Phylogenetic Relationships of Five Asian Schilbid Genera Including Clupisoma (Siluriformes: Schilbeidae)

Ethics

All the samples of fishes were bought from local fish dealers in Manzha Market in Menglun Town, Mengla County, Yunnan province, China. (21°56′07.30″N,101°14′56.54″E; elevation: 546m). As food fishes, no permits were required for sampling. All the samples were living in the natural body of water. The housing and husbandry conditions were unclear and all fishes were dead when obtained. Specimens were preserved using 70% ethanol in the Laboratory for Conservation and Utilization of Bio-resources, Yunnan University. All procedures followed corresponding regulations and by-laws and were approved by the Ethics and Experimental Animal Committee of Kunming Institute of Zoology, Chinese Academy of Science, China (KIZ_YP201002).

Sampling and outgroup selection

Seventeen individuals belonging to eight species of six catfish families were sampled (Table 1). Twenty-eight additional sequences from 28 species of 23 genera in 13 catfish families were downloaded from GenBank (Table 1). We used two species each from the Cypriniformes, Clupeiformes and Characiformes as outgroup taxa.

DNA Extraction, PCR and Sequencing

Primers were either designed based on sequences of Pangasiidae retrieved from GenBank by using Primer Premier 5.0 software (Premier Biosoft International), or they were adapted from literature (Table 2). Genomic DNA was isolated from tissue samples by standard phenol/chloroform extraction. PCR were performed in a 30μl reaction mixture containing 20–50 ng templates DNA, 1.2μM dNTP, 0.5μM of the forward and reverse primers, 0.15 units of EX-Taq DNA polymerase enzymes (TaKaRa) and 3 μl of 10× EX-Taq buffer. The amplification reaction was performed using 33 cycles of 30sec at 95°C, annealing at 66 to 55°C for 30sec, and extension of 72°C for 90sec, with an initial step of 4min at 95°C and a final step of 7min at 72°C. PCR products were purified on agarose gels and extracted (Watson BioMedical Inc. Shanghai) and sequenced with a BigDye DNA sequencing kit (ABI) on a 3730XL sequencer (ABI). The sequences were deposited in GenBank (accession numbers listed in Table 1).

Sequence analysis

De novo sequences were checked using BLAST [20] against the NCBI database (http://www.ncbi.nlm.nih.gov) to assess sequence similarity. They were aligned using ClustalX 1.83 and manually verified. DAMBE 4.1.19 [16] was used to identify unique haplotypes.

Phylogeny construction

Phylogenies were constructed using maximum likelihood (ML) via RAxML [17], Bayesian inference (BI) executed with MrBaves 3.2 [18], and maximum parsimony (MP) implemented in PAUP* 4.0b10 [19]. We selected the best-fitting models for ML and BI using the Akaike Information Criterion (AIC) [20, 21] as implemented in jModelTest 0.1.1 [22, 23]. BI analysis used four independent MCMC chains run simultaneously for 5 million generations while sampling one tree per 500 replicates, Burnin = 0, and Burninfrac = 0.10, 0.20, 0.30, 0.40, and 0.50. Two runs were conducted independently and the sampled trees were used to construct a 50% majority rule consensus tree after discarding the first 10% as burnin. Bayesian posterior probabilities (BPP), the frequencies of nodal resolution, were mapped on the BI tree. For MP and ML, nodal support was assessed using nonparametric bootstrap sampling [24] of 1000 pseudoreplicates.

Testing tree incongruence

The incongruence among different tree topologies was evaluated using the Approximately Unbiased (AU) test [25], as implemented in the CONSELV0.1i with default scaling and replicate values [26]. Site-wise log-likelihood values were estimated by PAUP*.

MtDNA

The concatenated mtDNA dataset comprised 2300 aligned sites: 626 from the COI fragment, 1137 from cytb, and 537 from the 16S rRNA fragment. The genes consisted of 41 unique haplotypes for 43 sequences of COI, 43 unique haplotypes among 46 sequences of cytb, and 33 unique haplotypes among 36 sequences of 16S rRNA. The combined alignment comprised 2300 positions, of which 941 (40.9%) were potentially parsimony-informative (Table 3).

Individual mtDNA gene analyses produced inconsistent topologies with low levels of nodal support, probably due to limited information harbored in a single gene. The trees constructed by analyses of the concatenated data using ML, MP and BI (Fig 1) were consistent for well supported nodes. The five Asian schilbid genera formed two monophyletic groups, one consisting of Clupisoma, Lades and Ailia (BI BPP = 1.0, ML BS = 100% and MP BS = 96%) and the other comprising Horabagrus and Pseudeutropius (BI BPP = 0.99, ML BS = 95% and MP BS = 56%). Clupisoma formed the sister taxon of Laides (BI BPP = 1.0, ML BS = 100% and MP BS = 100%). Excluding the MP tree, the two Asian schilbid groups rooted within the Bagridae. The superfamily Sisoridae, excluding the Aspredinidae, constituted a lineage referred to as “Big Asia” by Sullivan et al. (2006, 2008) [8, 9]. Within “Big Asia”, (Ailia (Laides, Clupisoma)) was the sister-group of the Sisoroidea (BI = 92%), while (Horabagrus, Pseudeutropius) was the sister taxon of the Bagridae (BI = 98%) (Fig 1).

NuDNA

The combined alignment of the nuclear genes RAG1 and RAG2 contained 3750 positions: 1430 from the RAG1 exon 1, 2), 1375 from the RAG1 exon 3, and 945 from RAG2 (Table 3). Among these, 1844 sites (49.3%) were potentially parsimony-informative (Table 3). The tree (Fig 2) displayed likelihood bootstrap proportions, parsimony bootstrap proportions and Bayesian posterior probabilities (BPP).

As with the mt-genes tree, the five Asian schilbid genera also showed the strongly supported monophyletic groups (Ailia (Laides, Clupisoma)) (BI BPP = 1.0, ML BS = 100% and MP BS = 100%) and (Horabagrus, Pseutropius) (BI BPP = 1.0, ML BS = 100% and MP = 99%). However, analyses of the nuDNA data consistently united them as sister taxa (BI = 94%, ML = 67% and MP = 66%) and rooted them in “Big Asia” with strong support (BI BPP = 1.0, ML BS = 100% and MP = 95%). Relationships among this group, the Bagridae, and the superfamily Sisoroidea were not well resolved.

Concatenated MtDNA and NuDNA

For a total evidence analysis, we have combined three mtDNA genes (COI, 16s and cytb) and two nuclear genes (RAG1 and RAG2). The three mtDNA fragments comprised 2300 aligned sites: 626 from the COI fragment, 537 from the 16S fragment, and 1137 from cytb; and the nuclear dataset consists of 3750 aligned bases: 1430 from the RAG1 (exon 1, 2) fragment, 1375 from the RAG1 (exon 3) fragment and 945 from RAG2 (Table 3). The concatenated datasets were comprised of six fragments including 6050 aligned sites.

The obtained nuDNA trees for the analyzed five Asian schilbid genera (Fig 3; ML and MP trees not shown) were somewhat similar to those of the mt genes trees. Analyses of both genomes resolved two strongly supported monophyletic clades: “Big Asia”, i.e., (Ailia (Laides, Clupisoma)) (BI = 100%, ML = 100% and MP = 100%) and (Horabagrus, Pseudeutropius) (BI = 100%, ML = 100% and MP = 100%). The genomes differed in that the clade (Ailia (Laides, Clupisoma)) did not associate with other taxa in former “Big Asia”. Further, (Horabagrus, Pseudeutropius) had a weakly supported relationship with the family Bagridae.

AU test

The AU test (Table 4) detected significant differences between the mtDNA and nuDNA datasets (P<0.05). Thus, the matrilineal history differed from that of biparental inheritance. We believe this result precluded combining the data sets for phylogenetic analysis inference because each genome had an independent history. However, we retained the result for readers who might be interested in concatenated data results.

Expanded dataset of Sullivan et al. (2006) [8]

To verify the results from the combined nuDNA dataset, we downloaded the RAG1 and RAG2 sequences of Sullivan et al. (2006) [8] from Siluriformes, to which we added our de novo sequences (Table 1). We reconstructed the ML, MP and BI trees (Fig 4A and 4B). The five Asian schilbid genera remained a monophyletic group with relationship within “Big Asia” shown as ((Aailia (Laides, Clupisoma)), (Horabagrus, Pseudeutropius)). However, this arrangement did not enjoy strong support (BI BPP = 0.88, ML BS = 59%, MP BS = 37%). At higher levels within “Big Asia,” the relationships among the genera in the Bagridae, and the superfamily Sisoroidea were poorly resolved.

Analyses of the expanded dataset further resolved relationships within Siluroidei sensu Sullivan et al. (2006) [8]. Their 13 strongly supported monophyletic linages (thick branches in Fig 4A) were recovered along with the further clustering of these groups into major clades(A, B, C and D (Fig 4B). Sullivan et al. (2006) [8] did not obtain interrelationships among their 13 lineages.

Phylogeny of Asian schilbid genera

Our analyses consistently support both African and Asian schilbids as monophyletic groups, and show that they are distantly related to one another. Thus, we confirm the non-monophyly of the Schilbeidae as recognized by Peng et al. (2005) [6], Hardman (2005) [7] and Sullivan et al. (2006, 2008) [8, 9].

Recognition of the groups (Ailia (Laides, Clupisoma)) and (Horabagrus, Pseudeutropius) foes not support the monophyly of the so-called “Big Asia” (Figs 1–4) as proposed by Sullivan et al. (2006, 2008) [8, 9]. Analysis of the combined mt gene data and the combined nuclear gene data suggest different suites of relationships among the two groups and other taxa. In the former analysis (Fig 1), the group (Ailia (Laides, Clupisoma)) appears as the sister taxon of the Sisoroidei, and the group (Horabagrus, Pseudeutropius) is the sister taxon of the Bagridae. In contrast, analyses of the combined nuclear data unite the two groups as sister subgroups (Fig 2). Analyses of the expanded dataset of Sullivan et al. (2006) [8] supports this relationship (Fig 4). Because AU testing does not reject either genomic tree, the two results may be equally reliable

Morphological and molecular phylogenetic studies of subsets of the Asian Shilbeidae have been undertaken by Mo (1991) [3], De Pinna (1993) [4], Diogo et al. (2004) [5], Peng et al. (2005) [6], Hardman (2005) [7] and Sullivan et al. (2006, 2008) [8], resulting in differing hypotheses of the relationships among these fishes. This might be in part an artifact of sampling, in particular, the absence of critical taxa. Our study is the first to detail the phylogenetic relationships for all nine recognized genera of Asian schilbids.

In a morphological study, Mo (1991) [3] concluded that the Asian schilbids including Clupisoma comprised two distinct groups: Ailia and the genera Horabagrus, Pseudeutropius and Platytropius. Our results from mtDNA analyses somewhat supports their result by Mo (1991) [3] did not clearly comment on the relationships of Clupisoma or specify which species of Platytropius were examined. He claimed Ailia was associated with the Clariidae and Heteropneustidae while Horabagrus, Pseudeutropius and Platytropius were closer to the Bagridae and Pangasiidae, which differs from our results. We did not have access to De Pinna’s (1993) [4] unpublished dissertation. Thus, we do not know if he examined Clupisoma. Researchers citing his dissertation state that he assigned Horabagrus to its own family because it was distinct from both the Schilbeidae and Bagridae [8]. Further, De Pinna (1993) [4] proposed that all schilbids (including African species) constituted a monophyletic group with the subgroup (Schilbinae (Ailiinae, Laides) being closer to the Pangasiidae than to the Shibeidae (see Fig 2 of Hardman, 2005) [7]. In contrast to our findings, and using a less complete set of Asian schilbids than included in the present study, De Pinna concluded that the Shilbeidae was monophyletic. Diogo et al. (2004) [5] examined Asian Ailia, Laides and Pseudeutropius, and African Schilbe and Siluranodon, and similar to De Pinna obtained results that differed from ours, concluding that the Schilbeidae exclusive of Horabagrus was monophyletic and its sister-group was the Pangasiidae. Unlike Pinna (1993) [4], Diogo et al. (2004) [4] did not propose intergeneric relationships among Ailia, Clupisoma, Horabagrus, Laides and Platytropius.

The molecular phylogenetic studies of Peng et al. (2005) [6] failed to resolve the relationships of Asian schilbids because they sampled Asian Clupisoma only, although they suggested that Chinese schilbids might be closest to either the Bagridae or Siluridae. Hardman (2005) [7] resolved the relationships as (Pseudeutropius (Horabagrus, Clupisoma)) and assigned these genera to the Horabagridae created by De Pinna. However, owing to absence of Ailia and Laides, his study failed to provide an overall phylogenetic scenario of the five genera of Asian schilbids. Further, his resolution of the relationships of Clupisoma differed from ours.

Sullivan et al. (2006, 2008) [8, 9] clustered Ailia with Ladies, and Horabagrus with Pseudeutropius with strong support. Both groups belonged to “Big Asia.” The group (Ailia, Laides) was weakly placed as the sister taxon of the Sisoroidea and the group (Horabagrus, Pseudeutropius) was weakly supported as the sister taxon of Bagridae in their MP and ML trees. Thus, their results are similar to ours based on mtDNA analyses. They could not place Clupisoma owing to its absence in their analyses.

In summary, we propose that 1) the group (Ailia (Laides, Clupisoma)) is monophyletic and 2) its sister-group, based on nuDNA analyses, appears to be (Horabagrus, Pseudeutropius), although this hypothesis conflicts with the matrilineal genealogy based on mtDNA data. Our work specifies the phylogenetic position of Clupisoma, which heretofore was ambiguous, and our hypothesis differs from that of Hardman, which Sullivan et al. (2006, 2008) [8, 9] assumed to be true.

Tree sensitivity

Many factors affect the topologies of phylogenetic trees, including choice of outgroup, ingroup representation, the evolution of genes, long-branch attraction (LBA), and method of tree construction [27]. Two of these factors considerably affect the topologies of the trees for catfishes: choice of genome and taxonomic representation. Phylogenetic relationships based on the mtDNA and nuDNA differ significantly, a discovery termed cytonuclear discordance [28]. The resulting trees differ not only among the members of “Big Asia” but also among other catfishes (Figs 1 and 2). The conflict is not unusual [29, 30]. Our results reinforce the hypothesis that nuclear and mt genes may have different evolutionary trajectories.

The density of ingroup sampling also affects trees. The addition of 17 ingroup sequences (Table 1) to the dataset of Sullivan et al. (2006) [8] changes the topology of the tree greatly. It further resolves the relationships among the 13 lineages comprising the suborder Sisoroidei (Fig 4). Saitoh et al. (2006) [31], Wang et al. (2007) [32], Li et al. (2008) [33], Yang et al. (2010) [34], Telford and Copley (2011) [27] and Wang et al. [35] emphasized the importance of increasing the density of ingroup sampling. The present study provides support for this approach.

Taxonomic implications

Taxonomy should reflect historical relationships [36]. Based on his own analyses and those of Mo (1991) [3], Hardman (2005) [7] recognized the Horabagridae of De Pinna (1993) [4] as containing the genera Horabagrus, Pseudeutropius and Clupisoma. Sullivan et al. (2006) [8] followed this assignment. Our results support the recognition of the Horabagridae vis-à-vis Asian taxa, but with the exclusion of Clupisoma. The Horabagridae De Pinna (1993) [4] contains Horabagrus and Pseudeutropius only. We note that sometimes Horabagrus has been assigned to the Bagridae [3].

Recognition of the Horabagridae renders the Schilbeidae a polyphyletic family. The type genus of Schilbeidae, Schilbe, is native to Africa. Because African schilbids are not the sister group of Asian genera [3], and to obtain a taxonomy that reflects the phyletic history of these Asian catfishes, we formally erect a new family Ailiidae fam. nov. (type genus Ailia) for monophyletic Asian group comprised of the genera Ailia, Laides and Clupisoma. This results in recognition of the following taxonomy for these catfishes:

Class Actinopterygii

Order Siluriformes

Suborder Sisoroidei

Family Horabagridae: Horabagrus (Asia), Pseudeutropius (Asia)

Family Ailiidae fam. nov.: Ailia (Asia), Laides (Asia), Clupisoma (Asia)

Family Schilbeidae: Schilbe (Africa), Irvineia (Africa), Pareutropuis (Africa), Parailia (Africa), Siluranodon (Africa), Platytropius (Asia), Eutropiichthys (Asia),Neotropius (Asia), Proeutropiichthys (Asia), Silonia (Asia) [3]

We do not have any specimens of Horabagridae or Ailiidae, we obtained the morphological information of seven species within these two lineages from FishBase (http://www.fishbase.org/search.php?lang=English). Unfortunately, only one morphological trait was available for all seven species. The total numbers of soft rays of anal fin in Horabagridae ranged from 31 to 33, while the ones in Ailiidae ranged from 39 to 55 (S1 Table). These data are congruent with our hypothesis of a new family of Ailiidae. In addition, images displayed on the website show differences in body shape: the abdominal line of Horabagridae tends to be flat, while those of the Ailiidae curve. These data also show divergence between these two lineages (S1 Fig). The morphological differences correspond with the molecular evidence for a new family.

Undoubtedly, morphological evidence is crucial to propose a new family from within an established family. We encourage the acquisition of deeper morphology evidence or other disciplines to further test our hypothesis of the Ailiidae.

Accession Numbers

All the sequences by this study have been submitted to GenBank. The accession numbers together with the downloaded data were listed in Table 1.