Mitogenome phylogeny reveals Indochina Peninsula origin and spatiotemporal diversification of freshwater crabs (Potamidae: Potamiscinae) in China

Sampling and specimens

We collected 57 species of freshwater crabs in the family Potamidae, including 55 species of Potamiscinae and two species of Potaminae, between the years 2004 and 2019 from China and surrounding countries (Fig. 1). All specimens were stored in 95% ethanol and deposited in the Jiangsu Key Laboratory for Biodiversity and Biotechnology, College of Life Sciences, Nanjing Normal University (NNU), Nanjing, China. The names of the taxa sampled, and GenBank accession numbers are presented in the Supplementary material (Table S1). In addition, publicly available mitogenomes of 32 species were retrieved from GenBank (Table S1). Altogether, 89 species in 46 genera were analyzed, including 17 species in 12 genera as outgroup taxa (Table S1).

DNA extraction, sequencing and assembly

Total genomic DNAs were extracted from gill tissues with the Cell and Tissue DNA Extraction Kit (Generay Biotech) following the manufacturer’s protocol. The quality of samples was checked on 1% agarose gels and NanoDrop 2000 (Thermo Scientific), before sequencing on the Illumina HiSeq X Ten platform using a paired-end 150 bp protocol at Novogene. After discarding reads with low quality, more than 13 000 000 clean reads were obtained for each sample. Two strategies were used for assembly and annotation. Mitogenomes of 37 species were assembled with Geneious 11.1.5 (Kearse et al., 2012) and annotated with MITOS2 webservers (Bernt et al., 2013) following Zhang et al. (2020b). Mitogenomes of other species were assembled and annotated using MitoZ 2.3 (Meng et al., 2019) with default settings. Two nuclear genes, Histone H3 and 28S rRNA, were amplified following Ji et al. (2016) and assembled using SeqMan II 5.05 (DNASTAR Inc., Madison, WI, USA) with manual inspection.

Phylogenetic analyses

Each protein coding gene (PCG) was separately aligned employing codon-based alignment using the MAFFT algorithm (Katoh and Standley, 2013) on the TranslatorX online platform (Abascal et al., 2010). Three rRNAs, mitochondrial 12S rRNA, mitochondrial 16S rRNA and nuclear 28S rRNA, were aligned using MAFFT 7.310 with G-INS-i strategy (Katoh and Standley, 2013). All alignments were trimmed with Gblocks 0.91b (Castresana, 2000), and concatenated with custom python scripts. Substitution saturation was assessed using DAMBE 7.2.137 (Xia et al., 2003; Xia and Lemey, 2009) separately for each codon position of 13 mitochondrial PCGs. In both analyses, the observed index of substitution saturation (Iss) values were significantly lower than the critical index of substitution saturation (Iss.c) values. Five matrices were constructed. Matrix 1 contains two mitochondrial rRNAs and all three codon positions of 13 mitochondrial PCGs. In order to avoid the effect of substitution saturation at third codon positions, Matrix 2 was constructed, containing two mitochondrial rRNAs and the first and second positions of 13 mitochondrial PCGs. Matrix 3 consists of amino acid sequences of 13 mitochondrial PCGs. Matrix 4 was constructed from the subset of Matrix 1 by excluding distantly related outgroups (i.e., marine crabs), to minimize long branch attraction artefacts. Matrix 5 contains two nuclear genes, Histone H3 and 28S rRNA.

The best partitioning scheme and substitution model were selected utilizing a greedy heuristic algorithm implemented in PartitionFinder 2.1.1 (Guindon et al., 2010; Lanfear et al., 2012, 2017). Partitions were predefined according to codon positions for PCGs. Different models were evaluated based on the Bayesian Information Criterion.

Phylogenetic trees were inferred by Maximum Likelihood (ML), Bayesian Inference (BI) and Maximum Parsimony (MP) analyses. The ML tree was reconstructed using IQ-TREE 1.6.12 (Nguyen et al., 2015) with 1000 ultra-fast bootstrap replicates (Minh et al., 2013). BI was done by MrBayes 3.2.7 (Ronquist et al., 2012). We run four independent Markov Chain Monte Carlo (MCMC) of 20 000 000 generations sampling every 1000 generations. Convergence of four MCMC chains and effective sample size (ESS) were checked in Tracer 1.6 (Rambaut et al., 2018). All parameters had ESS over 200. The first 10% of MCMC chains were discarded as burn-in. The MP analysis was conducted using MPBoot 1.1.0 (Hoang et al., 2018) with 1000 bootstrap replicates and using TNT 1.5 (Goloboff et al., 2008; Goloboff and Catalano, 2016) using the ‘New Technology search’ method (Sectorial Search, Ratchet, Drift and Tree Fusing) with 1000 bootstrap replicates. All trees were visualized and edited with iTOL (Letunic and Bork, 2019).

Divergence time estimation

Divergence times were estimated using BEAST 2.6.0 (Bouckaert et al., 2014) with Matrix 1. To select the best clock model, a strict clock model and a lognormal relaxed clock model were compared through a path sampling analysis (Lartillot and Philippe, 2006) with 100 steps of 1 000 000 generations. For the dating analysis, the Yule process and a strict clock model were used. We incorporated three calibration points under a log-normal prior distribution of the age of respective fossils (Drummond et al., 2006). The split between Eriocheir hepuensis and E. japonica dated to the early Pliocene (5.3–3.6 Mya), delimited by the fossil record of E. japonica (Karasawa, 2000). The age of the genus Scylla was calibrated to the late Oligocene (28.4–23.0 Mya) according to the Scylla costata fossil (Rathbun, 1919). The Acanthopotamon martensi fossil from the latest Pliocene (~2.6 Mya) was used to constrain Paratelphusula (Klaus et al., 2017), because Acanthopotamon and Paratelphusula are closely related (Klaus et al., 2019). Four independent MCMC chains with 50 000 000 generations each were implemented. Trees were sampled every 1000 generations. Four runs were checked using Tracer 1.6 and combined using LogCombiner with discarding first 10% MCMC chains. A maximum clade credibility tree was summarized in TreeAnnotator. Trees were visualized in Figtree 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/).

Ancestral area reconstruction

Due to the lack of molecular sequences of potamiscine crabs outside China, 16S rRNA gene sequences of species used by Shih et al. (2009) were combined with 16S rRNA sequences of species included in this study. A phylogenetic tree based on this combined dataset was constructed using BEAST 2.6.0 as described above and used for the following reconstruction. Ancestral distribution areas were estimated employing Bayesian Binary Method (BBM) and Statistical Dispersal-Vicariance Analysis implemented in RASP 4.2 (Yu et al., 2020). Eight districts were defined, Southwest mainland China (A), Southeast mainland China (B), Central China (CC; C), West Pacific islands (D), Indochina Peninsula (E), Sunda shelf islands (F), Socotra island (G) and Philippine islands (H). According to our results (Figs 2, S1–S12) and Shih et al. (2009), genetically closely related Potamiscinae species are also geographically close. Species distributed in Sunda shelf islands (F), Socotra island (G) and Philippine islands (H) separately form three distinct clades (Shih et al., 2009). Delimitation of districts in East Asia was mainly based on reconstructed phylogenetic relationships and recognized biogeographical regions of freshwater crabs (Dai, 1999; Shih and Ng, 2011) and of freshwater fishes, amphibians and freshwater crabs (Huang et al., 2020c). Specifically, Southwest mainland China district (A) refers to Hengduan Mountains Region and Southwest Karst Area; Southeast mainland China district (B) refers to Pearl River Basin and Wuyi Mountains region; CC district (C) refers to Yangtze River Basin; West Pacific islands district (D) refers to Hainan Island, Taiwan Island and Ryukyu Islands (Fig. 4).

Data properties

All sequences were deposited in GenBank (accession numbers are given in Table S1). The resultant Matrix 1 and Matrix 4 both consist of 11 555 bp sites, of which 6501 bp and 5930 sites are parsimony informative, respectively. Matrix 2 consists of 8146 bp sites, of which 3172 bp sites are parsimony informative. Matrix 3 consists of 3357 amino acids, of which 1491 amino acids are parsimony informative. Matrix 5 consists of 1155 bp sites, of which 173 bp sites are parsimony informative.

Phylogenetic inferences

Four main clades were recognized in the ML, BI, and MP trees inferred from all five matrices (Figs 2, S1–S12). There is no sign that phylogenetic inferences were biased either by substitution saturation (Matrix2, Figs S1–S3) or by long branch attraction (Matrix 4, Figs S7–S9). Analyses of Matrix 1 (Fig. 2), Matrix 2 (Figs S1–S3), and Matrix 4 (Figs S7–S9) produced nearly identical topology, which differs in several deep nodes from that of Matrix 3 (Figs S4–S6). This may be due to the conservation of protein sequences (Matrix 3) for phylogeny reconstruction at the intergeneric level within a subfamily. In addition, the relationships among the four clades were poorly resolved with two nuclear loci (Matrix 5, Figs S10–S12), probably because of the limited number of parsimony informative sites.

Chinese potamiscine crabs were grouped into two major clades, one of which is further divided into three clades (Fig. 2). The four clades are: (i) Southwest China (SWC), (ii) Indochina-Southwest China (ISWC), (iii) CC, and (iv) South China-adjacent Islands (SCI), which are basically congruent with Shih et al. (2009). Species in the SWC clade are mainly found in southwestern China, except for Aparapotamon gracilipedum which occurs in Central China (Henan Province; Fig. 1). The relationships within the SWC clade are largely unresolved. The ISWC clade is restricted to the southwest border region of China (Fig. 1), and has three sub-clades within it: a clade consisting of Pupamon prabang sister to Potamiscus yiwuensis and Potamiscus montosus; a clade consisting of Indochinamon edwardsi sister to Eosamon lushuiense and E. tenchongense; and a clade comprising the rest of the species of Indochinamon clustered together with Iomon luangprabangense. The CC clade is widespread (Fig. 1) and consists of species of Sinopotamon nested within which is a well-supported sub-clade comprising Tenuilapotamon, Acatiapotamon, Latopotamon and Vadosapotamon (MLBS = 100, PP = 1, MPBS = 100). The SCI clade comprises three sub-clades (Fig. 2) each with a distinct geographic distribution. The first sub-clade (MLBS = 100, PP = 1, MPBS = 99) includes Candidiopotamon (endemic to Taiwan island and the Ryukyu islands) and Geothelphusa (endemic to Taiwan island). The second sub-clade (MLBS = 100, PP = 1, MPBS = 99) comprises genera from Hainan island (Hainanpotamon, Apotamonautes, and Neotiwaripotamon), while the third sub-clade (MLBS = 100, PP = 1, MPBS = 99) consists of genera from southeastern China (Chinapotamon, Huananpotamon, Nanhaipotamon, Sinolapotamon, Bottapotamon and Qianguimon).

Tempo-spatial biogeography

The divergence time analysis suggested that the split between the subfamilies Potaminae and Potamiscinae occurred about 61.3 Mya (95% HPD: 68.6–55.0 Mya; Fig. 3), which is compatible with the conclusion of Tsang et al. (2014) for these freshwater crabs. The common ancestor of all four potamiscine clades was most likely found on the Indochinese Peninsula (Figs 4 and S13). The SWC clade separated from the other three potamiscine clades around 47.2 Mya (95% HPD: 52.7–42.5 Mya), and its diversification within southwestern China started around 29.6 Mya (95% HPD: 33.1–26.5 Mya). The other three potamiscine clades each dispersed independently northward from the Indochina peninsula. The ISWC clade began to diversify at 27.2 Mya (95% HPD: 30.4–24.5 Mya; Fig. 3) while the CC clade diverged from the SCI clade around 36.1 Mya (95% HPD: 40.3–32.5 Mya). The diversification of the CC clade occurred around 35.9 Mya (95% HPD: 40.2–32.3 Mya) expanding first into southwestern China and then into CC (Fig. S2). The separation of the three sub-clades within the SCI clade occurred between 36.1 and 34.4 Mya (Fig. 3).

Evolutionary history of potamiscine crabs in China

China has the highest species richness of freshwater crabs (Cumberlidge et al., 2011). Several hypotheses have been proposed to explain the origin of freshwater crabs in China (Bott, 1970; Du, 1993; Dai, 1999; Shih and Ng, 2011). Ancestral area reconstruction, in this study, supports the Indochina Peninsula origin hypothesis (Dai, 1999). The pioneering potamiscine freshwater crabs, the SWC clade, came to Yunnan (Shih et al., 2009; Shih and Ng, 2011), during the Eocene (47.28 Mya, 95% HPD: 52.76–42.57 Mya, Fig. 3). During this period, the tropic zone of southern China became humid (Sun and Wang, 2005; Quan et al., 2012), allowing northward movement of freshwater crabs from the Indochina Peninsula to southwestern China. Similar pattern was also observed in other species (e.g., Xu et al., 2015; Janssens et al., 2016). With the development of humid conditions in southern China (Sun and Wang, 2005; Herman et al., 2017), other clades of potamiscine freshwater crab migrated to China.

The diversification pattern of potamiscine crabs in China is complex. The different clades of potamiscine crabs each appears to have a disparate biogeographic history. The CC and SCI clades both began to diversify during the late Eocene or early Oligocene at a time when the East Asian monsoon was first initiated, that, together with the uplift of the Qinghai-Tibetan Plateau (QTP), greatly expanded freshwater habitats (Kutzbach et al., 1993; Li and Fang, 1999; Sun and Wang, 2005; Harris, 2006; Song et al., 2009; Herman et al., 2017). During the Eocene, most regions of southern China were arid and only the southernmost regions were humid (Sun and Wang, 2005), but from the Eocene to the Oligocene the climatic conditions in southern China changed from arid to humid (Sun and Wang, 2005; Herman et al., 2017). The impact of significantly increased precipitation on the availability of freshwater habitats facilitated the northward dispersal of potamiscine crabs from the Indochinese Peninsula to China and their subsequent local diversification. These freshwater crabs now dominate the vast monsoon region of mainland Southeast Asia, and have become the predominant subfamily of extant freshwater crabs in China (Chu et al., 2018). Within China, potamiscine freshwater crabs have come to dominate the macroinvertebrate biomass of mountain streams. After their northward dispersal into southwestern China, the CC and SCI clades colonized suitable unoccupied novel heterogeneous freshwater niches in the warmer central and southern regions of the country (Vellend et al., 2007; Ptacnik et al., 2010; Mintenbeck et al., 2012).

The SCI clade includes species found on Hainan island, Taiwan Island and the Ryukyu islands that were once part of the landmass of mainland Asia (Kimura, 1996; Maruyama et al., 1997; Zhao et al., 2007). Shih et al. (2009) speculated that the island lineages that are part of the SCI clade split from the continental lineages when these offshore island groups first. However, our results suggest that the island lineages of these potamiscine crabs split from their continental relatives ca 30 Mya, long before the opening of the Qiongzhou Strait and the Okinawa Trough (Kimura, 1996; Maruyama et al., 1997; Zhao et al., 2007). Given the poor dispersal abilities of freshwater crabs, the formation of a strong geographic barrier might not be a prerequisite for the separation of freshwater crabs on these islands. The low dispersal abilities of freshwater crabs living on the edges of the continental landmass before the formation of these islands meant that they were unlikely to maintain gene flow with distant continental populations further inland, and would have become genetically isolated anyway. The subsequent marine transgressions that led to the separation of these islands from the mainland would have only strengthened the geographic and genetic isolation of these species (Giri and Collins, 2014; Parvizi et al., 2018), resulting in further speciation on these islands.

Several studies have demonstrated that the Hengduan Mountain region (HMR) in southwestern China is a biodiversity hotspot for a number of different groups, including the freshwater crabs (López-Pujol et al., 2011; Lei et al., 2014; Sun et al., 2017; Chu et al., 2018). Species of potamiscine freshwater crabs found in SWC belong to either the SWC clade or the ISWC clade. Although these two clades originated independently, the timing of their diversification in China is estimated to have occurred concurrently during the Oligocene. The extensive uplift of the QTP triggered the speciation of both plants and animals (Lei et al., 2014; Wen et al., 2014; Xing and Ree, 2017; Sun et al., 2018) including the local diversification of potamiscine crabs in the adjacent HMR just southeast of the QTP. The HMR began the first phase of rapid uplift during the Oligocene (Wang et al., 2012), trapping the northward flow of warm moist air from both the Indian and Pacific Oceans. The steep, rugged and forested Hengduan Mountains in SWC range from elevations between 1300 and 6000 m. The deep valleys are drained by major rivers from the eastern Tibetan Plateau. The HMR were free from glaciation during the Pleistocene ice ages and their dense relatively isolated forests today support a rich biological diversity. The mountainous land in this area became warm and moist, which created additional suitable freshwater habitats for freshwater crabs, while the formation of steep hot dry valleys limited dispersal and gene flow between freshwater crab populations (Zhang, 1992; Yang, 2000; Ma and McConchie, 2001). Populations of crabs that became isolated in the streams and rivers draining the numerous steep valleys of this vast inaccessible forested mountain range underwent extensive in situ species diversification over time, which may explain the large number of endemic species with highly restricted distributional ranges found today in the HMR (Dai, 1999; Chu et al., 2018).

Taxonomic implications

The CC, SCI, and ISWC clades originated from a common ancestor, but the SWC clade originated independently and diversified locally in the HMR. The SWC clade includes ten extant genera: Aparapotamon, Arquatopotamon, Artopotamon, Parvuspotamon, Pararanguna, Ruiyupotamon, Tenuipotamon, Semicircularum, Trichopotamon, and several species of Potamiscus. The habitats of these species are distinctly unique, inhabiting mountain streams at relatively high elevations (1500–3000 m). Members of the SWC clade share morphological traits such as a third maxilliped without a flagellum on the exopod and a relatively small adult body size range that support their monophyletic origin, but it is difficult to identify other shared characters that are unique to this clade.

The phylogenetic relationships obtained in the present study revealed that the current taxonomic treatments of six genera (Aparapotamon, Indochinamon, Potamiscus, Sinopotamon, Tenuilapotamon and Tenuipotamon) need to be refined.

Only eight out of the 11 species of Aparapotamon included in present study formed a monophyletic clade, with the other three species belonging to two separate lineages. For example, eight species grouped together in a clade that was sister to Lophopotamon yenyuanense, while A. emineoforaminum formed a separate clade, as did A. molarum and A. inflomanum that formed a clade that was sister to Ruiyupotamon. This molecular classification into three lineages is supported by three different shapes of the G1 (male first gonopod) terminal article in each lineage (cf. Dai, 1999). For example, the G1 terminal article is either a straight or nearly straight cylindrical shape in the clade with eight species of Aparapotamon (cf. Dai, 1999), while the G1 terminal article has a unique disciform protrusion at the tip in the clade with A. molarum and A. inflomanum, and the G1 terminal article is elongated in A. emineoforaminum (cf. Dai, 1999). In summary, Aparapotamon should now be restricted to eight species, including the type species A. grahami, while two new genera should be recognized, one that includes A. molarum and A. inflomanum, and one for A. emineoforaminum (in preparation).

Indochinamon is an enigmatic genus with 39 described species that are widely distributed in northeastern India, the Indochinese Peninsula, and southwestern China (Yeo and Ng, 2007; Chu et al., 2018; Naruse et al., 2018; Zhang et al., 2020a). Our results suggest that Indochinamon is not monophyletic because two other genera (Eosamon and Iomon) are nested within it. Eosamon is sister to Indochinamon edwardsi and Iomon is sister to Indochinamon chinghungense and Indochinamon daweishanense. Indochinamon. Eosamon, and Iomon are superficially very similar and were recognized as separate genera from species previously assigned to Potamon sensu lato which included species from both the Potamiscinae and the Potaminae (Yeo and Ng, 2007). Further comprehensive molecular and morphological studies are needed to resolve the classification of Indochinamon and other related genera (such as Eosamon, Iomon, Doimon and Beccumon).

Potamiscus is another enigmatic genus with a complex taxonomic history, and there have been several attempts to clarify its classification (Brandis, 2000; Yeo and Ng, 2007; Ng et al., 2008). For example, some species originally identified as Potamiscus have now been transferred to different genera, e.g., Kukrimon and Quadramon (Yeo and Ng, 2007). Our results indicate that Potamiscus as currently configured, is not monophyletic, because species currently assigned to this genus were divided into three groups in two clades in the present study. For example, P. yiwuensis and P. montosus, together with Pupamon prabang, were included in the ISWC clade, an arrangement that is supported by morphological characters such as a conical G1 terminal article in P. yiwuensis and P. montosus that is not shared by other species of Potamiscus, but which is similar to that found in species of Indochinamon. Within the SWC clade, Potamiscus loshingense, P. elaphrius and P. yongshengensis are closely related species from southwestern China, while P. motuoensis (sister to Tenuipotamon) is narrowly endemic to Tibet.

It is widely accepted that Sinopotamon s. l. is polyphyletic (Ji et al., 2016; Shih et al., 2016; Chu et al., 2018; Zhang et al., 2020b). Shih et al. (2016) proposed that Sinopotamon s. l. be split into two genera, Sinopotamon s. str. and Longpotamon, based on molecular data (partial sequences of 16S and CO1), and morphological characters (the in vivo position of the male first gonopod). Our results, however, do not support the monophyly of Sinopotamon s. str. and Longpotamon as currently configured (Shih et al., 2016; Fig. 2). For example, we found no support for the assignment of S. parvum to Longpotamon or for the assignment of S. baokangense to Sinopotamon s. str. In addition, our results indicate that Sinopotamon s. str. and Longpotamon each comprise two lineages, and that S. honanense and S. changanense do not conform to the genus diagnosis for Longpotamon with respect to the structure of the male first gonopod (Chu et al., 2018). It is clear that further study with more extensive sampling of Sinopotamon s. l. is needed to resolve these questions.

Our results also indicated that Tenuilapotamon as currently configured is not monophyletic and actually comprises two lineages. For example, T. joshuiense does not group together with T. latilum latilum, and instead the former taxon shares a subclade with S. parvum and Acartiapotamon inflatum. In addition, T. joshuiense can be distinguished from other species of Tenuilapotamon on the basis of several morphological characters (a G1 with a straighter terminal article, a less convex basal region of the G1 terminal article, and a shorter terminal article whose tip does not reach the tubercle of the pleonal lock in situ). Further studies are needed to fully address the taxonomic status of Tenuilapotamon, Acartiapotamon and Sinopotamon parvum.

Our molecular analyses also indicated that Tenuipotamon purpura is phylogenetically distinct from the other species currently assigned to this genus (T. baishuiense, T. huaningense, T. panxiense, T. tonghaiense, T. xingpingense and T. yuxiense) that all grouped together in a different well-supported clade. Dai (1999) designated T. purpura as the type species of Tenuipotamon which means that the other six species currently included in Tenuipotamon need to be transferred to a new genus (in prep). Characters that distinguish these two lineages are the adult size range and the tip of the male first gonopod terminal article (T. purpura is smaller species with a straighter first gonopod terminal article tip than the other species), and the distributional range (where T. purpura is restricted to southeastern Yunnan while the other six species are only found in central Yunnan).