Testing the phylogenetic hypotheses of Stevardiinae Gill, 1858 in light of new phenotypic data (Teleostei: Characidae)

2.1 Morphological characters

A total of 176 phenotypic characters were analyzed. Characters are listed in Appendix 1. Most characters were osteological (115), with the remaining ones derived from sperm (42), external features (14), and histology (5).

The cleared and staining protocol followed Taylor and Van Dyke (1985). Whenever possible, two or more specimens of each species were prepared to check for anomalous features and to allow, whenever possible, the evaluation of sexually dimorphic characters. Osteological terminology follows Weitzman (1962) with the modifications listed by Vari and Harold (2001). The complete list of examined material and museum acronyms is included in Appendix 2.

Testes of specimens, previously fixed in 4% formaldehyde and stored in 70% alcohol, were gradually rehydrated in a decreasing ethanol concentration of distilled water. Once rehydrated, the material was refixed overnight in 2% glutaraldehyde and 4% paraformaldehyde in 0.1 M pH 7.2 Sorensen phosphate buffer. Samples were post-fixed in the dark for 2h in 1% osmium tetroxide in the same buffer, stained in block with an aqueous solution of 0.5% uranyl acetate for 2 h, dehydrated in acetone, embedded in araldite, sectioned, and stained with a saturated solution of uranyl acetate in 50% ethanol and lead citrate. Electron micrographs were obtained using a Carl Zeiss LEO 906 (Carl Zeiss, Germany) transmission electron microscope.

Ovaries were removed, fixed in 10% buffered formalin, and dehydrated in different grades of ethanol and clarified xylene. After processing, the tissues were embedded in paraffin, sectioned to a thickness of 5 µm, and stained with hematoxylin and eosin.

2.2 Molecular genetic data

DNA sequence data was entirely obtained from public repositories, mostly from GenBank (www.ncbi.nlm.nih.gov/nuccore) except for one sequence provided by Arcila et al. (2017) (Data S1). Four mitochondrial and four nuclear markers were analyzed. Mitochondrial sequences included fragments of the cytochrome c oxidase subunit 1 gene (COX1), cytochrome b gene (CYTB), and the ribosomal 12S and 16S rRNA genes. Nuclear markers were all DNA sequences of protein-coding genes, including portions of myosin heavy chain 6 (MYH6), patched domain containing 1 (PTCHD1), recombination-activating gene 1 (RAG1), and recombination-activating gene 2 (RAG2). Molecular data were aligned with Muscle (Edgar, 2004) using default settings. Gaps were considered as missing data. Potential contamination or mistakes in the identification of GenBank samples were detected through the use of BLAST (Altschul et al., 1990) and parsimony analyses of each molecular partition. When two or more reliable sequences for the same marker and species were available, they were merged using IUPAC codes for polymorphisms with Asado software (Nixon, 2004). Datasets for each individual marker including their GenBank accession numbers are provided as Data S1. When possible, voucher specimens for GenBank sequences were traced back to their ichthyological collections or public databases to account for any changes in the identifications since the publication of the sequences and correct them accordingly. Alignments for each marker are provided as Data S2.

2.3 Dataset and analysis

The complete matrix is available in TNT format as Data S2. The morphological partition is provided as Data S3 and available online at Morphobank P3663 (O’Leary & Kaufman, 2011, 2012). Missing entries in the data matrix are represented by “?” when the character states could not be evaluated, or by “*” for inapplicable character states.

The complete dataset is composed of 6309 characters, with 176 in the morphological partition and the remaining from genetic data. The number of informative characters is 1338, of which 169 are from morphology, 303 from CYTB, 216 from COX1, 93 from MYH6, 60 from PTCHD1, 215 from RAG1, 153 from RAG2, 35 from 12S, and 94 from 16S. The dataset has about 48% of missing entries, mainly occurring for species lacking data for some or all the molecular markers.

Phylogenetic analyses were done under the parsimony criterion with TNT (Goloboff et al., 2008). Searches included tree fusing, sectorial searches (Goloboff, 1999), and parsimony ratchet (Nixon, 1999), stopping each round after reaching three hits to the same optimum. Analyses were performed under extended implied weighting (Goloboff, 2014), which reduces the influence of homoplastic characters with many missing entries relative to equally homoplastic characters with more complete scoring. This analytical choice prevented characters from assuming artificially high weights due to lack of data rather than for congruence with the remaining species. Thus, this method solves a potential problem that arises when the original implied weighting (Goloboff, 1993) is applied to the analysis of combined datasets (Goloboff, 2014). Extended implied weighting also allows weighting each character against the homoplasy of each partition or smaller subdivision, instead of the homoplasy of each individual character. This is useful for molecular data, in which the homology of each phylogenetic character is relative, because it results from an alignment of contiguous and ordered DNA sites. This contrasts with most morphological features, which are recognizable individually, even if dislocated from the specimen. Also, in DNA data homoplasy is commonly concentrated in sets of contiguous sites (see Goloboff, 2014, for details and further justification of the method).

As in Mirande (2017, 2019) and Terán et al. (2020), different weighting schemes were used. Each individual DNA character was weighted according to its own homoplasy (SEP) or to the average homoplasy of a codon (three contiguous columns) (COD), 10 codons (30 contiguous columns) (GRP), or entire markers (BLK). Also, a weighting scheme grouping all first, second, and third codon positions as separate sets in coding (non-ribosomal) sequences (POS) was used. Each morphological character was weighted according to its own homoplasy in all cases. Ribosomal DNA data were weighted according to the homoplasy of the entire partitions in the POS scheme. In addition to the weighting schemes, the regular and extended implied weighting methods allow the user to define weighting strengths (K values), which determine how abruptly a character reduces its influence in the analysis as a function of their number of homoplastic steps (Goloboff, 1993, 2014).

Each of the five abovementioned weighting schemes is herein combined with three different weighting strengths, in which an average character has about 84, 87, and 90% of the weight of a perfectly hierarchic one. This corresponds to concavity constant K values of 7, 9, and 12, respectively. This combination of weighting schemes and strengths results in 15 analytical conditions. The number of analyzed species represents an arguably poor sampling for homoplasy and, hence, precludes the use of stronger (lower) values of K, which in exploratory searches produced obviously artefactual results, as the non-monophyly of the Stevardiinae. Hereafter, each particular search strategy is referred to as its scheme followed by its strength (e.g., POS9 represents POS scheme and K-value = 9).

A search under equal weighting was done for comparative purposes. A metacriterion of stability estimated through SPR distances (Goloboff, 2008) was used to select among the most parsimonious trees (MPTs) obtained under different weighting conditions (see Mirande, 2009, for details). Support was estimated through symmetric resampling (probability of change of 0.33) with values expressed as differences of frequencies (GC values in TNT) (Goloboff et al., 2003).

2.4 Institutional abbreviations

Abbreviations for institutions and collections are as follows: ANSP, Academy of Natural Sciences of Drexel University, Philadelphia; CAS, California Academy of Sciences, San Francisco; CPUFMT, Coleção de Peixes da Universidade Federal de Mato Grosso, Cuiabá; LIRP, Laboratório de Ictiologia de Ribeirão Preto, Universidade de São Paulo, Ribeirão Preto; MCZ, Museum of Comparative Zoology, Washington D.C.; MZUSP, Museu de Zoologia da Universidade de São Paulo, São Paulo; MNRJ, Museu Nacional da Universidade Federal do Rio de Janeiro, Rio de Janeiro; MCP, Museu da Pontifícia Universidade Católica do Rio Grande do Sul, Porto Alegre; MUSM, Museo de Historia Natural de la Universidad Mayor de San Marcos, Lima.

4.1 Landonini Weitzman & Menezes, 1998, new combination—Type genus Landonia Eigenmann & Henn, 1914

The tribe Landonini, as herein recognized, contains three genera: Eretmobrycon, Landonia, and Markiana. The close relationship between Eretmobrycon and Markiana was first pointed out by Thomaz et al. (2015), who proposed a new tribe Eretmobryconini. According to them, the Eretmobryconini was the sister group of the remaining Stevardiinae. Thomaz et al. (2015), however, did not include Landonia in their analysis and considered it an incertae sedis genus in the Stevardiinae. The monophyly of Eretmobryconini, as well as its position as the sister group of all the other tribes of Stevardiinae, was corroborated by Mirande (2019), who also did not include Landonia and considered this genus as tentatively belonging in the Diapomini, which present study contradicts. Vanegas-Ríos (2018) analyzed Landonia for the first time in a phylogenetic context and found this genus as the sister group of Phenacobrycon, in a different clade than in the present contribution. Our results constitute an alternative hypothesis to the one by Vanegas-Ríos (2018), based on a different set of morphological characters and adding information from DNA/rRNA sequences. Even if our analysis cannot logically refute the hypothesis by Vanegas-Ríos (2018) because both proposals were obtained from different data, he did not propose a tribal classification for Landonia, which was part of his outgroup. Given the high support we found for the inclusion of Landonia in the clade of Eretmobrycon, the absence of the former genus in previous analyses (Mirande, 2019; Thomaz et al., 2015), and the temporal priority of Landonini Weitzman & Menezes, 1998 over Eretmobryconini Thomaz et al. (2015), we propose herein the Eretmobryconini as a junior synonym of Landonini. Mirande (2019) diagnosed the Eretmobryconini based on eight morphological synapomorphies and proposed as additional characters the presence of iii, nine dorsal-fin rays, four teeth in the inner premaxillary row, the infraorbital three reaching the preopercle, the presence of two uroneurals, and the anteroventral angle of the articulation of the second and third infraorbitals. Most (if not all) of these characters are shared also by Landonia, so that the inclusion of this genus would not require major changes in the diagnosis of the tribe.

4.2 Xenurobryconini Myers & Böhlke, 1956 sensu Thomaz et al., 2015—Type genus Xenurobrycon Myers & Miranda Ribeiro, 1945

Based on the similarity of their caudal fins, Myers and Böhlke (1956) proposed the tribe Xenurobryconini for the genera Tyttocharax and Xenurobrycon. Weitzman and Menezes (1998), analyzing mainly the anatomy of the caudal fin in the Glandulocaudinae (sensu Weitzman & Fink, 1985), expanded the composition of Xenurobryconini by including in that tribe the genera Argopleura, Chrysobrycon, Iotabrycon, Ptychocharax, Scopaeocharax, Tyttocharax, and Xenurobrycon. Xenurobryconini was redefined by Thomaz et al. (2015), who excluded Argopleura, considered incertae sedis, and Chrysobrycon, which was transferred to Stevardiini. Mirande (2019) corroborated the monophyly of Xenurobryconini as proposed by Thomaz et al. (2015), with the exception that he also obtained Cyanogaster Mattox et al., 2013 in this clade. However, the actual relationship of Cyanogaster with the Xenurobryconini or even with the Stevardiinae is uncertain and subject of ongoing studies by one of the authors (JMM).

4.3 Phenacobryconini Weitzman & Menezes, 1998—Type genus Phenacobrycon (Eigenmann, 1914)

The tribe Phenacobryconini was proposed by Weitzman and Menezes (1998) to include only the nominotypical genus. The present study corroborates the validity of this tribe, given that Phenacobrycon is not found in a clade with other nominotypical genera having temporal priority. Indeed, according to our results, P. henni is the sister group of a large clade including most stevardiins. Nine morphological autapomorphies are obtained for the monotypic tribe Phenacobryconini, one of them related to caudal-fin pouch-scale morphology (see Results). According to Weitzman and Fink (1985), P. henni has an enlarged caudal scale on the lower caudal-fin lobe. However, the pouch scale present in P. henni is not as hypertrophied as in Acrobrycon, Diapoma, and Planaltina Böhlke (see character 113:0).

Thomaz et al. (2015) and Mirande (2019) did not include Phenacobrycon in their analyses and considered the genus as incertae sedis in Stevardiinae. Vanegas-Ríos (2018) found Phenacobrycon as the sister group of Landonia but did not propose a tribal classification for those genera. According to our results, Phenacobryconini is the sister group of a clade composed of the tribes Creagrutini, Diapomini, Glandulocaudini, Hemibryconini, and Stevardiini. The resurrection of Phenacobryconini does not contradict previous analyses, given that Thomaz et al. (2015) and Mirande (2019) did not analyze P. henni and in the relationships obtained by Vanegas-Ríos (2018), the Phenacobryconini may have been attributed to the sister group of a monotypic Landonini.

4.4 Stevardiini Gill, 1858 sensu Thomaz et al., 2015—Type genus Stevardia Gill, 1858

The Stevardiini was given the level of tribe by Thomaz et al. (2015) to group the Corynopomini and Hysteronotini of Weitzman and Menezes (1998) plus Chrysobrycon, included in the Xenurobryconini (Weitzman & Menezes, 1998). Although Thomaz et al. (2015) did not include Hysteronotus and Pterobrycon in their analysis, they agreed that both genera belong in the Stevardiini and suggested further studies to better understand their relationships with other members of the tribe.

In the present study, Hysteronotus megalostomus Eigenmann, 1911 and Pterobrycon myrnae Bussing, 1974 are analyzed and our results corroborate the hypothesis by Thomaz et al. (2015). The Stevardiini are composed of six genera distributed in two clades. The first clade has Chrysobrycon as the sister group of Hysteronotus and Pseudocorynopoma. The genus Chrysobrycon was proposed by Weitzman and Menezes (1998) to include two species previously described as Hysteronotus hesperus Böhlke, 1958 and Hysteronotus myersi Weitzman & Thomerson, 1970. In the proposal by Weitzman and Menezes (1998), Chrysobrycon was the sister genus of the remaining xenurobryconin genera. The present study strongly contradicts that hypothesis. The similarity between the pouch scale present in males of H. megalostomus and Pseudocorynopoma doriae Perugia, 1891 was first mentioned by Weitzman and Thomerson (1970). The relationship between Hysteronotus and Pseudocorynopoma, primarily described by Weitzman and Menezes (1998), was corroborated by Ferreira et al. (2011), Thomaz et al. (2015), and Mirande (2019).

The second clade obtained in the Stevardiini has Pterobrycon as the sister group of Corynopoma and Gephyrocharax. Relationships among these three genera were first proposed by Weitzman and Menezes (1998), who redefined the tribe Corynopomini (compare with Burns et al., 1995; Weitzman et al., 1988). The hypothesis by Weitzman and Menezes (1998) showed Pterobrycon and Corynopoma as sister taxa. Our results do not corroborate this relationship. Herein, Gephyrocharax and Corynopoma are recovered as sister genera in the final consensus topology, which agrees with the results obtained by Vanegas-Ríos (2018). According to that author, the clade formed by Corynopoma and Gephyrocharax was supported by the presence of spurs formed by hypertrophy of the ventral procurrent rays of the caudal fin. Thomaz et al. (2015) and Mirande (2019) also corroborated a close relationship between Corynopoma and Gephyrocharax, but they did not include Pterobrycon in their analyses.

4.5 Glandulocaudini Eigenmann, 1914 sensu Menezes & Weitzman, 2009—Type genus Glandulocauda Eigenmann, 1911

In the present study, the monophyletic condition of the Glandulocaudini as proposed by Thomaz et al. (2015) is corroborated, including Mimagoniates as the sister group of Glandulocauda and Lophiobrycon. Previous studies (Castro et al., 2003; Menezes & Weitzman, 2009) considered Lophiobrycon as sister group of a clade composed of Glandulocauda and Mimagoniates. Camelier et al. (2018), in a phylogenetic study of the Glandulocaudini based on DNA/rRNA data, found that Glandulocauda was not monophyletic, since that study reconstructed Glandulocauda melanopleura (Ellis, 1911) as sister to Lophiobrycon weitzmani Castro et al., 2003 rather than being included in a clade with Glandulocauda caerulea Menezes & Weitzman, 2009, which turned out to be closely related to Mimagoniates. In our analysis, G. melanopleura and L. weitzmani are sister groups but we did not analyze G. caerulea.

As mentioned earlier, the Glandulocaudini are supported by 40 synapomorphies, of which three are features of spermatozoa. The first sperm synapomorphy is the absence of a depression in the nuclear contour named the nuclear fossa (143:1). The nuclear fossa may be present or absent in species that have type II spermiogenesis. When present, the nuclear fossa can be classified according to the appearance of the contour of the nuclear membrane in longitudinal section: regular (subdivided into single or double concavity) or with irregular projections that penetrate the nucleus. Among the Stevardiinae, the loss of the nuclear fossa is homoplastic in the clade composed of Xenurobrycon and Tyttocharax. The absence of the midpiece (150:1) and the presence of a cytoplasmic canal restricted only to the apical region of the nucleus (174:0) represent the last two sperm synapomorphies. At the end of spermiogenesis, due to nuclear elongation, the midpiece disappears and the cytoplasmic canal remains restricted in the apical region of the nucleus. The mitochondria are anchored in the concave face of the nucleus throughout its length, and there are several mitochondria organized as a row along the nuclear outline. The absence of a midpiece is a homoplastic character in Pseudocorynopoma and the clade of Scopaeocharax, Tyttocharax, and Xenurobrycon. For further explanations about characid sperm and spermiogenesis and how phylogenetic characters were defined for this kind of data, read Quagio-Grassiotto et al. (2020).

4.6 Creagrutini Miles, 1943 sensu Thomaz et al., 2015—Type genus Creagrutus Günther, 1864

Our results corroborate the monophyly of Creagrutini as proposed by Thomaz et al. (2015), where Carlastyanax and Creagrutus were obtained as sister groups. The relationship between Carlastyanax and Creagrutus was first proposed by Mirande et al. (2013). Recently, Mirande (2019) expanded the Creagrutini to include Planaltina britskii Menezes et al., 2003, Lepidocharax burnsi Ferreira et al., 2011, and Microgenys minuta Eigenmann, 1913. In his analysis, Creagrutini was represented by two clades: one including L. burnsi and P. britskii, and the other having M. minuta as the sister group of Carlastyanax and Creagrutus. Although herein we did not include M. minuta, the Creagrutini sensu Mirande (2019) are strongly contradicted by the inclusion of Lepidocharax burnsi and Planaltina britskii in the Diapomini.

4.7 Hemibryconini Géry, 1966 sensu Thomaz et al., 2015—Type genus Hemibrycon Günther, 1864

The tribe Hemibryconini as defined by Thomaz et al. (2015) includes Acrobrycon, Hemibrycon, and tentatively the genus Boehlkea Géry, 1966. We did not include Boehlkea in our analysis, but according to Bertaco and Malabarba (2010), Boehlkea and Hemibrycon share some morphological similarities and are probably closely related.

Acrobrycon was included in the Diapomini, together with Planaltina Böhlke, 1954 and Diapoma Cope, 1894 by Weitzman et al. (1988) based on the presence of two attributes of the pheromone organs overlying the basal portions of the caudal fin: (1) mature males and females possessing organs of equivalent size; (2) presence of three or more series of scales immediately ventral to the lateral line scale series, which form the dorsal border of the pouch opening (see also Arcila et al., 2013). These characters were included in our analysis; however, the tribe Diapomini as proposed by Weitzman et al. (1988) was strongly contradicted.

4.8 Diapomini Eigenmann, 1909 sensu Thomaz et al., 2015—Type genus Diapoma Cope, 1894

The monophyly of the Diapomini sensu Thomaz et al. (2015) is corroborated in the present study. According to our analysis, it includes two clades (Figure 1). The first one shows the non-monophyly of Lepidocharax Ferreira et al., 2011, due to L. burnsi being the sister group of Planaltina. The first hypothesis of relationships of Lepidocharax was erected by Ferreira et al. (2011), who proposed the genus to be sister to all the remaining Stevardiinae (sensu Mirande, 2010), pointing out several similarities of morphological characters and sperm ultrastructure between Lepidocharax and Planaltina. In Thomaz et al. (2015), Lepidocharax appeared as the sister group of the Diapomini and it was considered as incertae sedis in this tribe. Planaltina species were not included in the analysis by Thomaz et al. (2015); however, they tentatively included it in the Diapomini. In Mirande (2019), Lepidocharax and Planaltina were also obtained as closely related to each other but placed within Creagrutini. In the present study, the clade comprising these genera is supported by 11 morphological synapomorphies: (1) presence of a deep groove above eyes containing many neuromasts (1:2), (2) ascending premaxillary process reaching as much as 1/3 of nasal bone length (18:1), (3) lateral wings of urohyal reaching only the middle longitudinal portion of the bone (70:1), (4) origin of first dorsal-fin ray inserted at vertical through anal-fin origin (89:1), (5) anteriormost hemal postzygapophyses present in transitional vertebrae (106:1), (6) nuclear fossa located laterally almost to tip of nucleus (145:3), (7) membranous vesicles in intermediate region of midpiece present (168:0), (8) membranous vesicles in posterior region of midpiece present (169:0), and absence of tubules and vesicles interconnected in (9) anterior (170:1), (10) intermediate (171:1), and (11) posterior (171:2) regions of midpiece.

The second clade of Diapomini is subdivided into two other clades. The first of them is composed of Bryconadenos, Knodus Eigenmann, 1911, Phallobrycon Menezes et al., 2009 and Rhinobrycon Myers, 1944. Rhinobrycon is the sister genus of the remaining ones, Knodus is not monophyletic, and Phallobrycon is the sister taxon of Bryconadenos. We prefer not to follow the classification proposed by Thomaz et al. (2015) concerning the synonymy of Bryconadenos with Knodus, given that neither their analysis nor ours inspected sufficient species of this clade or putatively related genera to support such nomenclatural changes. Also, Thomaz et al. (2015) did not provide an updated diagnosis of Knodus. A close relationship between Knodus and Rhinobrycon was also reported by Thomaz et al. (2015) and Mirande (2019) who did not include Phallobrycon in their studies. Thomaz et al. (2015) considered Phallobrycon as incertae sedis in Stevardiinae.

The sister group of the clade composed of Bryconadenos, Knodus, Phallobrycon, and Rhinobrycon is composed of several genera and supported by eight synapomorphies, of which one is morphological: possession of 18 or fewer branched anal-fin rays (91: 1). This clade is subdivided into two other clades, composed respectively of Attonitus Vari & Ortega, 2000 plus Ceratobranchia Eigenmann, in Eigenmann et al. (1914) and a large group formed by Bryconacidnus, Myers in Eigenmann and Myers (1929), Bryconamericus Eigenmann in Eigenmann et al. (1907), Diapoma, Monotocheirodon Eigenmann & Pearson, 1924, Nantis, Odontostoechus, Piabina Reinhardt, 1867, and Rhinopetitia Géry, 1964.

The close relationship of Attonitus and Ceratobranchia obtained in the present study corroborates the hypotheses by Thomaz et al. (2015) and Mirande (2019) in which Bryconamericus pachacuti Eigenmann, 1927 and Bryconamericus pectinatus Vari & Siebert, 1990 (transferred to Bryconacidnus by Thomaz et al., 2015) appeared as related to Attonitus and Ceratobranchia. Bryconamericus pachacuti was not included in the present study, and the status of Bryconacidnus will be discussed below. In the analysis by Thomaz et al. (2015), Ceratobranchia was obtained as the sister group of Attonitus, B. pachacuti, and B. pectinatus, while in Mirande (2019), Ceratobranchia was the sister group of B. pectinatus and B. pachacuti nested with Attonitus. The clade of Attonitus and Ceratobranchia is herein supported by 25 synapomorphies, of which seven are morphological: (1) two premaxillary tooth rows (31:1), (2) inner premaxillary teeth with maximum of three cusps (35:1), (3) inner premaxillary teeth with base narrower than crown (36:0), (4) dentary teeth with three or more cusps (53:1), (5) contact ectopterygoid-quadrate absent (54:1), (6) palatine bone one half as long as ectopterygoid or shorter (55:2), and (7) presence of eight branched dorsal-fin rays (88:1).

The remaining Diapomini comprises a clade including Diapoma plus Piabina and a larger monophyletic group formed by Bryconacidnus, Bryconamericus, Monotocheirodon, Nantis, Odontostoechus, and Rhinopetitia. The close relationship of Diapoma and Piabina was already proposed by Thomaz et al. (2015) but also includes Piabarchus Myers, 1928, which was not analyzed herein. In Mirande (2019), Piabina was the sister group of a more comprehensive clade that also included Diapoma. This clade is supported by nine synapomorphies, one of them morphological: (1) All anal-fin proximal radials posterior to first hemal spine (93:0).

The final clade within Diapomini clade includes Bryconacidnus plus Bryconamericus as the sister group of Monotocheirodon, Nantis, Odontostoechus, and Rhinopetitia. In Thomaz et al. (2015), a group of terminal taxa attributed to the genus Bryconacidnus formed the sister group of Attonitus and B. pachacuti. Mirande (2019) included B. pectinatus (as Bryconacidnus pectinatus) in his analysis, and this species appeared as related to Ceratobranchia, “Bryconamericus” diaphanus (Cope, 1878), “Bryconamericus” pachacuti, and the species of Attonitus. Bryconamericus was related to Diapoma, Piabarchus, and Piabina in Thomaz et al. (2015). In Mirande (2019), Bryconamericus, together with Piabarchus Myers (not included here), was the sister group of Nantis and Odontostoechus, as in the present study. According to Menezes and Netto-Ferreira (2019), the species of Rhinopetitia were more closely related to Bryconacidnus, Monotocheirodon, and Odontostoechus than to the remaining species of the Stevardiinae. That study reconstructed Bryconacidnus as the sister group of the other three genera, which the present study corroborates. This clade is supported by 14 synapomorphies, one of them morphological: scales extended posteriorly only to base of caudal fin (111:0).

Hypobrycon, Nantis, and Odontostoechus had been synonymized with Bryconamericus by Thomaz et al. (2015) due to the relationships they obtained for Bryconamericus exodon Eigenmann, 1907, the type species of the genus. Mirande (2019: 16) resurrected these three genera, after a comprehensive phylogenetic analysis of the Characidae. Bryconamericus exodon is very similar to B. stramineus Eigenmann, 1908 and, among the available information, only some DNA data exclude it from a sister group relationship with that species. Available sequences attributed to B. exodon in the literature and deposited in GenBank were produced by Oliveira et al. (2011) from specimens in lot LBP 7123, Pereira et al. (2013) from lots LBP 2893, 6392, 7111, and 19824, Thomaz et al. (2015) from a single specimen in the lot UFRGS 13571, and García-Melo et al. (2019) from lots LBP 5118, 12926, 13408, 13463, 13565, and 13598. Among these specimens, the lots from Oliveira et al. (2011) and Pereira et al. (2013) are currently assigned to either Bryconamericus iheringii Boulenger, 1887 or B. aff. iheringii in the collection databases, while the voucher from Thomaz et al. (2015) and most lots from García-Melo et al. (2019) are still catalogued as B. exodon. One of the authors (JMM) corroborated the current identification of some of the LBP lots. All the information attributed to B. exodon as analyzed by Thomaz et al. (2015) is sequences obtained from a single specimen still identified as B. exodon (UFRGS 13571). Analysis of sequences from this individual resulted in several taxonomic changes in the genus, but the identification of the specimen itself is suspect. Most of these sequences are very similar to DNA/rRNA samples of B. stramineus and Piabarchus analis (Eigenmann, 1914) (to B. stramineus RAG1: 99.03%, RAG2: 98.73%, 12S: 98.57%; to P. analis 16S: 99.28%), and indeed, B. exodon nests with those two species in gene-by-gene parsimony analyses. However, the COX1 sequence of B. exodon analyzed by Thomaz et al. (2015) has a similarity of 99.23% with a GenBank sample of Bryconamericus ikaa Casciotta et al., 2004. Indeed, the 100 most similar samples in GenBank database are from species in the clade of B. iheringii, rather distantly related to B. exodon among the Stevardiinae. The closest presumably true B. exodon COX1 sequence in GenBank has a similarity of 90.98% with this one, analyzed by Thomaz et al. (2015). Parsimony analyses of the complete dataset of Thomaz et al. (2015), under equal and implied weighting using medium-to-mild concavity values (K), resulted in the sister group relationship of B. exodon with the clade of B. iheringii, which included also Hypobrycon, Nantis, and Odontostoechus, in congruence with the results of Thomaz et al. (2015) under model-based analyses. Strong weighting against homoplasy (K < 10), instead, put B. exodon in the clade of B. stramineus and P. analis. Either removing the COX1 sequence of B. exodon from the dataset of Thomaz et al. (2015) or replacing it with one from García-Melo et al. (2019) resulted in a close relationship of B. exodon with B. stramineus and P. analis under any analytical condition. This makes evident a contamination of the COX1 sequence of B. exodon in Thomaz et al. (2015), which consequently invalidates the synonymy of Hypobrycon, Nantis, and Odontostoechus and implies the exclusion from Bryconamericus of the species found in the “Bryconamericus iheringii clade” of Mirande (2019). Among these three genera, Odontostoechus Gomes, 1947 has priority over Hypobrycon Malabarba & Malabarba, 1994 and Nantis. However, the generic reassignment of the species of the “Bryconamericus iheringii clade” (Mirande, 2019) to Odontostoechus still depends on the corroboration by broader analyses of the relationships herein obtained for Monotocheirodon Eigenmann & Pearson and, eventually, the position of Othonocheirodus Myers, 1927, which has yet to be included in any phylogenetic analysis. Both genera have temporal precedence over Odontostoechus.

This contribution constitutes a rather independent test for the phylogenetic relationships in the Stevardiinae. It also includes a set of characters related to sperm morphology and development for the first time in a phylogenetic analysis of the subfamily. This set of characters is particularly interesting for the Stevardiinae given that this clade includes species with insemination and presumably correlated variation in the morphology of sperm cells (e.gBurns & Weitzman, 2005; Javonillo et al., 2007; Pecio et al., 2005; Quagio-Grassiotto et al., 2020). Among the stevardiins previously included in phylogenetic analyses, insemination was reported for the species of Glandulocaudini, Stevardiini, Xenurobryconini, and Acrobrycon (Hemibryconini), Argopleura (incertae sedis), Attonitus, Bryconacidnus, some species of Bryconamericus, Diapoma, and Knodus (Diapomini), and some species of Creagrutus (Creagrutini) (Thomaz et al., 2015). The evolution of insemination, as optimized by parsimony on the tree topology obtained by Thomaz et al. (2015) would imply at least seven instances of parallel evolution with reversals at the common ancestor of Hemibrycon and the tribes Creagrutini and Hemibryconini. In this analysis, we found the presence of insemination ambiguous at the common ancestor of the Stevardiinae and most parsimoniously reconstructed as present in the clade comprising all components of the subfamily except for the Landonini. The ancestral state of the Landonini is also ambiguous given that insemination is present in Landonia, absent in Markiana, and unknown in Eretmobrycon. Thus, in one of the most parsimonious reconstructions of the character, the insemination had a single origin in the Stevardiinae, with multiple reversals to external fertilization. Spermatozeugmata are unencapsulated sperm packets present in many inseminating characid species (Pecio et al., 2005), but not in all (e.g., inseminating species of Diapoma, and species of Attonitus and Gephyrocharax). Therefore, not only their distribution in the Stevardiinae is more restricted than insemination, but also there are inseminating species in which this feature is unknown (e.g., Landonia), which renders the reconstruction of this character somewhat tentative at this point. According to the available data, the spermatozeugmata had also an origin in the clade including all the Stevardiinae excepting the Landonini, followed by four reversals to absence and an additional gain in the clade of Bryconadenos and Phallobrycon. However, as this feature is unknown for Landonia, the presence of spermatozeugmata in the common ancestor of all the Stevardiinae is still possible. We have not yet completely understood the evolutionary mechanisms and conditions that allowed the great diversification of the Characidae and particularly of the Stevardiinae, but is highly probable that insemination was one of them. Its comprehension may provide a clue to understand the evolution of this subfamily.

As mentioned, this dataset is somewhat exploratory in its search for additional characters informative about internal relationships of the Stevardiini. Therefore, we intentionally made no formal taxonomic changes in the cases wherein our results contradicted previous analyses. In the future, it will be valuable to expand the coding of characters from sperm and spermiogenesis to a broader extent in the family and also to combine the set of morphological characters herein proposed with those from the literature (Mirande, 2019; Vanegas-Ríos, 2018). However, that would require a minute comparison character by character and state by state to prevent redundancies and, after that, a huge work coding the missing entries derived from the different taxon samplings of the involved approaches. This contribution provides an intermediate step toward that broader goal.

The inspected characters behaved as expected, recovering most of the relationships in the literature and providing additional synapomorphies for most clades. The monophyly of the Stevardiinae and most of its tribes, as proposed in the literature, were corroborated. The different set of characters and taxa analyzed herein provided new morphological synapomorphies for the subfamily and their tribes and also imply some forthcoming taxonomic changes in the Stevardiinae. The most important taxonomic finding with respect to previous classifications is the recognition of the tribes Landonini, as a senior synonym of Eretmobryconini, and the monotypic Phenacobryconini. However, the main focus of this analysis was on the phenotypic characters and especially on those related to insemination, sperm, and spermiogenesis. This analysis results in a reinterpretation of the evolution of the presence of insemination and partially correlated characters, as the spermatozeugmata. Until now, several parallel acquisition of those reproductive traits in the Stevardiinae were suggested or, at least, implicitly considered. Our results are compatible with a single acquisition of insemination and related features, with several reversals in the subfamily. This hypothesis, however, should be further tested, given that information about reproductive strategies and morphology of the associated organs and reproductive cells is missing for many species in the Stevardiinae and Characidae in general. Also, potentially informative details about sperm and spermiogenesis remain to be studied and coded for many species. This contribution serve as step forward in that sense.