Diversification of myrmecophilous Clavigeritae beetles (Coleoptera: Staphylinidae: Pselaphinae) and their radiation in New Caledonia

Clavigeritae (Coleoptera: Staphylinidae: Pselaphinae) comprise a supertribe of specialized myrmecophile rove beetles that display numerous behavioural, chemical and anatomical modifications associated with a socially parasitic lifestyle. Due to the extreme morphological diversity of clavigerites, their systematic relationships, patterns of character evolution and host ant use have been challenging to infer. Here, we resolve deep divergences within Clavigeritae by assembling a molecular dataset encompassing the breadth of tribal diversity. We classify Clavigeritae into six tribes: extinct Protoclavigerini, and recent Tiracerini, Mastigerini sensu nov., Clavigerini sensu nov., Lunillini sensu nov. and Disarthricerini stat. nov. The previously recognized subtribes Clavigerodina, Apoderigerina, Dimerometopina, Hoplitoxenina, Miroclavigerina, Theocerina and Thysdarina are demonstrated as polyphyletic or highly derived internal clades of Clavigerini and are hereby synonymized with the latter. Colilodion – a highly enigmatic taxon proposed to be the earliest‐branching lineage of recent Clavigeritae – is revealed to be unrelated to Clavigeritae and is transferred to Pselaphitae. We provide a systematic treatment of newly‐discovered endemic genera from New Caledonia and infer their phylogenetic affinity to the Australian tribe Tiracerini. The zoogeographic distribution of early‐branching Clavigeritae lineages in India and the Australian region indicates a possible Gondwanan origin of the supertribe. Extant clades diversified from the Eocene onwards, correlated with the rise of modern ants. We present evidence for island radiations of Clavigeritae in both Madagascar and New Caledonia during the Oligocene‐Miocene. Using a newly‐created morphological dataset, we analyse patterns of character evolution and demonstrate widespread convergence in morphology that extends to virtually all traits. Counterintuitively, however, characters postulated to be involved in beetle‐ant communication are amongst the most invariant, and least convergent. The host ant spectrum of many Clavigeritae genera and species is broad, contradicting co‐cladogenesis with hosts and instead implying widespread host switching. We speculate whether morphological variability in Clavigeritae is truly adaptive, as opposed to the product of ‘morphological drift’ of body parts under weak selection inside ant colonies, in species with potentially very small effective population sizes.


Introduction
Pselaphine rove beetles (Coleoptera: Staphylinidae) comprise a clade of ∼10 000 species of minute (1-5 mm), predatory, litter dwelling beetles (Park, 1942;Chandler, 2001;Parker, 2016aParker, , 2016b. Within this speciose subfamily, a pervasive evolutionary trend exists in which numerous lineages have evolved into 'myrmecophiles' -species that engage in symbiotic relationships with ants (Parker, 2016a). Among these symbiont clades, the supertribe Clavigeritae represents the largest radiation, with ca. 370 described species, all of which are believed to be obligately dependent on ant colonies (Kistner, 1982;Parker & Grimaldi, 2014;Parker, 2016a). Clavigerites are tightly integrated into host ant societies where they live as social parasites. The beetles are accepted by their hosts seemingly as nestmates and show numerous behavioural and chemical adaptations for this way of life. Clavigerites interact intimately with worker ants, including receiving food directly via oral trophallaxis, and being licked, groomed and carried around the nest (Kistner, 1982;Parker & Grimaldi, 2014;Parker, 2016a). Most species possess trichomes -clumps of thick setae associated with large glands (the so-called 'Wasmann' glands). These structures are positioned near the base of the abdomen and their secretions are highly attractive to host ants, likely promoting the beetle's acceptance inside the nest (Park, 1932(Park, , 1942Akre & Hill, 1973;Cammaerts, 1992;Parker & Grimaldi, 2014). Anatomically, clavigerite beetles are amongst the most specialized myrmecophiles known: the mouthparts are strongly reduced and internalized to facilitate trophallaxis (Akre & Hill, 1973;Jałoszynski et al., 2020), and the body as a whole is strengthened via extensive fusions of segments. The tergites are combined into a single 'tergal plate', and the antennae likewise show a reduction of segments, with all modern species possessing between three and six flagellomeres. In addition to the trichomes, a diversity of smaller glandular structures decorate the integument and appendages (Cammaerts, 1974;Hill et al., 1976). Some of the most anatomically extreme genera are additionally eyeless and wingless (Akre & Hill, 1973;Jałoszynski et al., 2020).
Despite their fascinating biology, the evolutionary history and phylogenetic relationships of Clavigeritae remain poorly resolved. Key insights into the origin of this clade of myrmecophiles came with the recent discovery of a stem-group clavigerite, Protoclaviger trichodens Parker & Grimaldi, in Indian Cambay amber (ca. 52 million years old) (Parker & Grimaldi, 2014). Protoclaviger shares many synapomorphies with recent Clavigeritae, including trichomes and modified mouthparts that demonstrate unambiguously that it was a myrmecophile. However, Protoclaviger is notable in its possession of only a subset of the myrmecophilous adaptations seen in modern species: its abdominal tergites remain unfused and its antennal segments are only partially fused, bearing eight antennomeres. Its mouthparts also protrude more prominently beyond the oral cavity. Protoclaviger reveals that the Clavigeritae-ant symbiosis dates to at least the Early Eocene, and fossil-dated molecular trees produced in the same study indicate an even older emergence of the beetle-ant association, in the Late Cretaceous (Parker & Grimaldi, 2014). Such a scenario implies that the radiation of crown-group clavigerites correlates with -and was potentially catalysed by -the Cenozoic diversification of modern ants and their rise to ecological dominance (Barden, 2017).
Modern Clavigeritae are distributed in all zoogeographical regions except Antarctica (Newton Jr & Chandler, 1989), and their cladogenesis has been coupled to dramatic morphological divergence. In addition to being heavily modified relative to ancestral free-living pselaphines, the body plan of clavigerites is also extremely variable between species. Virtually every aspect of external anatomy is prone to radical variation, with exaggerated modifications of head, antennal, leg and dorsal abdominal morphology being common (Parker & Grimaldi, 2014). As a result, Clavigeritae are amongst the most morphologically diverse clades within Pselaphinae, and indeed Staphylinidae as a whole. Historically, this variability has obscured the systematics of these beetles and resulted in a highly fragmented tribal and generic classification: 114 genera have been described, of which 68 are monospecific (Hlaváč, personal database). These genera are classified into four tribes and 11 subtribes, but the monophyly of most of these higher taxa has never been rigorously assessed (Parker & Grimaldi, 2014). Many of these groups were created based on a limited number of characters, including features that are systematically indefensible such as the number of antennomeres -a character prone to extensive variation and convergence across the supertribe.
The monophyly and phylogenetic placement of the Clavigeritae were evaluated by Parker & Grimaldi (2014) and Parker (2016b), as well as by a reanalysis of these same data by Lü et al. (2020). All three studies recovered a monophyletic Clavigeritae that was most closely related to Arhytodini and Pselaphini -two pselaphine tribes currently placed in the supertribe Pselaphitae. However, outgroup sampling was limited in these analyses, precluding resolution of Clavigeritae's definitive sister group. Through morphological cladistic analysis, Parker & Grimaldi (2014) further resolved Protoclaviger as sister group to all extant Clavigeritae including the tribe Colilodionini. This latter tribe was erected for the single genus Colilodion Besuchet (Besuchet, 1991) -a taxon that appears to combine certain characters of both Pselaphitae and Clavigeritae. Prior to the discovery of Protoclaviger, Colilodion had been proposed as the sister group of the remaining Clavigeritae (Besuchet, 1991). Parker & Grimaldi (2014) questioned this proposal, suggesting a position either outside of Clavigeritae, or as a highly modified lineage within this supertribe that bears some secondary character state reversals to a more pselaphite-like morphology. The authors emphasized the need for a molecular resolution to the true phylogenetic affinity of Colilodion, pending collection of new specimens of this exceptionally rare genus.
Due to limited taxon sampling by Parker & Grimaldi (2014), the internal topology of Clavigeritae was not fully resolved, but certain findings nevertheless emerged. These included strong evidence for a monophyletic radiation of clavigerites in Madagascar -an island containing an extremely high diversity of members of this supertribe, with 29 endemic genera (Hlaváč, personal database). Second, the genus Tiracerus Besuchet -at that time the only genus within the tribe Tiracerini, was found to be sister to all other Recent Clavigeritae. Notably, Besuchet (1991) alluded to the existence of potentially multiple genera from New Caledonia that share characters with the Australia-endemic Tiracerus, implying an unchartered component to the diversity of Clavigeritae that is relevant for resolving the supertribe's early cladogenesis.
In the present contribution, we build on the findings of Parker & Grimaldi (2014) to address some key outstanding challenges in the systematics of Clavigeritae. Focusing primarily on early-branching relationships within the supertribe, we evaluate the phylogenetic position of the enigmatic Colilodion, and have accumulated extensive New Caledonian material, including multiple new genera that show affinities with the early-diverging tribe Tiracerini. We describe these critical taxa and place them molecularly in the Clavigeritae tree. In addition to providing molecular resolution of early Clavigeritae relationships, we present a newly-created morphological dataset for the Clavigeritae as a whole. These morphological data enable us to demonstrate empirically the challenges of using morphological characters for exploring systematic relationships in this phenotypically extreme group of insects. The new dataset has also permitted us to investigate patterns of character evolution and convergence across the clade with respect to different parts of the body. Finally, we also summarize the known data about host ant specificity of Clavigeritae, leading us to hypothesize how clavigerites may have host-switched so frequently during their evolutionary history.

Historical classification and zoogeography of Clavigeritae
Due to the diverse morphology of Clavigeritae, the higher classification has often changed rapidly upon discovery of new taxa (see File S1, Supporting information). The first representative of the group was found by J. D. Preyssler in an ant colony under a stone in Prague (Czech Republic) in 1788 -the first documented discovery of a myrmecophilous beetle. Two years later, Preyssler (1790) described it as Claviger testaceus. Leach (1815) first introduced Claviger in the system of Coleoptera: he placed it as the tribe 'Clavigerides' in the family 'Phalacrurida', together with the tribe 'Pselaphides'. Dalman (1826) described the second member of the group, Articerus armatus from Quaternary copal from Tanzania; both Claviger and Articerus were treated as 'Clavigerines' in Pselaphinae by Aubé (1833). Hope (1836) was the first to give Clavigeridae family status, which was followed by Saulcy (1874), Reitter (1882Reitter ( , 1885 and other authors until the end of the 19th century. Schaufuss (1872Schaufuss ( , 1890 attempted the first internal classification of clavigerines, establishing five tribes: Adranini, Articerini, Clavigerini, Clavigerodini and Clavigeropsini. His classification was ignored by Raffray (1904aRaffray ( , 1904bRaffray ( , 1905aRaffray ( , 1905b in his Catalogue et Genera des Psélaphides in which he recognized 41 genera of clavigerines and included them in Pselaphidae. Jeannel (1949), also ignoring Schaufuss's classification, established a new tribal system with four tribes: Clavigerini, Disarthricerini, Fustigerini and Miroclavigerini. Later, based on his study of clavigerines from Madagascar, Jeannel (1954) found this subdivision unsatisfactory and proposed a new tribal system with 12 tribes. Subsequently, he subdivided the largest tribe, Fustigerini, into five groups (Jeannel, 1959). Célis (1969Célis ( , 1970 used Jeannel's system for his studies of material from Africa and Madagascar, based on which he added three additional tribes (Lunillini, Hoplitoxenini and Neoceratopsini) and established two subtribes of Neocerini (Neocerina and Theocerina). Besuchet (1986) created a new tribe, Tiracerini, for the Australian genus Tiracerus which included all Australian species previously classified in Articerus.
The discovery of the morphologically aberrant genus Colilodion caused a sudden revolution in the classification of the group and led Besuchet (1991) to reclassify the whole clade into three tribes: the speciose Clavigerini, monogeneric Australian Tiracerini and the monogeneric Colilodionini. All remaining tribes were downgraded to subtribes of Clavigerini, some of which were later synonymized (Besuchet, 2008;Hlaváč, 2011;Hlaváč et al., 2013;Hlaváč & Nakládal, 2016). Newton Jr. & Thayer (1995 demonstrated that pselaphines are a clade within the rove beetles (Staphylinidae) and reduced them from family to subfamily rank. Correspondingly, the rank of the clavigerines changed to the supertribe Clavigeritae, in order not to alter the internal classification proposed by Besuchet (1991) and subsequent authors. Parker & Grimaldi (2014) added the extinct tribe Protoclavigerini, established for the Early Eocene fossil genus Protoclaviger, considered to be the sister taxon to all extant Clavigeritae. Hence, four tribes are currently recognized, and Clavigerini is divided into 11 subtribes, the validity of some of which is questionable (Parker & Grimaldi, 2014). One subtribe in particular, Clavigerodina, is highly problematic, being a dumping ground for taxa that resist easy assignment to other subtribes.
Zoogeographically, Clavigeritae occur globally except for the polar areas but are absent from all Atlantic and many Indian and Pacific islands, including New Zealand (see File S15 for details). They are present in La Réunion (Célis, 1974), Seychelles  and Fiji and Vanuatu (Besuchet & Hlaváč, 2011). The highest species diversity is known from sub-Saharan Africa (39 genera, 87 species and subspecies) and Madagascar (37 genera, 74 species). The Palaearctic region hosts nine genera and 61 species, but the real diversity may be lower, because many species of Claviger and Diartiger Sharp require revision. The Australian fauna consists of 12 genera and 67 species but also requires a thorough revision. The Oriental region hosts 19 genera and 43 species. Central and Southern America hosts only two genera and 32 species, while North America is the most species-poor region with two genera and eight species. The reported local diversity strongly corresponds to the sampling effort and taxonomic activity in each region, and the supertribe's true diversity is likely much higher, especially in tropical areas. Additionally, a large number of undescribed species have amassed in museum collections, in particular from the Neotropical, Oriental and Australian regions.

Material and methods
Data used for the analyses Molecular data. We adopted the data from Parker & Grimaldi (2014), which includes 34 representatives of the Clavigeritae, and added five newly sequenced taxa: the genus Colilodion as a contentious early-branching member of the Clavigeritae (Besuchet, 1991), and four of five new genera that are endemic to New Caledonia; these are formally described in the Systematic Part below. The molecular dataset includes representatives of all currently recognized extant tribes and subtribes of Clavigeritae except Disarthricerina and Lunillina. Laboratory protocols for these new sequences follow those of Parker & Grimaldi (2014). Of the outgroup taxa, we kept the members of Pselaphitae used by Parker & Grimaldi (2014) (i.e. Caccoplectus Sharp and Rhytus Westwood for Arhytodini, Curculionellus Westwood and Pselaphogenius Reitter for Pselaphini), but reduced the number of the nonmacrosceline groups as their sequences were nontrivial to align with the Clavigeritae data. Ultimately, we kept only Lasinus Sharp, Bryaxis Kugelann and Faronus Aubé as more distant outrgroups from the original dataset and added Dasycerus Brongniart (Staphylinidae: Dasycerinae) as the taxon rooting the tree. Data were edited and aligned in Geneious (Biomatters Limited, New Zealand) using a combination of MUSCLE (Edgar, 2004) and Geneious algorithms and translational alignment in the case of protein-coding genes. In most cases, Clavigeritae + Arhytodini + Pselaphini were aligned first, and the more distant outgroups were added in a subsequent step. In all ribosomal genes, many hard-to-align and hypervariable regions were present; these were removed from the final dataset and only the fragments with reliable alignments were kept for the analysis. The final dataset for 47 taxa consists of 3719 bp of two mitochondrial (cytochrome oxidase I, cox1, 827 bp; 16S rDNA, 485 bp) and three nuclear fragments (18S rDNA, 1349 bp; 28S rDNA, 536 bp; wingless, wg, 522 bp). The list of GenBank accession numbers for all sequences used is listed in File S2. The final DNA dataset in FASTA format is available in File S3.
Morphological data. We prepared a dataset of 82 morphological characters coded for 40 taxa of Clavigeritae including Colilodion and the Eocene fossil Protoclaviger trichodens, as well as 14 non-clavigerine taxa; Dasycerus was included to root the tree. The dataset was newly built to examine morphological diversity within Clavigeritae, and thus does not overlap with that of Parker & Grimaldi (2014) which was designed solely for evaluating the placement of Protoclaviger. Taxon selection aimed to include representatives of all clavigerinae tribes and subtribes, and to be compatible with the molecular dataset that was acquired independently. We coded 23 characters from the head + antennae, 30 thoracic characters, 20 abdominal characters including male genitalia, and 9 leg characters. In some taxa, certain characters could not be examined due to absence of males or the inaccessibility of characters in the fossil; all these cases are coded as '?'. The final morphology matrix in TNT format is available in File S4. Morphological terminology follows Chandler (2001), except in the use of 'ventrite' instead of 'sternite' when discussing thoracic structures. Specimens were examined with a Leica S8APO stereomicroscope with diffuse lighting at magnifications up to 128×. Habitus images were taken with a Canon EOS 5D in combination with a Canon MP-E65 1-5× macro lens. Resulting images are focus stacks, aligned and stacked with Zerene Stacker and then postprocessed in Adobe Photoshop. Electron micrographs were taken from gold-coated samples using a JEOL 6380 LV scanning electron microscope. Specimens examined for the morphology matrix are from the collection of the first and second authors.
Combined datasets. Combined DNA + morphology datasets were used for analyses combining both data sources, those testing the position of the taxa without DNA data, and for measuring the fit of morphological data on the molecular tree. In some genera, the species included in the molecular and morphological datasets were not identical but were combined since the characters used are assumed to be invariable within the genus. See File S5 for information regarding which species were used for the combined dataset.

Phylogenetic analyses
Maximum parsimony analyses and character mapping. All unconstrained analyses were performed with TNT (Goloboff et al., 2008) using Traditional Search with 1000 replications and a maximum of 100 trees saved per replicate. All characters were unordered and of equal weight. Standard bootstrap was calculated using 100 replicates. See below for constrained analyses to test the position of selected taxa. Morphological characters were mapped onto trees in Winclada (Nixon, 2002).
Maximum likelihood analysis. Maximum likelihood (ML) analysis was performed in IQ-Tree 1.6.10 (Nguyen et al., 2015) under default settings and with 1000 ultrafast bootstrap replicates, using the best-fitting substitution models selected by Par-titionFinder2 (Lanfear et al., 2017). The dataset used and nexus file defining the partitions and their substitution models are available in File S6.
Bayesian analyses. All Bayesian analyses were performed in MrBayes 3.2.6 (Ronquist et al., 2012). Molecular data were subdivided by gene fragments and by codon position (for protein-coding genes); we used a mixed substitution model (Ronquist et al., 2012), and search consisted of two independent runs each with 4 chains and 25 million generations, sampled every 1000 generations. Convergence and sufficient effective sample size of all parameters were checked in Tracer 1.7 (Rambaut et al., 2018); the first 25% of trees were removed as burn-in. For the mixed DNA + morphology analysis, morphology was defined as a single partition and analyses under the Mk model for morphology (Lewis, 2001) with the correction for only variable characters scored; we assumed equal state frequencies and gamma-distributed rate variation across characters. Nexus files used for DNA and DNA + morphology analyses are available in Files S7 and S8.
Evaluating position of non-DNA taxa. To test the phylogenetic position of selected taxa for which no DNA data were available, including a reanalysis of the position of Protoclaviger trichodens, we employed a topology-constrained analyses of morphology-only data using maximum parsimony and Bayesian inference, as introduced by Fikáček et al. (2020). Topology constraints followed the results of the ML analysis of molecular data. The maximum parsimony analysis was run in TNT using exhaustive search after imposing the topology constraints. Bayesian analyses were performed in MrBayes, constraining the topology by a set of partial constraints prepared in R using createMrBayesConstraints command in the paleotree library (Bapst, 2012); the analysis was run for 1 000 000 generations, and the first 25% of trees were removed as burn-in. The position of the analysed taxon in the final majority rule tree was inspected, along with posterior probabilities of the alternative placements visualized using the R script by Fikáček et al. (2020). Nexus files with MrBayes commands and constraints for the Protoclaviger analysis are available in File S9.

Molecular dating.
To understand the timing of the diversification of the Clavigeritae in New Caledonia, we dated the tree in MrBayes, using three fossil calibrations adopted from Parker & Grimaldi (2014) and implemented as offset-lognormal priors (data in brackets = minimum, mean and standard deviation of the prior in nonlogarithmic scale): (i) Protoclaviger trichodens from Cambay amber (52 Ma [prior = 20, 40.8, 2]. In addition, we constrained the root age (i.e. the divergence of Pselaphinae and Dasycerinae) by uniform prior with minimum of 135 Ma and maximum of 190 Ma, based on the stem age of Pselaphinae and Staphylinidae revealed by Lü et al. (2020). Each analysis consisted of two independent runs each with 4 chains and 25 million generations, sampled every 1000 generations; the first 25% of trees were removed as burn-in. Two alternative positions of Protoclaviger were revealed in the analyses: (i) as sister to all modern Clavigeritae, and hence constraining their stem age or (ii) as member of the crown Clavigeritae and hence constraining their crown age. Time tree analyses reflecting both alternative placements of Protoclaviger were run and the results compared. Detailed parameter settings for the analysis are available in File S10.

Morphological evolution in Clavigeritae
The extreme morphological diversity of Clavigeritae encompasses a variety of traits that have evolved convergently in two or more lineages. To evaluate the amount of convergence, we first mapped the morphological characters onto the molecular tree and inspected the reconstructed evolution of each character and its retention index (RI) using Mesquite (Maddison & Maddison, 2019). To test whether the evolution of morphological characters follows phylogeny, we performed permutation tests using the dataset containing Clavigeritae taxa only (i.e. without outgroups and Colilodion). We generated 1000 random trees using the birth-death model in Mesquite, mapped morphological characters onto each of these trees, and plotted their lengths (i.e. the number of character state changes required) using the core functions of R (R Core Team, 2020). The length distribution of random trees simulates phylogeny-independent evolution. This distribution was compared to the length produced by mapping character state changes of onto the morphology-based and molecular trees (the ML molecular topology was, in all cases, taken as a proxy of the real phylogeny of the group). If these actual tree lengths fall within the distribution of lengths produced by mapping character states to random trees, character evolution cannot be distinguished from a random (i.e. phylogeny-independent) pattern. Conversely, if the length of the molecular tree is shorter than the 95% left-tail percentile of the random tree distribution, then character evolution is nonrandom, and follows the phylogeny (at least partially). Various character sets were used for the tests: (i) all characters; (ii) all characters without the 'best-performing' ones (these were selected by their RI values, RI > 0.3 was arbitrarily selected as a threshold; 19 characters were excluded from this dataset); (iii) characters corresponding to a particular body part (head, thorax, abdomen without genitalia, male genitalia, appendages) and (iv) characters of superficial trichomes and setation which are potentially at the interface of physical interactions between the beetle and ants. The list of the characters in each of these groups is provided in File S11.

Systematics
The aedeagus was studied using a Zeiss transmitted-light microscope at magnifications of up to 500×. The aedeagus was dissected and preserved in Euparal on a plastic card pinned together with the specimen. All drawings were made using a drawing tube. Body dimensions were measured as follows: length of head capsule = distance from the occipital constriction to the anterior margin of the frontal rostrum (an anterior part of the head extending beyond the antennal insertions and raised above the plane of the clypeus and labrum); elytral length = length of elytron along the suture; body length = combined length of the head, pronotum, elytra and abdomen. Width always refers to the maximum width of a given structure. Label data are cited verbatim. Red holotype or paratype labels were attached to the examined type specimens. The holotypes are deposited in Muséum National d'Histoire Naturelle, Paris, France (MNHN). For a list of depositories of paratypes, see File S17.

Online data deposition and registration required by International Committee on Zoological Nomenclature (ICZN)
All data used for the analyses and complete results are available in Files S1-S17. The same files were also uploaded to the Zenodo research depository (https://zenodo.org/) under http://doi.org/10.5281/zenodo.4422103. Newly generated DNA sequences were submitted to GenBank under accession numbers MW417247-MW417248, MW423303-MW423313 and MW440696-MW440698. To make the new taxa available from the date of the online publication of this paper, the paper was registered in ZooBank under the following ID: http:// zoobank.org/urn:lsid:zoobank.org:pub:0EAF81B6-D9EA-48E0-BE4C-366FEDFA0110.

Results
Trees resulting from all analyses performed are available in File S12. Below, we summarize our principal new findings.

Monophyly of Clavigeritae
Both ML and Bayesian analyses based on DNA data revealed that all modern Clavigeritae except Colilodion form a strongly supported monophylum (posterior probability pp = 1.0, bootstrap b = 1) (Fig. 1). Colilodion is in all analyses revealed as the moderately supported sister group of arhytodine taxa (Caccoplectus + Rhytus; pp = 0.86, b = 78). The included arhytodine taxa plus Colilodion are revealed as sister to Clavigeritae (pp = 0.69, b = 73) and together form a strongly supported monophylum with Pselaphini (pp = 1.0, b = 100). As we argue in the Discussion section, this finding, as well as the discovery of Protoclaviger (Parker & Grimaldi, 2014), lead us to remove Colilodion from the Clavigeritae and refute the scenario proposed by Besuchet (1991) that this genus bears 'transitional' morphology between Clavigeritae and outgroup Pselaphinae.
Maximum parsimony analysis based on morphology data revealed Colilodion as a member of the Clavigeritae (see File S12). Bayesian analyses combining DNA and morphology data revealed Colilodion as sister to Clavigeritae (see File S12). Both topologies agree with the original morphology-based ideas by Besuchet (1991); however, we believe these outcomes stem from the challenge of employing morphology to resolve relationships within Clavigeritae. We refute these outcomes on the basis of the molecularly-inferred placement of Colilodion, as well as on the grounds of morphological evolution of certain key characters (see Discussion).

Internal topology of Clavigeritae
Three principal clades are revealed in the Clavigeritae by both ML and Bayesian DNA analyses (Fig. 1), all established as The Asian genera Diartiger and Triartiger Kubota form a strongly supported monophylum (pp = 1, b = 100) that is weakly supported as a sister group to the North American Adranes LeConte in the ML analysis (b = 64). The topology of the dated Bayesian analyses (Fig. 3) differ slightly from the nondated one, most importantly in placing Tiraspirus gen.n. as the earliest-divergent New Caledonian genus (in agreement with the ML analyses) and in grouping the North American Adranes with Palaearctic Claviger.
The morphology-based maximum parsimony analysis produced a largely unresolved topology that was wholly incongruent with the molecular trees ( Fig. 2A, B). Constrained and combined analyses of modern taxa for which no DNA data are available revealed a reliable position only for Dzumaca gen.n. (forming a strongly supported monophylum with Ziweia gen.n. and Tiracaleda gen.n., pp = 0.95), Novoclaviger Wasmann (sister to Articerodes Raffray, pp = 0.95) and Pararticerus Jeannel (sister to Articerodes, pp = 0.96). The position of all other taxa differed between Bayesian and maximum parsimony analyses, and multiple alternative placements, each with a low probability was revealed in Bayesian topology-constrained analyses.

Position of Protoclaviger and dating analysis
Protoclaviger was revealed as a sister taxon to all modern Clavigeritae except Colilodion in the maximum parsimony analysis (b = 27) (Fig. 2C). In contrast, Bayesian analyses combining DNA + morphology and the topology-constrained ones all revealed Protoclaviger as a member of crown-group Clavigeritae, forming a monophylum with non-tiracerine taxa (pp = 0.8) in the combined analysis (Fig. 2D) or with Clavigerini (pp = 0.58) in the topology-constrained analysis. Mapping of alternative placements revealed a moderately high support for a placement of Protoclaviger as sister to Mastigerini+Clavigerini (pp = 0.68; see File S12). We consider any crown-group position of Protoclaviger as highly unlikely based on its many plesiomorphic character states (Parker & Grimaldi, 2014). These include antennae (eight antennomeres compared to six or fewer in all modern Clavigeritae), abdominal tergites (tergites 4-6 are unfused in Protoclaviger but fused in all modern Clavigeritae), maxillary palpi (protruding from oral cavity in Protoclaviger but barely so in modern Clavigeritae), abdominal intersegmental membranes (distinct between ventrites in Protoclaviger, reduced to absent in modern Clavigeritae) and antennal scape (large and dorsally visible in Protoclaviger, reduced and dorsally obscured by shelves of frontal rostrum in modern Clavigeritae). None of these character states is known from any modern member of Clavigeritae, indicating that the position of Protoclaviger as a crown clavigerite with numerous reversals is highly unlikely.
We used Protoclaviger to calibrate our time tree analyses, using it either as the calibration of the stem age or crown age of the Clavigeritae. The results are summarized in Table 1. The time tree in which Protoclaviger was placed as the sister to modern Clavigeritae (i.e. the position we consider most likely), and hence calibrating modern Clavigeritae's stem age, is illustrated in Fig. 3, while both trees are shown together in File S12. As expected, both time trees differ especially in the estimates of the stem and crown ages of the Clavigeritae: 59.1 and 40.5 Ma when Protoclaviger is considered as sister to modern taxa, compared to 69.6 and 59.3 Ma when Protoclaviger is considered as a member of the crown-group. The differences between analyses are less apparent in divergence times of internal Clavigeritae clades, which are estimated to be 5-10 Ma younger in the analysis where Protoclaviger is sister to modern taxa than when Protoclaviger is placed in the crown-group. Both analyses indicate that the New Caledonian clade of Tiracerini diverged from the Australian relatives in the Late Eocene or Early Oligocene (39.1/28.1 Ma) and started to diversify in New Caledonia in the Early Miocene (21.5/16.6 Ma). The Madagascan clade diverged from the continental relatives during the Oligocene (29.5/22.3 Ma) and started to diversify shortly after (26.7/20.1 Ma).

Evolution of morphological characters
When mapped onto the ML molecular tree, most characters show low RI values, indicating a high amount of convergent evolution and more character changes than expected under a maximum parsimony criterion. The same trend is seen when characters are grouped by body parts. Nineteen characters show RI > 30, indicating a partial correlation with the ML topology (File S13); these characters are related to head subdivision (char. 2), the form of the buccal cavity (char. 12), the position of tentorial pits (char. 15), the presence/absence of antebasal and lateral pronotal foveae (chars. 25, 26), the position of mesocoxae (char. 34), setation of the metaventrite and elytra (chars. 37, 39, 63), presence/absence of sutural stria, lateral ridges and humeral callosity on elytra (chars. 42, 44, 48), male secondary characters of legs (chars. 72, 73) and male genital characters (chars. 76, 79, 80, 81). When all morphological characters are mapped on the ML tree (Fig. 4A), the tree is significantly shorter than the length distribution of characters mapped onto random trees. However, this result is clearly driven by the 19 characters listed above: the ML tree length falls within the distribution of random trees when these characters are not considered (Fig. 4B, File S14). These findings indicate that the evolution of most morphological characters is not correlated with phylogeny and is highly convergent and hence likely driven by other factors. When body parts are compared, characters of the head, abdomen and appendages produce ML tree lengths indistinguishable from the random tree distribution (Fig. 4C, E, F). In contrast, characters of the thorax, male genitalia and likely also surface trichomes involved in beetle-ant interactions produce a shorter ML tree length than the random tree distribution (Fig. 4D, G, H), indicating a correlation with the phylogeny, although many such characters still exhibit convergent evolution in two or more lineages (compare Fig. 4I, J).

Colilodion: Morphological convergence in a non-clavigerine myrmecophile
Colilodion is an Oriental genus with nine known species (Fig. 1M). Although it has never been directly collected in association with ants, it exhibits most morphological adaptations for obligate myrmecophily found in the Clavigeritae: (i) antennae with three to four antennomeres, including a hugely enlarged apical segment; (ii) strongly reduced mouthparts; (iii) fused abdominal tergites and (iv) presence of trichomes on the abdomen (as well as other parts of the body). However, Besuchet (1991) and Löbl (1994) noticed that it also bears characters typical of Pselaphitae that are absent from Clavigeritae: (i) tarsomere I short and tarsomere II elongate (both tarsomeres short in all Clavigeritae); (ii) abdominal ventrite VII visible externally (internalized in all Clavigeritae) and (iii) aedeagus with distinct parameres (parameres are completely absent in all Clavigeritae). Based on Colilodion's mixture of character states, Besuchet (1991) proposed that this genus represents a 'missing link' between Clavigeritae and Pselaphitae and is hence the likely sister group of the remainder of the supertribe. However, doubt was cast over this hypothesis with the discovery of Protoclaviger by Parker & Grimaldi (2014). This fossil taxon possesses some derived character states typical of clavigerites that are missing in Colilodion, including short tarsomeres I and II, and a truncate antennal apex with setose cavity. However, Protoclaviger otherwise exhibits a higher number of plesiomorphic states than does Colilodion. This led Parker & Grimaldi (2014) to consider whether Colilodion may be a highly modified clavigerite with some features that are convergent with Pselaphitae.
Our present molecular analysis (Fig. 1) likewise refutes Besuchet's (1991) hypothesis, but instead reveals that Colilodion is not a clavigerite at all: its similarities with Clavigeritae are instead due to convergence. Our molecular analysis reveals that Colilodion is more closely related to the genera Caccoplectus and Rhytus of the tribe Arhytodini (supertribe Pselaphitae). Colilodion is hence a prime example of morphological convergence in response to a similar lifestyle (assumedly myrmecophily). We therefore remove Colilodionini from Clavigeritae. However, given our limited outgroup sampling, we refrain from synonymizing Colilodionini with Arhytodini until a more comprehensive assessment of Colilodion's sister taxon can be carried out. We therefore leave Colilodionini insertae sedis within Pselaphitae.

Systematics of Clavigeritae
The traditional classification of Clavigeritae established by Jeannel (1954), Célis (1969Célis ( , 1970, and Besuchet (1986Besuchet ( , 1991 is based on the external morphology, in particular the number of antennomeres, the presence of a constriction or carina on the head, the form of the pronotum and the elytra, the structure of the composite tergite and the position of trichomes. Our analyses reveal that most of these characters are highly convergent and have evolved multiple times independently within Clavigeritae. Consequently, the subtribes represented in our molecular analyses are either polyphyletic (Mastigerina, Clavigerodina, Clavigerina, Apoderigerina) or highly morphologically derived internal branches rather than deeply divergent phylogenetically isolated clades (Miroclavigerina, Thysdarina, Theocerina, Hoplitoxenina). These latter four subtribes are moreover part of the Madagascar adaptive radiation of Clavigeritae characterized by an explosion of morphological diversity (see below). On the basis of these findings (Fig. 1), we hereby update the internal classification of the Clavigeritae as follows (see also  Tiracerini is a strongly supported clade containing the Australian genus Tiracerus and all New Caledonian genera, as previously supposed by Besuchet (1991). In contrast to remaining clavigerite clades, Tiracerini is characterized by a diverse morphology of the male genitalia (Figs 5B-D, 7), including a strong asymmetry in the early-branching Tiracerus, Tiraspirus and Tiramieua (Fig. 7K-L, N-O). The newly described New Caledonian genera comprise a putative monophyletic group, although greater sampling of both the New Caledonian and Australian fauna will be needed to reliably resolve internal relationships in this tribe.
Mastigerini Clavigerini is the largest clade of Clavigeritae, distributed worldwide and defined here as containing genera previously classified in Clavigerina and Clavigerodina plus all Madagascar genera irrespective of their previous subtribal placement. Presently, it contains 97 genera and 312 species and subspecies.
Lunillini includes three monotypic genera, Eurycheiles Jeannel, Ischyroceros Reichensperger and Lunilla Jeannel, distributed in the Afrotropical region including South Africa. They are all diagnosed by transverse robust antennomeres I and II, and antennomere III long, composed of three to four fused but still slightly discernable segments.
Disarthricerini includes two oriental genera, Disarthricerus Raffray with three species and the monotypic Kurbatoviella Hlaváč. The assignment of Kurbatoviella to Disarthricerini is tentative (Hlaváč, 2010) and needs confirmation by molecular data. The topology-constrained morphological analyses indicated Kurbatoviella as close to the Tiracerini, which can be confirmed or rejected only by the DNA analysis.
The proposed changes in classification (Table 2) are surely not the final word; we consider this a starting point for future, more detailed analyses and studies. Our taxon sampling is limited, especially for the Clavigerini, for which DNA data are only available for 20% of all genera. The inclusion of additional genera and species, as well as other DNA markers, will hopefully result in a more robust phylogeny allowing for a clearer understanding of intratribal phylogenetic relationships. This is a prerequisite for establishing a phylogenetically accurate internal classification, especially of the largest tribe, Clavigerini.

Zoogeography and island radiations of Clavigeritae
Our dating estimates for both stem-and crown-group ages implementing Protoclaviger as sister to modern Clavigeritae are considerably younger than those previously proposed (Parker & Grimaldi, 2014), presumably due to fewer total calibrations points on account of the reduced number of nonclavigerite taxa included in our analysis. The analysis is nevertheless consistent with a radiation of clavigerites that began during the Eocene, correlated with the rise of modern ants (Grimaldi & Agosti, 2000;LaPolla et al., 2013;Barden & Grimaldi, 2016;Barden, 2017). Our molecular trees revealed three principal clades of Clavigeritae that originated during the Eocene (52-31 Ma). The earliest-diverging lineage, Tiracerini is a clade confined to the Australian region, with a single Australian genus (Tiracerus) and an additional radiation in New Caledonia, described below. The phylogenetic position of Tiracerini indicates a Gondwanan origin of the Clavigeritae, which is further corroborated by the presence of the closest-known sister-group of all extant Clavigeritae (Protoclaviger) in Early Eocene Cambay amber from Gujarat, India. The Indian subcontinent started to separate from East Gondwana ca. 120 Ma and in the Early Eocene it was approaching contact with the Asian continent (Royden et al., 2008;Verma et al., 2016;Krause et al., 2019). Following the basal divergence of Australian Tiracerini, the remaining crown-group taxa underwent a further split into the two clades: the moderately speciose Oriental and Australian Mastigerini (five genera and ten species) and the vast Clavigerini with a worldwide distribution. Fully understanding the complex biogeography of Clavigerini will require more extensive taxon sampling as well as a more stable internal topology of the tribe.
Notably, our analyses reveal two island radiations of Clavigeritae (Fig. 3): one in Madagascar (34 genera and 70 species) and another in New Caledonia (five genera and five species known to date, but potentially larger). The Clavigeritae fauna of Madagascar represents nearly 20% of all known species of Clavigeritae and is also diverse morphologically. Due to the considerable morphological differences, the endemic Madagascar genera were previously not considered closely related and were classified in many, often endemic, subtribes. Multiple independent colonizations of Madagascar were therefore expected. Our analyses, in agreement with those of Parker & Grimaldi (2014), revealed that the Madagascan fauna forms a monophyletic group with a most recent common ancestor that colonized Madagascar at the Oligocene-Miocene boundary. The clade subsequently diversified from the Early Miocene (ca. 20 Ma) onwards. Clavigeritae hence parallels the radiation of the obligately myrmecophilous paussine beetles (Carabidae) on Madagascar. The paussine radiation is however of much younger origin, dating back only 2.6 Ma (Moore & Robertson, 2014).
The radiation of the Clavigeritae in New Caledonia is a novel discovery of this study. The presence of three undescribed genera of the tribe Tiracerini was reported for New Caledonia by Besuchet (1991), but these taxa were never described. We now report the presence of at least five monotypic genera, all forming a monophyletic group (Fig. 2). The ancestor of this group colonized New Caledonia between ca. 28.1-16.6 My ago, that is relatively long after the re-emergence of the island 37 My ago (Grandcolas et al., 2008;Nattier et al., 2017;Giribet & Baker, 2019). The species diversity resulting from this radiation is seemingly much lower than in Madagascar, perhaps the result of the less diverse ant fauna in New Caledonia or the smaller area of the island (more so than to the age of the radiation itself; Table 3, Fig. 3A-C). However, the real species diversity of the New Caledonian clade may be higher than reported here, and discoveries of additional species are expected with greater sampling effort. The five New Caledonian genera and species are treated in detail in the Systematic Part below.
There may be smaller island radiations of Clavigeritae which are not visible in our data due to the absence of DNA-grade material. For example, Fiji hosts 6 genera and 11 species of Clavigeritae. As in New Caledonia and Madagascar, the Fijian endemic genera are morphologically very diverse (Besuchet & Hlaváč, 2011). The monophyly of the Fijian fauna and the age of this possible radiation need to be tested. The oldest parts of the Fiji Archipelago emerged ca. 28 Ma (Beatty et al., 2017), but many ant groups only radiated in Fiji during the Miocene, ca. 14-9 Ma (Economo et al., 2015;Matos-Máravi et al., 2018;Liu et al., 2020). The possible radiation of Fijian Clavigeritae may hence be relatively young. It remains to be explained why islands, when colonized by the myrmecophilous species, tend to host radiations resulting in a boom of morphological diversity. New Caledonia was colonized by Clavigeritae ca. 10-20 My after its emergence when the local fauna was well-established and saturated already, which in theory decreases the probability of successful colonization by a new-comer. However, a well-established fauna of host ants may be a precondition for successful colonization by inquiline clavigerines, causing delayed colonization. This also corresponds well to the relatively late colonization of Madagascar by the Clavigeritae and Paussinae beetles, only possible after ants colonized Madagascar and the ant fauna diversified.
Monophyly and endemicity of island faunas may arise from the low dispersal potential of these symbiotic insects. Clavigerines are believed to be completely dependent on host ants, and unable to pursue a free-living existence outside of host colonies except for a short migration period. Adults of many species are winged and evidently must disperse to some degree, but successful long-distance dispersal to islands -where the beetles must encounter and adapt to novel ant hosts for long term viability -must surely be a relatively rare occurrence. This is reflected in the absence of clavigerites on the majority of small islands as well as some ant-impoverished large islands, such as New Zealand. It also follows that successful dispersal of clavigerites from islands on which they do become established is, in turn, highly unlikely. This may explain why subsequent cladogenesis of clavigerites is strongly constrained within the boundaries of a given island, leading to endemic radiations. This explanation may also apply to other obligate myrmecophiles that have radiated on islands, such as paussine carabids (Moore & Robertson, 2014). Table 2. An updated classification of the Clavigeritae. Genera in bold were available for the molecular analyses performed in this study. Genera marked by an asterisk are endemic to Madagascar and are (likely) the part of the Madagascan radiation of Clavigerini. For a detailed list of genera with their distribution and number of species, see File S15.

Host ant use in Clavigeritae
Although all Clavigeritae species are believed to be obligate myrmecophiles, interactions with ants have only been studied for a handful of taxa (Donisthorpe, 1927;Park, 1932Park, , 1942Akre & Hill, 1973;Cammaerts, 1996Cammaerts, , 1999Akino, 2002). For the majority of species, adaptive myrmecophilous traits such as trichomes, fusions of segments and modified mouthparts provide indirect evidence of obligate myrmecophily for these species (Parker, 2016a). Additionally, many species have been directly collected from colonies of their putative host ants: ant associations are currently known for 55 genera (i.e. half of known genera) and 197 species and subspecies (i.e. 52.5% of known species/subspecies). In total, 27 extant ant genera and 86 ant species have been recorded as hosts for Clavigeritae (Table 4, File S16), and these hosts span five formicid subfamilies: Dolichoderinae, Ectatomminae, Formicinae, Myrmicinae and Ponerinae. Representatives of Tiracerini have been recorded from all five of these subfamilies, Clavigerini from Dolichoderinae, Formicinae, Myrmicinae and Ponerinae; Mastigerini with Dolichoderinae and Myrmicinae; Disarthricerini only from Formicinae and Lunillini from Myrmicinae. Myrmicinae and Formicinae genera are the most frequent hosts, with the myrmicine genus Crematogaster Lund and formicine Lasius Fabricius hosting the largest array of Clavigeritae genera and species.
The rather loose relationship of many genera and species of Clavigeritae with their host ants indicates that host switches are Table 3. The number of species and genera of Clavigeritae and ants in New Caledonia, Madagascar, and Fiji, that is, islands in which adaptive radiation of the Clavigeritae is proven or expected. The information about the ant diversity adopted from Fisher & Peeters (2019) for Madagascar, Ward & Wetterer (2006) for Fiji and Jennings et al. (2013)  rather frequent, even occurring intraspecifically. Which host ant a given beetle utilizes may depend more on the ant's biology than on the ant's phylogenetic position. Host range may also be influenced by the distribution range of available hosts in a particular locality, with Clavigeritae in temperate regions (e.g. Claviger or Adranes) potentially more limited in the choice of hosts than beetles from tropical and subtropical areas (e.g. Fustiger and Tiracerus). It is also unclear whether all ants recorded for particular Clavigeritae species refer to 'true' host ants, in which the beetles survive and complete their life cycle, and to what extent these host records are incidental. What is presumably the case, however, is that the species richness of recent Clavigeritae -one of the most speciose clades of myrmecophiles -has likely been contingent on their capacity to utilize diverse ants as hosts. Host switching opens up potential niche space and provides opportunities for speciation; it can also help to circumvent coextinction with hosts (Zhou et al., 2019). Host switching appears to be characteristic of the most species-rich groups of obligate myrmecophiles, including Paussinae (Moore & Robertson, 2014) and Hetaeriinae (Histeridae) (Zhou et al., 2019). In contrast, highly-host specific clades that show negligible host switching, such as the many convergent lineages of Aleocharinae (Staphylinidae) that utilize specific army ant species as hosts (Maruyama & Parker, 2017), are closely tied to the evolutionary fate of a single host. Their own diversification is strongly contingent on their host's cladogenesis, and consequently, such groups can be very species poor.

Morphological evolution of Clavigeritae
Our analyses reveal a strong conflict between molecular and morphological characters. Although our molecular analyses are affected by limited taxon and gene sampling, several of the recovered clades are well-supported and correspond to zoogeographic distribution. Hence, they seem likely to reflect the true phylogeny of the group. The conflict seems therefore to arise from highly convergent evolution of many morphological characters. This conclusion is corroborated by the maximum parsimony analyses of morphological characters, which resulted in more than 1500 equally parsimonious trees and a largely unresolved consensus tree ( Fig. 2A, B). Our randomization tests (Fig. 4A-H) demonstrated clearly that the molecular topology does not, on average, score better than a randomly generated tree when morphological characters are mapped onto it. We found that when different body parts are compared, a somewhat correlated relationship with the molecular phylogeny is revealed for some characters of the head and thorax, those of the male genitalia and likely also trichomes on the body surface. Male genitalia are likely under sexual selection more than under selective pressures from the environment or host ants. They are likewise relatively refractory to these selective pressures, being internalized and hence physically isolated from the environment, and not involved in ant interactions. Moreover, genitalia are diverse in the Tiracerini only; in the remaining tribes (i.e. the majority of Clavigeritae) they are very uniform (Fig. 5). For these reasons, limited amounts of homoplasy in these structures is not surprising. On the other hand, the relatively low amount of homoplasy is surprising for trichomes on the dorsal body surface -structures at the interface of beetle-ant communication and physical interaction, and hence conceivably under strong selective pressure. The tight association and coevolution with ants may be the main driver of morphological evolution in Clavigeritae. The broad host ant spectrum of many genera and species of Clavigeritae, in some cases spanning four different ant subfamilies, contradicts this hypothesis: host switches do not necessarily induce morphological changes as expected under a parallel diversification scenario (Fig. 6A). Rather, rampant host switching across the supertribe implies that clavigerites may have evolved a principal solution to achieving social integration that is potentially effective across a broad spectrum of ant species (Fig. 6B). Consistent with this hypothesis, Donisthorpe (1927) noted that specimens of Claviger survived well when experimentally introduced to nests of novel hosts. Such a versatile mechanism of host exploitation may consequently be reflected in the relative morphological stasis of specific body structures that mediate beetle-ant behavioural interactions.
Why some other parts of the body are then seemingly free to evolve so dramatically is unclear; perhaps this marked variation between species is unrelated to myrmecophily but adaptive for other aspects of these beetles' life histories. Alternatively, it might not be the result of adaptation at all, but instead stem from nonadaptive 'morphological drift' of structures in environments where there is only a weak stabilizing selective force to maintain the mean trait value (Fig. 5B). For example, convergent variation in the extent of fusion within the antennal flagella, from one to four flagellomeres, may arise from an absence of strong selection to maintain a precise number of segments. Such a phenomenon may be predicted to occur inside ant colonies, which, despite their many novel ecological pressures, may nevertheless be relatively free of many of the selective constraints on trait performance that are experienced by free-living species in the external environment. Weak stabilizing selection is also expected in Fig. 6. Alternative models of ant-beetle coevolution. A, parallel diversification + host switching model predicts that each beetle species interacts with one or few closely related host ants; beetle cladogenesis partially mirrors that of the ants, and morphological diversity is generated by rare host switches to distantly related ant lineages (see red and blue beetles). B, morphological drift + promiscuous exploration model, in which a large amount of morphological diversity is generated neutrally under relaxed selection, and the resulting morphotypes are filtered for those broadly compatible with a wide spectrum of unrelated host ants, leading to rampant host switching. In this scenario, drift may also generate morphotypes of low fitness or that exhibit incompatibility with available hosts (depicted as extinct lineages). [Colour figure can be viewed at wileyonlinelibrary.com].
populations of rare species with tiny effective population sizes and hence high genetic drift (Lynch, 2007). Clavigerites may be archetypal examples of such naturally rare species. Examining patterns of gene flow and dispersal efficacy of obligate myrmecophiles such as clavigeritates, and estimating their effective population sizes, will help illuminate how these insects have diversified in the face of the population-level challenges associated with a socially parasitic lifestyle.

Clavigeritae Leach, 1815
Diagnosis. Antennae with three to eight antennomeres (three to six antennomeres in extant taxa), terminal antennomere usually longest (shorter than penultimate antennomere only in Thysdariella Hlaváč), apex of terminal antennomere at apex truncate, covered with short, thick setae or simply rounded and setose in Disarthricerina. Mouthparts reduced, barely visible, buccal cavity transverse with or without lateral cavities for the accommodation of maxilla. Abdomen with visible tergites 1-3 (segments IV-VI) either free (Protoclaviger) or fused into composite tergite (tergal plate of Chandler, 2001) which usually bears various kind of basal impression (all extant species). Elytra and abdomen with tufts of specialized trichomes (most species), sometimes trichomes completely absent (Disarthricerini). Legs with elongate meso-and metatrochanters; third tarsomere longer than first and second tarsomere combined; tarsus with single tarsal claws. For the diagnosis of the supertribe see also Chandler (2001: 534) and Parker & Grimaldi (2014).
Composition. Six tribes, 3 subtribes, 114 genera with 372 species and subspecies. Many taxa are waiting for description.
Biology. All the members of Clavigeritae are obligate myrmecophiles.
Key to tribes and genus groups of Clavigeritae 1. Antennae with more than six antennomers, visible tergites 1-3 (segments IV-VI) free. . . . . . . . . †Protoclavigerini -Antennae apparently with two to six antennomeres (in reality always with three to six antennomeres, but the scapus is sometimes minuscule and difficult to see), visible tergites 1-3 (segments IV-VI) fused into composite tergite. . antenna with three to six antennomeres; aedeagus with median lobe bulbous, with well-developed basal capsule, with dorsal diaphragm opening circular or elliptical ( Fig.5E -H)

Tribe †Protoclavigerini
Diagnosis. Body length about 1.6 mm. Antennae with eight antennomeres, scape large and exposed, terminal antennomere about twice as large as VII, maxillary palpi emerging well outside buccal cavity, abdomen with unfused tergites IV-VI, paratergites IV and V with paired hook-like trichomes, paratergite VI with smaller trichomes (Parker & Grimaldi, 2014).

Tribe Tiracerini Besuchet, 1986
(Figs 5B-D, 7-10) Diagnosis. Body length 1.0-3.0 mm. Trichomes always only on abdomen. Antennae with three antennomeres, scape always very small, terminal antennomere long, clearly longer than head, apex of terminal antennomere truncate, covered with short, thick setae. Mouthparts reduced, barely visible, buccal cavity transverse with or without lateral cavities for the accommodation of maxilla. Composite tergite with basal impression of various shape and size. Aedeagus symmetrical or asymmetrical, with (Tiracerus) or without parameres (all other genera), dorsal diaphragm opening absent.
Composition. Six genera with 53 species including those described in the present paper are known. More new species will be described from New Caledonia later.
Composition. Monogeneric group, the only genus Tiracerus has 48 described species.
Distribution. Endemic to Australia.
Tiracaleda genus group (Figs 7A-I, 8-9) Diagnosis. Body length 1.25-2.0 mm. Buccal cavity with lateral cavities for the accommodation of maxilla. Pronotum lacking sulci and foveae. Composite tergite with basal impression, with one to two small trichomes on each side. Aedeagus elongate, symmetrical in dorsal view, subdivided in two parts, basal part much longer than apical part, apical part formed by bifurcate lobe; ventrolateral foramen absent, dorsal phallobase diaphragm absent.
Distribution. Endemic to New Caledonia.
Sexual dimorphism. Female unknown. Very likely females lack the bunch of long setae in the basal third of the protibia.
Etymology. Named after the type locality of its only species, Mt. Dzumac. Gender feminine.

Dzumaca monteithi
Diagnosis. As for the genus. Biology. Unknown, three specimens were collected by sifting leaf-litter in rainforest and one by Malaise trap.
Sexual dimorphism. Males differ from females by triangular, sharply projecting mentum, strongly elongate posterior metaventral process and a large baso-median spice on metatibiae.
Etymology. The name combines Tira-referring to Tiracerini and -caleda referring to New Caledonia. Gender feminine.
Sexual dimorphism. Legs of females simple, lacking preapical spur on mesotibiae and preapical spine on mesotrochanters.
Etymology. Named after Ziwei Yin, a specialist on Pselaphinae from the Shanghai Normal University, China. Gender feminine.
Description. Body (Fig. 7G) light yellowish-brown, head, pronotum and elytra with fine microsculpture, composite tergite smooth, shiny. Length 1.45-1.60 mm. Head widest across eyes at their posterior margin, convergent behind eyes to posterior margin of head capsule, head capsule short, separated from short, nude, shagreened neck region by well-defined occipital constriction, rostrum obtuse, straight, temples short, about as long as length of eyes, clearly shorter than rostrum, frontal fovea absent, vertexal foveae present. Clypeus well visible dorsally, subrectangular, well-separated from frons by large antennal cavities. Mouth part rudimentary but well-developed, completely hidden in elliptical buccal cavity. Venter of head with gular plate large, almost flat, with two largely separated tentorial pits. Antennae long, with three antennomeres, scape and pedicel small, subequal in length, scape barely visible dorsally, terminal antennomere very long, slender, evenly expanded to apex, at apical third slightly curved, about 12 times as long as pedicel, elliptical in cross section, obliquely truncate at apex with dense, long setae. Thorax. Pronotum clearly wider than long, basal margin with two lateral impressions and weakly defined median lobe, with well-defined antebasal sulcus, lacking median longitudinal sulcus and foveae. Prosternum about half as long as pronotum, procoxal cavities closely separated. Hypomera large, hypomeral carinae absent. Mesoventrite more than twice shorter than metaventrite, fused together, with two arch-like mesoventral carinae which divide mesoventrite in three parts, mesocoxae well-separated by short, wide isthmus, mesoventral and anterior metaventral process with bunch of setae. Metaventrite large, about 1.75 times as wide as long, metacoxae widely separated by wide, short, largely concave posterior metaventral process. Elytra wider than long and longer than pronotum, 1.20-1.40 times as wide as pronotum, lacking basal foveae, striae or carinae, humeri well-defined, posterior corners rounded, posterior margins of each elytron with wide, low median lobe, lacking trichomes. Abdomen elongate, much longer than elytra and longer than wide, composite tergite (IV-VI) smooth, with deep transverse depression which is expanded in middle to discal area, basal depression bearing two small trichomes, with four well-developed and long paratergites, first visible paratergite with trichome in basal part, posterior third of each side with four long and stout macrosetae. First visible sternite (III) short, covered by dense setae, in midline almost four times shorter than second (IV), second visible sternite about as long as remaining visible sternites, lacking median basal carina and with two well-defined, deep basolateral foveae. Aedeagus (Fig. 7H, I Composition. Two monospecific genera (Tiramieua, Tiraspirus), both described below.
Distribution. Endemic to New Caledonia.
Sexual dimorphism. Female unknown. Etymology. The name is a combination of Tir-referring to Tiracerini and -amieua referring to Col d'Amieu, the type locality of the only species. Gender feminine.
Description. Body (Fig. 7J) light yellowish-brown, head smooth, pronotum and elytra with fine microsculpture, composite tergite smooth, shiny. Length 1.55 mm. Head widest behind eyes, after strongly convergent to posterior margin of head capsule, head capsule short, separated from short, nude, shagreened neck region by well-defined occipital constriction, rostrum obtuse, straight, temples short, about as long as length of diameter of eyes, clearly shorter than rostrum, frontal fovea hy, hypomeron; md, mandible; mn, mentum; msvmc, mesoventral median carina; msvp, mesoventral process; occ, occipital constriction; pd, pedicel; pmtvp, posterior metaventral process; pstp, posterior prosternal process; sc, scape; sst, sutural stria; tp, tentorial pits; tVIII, abdominal tergite VIII; IV-VIII, abdominal sternites. absent, vertexal foveae present. Clypeus not visible dorsally, very short, not separated from frons. Mouth part rudimentary but well-developed, completely hidden in semicircular buccal cavity. Venter of head smooth, with large, almost flat gular plate, with two largely separated tentorial pits. Antennae long, with three antennomeres, scape minuscule, not visible dorsally, pedicel almost twice as long as scape, antennomere III long, slender, strongly clavate at apex, at apical third slightly curved, about six times as long as pedicel, semicircular in cross section, truncate at apex with dense, long setae. Thorax. Pronotum clearly wider than long, cordiform, with well-defined median antebasal fovea, lacking lateral foveae and sulci. Prosternum about two third of length of pronotum, hypomera large, median part of prosternum separated from hypomera by hypomeral carinae, median part rough, procoxal cavities closely separated. Mesoventrite about three times shorter than metaventrite, fused together, mesoventral carinae absent, mesocoxae well-separated by short, wide isthmus, mesoventral process small, narrow, anterior metaventral process clearly wider. Metaventrite large, about twice as wide as long, metacoxae widely separated by wide, short, straight posterior metaventral process. Elytra wider than long and longer than pronotum, about 1.50 times as wide as pronotum, lacking basal foveae, striae or carinae, humeri well-defined, posterior corners sharp, posterior margins largely rounded, lacking trichomes. Abdomen elongate, much longer than elytra and much longer than wide, composite tergite (IV-VI) smooth, with transverse basal depression and with deep basal rectangular basin in basal half of abdomen, bordered by lateral carinae, with two lateral tubercles, lacking trichomes, with four long paratergites, first visible paratergites with lateral projection and trichome in basal part, composite tergite with sparse, long macrosetae. First visible sternite (III) short, in midline about five times shorter than second (IV) and clearly shorter than third (V), second visible sternite more than twice as long as remaining visible sternites, lacking median basal carina and with two well-defined basolateral foveae. Aedeagus (Fig. 7K, L) elongate, asymmetric in dorsal view, basal part and apical lobe not separated, ventrolateral foramen absent, dorsal phallobase diaphragm absent. Legs long and slender, fore and hind legs simple, mesofemora with minuscule median tooth, mesotibiae with long, sharp prebasal spine, mesotrochanters with sharp, very long preapical spine and with bunch of long setae. slightly depressed, depression smooth, lacking hypomeral carinae, median part of prosternum with rough, mesh-like structure, procoxal cavities closely separated. Mesoventrite (Fig. 10G) about three times shorter than metaventrite, fused together, with two mesoventral median carinae (Fig. 10G: msvmc) which divide mesoventrite in six parts, mesocoxae well-separated by short, wide isthmus, mesoventral and anterior metaventral process ( Fig. 10G: msvp, amtvp) with bunch of setae. Metaventrite (Fig. 10G) large, about 1.60 times as wide as long, metacoxae widely separated by wide, short, largely concave posterior metaventral process (Fig. 10G: pmtvp). Elytra (Fig. 9H) clearly longer than wide and almost twice as long as pronotum and twice as wide as pronotum, lacking basal foveae, sutural striae entire (Fig. 10H: sst), discal striae absent, humeri weakly-defined, posterior corners rounded, posterior margins slightly concave in middle, lacking trichomes. Abdomen elongate, much longer than elytra and longer than wide, composite tergite (IV-VI) smooth, with transverse basal depression and with deep basal rectangular basin in basal half of abdomen, bordered by lateral carinae interrupted by small trichomes, with three well-developed and long paratergites, first visible paratergites with large trichome, posterior part of composite tergite with about eight to ten long and stout macrosetae. First visible sternite (III) short (Fig. 10I), covered by dense setae, in midline more than three times shorter than second (IV) and about as long as third (V), second visible sternite clearly shorter than remaining visible sternites, with short median basal carina and with four well-defined basolateral foveae ( Fig. 10I: mbc, blf ). Aedeagus (Fig. 7N, O) elongate, asymmetric in dorsal view, basal part and apical lobe not separated, ventrolateral foramen absent, dorsal phallobase diaphragm absent. Legs moderately long and slender, fore and hind legs simple in both sexes, mesotibiae in males with small median spine and large preapical spur, mesofemora with tooth behind middle, mesotrochanters with sharp, long preapical spine (Fig. 10E). Diagnosis. As for the genus. Biology. Unknown, all known specimens were collected in rainforests by flight intercept traps, pitfall traps, by sifting litter and by Pyrethrum fogging of trees and logs.

Tribe Clavigerini
Diagnosis. Small to large clavigerines, size 1.00-3.00 mm. Trichomes always on elytra and abdomen. Antennae with three to six antennomeres, scape always very small, terminal antennomere from short to very long, apex of terminal antennomere rounded or truncate covered with short, thick setae. Mouthparts reduced, barely visible, buccal cavity transverse with lateral cavities for the accommodation maxilla. Composite tergite with basal impression of various shape and size, aedeagus from symmetric to with slightly asymmetric apex of median lobe, divided in basal and apical part, portion of basal and apical parts varying, dorsal diaphragm opening present.
Composition. Ninety-seven genera (Table 2) with 308 species (i.e. 83% of all described species of Clavigeritrae). Many new genera and species are awaiting description.
Distribution. All zoogeographic regions. Biology. Highly specialized genera Claviger, Diartiger and Adranes can be almost exclusively collected in nests of host ants. Tropical Clavigerini are living with many different ants (Table 4, File S16), but many species are mainly collected by sifting of leaf-litter and by flight intercept traps in well-preserved primeval forests. Ultraviolet trap is also giving some results, but it is less effective.

Tribe Disarthricerini
Diagnosis. Body length 1.20-1.30 mm. Body lacking trichomes. Antennae with three antennomeres, terminal antennomere considerably longer than antennomere 1 and 2 combined, apex of terminal antennomere rounded. Mouthparts reduced, barely visible, buccal cavity transverse with lateral cavities for the accommodation maxilla. Composite tergite lacking basal transverse impression, with deep basolateral foveae. Aedeagus symmetric, divided in basal and apical part, basal part longer than apical part, dorsal diaphragm opening present.
Distribution. Malay Peninsula and Borneo. Biology. Disarthricerus moultoni Bryant was collected on the ground, under dead leaves in the nest of Paratrechina butteli (Forel) (Bryant, 1915). Other specimens were collected at light and in rotten wood or under stones with unknown ant .
Comments. No DNA-grade specimens were available for our analyses for the tribe.

Tribe Lunillini
Diagnosis. Body stout, length 1.50-1.80. Trichomes present on abdomen. Antennae with three antennomeres, scape and pedicel robust, clearly transverse, terminal antennomere about as long as head, considerably longer than I and II combined, apex of terminal antennomere truncate covered with short, thick setae. Mouthparts reduced, barely visible, buccal cavity transverse with lateral cavities for the accommodation maxilla. Composite tergite with basal transverse impression, with deep basolateral foveae. Aedeagus symmetrical (at most slightly bent, Fig. 6G), divided in basal and apical part, basal part as long as or longer than apical part, dorsal diaphragm opening present.
Composition. Three monospecific genera, Eurycheiles, Ischyroceros and Lunilla. No undescribed species are known to us.
Distribution. Sub-Saharan Africa. Biology. Most specimens were collected in The Democratic Republic of Congo by sifting of debris of wet leaf-litter and soil around nests of small, thin unidentified ants; ant colonies had about 200-300 ant specimens (Coulon, 1982). Lunilla was collected with Pheidole. Ischyroceros was sifted from the leaf-litter in the forest in Malawi (Hlaváč & Baňař lgt., 2018).
Comments. No DNA-grade specimens were available for our analyses for the tribe.

Tribe Mastigerini
Diagnosis. Body stout, length 1.2-1.7 mm. Trichomes present on abdomen. Antennae with three to six antennomeres, scape and pedicel minuscule, terminal antennomere short, shorter than length of head, apex of terminal antennomere rounded (in Longacerus, Pseudacerus) or truncate (in Mastiger, Tapas, Tasmiger) covered with short, thick setae. Mouthparts reduced, barely visible, buccal cavity transverse with lateral cavities for the accommodation maxilla. Composite tergite with basal transverse impression, with deep basolateral foveae. Aedeagus from symmetric to slightly asymmetric apex of median lobe, divided in basal and apical part, basal part longer than apical, dorsal diaphragm opening present.
Biology. The majority of known material was collected by sifting of leaf-litter or by flight intercept trap in primeval tropical forests. The host ants are known for one species of Tapas (Iridomyrmex gracilis (Lowne)) and one species of Tasmiger (Rhoptromyrmex wroughtonii Forel).

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Appendix S1: Historical changes of the higher classification of the Clavigeritae S2. List of GenBank accession numbers of all sequences used for DNA analyses.
S3. DNA alignment used for molecular analyses (in FASTA format).

S4.
Morphology matrix used for morphology-based and mixed analyses (in TNT format).

S5.
Description of the composition of the combined DNA + morphology dataset.

S6.
Dataset used for IQ-Tree analysis and the description of the partitions. S7. Nexus file used for the MrBayes analysis of the molecular data.
S8. Nexus files used for the MrBayes analysis of the mixed (DNA + morphology) data.

S9.
Nexus files used for the MrBayes topology-constrained analysis of the position of Protoclaviger. S10. Nexus file used for the MrBayes time-tree analysis, showing how calibration points were incorporated. S11. Groups of morphological characters used for the analyses of phylogenetic signal and their retention indexes.

S12.
Results of all phylogenetic analyses performed. S13. Lengths of 1000 random trees for each group of characters. S14. Results of the randomization tests of the phylogenetic signal in morphological characters. S15. List of genera of the Clavigeritae, number of their species and distribution. S16. Summary of known interactions of Clavigeritae with ants.

S17. Descriptions of new species of New Caledonian
Tiracerini.