Field procedures were approved by CEUA-UFRJ (Ethics Committee for Animal Use of the Federal University of Rio de Janeiro; permit numbers: 065/18 and 084/23) and collecting permits were given by ICMBio (Instituto Chico Mendes de Conservação da Biodiversidade; permit number: 38553-11). For details about specimen euthanasia, fixation, and conservation, see Costa et al. (2023b). Specimens were deposited in the ichthyological collection of the Instituto de Biologia, Universidade Federal do Rio de Janeiro (UFRJ). Comparative material is listed in Barbosa and Costa (2008), Costa and Katz (2021), and Costa et al. (2023b). A complete list of specimens used in morphological comparisons, with their respective catalogue numbers and locality coordinates, appears in Suppl. material 1.

Methods for taking and describing morphological characters were according to recent studies on Paracambeva, including morphometric and meristic data following Costa (1992) and Costa et al. (2020b), osteological preparations following Taylor and Van Dyke’s (1985), latero-sensory system nomenclature following Arratia and Huaquin (1995) and Bockmann and Sazima (2004), bone terminology following Costa (2021) and Kubicek (2022), and fin ray formulae following Bockmann and Sazima (2004).

Methods for DNA extraction, amplification, and sequencing followed the most recent phylogenetic analysis of Paracambeva (Costa et al. 2023b), with PCR reactions performed in 45 μl with the following reagent concentrations: 5× GreenGoTaq Reaction Buffer (Promega), 1.0 mM MgCl₂, 1 μM of each primer, 0.2 mM of each dNTP, 1 u of Promega GoTaq Hot Start polymerase, and 50 ng of total genomic DNA; thermal profile: 95 °C for 5 min; 35 cycles of 94 °C for 1 min; 50–60 °C for 1–1.5 min; and 73 °C for 7 min; sequencing reactions performed in 20 μL reaction volumes containing 4 μL BigDye, 2 μL sequencing buffer 5× (Applied Biosystems), 2 μL of the PCR products (30–40 ng), 2 μL primer, and 10 μL ultrapure water; and the thermal profile was 35 cycles of 30 s at 95 °C, 30 s at 55 °C, and 1.5 min at 73 °C. Primers used for mitochondrially encoded genes were: Cytb Siluri F and Cytb Siluri R (Villa-Verde et al. 2012) for cytochrome b (CYTB); FISHF1 and FISHR1 (Ward et al. 2005) for cytochrome c oxidase I (COX1). Primers for the nuclear encoded genes were: MYH6 TRICHO F and MYH6 TRICHO R (Costa et al. 2020c) for myosin heavy chain 6 (MYH6), and RAG2 TRICHO F and RAG2 TRICHO R (Costa et al. 2020c) for recombination activating 2 (RAG2). MEGA 11 (Tamura et al. 2021) was used to read and interpret sequencing chromatograms, to perform the sequence annotation, and to translate DNA sequences into amino acid residues to verify the absence of premature stop codons or indels. GenBank accession numbers are in Table 1.

The terminal taxa for the phylogenetic analyses include all species of Paracambeva, including the new species herein described, and all species of the subgenus Trichomycterus. The remaining species of Trichomycterus and outgroups included in the analysis are the same as those used in Costa et al. (2023b). See Costa et al. (2023b) for justification for outgroup selection. Each gene dataset was aligned using the Clustal W algorithm (Chenna et al. 2003) implemented in MEGA 11 and analysed for determination of the optimal partitioning and evolutionary models (Table 2) using the Corrected Akaike Information (AICc) in PartitionFinder 2.1.1 (Lanfear et al. 2016). Phylogenetic analyses followed the methods described in Costa et al. 2023b), comprising Bayesian Inference performed with Beast 1.10.4 (Suchard et al. 2018), using the Yule process as the tree prior (Gernhard, 2008), two independent Markov chain Monte Carlo (MCMC) runs with 9 × 10⁷ generations with a sampling frequency of 1000; determination of the convergence of the MCMC chains and the proper burn-in values through the stationary phase of the runs using the effective sample size analysed with Tracer 1.7.1 (Rambaut et al. 2018), combination of generated trees using LogCombiner v.1.10.4 (Suchard et al. 2018) following a 25% burn-in at the outset of each run, and calculation of the consensus tree and Bayesian posterior probabilities using Tree Annotator version 1.10.4 (Suchard et al. 2018); and Maximum Likelihood analysis performed with IQTREE 2.2.2.6 (Minh et al. 2020), with support node expressed by ultrafast bootstrap (Hoang et al. 2018) and the Shimodaira-Hasegawa-like approximate likelihood ratio test (SH-aLRT; Guindon et al. 2010), both using 1000 replicates. In order to evaluate the contribution of each genetic locus dataset analysed, we generated individual trees for each of the nuclear genes and one for the mitochondrial genes, using ML as above described.

The divergence time analysis was conducted in Beast 1.10.4 using the same dataset, partitions, evolution models, and parameters as described above. Additionally, the analysis incorporated a lognormal uncorrelated relaxed clock model and a Yule speciation process as the tree prior (Gernhard 2008). Calibration points were established as follows: the origin of the Trichomycteridae with a normal prior distribution (mean = 106 MA, SD = 5.0), following the estimative of Betancur-R et al. (2015), which is often used as an indirect calibration strategy in other studies on trichomycterids (e.g., Ochoa et al. 2017; Vilardo et al. 2023); and the origin of the genus Corydoras with a lognormal prior distribution (mean = 55 MA, SD = 1), based on the dating of Corydoras revelatus Cockerell, 1925, the oldest known fossil of callichthyid catfishes. MCMC chains were assessed to verify convergence by evaluating the effective sample size of the runs in Tracer 1.7.1. The time-scaled tree was obtained using Tree Annotator version 1.10.4 to generate the consensus tree.

All phylogenetic analyses resulted in identical tree topologies (Fig. 5), in which T. antiquus is supported as sister to T. itatiayae, a species endemic to another region of the RPSB, with broad occurrence in streams draining the Maciço de Itatiaia, a massif forming a distinct nucleus of the Serra da Mantiqueira, and the adjacent Serra da Bocaina (Fig. 4). The clade comprising T. antiquus and T. itatiayae, hereafter the Trichomycterus itatiayae group, is supported as sister to a clade comprising all other species of Paracambeva known as the T. reinhardti species group (Costa 2021; Costa and Katz 2021). The independent analysis of individual gene trees indicated that all loci collaborated for this topology, since all gene trees corroborated monophyly of Paracambeva and both the RAG2 tree and the mitochondrial locus tree corroborated monophyly of the clade comprising T. antiquus and T. itatiayae, which is supported in both trees as sister to a clade comprising the remaining species of Paracambeva (Suppl. material 2). The age of the Paracambeva lineage was estimated at about 24.5 Ma, Late Oligoceno (95% HPD age interval 15.68–36.01), whereas according to the analysis, the divergence between the T. itatiayae group and the T. reinhardti group occurred at about 14.9 Ma, Middle Miocene (95% HPD age interval 7.16–24.84), and between T. antiquus and T. itatiayae at about 9.8 Ma, Late Miocene (95% HPD age interval 4.74–16.87).

Figure 5. 

Time-scaled tree obtained from the Bayesian analysis in Beast for 28 species of Trichomycteridae and 4 species as outgroups, using a multigene data set (COX1, CYTB, RAG2, and MYH6 with a total of 3144 bp). Asterisks (*) indicate maximum support values, and dashes (-) indicate support values below 50. Black stars indicate the calibration points, and the coloured bars below the tree represent the geological epochs. Numbers above branches indicate posterior probabilities of the Bayesian Inference, followed by SH-aLRT support (%) and ultrafast bootstrap support (%) of the Maximum Likelihood (ML) analysis; blue numbers below nodes indicate its median age; Red dots after species names indicate the syntopic species discussed in this paper.

The presence of a broad black longitudinal stripe along the midline of the flank in juveniles, which often gradually becomes diffuse and fragmented into small spots in adults, combined with dark spots on the dorsum, occurs in most species of Paracambeva (Costa 2021; Costa and Katz 2021; Costa et al. 2023b), as well as in most species of the T. nigroauratus group of the subgenus Trichomycterus (e.g., Barbosa and Costa 2008; Costa et al. 2022). However, the specific colour pattern described here for T. antiquus and T. maculosus above about 70 mm SL, including a narrow stripe that posteriorly breaks into small spots, combined with the scarcity or absence of dark spots on the ventral part of the flank, occurs only in these two species and in T. itatiayae (Barbosa and Costa 2008: fig. 5). According to our analysis, this specific pattern would have first appeared in the ancestor of the clade comprising T. antiquus and T. itatiayae, around 14.9 Ma (95% HPD age interval 8.87–23.64), and a second time in the exclusive ancestor of T. maculosus, around 2.3 Ma (95% HPD age interval 0.85–4.56) (Fig. 5; Suppl. material 3).

According to a recent biogeographical analysis, the most recent common ancestor of Paracambeva lived in an area presently occupied by the upper Rio Paraná, upper Rio São Francisco, and Rio Paraíba do Sul basins (Vilardo et al. 2023). Presently, Paracambeva occupies a large region encompassing these three basins, with the T. itatiayae group occurring in a portion of the RPSB between the Rio do Peixe and the area close to the Serra da Bocaina and the T. reinhardti group occurring in a broad area of the upper Rio Paraná and upper Rio São Francisco basins (Fig. 4). Timing estimates support a Late Oligocene origin for the Paracambeva lineage and a Middle Miocene age for the split between the T. itatiayae and T. reinhardti groups (Fig. 5). Despite conclusive hypotheses about the time of origin of the present configuration of these basins that are not yet available, geological data point to a past connection between the upper section of the Rio Paraíba do Sul, above the area close to the Serra da Bocaina, and the Rio Paraná basin (King 1956; Riccomini et al. 2010), and between the upper section of the latter basin and the upper section of the Rio São Francisco basin (Rezende et al. 2018). Therefore, the present distribution of Paracambeva and our estimates for the time of origin of this subgenus fit into this model of river basin evolution, with the rupture between the upper Paraíba do Sul and the Paraná-São Francisco basin that would have occurred in the Paleogene (Riccomini et al. 2010) corresponding to the divergence between the T. itatiayae and T. reinhardti groups, which preceded the rupture of the connection between the upper Rio Paraná and upper Rio São Francisco basins during the Middle Miocene (Rezende et al. 2018), corresponding to the wide distribution of the T. reinhardti group in these basins (Fig. 4; Costa and Katz 2021).

On the other hand, temporal estimates indicated the origin of the subgenus Trichomycterus lineage in the early Miocene. The biogeographical analysis performed by Vilardo et al. (2023) indicated that the MRCA of the subgenus Trichomycterus inhabited an ancestral area comprising only the Rio Paraíba do Sul, contrasting with its present distribution that includes both this basin and smaller coastal basins (Fig. 4; Costa 2021). However, geological data support a Cenozoic configuration of the RPSB different from the present one, in which its lower course corresponded to the present lower course of the Rio São João (Riccomini et al. 2010), which is now an isolated coastal basin. Thus, geological data are congruent with the hypothesised ancestral area of the subgenus in the Rio Paraíba do Sul alone, which in the past did not include the present upper course but included a present coastal basin in its lower course. The upper course of the RPSB probably was blocked for dispersion of trichomycterine catfishes that live in fast flowing streams by the great depression in its main channel responsible for the formation of the paleolake Tremembé during the Oligocene (Riccomini et al. 2004). Thus, it is possible that these two subgenera lineages, Paracambeva and Trichomycterus, were not in contact in the upper Rio Paraíba do Sul before the Middle Miocene. The origin of the specific colour pattern shared by the syntopic T. antiquus and T. maculosus is estimated to have occurred first in the T. itatiayae group during the Middle Miocene and much later in T. maculosus during the Late Pliocene in T. maculosus, which is compatible with the hypothesis of a more recent occupation of the upper RPSB by the subgenus Trichomycterus.

Interestingly, the other case involving similarly coloured syntopic species of trichomycterines from eastern South America involves T. itatiayae, the sister group of T. antiquus, and T. nigroauratus, the sister group of T. maculosus (Fig. 5). These two species are endemic to streams of the RPSB draining the Serra da Mantiqueira and the adjacent Serra da Bocaina (Barbosa and Costa 2008) and are commonly found associated with bottom leaf litter (Costa 2021). Juvenile specimens of these species share a colour pattern consisting of a broad black longitudinal stripe along the flank midline, whereas larger specimens assume a different colour pattern (Barbosa and Costa 2008). A broad black longitudinal stripe is present in all other species of Paracambeva (Costa and Katz 2021; Costa et al. 2023b), therefore already present in the MRCA of this subgenus, whereas this colour pattern is present in part of the species of the T. nigroauratus group among species of the subgenus Trichomycterus (Costa et al. 2022), thus arising after the initial diversification of this group. Considering the estimated age of Paracambeva in the Late Oligocene and the initial diversification of the T. nigroauratus group in the Late Pliocene, over 20 million years later, the most plausible hypothesis is that the colour pattern in the T. nigroauratus group had arisen after this group was in contact with species of Paracambeva in RPSB.

The sympatric occurrence of distantly related species of Neotropical catfishes exhibiting similar derived colour patterns has been often aprioristically considered as primary evidence of mimetic association (Axenrot and Kullander 2003; Alexandrou et al. 2011; Slobodian and Bockmann 2013), but may also be a result of evolutionary convergence for adaptation to live in special habitats like those occurring among sympatric psammophilic species (Zuanon et al. 2006; Costa et al. 2020c). Therefore, the occurrence of two species of Trichomycterus in the same habitat sharing the same colour pattern (Figs 1A, 6) but belonging to two distantly related subgenera could suggest a case of mimetic association or convergence for adaptation to live in a gravel bottom where both species were found, since their colour pattern is cryptic in this habitat. However, direct evidence to explain syntopic trichomycterines with similar colour patterns is not available for any of these hypotheses.

Figure 6. 

Trichomycterus (Trichomycterus) maculosus, UFRJ 13676, 97.6 mm SL, left lateral view.

In the case of evolutionary convergence for adaptation to live in special habitats, a possible explanation is that the colour pattern gives these two species some cryptic advantage in their habitat against predators since the colours are similar to the gravel substrate where they live. In the case of mimetic associations, including both Batesian and Müllerian mimicry, the model species (Batesian) or both species (Müllerian) have effective anti-predation features, which among catfishes usually comprise venom glands associated with fin spines (Wright 2009; Harris and Jenner 2019). For example, in mimetic associations involving siluroids with a well-developed pectoral-fin spine, the main anti-predation morpho-physiological attribute is the presence of an axillary venom gland associated with a pungent pectoral-fin spine (e.g., Greven et al. 2006; Wright 2011; Carvalho et al. 2021). The defence mechanism involves not only glands and the pectoral-fin spine but also special muscles and connective tissue (Wright 2015; Harris and Jenner 2019). However, mechanisms for anti-predation in trichomycterines are still unknown, and potential trichomycterine predators living in the Rio do Peixe drainage have not been presently recorded, although the Neotropical otter Lontra longicaudis (Olfers, 1818) and the catfish Steindachneridion parahybae (Steindachner, 1877) today rare or absent in the region, were until recently potential predators.

Although anti-predation features consisting of venom glands associated with fin spines occur in most catfish lineages (Wright 2009, 2015), they are unknown among trichomycterids. Unlike other catfishes, trichomycterids do not have pectoral and dorsal spines. In a survey on the presence of venom axillary glands in catfish lineages, Wright (2009) did not detect them in ‘Trichomycterus’ areolatus Valenciennes, 1846, concluding that these glands are not present in trichomycterids. However, a supposed axillary gland (e.g., Eigenmann 1918) or suprapectoral adipose organ according to Myers and Weitzman (1966) and axillary organ according to de Pinna (1989), situated at the same place as the venom axillary glands of other catfishes and having a similar orifice, have been superficially described for candirus and other trichomycterids (e.g., Eigenmann 1918). This supposed axillary gland, comprising a sack-like protuberance above the pectoral fin and just below the anterior pores of the lateral line of the flank, is also present in trichomycterines (Fig. 2), which is proportionally smaller than in other trichomycterids (e.g., sarcoglanidines, Myers and Weitzman 1966), with an orifice that is more conspicuous in juvenile specimens below about 40 mm SL, often having a great concentration of melanophores around its margin. Both T. antiquus and T. maculosus have a small axillary gland-like protuberance below the short lateral line canal and above the pectoral fin (Fig. 2). The absence of a pectoral spine in trichomycterids also imposes a limitation on the possibility of the axillary organ acting as a venom gland. Nonetheless, the stinging action of the opercular odontodes, which are located close to the axillary gland-like protuberance (Fig. 2) and may become bristly when the fish is molested, could be performed as the pectoral spine of other catfishes. However, a more detailed morphological study at the histological and biochemical level is necessary to investigate the presence of glandular tissue and toxic substances in T. antiquus and close relatives, which is not presently possible with the small sample of specimens currently available. Future studies are necessary to check what hypothesis best explains the unexpected syntopic occurrence of similarly coloured T. antiquus and T. maculosus.