmasses are few in comparison and target Europe (Meregalli et al. 2013) or Asia (Grebennikov 2014a,b, 2015b, 2016b; Grebennikov & Kolov 2016; Grebennikov & Morimoto 2016). The present work is the second DNA-based phylogeographic attempt utilizing flightless sub-Saharan weevils, following that of Grebennikov (2015a). Much of the delay in using these otherwise highly informative organisms might be apportioned to their repellently inadequate taxonomy. The latter was not based on phylogenetic principles and presently consists of poorly documented names, often synonymous at the species-, and even more so at the genus-group levels. It is symptomatic that the herein reported analysis deals with the genus entirely new to science and thus not haunted by the problem of the split identity, while two other weevil-focussed papers from the same litter-sampling program (Grebennikov 2015a, 2016a) target the previously monotypic genera. Other seemingly equally informative genera of weevils containing handfuls of named species and rediscovered in the samples (among them numerous, large and charismatic Typoderus) can be reported not before identity of the historical names becomes known through the type specimen examination. With relatively few names among the sub-Saharan litter weevils, this might be a manageable task, although preliminary results suggests that species- (Grebennikov 2015a) and particularly genus-level synonymy is rampant, with at least one exceedingly common and widespread “genus” having over 10 unrecognized synonyms. Similar taxonomic chaos was recently reported for flightless and phylogeographically informative Catapionus Schoenherr, 1842 and Notaris Germar, 1817 weevils in Asian highlands (Grebennikov 2016b; Grebennikov & Kolov 2016), although the main challenge there was the elusive identity of the species-, rather than that of the genus-group names. Once the taxonomic impediment is overcome one way or another (see above), flightless weevils (and indeed any among the low-dispersing and numerically abundant invertebrates) will be ready to have their phylogeographic potential fully unlocked.
Wet forests of the Eastern Arc Mountains: the gem of mainly untapped biodiversity. The Eastern Arc Mountains are arguably among the most biodiverse places on the Earth. Such an assessment is difficult to substantiate with reliable faunal and floral data, since the planet’s biodiversity is unevenly and incompletely documented. The latter statement is corroborated by the herein reported discovery of the narrowly endemic weevil genus entirely new to science. Tanzania, however, emerged as the most biodiverse continental country for chameleons (Tilbury 2010), with the majority of their diversity consisting of highly endemic species inhabiting EAM forests, and their number steadily increasing (Ceccarelli et al. 2014). Faunal surveys in EAM predominantly targeted the taxonomically better-known vertebrates (Rovero et al. 2014), and among them the relatively low-dispersing and habitat dependant amphibians and non-avian reptiles (= “herpetofauna”; Menegon et al. 2008). Invertebrates, and particularly the “cryptofauna” (Lawrence 1953; Leleup 1965) for the forest floor and the upper soil layer remain practically unknown. The most preliminary assessment of Arthropod diversity as seen when sorting the herein reported 130 litter samples suggests presence of numerically overwhelming and genetically diversified mesofauna, much of which is seemingly similar to Lupangus in having pronounced fine-scale phylogeographic structure. Only few among the sampled specimens were reported, mainly beetles (first apterous male of Lycidae by Bocak et al. 2014) and true bugs (Ulugurocoris Štys & Baňař 2013, the first Afrotropical Aenictopecheidae, by Štys & Baňař 2013; extremely sexually dimorphic Xenocaucus China & Usinger, 1949 assassin bugs, by Weirauch et al. 2017), and in every case new taxa and/or informative phylogeographic patters were detected. Other numerically abundant groups of the forest floor inhabitants, such as Formicidae or Acari, remain mainly unsorted and underutilized. They, and not the an-thropocentrically more appealing vertebrates, constitute most of the genetic diversity in EAM and contain great and still mainly untapped potential for research on evolutionary biology.