For the monotypic genus Stephoblemmus, S. humbertiellus is only recorded in the Yunnan Province in China. We collected eleven individuals of S. humbertiellus from two sites (Table S1). For the genus Mitius, four species had been recorded in mainland China: M. brevipennis, Mitius eryuanensis Yuan, Xie & Liu, 2021, Mitius flavipes (Chopard, 1928), Mitius minor (Shiraki, 1911). We managed to collect Mitius specimens from ten provinces in mainland China—a broad geographic range covering most of Southern China (Fig. S1). We also obtain a few samples from surrounding regions (e.g., Mitius enatus Gorochov, 1994 from Thailand and M. minor from Japan; see Table S1 for the list of specimens in this study).

The specimens were first preserved in 100% ethanol during fieldwork. In the laboratory, specimens have one leg stored at –4°C for molecular sampling, with the rest pinned and dry-preserved. All specimens studied in the article were deposited in the Museum of Flora and Fauna of Shaanxi Normal University, Xi’an, China (SNNU).

Genitalia was prepared by placing dissected genitalia into a solution of alkaline protease (0.2 g/ml, AOBOX, Beijing, China) with a water temperature of 40–50°C for 48 hours. Whole-body and head specimen photographs were obtained using a VHX-6000 Super-high magnification lens zoom 3D microscope (Keyence, Japan). The details of the ovipositor were obtained using JEC-6500 lon sputtering instrument (Hitachi, Japan) and TM3030Plus tabletop electron microscopy (Jeol, Japan). Photos of genitalia and quantitative measurements were obtained using the ToupCam digital camera and bundled software (ToupTek, Hangzhou, China). All measurements are in millimeters (mm). We use the following abbreviations of measurements: BL body length (from head to tip of abdomen); HL head length; HW head width; PL pronotum length; PW pronotum width (maximum width of pronotum); FWL tegmen (forewing) length; HFL hind femur length; OL ovipositor length.

Total genomic DNA was extracted from the leg muscle preserved in ethanol using the TIANamp Genomic DNA kit (Tiangen Biotech, Beijing, China), as directed by the manufacturer. One mitochondrial marker (cytochrome c oxidase subunit 1 [COX1]) and two nuclear markers (18S rRNA [18S], 28S rRNA [28S]) were amplified and sequenced. We used the primers in Hillis and Dixon (1991), and the primer sequences are provided in Table S2. PCR amplifications started with pre-denaturation at 94°C for 2min, followed by 40 cycles of denaturation at 94°C for 30s, annealing at 42/56/57°C (for COX1, 18S, and 28S, respectively) for 30s, elongation at 72°C for 1 kb/min, and ended with a final elongation step at 65°C for 10 min. The sequence data was then manually edited in GENEIOUS PRIME 2021.1.1 (https://www.geneious.com).

To investigate relationships between the Mitius, Stephoblemmus, and Loxoblemmus, we compiled a dataset of COX1, 18S, and 28S with nine species. For the genera Mitius and Stephoblemmus, we selected a total of eight morphologically distinct species, and for Loxoblemmus, we selected one representative species. Moreover, we selected representative species from 34 genera in the subfamily Gryllinae (Table S1). One individual per species was randomly selected from Mitius and Stephoblemmus. For Loxoblemmus, we downloaded the sequences of the three genes (COX1, 18S, and 28S) of L. campestris from GenBank (see Table S1 for accession number). For other genera in the subfamily Gryllinae, searching the GenBank database with the keywords “Gryllinae and COX1/18S/28S” identified 49, 45, and 36 Gryllinae species (including those labeled as “unknown” species) with COX1, 18S and 28S records, respectively (as of 9/11/2022). We picked one representative species for each genus. Since mitochondrial genes usually have higher mutation rates and more informative sites, priority was given to species with COX1 records. If a genus has multiple species with COX1 records, we randomly selected one species with the 18S and 28S sequences. For outgroups, only the families Mogoplistidae, Oecanthidae, Phalangopsidae, and Trigonidiidae had sequence records for the three genes. We selected one species from each subfamily in these families. In total, there are 43 species from 37 genera in Gryllidae and eight from four other families in Grylloidea as outgroups (Table S1).

We performed the maximum-likelihood (ML) phylogenetic analysis using the concatenated mitochondrial and nuclear genes in IQ-TREE v1.6.12 (Nguyen et al., 2015) with the best-fitting substitution model for each gene selected by ModelFinder (Kalyaanamoorthy et al., 2017). The nodal supports were estimated using ultrafast bootstrapping with 1,000 replicates (Hoang et al., 2018). We also conducted the Bayesian inference with the same dataset in MRBAYES 3.2.1 (Ronquist et al., 2012) and estimated the majority-rule consensus tree by Markov chain Monte Carlo (MCMC) sampling. The concatenated sequences were first partitioned into genes, and three codon positions in COXI (i.e., five partitions in total), and PARTITIONFINDER v2.1.1 (Lanfear et al., 2017) identified the best substitution models were selected according to the Bayesian information criterion (BIC). Samples were drawn every 1,000 steps over 10 million steps, with the initial 25% discarded as burn-in. We used FIGTREE v1.4.4 (Rambaut, 2021) to visualize and manipulate the phylogenies.

BEAST v1.10.4 (Suchard et al., 2018) was used for molecular clock dating with an uncorrelated log-normal relaxed clock model and a Yule tree prior. Three fossils were used for age constraints. The fossil of Protogryllus grandis Zeuner, 1937, considered the oldest fossil of Grylloidea (Sharovm, 1971; Zeuner, 1937), was used to assign an exponential prior for the stem Grylloidea with a minimum bound of 199.00 million years ago (Ma; BEAUti setting: Mean = 16, Offset = 200.15). The fossil of Pseudarachnocephalus dominicanus Gorochov, 2010, the oldest fossil of Mogoplistinae (Gorochov, 2010), was used to specify an exponential prior for the stem Mogoplistinae with a minimum bound of 98.17 Ma (BEAUti setting: Mean = 26.8, Offset = 100). Lastly, an exponential prior for the stem Gryllinae with a minimum bound of 56.00 Ma (BEAUti setting: Mean = 11.4, Offset = 55.8) was set according to the age of the fossil Gryllus vociferans Cockerell, 1925, the oldest fossil of Gryllinae.

Substitution models were selected under PARTITIONFINDER 2.1.1 (Lanfear et al., 2017) with the ‘beast’ set of models. For the Markov Chain Monte Carlo (MCMC) run, samples were drawn every 10,000 steps over a total of 50 million steps, with the first 25% of samples discarded as burn-in. The BEAGLE library was used to improve and speed up the likelihood calculation (Ayres et al., 2012). We used TRACER v1.7.1 (Rambaut et al., 2018) to ensure that all ESS values were above the recommended threshold (200). The time-scaled tree was generated using TreeAnnotator in BEAST and visualized in FIGTREE v1.4.4 (Rambaut, 2021).

For species delimitation analysis, our dataset includes 38 specimens of Mitius and Stephoblemmus (33 specimens from our collection, for which we have COXI, 18S, and 28S sequenced, and five COXI sequences from the Genebank (Table S1)). Acheta domesticus was included as the outgroup.

We first applied four single-locus species delimitation approaches with the COX1 sequences to discover the putative species boundaries in Mitius and Stephoblemmus: automatic barcode gap discovery (ABGD; Puillandre et al., 2012), generalized mixed Yule-coalescent (GMYC; Fujisawa & Barraclough, 2013), Bayesian Poisson tree process model (bPTP; Zhang et al., 2013) and Refined Single Linkage (RESL; Ratnasingham & Hebert, 2013). Then, we added the other two nuclear genes and used a multi-locus species delimitation method to validate these putative species delimitation hypotheses: Bayesian Phylogenetics and Phylogeography (BPP; Yang, 2015).

ABGD is a genetic distance-based method that does not require prior species assignment (Puillandre et al., 2012). We performed ABGD analyses online (https://bioinfo.mnhn.fr/abi/public/abgd) based on Kimura’s 2-parameter model (K2P; Kimura, 1980), with a minimum intraspecific variability of Pmin = 0.001, maximum intraspecific variability of Pmax = 0.1 and a minimum gap width of X = 1.5. RESL analysis was performed on the BOLD (Barcode of Life Data) platform, which can group COX1 barcode sequences into clusters. We used this platform to assign our COXI sequences to MOTUs (molecular operational taxonomic units). The bPTP analysis (Zhang et al., 2013) was conducted on the bPTP website (https://species.h-its.org/ptp). We estimated an ML tree with the COXI dataset using IQtree and used it as the input after pruning the outgroup. To check for convergence, we ran two independent MCMC analyses. Each run is 1000,000 steps in total and was sampled every 10,000 steps, and the first 10% was discarded as burn-in. For GMYC, another tree-based species delimitation method, we conducted the analysis using the single-threshold model in the “splits” package (Ezard et al., 2009) in R v3.6.3 (R Core R et al., 2020). We inferred the ultrametric guide tree using BEAST v1.10.4 (Suchard et al., 2018) with a strict clock model and a birth-death tree prior, a suitable tree prior for data sets with a mixture of intra- and inter-species sampling (Ritchie et al., 2017). For calibration, we used our molecular dating results: a normal prior (mean 11.16, standard deviation 6.40) for the MRCA (most recent common ancestor) of clade Mitius/Stephoblemmus; a normal prior (mean 13.47, standard deviation 7.11) for the MRCA of Mitius/Stephoblemmus and A. domesticus. The posterior distribution was estimated by MCMC sampling, with samples drawn every 1,000 steps over a total of 10 million steps, discarding the first 10% as burn-in. We used TRACER v1.7.1 (Rambaut et al., 2018) to ensure that all ESS values were above the recommended threshold (200). The time-scaled tree was generated using TreeAnnotator in BEAST.

For multi-locus species delimitation, we conducted unguided Bayesian species delimitation in BPP v4.3.8 (Yang, 2015) to explore different species-delimitation models while allowing changes in species-tree topology (A11 analysis; (Yang & Rannala, 2014)). For the priors of population sizes (θ) and divergence times (τ), we used inverse-Gamma priors with α = 3. The β parameter was adjusted according to the mean estimate of nucleotide diversity for θ (β = 0.01) and node height for τ (β = 0.001). We performed the analyses using the reversible-jump MCMC algorithm 0 (ε = 2) and algorithm 1 (α = 2, m = 1), with duplicated runs to check for convergence. Analyses were run for 20,000 MCMC steps, with samples drawn every five steps and the first 20% discarded as burn-in.

We used the concatenated COXI, 18S, and 28S sequences, with a total of 2492 sites. Our ML and BI tree-estimation analyses obtained similar tree topologies (Fig. S2 and S3). Both results showed Mitius and Stephoblemmus form a monophyletic clade with the highest support (PP of 1 and BS of 100), and the species of Stephoblemmus is nested within the genus Mitius, suggesting that Mitius and Stephoblemmus are synonyms. Moreover, although the tree topology inferred using the Bayesian relaxed-clock analysis (Fig. 2) is slightly different from that inferred using maximum likelihood and Bayesian analyses (Fig. S2; Fig. S3), all analyses supported the monophyly of the genus Mitius and Stephoblemmus, and have consistent internal species relationships. Our divergence dating result placed the stem age of the genus Stephoblemmus (= Mitius) in the Tortonian at 9.18 Ma with a 95% credibility interval (95% CI) of 15.21 to 4.08 Ma, the stem age of S. enatus at 3.5 Ma with 95% CI of 8.97 to 1.79 Ma, and the stem age of S. minor and S. humbertiellus at 1.81 Ma with 95% CI of 4.38 to 1.45 Ma.

Different species delimitation methods yielded similar results (Fig. 3). Single-locus methods, ABGD, bPTP, and RESL, identified three MOTUs (molecular operational taxonomic units; Fig. 3). One comprised of all S. minor individuals (yellow clade on Fig. 3). One includes both S. humbertiellus and M. brevipennis individuals (green clade on Fig. 3), suggesting that not only the two genera should be combined, the two species are also synonyms. The last MOTU has individuals of M. flavipes and M. enatus. The result from GMYC was slightly different—it recognized one lineage of S. minor as an additional MOTU. The multi-locus approach BPP supported the three-MOTU hypothesis with a high posterior probability of 0.90 (Fig. 3).

Mitius was established by Gorochov in 1985 and diagnostic features were their smaller body and genitalia shape. The type species of Mitius was Gryllus flavipes (= Mitius flavipes) Chopard, 1928 which possess rounded frons. Later, based on genitalia features, some species with special head shapes were also moved into the genus. Loxoblemmus castaneus Chopard, 1937 (= S. castaneus) is very similar to M. blennus (= S. blennus), except for the truncated frons in males, and was moved into the Mitius (= Stephoblemmus) by Gorochov (1996) for the genitalia features. For the same reason, Loxoblemmus vaturu Otte & Cowper, 2007 (= S. vaturu) and Mitius brevipennisYuan et al., 2022 (= S. brevipennis) were also separately assigned to the genus (Qiao et al., 2020; Yuan et al., 2022). Stephoblemmus was established by Saussure (1877) and their difference is more exaggeratedly truncated frons and forms a lamellar extension at the ends. Because of the peculiar head shape and the very small number of species (only the monotypy has been found since its discovery), no one has ever associated Stephoblemmus with Mitius. At most, their relationship with Loxoblemmus has been considered (Chopard, 1933; Qiao et al., 2020).

In this study, we found that the male external genitalia of genera Stephoblemmus and Mitius belong to a similar type while being distinct from Loxoblemmus. Stephoblemmus and Mitius both possess a unique and distinct epiphallic median lobe (Fig. 6) but for Loxoblemmus, its epiphallus with a pair of median lobes. In addition, their forewing patterns reflect similar species associations. We found that the apical fields of tegmina of species of both genera are shorter than half of the basal field, and that of Loxoblemmus is longer than half of the basal field (Fig. 5E–G). In the above, based on morphological characters, we argue that the genus Stephoblemmus and Mitius are extremely closely related.

To date, there are ten species worldwide (Cigliano et al., 2023) According to the shape of the frons, all the species could be divided into two groups. However, there are still many problems with the classification of these species. For example, there may be intraspecific species morphological variation (e.g., coloration, frons, hind wings, body size, and male genitalia, etc.) in species M. brevipennis (= S. brevipennis), M. enatus (= S. enatus), M. minor (= S. minor), etc.; and there may be problems with the species relationship in species of M. enatus (= S. enatus), M. eryuanensis, M. minor (= S. minor), M. blennus (= S. blennus), and M. splendens, etc.

Here, we used five methods to delimit Mitius (= Stephoblemmus) species and found that those species with truncated frons (S. humbertiellus) are mixed up with those species with spherical frons. Thus, although the type of frons can identify Mitius (= Stephoblemmus) species, it cannot naturally classify them. In this study, we collected Mitius (= Stephoblemmus) specimens from ten provinces, a broad geographic range covering most of Southern China (Fig. S1). We also obtain a few samples from surrounding regions (e.g., M. enatus from Thailand and M. minor from Japan; see Table S1 for the list of specimens in this study). Although the methods used to distinguish species differed, the results of all methods except GMYC were the same and identified three different MOTUs. The GMYC methods define species based on an ultrametric guide tree (Reid & Carstens, 2012). Thus, GMYC shows S. minor from Hainan is different from S. minor elsewhere on MOTU, this may be due to the large gap in divergence times. Accordingly, Stephoblemmus includes eight species, and four of them can be found in China.

All models consider M. enatus (= S. enatuscomb. n.), M. eryuanensis, and M. flavipes in China as MOTU, and morphological identification also supports this result. Yin & Liu (1995) identified the species from Yunnan as M. flavipes and noted differences with S. enatuscomb. n. in the color of posterior tibiae and the first section of posterior tarsus. However, we observed about twenty specimens of M. enatus (= S. enatus) collected from Yunnan, Guangxi, Hainan, and Hong Kong, and found the coloration of the hind leg is varied, and the hind tibiae and tarsus might color yellow (Fig. 7C) or dark brown (Fig. 7D). Therefore, Chinese M. flavipes recorded in Yunnan Province should be a misidentification. Yuan et al. (2021) distinguished M. eryuanensis and S. enatuscomb. n. through the presence or absence of bifurcation of epiphallic lateral lobes and hind wings. We found hind wings are present in some specimens or absent in others (indeed, hind wings always fall off in cricket adults) (Fig. 7A–B), and the bifurcate situation of the epiphallic lateral lobe is dependent, and the angle between the apex of lobe and its dorsal side would determine whether we observe the branches. When this angle is right or acute, we could find the epiphallic lateral lobe bifurcated; if it is obtuse, we could not observe it (Fig. 7H–N). M. eryuanensis should be a junior synonym of S. enatuscomb. n.

Besides, the results of all methods except GMYC consider S. minorcomb. n., M. splendens and M. blennus as a MOTU. In terms of morphology, Yang and Yang (1995) recorded M. blennus and M. splendens in Taiwan and specified that they were slightly different in body size and male genitalia. But the diversity of S. minorcomb. n. has been observed and it includes the situations of body size (the minimum body length is 9.25 mm, the maximum is 11.35 mm, and the average is 10.52 mm) and male genitalia (in dorsal view, epiphallic lateral lobe has two types of apical margin: truncated or arc-like margin (Fig. 8H–J); in ventral view, epiphallic lateral lobe also possesses two types of protrusion: only one truncate projection or two small protrusions (Fig. 8K–N)) in M. splendens (body length: 10.64–12.37 mm; Fig. 8C, G and I) and M. blennus (body length: 9.46–9.78 mm; Fig. 8E, H and J). Accordingly, we conclude that S. minorcomb. n. and M. splendens should be one species, and M. blennus in Taiwan should be a misidentification.

In addition, M. brevipennis and S. humbertiellus both with truncated frons. We compared the holotype of M. brevipennis with S. humbertiellus in the same location, the genitalia are very similar (Fig. 6A–F) but the male frons is more prominent in S. humbertiellus (Fig. 5A–D). Furthermore, all of the barcoding analysis methods that supported M. brevipennis and S. humbertiellus were combined into one MOTU. Therefore, we think M. brevipennis and S. humbertiellus should be one species.

The study reveals that the evolution of head shape in field crickets has not only undergone a long and complex historical process but also exhibits a complicated and variable diversity (Fig. S5). The frons of both Stephoblemmus and Loxoblemmus crickets appear to be specialized, but their basic structure is not similar, and they are not related in morphological changes. Stephoblemmus has a predominantly morphological variation on the dorsal side of the frons, with clypeus and labrum normal as in common crickets, and almost no very specific flattening; its compound eyes are normal, with most of the two compound eyes visible in frontal view; the antennal sockets are also normal, and not only are their relative positions similar to those of common crickets, but the size of the antennal sockets is also similar to those of common crickets. In contrast, the entire face of Loxoblemmus is flattened, with the frontal, clypeus, and labrum all flattened; meanwhile, the compound eyes are located on the dorsal side of the face, and only a few of them are visible in the frontal view; the antennal sockets are smaller and located on or near the dorsal side of the face. Therefore, the special head shape of Stephoblemmus should be the result of independent evolution. In terms of divergence time (Fig. 2), the clade in which these two genera are located diverged in 9.18 Ma, and within both clades, the taxa adjacent to these two genera are also dominated by common hemispherical heads.

The special head shape of Stephoblemmus is not only the result of independent evolution but has also undergone several independent evolutions within the genus. Although we do not have the molecular data for the S. castaneuscomb. n. and S. vaturucomb. n., but based on geographic distribution (S. castaneuscomb. n. distributed in Malesia and S. vaturucomb. n. distributed in Fiji), it is likely that truncated frons have evolved independently multiple times within the genus Stephoblemmus. In addition, the particular head features of Stephoblemmus may not be stable within the species. Not only are there individual differences in head characteristics, but there is also male and female dimorphism. Within the species of S. humbertiellus, the males’ frons are truncate, and in most individuals, the apex of the frons expand dorsally in a lamellar form; however, in a few individuals the frons are only slightly convex dorsally, as in Loxoblemmus species. Sexual dimorphism of frons in S. humbertiellus is also observed. The males of this species present with frons specialization as described previously, while the females have a normal hemispherical head with no shape specialization. Therefore guess that the presence of the trait might be related to sexual selection in this species.

Sexual selection is responsible for the evolution of various sexually dimorphic traits such as elaborate weapons, ornaments, and behaviors (Andersson, 1994). In cricket studies on sexual dimorphism, the size, morphology and pigment of a male’s head may all be associated with positive selection for more aggressive behavior (Sean et al., 2008). In studies of Loxoblemmus species, it has also been speculated that this particular head style is used as a weapon for male competition, but studies have been unable to confirm the involvement of this feature in male competition (Kuriwada, 2016). Although the specific function of the head trait is unknown, this trait also shows polymorphism within species. Compared to dimorphic traits, polymorphic traits may reflect more context of the occurrence of traits, such as the presence of transition states or the presence of diversity. However, there are few discussions on polymorphic traits in field crickets, and this study provides new material and information for related discussions.