The specimens used in this study were borrowed from the following collections, with curators in parentheses:

NHM – The Natural History Museum, London, England (David Notton); CASC – California Academy of Sciences, San Francisco, U.S.A. (Robert Zuparko); HNHM – Magyar Természettudományi Múzeum, Budapest, Hungary (Gellért Puskás); MCSN – Museo Civico di Storia Naturale “G. Doria”, Genova, Italy (Roberto Poggi); MNHN – Muséum National d’Histoire Naturelle, Paris, France (Claire Villemant); MRAC – Musee Royal de l’Afrique Centrale, Tervuren, Belgium (Eliane De Coninck); QSBG – Queen Sirikit Botanical Garden, Chiang Mai, Thailand (Wichai Srisuka); UFES – Universidade Federal do Espírito Santo, Vitória, Brazil (Celso Azevedo); USNM – National Museum of Natural History, Washington D.C., U.S.A. (David Furth); ZMBH – Museum für Naturkunde, Berlin, Germany (Frank Koch).

The images were obtained using a Leica MZ80 Stereomicroscope attached to a Leica DFC 495 video camera and captured with LEICA LAS (Leica Application Suite V3.6.0) by Leica Microsystems (Switzerland), using a dome illumination system described by Kawada and Buffington (2016), and combined using HELICON FOCUS (version 4.2.9). Illustrations and plates were edited for adjustments (e.g., levels, shadows/highlights).

The terms applied to the structures follow Lanes et al. (2020) and Barbosa and Azevedo (2011), integument terminology follows Harris (1979). Abbreviation: VOL = vertex-ocular line in dorsal view.

The ingroup is composed by males of 61 species (Table 1). The species analyzed correspond to over a third of the 182 species in Mesitiinae. Species selection aimed to cover the maximum possible morphological diversity in each genus to facilitate possible taxonomic decisions. Character definition was based on males for three main reasons: (1) the current classification by Argaman (2003) was based on males; (2) the male hypopygium and genitalia offer a range of characters not available for females; and (3) lack of conspicuous sexual dimorphism between male and female.

Except for Australomesitius Barbosa & Azevedo, 2016, known only from the female, 17 out of the 18 genera currently included in the subfamily were sampled. The outgroup includes representatives of all extant subfamilies of Bethylidae (Table 1). The BethylinaeBethylus cephalotes (Förster, 1860) was used for rooting the tree.

Table 1 should be included here associated with subchapter 2.4, landscape, and maximally page-filling.

A total of 112 characters (Appendix 1) were analyzed. Many of them were taken from descriptions in Móczár (1970, 1971), Nagy (1969, 1972), and Argaman (2003); additionally, new characters are proposed here for the first time.

The character matrix (Table S1) was produced using DELTA software (Dallwitz et al. 1993). All characters were treated as unordered. Inapplicable characters were coded as “?”.

The searches for the most parsimonious trees were carried out in TNT version 1.5 (Goloboff et al. 2016, using the Ratchet, Sectorial Searches and Tree-Fusing searching strategies (Goloboff 1999, Nixon, 1999). Parameters were as follows: collapsing rules selected for TBR; random seed set to 0; Sectorial Search in default mode; 200 iterations of Ratchet; 20 cycles for Drift; 10 rounds for Tree Fusing.

It has been argued that results based on characters properly weighted are to be preferred over those with all characters equally weighted (Farris 1969, Goloboff 1993, Goloboff et al. 2008a). Implied weighting is the most widely used method for attributing different weights during tree search, as it is independent of previous analyses and weighting schemes unlike, for example, successive weighting (e.g., Farris 1969). The weighting against homoplasy under implied weighting is related to a constant k — the lower the value of k, the higher the strength against homoplasy Goloboff et al. (2008b). Here, we used the TNT script setk.run, written by Salvador Arias (Instituto Miguel Lillo, San Miguel de Tucuman, Argentina), to calculate the value of k. The script returned a value of k = 11.674805 for our data set.

Bayesian analyses were conducted in MRBAYES 3.2.7 (Ronquist et al. 2012). We used the Mk model to morphological data, with correction for ascertainment bias (lset coding = variable), since autapomorphic characters were included. We first conducted an analysis without partitioning the original matrix, accounting for among-character rate heterogeneity using a discrete Gamma distribution with four rate categories (lset rates = gamma) and the prior on branch lengths described by an exponential distribution with scale parameter = 10 (prset brlenspr = Unconstrained:Exp(10)). We also conducted a similar analysis partitioning characters according to their degree of homoplasy. For this purpose, we retrieved homoplasy scores from implied-weighting analyses in TNT (see above). These values, derived from Goloboff’s measure of homoplasy, are normalized between 0 and 1, with the lowest value representing no homoplasy (Goloboff et al. 2008b). Branch lengths were maintained linked among partitions, and site-specific rates within partitions were not considered, as suggested by Rosa et al. (2019). MCMC analyses ran for 5,000,000 generations, sampling every 1,000, with four chains, and two independent runs. Convergence was assessed with Tracer 1.6 (Rambaut et al. 2018). Trees shown are majority-rule consensus trees (Contype = Allcompat).

The implied weighting analysis using k = 11,674805 resulted in one most parsimonious tree, with 675 steps, fit = 23,72727, consistency index (CI) = 0.19, and retention index (RI) = 0.61 (Figs 3–4; S1). The tribal classification proposed by Argaman (2003) was not supported, corroborating the classification taken by Azevedo et al. (2018).

Twenty-four characters were found as synapomorphies for the subfamily, 14 of them are exclusive transformations for Mesitiinae as listed below:

Ch. 5:1 malar space projected (Fig. 1B);

Ch. 23:0 contour of eye protruding (Fig. 1A);

Ch. 28:1 anterior depression of occiput present (Fig. 1C);

Ch. 31:0 ventral half of mesoccipital carina angled - (Fig. 1E);

Ch. 41:0 notauli of mesoscutum convergent posteriorly (Fig. 1F);

(Ch. 49:1 metapostnotal depression present (Fig. 2A);

Ch. 50:1 connection between central depression and triangular lateral depression of metapectal-propodeal disc (Fig. 2A);

Ch. 72:1 space between tegula and mesoscutum present (Fig. 1F);

Ch. 77:1 prestigmal abscissa of radial 1 of forewing present (Fig. 2C);

Ch. 85:1 constriction between metasomal sternite I and II present (Fig. 2B);

Ch. 87:1 lateral ventral lap metasomal tergum I present (Fig. 2B);

Ch. 89:1 metasomal segment II longer than others (Fig. 2B);

Ch. 101:1 projection of genital ring present (Fig. 2F);

Ch. 105:1 fusion between gonostipes and basivolsela present (Fig. 2F).

The other 10 characters states found as synapomorphies are not exclusive for Mesitiinae, but contributed to define the subfamily:

Ch. 7:1 orientation of malar space parallel (Fig. 1C);

Ch. 8:1 inner keel of mandible present (Fig. 1C);

Ch. 14:1 torulus and median clypeal carina fused (Fig. 1A);

Ch. 20:0 flagellomeres 1-11 slender (Fig. 1A);

Ch. 21:0 eye small (Fig. 1B);

Ch. 26:2 anterior ocellus crossing supra-ocular line (Fig. 1A);

Ch. 37:0 anterior margin of propleuron angled (Fig. 1E);

Ch. 81:0 hind wing with three distal hamuli (Fig. 2D);

Ch. 94:1 hypopygium as long as wide (Fig. 2E);

Ch. 106:1 cuspis with two arms (Fig. 2F).

Bayesian analyses largely corroborated the backbone of the relationships retrieved in parsimony (Figs S2, S3). Results from unpartitioned and partitioned analyses differed. In both analyses the genus Anaylax was not recovered as monophyletic, with Anaylax simplicitus being recovered as a distinct lineage relative to other species of the genus included in the present account. The unpartitioned analysis recovered Incertosulcus krombeini as sister group to Metrionotus, while in the analysis using partitioning by homoplasy score it was recovered nested within Metrionotus. However, the posterior probability of the clade Metrionotus + I. krombeini was very low in both cases (i.e., < 0.42). Both analyses also recovered different taxa as the sister group to all other mesitiine lineages: Moczariella centenaria in the unpartitioned analysis and the genus Bradepyris in the partitioned analysis.

The interpretation of topologies obtained allowed us to propose 17 nomenclatural changes: two new genera, one genus synonymy, three revalidations in species status, and 11 new specific combinations (Figs 5–6; Table 2, S2). Because a recent review for diagnostic characteristics for Bethylidae genera was published by in Azevedo et al. (2018), we describe here only the diagnostic characteristics for the new genera proposed and the changes for the genera reinterpreted in this work.

Argaman (2003) mentions that the posterior oblique sulcus of mesopleuron was an autapomorphy for Mesitiinae, but this was not found as a subfamily synapomorphy. From twenty-five subfamily diagnostic features defined by Azevedo et al. (2018), eight were found as synapomorphies: malar space projected (#5:1); inner keel of mandible present (#8:1); torulus and median clypeal carina fused (#14:1); eye small (#21:0); contour of eye protruding (#23:0); anterior depression of occiput present (#28:1); anterior margin of propleuron angled (#37:0); notauli convergent posteriorly (#41:0). The metapostnotal depression present (#49:1), metapostnotal depression and paraspiracular sulcus metapectal-propodeal disc present (#50:1), space between tegula and mesoscutum present (#72:1), and anterior margin of propleuron angled (#37:0) were found as synapomorphies for the first time. The metasomal segment II longer than others the only feature previously found as a Mesitiinae synapomorphy, by Sorg (1988).

Although Argaman (2003) stated that his tribal classification was based on monophyletic grouping, he did not publish this phylogenetic analysis, and analyses by Azevedo et al. (2018) did not support his results. Argaman (2003) described Triglenusini based on the characters shared by Bradepyris Kieffer, 1905, Pseudomesitius (Duchaussoy, 1916 [1914], and Triglenus Marshall, 1905. The two latter genera were treated as junior synonyms of Bradepyris by Barbosa & Azevedo (2015), and our analysis showed that Triglenusini is a paraphyletic grouping. The Mesitiini were described based on the characters shared by Anaylax Móczár, 1970a, Clytrovorus Nagy, 1972, Incertosulcus Móczár, 1970aa, Itapayos, Mesitius Spinola, 1853, Metrionotus Móczár, 1970a and Parvoculus Móczár, 1970d, and were recovered as polyphyletic in our analysis. The Heterocoeliini were described based on characters shared by Botoryan, Codorcas Nagy, 1972 (junior synonym of Heterocoelia), Gerbekas, Hamusmus (junior synonyms of Heterocoelia), Heterocoelia Dahlbom, 1854, Pycnomesitius Móczár, 1971b, Sulcomesitius Móczár, 1970c and Ukayakos (junior synonyms of Heterocoelia); however, Heterocoeliini were not supported as a monophyletic group, being recovered as polyphyletic instead. The tribe Domonkosini was based on characters shared by Domonkos (junior synonym of Incertosulcus), Pilomesitius Móczár, 1970b, Topcobius Nagy, 1972 (junior synonym of Sulcomesitius) and Zimankos, and was not supported as a monophyletic group, the evidence indicating that Domonkosini are a polyphyletic grouping. The same relationships were also observed in the Bayesian analysis. Therefore, we corroborate the decision of Azevedo et al. (2018) and maintain Mesitiinae without tribal classification.

Among genera represented by more than one terminal in parsimony analyses, five were found to be monophyletic: Astromesitius Barbosa & Azevedo, 2019, Bradepyris, Clytrovorus, Itapayos, and Mesitius; three were found to be paraphyletic: Botoryan, Metrionotus, and Pilomesitius; and eight were found to be polyphyletic: Anaylax, Gerbekas, Heterocoelia, Incertosulcus, Parvoculus, Pycnomesitius, Sulcomesitius, and Zimankos. Additionally, the paraphyly of Metrionotus and Pilomesitius and the polyphyly of Gerbekas, Heterocoelia, Pycnomesitius, Sulcomesitius, and Zimankos required new combinations (see below). Botoryan was found forming a clade nested within Zimankos and therefore we treat it as junior synonym of the latter. Anaylax simplicitus Barbosa & Azevedo, 2011 and Incertosulcus krombeini Móczár, 1970 were both recovered as distinct lineages, not clustering with other species of their respective genera. Mesitius krombeini was retrieved as closely related to the clade formed by Zimankos + Botoryan. Sulcomesitius nepalensis Móczár, 1986 clustered with species of Metrionotus. Gerbekas laosensis Móczár, 1975 was recovered as related to four species of Sulcomesitius.

The topologies obtained allowed the identification of morphological characters which potentially played important roles during the diversification of Mesitiinae, including sculpture of frons (#24); sculpture of dorsal pronotal area (#33); presence of median pronotal line (#36); presence of posterior propodeal projection (#63); length of hypopygium (#94); shape of posterior hypopygeal margin (#96); and length of hypopygium branches (#98). These characters are unique to the subfamily and allow us to hypothesize about their evolution. These hypotheses are largely based on convergent characteristics shared between Mesitiinae and Chrysidinae (Chrysididae) (Argaman, 2003).

From the 19 genera proposed for Mesitiinae, 15 exhibit roughly sculptured frons (character #24, state 1) and pronotal area (character #33, state 1), with foveolate patterns (Figs 1A, D), while only two genera completely lack these features. Mesitiinae attack beetle larvae of Chrysomelidae (Coleoptera) (Argaman 2003), the author found them living into ant nets, hence the thick and robust integument of Mesitiinae is presumably associated with the lifestyle of hosts. Michener (2000) postulated that the rough sculpturation (lamellae, carinae and foveolation) and projections could be related to the strengthening of the integument, providing defensive mechanisms for vulnerable areas such as the neck, base of metasoma, and other membranous regions in kleptoparasitic bees. Similar integumental sculpturation is observed in Nyssonini (Crabronidae), a group of apoid wasps that also exhibit kleptoparasitic behavior (Bohart and Menke 1976) and Chrysidinae (Chrysididae), which attack bees and aculeate wasps (Kimsey 1992). The same and convergent features can be observed in Mutillidae (Ronchetti and Polidori 2020). Thus, the integumental thickening in Mesitiinae seems to be associated with defense against their aculeate hosts, as mentioned by Lucena and Almeida (2022?) for Chrysidinae.

Cryptocephalini and Clytrini (Cryptocephalinae) leaf beetles have close association with ant nests. The larval stages remain in the ant nest in a positive interaction. The cocoon brought by the beetle mother is carried by the ants into the nest to complete its development (Agrain et al. 2015). Thus, to reach their beetle hosts, the Mesitiinae need to enter the ant nests. The convergent behavior among all taxa above is that all of them have dangerous hosts (bees and ants), as cited Thus, dense foveolation and thick integument could be associated with defenses.

The median pronotal line (Character #36, state 1) characteristic of many mesitiine lineages (Fig. 1D) is associated with the pronotum-postoccipital muscle (Vilhelmsen et al. 2010), which has its origin at the internal ridge associated with this impression. This ridge could increase the anchorage insertion point allowing stronger contractions. The muscle is the pronotal elevator of the head, so the increase of power for this muscle allows more possibilities for head movements. Nagy (1968) described the parasitoid behavior of females of Mesitiinae and recorded that they steal pre-pupal beetles from the ant nests using their mandibles. Therefore, the increased range of head movements could be adaptative in the context of the female parasitoid behavior.

Among bethylids, the posterior projections on the propodeum (Fig. 2A) are exclusive for Mesitiinae (Character #63, state 1). However, it is absent in Anaylax, Astromesitius, Bradepyris, Clytrovorus, Hadesmesitiusgen. nov., and Moczariella. Argaman (2003) argued that this posterior projection could facilitate the opening of the cocoon wall during adult emergence, but this was never confirmed for Mesitiinae species. Perhaps a more plausible hypothesis is that the posterior propodeal projection is associated with defense against ants, with the projection protecting the base of metasoma, preventing damage to the petiole.

On the other hand, the musculature could indicate another adaptation associated with this structure. The muscle T1-S/T2 has its origin at the posterior corner of the metapectal-propodeal complex and inserts at anterior margin of second metasomal segment. We dissected some Mesitiinae specimens with this projection and observed that the T1-S/T2 muscle has its origin inside the projection, thus increasing the anchorage insertion point of this muscle and giving it more strength. Additionally, this muscle is related to the sternal torsion of the metasoma (Mikó et al. 2007).

All mesitiine wasps have the second metasomal segment longer than the others, an exclusive feature for the subfamily (Barbosa and Azevedo 2011), which was also recovered herein as a synapomorphy for Mesitiinae. The great degree of metasomal segment modification, mainly the length of the third segment longer than others, is associated with oviposition, copulation and defense. Moreover, an additional muscle row was recorded in association with this segment expansion (Kimsey 1992), which is an additional anchorage point for the T1-S/T2 muscle into the posterior propodeal projection.

There are several shapes of hypopygium exclusive to Mesitiinae, which are described by seven characters in the present analyses (Characters #94 to #100).

Clade A (Fig. 4 and 6) includes seven genera, comprising 113 described species, which represent 63.8% of the total diversity of the subfamily. In this clade are also included the largest species of Mesitiinae.

Muscles located at the base of male genitalia are responsible for movements such as protraction as well as copulation, being inserted at the anterior region of the hypopygium, including the spiculum and anterolateral apodeme. Therefore, the contraction and relaxation between the genitalia base and the spiculum provides the movement of genitalia structures. Thus, the shape and size of the spiculum have direct association with insertion of muscles in the genitalia, affecting the kind and potential of its movements that is, with more and diversified muscle insertions, structures will be capable of performing more complex movements.

The modification of the length of the hypopygial branches is associated with the deformation of the hypopygium, which results from the contraction of the muscles. More muscles inserted at a longer spiculum promote a higher degree of hypopygium deformation, hence the long branches are associated with long median indentation, providing the hypopygium with an area of deformation, giving the structure more flexibility. This is also associated with a wide spiculum. On the other hand, short branches are associated with a simple acute spiculum, since the muscle contraction provides less deformation to the hypopygium, without the need of a deformation area.

The hypopygium shape is associated with muscle insertion and hence it could provide specific functions and adaptations for each genus. Schulmeister (2003) recognized these muscles and named them as “a”, “b” and “c”. These muscles originate in the gonocondyle in the cupula and inserts at the spiculum and laterally at the ninth sternite (= hypopygium). The cupula is attached to the male genitalia base, and the muscles among these sclerites promote some movement of the genitalia, thus the muscles gonocondyle-spiculum (a), laterally of gonocondyle-spiculum (b), and gonocondyle-laterally ninth sternite (c), have indirect action in male genitalia action (Schulmeister 2003). These muscles expose the male genitalia by the elevation of the basal margin of the cupula (Boudinot 2013), thus probably a larger insertion point could increase the torque movement.

Presently, Mesitiinae are known from warm regions of the Old World, encompassing all of its four zoogeographical regions: Afrotropical (including Madagascar), Australian, Oriental, and Palearctic (Fig. 7). Unfortunately, Australomesitius mirus Barbosa & Azevedo, 2006, the single species of the family ever recorded in Australia, could not be included in this analysis. However, its position can be inferred based on the shape of the apex of median clypeal carina, arched [in profile], the presence of fusion between sublateral and inner discal carina of propodeal disc, and the orientation of inner discal carina of propodeal disc not parallel with median carina. Observing the distribution patterns among Mesitiinae lineages, there is an apparent association of the early-diverging lineages (e.g., the genera Bradepyris and Moczariella) with the Palearctic region. Additionally, thirteen of nineteen Mesitiinae genera have species recorded from the Palearctic region. From the remaining six genera, Australomesitius is endemic of Australia and Pilomesitius is endemic of Madagascar, while the other four are recorded in the Oriental and Afrotropical regions.

Based on this pattern, it may be that the early diversification of Mesitiinae occurred in the Palearctic region, with lineages progressively occupying the adjacent Oriental and Afrotropical regions and, later, Australia and Madagascar. Occupation of Madagascar likely occurred several times independently within the subfamily, while only a single lineage was able to reach Australia. In the future, a dated phylogeny of the family allied to the exploration of its relationships among other bethylid lineages and discoveries regarding fossil history may provide valuable evidence for a detailed approach on biogeography and diversification of Mesitiinae.