A total of 35 species in 15 genera in three tribes of Proctotrupinae were sampled as ingroups (Table 1). Taxa were chosen to cover the taxonomic range of genera in the subfamily Proctotrupinae. The type species were included in 12 genera. Outgroup taxa were selected based on the phylogenetic relationships shown in Blaimer et al. (2023). These outgroups were four species in two genera: Vanhornia hikosanensis Abe, Yamagishi & Konishi, 2024 and V. yurii Timokhov & Belokobylskij, 2020 in Vanhorniidae Crawford, 1909, and Helorus ruficornis Foerster, 1856 and unidentified species in Heloridae Foerster, 1856.

Voucher specimens were deposited in the following institutions and detailed information on all the samples is provided in Table SS1.

This study is based on specimens deposited in the following institutions: ELKU – Entomological Laboratory, Faculty of Agriculture, Kyushu University, Fukuoka, Japan; NBC – Nagoya Biodiversity Center, Nagoya, Japan; NARO – Insect Museum, National Agriculture and Food Research Organization, Tsukuba, Japan; EUM – Ehime University Museum, Matsuyama, Ehime, Japan; KPMNH – Kanagawa Prefectural Museum of Natural History, Odawara, Japan; OMNH – Osaka Museum of Natural History, Osaka, Japan; SEHU – Hokkaido University Insect Collection, Sapporo, Japan; TARI – Taiwan Agricultural Research Institute, Taichung, Taiwan.

All terms of proctotrupid morphology followed Townes and Townes (1981), Ernst et al. (2013), Butcher and Quicke (2023), Dal Pos et al. (2023) and the Hymenoptera Anatomy and Ontology (HAO) (Yoder et al. 2010).

Morphological data were scored for the included taxa based on physical examination of all specimens. Biological characteristics were coded according to previous records and were included in the matrix. The character matrix is presented in Table S2.

Photographs, except for the dissected parts, were taken using a Canon MP-E65 mm macro lens mounted on α Sony 7R IV digital camera. Individual photos were stack-combined with a Zerene Stacker (Zerene Systems LLC) and processed in Adobe Photoshop CC (Adobe).

We dissected a part of specimen in our taxon sampling to examine the female terminalia. The dissected parts were sorted using 10% KOH solution and mounted in Euparal on a glass slide. Photographs of the dissected parts were taken using Olympus SZX7 and Nikon ECLIPSE Ci–L microscopes.

For DNA extraction, 29 species in 15 genera within the ingroup and four species in two families within the outgroup were used (85% of OTUs). Voucher specimens were included among the examined specimens used to construct the matrix (Tables S1, S3). DNA was extracted from a single leg of each specimen using the protocols in the DNeasy Blood and Tissue kit (Qiagen, Japan). Four gene regions, mitochondrial 16S, nuclear 18S, 28S D2–3, and RNA polymerase II (POLII), were amplified. PCRs were carried out using 10 μL reactions containing 1 μL DNA, 0.3 μL of forward primer, 0.3 μL of reverse primer, 5 μL KOD One Mix Blue (Toyobo, Japan) and 3.4 μL RNAse free water. The primers used are listed in Table S4. The PCR products were purified using ExoSAP-IT Express (Thermo Fisher Scientific, Tokyo, Japan) and subjected to Sanger sequencing at Azenta Life Science (Tokyo, Japan). Sequencing data were deposited in the DDBJ under the accession numbers listed in Table S3.

All sequencing data were aligned by L-INS-i method (Katoh et al. 2005) under the auto strategy setting using Mafft 7 (Katoh et al. 2019) (online). The aligned data were trimmed using option “-automated1” that uses a heuristic selection of the automatic method based on similarity statistics in Trimal 1.5 (Capella-Gutierrez et al. 2009).

Maximum likelihood (ML), Bayesian inference (BI), and maximum parsimony analyses were conducted. The maximum parsimony analysis was conducted based on morphological data. ML and BI analyses were conducted based on DNA and total evidence combined with DNA and morphological data. The best substitution models for all the analyses were estimated using Modelfinder (Kalyaanamoorthy et al. 2017).

For ML, the best-fit models for each gene region are represented in Table S5. The analysis was performed using Iqtree ver. 2.4 (Minh et al. 2020), with 1000 replicates of an Ultrafast bootstrap (Hoang et al. 2018) and an SH-like approximate likelihood ratio test (Guindon et al. 2010). We conducted the analysis 11 times individually, checked that each results showed same topology, and selected the tree that had the median log-likelihood.

For BL, the best fit models were searched using option “-mset mrbayes”. The proposed models are presented in Table S5. BI analysis was conducted using Mrbayes ver. 3.2.7a (Ronquist et al. 2012). All parameters were unlinked between each partition. We run two independent runs with eight chains each for 50 million generations, sampling 5000 generations. After verifying the average standard deviation of split frequencies (ASDSF) less than 0.01, we excluded the first 25% of sampled trees as burn-in for diagnosis, and we used same burn-in to summarize the parameters and trees. We used Tacer v1.7.2 (Rambaut et al. 2018) to check that the Markov chain Monte Carlo (MCMC) runs reached a state of convergence and the effective sample size (ESS) was over 200 in all parameters.

The maximum parsimony analysis was conducted under both non-additive and implied weighting using TNT ver 1.6 (Goloboff and Morales 2023). The equal weighting analysis was conducted under default settings, except for the following: memory was set to 99,999 trees, 30,000 replicates, and 3,000 trees saved per replication in the traditional searches. The implied weighting analysis was conducted using an implied weighting search (k=5.0000) under the same settings in the equal weighting analysis. The GC values were calculated using 1,000 replicates in both analyses. The supporting unambiguous characters were shown by Asado ver. 2.0 (previously named Winclada 1.00.08) (Nixon 2002).

To mapping the characteristics on the estimated tree, we performed ancestral state reconstructions (ASR) in Mesquite v3.80 (Maddison and Maddison 2023) on three biological characters, host taxon, habitat of host, and parasitism. The character states were based on reference data and listed in Table S6. We used the BI tree based on total evidence as topological tree. We used a maximum likelihood reconstruction with equal rate model (Mk1).

(see Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17; Table S2 for the matrix and ASR (Table S6))

In total, 157 states from 65 morphological characters and eight states from three biological characters were coded.

Antenna

0 Shape of flagellomere 9 and 10: 0, bullet-shaped (Fig. 7A); 1, rectangular (Fig. 7B).

1 Tyloid on flagellum in male: 0, tyloid absent; 1, absent on F11; 2, present on F11.

2 Shape of tyloid on flagellum in male: 0, tyloid absent; 1, ridge (Fig. 7C); 2, round (Fig. 7D).

Head (Fig. 1)

3 Length of face compared to inner distance between eyes: 0, 1.1<; 1, 1.1–1.3; 2, <1.3

4 Shape of clypeus, as seen in ventral view: 0, normal or weakly protruded; 1, strongly protruded (Fig. 7E).

5 Ventral margin of clypeus: 0, almost straight (Fig. 7F); 1, curved (strongly concave in Disogmus (Fig. 8A), weakly convex in Nothoserphus Brues, 1940 (Fig. 8B)).

6 Length of ventral margin of clypeus compared to inner distance between eyes: 0, 0.4≦; 1, <0.4.

7 Mandible, as seen in anterior view of head: 0, weak and thin (Fig. 8A); 1, stout (Fig. 8C).

8 Subapical teeth on mandible: 0, absent; 1, present (Fig. 8D).

9 Gena: 0, not bulge; 1, bulge, almost carinate (in Nothoserphus, Fig. 8E). This character called “cheek” in Townes and Townes (1981) and He and Xu (2015).

10 Malar sulcus: 0, absent; 1, present (sometimes incomplete).

11 Shape of temple in dorsal view: 0, almost flat (Fig. 9A); 1, round bulge (Fig. 9B).

12 Distinct ridge between antennal foramina: 0, absent; 1, present (Fig. 9D).

13 Structure between antennal foramina: 0, absent; 1, foveate; 2, small projection (Fig. 9C); 3, short carina (different structure from clear ridge in Char.12).

14 Fovea on vertex: 0, absent; 1, present (in Nothoserphus, Fig. 9E).

15 Occipital carina: 0, incomplete (only dorsal half or shorter); 1, complete or almost complete.

16 Margin of occipital carina: 0, smooth; 1, with wrinkles (in Maaserphus Lin, 1988, Fig. 9F).

17 Hypostomal carina: 0, distinctly separate from occipital carina (Fig. 10A); 1, reaching or almost reaching occipital carina (Fig. 10B).

Mesosoma (Figs 2, 3)

18 Dorsal part of pronotum: 0, smooth; 1, partly foveolate (sometimes weakly); 2, rugose.

19 Ratio of pronotal length (from dorso-lateral margin to ventro-lateral margin) to width (from ventro-lateral margin of pronotal shoulder to postero-lateral margin of pronotum) (Fig. 2): 0, 1.2<; 1, ≦1.2.

20 Inner pit of pronotal shoulder: 0, absent (Fig. 3A); 1, present (Fig. 10C). In Proctotrupidae, pronotum sticks out dorso-laterally (Figs 2, 3A) and it is called “pronotal shoulder”.

21 Lateral margin of pronotal shoulder: 0, not carinate, fused with anterior part and lateral part; 1, developed and carinate (Fig. 10D).

22 Lateral apex of pronotal shoulder: 0, round (Fig. 10E); 1, weakly projecting (Fig. 10F); 2, projecting (Fig. 10D).

23 Longitudinal groove on pronotum: 0, absent; 1, present (Fig. 10F).

24 Epomia: 0, reaching anteriorly and not connecting to pronotal shoulder (Fig. 10F); 1, reaching dorsally and connecting pronotal shoulder (Fig. 10D).

25 Mesothoracic spiracle: 0, touching postero-lateral margin of pronotum (Fig. 10D); 1, separated from postero-lateral margin of pronotum (Fig. 10E).

26 Length of notaulus: 0, notaulus absent; 1, short (same as length of tegula, Fig. 3A); 2, long (reaching to mid length of mesoscutum, Fig. 11A, B).

27 Shape of notaulus: 0, notaulus absent; 1, straight; 2, more or less curved.

28 Horizontal groove on mesopleuron (Fig. 2): 0, absent; 1, present. This character listed in HAO as transepisternal line (see discussion).

29 Shape of horizontal groove: 0, straight (Fig. 11C, D); 1, weakly curved; 2, curved or strongly curved (Fig. 12A).

30 Mesepimeral sulcus: 0, smooth; 1, punctate to foveolate, each fovea clearly separate (Fig. 11C); 2, strongly foveolate, each fovea close and almost contiguous (Fig. 11D). This character is called as mesopleural suture in Townes and Townes (1981) and He and Xu (2015).

31 Fovea on mesopleural suture: 0, absent (mesopleural suture smooth); 1, present in dorsal half (Fig. 11C); 2, almost complete to complete (Fig. 11D).

32 Mesodiscrimen: 0, smooth (without foveation); 1, posterior half foveolate; 2, evenly foveolate (Fig. 12B). Sometimes the foveation is weak.

33 Anterior discrimenal pit: 0, absent; 1, present, but weak; 2, present.

34 Metapleural smooth area: 0, absent (Fig. 12C); 1, narrow, cover 0.1 times of metapleuron or less (Fig. 12D); 2, large, cover 0.5 times of metapleuron or more (Fig. 12E).

35 Metapleural carina (carina connecting dorsal margin of metapleural smooth area and antero-dorsal margin of propodeum): 0, absent (Fig. 12E); 1, present (sometimes weak or incomplete) (Fig. 13A).

36 Smooth area on dorsal part of propodeum: 0, absent (dorsal part of propodeum areolate) (Fig. 13B); 1, narrow, cover 0.2 times of dorsal part of propodeum or less (Fig. 13C); 2, large, cover 0.5 times of dorsal part of propodeum or more (Fig. 13D).

37 Dorsal propodeal carina (longitudinal median carina on smooth area): 0, absent; 1, present (Fig. 13D).

38 Ratio length of smooth area on dorsal part of propodeum to width: 0, smooth area absent or very narrow; 1, 0.8≦; 2, ≦1.0.

Leg

39 Maximum length of hind tibial spur compared to length of hind basitarsus: 0, short (less than 0.4 times); 1, normal (0.4–0.75 times as long); 2, long (over 0.75 times) (in Cryptoserphus Kieffer, 1907, Fig. 13E).

40 Black tooth base of tarsal claws: 0, absent; 1, present (in Exallonyx, Fig. 13F)

Venation (Fig. 4)

41 Vein r-rs: 0, absent (in Brachyserphus, Fig. 14A); 1, very short (Fig. 14B); 2, relatively long (Figs 4, 14C).

42 Ratio of length (depth) of stigma to width: 0, 0.7< ; 1, ≦0.7 (stigma wide).

43 Vein R1: 0, reaching apical margin of RS; 1, reaching beyond RS (Fig. 14C).

44 First and second discal cell: 0, confluent (1m-cu absent); 1, separate (1m-cu present, Fig. 4).

Metasoma (Figs 5, 6)

45 Longitudinal carina on antero-ventral part of petiole (aps in Fig. 5A, B): 0, absent (antero-ventral part of stalk smooth in Disogmus, areolate in Nothoserphus) (Fig. 14D); 1, present (Fig. 14E, F).

46 Petiole: 0, stalk formed; 1, segmented (in Heloridae)

In Proctotrupidae, petiole is formed by anterior part of synsternite and isn’t segmented. In this paper, we distinguished between these character states. We defined the petiole derives from synsternite like Proctotrupidae as “petiole”, and one from the segment confused tergite 1 and sternite 1 like Heloridae as “segmented petiole”.

47 Petiole (Fig. 5): 0, absent (in outgroup); 1, present.

48 Petiole visibly: 0, hidden by syntergite (Fig. 15B); 1, visible (Fig. 15A, C).

49 Ratio of length of petiole to width: 0, stalk hidden by syntergite; 1, short (L≦W); 2, long (L>W).

50 Antero-ventral margin of petiole: 0, weak, fused with anterior part (Fig. 15A); 1, developed, not fused with anterior part (Fig. 15B, C).

51 Setae on ventrolateral part of syntergite: 0, absent; 1, present.

52 Antero-dorsal margin of syntergite: 0, not carinate (Fig. 15D); 1, distinct carinate (Fig. 15E).

53 Multiple grooves on antero-dorsal part of syntergite: 0, absent; 1, present, but very short (almost foveae) (Fig. 15E); 2, present (Fig. 5C).

54 Ovipositor sheath (3rd valvula) visible and hardened in female: 0, less hardened (in outgroup); 1, hardened, clearly visible.

The 3rd valvula of Proctotrupidae shows specific characteristics (Masner 1993). We called this character “ovipositor sheath” following previous studies. In Proctotrupidae, ovipositor sheath is rigid, sclerotized, large, and covers ovipositor totally. We coded these characteristics as “hardened” in Char. 54. The characters of ovipositor sheath below (Char. 55–62) were coded only in Proctotrupidae. The states in outgroup were coded as “–”.

55 Length of ovipositor sheath compared to hind tibia: 0, short (0.4<); 1, normal (0.4–0.8); 2, long (<0.8).

56 Ratio of length of ovipositor sheath to width: 0, short (ovipositor sheath wide), 5.0<; 1, normal, 5.0–10.0; 2, long (ovipositor sheath narrow), <10.0.

57 Shape of ovipositor sheath in lateral view: 0, straight (Fig. 15F); 1; curved apically (Fig. 16A, B); 2, weakly and evenly curved (Fig. 16C); 3, distinctly and evenly curved (Fig. 16D).

58 Lateral surface of ovipositor sheath: 0, wrinkle absent; 1, wrinkled (Fig. 16A).

59 Setae on ovipositor sheath: 0, absent; 1, present (usually sparse).

60 Distribution of setae on ovipositor sheath: 0, setae absent; 1, present in dorsal and ventral (absent in apex); 2, present in dorsal, ventral, and apex; 3, present randomly.

61 Length of setae on ovipositor sheath: 0, setae absent; 1, setae on dorsal part almost same length of ones on ventral part; 2, setae on ventral part longer than ones on dorsal part.

62 Shape of apex of ovipositor sheath: 0, round (Fig. 16D); 1, tapered to the apex, but not pointed (Fig. 16E); 2, pointed (Fig. 16B).

63 Shape of gonostyle: 0, wide and blunt; 1, narrow and sharp (in Codrus, Fig. 16F).

64 Ratio of length of second valvifer to width: 0, long, <3.0 (Fig. 17A, B); 1, short, 3.0≦ (Figs 6, 17C, D).

Biology (for ASR)

1 Host taxa: 0, Coleoptera; 1, Diptera; 2, others.

The states of this character were based on the previous records. The reference are as follows: Nothoserphus afissae (Watanabe, 1954): Abe (2023); N. scymni (Ashmead, 1904): Abe and Seki (2021); N. mirabilis Brues, 1940: Poorani (2023); Cryptoserphus aculeator (Haliday, 1839) and C. flavipes (Provancher, 1881): Masner (1968); Mischoserphus arcuator (Stelfox, 1960): Sueyoshi and Abe, personal observation; Oxyserphus clypeatus (Ashmead, 1893): Abe (2022); Brachyserphus parvulus (Nees, 1834): Choi et al. (2012); Proctotrupes gravidator (Linnaeus, 1758), Codrus niger Panzer, 1873, Codrus nebriae (Watanabe, 1954), Parthenocodrus elongatus (Haliday, 1839), Phaneroserphus calcar (Haliday, 1839), and Phaenoserphus viator (Haliday, 1839): Townes and Townes (1981); Vanhorniidae: Timokhov et al. (2020); Heloridae: Zhang et al. (2020).

2 Habitat of host: 0, open land; 1, forest; 2, soil.

3 Parasitism; 0, solitary; 1, gregarious.

The states of this character were based on the previous records. The reference are as follows: Nothoserphus afissae: Abe (2023); N. scymni: Abe and Seki (2021); N. mirabilis: Poorani (2023); Cryptoserphus aculeator and Mischoserphus arcuator: Sueyoshi and Abe, personal observation; Oxyserphus clypeatus: Abe, personal comment; Brachyserphus parvulus: Choi et al. (2012); Proctotrupes gravidator, Codrus niger, Codrus nebriae, Parthenocodrus elongatus, and Phaenoserphus viator: Townes and Townes (1981).

The family Proctotrupidae was sister to the family Vanhorniidae in all analyses. Based on morphological characters, this relationship was supported by one synapomorphies (46.0) and two homoplastic state (char. 3.1 and 30.2) under both weights. The synapomorphy, petiole formed by synsternite 1 and not segmented (char. 46.0), is shared between Proctotrupidae, Vanhorniidae, and Pelecinidae (Masner 1993). This characteristic may be important for understanding the evolutionary pattern in Proctotrupoidea.

It was estimated that Vanhorniidae are sister to Proctotrupidae by previous phylogenetic studies of Hymenoptera based on molecular data (e.g. Heraty et al. 2011; Blaimer et al. 2023). Sharkey et al. (2012) estimated same relationships based on total evidence, but they didn’t mention the morphology between these families. Our analysis showed the same relationship based on both molecular and morphology.

The monophyly of Proctotrupidae was supported by all analyses. That has been estimated in previous studies (e.g. Heraty et al. 2011; Blaimer et al. 2023)ants, sawflies, and bees, but taxon sampling was limited (3–6 species in Proctotrupidae and one species in Vanhorniidae). In this study, we included 35 species belonging to 15 genera in Proctotrupidae and two species in Vanhorniidae. Our results strongly support the monophyly of Proctotrupidae.

Morphological analysis under both weights was supported by three synapomorphies (Figs 21, 23): horizontal groove on mesopleuron present (char. 28.1); petiole (derives from syntergite, called “stalk” in Townes and Townes (1981)) present (char. 47.1); ovipositor sheath visible and hardened in female (char. 54.1). The hardened ovipositor sheath (char. 54.1) is also shared with other genera and subfamilies in Proctotrupidae, except for the genus Heloriserphus in Heloriserphinae. (Townes and Townes 1981; He and Xu 2015; Kolyada and Mostovski 2017; Engel et al. 2022). In extinct taxa that kept the posterior part of metasoma, this trait was shown as well as in extant taxa (e.g. †Astarteserphus, †Cresogmus, and †Gurvanotrupes) (Zhang and Zhang 2001; Engel et al. 2022; Rasnitsyn et al. 2022). Several taxa in Hymenoptera have similar characteristics of ovipositor sheath to Proctotrupidae, sometimes stout, sclerotized, or elongate (e.g. Braconidae (Centistes, van Achterberg 1985) and Xyelidae (Xyela, Blank et al. (2013)) However, they are not shown in other families of Proctotrupoidea (Masner 1993; Kim et al. 2016; Abe et al. 2024). We conclude that this character is a specific character of Proctotrupidae among Proctotrupoidea and suggest that a secondary loss occurred in Heloriserphus. The stalked petiole is also thought to be specific character of Proctotrupidae. The character is shared among other subfamilies or extinct genera in Proctotrupidae (petiole in †Astarteserphus following Engel et al. (2022)) (e.g. Townes and Townes 1981; Zhang and Zhang 2001). Based on our analysis, it was estimated that the long stalked petiole was secondarily lost in Cryptoserphini (Figs 21, 23; Table S2), transferring the genus Nothoserphus to Disogmini (discussed below). In other families of Proctotrupoidea , petiole is absent in Vanhorniidae (char. 47.0) and Pelecinidae, and the first metasomal tergite and sternite form a segmented petiole (char. 46.1) in others (Masner 1993; Kim et al. 2016; Abe et al. 2024). In the filed observation, Proctotrupidae use their whole metasoma when laying eggs (Huggert 1979; Abe and Hashizume 2022) that is not shown in Vanhorniidae (Marshall 2023). This implies that the two major diagnostic characters of Proctotrupidae, the stalked petiole and ovipositor sheath, are related with the structure and mechanism of ovipositor system.

The horizontal groove (Fig. 2) is shared among most genera in Proctotrupidae, however, Heloriserphus in Heloriserphinae and Apoglypha in Proctotrupini in Proctotrupinae do not have this character (Townes and Townes 1981). In Platygastridae, a similar structure in the same position is known as the “transepisternal line” (listed in HAO: http://portal.hymao.org/projects/32/public/ontology_class/show/6872). This character is known to connect with twomuscles of mesopleuron, pl2-3ax2 (anterior mesopleuron-third axillary sclerite of fore wing) and pl2-t2b (second mesopleuro-mesonotal) (Mikó et al. 2007; Mikó et al. 2021). The horizontal groove in Proctotrupidae appears to be the same character as transepisternal line. To clarify the homology and function, dissections and comparative anatomical exminations of mesopleural muscles are required.

Morphological analysis showed that the clade (Disogmini + Nothoserphus) was supported by one synapomorphy and one homoplastic state: clypeus strongly protruding in ventral view (char. 4.1), and inner pit of pronotal shoulder present (char. 20.1). These states are the efficient diagnostic characteristics of this clade. Although they did not support this clade, two characteristics, char. 5.1 (ventral margin of clypeus curved) and 45.0 (longitudinal carina on antero-ventral part of stalk absent) were shared only by Disogmini and Nothoserphus in Proctotrupidae (Table S2). It was estimated that these states are also efficient diagnostic characters of this clade.

The clade of Disogmus was supported by two homoplastic states in the morphological analysis (Figs 21, 23): 19.0 and 22.2. Compared to the definition of Nothoserphus (Townes and Townes 1981; Lin 1987), we suggest that state 19.0, length of pronotum less than 1.2 times as width in lateral, is useful as a diagnostic character of this genus. The state 51.0, ventral part of syntergite with setae, supported the clade under equal weight. This characteristic is mentioned as a diagnostic character of Disogmus in Townes and Townes (1981), but it was shared by N. mirabilis, the type species of Nothoserphus, in our observations.

This clade of Nothoserphus was supported by three synapomorphies and seven homoplastic states (Figs 21, 23): 3.0, 7.0, 9.1, 10.0, 13.1, 14.1, 34.0, 37.0, 49.1, and 53.0. Comparing the definition and diagnosis of this genus to other genera (Townes and Townes 1981), two synapomorphies, foveae present between antennal foramina (char. 13.1), and vertex with fovea (char. 14.1), are additional useful characters defining Nothoserphus.

Although not scored as a phylogenetically valuable character, the genus Nothoserphus has some unique traits compared to other Cryptoserphini: notaulus is long (char. 26.2) except in N. boops group (short or absent in Cryptoserphini); petiole is visible in dorsal view (char. 48.1) (invisible in Cryptoserphini) (Townes and Townes 1981; Lin 1987; the matrix shown in Table S2). This genus is known as parasitoid of Coccinellidae (Abe 2023) and its habitat is in open land (bio. Char. 2.0) exceptionally (in the forest in other Cryptoserphini; Fig. S6). Based on these results, we transferred Nothoserphus from Cryptoserphini to Disogmini.

The monophyly of Cryptoserphini, excluding Nothoserphus, was strongly supported by total evidence (Figs 24, 25) and was weakly supported based on morphology (Figs 20, 22), but was not supported based on molecular data (paraphyletic, Figs 18, 19). Two genera, Oxyserphus and Brachyserphus, were placed outside of the rest of Cryptoserphini in all analyses except morphological analysis under equal weight. In the molecular analysis, Cryptoserphini except Nothoserphus was paraphyletic with strong support values. Considering these results, it is estimated that the rest of Cryptoserphini are paraphyletic and that two genera, Oxyserphus and Brachyserphus, branched earlier. We did not include some genera in the Southern Hemisphere (e.g. Apoglypha, Serphonostus and Sminthoserphus) in the analysis. However, further research is required to confirm these hypotheses.

The monophyly of the remaining Cryptoserphini was weakly supported except Tretoserphus in all analyses.

The genus Tretoserphus was not monophyletic. In morphological analysis, we included 19 characters (4.0, 7.1, 8.1, 10.1, 15.1, 17.1, 18.0, 26.1, 28.1, 30.1, 31.2, 34.2, 35.1, 39.0, 41.2, 43.0, 48.0, 53.2, 55.1/2) that were mentioned in the definition of this genus (Townes and Townes 1981). However, they were not scored as valuable states in the analysis. The species T. laricis was sister to Phoxoserphus in all analyses except under equal weight and it was supported by three weak homoplastic states (char. 55.1, 56.1, 62.0) under implied weight. It is estimated that the genus Tretoserphus is relatively close to Phoxoserphus.

The monophyly of the genus Phoxoserphus was supported by five homoplastic states: 20.1, 22.2, 24.1, 33.1, and 35.0. We suggest that state inner pit between pronotal shoulder present (char. 20.1), and epomia connected to pronotal shoulder (char. 24.1), are useful as diagnostic characters of this genus that were not mentioned in the definition by Lin (1988).

The monophyly of Cryptoserphus was supported by one synapomorphy and three homoplastic states: mesothoracic spiracle separated from postero-lateral margin of pronotum (char 25.1); metapleural carina absent (char. 35.0); length of longer hind tibial spur over 0.75 times of hind tibia (char. 39.2). Compared with the definition of this genus, we suggest that the characteristics, mesothoracic spiracle not touching postero-lateral margin of pronotum (char. 25.1) and absence of metapleural carina (char 35.0), are efficient diagnostic characters. The monophyly of Mischoserphus was supported by two homoplastic states: 21.1 and 43.1. The two genera formed a monophyletic clade under both weights. This clade was supported by one synapomorphy (char. 31.0) and two homoplastic states (char. 7.0 and 17.1). Considering these results, we conclude that the genus Mischoserphus is monophyletic and close to Cryptoserphus. The result of character mapping showed that these genera shared host taxon (bio. char. 1.1: host is Diptera) (Fig. S5). Based on the previous studies, only these two genera are known to parasitize to Diptera in Proctotrupoidea (Masner 1993; Kolyada and Chemyreva 2019). This implies that this is an important trait when discussing host shifts in Proctotrupoidea.

The monophyly of Maaserphus was supported by two synapomorphies and two homoplastic states: 15.0, 16.1, 18.2, and 23.1. The longitudinal groove on pronotum in lateral (char. 23.1) was a synapomorphy in our results. However, the Australian genus Apoglypha in Cryptoserphini also has this groove (Townes and Townes 1981) and was not included in our analysis. This character may be useful in distinguishing Maaserphus, but additional studies including those on Apoglypha are needed to clarify the definition of this genus. Compared to the definition in other genera (Townes and Townes 1981), the state margin of occipital carina with wrinkles, (char. 16.1) is also efficient diagnostic characters of Maaserphus.

The monophyly of Brachyserphus was supported by two homoplastic states: 24.1 and 25.1. Kolyada and Mostovski (2017) suggested that the absence of malar sulcus is a diagnostic character of Brachyserphus, however, all species of this genus in our analysis had a malar sulcus (char. 10.1). Following the figures in Kolyada and Mostovski (2017), another Oriental and Neotropical genus Trachyserphus has the states mesothoracic spiracle separatred from postero-lateral margin of pronotum (char. 25.1) same as Brachyserphus. A detailed morphological review of Brachyserphus and other closely related genera is required to define the genus.

The monophyly of Proctotrupini was supported by two synapomorphies and three homoplastic states in the morphological analysis (Figs 21, 23): 3.2, 19.0, 25.1, 27.0, and 64.1. The biological state 3.1, gregarious parasitoid, was shared among Codrus, Parthenocodrus, Phaneroserphus, and Proctotrupes (Table S6). Based on ASR, it was estimated as an ancestral state in Proctotrupini (Fig. S7). However, Exallonyx trifoveatus (Kieffer, 1904) which was not included in our analysis was reported as a solitary parasitoid (Williams 1932). It is estimated that gregariousness undergoes multiple changes across phylogeny.

The monophyly of Parthenocodrus was supported by seven homoplastic states (char. 0.0, 8.1, 32.0, 43.0, 44.0, 58.1 and 61.1) Figs 21, 23). Based on the molecular data, this genus was monophyletic, but the relationship between the two closely related genera, Exallonyx and Phaneroserphus, was not specified (Figs 18, 19). The genus Parthenocodrus is clearly distinguished by two characters, length of face 1.1 times or less than inner distance between eyes (char. 3.0, 3.2 in Exallonyx and Phaneroserphus), and distinct ridge between antennal foramen absent (char. 12.0, 12.1 in Exallonyx and Phaneroserphus). These characters did not support the clade as a valuable character in our analysis. A detailed examination of these genera is required to define Parthenocodrus.

The genus Phaneroserphus was paraphyletic under both weights (Figs 21, 23). This genus can be distinguished by having clear ridge between antennal foramen (char. 12.1) (Choi et al. 2016). In our analysis, this character is shared with Exallonyx. These genera can be specified only by black tarsal teeth (char. 40, discuss below). We concluded these two genera are close lineage.

The monophyly of the genus Exallonyx was supported by one synapomorphy and five homoplastic states: 11.1, 32.0, 38.2, 40.1, 55.1 and 58.1. This genus is clearly distinguished by black teeth on the fore and middle tarsal claws (char. 40.0) (Towners and Townes 1981; He and Xu 2015). This genus has the highest species diversity in Proctotrupidae (approximately 350 species, 50% of Proctotrupidae) (Kolyada and Chemyreva 2019). However, taxonomic revision is required to define this genus.

The monophyly of Codrus was supported by one synapomorphy and two homoplastic states: 11.1, 38.2, and 63.1.

The monophyly of Phaenoserphus was supported by three weak homoplastic states: 13.0, 39.0, 58.1. This genus is difficult to define and distinguish between species based on exoskeleton (Townes and Townes 1981). Therefore, a detailed examination of their morphologies is required.

The rigid and integument ovipositor sheath is a specific character of Proctotrupidae (discussed above). We coded seven characters related to the ovipositor sheath (char. 55–62). According to the analysis, five characters, char. 55, 56, 57, 58, and 61, supported the species or clade under both weights. However, all these characters were coded as homoplastic states, and were not valuable in supporting phylogenetic relationships. Character 56, ratio of length of ovipositor sheath to width, is an example of multiple substitutions. The state 56.1 supported (Cr. flavipes + Cr. longitarsis) as a homoplastic state under both weights. This state also supported the clade comprised T. laricis, Phoxoserphus, Maaserphus, and (O. clypeatus + Brachyserphus) under equal weight, and O. clypeatus and (T. laricis + Phoxoserphus) under implied weight. Another state (56.2) supported Pr. gravidator under both weights. This state also supported Maaserphus under equal weight, and the clade comprising Cryptoserphus, Maaserphus, Mischoserphus, Phoxoserphus, and Tretoserphus under implied weight. This convergence was also observed in our analysis. The state 57.1, ovipositor sheath curved apically, supported two species belonging to different tribes, Phoxoserphus iyokpe in Cryptoserphini and Phaenoserphus sp.1 in Proctotrupini under both weights. Although it was shown under equal weight, this state supported three another clades: P. gravidator, Parthenocodrus, and (O. clypeatus + Brachyserphus). Based on these results, we inferred that some characteristics of the ovipositor sheath, such as its proportion, surface sculptures, and setation, were evolutionarily labile.

When accessing a host, Proctotrupidae insert their abdomen including ovipositor sheath into the host body or habitat (Huggert 1979; Nakamura et al. 2013; Abe and Hashizume 2022; Enomoto and Abe, personal observation). Moreover, based on our analysis, some of the genera that attack the same host tend to have similar ovipositor sheath states (see the matrix in Tables S2, S6). For example, O. clypeatus and B. parvulus parasitize sap beetles (Nitidulidae) and share the following morphological states of ovipositor sheath: normal length; curved apically; setae present in dorsal and ventral parts, ventral ones longer than dorsal ones. Another example is about C. aculeator, C. flavipes, and M. arcuator. They parasitize fungus gnats (Mycetophilidae) and share the following morphological states of ovipositor sheath: clearly long; weakly and evenly curved; sparsely punctate laterally; sharp apex.

The evolution of some morphological traits is related to parasitoid hosts (Dal Pos and Sharanowski 2024), and convergence sometimes occurs based on host biology (Gauld and Mound 1982). Especially in the morphological characters related to oviposition, it is estimated that they are affected by the hosts and type of substrates (e.g. Aulacidae in Vilhelmsen and Turris (2011); Ceraphronoidea in Earnst et al. (2013); Cynipidae in Guiget et al. (2023); Ichneumonidae in Dal Pos and Sharanowski 2024; Orussidae in Vilhelmsen et al. (2001)).

In Proctotrupidae, the ovipositor sheath is implicated in the relationship between parasitic behavior and host habitat. It is possible that host-driven convergent evolution occurred among the Proctotrupidae lineages.