Ontogenetic changes in mouth morphology triggers conflicting hypotheses of relationships in characid fishes (Ostariophysi: Characiformes)

Bryconamericus lethostigmus is the type-species of the monotypic genus Odontostoechus, diagnosed in part based on the presence of a unique tooth series in the premaxilla. Recently a new proposal of classification of the Stevardiinae placed Odontostoechus as a junior synonym of a monophyletic genus Bryconamericus sensu stricto, a genus characterized by the presence of two tooth series. Bryconamericus lethostigmus is redescribed herein and the single tooth series in the premaxilla is demonstrated to originate from merging of the external tooth row with the inner row during ontogeny refuting primary hypothesis of homology between the mouth morphology of B. lethostigmus and the genera Bryconacidnus, Ceratobranchia, Monotocheirodon, Othonocheirodus, Rhinopetitia and Rhinobrycon. A phylogeographic analysis indicated that the pattern described for the sympatric species Diapoma itaimbe is not mirrored by B. lethostigmus. The results also do not support the hypothesis of a new species in the rio Araranguá drainage.


Introduction
Bryconamericus lethostigmus (Gomes) was described as the type-species of the monotypic characid genus Odontostoechus Gomes, originally assigned to the subfamily Cheirodontinae due to the presence of a single tooth series in the premaxilla. Gomes (1947) proposed the genus Odontostoechus as related to Distoechus Gomes (= Deuterodon Eigenmann), Othonocheirodus Myers and Monotocheirodon Eigenmann & Pearson based on the structure of the mouth, and, in the case of the last two genera, by the presence of two rows of gill rakers on the lower ramus of the branchial arches. Indeed, Böhlke (1954) suggested a new Cheirodontinae tribe, the Monotocheirodontini, composed by Odontostoechus lethostigmus, Othonocheirodus and Monotocheirodon, based on similarities of the mouth, while Géry (1977) placed Odontostoechus as a junior synonym of Othonocheirodus. However, Malabarba (1998) considered Odontostoechus as a possibly valid genus encompassing several Bryconamericus-like species. This author suggested that the single tooth series in Odontostoechus originates during ontogeny where the premaxilla forms from the merging of the two tooth series, and therefore, that the presence a single tooth series is not homologous in regard to other Cheirodontinae species, placing the genus as incertae sedis in Characidae.  Malabarba, Weitzman (2003) provided a better resolution for the phylogenetic relationships among the species described by Gomes by diagnosing a monophyletic group of characid fishes, referred therein as Clade A that included Odontostoechus lethostigmus. This hypothesis was corroborated by molecular (Javonillo et al., 2010;Oliveira et al., 2011) and morphological (Mirande, 2010) phylogenies that support the monophyly of Clade A. Mirande (2010) named that clade Stevardiinae, which was further corroborated by Baicere-Silva et al. (2011), who found that the spermiogenesis process in Odontostoechus lethostigmus is homologous with regard to other Stevardiinae.
Bryconamericus lethostigmus is a specialized riverine fish from South Brazilian coastal drainages, originally described from the rio Maquiné (rio Tramandaí drainage). Morphologically similar populations have been found on the northern neighbour drainages of rio Três Forquilhas (also from rio Tramandaí drainage), rio Mampituba and rio Araranguá. Malabarba (1998: 232) suggested that the population from the rio Araranguá may represent a new species. These river drainages comprise the Tramandai-Mampituba Freshwater Ecoregion (unit 335 - Abell et al., 2008;Fig. 1), an area of high endemism of fish species because of the congruent distributional pattern among several species that are solely found among these three drainages (Malabarba, Isaia, 1992). Recently, Hirschmann et al. (2015) showed that Diapoma itaimbe (Malabarba & Weitzman, 2003), another member of Stevardiinae, has a very strong genetic structure among these drainages, and that the rio Araranguá population is the most distinct among them. Bryconamericus lethostigmus and D. itaimbe are sympatric and share the same habitats with clear and cold waters over rocky substrates, leading to the hypothesis that B. lethostigmus may also have strong population structure among these drainages.
The alternative hypotheses of phylogenetic relationships of Bryconamericus lethostigmus and the possibility that one of its populations actually represents a new species demonstrate the need to redescribe this species and delimit its range of morphological and genetic variation. In this study, we redescribe Bryconamericus lethostigmus based on the examination of several populations including that of rio Maquiné from where the species was originally described (Gomes, 1947). Next, we describe and analyse the mouth shape and the origin of a single series of teeth in the premaxilla in B. lethostigmus, a species of a group characterized by the presence of two tooth series, and discuss its phylogenetic implications and possible homologies with other stevardiines. Finally, we analyse the genetic variation found in the populations of B. lethostigmus from the Tramandai-Mampituba ecoregion (Abell et al., 2008) to test if this species displays the same phylogeographic structure observed for D. itaimbe (Hirschmann et al., 2015), and if the rio Araranguá population represents a different species (Malabarba, 1998: 232).  (Abell et al., 2008). Squares represent localities of collection of B. lethostigmus for DNA analysis. Each mark may represent more than one lot.

Material and Methods
Morphological analyses. Measurements and counts were taken following Fink, Weitzman (1974), except for analfin base length measured between the origin of the first and last anal-fin rays; dorsal-fin base length measured between the origin of the first and last dorsal-fin rays; maxillary length measured from the contact with the premaxilla to its posterior tip; premaxillary length measured from the anterior tip to the contact with the maxilla; mouth width measured in the largest mouth gap. Measured specimens come from all four river drainages inhabited by this species (rio Araranguá -ARA; rio Mampituba -MAM; rio Três Forquilhas -FOR; and rio Maquiné -MAQ). Counts of vertebrae, teeth and procurrent caudal-fin rays were taken from cleared and stained specimens (c&s) prepared according to Taylor, Van Dyke (1985). We included the four vertebrae of the Weberian apparatus in vertebral counts, and counted the terminal centrum as a single element. We took measurements point to point with a calliper on the left side of specimens. We recorded all measurements as percents of standard length (SL) except for subunits of the head, which we recorded as percents of head length (HL). Sex of adult specimens of Bryconamericus was determined by the presence (males) or absence (females) of bony hooks in fin rays. Differences in measurements and counts among the four populations were tested with the Kruskal-Wallis test and post-test multiple comparisons using pgirmess package (Giraudoux, 2013) in R (R Development Core Team, 2013). We used the software Past (Hammer et al., 2001)

Molecular analyses.
We used tissue samples from 45 specimens of Bryconamericus lethostigmus collected throughout its distribution and maintained in 96% ethanol in the fish collection at Departmento de Zoologia, Universidade Federal do Rio Grande do Sul (UFRGS). Tissue samples are identified as "TEC" in the list of examined material and include individuals from all four river systems in the range of this species (rio Araranguá -ARA; rio Mampituba -MAM; rio Três Forquilhas -FOR; and rio Maquiné -MAQ).
DNA extraction followed a modified salt precipitation protocol (Medrano et al., 1990). For each sample we used PCR to amplify two mitochondrial genes: cytochrome oxidase subunit I (COI) and the NADH dehydrogenase 2 (ND2) and three nuclear genes: SH3 and PX domain-containing 3-like protein (SH3PX3), S7 ribosomal protein intron 2 (S72) (Cooke, Beheregaray, 2007) and myosin heavy chain 6 gene (Myh6) (Li et al., 2007). PCRs were carried out in 20μl reactions containing 10-50ng DNA, 0,2μM of each primer, 0,2mM of each dNTP, 1x Buffer, 1,5μM MgCl 2 and 1U Platinum Taq DNA polymerase (Invitrogen, São Paulo, BR). PCR conditions and primers are presented in (Appendix S1 -Available only as online supplementary file accessed with the online version of the article at http://www. scielo.br/ni). PCR products were checked by electrophoresis in agarose gel, purified using EXOSAP (Exonuclease I and Shrimp Alkaline Phosphatase GE Healthcare®, Piscataway, USA) and sequenced in both directions in Macrogen Inc, Seoul, South Korea.
Forward and reverse reads were assembled and visualized using Geneious 5.6.7 (Drummond et al., 2012). The consensus sequences were automatically aligned using the software CLUSTALW (Thompson et al., 1994) in BIOEDIT 7.1.3.0 (Hall, 1999) with default parameters. The mitochondrial coding genes COI and ND2 were concatenated for all analyses. Basic descriptive statistics, such as nucleotide (π) and haplotype diversity (hd) as well as neutrality tests were calculated in the software DNASPv5 (Librado, Rozas, 2009). Genetic distances among populations were calculated using the Kimura 2-parameter (K2P) substitution model (Kimura, 1980) using MEGA 6 (Tamura et al., 2013). Genetic structure among populations was quantified through the Analysis of Molecular Variance (AMOVA) carried out in the program ARLEQUIN 3.5 (Excoffier, Lischer, 2010). For this analysis, we considered individuals sampled in each river drainage as coming from different populations.
Evolutionary relationship among haplotypes were estimated based on the median-joining method (MJN) (Bandelt et al., 1999) using the program NETWORK 4.1.0.8 (www.fluxus-engineering.com). Estimates the mtDNA coalescence time and the historical effective population size (Ne) using the BEAST 1.8.2 package (Drummond et al., 2012). The mtDNA data set was analysed assuming an evolutionary rate of 0.01/site/Myr (Bermingham et al., 1997;Reeves, Bermingham, 2006;Ornelas-García et al., 2008) and a strict clock model, which is a generally well justified for analysis within a species or among a few closely related species (Li, Drummond, 2011). To compare the evolutionary history of B. lethostigmus to that of a sympatric species, we estimated coalescence times and population sizes for D. itaimbe based on sequences obtained from GenBank (KP399679 to KP399733; KP406648 to KP406702; Hirschmann et al., 2015). Based on the Bayesian Information Criterion in PartitionFinder (Lanfear et al., 2012) we assumed the HKY+I substitution model for COI and ND2. Ten million Markov chain Monte Carlo (MCMC) steps were performed for coalescence time estimation and samples were collected every 1,000 steps. The efficiency of the chain was assessed in Tracer 1.5 (Rambaut, Drummond, 2009) with 10% burn-in.  (Gomes, 1947)

Figs. 2-6
Odontostoechus lethostigmus Gomes, 1947: 07-12  Diagnosis. Bryconamericus lethostigmus is distinct from all other species of the genus by the following autapomorphy: presence of two rows of teeth in the premaxilla in small specimens (up to about 30 mm SL) progressively merging in one tooth row in the premaxilla in large specimens (more than about 40 mm SL) (vs. two tooth rows in the premaxilla regardless of body size). This species is distinct from all congeners by the atrophied upper lip in large specimens, leaving the premaxillary teeth exposed. This character is associated with its common name, "smiling tetra".
Description. Morphometric data is summarized in Tab. 1. Body moderately elongate and compressed. Dorsal profile slightly convex from head until dorsal-fin origin, nearly straight from posterior dorsal-fin base to adipose fin and slightly concave from adipose-fin base to caudal-fin origin. Ventral body profile slightly convex from head to anal-fin origin, straight along anal-fin base and slightly concave from posterior anal-fin base to caudal-fin origin. Greatest depth at dorsal-fin origin or somewhat anterior. Caudal peduncle slightly longer than deep. Dorsal and ventral profiles of caudal peduncle slightly concave. Mouth large, subterminal, lower jaw shorter than upper jaw and upper lip atrophied in large specimens, leaving the premaxillary teeth exposed (Fig. 3). Snout profile rounded. Premaxilla with two tooth rows in small specimens (up to about 30 mm of SL) and one tooth row in large specimens (more than about 40 mm SL) . Single tooth row corresponding to teeth of both rows merged into single series with more teeth than inner or outer series of teeth counted alone. Merging of tooth rows gradual; specimens with 30 to 40 mm of SL with well defined double or single series, or in most cases with teeth of inner and outer rows partially merged in single series . When present, inner row with four teeth with five cusps and outer row with three teeth with three cusps; teeth of inner row pedunculate and wider distally than teeth of outer row. Single premaxillary tooth series with seven, rarely five, six or eigth teeth, equal in size, pedunculate and anteroposteriorly compressed, with five cusps and sometimes one tooth with six cusps. Maxilla with three to seven teeth with three to five cusps; number of maxillary teeth increasing with body size (Fig. 5). Last tooth or two posteriormost teeth may be conical in small specimens. In large specimens two anterior maxillary teeth with almost same size of premaxillary teeth and all maxillary teeth exposed . Dentary with eight to eleven teeth, usually with five cusps, decreasing gradually in size posteriorly; last three teeth small with fewer cusps; last one or two teeth conical in some specimens (Fig. 6).   Intermediate stages of ontogenetic changes in the number of tooth rows of the premaxilla with both specimens showing two tooth series. a. smaller specimen with the outer series of teeth larger than those of inner series, and teeth irregularly arranged in two series; b. larger specimen with premaxilla expanded laterally and teeth of the outer and inner series equal in size and interposed at their bases, but not fully forming a single series. Dorsal-fin rays ii, 8 rarely 7, 9 or 10 (mode = 8; n = 318). Dorsal-fin insertion slightly posterior to ventral-fin origin. Adipose fin present. Anal-fin rays iii-v, usually iv or v, 13-19 (mode = 16, n = 317). Pectoral-fin rays i, 9-14 rarely 8 or 15 (mode = 12; n = 314). Pelvic-fin rays i, 6-8 rarely 5 or 9 (mode = 7; n = 316). Caudal fin forked, margin of lobes rounded and equal size. Principal caudal-fin rays 19, rarely 17, 18 or 20 (mode = 19; n = 301); 12-14 procurrent caudalfin rays dorsally (mode = 14; n = 8) and 9-13 ventrally (mode = 12; n = 8).
Statistical results. Some measurements and counts showed significant differences on means among the populations (Tab. 2), but without any repeatable pattern to distinguish any or a group of populations from all remaining populations. In agreement to the above mentioned results, PCA revealed no differences on measurements of specimens among populations (Fig. 7). Thus, populations from different river basins showed no morphologically significant differences among them.

Color in alcohol.
General ground body color yellowish olive (Fig. 2a) Color in life. Life color described from a specimen from rio Tramandaí drainage (Fig. 2b) (Marques et al., 2002: 28;Reis et al., 2003: 127) mainly due to habitat degradation. In the lastest list of endangered species from Rio Grande do Sul State, Brazil, B. lethostigmus was not classified in any threatened categories (Rio Grande do Sul, 2014) given that its population is apparently stable and no eminent threats were identified. The populations from the rio Araranguá, rio Urussanga and Mampituba drainages were not included in the list of threatened species from Santa Catarina State (CONSEMA, 2011). This species was also not included in the national list of threatened species (MMA, 2014).
We have tried to collect additional specimens of B. lethostigmus in the rio Urussanga and tributaries, the northernmost record of the species, but we were unable to find new specimens in this drainage. This river basin is currently very impacted by coal mining activities, which may affect the occurrence of the species.

Molecular results.
A total alignment of 1,700 base pairs (bp) was obtained for the mitochondrial genes (COI, 676 bp and ND2, 1,024 bp). Nucleotide and haplotype diversity were 0.0029 and 0.98, respectively. A total of 57 polymorphic sites defining 38 different haplotypes for B. lethostigmus were found. Neutrality tests were significant, rejected the null hypothesis of constant population size and/or no natural selection (Tajima's D -2.3318, P<0.01; Fu's FS -25.65, P<0.01). We obtained an alignment of 328 bp for S72, 700 bp for SH3PX3, and 759 bp for Myh6, but no variability was observed and, therefore, excluded from further analyses. New sequences generated in this study were submitted to GenBank (KX815171 to KX815260).
Mitochondrial data showed no association between haplotypes and drainages (Fig. 8) and no strong genetic structure among predefined populations. The haplotype network does not show any phylogenetic structure, and most haplotypes are connected by one or two mutations. However, there are no shared haplotypes among drainages, except for two haplotypes shared between FOR and MAM. Thus, the lack of geographic structure is due to a shallow genealogical structure rather than shared haplotypes. In agreement to the haplotype network, the AMOVA (Tab. 3) showed low isolation among river systems for the mtDNA data (F ST =0.08). Mitochondrial DNA genetic distance among populations (about 0.3%) was also very low (Tab. 4). The mtDNA coalescence time of B. lethostigmus was around 0.34 millions of years ago (Ma) (95% Highest Posterior Density [HPD] 0.15-0.58 Ma), significantly more recent than the estimate for D. itaimbe, which was around 2 Ma (95% HPD 0.97-3.5 Ma). The coalescence time for the MAM+FOR group was very similar for both species (0.32 Ma -95% HPD 0.14-0.56 Ma; and 0.31Ma -95% HPD 0.12-0.55 Ma for B. lethostigmus and D. itaimbe, respectively). Effective population size estimates were similar for both species, being 1.2 million (95% HPD 0.55-2.23 million) effective females for B. lethostigmus and 0.86 million (95% HPD 0.40-1.63 million) effective females for D. itaimbe. (Gomes, 1947 Fig. 8. Median-joining networks among haplotypes of Bryconamericus lethostigmus samples inferred by the concatenated mtDNA dataset (COI+ND2). Each circle represents a unique haplotype with circle sizes being proportional to their frequencies.

Material examined. Odontostoechus lethostigmus
Each colour represents a population. Crossed markers indicate the number of mutations between haplotypes.

Discussion
Both morphological and molecular data refute the hypothesis that the population from the rio Araranguá possibly constitutes a new species (contra Malabarba, 1998). We have found no clear morphological differences to distinguish the rio Araranguá or any of the studied populations from all other populations. We found statistical differences in the mean of some measurements or counts in pairwise comparisons, but these differences were not consistent among all populations to allow the diagnosis of any distinct population (Tab. 2). For example, the mean dorsal-fin depth is statistically different between the rio Araranguá and rio Três Forquilhas populations, but does not distinguish any of these populations from both the rio Maquiné and rio Mampituba populations. The shallow genealogical depth of the mtDNA combined with its lack of geographic structure, and the lack of genetic variation in the nuclear markers are also strong indicators that all B. lethostigmus populations constitute a single species.
Genetic Structure and Evolutionary History. Considering that B. lethostigmus and D. itaimbe occur in the same drainages, are syntopic, and share a similar life history, the different phylogeographic pattern found for both species is striking. While D. itaimbe shows a strong genetic structure with well-defined mtDNA clades (Hirschmann et al., 2015), B. lethostigmus showed no genealogical structure, and only a very weak genetic structure based on mtDNA data. Such phylogeographic difference between these species may be explained by three different scenarios, or a combination of them: (1) Different mtDNA evolutionary rates between the two species; (2) Different colonization times of the two species in these drainages; and (3) Gene flow among B. lethostigmus populations through paleaodrainages hampering population structure.
Given the differences on time to most recent common ancestor (TMRCA) for mtDNA, but the similar historical effective population size estimates, one could propose that B. lethostigmus had a mtDNA evolutionary rate ten times slower than D. itaimbe. Nabholz et al. (2008Nabholz et al. ( , 2009 showed that mtDNA evolutionary rate, based on cytochrome b sequence, exhibit a thirty-fold range variation across bird lineages and of hundred-fold variation across mammal lineages. However, such wide evolutionary rate variation was found among distinct taxonomic groups showing differences in generation time and metabolic rate. On the other hand, B. lethostigmus and D. itaimbe are relatively closely related species, with similar generation time and metabolic rates. Tringali et al. (1999) found only a 0.5-fold difference in mtDNA evolutionary rate for fish species of the same genus. Thus, a difference on evolutionary rate among B. lethostigmus and D. itaimbe seems an unlikely explanation for their distinct phylogeographic patterns.
Therefore, if these two species have similar mtDNA evolutionary rates, they probably had different evolutionary histories, as illustrated by their different TMRCA for mtDNA lineages, despite the similar historical effective population sizes. One possibility is that the ancestor of these two species colonized these drainages in different periods. The ancestor of D. itaimbe may have arrived in this ecoregion around 2 Ma, while the ancestor of B. lethostigmus only arrived around 0.34 Ma, and therefore, the lack of genetic structure among B. lethostigmus population may be due, at least in part, to a shorter period of evolutionary time. The ancestors may have reached their current distribution by headwater capture or through potential connections on the coastal plain during marine regressions. Based on present data it is not possible to state which dispersal route was followed by these species.
Another possibility for the lack of genetic structure in B. lethostigmus is a more intense gene flow among populations over time through palaeodrainages during the marine regressions compared to D. itaimbe. The reconstruction of palaeodrainage availability over time (Fig. 9) suggests that these drainages have been isolated only a short period of time, whereas during most of the time, these drainages had connections, as demonstrated by Thomaz et al. (2015b). Under this hypothesis, B. lethostigmus would be more efficient in maintaining genetic connectivity among populations than D. itaimbe. A caveat to this hypothesis is that if this species had used it these connections until very recently, this would result in a larger amount of shared haplotypes among drainages, which was not observed. There are two mitochondrial haplotypes shared between FOR and MAM populations of B. lethostigmus, while the other populations share no haplotypes. This is the same pattern observed for D. itaimbe. This common pattern of haplotype sharing observed for both species may be related to a single event that allowed gene flow between these two drainages, most probably headwater capture (Hirschmann et al., 2015), a process still active in the rivers of the coastal region (Ribeiro, 2006). According to our results, this event may have occurred around 0.3 Ma, probably close to the dispersal of the ancestor of B. lethostigmus across coastal drainages. Fig. 9. Variation in sea level for the last 500 ka shown according to Miller et al. (2011).
The hypothesis of headwater capture instead of dispersion through low land connections to explain the sharing haplotypes between the rio Três Forquilhas and rio Mampituba may be further supported by the lack of shared haplotypes between FOR and MAQ populations, which are currently connected through freshwater lagoons. The region is widely visited for field work, and there are no capture records for this species in lagoons. Therefore, it is likely that lagoons act as barrier to dispersion of B. lethostigmus between river systems similarly to what was proposed for D. itaimbe (Hirschmann et al., 2015).

Mouth Morphology.
The presence of a single tooth series in the premaxilla has been a key character in Eigenmann's (1915) definition of the characid subfamily Cheirodontinae, that have had up to 56 genera assigned along its history (Malabarba, 1998: appendix B), including Odontostoechus. The single tooth series in the premaxilla, however, has been demonstrated by direct homology (Malabarba, 1998) or by parsimony analyses (e.g. Paracheirodon Géry by Weitzman, Fink, 1983; Charax Gronow by Lucena, 1987; "the Rosy Tetras Clade" by Weitzman, Palmer, 1997;Xenurobryconini by Weitzman, Fink, 1985;Netto-Ferreira et al., 2013) to originate several times and independently in different genera of the Characidae. Malabarba (1998) identified three non-homologous origins of a single tooth series in Characidae: (1) the merging of both rows into a single one; (2) the reduction of the anterior tooth row, with one or even two teeth remaining and not perfectly aligned in a single series; and, (3) the complete loss of the anterior tooth row. According to Malabarba (1998) the single tooth row in the B. lethostigmus originated with the merging of the external tooth row and the inner row during ontogeny. As such this character is non-homologous with the single tooth row observed ontogenetically in Cheirodontinae. This observation is corroborated in this study where we provided the description and illustration of the merging of the two tooth rows in a single series along ontogenetic development without tooth loss in B. lethostigmus .
Among stevardiines, relationships of B. lethostigmus have been alternatively assigned to two groups of genera. Phylogenies based on comprehensive samples of characid taxa (Mirande, 2010;Javonillo et al., 2010;Oliveira et al., 2011;Thomaz et al., 2015a) have found the type-species of Odontostoechus grouped to at least one representative of the genus Bryconamericus sensu stricto (Thomaz et al., 2015a), including the genera Hypobrycon and Nantis (currently junior synonyms of Bryconamericus). Alternatively, scenarios based on primary hypotheses of homology through direct comparisons of characteres related to mouth morphology have resulted in completely different hypotheses, grouping B. lethostigmus to genera such as Bryconacidnus, Ceratobranchia, Monotocheirodon, Othonocheirodus, Rhinopetitia, and Rhinobrycon (Böhlke, 1954;Menezes et al., 2013;Netto-Ferreira et al., 2014). However, these hypotheses are based on direct observations and have not been subjected to congruence tests of homology (de Pinna, 1991).
We herein refute the hypothesis of primary homology between mouth modifications found in B. lethostigmus and in the stevardiine genera Bryconacidnus, Ceratobranchia, Monotocheirodon, Otonocheirodus, Rhinopetitia, and Rhinobrycon. The tooth row reduction in the premaxilla during ontogeny observed in B. lethostigmus represents a unique condition among the Stevardiinae. The juvenile stage in B. lethostigmus, characterized by two premaxillary tooth series, is identical and homologous to the juvenile and adult stages in the species of Bryconamericus sensu stricto. The adult stage in B. lethostigmus, however, is characterized by a single tooth series that includes the teeth of both rows merged, is a novelty and an autapomorphy of this species. It can be characterized both by the heterochronic process of peramorphosis related with the addition of one step on the development of the premaxilla as well as by heterotopy by the changes in the spatial distribution of the teeth of the inner and outer series in a final stage of growth (Zelditch, Fink, 1996). We can recognize the final condition in B. lethostigmus is the result of a causal sequence of events (merging tooth series) -feature A induces feature B -and it is not only a temporal change.
Although mouth shape in adult B. lethostigmus remarkably resembles that of Monotocheirodon (Menezes et al., 2013: figs. 2, 10, 14) they have different origins and are homoplastic. The mouth modifications in Monotocheirodon species does not originate from ontogenetic changes in mouth morphology. Instead, the single tooth series in the premaxilla of Monotocheirodon species apparently originates from the complete loss of the external tooth series, since all the species retain a single series of four teeth that seems to be homologous to the four teeth in the inner premaxillary tooth series found in the stevardiines (Menezes 14 e160073 [14] et al., 2013, fig. 14, illustrated three premaxillary teeth in Monotocheirodon kontos, but the description refers to the presence of four teeth in all specimens).
Another diagnostic character of B. lethostigmus is the upper lip atrophied in large specimens, leaving the premaxillary teeth exposed. Its popular name, "smiling tetra", was due to this character. However, as this character is apparent only in large specimens, young and small specimens do not "smile". The atrophied upper lip is shared with Deuterodon stigmaturus (Gomes), which is sympatric to B. lethostigmus, D. rosae (Steindachner), Henochilus wheatlandii Garman, Psalidodon gymnodontus Eigenmann, Bryconacidnus, Ceratobranchia, Monotocheirodon, Othonocheirodus and Rhinopetitia (Lucena, Lucena, 2002;Netto-Ferreira et al., 2014). In D. stigmaturus, the atrophied upper lip exposes the upper maxilla as in B. lethostigmus. Lucena, Lucena (2002) also observed that this character is an independent acquisition on these species. On the other hand Netto-Ferreira et al. (2014) suggested that this character plus the morphological similarity of outer series teeth of the premaxilla were strong evidence towards a relationship among the stevardiines B. lethostigmus, Bryconacidnus, Ceratobranchia, Monotocheirodon, Othonocheirodus, Rhinobrycon and Rhinopetitia. Again, we refute this hypothesis of primary homology due to differences in the origin of the single tooth series and lip reduction during the ontogenetic development of these species. This hypothesis also fails considering the phylogeny based on molecular data presented by Thomaz et al. (2015a), showing Bryconamericus lethostigmus more closely related to other Bryconamericus species than to the genera listed above.