Mitogenomic phylogeny, biogeography, and cryptic divergence of the genus Silurus (Siluriformes: Siluridae)

Abstract

The genus Silurus, an important group of catfish, exhibits heterogeneous distribution in Eurasian freshwater systems. This group includes economically important and endangered species, thereby attracting considerable scientific interest. Despite this interest, the lack of a comprehensive phylogenetic framework impedes our understanding of the mechanisms underlying the extensive diversity found within this genus. Herein, we analyzed 89 newly sequenced and 20 previously published mitochondrial genomes (mitogenomes) from 13 morphological species to reconstruct the phylogenetic relationships, biogeographic history, and species diversity of Silurus. Our phylogenetic reconstructions identified eight clades, supported by both maximum-likelihood and Bayesian inference. Sequence-based species delimitation analyses yielded multiple molecular operational taxonomic units (MOTUs) in several taxa, including the Silurus asotus complex (four MOTUs) and Silurus microdorsalis (two MOTUs), suggesting that species diversity is underestimated in the genus. A reconstructed time-calibrated tree of Silurus species provided an age estimate of the most recent common ancestor of approximately 37.61 million years ago (Ma), with divergences among clades within the genus occurring between 11.56 Ma and 29.44 Ma, and divergences among MOTUs within species occurring between 3.71 Ma and 11.56 Ma. Biogeographic reconstructions suggested that the ancestral area for the genus likely encompassed China and the Korean Peninsula, with multiple inferred dispersal events to Europe and Central and Western Asia between 21.78 Ma and 26.67 Ma and to Japan between 2.51 Ma and 18.42 Ma. Key factors such as the Eocene-Oligocene extinction event, onset and intensification of the monsoon system, and glacial cycles associated with sea-level fluctuations have likely played significant roles in shaping the evolutionary history of the genus Silurus.

INTRODUCTION

Understanding phylogenetic relationships, biogeographic history, and species diversity is not only fundamental to evolutionary biology and ecology but also to conservation biology (Alfaro et al., 2009; Donoghue, 2008; Smith et al., 2005; Winter et al., 2013). However, this task is challenging with respect to globally distributed taxa. Notably, widely distributed taxa have generally undergone multiple colonization and/or speciation events by virtue of topographic and climatic changes, resulting in complex (and often unresolved) phylogenetic relationships (Crisp et al., 2009; Hotaling et al., 2021; Jetz et al., 2014; Murphy & Austin, 2005; Swenson, 2011). Furthermore, species diversity has often been underestimated due to limited spatial, hence taxonomic, coverage and potential cryptic diversity (Bickford et al., 2007; Cicconardi et al., 2013; Pfenninger & Schwenk, 2007; Sholihah et al., 2021a, 2021b, Siler et al., 2011; Utami et al., 2022). Consequently, to accurately assess the biodiversity of widely distributed taxa and to improve our understanding of their evolutionary history, it is necessary to conduct phylogenetic and biogeographic analyses, as well as large-scale evaluations of species diversity using molecular approaches.

The genus Silurus, an important group of catfish species belonging to the family Siluridae, is found throughout the freshwaters of Eurasia. The genus includes notably large species (e.g., Silurus glanis Linnaeus 1758), which can exceed 2 m in length and weigh up to 100 kg. To date, researchers have described 16 species worldwide (Britz & Win, 2010; Hibino & Tabata, 2018; Li et al., 2019), including 13 species in East Asia and three species in Europe and Central and Western Asia (https://www.fishbase.org/) (Lu et al., 2023). Both Silurus asotus Linnaeus 1758 and S. glanis are widely distributed across the Eurasian continent and are considered of economic importance in many countries (https://www.fishbase.org/), whereas the remaining 14 species exhibit more narrow distribution ranges, with over half classified as endangered or threatened (Jiang et al., 2016; EU Habitat Directive IV). For example, Silurus mento Regan 1904, once abundant in Dianchi Lake in Yunnan, China, has long been commercially fished, resulting in a dramatic decline in its population since the 1970s, remaining scarce for decades (Chu & Chen, 1989; Yang et al., 2008). As such, various species within the genus have attracted substantial research interest due to their economic value and precarious conservation status (Copp et al., 2009; Gao et al., 2011; Leonardos et al., 2009; Loannis et al., 2007; Luo & Xie, 2008; Yang et al., 2008).

Evaluating phylogenetic relationships and species diversity within a genus are fundamental steps toward the implementation of conservation measures for endangered and threatened species. Previous studies have explored the phylogenetic affiliations of Silurus species using complete mitochondrial genome (mitogenome) sequences (Kappas et al., 2016; Kishimoto et al., 2022; Mabuchi et al., 2020; Park et al., 2020; Schedel et al., 2022; Székvári et al., 2021; Vittas et al., 2011; Wang et al., 2015a, 2015b; Wu et al., 2021; Yang et al., 2019; Zeng et al., 2011). Nevertheless, these studies primarily relied on a limited number of species and restricted sampling ranges, leaving many taxa unexamined and the overall phylogeny unresolved. Furthermore, the diversity within the genus remains only partially understood, as evidenced by the identification of three new species (i.e., Silurus duanensis Hu, Lan & Zhang 2004, Silurus longibarbatus Li, Li, Zhang & He 2019, and Silurus tomodai Hibino & Tabata 2018) in the past two decades (Hibino & Tabata, 2018; Hu et al., 2004; Li et al., 2019).

The biogeographic history of a taxon generally results from complex interactions between past ecological processes and geological events, acting together over time (Hoorn et al., 2010; Hubert & Renno, 2006; Olden et al., 2010). Deciphering the biogeographic history of the genus Silurus could help elucidate the patterns of diversification within the group and to clarify the various processes that have influenced these patterns across different spatial and temporal scales. A deeper understanding of the biogeographic history of a taxon not only sheds light on the mechanisms underlying uneven distribution patterns but also provides opportunities to test hypotheses regarding the origins of species and their long-distance dispersal. To date, however, a comprehensive biogeographic study of the genus Silurus has yet to be conducted, with the origin and spatial expansion of the genus remaining largely unexplored.

As a freshwater fish lineage, the evolutionary history of Silurus is likely influenced by major biogeographic barriers in East Asia, such as the Korean Strait, Nanling Mountains, and Baekdudaegan Mountains (Chen et al., 2022; Feng et al., 2023; Kim et al., 2012). In addition, fragmentation and multiple vicariance events are expected, given the high number of endemic species with narrow distribution ranges. This pattern is also expected to favor the emergence of high levels of cryptic diversity within ancient and widely distributed species, challenging traditional taxonomies and calling into question current estimates of species diversity. For example, genetic analyses using the mitochondrial cyt b gene have revealed that S. asotus populations in several major Chinese watersheds consist of multiple genetic lineages (Xu et al., 2017), suggesting an underestimation of species diversity in the genus. Consequently, it is expected that other species within Silurus are still awaiting discovery.

In this context, we sequenced and annotated 89 new mitogenomes from Silurus species collected in China (83 specimens involving six species), Uzbekistan (five S. glanis specimens), and Germany (one S. glanis specimen). Combined with 20 previously published Silurus mitogenomes (11 species), we assembled a dataset consisting of 109 mitogenomes representing 13 recognized species to investigate the phylogenetic relationships and divergence history within the Silurus lineage. Based on extensive spatial coverage and multiple samples from different biogeographic regions for widely distributed species, we further explored the phylogeographic patterns in the genus. We aimed to: (a) reconstruct a robust mitochondrial phylogeny for the genus Silurus; (b) infer the biogeographic history of the genus and identify its origin, timing and pathways of dispersal; and (c) assess species diversity within the genus using standardized DNA sequence-based species delimitation.

MATERIALS AND METHODS

Sample collection and DNA extraction

We collected 89 Silurus samples belonging to eight morphospecies from 20 localities in Eurasian drainages (Supplementary Table S1; Figure 1). We failed to obtain samples of three species, including Silurus burmanensis Thant 1966 (only found in Myanmar), S. mento (only found in Dianchi Lake, Yunnan, China), and Silurus triostegus Heckel 1843 (only found in Iraq), due to their very restricted distribution ranges and limited information concerning their abundance and ecology (Britz & Win, 2010; Hibino & Tabata, 2018; Yang et al., 2008; www.fishbase.org). The 24 localities covered most of the geographic distribution ranges of the Silurus species. For widespread species, multiple samples were collected from different localities. Samples of fin and/or muscle tissue were obtained and preserved in 95% ethanol. Total genomic DNA was extracted from fin or muscle samples of each specimen using a Qiagen DNeasy Tissue Kit (Hilden, Germany) according to the manufacturer’s instructions.

Figure 1.

Geographic distribution of 13 valid Silurus species used in the present study

Phylogenetic tree at left bottom corresponds to topology inferred in the present study and different shapes and colors indicate distinct morphological species. Shapes in dashed ellipses represent identical localities. Detailed sampling information of Silurus specimens can be found in Supplementary Table S1.

Mitogenome sequencing, sequence quality control, assembly, and annotation

To maximize the efficiency of genome assembly, 1 µg of DNA from each sample was used for library construction. The libraries were prepared using an Illumina SeqTM DNA Sample Prep Kit (Illumina, USA) with an insert size of 300–500 bp. Each library was sequenced on the Illumina HiSeq 4000 platform (Illumina, USA), generating 150 bp paired-end reads.

All sequence reads were filtered using the NGS QC Toolkit v.2.3.3 (Patel & Jain, 2012). Cutoff values of 50 for the percentage of read length and 20 for the PHRED quality score were used for editing reads. MITOZ v.2.3 was used to assemble high-quality reads into contigs (Meng et al., 2019). Mitogenome annotations were performed using MitoAnnotator v.3.32 (Iwasaki et al., 2013). Gene boundaries were manually checked by aligning homologous annotated gene sequences of published Silurus species using MEGA v.6 (Tamura et al., 2013). To ensure the quality of the assembled mitogenomes, mitochondrial contig mapping was performed using Geneious R11 (Kearse et al., 2012). All novel mitogenomes were deposited in GenBank under accession numbers OQ791283, OQ835313–OQ835316, OQ843503–OQ843543, and OQ851421–OQ851463 (Supplementary Table S1).

Mitogenome alignment

We downloaded 19 publicly available Silurus mitogenomes from the NCBI database, representing 10 morphological species (Supplementary Table S1), although relevant DNA sequences were not found for S. burmanensis, S. mento, or S. triostegus. A total of 35 localities from Eurasian drainages were included, covering most of the distribution ranges of Silurus species. In addition, four Siluridae species belonging to different genera (Ompok bimaculatus Bloch 1794, Pterocryptis cochinchinensis Valenciennes 1840, and Kryptopterus vitreolus Ng & Kottelat 2013), as well as a representative of four Siluriformes families (Ictaluridae, Cranoglanididae, Clariidae, and Saccobranchidae) and two Cypriniformes (Danio rerio Hamilton 1822 and Cyprinus carpio Linnaeus 1758) were used as outgroups (Supplementary Table S1).

The mitogenomes were aligned using MUSCLE (Edgar, 2004), with 22 transfer RNA (tRNA) genes, 13 protein-coding genes (PCGs), and two ribosomal RNA (rRNA) genes manually extracted using MEGA v.6. The control region from the final matrix were excluded as it presented repetitive patterns that are generally poorly assembled using short reads. For the PCGs, stop codons were removed from each sequence.

Phylogenetic analyses

To determine the saturation of each gene sequence, we performed substitution saturation using DAMBE v.6.4.20 (Xia, 2013; Xia & Lemey, 2009). No substitution saturation was found for the 13 PCGs, 12S rRNA, 16S rRNA, or 22 tRNAs, even in the third positions of the 13 PCGs (Table 1). Therefore, 37 mitochondrial genes were concatenated as a “supergene” dataset (15 656 bp) using local Perl scripts. The sequence alignment was partitioned into 42 subsets: a total of 39 sets for three codon positions of each PCG, two for the 12S and 16S rRNA genes, and one for the 22 tRNA genes. The best partitioning schemes and corresponding nucleotide substitution models for the supergene dataset were chosen using PartitionFinder v.2.1 (Lanfear et al., 2012). Scores for different models were calculated based on the Bayesian Information Criterion.

Table 1. Saturation tests of three codon positions in PCGs, 12S rRNA, 16S rRNA, and combined tRNAs using Xia’s method in DAMBE5.

Gene region	Symmetrical tree			Asymmetrical tree
	Iss	Iss.c	P	Iss.c	P
All positions of 13 PCGs	0.120	0.818	<0.0001	0.572	<0.0001
1st positions of 13 PCGs	0.063	0.809	<0.0001	0.555	<0.0001
2nd positions of 13 PCGs	0.021	0.809	<0.0001	0.555	<0.0001
3rd positions of 13 PCGs	0.312	0.809	<0.0001	0.555	<0.0001
12S rRNA	0.166	0.747	<0.0001	0.440	<0.0001
16S rRNA	0.126	0.782	<0.0001	0.506	<0.0001
22 tRNAs	0.084	0.777	<0.0001	0.497	<0.0001

Maximum likelihood (ML) and Bayesian inference (BI) reconstructions of the phylogenetic relationships within Silurus were performed using the best partitioning schemes (Table 2). The GTR+I+Γ model was used to perform bootstrapping in RAxML-VI-HPC v.2.0.1 (Stamatakis, 2006), with a random seed and 1 000 replicates. The BI analyses were performed in MrBayes v3.1.2 (Ronquist & Huelsenbeck, 2003) with two independent runs, each with four Markov Chain Monte Carlo (MCMC) chains. All nucleotide substitution models were unlinked across partitions, and different partitions were allowed to evolve at different rates. An independent run was performed for 20 million generations, sampling every 1 000 generations, with the first 25% discarded as a burn-in, while the remaining samples were used to summarize Bayesian posterior probabilities (BPPs). Convergence of the BI analyses was assessed based on the average standard deviation of split frequencies being less than 0.01, and the potential scale reduction factors (PSRF) being close to 1.0 for all parameters. The convergence of BI analyses was checked using Tracer v.1.5 (Rambaut & Drummond, 2007). Genetic divergence among observed clades and/or subclades was calculated in MEGA using the Kimura two-parameter model (Kimura, 1980).

Table 2. Best partitioning schemes and corresponding nucleotide substitution models of all mitogenomes used in phylogenetic inferences within Silurus.

Scheme number	Partition scheme	Best model
1	ATP6_1st, ND1_1st, ND2_1st, ND3_1st, ND4_1st, ND5_1st, CYTB _1st	GTR+I+G
2	ATP6_2ed, ATP8_2ed, ND1_2ed, ND2_2ed, ND3_2ed, ND4_2ed, ND5_2ed	GTR+I+G
3	ATP6_3rd, COIII_3rd, ND1_3rd, ND2_3rd, ND3_3rd, ND4_3rd, ND4L_3rd, ND5_3rd, CYTB_3rd	GTR+I+G
4	ATP8_2ed, 22tRNAs	HKY+I+G
5	ATP8_3rd, 12sRNA, 16sRNA	GTR+I+G
6	COIII_1st, COI_1st, COII_1st, ND4L_1st	TIMEF+I+G
7	COI_2ed, COII_2ed, COIII_2ed, ND4L_2st, CYTB_2ed	HKY+I
8	COI_3rd, COII_3rd	TRN+I+G
9	ND6_1st, ND6_2ed	HKY+I+G
10	ND6_3rd	TRN+G

Delimitation of molecular operational taxonomic units (MOTUs)

Three PCGs, i.e., COI, CYTB, and ND2, were selected for delimitation analyses. Distance- and tree-based methods were applied to determine whether multiple MOTUs were present among the most widespread species, S. asotus and Silurus microdorsalis Mori 1936. Silurus lithophilus Tomoda 1961, S. tomodai, and Silurus soldatovi Nikolskii & Soin 1948 were included in delimitation analyses due to their close phylogenetic relationships with the S. asotus clades. First, genetic distances among different clades/subclades determined using the K2P model were input into MEGA. Second, automatic barcode gap discovery (ABGD) was performed using the default value for the relative gap width (X=1.5) and K2P distance (Puillandre et al., 2012). Third, the Poisson tree process (PTP; Kapli et al., 2017) in its multiple rates version was run using the web server (http://mptp.h-its.org). An ML tree was constructed in RAxML-VI-HPC using the GTR + I +Γ model selected by MrModeltest (Nylander, 2004). Fourth, the general mixed Yule-coalescent (GMYC) model (Fujisawa & Barraclough, 2013) based on single- and multiple-thresholds was run using the GMYC web server (http://species.h-its.org/gmyc/). A fully resolved ultrametric gene tree used in the GMYC analysis was built in BEAST v.1.8 (Drummond & Rambaut, 2007) by employing a Yule pure-birth model tree with the GTR + I + Γ substitution model and an uncorrelated relaxed lognormal clock model. Two MCMC chains of 50 million steps, sampled every 1 000 steps, were run and convergence was assessed using Tracer v.1.5.

Dating analyses

Divergence dating with an approximate likelihood calculation was performed using MCMCTREE in PAML v.4.9 (Yang, 2007). Due to the low genetic divergence among sequences within MOTUs, one representative sequence for each MOTUs was chosen. Both S. soldatovi and S. lithophilus were included, although they clustered with different morphospecies with low divergence. To estimate divergence times, a BI phylogenetic tree was reconstructed using selected specimens with MrBayes v.3.1.2 with the same parameters used for BI analyses above and the best partitioning schemes selected by PartitionFinder (Table 3). For this reconstruction, two outgroups outside Siluriformes (Danio rerio and Cyprinus carpio) were added.

Table 3. Best partitioning schemes and corresponding nucleotide substitution models of selected mitogenomes per MOTUs in Bayesian inference used in dating analyses.

Scheme number	Partition scheme	Best model
1	ATP6_1st, ND1_1st, ND2_1st, ND3_1st, ND4_1st, ND4L_1st, CYTB _1st	GTR+I+G
2	ATP6_2ed, ND1_2ed, ND2_2ed, ND4L_2st, ND3_2ed, ND4_2ed	GTR+I+G
3	ATP6_3rd, ND1_3rd, ND2_3rd, ND3_3rd, ND4_3rd, ND4L_3rd, CYTB_3rd	GTR+I+G
4	ATP8_2ed, 22tRNAs, ATP8_2ed, ND5_1st,	TVM+I+G
5	ATP8_3rd, ND5_3rd	GTR+I+G
6	COIII_1st, COI_1st, COII_1st	TIMEF+I+G
7	COI_2ed, COII_2ed, COIII_2ed, CYTB_2ed	K81UF+I
8	COI_3rd, COII_3rd, COIII_3rd, ND5_2ed	GTR+I+G
9	ND6_1st, ND6_2ed	K81UF+I
10	ND6_3rd	TRN +G
11	12sRNA, 16sRNA	GTR+I+G

One fossil calibration and three estimated calibrations were employed, including the: (i) most recent common ancestor (MRCA) of the root was the oldest fossil of the crown-group Ostariophysi, the stem gonorynchiform †Rubiesichthys gregalis (Benton et al., 2015; Fara et al., 2010) from the Berriasian of Spain (145.5–140.2 million years ago (Ma)). A lower bound of 140.2 Ma and an upper bound of 145.5 Ma were employed as age constraints of the root. (ii) divergence time between Cyprinidae and Danionidae was estimated at 59–122 Ma (www.timetree.org). (iii) divergence time between Siluridae and Cranoglanididae was estimated at 49–66 Ma (www.timetree.org). (iv) first appearance of the African Clariidae was in the Lower Eocene (34–56 Ma), lower age constraint of the clade (Clarias, Heteropneustes)=34 Ma, and soft upper bound=140.2 Ma. The overall substitution rate was estimated using the BASEML program in PAML with a root age of 145.5 Ma. The first 200 000 cycles in MCMCTREE were discarded as burn-in, with sampling every 50 cycles to obtain a total of 20 000 samples. Three independent MCMC runs were conducted with different random seeds to check the convergence of each run. The distributions of the parameter values from the MCMC samples were assessed using Tracer v.1.6, with the values of effective sample size exceeding 200.

Ancestral area reconstruction

The dated BI tree described above was used for biogeographical analyses and Reconstruct Ancestral State in Phylogenies (RASP v.4.2 (Yu et al., 2020)) was used for ancestral area reconstructions. The optimal model was selected using the model selection function in the BioGeoBEARS package within RASP v.4.2, which compared AIC values to determine the best fit for the data. Geographic units for biogeographic reconstructions were based on recognized biogeographical regions separated by well-known biogeographic barriers in Eurasia, i.e., Qinghai-Xizang Plateau and Korea Strait. Given that, three regions were defined: China and Korean Peninsula (A), Europe and Central and Western Asia (B), and Japan (C).

RESULTS

Taxonomic coverage and mitogenome characteristics

The 89 newly sequenced mitogenomes of Silurus were circular, with lengths ranging from 16 419 to 17 610 bp and average sequencing depths ranging from 556× to 3 254× (Supplementary Table S1). Among these 89 new mitogenomes, a total of 43, seven, six, one, six, seven, 13, and six were generated for S. asotus, S. duanensis, S. glanis, S. graham Regan 1907, Silurus lanzhouensis Chen 1977, S. longibarbatus, Silurus meridionalis Chen 1977, and S. microdorsalis, respectively. The mitogenomes contained 13 PCGs, 22 tRNA genes, two rRNA genes (12S and 16S RNAs), and a putative control region. The observed variations in length were mostly due to the variable length of the control region. Gene order was identical to that of other Osteichthyes.

Phylogenetic relationships

The length of the “supergene” dataset containing the 13 PCGs, 22 tRNAs, and two rRNAs reached 15 656 bp. The best-fit partitioning scheme determined by PartitionFinder for the 116 mitogenomes (including Clarias fuscus Lacepède 1803, Cranoglanis bouderius Richardson 1846, Heteropneustes fossilis Bloch 1794, Ictalurus punctatus Rafinesque 1818, Kryptopterus vitreolus, Ompok bimaculatus, and Pterocryptis cochinchinensis as outgroups) included 10 partitions, with corresponding evolutionary models (Table 2). The genus Silurus formed a strongly supported monophyletic group in the BI and ML trees, with a posterior probability of 1.0 and bootstrap score of 100 (Figure 2). The ML and BI trees yielded identical topologies and resolved eight well-supported clades: Clade A, most specimens of S. asotus and all specimens of S. soldatovi; Clade B, two S. asotus individuals, S. lithophilus, and S. tomodai; Clade C, S. lanzhouensis; Clade D, S. meridionalis, S. duanensis, and S. grahami; Clade E, Silurus biwaensis Tomoda 1961; Clade F, Silurus aristotelis Garman 1890; Clade G, S. glanis; and Clade H, S. longibarbatus and S. microdorsalis (Figure 2). Both analyses supported S. longibarbatus and S. microdorsalis as sister taxa, occupying the basal position of all remaining Silurus species, with high bootstrap values. The monophyly of S. microdorsalis, S. longibarbatus, S. glanis, S. aristotelis, S. biwaensis, S. grahami, S. duanensis, S. lanzhouensis, and S. meridionalis was highly supported.

Figure 2.

Phylogenetic trees based on 108 Silurus mitogenomes and nine outgroups

A: Maximum-likelihood (ML) tree. B: Bayesian inference tree. Values on branches indicate bootstrap proportions from ML analysis and Bayesian posterior probabilities. Triangles and rectangles indicate two pairs and one pair of mandibular barbels, respectively.

For the S. asotus complex, two major clades (A and B) and three subclades (A1, A2, and A3) within Clade A were observed. With respect to S. microdorsalis, two sublineages (H1 and H2; Figure 2) with relatively high genetic divergence (K2P distance of 3.75%) were recovered. The phylogenetic trees showed that S. soldatovi was clustered with S. asotus within subclade A1 (Figure 2). Furthermore, S. lithophilus and S. tomodai were nested within S. asotus in Clade B and displayed shallow genetic divergence, with K2P distances ranging from 0.15% to 1.11% among S. lithophilus, S. tomodai, and the closest mitogenome of S. asotus.

MOTU delimitation in S. asotus complex and S. microdorsalis

The K2P, ABGD, and PTP analyses indicated that S. asotus and S. microdorsalis could be delimited into four and two MOTUs, respectively (Figure 3A). GMYC analyses yielded congruent delimitations (only single threshold model was presented in Figure 3A), except for the identification of a fifth MOTU within S. asotus (Figure 3A). Using a majority rule consensus, four MOTUs (Sa1 to Sa4) and two MOTUs (Sm1 and Sm2) were delimited in the S. asotus complex and S. microdorsalis, respectively. The average K2P genetic distance between the MOTUs ranged from 2.37% to 6.68% in the S. asotus complex and reached 4.49% for S. microdorsalis (Table 4).

Figure 3.

DNA-based species delimitation and distribution patterns of identified putative species of Silurus asotus and S. microdorsalis

A: DNA-based species delimitation. B: Distribution patterns of identified putative species of S. asotus. C: Distribution patterns of identified putative species of S. microdorsalis. Bars with different colors indicate different valid species. MOTU delimitation schemes collected from ABGD, PTP, and GMYC algorithms, including consensus delimitation scheme. BM, Baekdudaegan Mountains; NL, Nanling Mountains; YTR, Yangtze River; PR, Pearl River.

Table 4. Genetic K2P distances between MOTUs of Silurus asotus and S. microdorsalis.

Taxon 1	Taxon 2	Genetic distance (%)
Sa1	Sa2	2.38
Sa1	Sa3	3.02
Sa2	Sa3	3.64
Sa1	Sa4	6.22
Sa2	Sa4	6.68
Sa3	Sa4	6.30
Sm1	Sm2	4.49

Within the S. asotus complex, MOTU Sa1 showed widespread distribution, covering most rivers in East Asia; MOTU Sa2 was observed in the Pearl River only; MOTU Sa3 was found in the Yangtze River and Pearl River; and MOTU Sa4 was observed in the Yangtze River and Japanese rivers (Figure 3B). As the type locality of S. asotus was not reported in detail in its original description, it was not possible to identify the MOTU in the type locality of the species. The two S. microdorsalis MOTUs were found in Northeast China (Sm1) and South Korea (Sm2) at the species type locality (Figure 3C), showing relatively close spatial distances (Figure 3C).

Divergence time estimates and biogeography

Divergence age estimates suggested that the split between Silurus and closely related genera occurred approximately 37.61 Ma (95% highest posterior density (HPD), 27.93–47.72 Ma; Figure 4). The divergences within the eight clades ranged from 11.56 Ma (95% HPD, 7.23–16.41 Ma) for Clade A to 29.44 Ma for Clade H (95% HPD, 21.64–37.66 Ma). The MOTUs within S. asotus diverged between 3.71 Ma (95% HPD, 1.92–5.88 Ma) and 11.56 Ma (95% HPD, 7.23–16.41 Ma), and the two S. microdorsalis MOTUs diverged approximately 8.74 Ma (95% HPD, 3.23–15.51 Ma).

Figure 4.

Chronogram of selected species derived from divergence time estimations in MCMCTREE and ancestral area reconstructions

Black dots (CP1–4) refer to calibration points. Shaded blue horizontal bars show 95% highest posterior densities (HPD) for divergence times. Pie charts depict ancestral geographic ranges of corresponding nodes inferred with BioGeoBEARS using the BAYAREALIKE model, with different colors representing different biogeographic regions. On the timescale, Cre., Cretaceous; Pal., Paleogene; Eoc., Eocene; Oli., Oligocene; Mio., Miocene; Pli., Pliocene; Ple., Pleistocene.

The best-fit model in BioGeoBEARS was BAYAREALIKE. The common ancestor of all Silurus species was most likely originated from China and the Korean Peninsula (88.2%; Figure 4). The Silurus species spread to Europe and Central and Western Asia from China and the Korean Peninsula between 21.78 Ma and 26.67 Ma (Figure 4). Furthermore, Silurus species dispersed independently eastward to Japan at least three times. In turn, S. asotus recolonized China and the Korean Peninsula approximately 0.45 Ma.

DISCUSSION

Phylogenetic reconstructions

Most previous phylogenetic studies of Silurus have focused on the mitogenomic structures of individual species or inferred phylogenetic positions using a limited number of taxa (Kappas et al., 2016; Kishimoto et al., 2022; Mabuchi et al., 2020; Park et al., 2020; Székvári et al., 2021; Vittas et al., 2011; Wang et al., 2015a, 2015b; Wu et al., 2021; Yang et al., 2019; Zeng et al., 2011), providing only a partial view of the phylogenetic relationships within the genus. In the present study, we established a comprehensive phylogenetic framework for Silurus, utilizing the largest spatial and taxonomic coverage of whole mitogenome sequences to date.

The recovery of Silurus as monophyletic is consistent with previous phylogenetic analyses (Kappas et al., 2016; Park et al., 2020; Schedel et al., 2022; Székvári et al., 2021; Wang et al., 2015a; Wu et al., 2021; Yang et al., 2019). Our study confirmed strong congruence between molecular delimitations and morphology-based taxonomic hypotheses in nine species, i.e., S. microdorsalis, S. longibarbatus, S. glanis, S. aristotelis, S. biwaensis, S. grahami, S. duanensis, S. lanzhouensis, and S. meridionalis, and robustly supported S. microdorsalis and S. longibarbatus as sister taxa located in the basal position of the Silurus phylogeny (Figure 2). The sister relationship between S. microdorsalis and S. longibarbatus was unexpected due to their distant distribution ranges, separated by at least 2 000 km, and remarkable morphological differences, including counts of mandibular barbel pairs and anal-fin rays (Chu et al., 1999; Li et al., 2019).

Surprisingly, S. soldatovi was clustered within the S. asotus subclade A1, exhibiting very shallow genetic distances. This finding contrasts with previous phylogenetic studies (Alam et al., 2019; Mabuchi et al., 2020; Schedel et al., 2022; Yang et al., 2019), which may be attributed to their use of a limited number of taxa and circumscribed sample ranges, thereby overlooking important intraspecific genetic variations in widespread species, thus leading to the classification of S. soldatovi as monophyletic. Our results suggest that S. soldatovi may have recently diverged from the closely related S. asotus, or that S. asotus may comprise a complex of cryptic taxa, given the significant morphological differences between the two species despite their intertwined mitochondrial genealogies (Chu et al., 1999). The most salient morphological distinction is that S. asotus possesses one pair of mandibular barbels, whereas S. soldatovi possesses two (Chu et al., 1999). Furthermore, the phylogenetic trees placed S. lithophilus, S. tomodai, and S. asotus within Clade B, exhibiting shallow divergence. While these species were reciprocally monophyletic, their phylogenetic relationships showed only moderate statistical support. These phylogenetic relationships are consistent with previous findings (Mabuchi et al., 2020; Schedel et al., 2022). Dating analyses indicated that the three species within Clade B diverged approximately 2 Ma, in line with the conclusions reported in Tabata et al. (2016). These results underscore the close relationship among S. lithophilus, S. tomodai, and the lineage known as S. asotus, suggesting that S. asotus is clearly polyphyletic. Future research should incorporate larger population-level sample sizes, nuclear gene analysis, and detailed morphological descriptions to clarify the phylogenetic relationships among these closely related species and unravel the complex evolutionary history of the S. asotus complex.

Biogeographic history

The BioGeoBEARS reconstructions supported the origin of Silurus in China and the Korean Peninsula. This scenario is consistent with the highest species richness observed in these regions, which host 10 of the 16 known species of Silurus (www://fishbase.org/; accessed on 1 January 2023). Our results suggested that Silurus appeared during the late Eocene, coinciding with the Eocene-Oligocene extinction event (Hren et al., 2013; Ivany et al., 2000), which led to the creation of many empty ecological niches, likely facilitating the emergence of Silurus (Hren et al., 2013; Ivany et al., 2000). In addition, the substantial diversification of Silurus occurred during the Miocene, coinciding with the Mid-Miocene Climatic Optimum (MMCO) and intensification of the Asian monsoon (Clift et al., 2008; Favre et al., 2015; Guo et al., 2002; Lu & Guo, 2014). Previous studies have also noted that the intensification of the monsoon was accompanied by intense diversification in the East Asian fish fauna (Cheng, 2015; Clift & Plumb, 2008; Feng et al., 2023).

We inferred three major colonization events in the genus Silurus from the late Oligocene to the Miocene (Figure 4). Two dispersal events occurred, one from China and the Korean Peninsula to Europe and Central and Western Asia, and another from China and the Korean Peninsula to Japan between the late Oligocene and early Miocene. This period coincides with the onset of the East Asian monsoon climate, when heavy rainfall induced by the summer monsoon led to increased continental erosion rates, potentially accelerating dispersal between China and the Korean Peninsula, and Europe and Central and Western Asia (Colin et al., 2010; Liu et al., 2012). Moreover, the increased rainfall may have enhanced fluvial discharge into the marginal marine environment, creating dispersal pathways between the Eurasian continent and islands such as Japan (Colin et al., 2010; Liu et al., 2012). A subsequent dispersal event from China and the Korean peninsula to Japan occurred approximately 11.19 Ma (Late Miocene) (Figure 4), likely driven by periodic connections of rivers between the Asian mainland and Japan, induced by recurrent marine transgression and regression in East Asia (Kocsis & Scotese, 2021; Scotese & Wright, 2018). Feng et al. (2023) reported similar findings, examining the roles of marine transgression and regression in East Asia on the dispersal of endemic East Asian carp from the Asian mainland to Japan.

Two dispersal events between the Asian mainland and Japan during the Pleistocene were also identified. Initially, S. asotus in Japan (Sa4) recolonized the Asian mainland during the Late Pleistocene, although this conclusion is somewhat tentative due to the inclusion of only one mitogenome. Subsequently, a third colonization from the Asian mainland to Japan occurred during the early Pleistocene. Various studies have demonstrated that East Asia underwent multiple large-scale glaciation events during the Pleistocene, leading to periodic fluctuations in sea levels (von Wissmann, 1937; Voris, 2000; Zachos et al., 2001). Similar patterns of dispersion are evident among freshwater fish taxa between the Asian mainland and Japan (Tsao et al., 2016) and/or other islands such as Taiwan and Hainan (Chen et al., 2020; Gao et al., 2023; Wang et al., 2000; Yang et al., 2012b).

We argue that paleoecological events, including the onset and intensification of the monsoon system, recurrent marine transgression and regression, and glacial cycles have significantly influenced the biogeographic history of the genus Silurus. However, given that S. triostegus in Europe and Central and Western Asia was not included in the current study, some biogeographic events may have been overlooked.

Phylogeographic patterns within S. asotus and S. microdorsalis

Many cryptic species with identical morphological traits have been recognized using molecular tools (Bickford et al., 2007). In our study, species delimitation analyses revealed four (Sa1 to Sa4) and two (Sm1 and Sm2) MOTUs in S. asotus and S. microdorsalis, respectively, characterized by substantial genetic divergence (K2P distance >2%; Table 1). In this context, S. asotus was not monophyletic as Sa4 was closely related to S. tomodai and S. lithophilus (Figure 2), suggesting that the taxonomic characters used to separate S. asotus from other closely related species are varyingly conserved among species and represent a mosaic of converging and/or ancestral character states. These results highlight the need to re-examine the utility of these morphological characters in taxonomy considering our current phylogenetic reconstruction.

The four observed MOTUs in S. asotus and their overlapping distributions suggest that the S. asotus complex has experienced a complex evolutionary history. Xu et al. (2017) also reported overlapping distributions for members of the four observed MOTUs using mitochondrial cyt b fragments. MOTU Sa4 is believed to have initially emerged in the Japanese archipelago during the Late Miocene, isolated by the Korean Strait, and then dispersed to the Asian mainland during the Late Pleistocene due to falling sea levels (von Wissmann, 1937; Voris, 2000; Zachos et al., 2001). MOTU Sa2 and Sa3 are thought to be the byproducts of the rapid uplift of the Qinghai-Xizang Plateau from the Late Miocene (8.0 Ma) to Late Pleistocene (Molnar, 2005). The rapid rise of the plateau significantly influenced the regional climate and reshaped the landscape features of eastern Asia, including the drainage systems (Jia et al., 2003). This geological change is hypothesized to be a crucial driving force in the formation and colonization of many fish taxa in East Asia (Guo et al., 2005; He & Chen, 2006; He et al., 2001; Yang et al., 2012a). Furthermore, MOTU Sa2 underwent local divergence in the Pearl River and was unable to disperse northward due to the Nanling Mountains barrier (Figure 5A). The most recently diverged MOTU, Sa1, has become the most widely distributed species in the East Asian ranges, which may be due to recent population expansion after the Late Pleistocene glacial retreat (Gao et al., 2012) and/or frequent summer flooding (Chen et al., 2017).

Figure 5.

Dispersal routes of Silurus inferred from biogeographic analyses at different time scales

A: 25–20 Ma. B: 20–15 Ma. C: 15–10 Ma. D: 4–0 Ma. A, B, and C on each map represent three recognized biogeographical regions, i.e., China and Korean Peninsula, Europe and Central and Western Asia, and Japan, respectively.

Regarding S. microdorsalis, the discovery of two MOTUs exhibiting deep divergence was unexpected, given the restricted distribution of the species and the relatively close distributions of the two MOTUs (<800 km). Divergence dating suggested that the two MOTUs split during the late Miocene, implying a period of long-term isolation. The Baekdudaegan Mountain range appears to have acted as a biogeographic barrier, facilitating the separation of the two MOTUs. Previous reports have confirmed that the Baekdudaegan Mountain range has played an important role in the divergence and/or speciation of multiple freshwater fish species in the Korean Peninsula (Jeon et al., 2022; Kim et al., 2012).

Given that the formal taxonomy of DNA-based cryptic lineages often lacks stability (Satler et al., 2013), the MOTUs identified herein merely conform to the primary species hypotheses. Additional morphological and phylogenetic evidence is required to clarify the taxonomic status of each MOTU under the criteria of secondary species hypotheses. Furthermore, the species diversity within the genus Silurus has likely been seriously underestimated, necessitating further investigation with expanded sample ranges and sizes.

CONCLUSIONS

Our study employed a comprehensive phylogenetic framework and biogeographic history to examine cryptic diversity within the genus Silurus. Phylogenetic trees resolved eight main clades and indicated that S. soldatovi, S. lithophilus, and S. tomodai were not monophyletic, as they clustered with different lineages of S. asotus. Biogeographic reconstruction identified China and the Korean Peninsula as the regions of origin, with multiple dispersal events triggered by the Eocene-Oligocene extinction event, onset and intensification of the monsoon system, recurrent marine transgression and regression, and glacial cycles. In addition, DNA-based delimitation identified four and two MOTUs in S. asotus and S. microdorsalis, respectively, although further research is required to clarify their taxonomic status. These findings not only improve our understanding of the phylogenetic relationships and evolutionary history of Silurus species but also pave the way for future ichthyological research on this group.

SUPPLEMENTARY DATA

Supplementary data to this article can be found online.

Funding Statement

This study was supported by the National Natural Science Foundation of China (32000306), Project of Innovation Team of Survey and Assessment of the Pearl River Fishery Resources (2023TD-10), and Natural Science Foundation of Shaanxi Province (2023-JC-YB-325)