Integrative taxonomy reveals a new species of Cyphocharax (Characiformes: Curimatidae) from the Upper Paraíba do Sul River basin, Brazil

A new species of Cyphocharax is described from the Upper Paraíba do Sul River basin, São Paulo, Brazil based on integrated morphological and molecular delimitation criteria. It is morphologically distinguished from its congeners by the presence of a round, dark blotch at the midlength of the caudal peduncle not extending to the proximal portions of the median caudal-fin rays, 19–20 circumpeduncular scales, 34–41 perforated lateral-line scales, 6–7 longitudinal scale rows above and below the lateral line, greatest body depth corresponding to 34.7–39.9% of standard length (SL), and the caudal peduncle depth corresponding to 13.3–15.2% of SL. The lowest genetic distances between the new species and other congeners are: 2.5% from C. gilbert , followed by 3.0% from C. santacatarinae , and 3.2% from C. aff. gilbert . All species delimitation criteria employed herein corroborated the recognition of the new species. In addition, comments on its conservation status are provided.

In his revisionary study of the genus, Vari (1992) considered C. gilbert to be a widely distributed species occurring across eastern Brazilian coastal drainages from Bahia to Rio de Janeiro and eastern São Paulo. More recently, Melo et al. (2018) expanded the C. gilbert clade to include all species from coastal drainages from eastern Brazil and the Paraná River and La Plata system, as well as Cyphocharax corumbae (Pavanelli & Britski, 3/17 ni.bio.br | scielo.br/ni Guilherme M. Dutra, George Vita, Péricles V. Gentile, Luz E. Ochoa and Andre L. Netto-Ferreira 1999). Those authors considered C. gilbert to represent a single species. However, analyses of specimens from the Upper Paraíba do Sul River basin (eastern São Paulo), previously attributed to 'Cyphocharax gilbert' indicate that they belong to a distinct and undescribed species, which is described herein.

MATERIAL AND METHODS
Morphological analyses. Measurements and counts follow Vari (1992) and Melo, Vari (2014), with the addition of the number of circumpeduncular scales, scale rows around the least depth of the caudal peduncle (Fink, Weitzman, 1974). All measurements were taken point-to-point to the nearest 0.1 mm with digital calipers under a stereomicroscope with incident light, preferably on the left side. In the description, frequencies are given in parentheses after each count, and an asterisk indicates counts for the holotype. Two specimens were cleared and stained according to Taylor, Van Dyke (1985). Vertebral counts included the four vertebrae of the Weberian apparatus as individual precaudal centra, whereas the fused PU1+U1 was considered a single centrum.

Molecular analyses.
Total genomic DNA of seven specimens corresponding to the new species were extracted from muscle tissue according to the Ivanova et al. (2006) protocol (Tab. 1). We optimized PCR conditions to amplify a 657 base pair (bp) fragment of the mitochondrial Cytochrome c oxidase subunit I protein-coding gene (COI) using the primers FishF6 and FishR7 (Jennings et al., 2019). PCR reactions were run in a volume of 12.5 ul, with 1.25 μl of 10X buffer (10 mM Tris-HCl+15 mM MgCl2), 0.5 μl dNTPs (200 nM of each), 0.5 μl each 5 mM primer, 0.05 μl Platinum® Taq Polymerase (Invitrogen), 1 μl genomic DNA (10-50 ng), and 8.7 μl ddH2O. The thermo-cycler profile consisted of an initial denaturation (2 min at 95ºC) followed by 30 cycles of chain denaturation (30 sec at 94ºC), primer hybridization (30 s at 52ºC), nucleotide extension (30-60 sec at 72ºC), and a final extension (8 min at 72ºC). All PCR products were first visually identified on 1% agarose gel and purified using ExoSap-IT® (USB Corporation) following manufacturer instructions. Sequencing was performed at the sequencing facility Centro de Pesquisa sobre o Genoma Humano e Células Tronco at Universidade de São Paulo, São Paulo, Brazil.  Alignment and phylogenetic analyses. Consensus sequences for the new species were assembled and edited in Geneious Prime 2022.0.1 and aligned with 203 sequences of cytochrome oxidase (COI) from Melo et al. (2018) (Tab. S1) using the MUSCLE algorithm (Edgar, 2004). A total matrix with 210 samples (see supplementary material) and 657 bp was used to estimate a Maximum Likelihood (ML) tree in RAxML v. 8.0.19 (Stamatakis, 2014) on the 40 CPU 128GB Brycon server at LBP/UNESP. We conducted 20 Maximum Likelihood searches for the phylogenetic tree that best fit the data under the GTRGAMMA substitution model and used non-parametric bootstrapping to assess nodal support by allowing the program to automatically determine the number of replicates using the autoMRE criterion. To test the validity of the new species, we implemented three species delimitation approaches. First, we employed the Bayesian implementation of the Poisson Tree Process model (bPTP, Zhang et al., 2013), using the best ML tree as the input file in the PTP web server (http://species.h-its.org) with default settings. The second method was the Generalized Mixed Yule-Coalescent model (GMYC) (Pons et al., 2006;Fujisawa, Barraclough, 2013). This model aims to discern stochastic birth-death processes separating species from neutral coalescent processes occurring within species, by analyzing time intervals between branching events in a single time-calibrated gene tree (Pons et al., 2006). For this analysis, we estimated a calibrated tree using the lognormal relaxed molecular clock model, which assumes that the rates of molecular evolution are uncorrelated, but log-normally distributed among lineages as implemented in BEAST v1.8.2 (Drummond, Rambaut, 2007). We obtained an ultrametric gene tree with two independent runs of 50 million generations with one tree sampled every 1000th generation. BEAST log files were examined in Tracer v1.6 (Rambaut et al., 2018) to assess stationarity and parameter convergence. A maximum clade credibility tree was estimated in TreeAnnotator v1.8.1, based on the sampled trees after discarding the first 25% generations as burn-in. The ultrametric gene tree was used in the GMYC analysis using a single threshold via GMYC web server. Third, we employed the program Assemble Species by Automatic Partitioning (ASAP) which uses pairwise genetic distances to provide species partitions ranked by a scoring system that uses no biological prior insight about intraspecific diversity (Puillandre et al., 2021). Finally, to contextualize the results of the species delimitation analyses, we estimated pairwise genetic distances for a sub-matrix containing other Cyphocharax species in Mega version 11.0.10 using the Kimura 2-parameter model. Sabaj (2020). Abbreviations used in the text are: CS = cleared and counterstained specimens, SL = standard length, and HL = head length. Diagnosis. Cyphocharax tamuya differs from most congeners, except C. aninha Wosiacki & Miranda, 2014, C. gilbert, C. gillii (Eigenmann & Kennedy, 1903, C. gouldingi Vari, 1992, C. meniscaprorus Vari, 1992, C. mestomyllon Vari, 1992, C. modestus (Fernández-Yépez, 1948, C. naegelii (Steindachner, 1881), C. oenas Vari, 1992, C. platanus (Günther, 1880, C. santacatarinae (Fernández-Yépez, 1948), C. spilotus (Vari, 1987), C. spiluropsis (Eigenmann & Eigenmann, 1889), and C. spilurus (Günther, 1864) by the absence of distinctive pigmentation of the body and fins combined with a dark blotch on the midlength of the caudal peduncle [vs. possession of a reticulate color pattern formed by pigmentation on the scale borders plus a round dark blotch on the caudal peduncle in C. gangamon Vari, 1992; a longitudinal series of dark blotches on the scales of the flanks plus a caudal peduncle blotch extending from the base of the median caudal-fin rays to the vertical through the posterior limit of the adipose-fin base in C. sanctigabrielis; multiple series of dark stripes or spots running between most scale rows in C. cramptoni Bortolo & Lima, 2020, C. helleri (Steindachner, 1910, C. multilineatus (Myers, 1927), and C. pantostictos Vari & Barriga Salazar, 1990; a patch of dark pigmentation on the dorsal fin in C. notatus (Steindachner, 1908) and C. vexillapinnus Vari, 1992; a single dark midlateral stripe in C. boiadeiro Melo, 2017, C. corumbae, C. laticlavius Vari & Blackledge, 1996, C. saladensis (Meinken, 1933, and C. signatus Vari, 1992; a midlateral series of irregular patches of dark pigmentation along the lateral line in C. albula, C. jagunco, C. punctatus (Vari & Nijssen, 1986), and C. vanderi (Britski, 1980); a random pattern of small dark spots on the lateral and dorsolateral surfaces of the body in C. voga (Hensel, 1870); a midlateral spot of dark pigmentation ventral to the dorsal-fin base in C. biocellatus Vari, Sidlauskas & Le Bail, 2012; or the lack of a pronounced caudal-peduncle blotch in C. abramoides (Kner, 1858), C. aspilos Vari, 1992, C. derhami Vari & Chang, 2006, C. festivus Vari, 1992, C. leucostictus (Eigenmann & Eigenmann, 1889, C. magdalenae, C. microcephalus (Eigenmann & Eigenmann, 1889), C. muyrakytan Bortolo, Lima & Melo, 2018, C. nigripinnis Vari, 1992, C. pinnilepis, C. plumbeus (Eigenmann & Eigenmann, 1889, and C. stilbolepis Vari, 1992]. Cyphocharax tamuya can be distinguished from most congeners with a dark blotch on the caudal peduncle, except C. naegelii, by the presence of 19-20 circumpeduncular scales [vs. 16-18 in C. gilbert (n = 43), and 16 in C. modestus (n = 3), C. santacatarinae (n = 10), C. spilotus (n = 8), C. voga (n = 11), and 23-25 in C. platanus (n = 9)]. It differs from C. naegelii by possessing a round caudal peduncle blotch that does not reach the proximal portions of the median caudal-fin rays [vs. a horizontally elongate blotch that extends onto the median caudal-fin rays (Fig. 2)], the greatest body depth corresponding to 34.7-39.9% of SL (vs. 29.0-33.0% of SL), the caudal peduncle depth corresponding to 13.3-15.2% of SL (vs. 12.0-13.0% of SL), and the presence of 33 vertebrae (vs. 34). The new species differs from C. platanus by the presence of 34-41 pored lateral-line scales from the supracleithrum to the hypural joint (vs. 48-54), and the presence of 6-7 scale rows between the lateral-line scale series and the anal-fin origin (vs. 8-10). Furthermore, the new species differs from C. voga by the presence of 33 vertebrae (vs. 34-37), from C. Head compressed, pointed overall in lateral view. Mouth subterminal, located at horizontal through ventral margin of pupil. Upper jaw slightly longer than lower jaw. Nostrils close together and separated by thin flap of skin. Anterior nostril circular, near midpoint between snout tip and anterior margin of eye. Posterior nostril crescent shaped. Adipose eyelid more developed anteriorly, with vertically ovoid opening near center of eye. Eye on anterior one-half of head length. Branchial membranes joined at isthmus. Branchiostegal rays 4(2).

Coloration in alcohol.
Specimens retain silvery guanine deposits on infraorbitals and opercular series. Background coloration of body tan. Dark coloration on dorsal portion of head, maxilla, upper lip and infraorbital 1, lighter ventrally. Dark chromatophores on postorbital region of head slightly larger than those on snout. Overall pigmentation of latter portion of postorbital region consequently lighter than adjoining surrounding areas. Scale rows from predorsal to lateral-line series with dark chromatophores concentrated at focus of each scale. Scales below lateral line with few, sparse chromatophores, lacking distinctive pigmentation pattern. Lateral surface of body of small specimens (up to 69.7 mm SL; Fig. 2B) with faint dark midlateral stripe up to 2 scales deep running from vertical through dorsal-fin insertion to anterior margin of caudal peduncle blotch. Dark midlateral stripe inconspicuous or absent in specimens larger than 70 mm SL. Midlateral surface of caudal peduncle with dark round blotch, not extending onto proximal portions 9/17 ni.bio.br | scielo.br/ni Guilherme M. Dutra, George Vita, Péricles V. Gentile, Luz E. Ochoa and Andre L. Netto-Ferreira of median caudal-fin rays. Caudal peduncle blotch more conspicuous in juveniles (Fig.  2B). Distal tip of dorsal fin with discrete, dark chromatophores on rays and associated membranes. Anal-fin rays with dark chromatophores on distal tip, more concentrated on anteriormost rays. Adipose, caudal, pectoral and pelvic fins overall hyaline. Caudal fin with discrete dark chromatophores more concentrated on lower caudal-fin lobe in some specimens.
Geographical distribution. Cyphocharax tamuya is known from the Paraíbuna and Paraitinga rivers, two main headwater tributaries of the Paraíba do Sul River, São Paulo, Brazil (Fig. 3).  Etymology. The epithet "tamuya" honours the indigenous Tamuya people. The word Tamuya, from Tupinambá language, means "grandfather or the oldest". Linguistics variations to Tamuya include "Tamúya", "Tamujas" or "Tamuía", all keeping the same origin and meaning. The Tamuyas were driven extinct during the colonization of the Brazilian coast in the seventeenth century. Before their extinction, they lived in the area of occurrence of Cyphocharax tamuya. A noun in apposition.

Conservation status.
The Upper Paraíba do Sul River basin area was impacted in the 1970's by the construction of the Paraibuna Dam, which changed the water dynamics in the area. However, considering that Cyphocharax species inhabit lentic waters, the dam may not impact the population of C. tamuya in that area. In fact, part of the specimens examined herein originate from inundation zone above the dam and were collected after the dam's construction (e.g., MZUSP 21668). Furthermore, the species also occurs in at least three localities upstream of the dam. Thus, we recommend that C. tamuya be classified as Least Concern (LC), according to the International Union for Conservation of Nature (IUCN) categories and criteria (IUCN Standards and Petitions Subcommittee, 2019). Molecular species delimitation. The molecular dataset included 210 sequences of 657 base pairs in length. Stop codons, deletions or insertions were not observed in any sequences. In the matrix 386 positions were conserved and 271 were variable, with base compositions of 23% adenine, 27.9% cytosine, 30.1% thymine and 18.4% of guanine. The data were not saturated as evaluated via the Iss.c value. The maximum likelihood tree from RAxML shows the same topology obtained in the ultrametric tree from BEAST and the three delimitation analyses (Fig. 4). The bPTP approach delimited 95 lineages, and Cyphocharax tamuya was grouped with the remaining species composing the C. gilbert clade of Melo et al. (2018) with a posterior probability of 0.926 (Fig. 5). In contrast, the results from the ASAP analysis showed a low asap score of 4.00 with a p-value 0.0063, and delimited 101 species from the total dataset (six more than did bPTP). The ASAP analysis delimited all species belonging to the C. gilbert clade as unique lineages except for C. naegelii and C. platanus which were grouped in a single lineage. The best fitting model in the GMYC analysis received a likelihood score of 1471.86 in comparison with the null model of 1435.10. The GMYC analysis identified a total of 106 entities with a confidence interval of 103-111 and a threshold time of -0.007. This analysis recognized the new species as a differentiated entity, again grouped C. naegelii and C. platanus in   Numbers at right of nodes represent bootstrap support. the same lineage, and recovered two distinct lineages within C. modesus. Regarding the genetic distances from a sub-matrix containing the COI sequences from the all known species in C. gilbert clade (Tab. 3), the overall mean distance among species was 3%. The lowest genetic distance between C. tamuya and any other known species was 2.5% (from C. gilbert), followed by 3% from C. santacatarinae, and 3.2% from C. aff. gilbert. Vari (1989) hypothesized the Cyphocharax gilbert group as including C. gilbert, C. modestus, C. santacatarinae, and C. voga, based on the shared presence of a round caudal peduncle blotch. Vari also suggested C. naegelii and C. platanus as close relatives of the C. gilbert group based on overall similarity. Recently, Melo et al. (2018) corroborated the C. gilbert group as a well-supported monophyletic clade within Curimatidae, also including C. corumbae, C. naegelii, C. platanus, and C. spilotus, in agreement with the morphological hypothesis of Vari (1989). The maximum likelihood analysis presented herein recovered a similar topology to that proposed by Melo et al. (2018). Despite the methodological limitations of single locus analyses in phylogenetic inferences (e.g., Hajibabaei et al., 2006Hajibabaei et al., , 2007, it is important to note that the present analysis supports the assignment of C. tamuya to the C. gilbert clade and adds molecular evidence to the morphological similarity to support the hypothesis of a close relationship between these species.

DISCUSSION
In his revisionary study of Cyphocharax, Vari (1992) interpreted the conspicuous variation in the shape of the dark blotch on the caudal peduncle among the specimens of 'C. gilbert' as reflecting intra-specific plasticity. Upon the revalidation of C. albula, Dutra et al. (2017) presented evidence partially refuting Vari's delimitation of C. gilbert, which the description of C. tamuya herein further corroborates. In short, Vari's concept of C. gilbert subsumed several entities now recognized as distinct species.
The examination of specimens corresponding to C. gilbert across its distribution revealed that specimens from coastal rivers between the Paraíba do Sul and Doce rivers drainages present a dark round blotch on the caudal peduncle that never reaches the proximal portion of the median caudal-fin rays. The holotype of C. gilbert demonstrates a similar pattern. In contrast, at least part of the specimens distributed in coastal rivers of northeastern Brazil between the Mucuri River (Gomes et al., 2015), and the de Contas River (Vari, 1992), possess an elongate caudal peduncle blotch extending onto the median caudal-fin rays. This pattern differs from the holotype of C. gilbert and suggests that these northeastern specimens may not be conspecific with C. gilbert. The single locus (COI) evidence presented herein also supports a subdivision of Vari's concept of C. gilbert, by recovering a genetic difference (1.9%) between C. aff. gilbert from the Mucuri River and specimens of C. gilbert from closer to the type locality. That result supports Gomes et al. (2015), suggestion that C. aff gilbert represents another distinct species. Our single-locus phylogeny places this species as the sister group of C. tamuya, with this clade being the sister of C. gilbert from the Paraíba do Sul River basin and coastal rivers eastern Brazil between the Paraíba do Sul and the Baía de Guanabara, though this placement requires testing with a multilocus dataset. Further examination of some of the specimens that Vari included in his study (MZUSP 20686, 20687, 20729, 21668), revealed that a higher number of circumpeduncular scales easily distinguishes C. tamuya from C. gilbert and C. aff. gilbert. Vari (1992) did not examine this character.
Overall, the present results highlight the need for a comprehensive taxonomic review of the C. gilbert group using integrative taxonomic approaches. Such review would not only clarify the currently neglected diversity under this name, but also allow proper conservation strategies for the species occurring in deeply populated areas in Brazil, where rising urbanization, agriculture and river damming pressure the known and unknown components of ichthyological diversity.