Phylogeny and species delimitation of the genus Longgenacris and Fruhstorferiola viridifemorata species group (Orthoptera: Acrididae: Melanoplinae) based on molecular evidence

Phylogenetic positions of the genus Longgenacris and one of its members, i.e. L. rufiantennus are controversial. The species boundaries within both of L. rufiantennus+Fruhstorferiola tonkinensis and F. viridifemorata species groups are unclear. In this study, we explored the phylogenetic positions of the genus Longgenacris and the species L. rufiantennus and the relationships among F. viridifemorata group based on the 658-base fragment of the mitochondrial gene cytochrome c oxidase subunit I (COI) barcode and the complete sequences of the internal transcribed spacer regions (ITS1 and ITS2) of the nuclear ribosomal DNA. The phylogenies were reconstructed in maximum likelihood framework using IQ-TREE. K2P distances were used to assess the overlap range between intraspecific variation and interspecific divergence. Phylogenetic species concept and NJ tree, K2P distance, the statistical parsimony network as well as the generalized mixed Yule coalescent model (GMYC) were employed to delimitate the species boundaries in L. rufiantennus+F. tonkinensis and F. viridifemorata species groups. The results demonstrated that the genus Longgenacris should be placed in the subfamily Melanoplinae but not Catantopinae, and L. rufiantennus should be a member of the genus Fruhstorferiola but not Longgenacris. Species boundary delimitation confirmed the presence of oversplitting in L. rufiantennus+F. tonkinensis and F. viridifemorata species groups and suggested that each group should be treated as a single species.

Yes -all data are fully available without restriction Abstract. Phylogenetic positions of the genus Longgenacris and one of its members, i.e. L.

43
Taxonomy is a process to take or collate decision continually. Any taxonomic decision 44 taken since the inception of zoological nomenclature in 1758 has relevance today, and on into 45 the future, no matter that decision was right or wrong [1]. The process of modern taxonomy can 46 be viewed as a taxonomic circle, and hypothesis established from any information should be tested with other sources of information, i.e. taxonomists must break out of the circle of inference in 48 species delineation work to raise the entity to species status [2].

49
Cryptic species usually refers to as one of two or more species that are morphologically

57
Despite the existence of cryptic species (overlumping), oversplitting may also exist especially 58 in early described species groups because of the lack of type comparison which usually lead to 59 repeated descriptions of the same species as different ones without actual morphological difference 60 [14]. Incorrect assignment of a species in genus or higher levels will also lead to description of the 61 same species as different ones because the comparison can't be made between the most close 62 relatives. In these cases, morphological revision is necessary to confirm the presence of 63 morphological differences among the closely related species. Moreover, other sources of data, 64 including geographical, biological, ecological, reproductive, behavioral and DNA sequence characters vary even among individuals from the same population. For example, specimens of each 85 species collected from the same locality on the same date exhibit similar pattern of variation in 86 tooth length (Fig 2), with median tooth longer than submedian and lateral teeth in some individuals 87 (Figs 2A, C, E and G), but nearly as long as (Figs 2B, D, F and H) or slightly shorter (Fig 2I) than 88 submedian and lateral teeth in other individuals, or with submedian teeth indistinct or even absent in 89 a few individuals ( Fig 2J). Therefore, it is difficult to identify specimens of F. viridifemorata group 90 using morphological characters only, and frequently the same specimen could be probably 91 recognized as different species by different identifiers.

92
Species delimitation using molecular data has attracted more and more attention from

113
Taxon sampling 114 A total of 152 individuals representing 3 families 9 genera and 14 species were sampled (S1 115   Table). At least five individuals from each population and as many populations as possible of the 116 widespread species were sampled whenever the specimens were available (S1 Table). The sample 117 and type locality of F. viridifemorata group were marked as in the map (Fig 3). Species assignation 118 of specimens was performed mainly following Li & Xia's [25] key to species plus geographical 119 information. For example, the specimens from type locality and neighboring places will be assigned 120 to the same species if there is no distinct difference between them. Partial COI sequences were from 121 our previous study (S2 Table) [14]. All specimens were preserved in 100% ethanol and stored at 122 room temperature.   Table). Each haplotype was blasted using MEGABLAST   To provide a profile for the setup of taxa and groups for calculating genetic distances, a sequences typically fall apart into a separate subnetwork for each Linnean species (but with a higher 179 rate of true positives for mtDNA data) and DNA sequences from single species typically stick 180 together in a single haplotype network [46]. Therefore, we constructed haplotype networks for

198
The trees inferred from COI and the combined alignments displayed similar topologies (Fig 4). 199 Nearly all species formed reciprocally monophyletic clades except F. tonkinensis+ L. rufiantennus 200 and F. viridifemorata groups. The main differences between the single COI gene tree and the 201 combined alignment tree were the placements of Emeiacris maculata, of which two clades did not 202 form monophyletic clade but were added in turn to the clade of its closest relative Paratonkinacris 203 vittifemoralis in COI gene tree (Fig 4A), and Apalacris tonkinensis, which is a member of the 204 subfamily Catantopinae but had a closer relationship to most of Melanoplinae members than 205 Tonkinacris sinensis in the combined alignment tree (Fig 4D)  As for NJ trees, the one deduced from single COI gene (S3 Fig A) Table), a 243 putative threshold for species assignment proposed by previous study (Herbert et al., 2003), with E.  Table).

250
The interspecific divergences of COI sequences within F. viridifemorata groups ranged from 251 1.00% to 2.03%, those between species of F. viridifemorata groups and F. tonkinensis were up to 252 5.53-6.08% and the one between F. tonkinensis and L. rufiantennus was 0.33%, but that between L. 253 rufiantennus and L. maculacarina was as high as 7.33% (S4 Table). The interspecific divergences 254 calculated from ITS1 and ITS2 sequences displayed similar distribution patterns (S5, S6 Tables), i.e. 255 species within F. viridifemorata group and F. tonkinensis+ L. rufiantennus group had much lower 256 between-species mean distances but the mean distances between other pairewise species were 257 distinctly much higher. For all of three alignments, the distances between species within Melanoplinae were constantly lower than those between species in Melanoplinae and that out of 259 Melanoplinae (Tables S4-S6).

1) Fruhstorferiola tonkinensis + Longgenacris rufiantennus group 263
Considering the high similarity between F. tonkinensis and L. rufiantennus, we sampled 15  Analysis with haplotype network led to a similar result. The numbers of COI haplotypes 287 detected in F. tonkinensis, L. rufiantennus and L. maculacarina were 12, 3 and 4, respectively (S7 288   Table). Among the 3 haplotypes detected in L. rufiantennus, the one represented by 11 individuals 289 was shared with F. tonkinensis, and the other two represented each by a single individual were 290 private for L. rufiantennus. In the net work from COI haplotypes (Fig 5B), haplotypes of  Table). In the net work from ITS1 296 sequences (Fig 5C), haplotypes of F. tonkinensis and L. rufiantennus formed a clade but no monophyletic subclade, and the 2 haplotypes of L. maculacarina did not connect into a single net work, but separated from each other. For ITS2 sequences, 3 haplotypes were detected for F. 299 tonkinensis and L. rufiantennus each, with two shared haplotypes (S9 Table). Haplotypes of all 3 300 species connected into a single net work together with haplotypes of F. viridifemorata group (Fig   301   5D), indicating a much lower evolution rate in ITS2 sequence.

302
In GMYC analysis based on COI sequences, 14 putative species were delineated from the 303 whole data set (S10 Table,   rufiantennus were delineated as the same species (Figs6C, S4 Fig; S10 Table). Samples of F.  In an earlier study, the relationship between F. kulinga and F. huayinensis was discussed using 313 single COI barcoding fragment, and the result did not support the validity of F. huayinensis [14].  Fig C).

330
Mean intraspecific variations within each species calculated from COI sequences were 331 distinctly less or slightly larger than 1%, and the largest pairwise intraspecific variation was as high 332 as 2.97% in F. kulinga, but still slightly less than 3%. Broad overlaps between intraspecific genetic 333 variations and interspecific divergences are found in all species pairs (Table 2). For ITS1 and ITS2 334 sequences, all intraspecific variations within population are distinctly less than 1% and only a few 335 ones between populations are slightly more than 1% (S3 Table). As for the interspecific divergences,

343
Haplotype network analysis detected no shared haplotype in COI sequences among the four 344 species (S7 Table), but shared haplotypes occur in ITS1 and ITS2 sequences among these species 345 (S8, S9 Tables). In the network from COI haplotypes (Fig 7B), all haplotypes are connected into a 346 large network in a maximum connection steps of 11 at 95%, but three of the four species do not 347 form reciprocally monophyletic clades. Although the three haplotypes of F. omei forms a so-called 348 monophyletic clades, the maximum mutational steps of haplotypes within F. omei reaches 4 steps, 349 slightly higher than the minimum mutational steps of haplotypes between F. omei and F. 350 viridifemorata. For ITS1 sequences, a haplotype shared by three species with high frequencies as 351 well as another one shared by two species with low frequencies are found (S8 Table). In the 352 network from ITS1 haplotypes (Fig 7C), still no species forms reciprocally monophyletic clades.

353
For IIS2 sequences, a haplotype shared by four species is found (S9 Table) and all haplotypes of F. 354 viridifemorata group and F. tonkinensis+L. rufiantennus group are connected into a single net work 355 as mentioned in the previous section (Fig 5D).

356
For the four putative species delineated in GMYC analysis (S10 Table), the putative species 9  (Figs 4A and D, S1 Fig, S2 Fig). Therefore, L. rufiantennus 375 should be regarded as a member of the genus Fruhstorferiola but not a member of Longgenacris no 376 matter according to morphological or molecular evidences.

377
As for the relationship between L. rufiantennus and F. tonkinensis, all analysis (NJ tree, genetic distance and haplotype network) lead to the same result that they should be the same species but not two independent species because all individuals of L. rufiantennus fall into the clade of F. 380 tonkinensis in NJ trees (Fig 5A, S3 Fig), the pairwise genetic distances within F. tonkinensis 381 completely overlapped with those between F. tonkinensis and L. rufiantennus (Table 1), the COI 382 haplotype of L. rufiantennus with highest frequency are shared with F. tonkinensis (S7 Table) and 383 all haplotypes of the two species formed a whole network under the 95% parsimony connection 384 limit (Fig 5B) , GMYC analysis delineated them as the same species (Fig 6C, S4 Fig; S9 Table).

387
Subfamily placement of the genus Longgenacris 388 The genus Longgenacris was originally placed in the subfamily Melanoplinae and considered 389 most similar to the genus Ognevia Ikonnikov, 1911 [15]. The phylogenetic position of the genus were added to the present study, and ITS region was employed in addition to COI sequence. 400 However, the increases of the sampled species and molecular markers did not lead to different result 401 from that of previous study [14]. It seemed that the resolution of the datasets were contributed 402 mainly by COI gene sequences, and ITS region had a much lower evolution rate than COI gene in  Table). Do the four MOTUs represent morphologically cryptic species or only ancient genetic polymorphism? Among species of F. viridifemorata group, the morphological characters 423 originally employed to describe the different species have been approved to be variable even within 424 populations of the same species (Fig 2), and most analyses of molecular evidences are congruent 425 with the result of morphological reexamination. As for the four MOUTs delineated by GMYC 426 analysis using COI gene (S10 Table), they didn't be supported by either morphologocial or 427 geographical informations. Furthermore, this approach tends to overestimate the number of species