In this study, a total of 24 species were reported, including two new species: a cryptic species in the T. minxianensis population and a cryptic species in the T. robusta population. The morphological and molecular data were consistent in 14 of the 22 species identified. The results showed that two cryptic species could be described in the biodiversity hotspot area, which reinforces the general view that a large amount of unrecorded diversity remains in the plateau loach. For example, only one haplotype has been identified in the clade of T. bleekeri and T. orientalis. Therefore, more specimens must be collected, sequences must be added, and the possibility of identifying more cryptic species cannot be ruled out.
Different numbers of MOTUs were identified in the four DNA barcode analysis methods: 17 different MOTUs were identified using the PTP and GMYC models, and 19 MOTUs were identified using the ABGD and BOLD methods. T. shiyangensis and T. leptosoma could not be distinguished by the PTP or GMYC model; however, the ABGD and BOLD methods allowed different MOTUs to be assigned to each species (Figure 3). The inconsistent results of the four methods may be due to differences in the methods used to distinguish species. The ABGD and BOLD methods are based on the genetic distance between species and distinguish species by the difference of intraspecific and interspecific genetic distances. The BOLD method defaults to a genetic distance threshold of 2.2% and ABGD of 2.8%, resulting in the same number of MOTUs defined by the two methods. Although the RESL in the BOLD system has a stronger taxonomic performance than that in the ABGD system and thus shows better species identification and MOTU assignment results (Ratnasingham and Hebert, 2013), the two methods in this study achieved the same results, which may be related to the identified species. A key aspect implicit in DNA barcoding analysis is the genetic distance threshold values used to define the MOTUs. COI genetic distance values from 1% (Hubert et al., 2008) to 2% (Keskin et al., 2013) have been considered the threshold values for fish DNA barcoding analysis. However, these values are derived from comparative analyses of species diversity in different aquatic ecosystems. For example, 2% is used to represent the DNA barcodes for the community of fish in certain rivers (Pereira et al., 2013). However, when DNA barcoding analysis was used for a group of closely related species (e.g., the same genus), a lower genetic distance value has been reported (Carvalho et al., 2011; Pereira et al., 2011, 2013). In particular, a low threshold value of 0.92% is needed to distinguish MOTUs in the genus Laemolyta (Anostomidae) (Ramirez and Galetti, 2015). Although most of the values obtained in this paper are above 1.47% (14 of 18 MOTUs, Table 2), the maximum threshold value of related species detected between the obtained MOTUs is 0.40%, and some species have shared haplotypes. The existence of haplotype sharing among different species of plateau fishes may be related to complex species differentiation mechanisms or convergent evolution associated with local adaptation (Shen et al., 2019; Chen et al., 2020). A lower threshold of genetic distance may be obtained when the genetic relationships between different species within a genus are analysed. Although this approach based on genetic distance analysis is easy to perform, it lacks phylogenetic content, uses artificial boundaries to distinguish species and lacks the objectivity of species evolution (Ortiz and Francke, 2016). The GMYC and PTP methods define species based on evolutionary trees. GMYC uses ultrametric trees to define species (Fujisawa and Barraclough, 2013), and PTP uses substitution calibrated trees to define species, which avoids the potential pitfalls of constructing time-calibrated species phylogeny. (Zhang et al., 2013). We believe that tree-based techniques are effective in identifying individual species because identifying a particular branch representing a particular species requires a threshold to represent the clade length and/or the pair distance used to distinguish differences between individuals (Gustafsson, et al., 2009; Vieites et al., 2009; Lim and Meier, 2011). Such thresholds are also required when DNA barcoding data are analysed using clustering methods and based on distance methods. A technical problem with clustering is that paired distances of three or more sequences need not be equal; therefore, strict thresholds are usually impossible to apply (Meier et al. 2006). Both methods defined 17 MOTUs in this study. Clearly, the accuracy of DNA barcoding methods depends largely on the target species being analysed (Pentinsaari et al., 2017)
The difference in the number of MOTUs detected by the different analysis methods was mainly seen in two pairs of MOTUs, with relatively low genetic distance observed between T. shiyangensis and T. stoliczkae (2.65%) and between T. leptosoma and T. papilloso-labiatus (1.47%). These relatively low genetic distance values may be related to the late differentiation of these MOTUs. Notably, the MOTUs of relatively recent origin had less time than species of distant origin to accumulate genetic differences, which hindered their correct identification, despite the species differing greatly in their morphological characteristics. T. papilloso-labiatus has an obvious swim bladder, while T. leptosoma does not (Zhao, 1984). The characteristics of the genetic diversity of these species are the same, and they both show a relatively high level of haplotype diversity (>0.5) and a relatively low level of nucleotide diversity (<0.5%) (Table 1). These characteristics indicate that after the differentiation of these species, the founder effect and environmental heterogeneity caused by water system changes led to the rapid accumulation of variation in the population, resulting in a high haplotype diversity index. The accumulation time of the nucleotide diversity index was much longer than that of the haplotype diversity index. In terms of geographical distribution, these two species are mainly distributed in the Shulehe River and Heihe River. The possibility of sympatric speciation exists; however, this supposition needs to be confirmed by further analysis.
An example of incompletely separated species was also found. T. minxianensis, T. robusta, T. pappenheimi and T. siluroides are not sufficiently differentiated by COI gene differences, and there are also shared haplotypes among the four species (Figure 5). These phenomena can be explained as frequent mitochondrial DNA introgression events before species differentiation (Feng et al., 2018) or phenotypic plasticity in fish (Robinson and Parsons, 2002; Thibert-Plante and Hendry, 2011). The morphological characteristics of T. hsutschouensis, which was identified as an independent species isolated from T. robusta, include bare and scaleless bodies and a relatively low ratio of body length to body height (Wang, 1991). T. robusta only has residual scales in specific parts of its body. The Jinghe River population of T. robusta has scales along the lateral line from the caudal fin to the front of the dorsal fin. Moreover, the Jinghe River population and other populations of T. robusta were clustered into two clades (Figure 3), and the genetic distance between the populations reached 7.9% (Table 2). These phenomena suggest the existence of cryptic species of T. robusta. Differences were not observed between T. minxianensis and T. minxianensis sp1 in the degradation of the swim bladder, whether the end of the pelvic fin reached the anus, the starting point of the dorsal fin and the pelvic fin relative to each other or the morphological measurement data. However, the scales of T. minxianensis sp1 were only found in the caudal peduncle, which is quite different from the scale pattern of T. minxianensis, in which all the body parts except the head have obvious round scales. The genetic distance between the two populations was 7.4% (Table 2), which indicated that cryptic species occurred in T. minxianensis. Similar to this example of incomplete species separation, Wang (1991) argued that the plateau loach groups without scales (T. hsutschouensis) come from scaly groups (T. minxianensis) following the degeneration of scales. The groups with remnant body scales (T. robusta) are the intermediate species between the two types. The presence or absence of scales marks a leap in the evolution of plateau loach populations. The cryptic species found in this study provide more evidence for this speculation.
The morphological characteristics and molecular characteristics were inconsistent in T. pseudoscleroptera and T. scleroptera. The two species have similar appearances but different internal anatomical structures. The anterior and posterior segments of the swim bladder of T. pseudoscleroptera are the same size, with a long pouch or oblong oval shape and no pyloric caecum. The posterior chamber of the swim bladder of T. scleroptera is developed, the anterior segment is thin, and the posterior segment is enlarged into a long pouch (Zhu et al., 1981). Without the comparison of internal anatomical structure, these species are easy to misidentify, and morphological identification may be incorrect (He et al., 2008). However, due to the low interspecific distance between the two species (0.40%), the two MOTUs could not be correctly distinguished. This inconsistency was also found between T. dalaica and T. stoliczkai. The posterior chamber of T. dalaica's swim bladder was oval, while the posterior chamber of T. stoliczkai 's swim bladder was degraded; thus, this feature can be used to accurately distinguish the two species.
As shown by the two cases reported here, DNA barcoding did not identify enough differences to distinguish similar species because the lineages were not completely divided into different clades. The reason for this phenomenon is the process of incomplete lineage sorting. Due to the extremely short time of species differentiation, ancestral traits are randomly fixed in the differentiated species (Fontenot et al., 2011; Leavitt et al., 2017). Similar phenomena have been found in Psorophora (Chan-Chable et al., 2016), Syngnathidae (Zhang et al., 2017) and Laemolyta (Ramirez et al., 2015), and mixed lineage cases are particularly common in plateau fish (Shen et al., 2018). In this sense, to find evidence of reproductive isolation, nuclear genetic and ecological data must be combined for further research (Mardulyn et al., 2011; Versteirt et al., 2015; Beebe, 2018).
Species with morphological characteristics that are not significantly different may be easily identified as a single species. For example, T. bleekeri and T. polyfasciata have very similar morphological characteristics and do not present significant differences in quantitative traits in different proportions of their bodies, and they have been identified as the same nominal species. Ding et al. (1996) believed that they should be divided into two different species based on molecular data and pointed out that the main distinguishing feature was that 10-12 wide dark brown horizontal stripes occurred on the side of the body. However, even among T. bleekeri individuals collected from the same site, the number of horizontal stripes on the side of its body can range from 0-10. Of the specimens collected from the Wenchuanhe River in Sichuan Province, most had 5-7 horizontal stripes, and almost none had more than 10. Therefore, the validity of T. polyfasciata is still questionable (He et al., 2008). In this study, the numbers of these two species of plateau loach collected were relatively small, with 10 T. bleekeri and 5 T. polyfasciata, and 7-9 horizontal stripes were observed on the sides of the fish bodies. Although the division into two different species was also not supported by morphology, the genetic distance between the two species reached 8.57%, which far exceeded the threshold of genetic distance within the species of 2% (Pereira et al., 2013). Therefore, these two species likely underwent genetic differentiation in terms of genetic material; however, due to the small size (the length of the collected sample is 5-8 cm), the morphological difference is not obvious; therefore, they have historically been regarded as one species. Obviously, the body colour or body markings of plateau loach may not be an effective classification feature for the identification of species and cannot be used as the main basis for identification.
Herzenstein (1891) identified T. papilloso-labiatus as a subspecies of T. strauchii, and this finding was also supported by Zugmeyer (1910). T. strauchii lack a developed mastoid process similar to that of T. papilloso-labiatus and only have a strong naked fold; however, the mastoid process on the upper lip of plateau loach inhabiting the Hexi Corridor is obviously a double line, while that on the lower lip is a blurred double line. Characteristics such as the mastoid process and a strong naked crease are continuously transitive in a geographical distribution without obvious boundaries. However, the appearance of significant double lines on the mastoid marks discontinuity in the variation, and relatively stable differences are also observed in a series of other morphological traits. Thus, T. papilloso-labiatus should be regarded as an independent species (Li and Chang, 1974; Zhao, 1984), which is also supported in the phylogenetic tree constructed in this study (Figure 3). T. strauchii and T. papilloso-labiatus are clustered into two different clades and should be independent species.
Limited differences are observed in the morphological characteristics between T. wuweiensis and T. scleroptera. Li and Chang (1974) regarded T. wuweiensis as an independent species based on 7 morphological traits. Zhu and Wu (1975, 1981) believed that a certain continuity occurs in the identification characteristics of these two species. However, after collecting specimens of T. scleroptera distributed in the Datonghe River, which is only one mountain away from the collection location of the T. wuweiensis specimens, Zhao (1984) believed that significant differences occurred between the two species in the number of pectoral fin rays, intestinal shapes and gill rakers, thus supporting T. wuweiensis as an independent species. In this study, T. wuweiensis and T. scleroptera clustered into different clades, and the two species were greatly differentiated, which also support the idea that T. wuweiensis is an independent species. The low genetic diversity of T. wuweiensis may be due to the short time since species differentiation and the low haplotype diversity, and nucleotide diversity may be caused by the founder effect and the narrow distribution area (the species is distributed only in the east and west Shiyanghe River tributaries).
T. shiyangensis, T. papilloso-labiatus and T. hsutschouensis are distributed in three inland river systems in the Hexi Corridor. The maximum intra-species genetic distance of these three species is more than 1%, which may be mainly due to the wide geographic distribution of the three species and the large population differentiation caused by the barriers created by the water systems. This phenomenon also appears in the sympatric distribution of Gymnocypris chilianensis, in which each geographic population is clustered into a single clade with large genetic differentiation (Zhao et al., 2011).
The different geographic populations of some widespread species are identified as different species or subspecies due to some more significant morphological differences. For example, T. stoliczkae was divided into 7 subspecies (Herzenstein, 1891) due to the differences in the number of gill rakers, the proportion of quantitative traits and the number of spiral loops of intestinal tubes with changes in altitude or water system. In this study, the samples were collected in three drainage systems (the Yellow River, Jialing River and inland rivers in the Hexi Corridor). The maximum genetic distance within the species was greater than 1.2% (Table 2). However, the samples of different water systems have shared haplotypes, which indicates that different geographic populations of T. stoliczkae in the surveyed area are from a common ancestor.
The membranous swim bladder of T. obscura is very developed, with a constriction in the middle, and its length accounts for approximately 2/3 of the abdominal cavity. In contrast to T. orientalis, its body surface has obvious spines and is regarded as an independent new species (Li, 2017). In this study, a relatively large number of samples (n=234) were collected in the distribution area. The phylogenetic tree showed that the samples from different water systems were clustered into different clades, the maximum genetic distance within the species was 2%, and the nucleotide diversity and haplotype diversity were relatively high (h=0.887, π=0.00777). These findings indicate that large differentiation occurs between the two geographically separated populations of T. obscura and the possibility of allopatric speciation. T. obscura and T. orientalis were also divided into two different monophyletic lines in the phylogenetic tree, which is consistent with the results of the analysis of Wu (2017).
Although only 3 specimens of T. sp1 were collected in the Liangdang section of the Jialing River, obvious differences were observed in morphological characteristics from other species of plateau loach. Thus, it should be identified as a new species that has not been reported; in addition, more specimens should be collected for further confirmation. T. sp2 was collected in the Jialing River and showed degeneration of the membranous swim bladder, leaving only a small chamber, an anus near the start of the anal fin, the end of the pelvic fin adjacent to the anus, a large spot on the back of the body, a spot on the side of the body and other morphological characteristics that were obviously different from those of the closely related species T. obscura. A detailed description of these newly discovered species is necessary to record the relationship between morphology and molecular identification criteria (Versteirt et al., 2015).