A third pelvic-less species of Cambeva from river basins draining the Geral mountain range in southern Brazil is described. It is distinguished from other congeners lacking pelvic fin and girdle, C. pascuali and C. tropeiro, by having six pectoral-fin rays, 20–23 dorsal procurrent caudal-fin rays, 15–20 opercular and 25–30 interopercular odontodes and a different colour pattern consisting of flank dark brownish-grey with two irregular horizontal rows of small pale yellow grey marks. Whereas available molecular evidence indicates that C. pascuali is more closely related to C. zonata, a species with well-developed pelvic fin, and C. tropeiro is more closely related to C. balios, another species also with well-developed pelvic fin; osteological data strongly suggest that the new species herein described is more closely related to C. diatropoporos than to other congeners. Therefore, this study indicates that the pelvic fin and pelvic-fin support have been lost independently in each of these three species of Cambeva, which corresponds to 11% of all describe species. This result highly contrasts with the closely-related trichomycterine genera Trichomycterus, in which only one in 50 species lost pelvic fin and girdle (0.2%) and Scleronema with all the nine included species having well-developed pelvic fin. These data suggest a stronger tendency to losing pelvic fin in Cambeva, but factors favouring this evolutionary event are still unknown.

Mountain biodiversity, osteology, Rio Uruguai basin, Serra Geral, systematics

The presence or not of pelvic fin was considered an important morphological character for generic delimitation in teleost fishes by 19^th century naturalists. Amongst the Trichomycterinae, one of the eight subfamilies of the Neotropical catfish Trichomycteridae, the two oldest genera were distinguished only on the basis of this character. Trichomycterus Valenciennes, 1832 was first diagnosed by its only included species, Trichomycterus nigricans Valenciennes, 1832, having pelvic fin (Valenciennes in Humboldt and Bonpland 1832), thus differing from the only other then-known trichomycterine Eremophilus mutisii Humboldt, 1805, in which pelvic fin is absent (Humboldt 1805). This classificatory scheme was followed without criticism for a long time. Over 100 years after, two other trichomycterids were placed in the genus Eremophilus, E. candidus Miranda-Ribeiro, 1949 and E. camposi Miranda-Ribeiro, 1957, just by lacking pelvic fin (Miranda-Ribeiro 1949, 1957). However, morphological and molecular studies have consistently shown that E. candidus is a member of the genus Trichomycterus (Barbosa and Costa 2003; Katz et al. 2018) and E. camposi belongs to the microcambevine genus Listrura de Pinna, 1988 (de Pinna 1988; Costa et al. 2020a), therefore constituting independent evolutionary events of pelvic fin loss amongst trichomycterids.

Recent taxonomical studies on the Trichomycterinae, which is a species-rich clade with over 220 species occurring in most river basins between southern Central America and Patagonia in southern South America (Katz et al. 2018), have revealed few other species also lacking pelvic fin. Costa and Bockmann (1993) have shown that some specimens of Ituglanis parahybae (Eigenmann, 1918) lack pelvic fin. More recently, five pelvic-less trichomycterines have been recorded from the Argentinean Andes (Fernández and Liotta 2016), as well as two species lacking pelvic fin being described from south-eastern and southern Brazil (Ferrer and Malabarba 2011; Ochoa et al. 2017a). These Brazilian species are presently placed in Cambeva Katz, Barbosa, Mattos & Costa, 2018, a genus comprising 25 valid species endemic to south-eastern and southern Brazil (Katz et al. 2018; dos Reis et al. 2019), as well as a single species from north-eastern Argentina (Terán et al. 2017).

Amongst the 26 species of Cambeva, only C. pascuali (Ochoa, Silva, Costa e Silva, Oliveira & Datovo, 2017) and C. tropeiro (Ferrer & Malabarba, 2011) lack pelvic fin and girdle (Ferrer and Malabarba; 2011 Ochoa et al. 2017b), whereas, in all other 24 remaining species, these structures are well-developed, including an ossified pelvic girdle and a pelvic fin with five rays. We herein describe a third species of Cambeva lacking pelvic fin and girdle, exhibiting some morphological features indicating that it is not closely related to the other two congeners lacking pelvic fin.

Morphometric and meristic data were taken following Costa (1992), with modifications proposed by Costa et al. (2020b); measurements are presented as percent of standard length (SL), except for those related to head morphology, which are expressed as percent of head length. Fin-ray counts include all elements; vertebra counts include all vertebrae except those participating in the Weberian apparatus; the compound caudal centrum was counted as a single element; counts of vertebrae and procurrent fin rays were made only in cleared and stained specimens; counts of principal-fin rays were made in all available specimens, except juveniles about 25 mm SL or less; counts of jaw teeth were approximate, due to their irregular arrangement, great number and frequent loss, making impossible accurate counts. Specimens were cleared and stained for bone and cartilage (C&S in lists of specimens) following Taylor and Van Dyke (1985); osteological characters included in the description are those belonging to structures that have informative variability for positioning the new species amongst congeners, including the mesethmoidal region, suspensorium and opercular apparatus and the parurohyal. Terminology for bones followed Bockmann et al. (2004), except for: the ‘antorbital’ and the ‘tendon-bone supraorbital’ that are here called ‘lacrimal-antorbital’ and ‘fronto-lacrimal tendon bone’, respectively, following Baskin (1973); ‘epibranchial 5 cartilage’, here replaced by ‘ceratobranchial 4 accessory element’ following Carvalho et al. (2013); ‘urohyal’, here substituted by ‘parurohyal’ following Arratia and Schultze (1990); ‘pleural rib’, here substituted by ‘rib’ following Britz and Bartsch (2003) that have shown to exist only a single rib type in teleosts. Osteological illustrations were made using a stereomicroscope Zeiss Stemi SV 6 with camera lucida. Cephalic laterosensory system terminology follows Arratia and Huaquin (1995), with modifications proposed by Bockmann et al. (2004). Specimens are deposited in the ichthyological collection of the Institute of Biology of the Federal University of Rio de Janeiro, Rio de Janeiro city and in the Centre of Agrarian and Environmental Sciences, Federal University of Maranhão, Campus Chapadinha (CICCAA). Comparative material is listed in Costa et al. (2020b), with the addition of Cambeva diatropoporos UFRJ 6913, 1 (C&S). Within the text, geographical names follow Portuguese terms used in the region, thus avoiding common errors or generalisations when tentatively translating them to English, besides making easier their identification in the field.

Phylogenetic relationships amongst species of Cambeva are still poorly known. With rare exceptions (Bockmann et al. 2004; Ferrer and Malabarba 2013), osteological data that have been broadly used to diagnose species and establish morphological phylogenetic hypotheses in trichomycterid groups (e.g. Baskin 1973), have been broadly omitted even in recent taxonomical studies on Cambeva (e.g. Wosiacki and Garavello 2004; Ferrer and Malabarba 2011; Ochoa et al. 2017b). Similarly, trichomycterid molecular phylogenies have included only a few species of this genus (Ochoa et al. 2017a, 2020; Katz et al. 2018). Therefore, the phylogenetic position of the new species herein described is still uncertain, but some derived osteological character states, shared by C. flavopicta and other congeners here examined or with available osteological data from literature, highly suggest possible relationships as below discussed.

At first glance, the absence of pelvic fin and girdle in all specimens of C. flavopicta could be considered as evidence of close relationships with the only two congeners sharing this condition, C. pascuali from the Rio Paranapanema drainage, Rio Paraná basin and C. tropeiro from the Lagoa dos Patos system. However, a recent unilocus phylogeny (Donin et al. 2020) indicates that C. pascuali and C. tropeiro are not closely-related species. The former species appears as being closer to C. zonata, a species from the Rio Ribeira do Iguape basin in south-eastern Brazil, whereas the latter is sister to C. balios (Ferrer & Malabarba, 2013), a species also endemic to the Lagoa dos Patos system and sharing with C. tropeiro a similar colour pattern consisting of rounded dark brown spots on the flank (Ferrer and Malabarba 2011: fig. 1, 2013: fig. 1). On the other hand, osteological evidence described below supports C. flavopicta as being closely related to species having a well-developed pelvic fin.

The long premaxilla recorded for C. flavopicta, with its longest length greater than the distance between the extremities of the mesethmoid cornua (Fig. 2A), is similar to the long premaxilla shared only by some congeners (Ferrer and Malabarba 2013: fig. 2) belonging to a molecularly well-supported clade from rivers of the Lagoa dos Patos system draining the Serra Geral (e.g. Katz et al. 2018; hereafter the C. balios group). Amongst species of the C. balios group, C. flavopicta shares with C. diatropoporos (Ferrer & Malabarba, 2013) and C. poikilos (Ferrer & Malabarba, 2013), a derived compact autopalatine with an almost indistinct posterolateral process (Fig. 2A; Ferrer and Malabarba 2013: fig. 2b, c), not present in any congener here examined. Amongst species herein examined, the presence of a short process on the anteroventral portion of the metapterygoid, just anterior to the articulatory cartilaginous block, is shared only by C. flavopicta and C. poikilos, but the metapterygoid of C. diatropoporos, which is possibly more closely related to C. flavopicta than to C. poikilos, was not illustrated in previous studies and specimens were not available to this study. According to Ferrer and Malabarba (2013: fig. 2c), C. diatropoporos has a long antorbital, about half the length of the fronto-lacrimal tendon bone length, a derived condition here recorded only for C. flavopicta (Fig. 2A). In addition, C. diatropoporos possesses a colour pattern similar to that of C. flavopicta, including flank dark brownish-grey with small pale yellow grey marks with varied shapes (Ferrer and Malabarba 2013: fig. 10), corroborating a hypothesis of close relationships between C. flavopicta and C. diatropoporos. Besides lacking pelvic fin and girdle, C. flavopicta also differs from C. diatropoporos and C. poikilos in having more dorsal procurrent rays in the caudal fin (21–23 vs. 15–19), more ventral procurrent rays in the caudal fin (16–17 vs. 10–15), the posterior margin of the caudal fin convex (vs. straight) and presence of a small filament on the tip of the first pectoral-fin ray (vs. absence); from C. diatropoporos by having the extremity of the parurohyal lateral process sharply pointed (Fig. 2C; vs. broad and rounded, Ferrer and Malabarba 2013: fig. 5b) and fewer pectoral-fin rays (6 vs. 7); and from C. poikilos by the presence of the anterior section of the infra-orbital canal (vs. absence).

As discussed above, this study indicates that the pelvic fin and pelvic-fin support have been lost independently in three lineages of the genus Cambeva, corresponding to three species in a total of 27 (11.1%). This result highly contrasts with the closely-related trichomycterine genus Trichomycterus, in which only one species, T. candidus, does not have pelvic fin and girdle, amongst a total of 50 valid species (0.2%), as well as Scleronema, the sister group of Cambeva (Katz et al., 2018), with all the nine included species having well-developed pelvic fin and girdle (Ferrer and Malabarba 2020). These data suggest a stronger tendency to losing pelvic fin in Cambeva than in these closely-related genera, but factors favouring this evolutionary event are still unknown. Amongst trichomycterids of the TSVSGM-clade (Costa and Bockmann 1994; Costa et al. 2020a), parallel loss of pelvic fin and reduction or loss of other fins in the subfamilies Glanapteryginae and Microcambevinae have been related to their interstitial habits (Costa et al. 2020a). On the other hand, conflicting field reports about the habitat of pelvic-less species of Cambeva do not indicate that they share similar habits. Specimens of C. flavopicta were found buried in terrestrial marginal vegetation of a fast flowing stream (see above), whereas specimens of C. pascuali were found under rocks and aquatic vegetation of a muddy bottom area bordered by riparian vegetation (Ochoa et al. 2017b), but no data are available about the specific habitat of C. tropeiro. Therefore, little is still known about the ecology and behaviour in species of the genus Cambeva, presently not allowing explanations involving habits for the evolutionary meaning of multiple pelvic fin losses.

We are grateful to João Antônio de Bittencourt Vitto for assistance during field expeditions. Instituto do Meio Ambiente, Santa Catarina and Instituto Chico Mendes de Conservação da Biodiversidade provided collecting permits. This work was partially supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; grant 307349/2015-2 to WJEMC). The manuscript benefited from a detailed review provided by F. Ottoni.