Phylogeny of the Oniticellini and Onthophagini dung beetles (Scarabaeidae, Scarabaeinae) from morphological evidence

Abstract A phylogenetic study was conducted to hypothesize relationships of most of the genera of the Oniticellini and Onthophagini for the first time using morphological characters from a diverse array of external and internal sclerites. The monophyly and sister relationship of both tribes was found using Bayesian and parsimony analyses with heavily to moderately weighted data. An alternative hypothesis based on parsimony analyses of unweighted or slightly weighted data show a paraphyletic Oniticellini without the Onthophagini, although recognition of the subtribe Helictopleurina as a tribe would eliminate non-monophyly. Of the three Oniticellini subtribes, the Helictopleurina and Drepanocerina are monophyletic. There is no support for the monophyly of the Oniticellina or the Onthophagini subtribe Alloscelina, as currently defined. The genus Liatongus is paraphyletic, while strong support was found for monophyly of the Madagascan genus, Helictopleurus. The genus Onthophagus is never monophyletic in any analysis performed. Two new subtribes are also proposed: Liatongina subtr. n. including the genus Liatongus and Attavicina subtr. n. including the genera Attavicinus and Paroniticellus. Topological evidence shows that the ancestral oniticellines and onthophagines were all coprophagous with alternative food sources evolving relatively recently. Both myrmecophily and termitophily probably evolved only once in the onthophagines. The phylogenetic analysis supports an African origin for the two tribes, with a relatively early age for the split of the Madagascar helictopleurines from the remaining oniticellines via dispersal. Furthermore, the presence of the oniticellines in the New World is hypothesized to be due to two relatively old dispersal events via Beringia and two relatively recent trans-Atlantic invasions of the Caribbean.


Introduction
The Oniticellini and Onthophagini contain close to one half of the roughly 7,000 species of scarabaeine dung beetles known worldwide (Schoolmeesters et al. 2010). Undoubtedly more species remain undocumented, perhaps as many as 1,000 or more, and certainly a large proportion of those, when described, will be placed in either the speciose onthophagine genus Onthophagus or another closely related group. In one estimate, Onthophagus alone contains 1765 species (Hanski and Cambefort 1991) while a recent estimate put the total at ~2500 species (Tarasov and Solodovnikov 2011). With their high species diversity, these two tribes are among the most important in regards to understanding the evolution of the Scarabaeinae dung beetles.
Many Oniticellini and Onthophagini (generic classification in Table 1) are tunnelers where dung is buried at the ends of tunnels created beneath a dropping and used as adult and larval food. But within these tribes the behaviors and food sources are far more diverse. For example, the oniticellines Oniticellus Serville, Tragiscus Klug, and possibly Paroniticellus Balthasar have species with nesting behavior known as dwelling, where brood cavities are excavated within the dung or in the soil adjacent to and in contact with the dropping (Davis 1989, Philips andBell 2008). Some of the smallsized onthophagines are kleptocoprids that steal dung from larger tunneling or rolling dung beetles. Nearly all oniticellines specialize on dung, especially those from large bodied herbivores (Cambefort 1991). In contrast, many onthophagines have alternative food and nesting habits. This includes taxa that feed on carrion, millipedes, fruit, or mushrooms Cambefort 1991, Brühl andKrell 2003). Several genera are also highly specialized to live with ants or termites and likely feed on nest detritus (Branco 1995(Branco , 1996Krell and Philips 2010). The number of times these alternative food sources have evolved is unknown.
The Onthophagini contain approximately 35 genera, the number depending upon the chosen taxonomic classification (Cambefort 1991, Davis et al. 2002, Philips 2011. Many taxa previously considered subgenera are now recognized as genera, including Proagoderus Lansberge, Diastellopalpus Lansberge, Digitonthophagus Balthasar, Strandius Euonthophagus Balthasar (Halffter andMatthews 1966, Davis et al. 2008). Subgenera are still recognized in both Onthophagus and Caccobius, and species groups defined, for example, by Arrow (1931), Balthasar (1963), Boucomont (1923), Branco (1992a) andd'Orbigny (1913) are commonly used. Within the Oniticellini, species groups have been defined, for example, within Liatongus and Helictopleurus (see Montreuil 2005 for those in the latter). In the Onthophagini, subtribes have yet to be proposed with the exception of the Alloscelina, while three are recognized in the oniticellines. But typically few characters for any of these groups justify their recognition and none have yet been studied in a broader phylogenetic context. Table 1. Generic classification of the Oniticellini and Onthophagini with the number of species in each genus currently recognized in the far right column. Taxa in bold are included in this study.
Inconsistencies in the classification of these two tribes have been frequent. For example, nine of the 35 generic level onthophagine taxa were formerly placed in Dichotomiini (Janssens 1939;Balthasar 1941;Branco 1989Branco , 1990Branco , 1992aBranco , 1992b or the Oniticellini (Kolbe 1905). The Alloscelina were previously placed as a subtribe of the Scarabaeini (Halffter and Edmonds 1982) or were considered a tribe by Janssens (1949). In a taxonomic proposal for the Drepanocerina (Krikken 2009), the previous species groups of Janssens (1953) were elevated to genera. Most recently, this classification has been updated by Roggero et al. (2015) with the creation of another two genera. Lastly, the monotypic Scaptodera (S. rhadamistus) recognized by some workers, has been previously cited as Liatongus (Paraliatongus) Reitter, Oniticellus Serville, or even Pseudoniticellus Janssens (e.g., Hanski and Cambefort 1991).
Based on evidence of relationships within the Scarabaeinae from a wide range of morphological characters (Zunino 1983, Luzzatto 1994, Philips et al. 2004b, Tarasov and Génier 2015 and some molecular studies (Villalba 2002, Ocampo and Hawks 2006, Monaghan et al. 2007, there is strong support for a single common ancestor for the Oniticellini and Onthophagini. But relationships between and within these two tribes and subtribes are very poorly known. Therefore, this morphological study was undertaken to hypothesize the phylogeny of this group, including a test of monophyly of the entire clade, each tribe, and the four subtribes of the oniticellines and onthophagines currently recognized. Also, a preliminary test of generic monophyly of Helictopleurus d'Orbigny, Liatongus Reitter and Onthophagus Latreille is performed. As is possible with all phylogenies, the evolution of the biology and biogeography of this clade of dung beetles is also hypothesized.

Materials and methods
The 41 ingroup species, including many rare taxa, represent the morphological and behavioral diversity found within these two tribes reasonably well ( Table 2). Representatives of 12 of the 26 genera of Oniticellini were studied, including both Old and New World representatives of Liatongus Reitter. Excluded from the study were two monotypic and rarely collected genera Drepanoplatynus Kirby, and Heterosyphus Paulian, as well as Sinodrepanus Simonis. Also, excluded were seven of the recently proposed Drepanocerina genera of Krikken (2009) Barbero et al. (2009a, b), and Roggero et al. (2015) whose species were previously placed in Drepanocerus; all of these genera were created after the data collection for this study and their relationships, based on morphological data, have recently been well studied (Roggero et al. 2015). Twenty-four of the 36 genera of Onthophagini were also represented in the data set. Of the 12 excluded genera, all are uncommon in collections and 10 are either monotypic or contain only two species.
To test generic monophyly, the study included three species of the Madagascar endemic Helictopleurus (from three species groups), two species of Liatongus Reitter (an Old and New World representative), and three species of Onthophagus Latreille (one endemic each from North America, Africa, and Australia). Three outgroup taxa were used to polarize character evolution in the ingroup and included a representative from the Onitini, Eurysternini, and Sisyphini. Previous analyses of the entire subfamily using morphology (Philips et al. 2004b) and evidence from molecular studies (Villalba 2002, Ocampo and Hawks 2006, Monaghan et al. 2007), support the Onitini or a lineage of this tribe (if the Onitini is paraphyletic) as the probable sister clade of the Oniticellini + Onthophagini, and the Sisyphini may also be closely related as well.
Dried specimens were initially relaxed in hot water and then cleared using lactic acid. Individual sclerotized body parts were observed on slides in glycerine and some of the larger structures were studied dry. An attempt was made to discover as many characters as possible from the various sclerotized structures without any bias as to what might be phylogenetically informative. Characters were only excluded if discrete states could not be adequately defined or if they were autapomorphic within a binary character. Five genitalic characters were used and more could have been included (as some recent dung beetle studies have done (e.g., Roggero et al. 2015, Tarasov andSolodovnikov 2011). Regardless, a very broad range and relatively high number (134) of Sukelus jessopi (Branco) Kenya both external and internal morphological characters were discovered (including those from sclerites not used in the previous two studies cited) and used to hypothesize relationships among taxa (see below). Illustrations of nearly all of the character states are in Philips et al. (2002Philips et al. ( , 2004aPhilips et al. ( , 2004b and are not duplicated herein. Morphological nomenclature follows primarily Lawrence and Britton (1994), and additionally that of Nel and Scholtz (1990), Doyen (1966) and Dellacasa et al (2010). For the parsimony analysis, WinClada (Nixon 1999) was used to enter data (Table  3) in Dada, the trees were first produced with Nona (Goloboff 1993a), and the character state distributions were analyzed with Clados (Nixon 2002). Characters were analyzed unordered and initially unweighted. One thousand repetitions with random taxon entry order in Nona were used several times to make certain the shortest trees were discovered, although all of the most parsimonious trees were found within the first 100 replications. Use of the Rachet in TNT (Goloboff et al. 2008) did not find any additional topologies with further searching, using the default parameters as follows: the perturbation phase used up-weighting and down-weighting probabilities of 4 and was stopped when 20 substitutions were made or 99% swapping was completed; the total number of itera tions was set at = 10 and the auto-constrained iterations at = 0 with alternate equal weights on.
Piwe weighting (parsimony with implied weighting) in TNT (Goloboff 1997) was used in an attempt to reduce the number of equally parsimonious trees. K values were tested from 1 through 50 using the default parameters. This program estimates character weights (i.e., character reliability based on degree of homoplasy) during the tree search and is based on searching for trees which have maximum total fit. The algorithm considers the fit as a concave function of homoplasy and, when comparing topologies, differences in steps occurring in characters which show more homoplasy, are less influential. In other words, if two trees of equal length are being compared, the tree topology supported by the character with less homoplasy is preferred (see Goloboff 1993b for more details). Analyses with lower K (or CO) values have the strongest weighting against those characters with homoplasy.
Weighting in morphological analyses has been considered unscientific by some (e.g., Kluge 1997) but valid by many others (e.g., Griswold et al. 1999, Platnick 2000. It could be argued that giving all characters equal weights is a form of unbiased weighting. Certainly some characters are more informative than others in any analysis and differential weighting is considered a reasonable technique to use in the present study (see Goloboff et al. 2008 for further discussion and citations within).
The standard model for morphological data and the default set of priors were used in the Bayesian analysis (Mr. Bayes version 3.2.1 for Windows 32 bit) (Ronquist et al. 2012). The topology was developed using the MCMC command with two simultaneous searches. Three separate runs were performed and for each 1,000,000 generations were typically needed to get below the 0.01 level of the standard deviation of split frequencies. Default burn-in values used were the first 25% from the cold chain. Plots of the likelihoods of sampled trees were examined to determine when the MCMC chains had reached the stationary distribution. The majority rule consensus tree was obtained from the remaining trees.
Posterior probability node support is shown on the tree from the Bayesian analysis. Bootstrap values were calculated in TNT from 1000 replicates using sampling with replacement, the traditional tree search setting, and collapsing of groups below a value of 1. Values are displayed at all nodes where support is at or above 0.5 or 50%. Consistency and retention indices (CI and RI) derived from unweighted data  are listed after each character within the descriptions. CI and RI values for weighted data (not shown) would result in different values for many characters depending upon the K value used; additionally the lower the K value the more likely characters would receive values of zero.
Biogeographic analysis was done using a Dispersal-Vicariance Analysis (Ronquist 1997) in S-DIVA (Yu et al. 2010) to hypothesize ancestral state distributions with the maximum areas at each node = 4 to reduce erroneously inferred vicariance events. The included areas were Afrotropical, Palaearctic, Oriental, Australasian, Madagascar, Nearctic, Neotropical, and Caribbean following those recognised for dung beetles in Davis et al. (2002).
Brief discussions on the evolution of morphology, nesting behavior, food sources, and biogeography are based on the best supported hypotheses of evolution. Rather than inclusion in the discussion, Table 4 is used to summarize the specific character support based on unweighted and/or weighted K = 10 topologies for the tribes, subtribes, and relevant clades. One should note that weighted tree support is based on synapomorphies that can be considered more reliable in defining clades. Table 4. Character support for selected clades, unweighted or weighted (K value =10) data.

Results
The hypothesis that the Oniticellini and Onthophagini constitute a monophyletic group is strongly supported by this study (Figs 1 -5). No outgroup taxa were found within the ingroup in any of the analyses. Analysis of the unweighted data using parsimony produced a total of six trees (strict consensus with character support in Figs 4A, 5), each with a length of 825 steps, and overall CI and RI of 0.29 and 0.50, respectively. Alternative topologies in the step-like tree are due to two polytomies within the onthophagines, one located approximately in the middle of the topology involving alternate relationships of six taxa and the other located at the apex in four probable termite associated taxa. The Bayesian analysis shows unresolved clades near or at the base in both the Oniticellini and the Onthophagini (Fig. 2). The tribe Onthophagini is monophyletic in all analyses. For the Oniticellini two results were discovered. The tribe is monophyletic in the Bayesian tree (Fig. 2) and the Piwe weighted trees with K values from 1-10 (Figs 1A, 3, 4B). In contrast, the . This topology is considered to be the best supported in this study. Bootstrap values above 50% found for nodes indicated. A Oniticellini.
A A parsimony tree using unweighted characters and the Piwe weighted trees with relatively high K values from 11-50 show the Helictopleurina as sister to the remaining oniticellines + onthophagines (Fig. 4A). K values of 44 and higher resulted in one of the six unweighted trees found. This was an expected result since with higher K values the weighting becomes increasingly reduced and eventually becomes equivalent to un- weighted data. Topologies found using K values = 11-30, are identical and similar to the unweighted data topology except that the two polytomies seen in the latter are fully resolved (Fig. 5, two small clade inserts).
K values of 8-10 also resulted in identical topologies ( Fig. 1) while those of 4-7 resulted in minor rearrangements of taxa within the Onthophagini. Rearrangements occurred at the base or in more derived clades, via taxa shifting up or down one node, or by adjacent taxa becoming sisters. As expected, weighting generally increases tree length compared to unweighted data, except in cases of minimal weighting (i.e., high K values) that result in the same topology found using unweighted data. But moderate weighting of K values = 8-11 only increased the tree length by two steps (825 to 827). In contrast, K values of 1, 2, and 3 increased the tree length by 31-64 steps (lengths were 889, 874, and 856 steps respectively). Additionally, analyses with the very lowest K levels show trees the most altered (Fig. 3) compared to the unweighted analyses. Therefore the best hypothesis is considered to be the fully resolved parsimony topology using K values of 8-10 ( Fig. 1) due to only moderate data weighting, the increase in tree length by only two steps, and congruence with the Bayesian analysis (Fig. 2).

Clades within the Oniticellini
In all analyses, the Helictopleurina and Drepanocerina are monophyletic (supported by four and six controverted synapomorphies respectively) while the Oniticellina is paraphyletic. The sister relationships of Tiniocellus + Scaptocnemis, Tragiscus + Oniticellus, and Anoplodrepanus + Euoniticellus are also invariant in all analyses. Attavicinus monstrosus (Bates), formerly placed in Liatongus (see Philips and Bell 2008), does not share common ancestry with the other Liatongus taxa. The clade consisting of Attavicinus + Paroniticellus is maintained except in the most extreme weighting (K = 1 (and 2, not illustrated); Fig. 3). In those two topologies, Paroniticellus is shifted to a more derived position in the tree and Attavicinus is a basal lineage and sister to all remaining Oniticellina + Drepanocerina. Also, the clade composed of Liatongus militaris as sister to Tragiscus + Oniticellus is shifted to an apical origin. The remaining oniticellines show only minor differences in arrangement in the trees with K values < 6 compared to that seen in with K values of 8-10.

Clades within the Onthophagini
This tribe is a well-defined monophyletic group based on both Bayesian and parsimony analyses, and using either unweighted or weighted data. In the unweighted topology, five states (one uncontroverted) support monophyly. The relationships found with K values = 8-10 are identical (Fig. 1B) and differ from those found at higher K values of [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] in that the basally originating Phalops and Digitonthophagus are sister taxa in the former instead of successive branches in the latter. There are also minor differences of the position of some taxa in the middle of the tree (from Onthophagus hecate through to Milichus). Trees based on stronger weighting with K = 1-7 values (see Fig.  3) place Digitonthophagus as basal and, in trees from K = 2-7, Onthophagus capella + Diastellopalpus as the next most apical clade. In the K = 4-7 trees, a third basal clade consisting of Proagoderus + Phalops is seen. The trees found with the most extreme weighting (K values of 1-3) compared to all others are the most different (Fig. 3); the topology found with K = 2 (not illustrated) contains elements of both. Changes include the shifting to a more apical origin of both Proagoderus and Phalops.
The probable symphile taxa (from Amietina to the tree apex) with the most derived features have strong support for monophyly; all included nodes are supported by a minimum of one uncontroverted and two controverted states. These lineages are not dung feeders but necrophilous (Amietina), myrmecophilous (Haroldius, Megaponerophilus, Alloscelus, and likely Eusaproecius and Pseudosaproecius) or termitophilous (Heteroclitopus, Pinacotasus, Stiptopodius, and probably Sukelus). Inquiliny most parsimoniously evolved only once in the common ancestor of Haroldius through to Sukelus. Support for this clade is based on five character states, including one uncontroverted. The Alloscelina (Alloscelus, Haroldius, and Megaponerophilus) is found within this larger clade but there is no support for its monophyly in any of the tree searches performed.

Discussion
These results and the morphological studies of Philips et al. (2004b), Edmonds and Halffter (1978) and Zunino (1983) and the molecular studies of Villalba et al. (2002), Ocampo and Hawks (2006) and Mlambo et al. (2015) support monophyly (or at least a close relationship) of each of these two tribes. In contrast, the molecular study of Wirta et al. (2008) does not support either a monophyletic Oniticellini or Onthophagini. The base of their tree contains two large clades of onthophagines, and between those lineages and the apical helictopleurines are a mix of clades composed of both the oniticellines and onthophagines. In the molecular study of Monaghan et al. (2007), the oniticellines and onthophagines were always a single clade based on parsimony or Bayesian 9 partition models. But in other tested models (Bayesian 7 partition, maximum likelihood, parsimony length invariant, and POY models), this group was paraphyletic. Additionally, the tribe Oniticellini was monophyletic in the Bayesian 7 partition analysis (their Fig.  2), but paraphyletic in the nine-partition, maximum likelihood, and three parsimony analyses via the inclusion of either two or four species of onthophagines (one parsimony tree shown in their Fig. 3). In contrast to the results here, the genus Helictopleurus was in a relatively derived position and hence, there is no support for a monophyletic Oniticellini without the inclusion of the helictopleurines. Lastly, the onthophagines were always polyphyletic. Similar problems of non-monophyly of these tribes are seen in Mlambo et al. (2015), using molecular evidence from four genes in a study of African taxa, where the tribe Onthophagini is paraphyletic without the Oniticellini.
In contrast to all other studies, Bai et al. (2011) found the sisyphines positioned within the oniticellines. This study relied on hind wing and body morphological characters and the result is most likely due to evolutionary convergence. Additionally, Gé-nier (2009) moved the genus Eurysternus as a subtribe to the Oniticellini, but evidence herein does not support this placement.

The Oniticellini Clades
Traditionally the Oniticellini are divided into three subtribes, the Drepanocerina, Helictopleurina, and the Oniticellina (e.g., Halffter and Matthews 1966). In this study the Drepanocerina and Helictopleurina are monophyletic. Recognition of the Helictopleurina as a tribe would eliminate possible paraphyly in the Oniticellini. This was done for perhaps the first time in Orsini et al. (2007), although their study provides little evidence for doing so with a single outgroup and only Madagascar species represented.
The two representatives in this study, Cyptochirus and Afrodrepanus, were found as sister taxa. The Drepanocerina may not be monophyletic based on preliminary studies mentioned but not discussed in Barbero et al. (2009b), perhaps due to the relative heterogeneous morphologies of the genera. Krikken (2009) does give a morphological description of the group but later notes that support for their monophyly is negligible. Roggero et al. (2015) using Anoplodrepanus as a single outgroup gives a relatively weak test for monophyly. But based on the results in this study and the hypothesized close relationship of taxa included within the Drepanocerina, there is support for the monophyly of the clade. Notably, Scaptocnemis and Anoplodrepanus have been placed by some authors in the Drepanocerina previously, but this study supports their placement in the Oniticellina as suggested earlier by Branco (2010) and Krikken (2009). The Drepanocerina as defined herein is therefore restricted to the Afrotropical-Oriental regions.

Helictopleurina
The Madagascar endemic helictopleurines are composed of only two genera, the speciose Helictopleurus with ~80 species and the monotypic Heterosyphus. The included species of Helictopleurus representing three subgenera are strongly supported as monophyletic. The rare Heterosyphus was not available for study. When Paulian (1975) described this genus he placed it in the Canthonini and discussed its unusual morphology, surmising that it was not closely allied with this clade. But he also stated that this taxon might be related to Helictopleurus based on the form of the head and legs. Similarities were noted with Sisyphus in the form and sculpture of the prothorax and that the presence of a visible scutellum separates it from Onthophagus, although this last character is present to various extent within species of Helictopleurus as well. Halffter and Edmonds (1982) may have been the first to place Heterosyphus in the Helictopleurina. The fairly well supported molecular study of Wirta et al. (2008) shows this taxon to be a derived Helictopleurus. Therefore, Heterosyphus Paulian should be considered a junior synonym of Helictopleurus d'Orbigny so that the species becomes Helictopleurus sicardi (Paulian).
The position of the Helictopleurina in the moderately or heavily weighted topologies supports the recognition of the group as either a monophyletic subtribe or tribe. In contrast, the unweighted or slightly weighted topologies support the creation of a new tribe to avoid a paraphyletic Oniticellini. But molecular evidence from Monaghan et al. (2007) in some of their analyses suggests that the helictopleurines are derived from within the oniticellines. Therefore although the results found in this study support the helictopleurines as either a subtribe or tribe, current molecular evidence suggests that this clade should continue to be recognized as a subtribe.

Oniticellina
The subtribe Oniticellina is paraphyletic, although some generic sister relationships are invariant (see in results above). Euoniticellus and Paroniticellus have been classified as subgenera of Oniticellus in the past (Halffter and Matthews 1966) but neither appears phylogenetically close to Oniticellus and their generic status is supported. Interestingly, some evidence exists that the monotypic Paroniticellus and Oniticellus, (perhaps only the subgenus Oniticellus, sensu stricto) are closely related, as larvae of both are characterized by numerous setae on the head capsule (Lumaret 1979). But immatures will first need to be examined over a broader range of species to confirm this potential synapomorphy. The Jamaican endemic Anoplodrepanus is sister to Euoniticellus and also probably a valid genus.
While an African Euoniticellus was included in this study, the single New World species E. cubiensis Laporte from the West Indies (Cuba, Jamaica, Isla de la Juventud, and the Bahamas (Woodruff 1973, M.A. Ivie pers. comm.)) was not. Zunino (1982) states that this species (referred to as Oniticellus cubiensis) "belongs though in a rather isolated position, to the genus Euoniticellus." A more recent phylogeny of Euoniticellus by Cambefort (1996) solidly supports its placement in this genus.
Several genera not included also deserve brief comments. The monotypic genus Scaptodera was removed from generic synonymy by Paulian (1986), but might be a Liatongus based on its previous placement. Ixodina is often considered a synonym of Drepanocerus (e.g., Davis et al. 2008) but Krikken (2009) and Roggero et al. (2015) list this genus as valid and part of the Drepanocerina. The species now placed in the Southeast Asian Yvescambefortius (Ochi and Kon 1996) was originally placed in Oniticellus by Gillet (1926) and later in Tiniocellus by Janssens (1953). Ochi and Kon (1996) hypothesized a close relationship of their new genus with the east African Scaptocnemis, which in this phylogeny is sister to Tiniocellus. Hence, it may be part of the Scaptocnemis + Tiniocellus clade.
Based on this study, a solution for monophyly within the Oniticellini subtribes outside of the Helictopleurina would be to recognize all genera (including the Drepanocerina) in a single subtribe, or alternatively, the recognition of the Drepanocerina and two or more subtribes. From the Bayesian analysis and the majority of parsimony topologies found (see Figs 1A, 4B), another possibility is a subtribe including Paroniticellus and Attavincinus, while the other would include all remaining genera together with the Drepanocerina. One could also divide the Oniticellina into three or more subtribes while continuing to recognize the Drepanocerina. Based on current evidence of relationships, a possible subtribal classification recognizing hypothesized monophyletic groups is as follows: Drepanocerina Diagnosis: This subtribe can be characterized by the following two characters: The mesonotal prescutum anteriorly with the transverse plate approximately ventrally directed is narrowly rounded apically and the posterior median tormal process (epitorma) of the epipharynx is smoothly and broadly tapered throughout its length. Based on a weighted topology (K=3) the three following characters also support this clade: The prothoracic apodeme has an incomplete oblique suture/carina, one or more "extra" internal sclerites are associated with the galea, and the metanotum scutellum posteriorly has a slight margin or remnant edge.
Oniticellina Diagnosis: This subtribe can be characterized by the following combination of characters: Maxilla internal accessory sclerite is curved near the proximal apex, the protibia on the proximal side has a line of setae associated with a carina, males have a longitudinal ridge on the pronotum laterally, and the mesosternal scutum is distinctly transverse in shape.
Note that based on the most heavily weighted parsimony topologies (Piwe K values = 1 and 2), Paroniticellus should be placed within the Oniticellina. Based on the Bayesian topology, Paroniticellus may need to be placed in its own or possibly one of the other subtribes, as its relationship to the other subtribes (excluding the Helictopleurina) is unclear.

The Tribe Onthophagini
This tribe in every analysis was a well supported monophyletic group. Philips et al. (2004b) also support the monophyly of this tribe based on morphological evidence. In contrast, molecular data in Monaghan et al. (2007), Wirta et al. (2008) and Mlambo et al. (2015) generally support a poly-or paraphyletic Onthophagini and indicate that some currently recognized onthophagines may belong elsewhere.
The topologies found herein with extreme weighting K = 1-3 topologies are considered less trustworthy compared to those found with higher K values. This conclusion is also supported by the molecular studies of Monaghan et al. (2007) and Wirta et al. (2008) that do not support either a relatively derived position of Proagoderus and Phalops. Further, Tarasov and Solodovnikov (2011) also found support for a basal origin of Proagoderus and Diastellopalpus as seen in the less strongly weighted topologies. An internal classification of the onthophagines, if desired and based on the evidence herein, would likely necessitate the creation of numerous subtribes to maintain monophyly in classification. Branco (1989Branco ( , 1990Branco ( , 1992aBranco ( ,b, 1995 placed several of these most derived genera together in a new group based on shared morphological features, such as antennomere number. His Pinacotarsus group includes eight genera (taxa in bold included in this study): Dorbignyolus, Eusaproecius, Heteroclitopus, Krikkenius, Pinacopodius, Sukelus (as Falcidius), Stiptocnemis, and Pinacotarsus. The Stiptopodius group, first defined by Boucomont (1923) and containing some of the genera in the Pinacotarsus group, was redefined by Branco (1995) and includes the genera Neosaproecius, Pseudosaproecius, Stiptopodius, and Stiptotarsus. Of the six genera included in this study, all form a monophyletic clade at the apex of the onthophagines in both the parsimony and Bayesian analyses. Branco hypothesized the Stiptopodius group as less derived than the Pinacotarsus group and this is approximately seen in Figs 1 and 2. But neither group is monophyletic in any of the topologies discovered in this study.

The Genus Onthophagus
As a very preliminary test of monophyly for Onthophagus, three species were studied, including one Nearctic, an Australian, and one Ethiopian. Also included were four taxa that are usually considered subgenera of Onthophagus (i.e., Digitonthophagus, Diastellopalpus, Proagoderus, and Strandius), as well as Euonthophagus, Hyalonthophagus, and Mimonthophagus. All will be considered Onthophagus, sensu lato in the discussion below.
In this study, Onthophagus, in either the strict or broad sense, is not monophyletic. Onthophagus, sensu stricto, typically appears in three clades. Onthophagus, sensu lato, appear in various topologies in no fewer than four to as many as nine separate lineages. Strandius often appears as part of a clade of Onthophagus while Euonthophagus, Hyalonthophagus and Mimonthophagus are justified as genera evolutionarily distinct from Onthophagus. In regards to Onthophagus, sensu stricto, molecular evidence from Monaghan et al. (2007) also does not support monophyly. Tarasov and Kabakov (2010) discuss the difficulty with placing species within the subgeneric classification. Furthermore, the creation of subgenera based on Asian species that are derived from African lineages will certainly create issues of monophyly for the latter.
Two phylogenetic studies have been done that concentrate mainly on Onthophagus. One on the "Serrophorus" complex (Tarasov and Solodovnikov 2011) used morphological data and included 39 species of Onthophagus. This work also explored the use of male endophallic characters in the group; from 74 parsimony informative characters used, 45 (~60%) were coded from male genitalia. Emlen et al. (2005) used a sample of 48 Onthophagus species, employing up to seven gene sequences. Monophyly of the included taxa is assumed in both studies and it is possible that this is not the case. Regardless, more taxonomically broad studies on this genus and with molecular data are certainly needed to clarify relationships among subgenera and species groups. More genera then are currently recognized will likely be needed to delimit monophyletic clades in this speciose genus.
From various sources (Wikispecies website accessed 3 N. 2015, Zunino 1979, Lumaret and Kim 1989, Tarasov and Solodovnikov 2011, Tagliaferri et al. 2012, the subgenera of Onthophagus that perhaps are most commonly accepted are listed below (taxa in bold included in this study). Excluding five that most dung beetle workers recognize at the generic level (

Onthophagines?
There are a number of scarabaeine species that historically have been placed in various tribes, reflecting difficulty in their classification. This often included some of the inquilinous scarabaeines with highly modified morphologies, such as flattened leg segments and the presence of trichomes (e.g., Krell and Philips 2010). The formerly recognized Alloscelina (or Alloscelini) was once placed within the Scarabaeini, which included all the "rollers" at that time (Halffter and Matthews 1966). This group included the currently recognized onthophagine genera Alloscelus, Megaponerophilus (a former subgenus of Caccobius), Haroldius (including its junior synonyms Ponerotrogus Silvestri and Afroharoldius Janssens, both synonymized by Paulian (1985), Formicdubius Philips and Scholtz, synonymized by Krell and Philips (2010)) and the genus Freyus Balthasar (now a junior synonym of the dichotomine Paraphytus Harold (Halffter and Matthews 1966)).
Two problematic genera are Haroldius Bocomont and Cassolus Sharp. Hanski and Cambefort (1991) classify Haroldius as a Canthonini, while Philips and Scholtz (2000) placed it within the Onthophagini. Cassolus is listed as a canthonine in Halffter and Matthews (1966), but as an onthophagine in Hanski and Cambefort (1991). In this study, Cassolus and Haroldius appear deep within the onthophagines as sister taxa and supported by 12 character states, including two uncontroverted. Together with their sister clade, they are supported by five characters, including one that is uncontroverted. But they may indeed not be onthophagines. In Bai et al. (2011) using morphological evidence, they found Cassolus + Parachorius positioned outside the onthophagine clade. A recently described species of the Oriental canthonine Parachorius bridges the morphological gap between these two genera, and both form a monophyletic clade according to Tarasov and Keith (2011). At this time, the effects of morphological convergence remain unclear and evidence from molecular data is probably needed to resolve the origins of both genera and their proper tribal classification.

Antennal Pits
In Philips et al. (2004b), the Onthophagini and Oniticellini shared what was thought to be a fairly strong and intriguing synapomorphy of antennal cavities in the first (proximal) and middle antennomeres of the club. It was also seen in the canthonine Epirinus hilaris Péringuey but is lacking in E. silvestris Cambefort and seven other canthonines examined in Philips et al. (2004b). Interestingly, E. hilaris plus three other Epirinus species may be sister to the sisyphines, onitines, oniticellines and onthophagines, as seen in the molecular based topology of Monaghan et al. (2007). Some onitine taxa later examined, notably Platyonitis smeenkorum Krikken, were also noted to share this feature, while others (e.g. Bubas sp., Heteronitis sp., and Onitis spp.) did not. Interestingly, this feature is also found in the dichotomines Ontherus Erichson, Scatimus and "Trichillum" but not in Scatrichus and "Pedaridium" (F. Génier pers. comm.). Recently, Tarasov and Solodovnikov (2011) discovered that the Onthophagus subgenus Parascatonomus has only a single cavity on the first antennomere of the club. As these cavities are absent in a reasonably broad selection of other dichotomiines and coprines (and hence it is probably not a plesiomorphic condition for dung beetles), this trait appears to have independently evolved numerous times. The function of these antennal cavities is unknown.

Nesting behavior in the dung feeders
Ancestral oniticellines and onthophagines are presumed to have been coprophagous as all of the basal lineages are dung feeders. Tunneling behavior in this clade is very similar amongst taxa and all are classified as Pattern I nesters in Halffter and Edmonds (1982), with a few notable exceptions. An alternative nesting strategy known only within the Oniticellini is dwelling or endocoprid behavior (classified as Pattern VII nesters), which can include nests within dung or nests in pits directly below the dung. Known only in Oniticellus and Tragiscus (Davis, 1989), it may also occur in Paroniticellus although this is currently unclear (see Matthews 1966 andPhilips andBell 2008 for comments). Excluding Paroniticellus, this behavior only evolved once in the common ancestor of the sister genera Oniticellus and Tragiscus.
Both tribes are characterized typically and most distinctively by the absence of a brood ball and larvae are supplied with a brood mass that is either modified to some degree or not. Additionally, male-female cooperation is absent or limited and brood care is not known to exist. These are simple behaviors in what one could argue is a derived lineage compared to some of the more ancient origin tunneling clades where complex behaviors evolved. For example, coprines construct brood ovoids, and have extensive male-female cooperation, and brood care. But simple nesting behavior appears to be a very successful strategy evolutionary; the oniticellines and onthophagines are some of the most successful dung beetles in many ecosystems in terms of number of species, and often in terms of abundance and biomass (Davis et al. 2008).

Inquiliny
The African endemic Stiptopodius and Pinacotarsus groups of onthophagines form a monophyletic clade as discussed above. Based on the results, all included species are probably either myrmecophilous or termitophilous. As evidenced by the relatively high diversity within this lineage, the association with social insects may be relatively ancient. In contrast, only the New World oniticelline, Attavicinus montrosus, is an inqui-line. This species is associated with leaf-cutter ant debris piles and has a very restricted distribution in central Mexico (Philips and Bell 2008). There is no strong evidence for the age of these associations at this time. Krikken (1982) stated that the onthophagine genera Dorbignyolus, Krikkenius and Pinacopodius are all termitophilous. It is also thought that several others, including Pinacotarsus, Heteroclitopus, Stiptocnemis and Sukelus, are also known or likely termitophilous (Branco , 1992a. Directly above this termitophilous clade in a less derived position are two taxa in the genera Eusaproecius and Pseudosaproecius. Pseudosaproecius is considered a generalist in regards to food source by Branco (1995), with two species collected on fish carrion and a third on human dung (and see Cambefort 1984). But he also notes that nine species have been collected only at light. Similarly, only a single record from dung of a Eusaproecius species (of seven described) is known (Branco 1992a). The isolated dung and carrion records are probably not typical, but more likely may be supplemental food sources. A total of 26 specimens of Neosaproecius trituberculatus (Frey) were only collected in flight intercept traps and never at any bait in Ghana (Philips, unpublished data). Hence, it seems probable based on common ancestry, all of the species in this clade are symphiles of some type.
Above these clades (in less derived positions) in successive steps are three additional lineages also associated with social insects. The first, Alloscelus, is thought to be closely related to Pseudosaproecius (due to similar male genitalia and form of sexual dimorphism) but they do not form a monophyletic clade. The second and third, Megaponerophilus and Haroldius + Cassolus, are mainly myrmecophiles, with the exception of Cassolus where only one species of nine is known to be an ant associate (Halffter and Matthews 1966).
In summary, the evidence in this analysis suggests that there may be a single origin of inquiliny in the onthophagines. Several genera with unknown habits are probably associated with either ants (Eusaproecius and Pseudosaproecius) or termites (Sukelus). The other four genera may also be considered as part of this clade although future studies may shift Haroldius and Cassolus out of the onthophagines. The evolution of myrmecophily preadapting the more derived taxa for an association with the Isoptera also seems plausible.

Necrophagy, mycetophagy, and frugivory
Necrophagy and mycetophagy are rare within the oniticellines and onthophagines and perhaps are recently evolved behaviors with several independent origins. For example, Liatongus rhinocerulus (Morón 1984) and some species of Onthophagus are specialists on basidiomycete mushroom fruiting bodies, although L. rhinocerulus has also been found on carrion and most of the Onthophagus are probably generalist feeders (Halffter and Matthews 1966). Carrion feeding is a specialty of species of Amietina, is known in Caccobius, and in New and Old World species of Onthophagus. There are many taxa associated with feeding on millipede carrion as well, particularly in Africa, perhaps indicating a more ancient origin of this habit. Other alternative feeding habits, such as fruit feeding, may be more recently derived. One should note that the use of any of these foods by adults does not necessarily indicate their use as larval brood food.

Biogeography
The Oniticellini and Onthophagini are postulated to be a relatively modern aged group of dung beetles by Cambefort (1991). Davis et al. (2002) noted that the antiquity of generic distribution patterns lent support for an old age of the entire tribal and subtribal lineage divergence within the Scarabaeinae. Estimates for an origin of the common ancestor of these two tribes are the late Mesozoic (Davis et al. 2008, Philips 2011). More precisely, Mlambo et al. (2015) place the origin at ~early Miocene and the Oniticellini at mid to late Miocene. An Afrotropical origin of the common ancestor of both tribes is most likely, based on this study as well as Philips et al. (2004b) and Monaghan et al. (2007). Based on the results herein, most of the lineages are restricted to the Afrotropical region (Fig. 6). While all of the early branching as well as most of the generic level lineages of onthophagines are restricted to the Afrotropics with dispersals to other regions found in typically more derived lineages, the oniticellines show a pattern of relatively ancient lineage shifts into Madagascar, Palaearctic, and Nearctic regions and, more recently, into the Orient and the Caribbean.
Another general trend of the Oniticellini appears to be successive invasions from the Ethiopian into the Palaearctic through to the Oriental region (including the Lesser Sunda Islands and Sulawesi) in many groups (e.g., Tibiodrepanus (Barbero et al. 2011)). Currently the Drepanocerina have eight genera endemic to Africa while the Oniticellina have two. Hence, 14 generic level clades that evolved in Africa have dispersed to other geographic regions. In the Palearctic and the Caribbean, the Oniticellina have a single endemic genus in each region (Paroniticellus and Anoplodrepanus, respectively), while the Drepanocerina have a single endemic (Sinodrepanus) in Southeast Asia and southern China (Davis et al. 2002). The Oniticellina are known from the Mediterranean region but, in contrast, no Drepanocerina are currently found there, although fossil deposits from England dated at 70-100,000 years BP indicate they were once present (Krikken 2009, Barbero et al. 2009b. Based on basal branching, the oniticelline ancestors that led to the Madagascan helicopleurines, the Palaearctic Paroniticellus, and the New World Attavicinus and Liatongus may be relatively old dispersal events. Dispersal of an ancestral oniticelline from Africa to Madagascar during the early to middle Cenozoic may be the most likely hypothesis for the presence of the helictopleurines (see Yoder and Nowak 2006) and the clade origin has been dated to between 34 and 21 Mya (Wirta et al. 2008). In contrast and based on the low diversity of the Madagascar onthophagines, their presence there must be due to one or more relatively recent dispersal events, not including the modern accidental introductions.
The New World oniticellines belonging to Attavicinus and Liatongus may each represent separate invasions by ancestral species from Asia to North America via Ber- ingia. The land connection between these continents has been present from the mid-Cretaceous up through the late Pliocene 3.5 Mya and several more times during the Pleistocene (Sanmartín et al. 2001). Cambefort (1991) suggests the colonization of North America by these ancestors is not older than the Pliocene. Based on comparison with probable dispersal events in natricine snakes (Guo et al. 2012), invasions may have occurred even earlier during the late Oligocene or early Miocene ~27 Mya. If dispersal via tropical forest was required, the most likely dispersal event occurred during the Eocene (~50-35 Mya), when the northern hemisphere was much warmer and humid than it is today and a continguous boreotropical forest linked the two continents. At the end of the Eocene, a mixed deciduous hardwood forest changed to conifers and the rise of the Rocky Mountains and Sierra Madre Occidental created a barrier between the western and eastern Nearctic. It is not yet clear if the restricted distribution of these taxa in western North America is due to a mountain high elevation barrier or some other cause.
The Mexican Attavicinus monstrosus generally appears as sister to Paroniticellus, a monotypic genus known from middle Asia and Turkey. This close relationship is difficult to explain compared to the latter's alternate position in the K = 1 or 2 extreme weighted trees, where it is deep within a paraphyletic Oniticellina. But links between the eastern/western Palearctic with western Nearctic are certainly known (Sanmartín et al. 2001) and therefore a sister relationship is plausible. This lineage is sister to all other Oniticellina and Drepanocerina included in the study, evidence for its antiquity. But no doubt further support is needed for broad acceptance.
The two Nearctic Liatongus species, L. californicus (Oregon and northern California) and L. rhinocerulus (northern and central Mexico) most likely share a most recent common ancestor based on very similar external morphologies and close but disjunct distributions. The one representative included in this study, L. californicus, nearly always appears as sister to a larger clade that has as its basal lineage an Old World species of Liatongus. Zunino (1982), states that these two North American species of Liatongus belong to the Southeast Asian L. phanaeoides group. Thus, the position of the two included species of this genus in this study (one New World and one Old World) as sister taxa in the K = 3 weighted tree may indicate an accurate evolutionary relationship.
The Jamaican endemic Anoplodrepanus and an African representative of Euoniticellus appear as sister taxa in all topologies discovered. It is a relatively derived lineage appearing at the apex of the Oniticellini topology with its sister clade of Cyptochirus + Afrodrepanus. Euoniticellus has an African origin with Palaearctic and Asian species and a single New World member, E. cubiensis (Cambefort, 1996). A link between the faunas of Africa and the Caribbean has been reported elsewhere for many groups of insects and studies give both vicariance and dispersal hypotheses (e.g., Brown 1978, Flint 1977, Liebherr 1988). Both Anoplodrepanus and E. cubiensis may represent two invasions of the New World; one by perhaps by a Euoniticellus ancestral species that reached Jamaica and evolved into what we now recognize as Anoplodrepanus. A second invasion of a Euoniticellus ancestral species is supported by the position of E. cubiensis in a phylogeny of the genus done by Cambefort (1996). From a study on lizards (Gamble et al. 2011), it is possible that these two dung beetle dispersal events happened sometime between 6.0 to 21.9 Mya (but see Wiens et al. 2009 and references within for some contrasting evidence on colonization age and diversification in frogs).
Few biogeographic studies have been completed on individual genera. But in one on Eodrepanus (Barbero et al. 2009b), the oldest lineages are west and central or perhaps east African and a major split between the African and Oriental clades is seen using a variety of analyses (Ochiai similarity matrix, parsimony analysis of endemicity, and dispersal vicariance analysis). Evidence also suggests the effects of an initial African-Palearctic split and later further vicariant events and at least one more dispersal event in Asia.
The Onthophagini dispersal pattern is similar to that seen in the Oniticellini, but continued into Australia and South America, as well as everywhere else where one finds dung beetles. The African, Asian, and Palaearctic regions contain a total of 30, 11, and 3 genera, respectively. But of the total 35 Onthophagini genera, most are endemic to either the African (22 genera) or Asian (4 genera) regions (Davis et al. 2002).
Onthophagus alone is the most widespread as well as the most speciose genus, comprising at least 45% of the species of Scarabaeinae. It is also the only onthophagine genus that has spread outside of the Ethiopian, Oriental, or Palaearctic regions. Based on Emlen et al. (2005), this group has its oldest lineages in Africa and its youngest lineages in Australia and the New World. Dispersal to the New World may have occurred earlier than in the Oniticellini before the rise of the Western Cordillera by the late Eocene, although extensive erosion largely eliminated this barrier by 30 Mya, allowing dispersal between the western and eastern Neactic (Sanmartín et al. 2001). In the New World, the Oniticellini include six species, while the Onthophagini have 139 species (Davis et al. 2002, Delgado et al. 2006. Moreover, Onthophagus spread into South America (95 species) and occur at least as far south as 40th parallel (Cambefort 1991), whereas the oniticellines are completely absent from this region. The high diversity of the onthophagines (compared to the oniticellines) may be due to earlier dispersal and longer time for speciation or perhaps other factors not yet clear.

Conclusions
This study is the first with a broad range of taxa from these two tribes that strongly supports the monophyly of the Oniticellini + Onthophagini, and the Onthophagini. The Oniticellini are also monophyletic in the Bayesian analysis (Fig. 2) and in all weighted analyses with K values of 10 or less (Figs 1A, 3). In contrast, the unweighted and weighted analyses with K values of 11-30, the tribe Oniticellini is paraphyletic without the onthophagines (Fig. 4A). This is considered a less plausible hypothesis based on the lack of congruence with trees supported with molecular data (Monaghan et al. 2007, Wirta et al. 2008, Mlambo et al. 2015. In all topologies, the Helictopleurina are monophyletic. Recognition of this clade as a tribe would eliminate the potential paraphyly of the oniticellines in unweighted and weighted topologies (K = 11-30). But, since molecular data shows the helictopleurines as more derived oniticellines and not a basally originating lineage, the group should continue to be recognized as a subtribe. The Drepanocerina is supported as monophyletic. The Oniticellina is paraphyletic without the Drepanocerina and therefore redefinition of the former and the recognition of additional subtribes is needed if one desires to maintain monophyly in the classification.
The three most thorough molecular studies on the Oniticellini and Onthophagini generally support the monophyly of the two tribes combined and the Oniticellini (Monaghan et al. 2007, Wirta et al. 2008, Mlambo et al. 2015. But in contrast to the results herein, the Onthophagini are paraphyletic without the Oniticellini. The lack of congruence and the odd placements of taxa in the two older studies reduces the confidence in these hypotheses. In Monaghan et al. (2007) and Wirta et al. (2008) the Helictopleurina are placed in a more apical position within the oniticellines in contrast to a basal position in the present results. Additionally in some analyses done in Monaghan et al. (2007), some of the onthophagines such as Digitonthophagus + Phalops are sister to the onitines + oniticellines and remaining onthophagines. There are also genera, such as Oniticellus and Tiniocellus, where species of each appear in separate clades. In Wirta et al. (2008), there is a mix of oniticelline and onthophagine taxa in the topology, as well as a species of Onitini positioned within this clade. In contrast, Mlambo et al. (2015) support the onitines as sister to the onthophagines + oniticellines as hypothesized by Philips (2004b). Lastly, the speciose genus Onthophagus, based on both molecular and morphological evidence, is not monophyletic and will no doubt need future subdivions. Clearly the assumption of monophyly of this genus or any of the species or generic groups proposed should not be assumed.
Certainly larger molecular data sets and perhaps large numbers of morphological characters from a broad range of taxa is needed to help stabilize our conclusions and understanding of the evolution of these two major tribes of Scarabaeinae. Morphological phylogenies have the advantage that rarely collected taxa can be included. Regardless, future molecular studies using new techniques will produce large amounts of data and will clarify the evolution of these tribes and major clade divergence times (but see Mlambo et al. 2015 for divergence estimates of Afrotropical lineages). Although further study is needed on these tribes and scarabaeine evolution in general, the picture of their extensive diversification is slowly becoming clearer.
review. I am grateful to Robert Neidlinger who photographed and carefully edited most of the automontage images of the specimens illustrated in the trees. Lastly, thanks to Harry Taylor and Max Barclay (Natural History Museum, London) for arranging photography of the Sukelus and Cassolus specimens.