First DNA analysis of pill scarabs ( Coleoptera : Hybosoridae : Ceratocanthinae ) reveals multiple paraphyly of Afrotropical

This paper is the fi rst attempt to resolve relationships among the Ceratocanthinae: Ceratocanthini pill scarab beetles using DNA sequences. It is focused on the Philharmostes group of seven Afrotropical genera: Baloghianestes (3 spp.), Callophilharmostes (1 sp.), Carinophilharmostes (1 sp.), Chaetophilharmostes (1 sp.), Cryptophilharmostes (3 spp.), Petrovitzostes (1 sp.) and Philharmostes (31 spp.). A phylogenetic analysis of 46 terminals and alignment of 2,913 bp from one mitochondrial and two nuclear fragments corroborates monophyly of this group, but rejects that of Philharmostes, the largest genus. The latter is paraphyletic with respect to at least four other smaller genera and consists of at least three distantly related clades. One of them, formed by Philharmostes ballerioi sp. n. from the Tanzanian Nguru (the type locality) and Kaguru Mountains, is sister to the rest of the entire Philharmostes group. The nominal genus Philharmostes is, therefore, a waste-basket taxon for accommodating members of this group that lack the distinct characters of the smaller genera. Pending further research, the phylogenetically inadequate generic taxonomy of the Philharmostes group is not modifi ed. Molecular clock analysis estimates separation of the mitochondrial lineages of two known populations of the new species at about 2.2 Ma, which corresponds with recurring shrinkage and expansion of African rainforest caused by climatic fl uctuations during the Pleistocene. Adults of all nominal ingroup genera are illustrated along with male and female body parts of the new species. Diagnostic and/or synapomorphic morphological characters of the Philharmostes group of genera are revised. Habitus images and other supplementary information on all sequenced specimens are available online at dx.doi.org/10.5883/DS-VGDS001 and dx.doi.org/10.5883/DS-VGDS004. ZooBank Article LSID: EFB8BEFA-0658-4CFB-8720-4A778D593BB6


INTRODUCTION
Ceratocanthini (the largest of the three currently recognized tribes in the subfamily Ceratocanthinae) constitutes a pantropi cal clade of 38 extant nominal genera with some 358 species (Ballerio & Grebennikov, 2016).They are best known for their ability to conglobate (= to pack their body into a tight spheroid) using interlocking exoskeletal structures, thus earning the entire subfamily the colloquial name "pill scarabs" (Howden & Gill, 2000).Their adults are mainly found in tropical forests at low altitudes.Flightless species are sometimes abundantly sifted from forest fl oor litter, while volant species are sampled by canopy fogging or using fl ight intercept traps.Adults and larvae of pill scarabs occur in rotten trees (Choate, 1987), sometimes in close proximity with ants and/or termites, however, this has never been documented as biologically signifi cant (Parker, 2016).All three main tropical regions of the World (Neotropical, Afrotropical and those in Asia and Australia) have highly distinct pill scarab faunas with no genera in

Specimen sampling and deposition
The majority of the specimens of Ceratocanthini included in this study were sifted from forest litter in Tanzania (list of localities and sample codes are in Grebennikov, 2017;Grebennikov & Heiss, 2018).In addition, (mainly outgroup) specimens were opportunistically accumulated from a variety of sources (see Acknowledgments).Unless otherwise stated, all herein reported specimens are stored in the Canadian National Collection of Insects, Arachnids and Nematodes in Ottawa, Canada (CNC, curator P. Bouchard).

Preliminary DNA barcoding
A total of 149 freshly sampled Afrotropical Ceratocanthini (and those received from other sources) were preliminary sorted into morphospecies and DNA barcoded (658 bp of COI-5' sequenced; Hebert et al., 2003a, b; all data are not shown).DNA barcoding followed the standard protocol of the "Canadian Centre for DNA Barcoding" at the University of Guelph, Canada (CCDB, http:// www.ccdb.ca).Alignment of the protein-coding COI was trivial and did not result in insertions/deletions (= indels), stop codons or frame shifts.Their Neighbour Joining (NJ) clustering (topology is not shown) was performed online using the Barcode of Life database (= BOLD, Ratnasingham & Hebert, 2007) engine (http:// www.boldsystems.org/).

Three-marker dataset
Select DNA barcoded specimens representing all the terminal PhG clusters (= putative species) in the NJ tree were additionally sequenced for two nuclear ribosome-coding regions: internal ribosomal spacer 2 (ITS2) and 28S rDNA (Table 1).Among seven nominal genera of PhG (= ingroup, for the purpose of this phylogenetic analysis), DNA sequences from 32 terminals were obtained, including multiple representatives of all three non-monotypic genera (Baloghianestes, Cryptophilharmostes, Philharmostes) and the type species of two monotypic genera (Carinophilharmostes and Petrovitzostes).No DNA-grade specimens could be obtained for the genera Callophilharmostes and Chaetophilharmostes so these taxa were not included in the analysis.To test the monophyly of PhG and place it into a broader phylogenetic framework, 12 other Ceratocanthini non-PhG genera, each represented by a single terminal, were added to the dataset.Monophyly of Ceratocanthini and their placement within monophyletic Hybosoridae are well established based on analyses utilizing larval morphology (Grebennikov & Scholtz, 2004;Grebennikov et al., 2004), DNA (Ocampo & Hawks, 2006) and adult morphology (Ballerio & Grebennikov, 2016).This hypothesis is herein accepted a priori by adding to the matrix two representatives of non-Ceratocanthini Hybosoridae (Table 2) to root the topology.All laboratory work was done in CCDB using protocols and primers described in Grebennikov (2018).All relevant laboratory data (such as electropherograms, sequences, specimen images and their localities) are available online in BOLD dataset dx.doi.org/10.5883/DS-VGDS001; the GenBank accession numbers are listed in Table 2.All DNA sequences used in this study were newly generated.Kolbe, 1895 (31, Figs 2A-E, 4, 5).Monophyly of PhG appears likely because of the number of shared diagnostic characters (Ballerio, 2000(Ballerio, , 2001) ) and its relatively high bootstrap support (85% in Ballerio & Grebennikov, 2016).The largest PhG genus (Philharmostes) is, however, hypothesized to be paraphyletic with respect to at least one (Baloghianestes, see Ballerio & Grebennikov, 2016) and perhaps more, small genera.This scenario is indeed likely, since species of all six small PhG genera and those of Philharmostes are sympatric and their biological preferences are similar.Moreover, three of the four monotypic genera (Callophilharmostes, Carinophilharmostes, Chaetophilharmostes) were established for nominal species originally described in Philharmostes (Paulian, 1968(Paulian, , 1977)), which is refl ected in their unwieldy names.It seems likely, therefore, that the species-rich genus Philharmostes is a paraphyletic waste-basket taxon used for those members of PhG that lack distinct diagnostic characters of the six other small genera.
This paper was triggered by the availability of many representatives of PhG recently collected in Africa, mainly Tanzania.Freshly collected specimens of four small genera of PhG, along with numerous "typical" Philharmostes from most of their distribution (West, East and South Africa, as well as Madagascar, map in Fig. 3) offered an opportunity of testing the taxonomically-implied reciprocal monophyly of nominal genera.Furthermore, two populations from the adjacent Nguru and Kaguru mountains in Tanzania (map in Fig. 6) belonging to a new species which in the present classifi cation should be included in Philharmostes are not similar to the rest of the genus (smaller in size, nearly spherical when enrolled and lacking dorsal eyes).The fi rst and main goal of this paper is, therefore, to perform the fi rst DNA-based phylogenetic analysis of Ceratocanthini pill scarabs focussing on PhG and to test two hypotheses: monophyly of the entire group and that of its largest genus Philharmostes, particularly with respect to the small genera; and to determine the placement of the new Nguru/Kaguru species.The second goal of this paper is to formally describe this new species and, considering its newly hypothesized sister-group relationships with the rest of PhG (see Results), thoroughly illustrate its external and internal structures for use in future comparative morphological studies.The third goal, also triggered by the remarkably ancient relationships of this new species, is to review all diagnostic and/or synapomorphic morphological characters of PhG, as previously defi ned, and to determine which of them might be detected in the new species.The fourth and fi nal goal is to test reciprocal monophyly of both neighbouring populations of the new species and to date their separation.This is intriguing, because this date might indicate when the currently small and compact wet rainforests on Nguru and Kaguru became separated by hot and dry savannah (presently the gap is about 65 km).Overall, this paper strives to shed new phylogenetic light on Ceratocanthini, a diversifi ed, easy-to-recognize and relatively neglected branch of the scarabaeoid beetles.

Alignment of ribosomal markers
Alignment of the ITS2 and 28S sequences was done using the MAFFT 7 online platform (Katoh et al., 2002;Katoh & Toh, 2008a) and the Q-INS-i algorithm (Katoh & Toh, 2008b) utilising the secondary structure information and resulted in the introduction of 638 and 93 indels, respectively (Table 1).To minimize bias, no parts of the alignments were excluded from the analysis.Three aligned single-fragment datasets were concatenated using Mesquite 3.11 (Maddison & Maddison, 2011) into a matrix of 46 terminals and 2,913 aligned positions containing 45% of the completely undetermined characters (mainly due to numerous indels in ITS2).

Analysis of the three-marker dataset (= phylogenetic analysis)
Phylogenetic analysis was done on the CIPRES Science Gateway online platform (Miller et al., 2010) using the Maximum Likelihood (ML) method.Phylogenetic trees were obtained using RAxML 7.2.7 (Stamatakis, 2006), with default parameters, un-less otherwise stated.The concatenated matrix was partitioned into three fragments (Table 1) and an independent GTR + G model (the only one implemented in RAxML) was applied to each data partition.The best scoring ML tree was selected among 1000 searches on the original alignment with different randomized parsimony starting trees.Support values were obtained based on 1000 bootstrap replicates (Felsenstein, 1985;Stamatakis et al., 2008).

Temporal analysis of DNA barcodes
The second analysis was designed to test the reciprocal monophyly of the Nguru and Kaguru populations of the herein described new species and if so, to estimate when they ceased exchanging genetic material.Lacking fossils and unambiguous biogeographical events to calibrate the phylogeny, a uniform substitution rate was implemented.A fl at molecular clock of 0.018 nucleotide substitutions per site per million years per lineage (subs/s/Myr/l) was applied to the DNA barcode fragment, which is in agreement with results obtained for other beetles (Papadopoulou et al., 2010; Andújar et al., 2012).A matrix was formed from 13 barcodes of the new species, of which six and seven were for the Nguru and Kaguru populations, respectively.Bayesian phylogenetic analysis in BEAST 1.8 (Drummond et al., 2012) was used to simultaneously estimate an ultrametric phylogenetic tree and ages of di-versifi cation.The TN93 evolutionary model (estimated in MEGA 7, Kumar et al., 2016) was applied and the MCMC chains ran for 10 million generations.Consensus trees were estimated using TreeAnnotator (Drummond et al., 2012) after discarding 25% of the initial trees as a burn-in fraction, after checking ESS of likeli- hood, evolutionary rates and root age values, and ensuring that the tree likelihood values had reached a plateau.Posterior probabilities were used as a measure of node support.No monophyletic groups were enforced prior to the analysis, and the rooting was done on the longest branch.

Morphological study
In order to illustrate the adult morphology of the members of PhG, two imaging strategies were followed.Firstly, high-resolution habitus images were generated for specimens of all seven nominal genera of PhG (Figs 1A-E, 2A-F).These images were of dry-mounted specimens and based on multiple imaging with subsequent digital deep-focus stacking.Unlike the majority of earlier pill scarab illustrations of artifi cially descended (= unrolled, fl attened) specimens (see, for example, Ballerio & Gre-bennikov, 2016), the images are of enrolled specimens, as they are most commonly seen by humans.Secondly, and in view of its phylogenetic position as the sister to the rest of PhG (see Results), morphological structures of the herein newly described species are extensively illustrated and those of both sexes compared.For this purpose, male (paratype 3751) and female (paratype 7055) of P. ballerioi sp.n. were disarticulated and extensively imaged (Figs 4A-Y, 5A-J).Disarticulated body parts were photographed submerged in glycerol, while the genitalia of both sexes were fi rst macerated in a warm 5-10% solution of KOH in water and then stained in Chlorazol black.Since the newly described species cannot be reliably sexed without dissecting the genitalia, which resulted in signifi cant damage to the body, the holotype was selected from structurally intact and, therefore, unsexed specimens.No specimens from Kaguru Mts. were dissected.

Morphological terms
The terms "genal canthus" and "dorsal ocular area" (Ballerio & Grebennikov, 2016), which refer to the horizontally oriented ridge completely dividing (in some members of PhG) the compound eye, are consistently referred to as "interocular bridge" and "dorsal eye", respectively, in line with their usage in other Fig. 3. Maximum Likelihood inference phylogram of Ceratocanthini pill bugs, which reveals that the genus Philharmostes is paraphyletic with respect to other members of the monophyletic Philharmostes Group (Analysis 1).Orange branch represents the herein described new species from Tanzania, sister to the rest of the Philharmostes Group.Digits at internodes are bootstrap values > 65%.Map indicates known distribution of the Philharmostes Group with all of its seven nominal genera found in the tropical belt of continental Africa, while only the nominal Philharmostes is additionally known from both South Africa and Madagascar (approximated from Ballerio & Grebennikov, 2016).Red dots denote imaged specimens (to scale).beetles (i.e.Gyrinidae: Beutel et al., 2017).Terminology of male genitalia and associated structures mainly follows D'Hotman & Scholtz (1990), therefore the term "basal piece" is used instead of "phalobasis" (Ballerio, 2016)."Genital segment" of D'Hotman & Scholtz (1990) is, however, referred to by its anatomical term "abdominal segment 9" (exact homology of its sclerites is un- known); its structures referred to as "manubrium" and "basal triangle" (i.e.Ballerio et al., 2011) are shown in Fig. 4Y."Apophyses of parameres" (Fig. 4Y) are illustrated in Ballerio (2000, Fig. 6).All other morphological terms are the same as in Ballerio & Grebennikov (2016), including "enrolment coaptations", which defi nes beetles' ability to fold their body into a tight spheroid without gaps or protruding appendages.

RESULTS
The ML phylogenetic analysis of 46 terminals analyzed using the concatenated 2,913 bp matrix produced a wellresolved topology (Fig. 3).This analysis recovered the monophyletic PhG (bootstrap support 100%) sister (86%) to the Afrotropical genus Congomostes Paulian, 1968.Monophyletic PhG consists of six clades (A-F, Fig. 3), all having 100% support (except clade C with 56%).The genera Cryptophilharmostes and Baloghianestes (clades A and D, respectively) are monophyletic (both with 100%).The genus Philharmostes consists of three not most closely related clades (C, E and F); Philharmostes from Madagascar form the basal dichotomy of clade C, while clades E and F consist of small-bodied species.The genera Petrovitzostes and Carinophilharmostes, each represented by a single terminal, form clade B (100%) sister (100%) to Philharmostes clade C; the resulting clade B + C is sister (90%) to the genus Carinophilharmostes.The herein described new species (clade F) is sister (66%) to the rest of PhG.
The Bayesian temporal analysis of 13 terminals of the new species using the 658 bp DNA barcode matrix, recovered two geographically structured clades from the Nguru and Kaguru Mountains (Fig. 6), both with 100% posterior probabilities.Separation of their mitochondrial strains is dated at 2.21 million years ago (Ma), while diversifi cation within them started 0.45 and 0.13 Ma, respectively (Fig. 6).

Genus Philharmostes Kolbe, 1895
Kolbe, 1895: 344.Diagnosis.In the key to genera of Afrotropical pill scarabs (Ballerio & Grebennikov, 2016) this species is identifi ed as belonging to the Philharmostes Group (couplets 15-20) by having (1.) the enrolment coaptation and (2.) broadly arcuate protibia, with the outer margin nearly even and without distinct teeth.Lacking the unique dorsal head trichome of Callophilharmostes, this species in lacking a dorsal eye will not key out to Philharmostes (all species of which have a dorsal eye), but to Baloghianestes (because of its similarly shaped dorsal head contour, which is unlike that of the quite dissimilar Carinophilharmostes).The new Tanzanian species differs from the exclusively West African Baloghianestes by having an even, uninterrupted and sharp lateral carina on each elytron; in the latter genus it is either absent (the type species) or notably serrate in dorsal view (other species).As its phylogenetic position is far outside Baloghianestes, it might be warranted to erect a new genus for this species; this, however, is postponed until its relationships are better understood.The new species is, therefore, provisionally assigned to the (already) non-monophyletic Philharmostes (see Discussion).
Etymology.This species is named after Alberto Ballerio in recognition of his extensive work on pill scarabs that prompted my interest in this tribe.
Distribution.Known only from the Nguru and Kaguru mountains in Tanzania (Fig. 6), two discrete sky-islands of wet, closed-canopy rainforests separated by at least 65 km of arid savannah and forming a part of the exceptionally biodiverse chain of the Eastern Arc Mountains.Altitude: 623-1,229 m at Nguru and 1875 m at Kaguru.

Monophyly and sister group of the Philharmostes group of genera
The herein recovered and strongly supported monophyly of the Philharmostes group of pill scarab genera is not surprising, since it was twice hypothesized based on morphological similarities (Ballerio, 2000(Ballerio, , 2001)), and then supported by a morphology-based quantitative analysis (Ballerio & Grebennikov, 2016).The sister group relationships between PhG and a clade represented here by Congomostes is more controversial.The bootstrap support of this clade (86%) is moderate, while the analysis is lacking representatives of the 21 remaining nominal Ceratocanthini genera.Previously, PhG was recovered as sister to a weakly supported clade of eight Afrotropical and Neotropical genera (Ballerio & Grebennikov, 2016), fi ve of which are included in the present analysis and do not form a clade.Considering these discrepancies, it is best to conclude that the sister group of PhG remains unknown.

Morphological synapomorphies of the Philharmostes group of genera
Since P. ballerioi sp.n. is sister to the rest of PhG, it offers an opportunity to review known morphological synapomorphies of the clade in the expectation that some of them might be detected and others not in the new species.The former, therefore, should be interpreted as those possessed by the Most Recent Common Ancestor (MRCA) of the herein re-defi ned PhG (that is, including P. ballerioi sp.n.), while the latter likely evolved later, during the period corresponding to the internode between the basal-most dichotomy of PhG and before MRCA of the clade consisting of PhG minus P. ballerioi sp.n. (= clades A-E in Fig. 3).

Inadequate generic taxonomy of the Philharmostes group of genera
The analysis presented indicates that the existing taxonomic arrangement of PhG in seven nominal genera is non-cladistic and must be modifi ed.The topology depicted in Fig. 3 shows that the nominal genus Philharmostes is a waste-basket taxon used to accommodate distantly related and morphologically uniform members of PhG, while its easy-to-diagnose smaller lineages are artifi cially elevated in at least four nominal genera (and likely six, since the genera Callophilharmostes and Chaetophilharmostes were not included in this analysis).This situation is commonly encountered in various branches of the Tree of Life, when a plethora of historical names are scrutinized phylogenetically.In such situations three nomenclatorial revisions are possible.Firstly, one may apply the oldest genus-group name (Philharmostes in this case) to the entire PhG and, consequently, synonymize all six smaller nominal genera with the latter.Secondly, one may split the nominal Philharmostes into at least three smaller genera corresponding to clades C, E and F (Fig. 3).The third solution is to apply a tree-based biological nomenclature approach of the Phylo-Code, when rank-free taxon names are linked directly to clades (de Queiroz & Gauthier 1990, 1992;Cantino & de Queiroz, 2014), as recently implemented among anole lizards (Dactyloidae, Poe et al., 2017).With pill scarabs, however, none of these solutions seems timely, since the clade remains in a state of phylogenetic obscurity, with hardly any among its 38 nominal genera phylogenetically sound.Considering that the very fi rst herein implemented DNA-based analysis revealed a non-phylogenetic taxonomy, similar results might be eventually found among these beetles.In such a situation the conservative solution is to adhere to the current interim generic taxonomy (Ballerio & Grebennikov, 2016) and continue testing the taxonomically implied groups until a well-resolved and densely sampled tree of pill scarabs becomes available for a taxonomic rearrangement.

Philharmostes ballerioi sp. n., sister to the rest of the Philharmostes group of genera
Clades such as that formed by P. ballerioi sp.n., i.e., much smaller in number of species compared to their sister clades (and often in their geographical area), are often erroneously called "basal" (but see Krell & Cranston, 2004), "primitive" or "ancestral" (but see Omland et al., 2008).Philharmostes ballerioi sp.n. is yet another such example, even if at a shallower level, being the only known representative of the lineage sister to the rest of the entire PhG embracing 41 extant species.Discovery and/or phylogenetic interpretation of such deeply-nested organisms are of particular evolutionary signifi cance, since following the principle of parsimony, their properties indicate ancestral states for binary characters having both states in the diversifi ed sister clade.At least two characters of P. ballerioi sp.n. were likely acquired convergently with those of some PhG members, namely (1.) antennae consisting of nine antennomeres (since their number in the sister clade varies between seven and ten, and the latter state is widespread in other pill scarabs and likely has been present in MRCA of PhG) and (2.) complete lack of hind wings (other PhG have hind wings varying between entirely absent and fully functional, and the latter state is widespread in other pill scarabs and likely has been present in MRCA of PhG).These hypotheses will be tested and their list likely expanded once still inadequately known adult morphological characters of PhG (particularly understudied genitalia of both sexes contributing only seven among 107 characters in Ballerio & Grebennikov, 2016) become optimized on an inclusive and well-supported PhG phylogenetic tree.

Philharmostes ballerioi sp. n.: phylogeographical aspects
The discovery of the phylogenetically signifi cant new species sister to a much larger clade was made in the Nguru and Kaguru mountains (Fig. 6), two adjacent localities in the Eastern Arc Mountains (= EAM) in Tanzania (Fig. 6) renowned for their exceptionally high biodiversity (Lovett & Wasser, 1993).The orographic effect of these highlands results in a reliable source of rainfall from the nearby Indian Ocean, and the moisture is suffi cient to support sky-islands of rainforest on the slopes of the EAM, which sharply contrast with the surrounding hot and dry lowlands.Such conditions enabled the moisture-dependent biota on EAM to survive throughout the dramatic climatic fl uctuations in the Pliocene-Pleistocene, when Afrotropical rainforest repeatedly shrank to about 10% of its present size (Fig. 4 in Hamilton & Taylor, 1991).The recently radiated neoendemics and species-poor clades of relics form two main groups of EAM endemics (Fjeldså & Lovett, 1997), and the new species likely belong to the latter.These considerations indicate that the separation between P. ballerioi sp.n. and the rest of PhG might predate the Miocene uplift of the central African plateau.The latter event caused aridifi cation of the East African climate and fragmentation of the rainforest belt which extended throughout the entire continent since at least the Middle Cretaceous (Fig. 9.24 in Kirk-Spriggs & Muller, 2017).
Lack of the hind wings in adults of the new Philharmostes indicates reduced dispersal capacity.Flightlessness is thought to be positively selected for in the island-type EAM forest, since it prevents the aridity-intolerant organisms from being swept by wind into nearby dry inhospitable areas (Grebennikov, 2008).The presence of the new species in at least two EAM localities is, therefore, likely a result of vicariance of their once widespread ancestor, rather than a relatively recent founder dispersal (Heads, 2014).This hypothesis is consistent with the reciprocal monophyly of both populations (Fig. 6) and the time of their separation estimated at about 2.21 Ma, when both currently isolated Nguru and Kaguru forests might have been connected during at least one of many recurring wet periods concurrent with the Pliocene-Pleistocene interglacial cycles.

Fig. 6 .
Fig. 6.Ultrametric time tree dating the separation of the Nguru and Kaguru populations of Philharmostes ballerioi sp.n. obtained using BEAST and 0.018 subs/s/Myr/l (Analysis 2).Numbers on the time scales are million years before present.Node bars represent 95% confi dence interval of age estimate.Terminal numbers are specimen number (fi rst four digits) followed by litter sifting sample code and GenBank accession number.The six sequenced Nguru specimens are the types, specimen 3749 is the holotype.Illustrated paratype 3758 has right elytron removed to reveal complete lack of hind wings.Map illustrates both known localities of P. ballerioi sp.n.

Table 1 .
DNA fragments used in phylogenetic analyses (total number of sequenced terminals, followed by minimal, maximal and aligned length of each fragment, and the fi rst and the last position of each aligned fragment in the concatenated matrix).

Table 2 .
DNA fragments and their GenBank accession numbers of the 46 terminals used in the phylogenetic analyses.All included Ceratocanthinae belong to the nominotypical tribe Ceratocanthini.