Specimens were euthanized by sub-merging them in a buffered solution of tricaine methanesulphonate (MS-222) at a concentration of 250 mg/L, for a period of 10 min, following the guidelines of the Journal of the American Veterinary Medical Association (AVMA Guidelines) (Leary et al. 2013) and European Commission DGXI consensus for fish euthanasia (Close et al. 1996, 1997). Molecular data were obtained from specimens fixed and preserved in absolute ethanol. Specimens used for morphological comparisons were fixed in formalin for a period of 14 days and then transferred to 70% ethanol. All the collected material was deposited in the fish collection of Institute of Biology, Federal University of Rio de Janeiro (UFRJ).

Total genomic DNA was extracted from muscle tissues using the DNeasy Blood & Tissue Kit (Qiagen), according to the manufacturer’s protocol. The analyses included a set of partial sequences of four nuclear genes: recombination activating 2 (RAG2), myosin heavy chain 6 (MYH6), SH3 and PX domain containing 3 (SH3PX3), glycosyltransferase (GLYT); and partial sequences of two mitochondrial encoded genes: cytochrome c oxidase I (COI) and cytochrome b (CYTB). Amplification of the target DNA fragments was made through the polymerase chain reaction (PCR) method, using the following primers: Cytb Siluri F, Cytb Siluri R, L5698-ASN, H7271-COI (Villa-Verde et al. 2012), Glyt_F577, Glyt_F559, Glyt_R1562, Glyt_R1464, myh6_F459, myh6_F507, myh6_R1322, myh6_R1325, SH3PX3_F461, SH3PX3_R1303, SH3PX3_F532, SH3PX3_R1299 (Li et al. 2007); RAG2 MCF, RAG2 MCR (Cramer et al. 2011), MHRAG2-F1 and MHRAG2-R1 (Hardman & Page, 2003). Double-stranded PCR amplifications were performed in 60 μl reactions with reagents at the following concentrations: 5× GreenGoTaq Reaction Buffer (Promega), 3.2 mm MgCl2, 1 μm of each primer, 75 ng of total genomic DNA, 0.2 mm of each dNTP and 1 U of standard Taq polymerase or Promega GoTaq Hot Start polymerase. The thermocycling profile was as follows: initial denaturation for 2 min at 94–95 °C; 35 cycles of denaturation for 1 min at 94 °C, annealing for 1 min–90 s at 48.0–64.0 °C and extension for 1–2 min at 72 °C; and terminal extension for 4 min at 72 °C. Negative controls were used to check on contaminations. The PCR products were then purified using the Wizard SV Gel and PCR Clean-Up System (Promega). Sequencing reactions were made using the BigDye Terminator Cycle Sequencing Mix (Applied Biosystems). Cycle sequencing reactions were performed in 20 μl reaction volumes containing 4 μl BigDye, 2 μl sequencing buffer 5× (Applied Biosystems), 2 μl of the amplified products (10 -40 ng), 2 μl primer and 10 μl deionized water. The thermocycling profile was: (1) 35 cycles of 10 seconds at 96 °C, 5 seconds at 54 °C and 4 minutes at 60 °C.

In-group included seven subfamilies of Trichomycteridae lineages. Since this study was directed to searching for relationships and monophyly of Trichomycterus, the analysis included a total of 50 terminal taxa presently placed in this genus, as well as 21 terminal taxa belonging to the trichomycterine genera Bullockia, Eremophilus, Ituglanis and Scleronema. Out-group selection was directed to sample representatives of other lineages of the Trichomycteridae, comprising six terminal taxa representing each of the remaining subfamilies, as well as representatives of other families of the Loricarioidea, including the nematogenyid Nematogenys inermis Guichenot, 1848 and the loricariid Pareiorhina rudolphi (Miranda-Ribeiro, 1911), in which the analyses were rooted. A list of specimens and its respective GenBank accession numbers is provided in Suppl. material 1. Sequences were aligned and edited in Mega 7 software (Kumar et al. 2016) using the ClustalW algorithm (Chenna et al. 2003). Gaps were considered as informative characters. The data set was partitioned by gene, and the best-fit evolutionary model selection was performed under the Akaike information criteria in the software jmodeltest 2 (Darriba et al. 2012) for each partition. Bayesian Inference (BI) and Maximum Likelihood (ML) analyses were conducted using the softwares Mrbayes 3.2 (Ronquist et al. 2012) and Garli 2.0 (Zwickl 2006), respectively. The BI analysis was conducted with the following parameters: two independent Markov chain Monte Carlo (MCMC) runs of two chains each for 20 million generations, with a tree sampling frequency of every 100 generations. The convergence of the MCMC chains and the proper burn-in value were assessed by evaluating the stationary phase of the chains using tracer v. 1.5 (Rambaut et al. 2013). The BI final consensus tree and its Bayesian posterior probabilities were generated with the remaining tree samples after removing the first 25% samples as burn-in. To test the support of the nodes in the ML analysis, 1,000 bootstrap (Felsenstein 1985) replicates were made.

Morphological comparisons were focused on the external morphology and osteological features of cleared and stained specimens prepared according to Taylor and Van Dyke (1985). Terminology for osteological nomenclature follows Arratia (1998). Measurements and counts follow Barbosa and Costa (2003). A list of material analysed appears in Suppl. material 2.

The concatenated matrix comprised 4380 bp after alignment (522 bp for COI, 857 for CYTB, 890 for GLYT, 542 for MYH6, 884 for RAG2 and 680 for SH3PX3). The best-fit evolutionary models found are shown in Suppl. material 3. The best log-likelihood score for the ML analysis was -31473.209522 and the mean for BI was -lnl -32013.626.

Both analyses exhibit similar topologies (Figs 1, 2). In all analyses, monophyly of the Trichomycterinae was highly supported, but nominal species of Trichomycterus did not form a single monophyletic group. Two major clades were found in Trichomycterinae. The first one, the trichomycterine clade a, is highly supported in both analyses (Figs 1, 2) and comprises species geographically restricted to an area encompassing eastern, south-eastern and southern Brazil, which have been traditionally placed in two genera, Scleronema and Trichomycterus. Both analyses highly corroborate three major subclades (Figs 1, 2), one comprising T. nigricans, the type species of Trichomycterus, and other congeners, therefore herein recognised as the true Trichomycterus; another corresponding to the genus Scleronema; and another clade, sister to Scleronema and comprising several nominal species of Trichomycterus, herein recognised as a new genus (see taxonomic accounts below).

Figure 1. 

Phylogenetic positioning of Cambeva among the Trichomycteridae, inferred by Bayesian Inference from the analysis of molecular data, total of 4380 bp comprising segments of nuclear genes for GLYT, MYH6, RAG2 and SH3PX3 and the mitochondrial genes COI and CYTB. Numbers on each node represent posterior probabilities.

The second trichomycterinae group, the clade b, is weakly supported in the BI analysis (Fig. 1), but not supported in the ML analysis (Fig. 2). It includes taxa endemic to a broad geographical area, including the Andean region, Amazon, and Patagonia, formally placed in the genera Bullockia, Eremophilus, Ituglanis and Trichomycterus, which is herein graphed “Trichomycterus” by being distantly related to that clade including the type species of the genus. Relationships among basal lineages of the clade b were weakly supported (BS < 50; BI < 95).

Figure 2. 

Phylogenetic positioning of Cambeva among the Trichomycteridae, inferred by Maximum Likehood from the analysis of molecular data, total of 4380 bp comprising segments of nuclear genes for GLYT, MYH6, RAG2 and SH3PX3 and the mitochondrial genes COI and CYTB. Numbers on each node are bootstrap percentages of the Maximum Likelihood analysis; asterisks indicate maximum support value and hyphen values under 50.

Since the first objective phylogenetic analysis of the Trichomycteridae by Baskin (1973), monophyly of the Trichomycterinae has been challenged by several authors. Besides not finding unique diagnostic features for the subfamily, supposed derived traits shared by some trichomycterines and species of other trichomycterid subfamilies made uncertain the position of some trichomycterine members. de Pinna (1989) suggested that the trichomycterine genus Scleronema is more closely related to the subfamily Sarcoglanidinae than to other trichomycterines, as already discussed by Myers & Weitzman (1966), due to the common possession of an enlarged maxilla; Arratia (1990) also reported a maxilla enlargement for the trichomycterine genera Bullockia and Hatcheria, besides species of ‘Trichomycterus’. Our results clearly indicate that taxa having long maxilla (e.g., sarcoglanidines, Bullockia, Scleronema) do not form a monophyletic group. Considering the psammophilic habits exhibited by species of Sarcoglaninidae and Scleronema, with specimens digging into sand or gravel (Zuanon and Sazima 2004, Schaefer et al. 2005, AM Katz and WJEM Costa pers. obs.), the long maxilla exhibited by these taxa may be interpreted as an adaptive convergence generating homoplasies in both groups.

Costa and Bockmann (1993) considered Ituglanis and Scleronema more closely related to the TSVSG clade, comprising the Tridentinae, Sarcoglanidinae, Vandelliinae, Stegophilinae and Glanapteryginae (Costa and Bockmann 1994), than to other trichomycterids based on a thin tip of the lateral process of the urohyal and a slightly reduced interopercular patch of odontodes. However, as noted by subsequent authors (e.g. Fernández and de Pinna 2005; Datovo and Bockmann 2010; the present study) other trichomycterines also exhibit in some extent those character states. More recently, Datovo and Bockmann (2010), in their myological analysis of trichomycterids, found evidence indicating that Ituglanis and Scleronema are more closely related to other Trichomycterinae than to the TSVSG clade. A similar result has been proposed in molecular phylogenies (e.g. Henschel et al. 2017, Ochoa et al. 2017). In our multigene analysis, using a broader genetic sample, the phylogenetic position of Ituglanis and Scleronema within a monophyletic Trichomycterinae is highly corroborated (Figs 1, 2). Ituglanis, with species distributed in the main tropical and subtropical Cis-Andean river basins of South America, did not appear as closely related to Scleronema, but sister to a clade comprising Andean, Patagonian and Amazon species of ‘Trichomycterus’ (Figs 1, 2).