Mitochondrial Phylogenomics Suggests Complex Evolutionary Pattern of Pronotal Foliaceous Mimicry in Hierodulinae (Mantodea: Mantidae), with Description of a New Species of Rhombodera Burmeister, 1838 from China

Simple Summary Rampant morphological homoplasy in Mantodea has confused systematists for a long time. The leaf-like pronotum was long thought to be one of the key characteristics of Rhombodera in Hierodulinae. In this study, we describe a new species of Rhombodera and discuss the phylogeny as well as complex convergent evolution of leaf-like pronotum in Hierodulinae based on the combination of morphological and mitogenome data. The results reveal the unnatural grouping of the current classification in Hierodula and Rhombodera, rendering the well-developed lateral expansion invalid as the diagnosis of Rhombodera. Mapping the pronotal-foliaceous trait onto our phylogenetic tree show that the origin in multiple lineages of this trait in Hierodulinae may have adaptive significance. The congruent results of the group division between mitochondrial phylogenomics and morphology of male genitalia also highlight the capacity of the use of mitogenomes in resolving the derived mantodean phylogenetic relationship. Abstract Hierodulinae is a species-rich mantid subfamily, with some species bearing a notable leaf-like pronotum. However, the evolutionary pattern and taxonomic significance of the leaf-like pronotum are largely unknown. Here, we present a phylogenomic analysis of the Hierodulinae genera Rhombodera Burmeister, 1838, and Hierodula Burmeister, 1838 based on mitochondrial genomes. We also describe a new species, namely Rhombodera hyalina sp. nov. from Guangxi, China. Our phylogenetic result, together with the evidence from male genitalia, suggests the division of the Oriental Hierodula and Rhombodera complex into three clades. We find a complex pattern on the evolution of the leaf-like pronotum, which is present in at least five lineages, respectively, of the above three clades.


Introduction
Hierodulinae Brunner de Wattenwyl, 1893 is the largest subfamily of Mantidae Latreille, 1802, containing 25 genera and more than 200 species, which widely occur in the Old World [1]. The adults of Hierodulinae show a typical praying mantis body shape with medium to large size, but some species possess a foliate pronotum that is related to leaf-mimicry. The specimens were collected in both daytime and night during 1982-2021. Nymphal individuals at the time of collection were reared for obtaining the adult and ootheca specimens within an RDN intelligent climatic chamber with the environmental parameters following settings: temperature (25 • C); relative humidity was (70 ± 5)% and photoperiods were L/D = 18/6 h with light 9000 lx. Adult individuals were euthanized with ethyl acetate and stored temporarily in 75% alcohol for examination. Male terminalia was prepared by macerating the apical portion of the abdomen in hot 20% KOH for 10 min to isolate the genital structures. Detailed external features and male genitalia were examined based on dry pinned specimens under an OPTEC (Chongqing, China) stereoscope and captured the image information using the corresponding OPT pro application. Other photographs were taken using a Canon EOS700D digital camera. All photographs were processed using Adobe Photoshop CC2017. Morphological terminology and abbreviation follow [10,12,22]. Measurements are in mm. All of the samples were preserved at the Entomological Museum of China Agricultural University (CAU).

DNA Extraction and Mitogenome Sequencing
Total genomic DNA was extracted from the thoracic and coxal muscle tissue of a single specimen using a DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). Sixteen Illumina TruSeq libraries with 450 bp average insert size were prepared and sequenced using the Illumina Hiseq2500 platform with 250 bp paired-end reads. Raw reads were trimmed of adapters and then the low-quality and short reads were removed with PRINSEQ [23]. Highquality reads were used for de novo assembly using IDBA-UD [24] with minimum and maximum k-values of 60 and 200 bp, respectively. Gene sequences of mitogenomes were annotated using Mito Z [25], and further corrected in Geneious v.10.1.3 [26] by aligning the sequences with those of homologous genes of other Mantodea species.

Phylogenetic Analysis
Each PCG sequence of mitochondrial DNA was aligned based by using codon-based multiple alignments under the MAFFT algorithm [27] on TranslatorX [28] online platform with the L-INS-i strategy. Sequences from two rRNA genes were separately aligned using the MAFFT v.7.0 online server with G-INS-i strategy [29]. MEGA v.6.0 [30] was used to check, concatenate, and output these alignments. Three datasets were finally outputted: (i) all 13PCGs and two rRNA genes (PCGRNA; 12754 bp); (ii) all 13 PCGs with third codon positions removed and two rRNA genes (PCG12RNA; 9171 bp), and all 13 PCGs translated to protein (AA; 3583 amino acids). The dataset PCGRNA failed to recover the monophyly of Mantidae under the best-fit GTR+F+I+G4 model ( Figure S6), thus was not included in the further analyses.
Two partitioning schemes (gene-partition and non-partition) were used for AA and PCG12RNA datasets to construct the phylogenetic trees under maximum likelihood (ML) methods and Bayesian inference (BI). For ML analyses, the optimal partitioning scheme and best substitution model for each partition were selected under the ModelFinder [31] in IQ-TREE 1.6.12 [32]. The "greedy" algorithm with branch length estimated as "unlinked" was set and the AIC, AICc, and BIC criteria were chosen. ML phylogenetic trees were constructed using the IQ-TREE webserver by an ultrafast bootstrap approximation approach with 1000 replicates with the best model estimated. Bayesian analyses were carried out using MrBayes v.3.2.77 [33]. Two simultaneous runs were conducted for the datasets, and trees were sampled every 1000 generations with the first 25% discarded as burn-in. The average standard deviation of split frequencies was thought to be stationary until below 0.01. Remarks. Rhombodera is currently known as non-monophyletic [10,11]. Our phylogeny shows strong signal clustering of H. latipennis, H. jianfenglingensis, H. chinensis, R. kirbyi, Rhombodera stalii Giglio-Tos, 1912, Rhombodera latipronotum Zhang, 1990, Rhombodera valida Burmeister, 1838, Rhombodera zhangi (Wang and Dong, 1993), Rhombodera megaera, Rhombodera hyalina sp. nov., and R. longa in one clade, recovering Rhomobodera as paraphyletic with respect to Hierodula. Species in this clade also share the similarity in robust pronotum of medium length with prozone 2.2-3.5 times as long as metazone, well-developed secondary distal process (sdp) in ventral phallomere not beyond the width of the main lobe of sclerite L4A, large afa in left phallomere mostly bilobed as two apical sharp lobes and elongate nearly parallel ( Figure 1G-R). These characters greatly meet the typical Rhombodera species rather than Hierodula species, indicating the problematic taxonomic position of H. latipennis, H. jianfenglingensis, and H. chinensis currently. A large number of Southeast Asia lineages without strongly dilated pronotum of Hierodula are also facing such a situation. Thorough revision and redefinition of the genera in Hierodula + Rhombodera complex are urgently needed to solve the above problem.
Rhombodera hyalina sp. nov. Type material. Holotype (Figure 2A,B and Figure 3A Diagnosis. This new species is very similar to R. latipronotum and R. valida ( Figure S1) but could be recognized from the latter two by the following features: (1) margins of the pronotum of the female slightly rippled dorsally and darkly pigmented when alive while the latter two are completely smooth and nearly unicolor dorsally ( Figures 2C and 4A).
(2) Forewing discoidal area completely hyaline in males. (3) Narrow pterostigma of both sexes. (4) Slender and long sdp and the possession of median secondary distal process (sdpm) in the ventral phallomere ( Figure 3E). (5) The mature nymphs of the new species show a ventrally upturned abdomen ( Figure 5D), while it is outstretched in R. latipronotum ( Figure 5E) and R. valida. The new species is also similar to Indochinese Rhombodera species R. megaera and R. zhangi but could be immediately distinguished by more dilated pronotum: the ratio length/width are about 1.4, 2.7, and 2.4 times of R. hyalina sp. nov., R. megaera, and R. zhangi, respectively.        Diagnosis. This new species is very similar to R. latipronotum and R. valida ( Figure  S1) but could be recognized from the latter two by the following features: (1) margins of the pronotum of the female slightly rippled dorsally and darkly pigmented when alive while the latter two are completely smooth and nearly unicolor dorsally ( Figures     Description. Measurements. See Table 1. Male (Figure 2A,B, Figure 3A,B,E and Figure  Left phallomere: Sclerite L4B spoon-shaped. Sclerite L2 long, fusiform in shape, slowly narrowing towards the distal end, and suddenly contracting near the tip. Apical process paa long, strongly curved near its end, and forms a hook-like apex. Phalloid apophysis of the sclerite L1 well developed, with the lobe of the anterior one (aafa) short pencil-tip shaped, while the posterior one (pafa) long and robust. Both the two lobes of phalloid apophysis have a sharp tip. Ventral phallomere: broad, roughly square-shaped. Left edge of sclerite L4A weakly sclerotized. Primary distal process (pda) membraneous, rounded, and prominent. Two distal processes of ventral phallomere: lateral secondary distal process (sdpl) much sclerotized, long sicklelike, its length approximately equals to the width of sclerite L4A; sdpm less sclerotized, much slender, nearly 1/3 length of sdpl. Right phallomere: main posterior lobe nearly triangular, ridge-like projection (pva) and tooth-like projection (pia) smoothly connect in the shape of claws, upward-curved.
Female ( Figure 2C,D and Figure 3C,D). Large-sized, light green. Head. Triangular, about 1.15 times as long as wide. Vertex convex, with three small ocelli in a triangular formation. Lower frons width nearly equal to height, pentagonal with arched superior margins. Pronotum very robust, roughly regular rhomboidal in shape, 1.3 times as long as wide. Anterior and posterior margins of pronotum on the same side form an angle of about 92 • . Postcervical plate transversal saddle-shaped. Margins of pronotum slightly uprolled dorsally and darkly pigmented when alive. Metazone about 2.21 times as long as prozone. Forelegs. Similar to male but much more robust, with 6-8 dorsal spines. Coxal lobes round, nearly symmetric. Trochanter unicolor in ventral view. Femoral brush ellipse-shaped, starting from 12th and ending just a little distally of 15th AvS. Spinal formula arranging as F = 4DS/15AvS/4PvS; T = 14AvS/11PvS. Wings. Well developed. Forewing leaf-like, green in costal area, partly hyaline in discoidal area, and gradually greening to anterior margin, about 2.5 times as long as wide. Pterostigma grayish white, very tiny and narrow. Hindwing hyaline, triangular, length 1.75 times width. Abdomen elongated oval-form, much more robust than male, uniformly green.
Ootheca ( Figure 5F). Barrel-like, with no extension at the residual process. Generally russet brown in color with thick beige external coating outside the external wall. The proximal end of the ootheca encircles the substrate completely, resulting in the ventral surface being perpendicular to the substrate. Flexible flaps in the emergence area reduced.
Etymology. The name refers to the completely hyaline discoidal area in the forewing of the male adults of the new species.
Biology. This species has one generation per year in Guangxi. They overwinter as a diapause stage (antepenultimate instar) in November to February, mature in April to early May, and lay oothecae in early June. Some nymphs also have dark markings on their forefemur which disappear after their final molt.

Phylogenetic Analysis
All topologies under ML and BI methods by PCG12RNA and AA datasets with different partition schemes successfully recovered the sister group of Mantidae and Hymenopoididae with high nodal support (BP ≥ 75%, BPP ≥ 90%) ( Figures S2-S5, S7-S10). In Mantidae, the four subfamilies included in our analysis, i.e., Mantinae, Tenoderinae, Vatinae, and Hierodulinae were respectively, recovered as a monophyletic clade, while their relationships were slightly different among some results from different datasets: the relationship (Mantinae + (Tenoderinae + (Hierodulinae + Vatinae))) was recovered only under our BI analysis of PCG12RNA-gene partition ( Figure S9). In contrast, the alternative hypothesis (Mantinae + (Hierodulinae + (Vatinae + Tenoderinae))) was supported in our AA datasets by BI and ML analyses as well as PCG12RNA-no partition by ML.
In Hierodulinae, our analyses support the monotypic genus Mekongomantis, with most of the results (with the exception of PCG12RNA-no partition by BI and AA-gene partition by ML placing the genus at the base of Rhomobodera) placing the genus as the sister group of Hierodula + Rhombodera complex with high nodal support (BP = BPP = 100%), which is consistent with Shi and Yuan and Shi et al. [20,34] The rest of the species were divided into three main lineages: membranacea group (H. membranacea and H. confusa), patellifera group (Rhombomantis fusca, R. longipennis, and H. patellifera), and Rhombodera group (R. kirbyi, R. stalii, R. latipronotum, R. valida, R. zhangi, R. megaera, H. latipennis, H. chinensis, H. jianfenglingensis, R. longa, and R. hyalina sp. nov.). Rhombodera group was recovered to be sister group to membranacea group + patellifera group by AA-no partition (BPP = 1), PCG12RNA-gene partition (BPP = 92%) under BI as well as AA-no partition (BP = 42%) and PCGRNA-no partition (BP = 61%) under ML. The different topology, i.e., membranacea group + (patellifera group + Rhombodera group), was recovered by PCG12RNA-gene/no partition, but with low nodal support (BP = 57%, 67%). The genus Rhombomantis was not recovered as a monophyletic group in AA analysis under ML and BI as well as in PCG12RNA-no partition under BI, and the genus Rhombodera was recovered as paraphyletic with respect to Hierodula in all analyses.

General Features of Newly Sequenced Mitogenomes
A total of sixteen complete or near-complete mitogenomes were assembled in this study. The length of them ranged from 17298 bp in R. valida to 15391 bp in H. latipennis table ( Table 2). All of the sequences share the identical gene arrangement to the putative ancestral insect. The mitogenomes of H. chinensis, H. confusa, H. membranacea, H. jianfenglingensis, and R. longipennis contained 13 protein-coding genes (PCGs), 22 transfer RNA genes (tRNAs), two ribosomal RNA genes (rRNAs), and a control region. Pseudogene trnM* was found at the 3 end of a large noncoding region (LNCs) between trnM and nad2 in R. longa, R. valida, R. kirbyi, R. latipronotum, R. hyalina sp. nov., and R. Stalli, which was also reported in seven other Hierodulinae species [35]. For those nearly complete mitogenomes, the trnI, trnQ, and trnM were not found in H. latipennis, and the trnI and trnQ were not found in R. zhangi, and R. megaera. A/T biased of mitogenomic nucleotide composition was also found in all sixteen newly sequenced Hierodulinae mitogenomes, ranging from 76.20% in R. kirbyi to 74.24% in R. valida from Java. In addition, the mitogenomes had an overall positive AT-skew (0.04 to 0.06) and negative GC-skew (−0.24 to −0.19) ( Table 2). Table 2. Base composition of sixteen Hierodulinae mitogenomes.

Taxonomic Implication for the Future Revision
The phylogenetic trees ( Figure 6 and Figures S2-S10) clearly divided the "Hierodula + Rhombodera" complex into three main clades: the clade "membranacea group", which includes type species of Hierodula, H. membranacea. This clade shares similarities of the slender pronotum, more than ten fine spinules arranging along the dorsal coxa of the forelegs, the punctiform pterostigma, the intersegmental coloration of the ventral abdomen, the ear-like phalloid apophysis (afa), and the extremely elongated sdpl as well as its base ( Figure 1B,C). Rhombodera extensicollis is also inferred to be part of this group because of the presence of the above characters [12,36], although still lacking molecular evidence currently. The "patellifera group" is represented by the species related to H. patellifera and the genus Rhombomantis. Similarities of this group are the relatively stubby pronotum, the simplified and not bilobed afa, and the long spear-shaped sdpl with its length slight beyond the width of the main lobe of the sclerite L4A ( Figure 1B-F). The clade with H. latipennis at the base (Rhombodera group) is a highly diverse group, including species both with and without lateral pronotal expansion. The similarities of this clade include the medium-length pronotum, the well-developed secondary distal processes, and the distinctly bilobed afa (anterior lobe, aafa, and posterior lobe, pafa) ( Figure 1G-R). At least one of the two lobes in Rhombodera group possesses a sharp apex; there is also a parallel tendency in the directions of the elongating of the two lobes in this group (vs. deeply forked afa in Tenoderinae and Vatinae). Combined with the above morphological and molecular evidence, the presence of well-developed lateral expansion should not be homologous in many species, and at least some species with weak lateral expansion might be very closely related to Rhombodera, rendering the well-developed lateral expansion invalid as the diagnosis of Rhombodera.
Two treatments could be applied to resolve the unnatural grouping of Rhombodera. One of them could erect new taxa or remove described taxa from synonyms to accommodate several subclades in the Rhombodera group, while male genitalia will be the main diagnosis of these taxa. The alternative treatment could be the rearrangement of the hieroduline species between existed genera in precedence. It may require a large number of species to be transferred from Hierodula to Rhombodera as well as expanding the definition of Rhombodera, and finally, Rhombodera will at least include the Rhombodera valida group and Hierodula chinensis group from the oriental realm. Our results are here inclined to support the second one in dealing with the unnatural grouping in the complex. Such treatment provides higher stability and fault tolerance, and the diagnosis of the three main clade of this complex will be quite simple, though the final results of this way may cause a great increase in the species number of Rhombodera. The further division of Rhombodera into several genera would be over-reliant on male genitalia because almost no other stable morphological characters could be concluded, which might reduce the practicality at the genus level. It is also worth noting that some synonymies of hieroduline taxa were established based on problematic species itself, and recovering the validity of these taxa is also an urgent problem.
Our study did not include the Australasian and Papua lineages such as Tamolanica and the species related to Hierodula majuscula of the Hierodula+Rhombodera complex. However, we conclude that it will not influence the validity of the above discussion because the ancestral position of these lineages [35] and distinct male genitalia [10] give this lineage relative independence compared to the oriental lineages.

Evolutionary Significance of the New Species
Our findings indicate that Rhombodera lineage also has species with the narrow pronotum, and the new species presents foliaceous pronotum again in a sub-clade without such character. Combining this new finding with our phylogeny, the pronotal foliaceous mimicry was observed in at least five lineages in oriental hieroduline species (Figure 6), suggesting a complex evolutionary pattern of pronotal morphology. The repeating appearance of leaf-like pronotum in multiple clades justifies the particularity and adaptive significance of this phenomenon in Mantodea.
x FOR PEER REVIEW 14 of 17    Table S1. Taxa used in present study from GenBank. Table S2. Collection information of the ingroups species sequenced in the present study [37][38][39][40]. Figure S1.  Figure S2. Topology of the ML analysis from AA-no partition dataset. Figure S3. Topology of the ML analysis from AA-gene partition dataset. Figure S4. Topology of the ML analysis from PCG12RNA-gene partition dataset. Figure S5. Topology of the ML analysis from PCG12RNA-no partition dataset. Figure S6. Topology of the ML analysis from PCGRNA-no partition dataset. Figure S7. Topology of the BI analysis from AA-no partition dataset. Figure S8. Topology of the BI analysis from AA-gene partition dataset. Figure S9. Topology of the BI analysis from PCG12RNA-gene partition dataset. Figure S10. Topology of the BI analysis from PCG12RNA-no partition dataset.