A redescription of Astyanax gymnodontus ( Eigenmann , 1911 ) , new combination , a polymorphic characid fish from the rio Iguaçu basin , Brazil

Astyanax gymnodontus (Eigenmann, 1911), new combination, a polymorphic characid fish endemic to the Rio Iguaçu basin in Southern Brazil, is redescribed. The species was originally described as belonging to the monotypic genus Psalidodon, distinguished from Astyanax by the possession of a single series of teeth on the premaxilla (vs. two series), by the presence of relatively compressed teeth on the premaxillary and dentary bones (vs. relatively bulky teeth), and by the absence of lips (vs. lips present). Examination of a large number of specimens from the rio Iguaçu basin that present dental and lip features that were intermediate between Psalidodon and Astyanax indicated that specimens attributed to Psalidodon show extreme characters observed within a polymorphic Astyanax species. Accordingly, the genus Psalidodon is synonymized under Astyanax. Specimens of “Astyanax sp. F” morphotype from the rio Iguaçu basin belong to the species Astyanax gymnodontus and display a regular tooth arrangement and morphology, and their lip development is similar to that of other Astyanax species.

In the Nupélia samples, specimens that perfectly fit the original description of P. gymnodontus were collected, as well as several Astyanax specimens.The species of these two genera can be separated by their teeth and lips; P. gymnodontus presents only one row of premaxillary teeth and no lips, while all Astyanax species possess two rows of premaxillary teeth, with five (rarely four) teeth in the inner row, three or four (rarely two or five) teeth in the outer row, and well developed lips (Garavello et al., 1997).Among specimens attributed to the Astyanax species, several which were temporarily called "Astyanax sp.F" by Sampaio (1988) and Garavello et al. (1997) present very similar overall features when compared with P. gymnodontus, except from their lip/ teeth characters.Nevertheless, many intermediate specimens between these two morphotypes were collected and analyzed as well.Their lips range from being well-developed to being poorly developed or absent.Each of these types of lips can be combined with one or two rows of well-positioned premaxillary teeth, or even with an inner series of two along with one, two, three or four teeth forming poorly defined outer row randomly arranged on one or both sides of the superior jaw.Garavello et al. (1997) arbitrarily categorized intermediate specimens with one single row of premaxillary teeth and both well and poorly developed lips as Psalidodon sp.The similarity in the overall features of the three morphotypes ("Astyanax sp.F", "Psalidodon sp.", and "P.gymnodontus") generated the suspicion that they could belong to a single species with a great buccal polymorphism.The main scope of this study is to test whether all these specimens are different species or whether they belong to a single polymorphic species.

Material and Methods
Meristic and morphometric data traditionally used in characid systematics were taken from 30 specimens in each of the three morphotypes mentioned above.Trusses composed by 21 measurements were also taken (Fig. 1), as suggested by Strauss & Bookstein (1982).
Traditional measurements are presented as ratios of standard and head lengths.These measurements, together with trusses, were analyzed using both principal components and canonical discriminant analyses, according to Humphries et al. (1981) and Reis et al. (1990), in order to detect any effective differences between the three morphotypes compared.In the principal components analysis, the first component always has the highest percentage of variation and, if the eigenvectors are all positive, this is interpreted as a consequence of the size differences among specimens, as was the case here.This indicates that each variable contributes differently to the size discrimination.The remaining components are interpreted as being representative of body form, according to Jolicouer & Mosimann (1960).In the second component, eigenvectors of the variables that contributed to the faint separation between both Psalidodon and the Astyanax morphotypes have the highest modular values indicated in bold in Table 1.For these variables, we calculated ratios against all other measurements in order to establish ratios that could diagnose the morphotypes, but all of the calculated ratios overlapped.
Canonical discriminant analysis maximizes differences between the previously defined morphotypes (Garavello & Reis, 1988), and in this study suggested several unusual ratios between the measurements.The pairs of variables for the ratios were selected by a Pearson's correlation between canonical variables and used residuals of the regression for each character in the principal components analysis.The ratios suggested by that analysis are shown in Table 2. Analyses treating traditional and truss measurements separately were also performed and resulted in even more overlapping results, and hence are not presented.Both analyses were calculated using SAS ® software and the routine proposed by Reis et al. (1990).
Seven specimens were radiographed, and five were cleared and stained for osteological observations following the methods of Taylor & Van Dyke (1985).In counting the total number of vertebrae, the fused PU1+U1 was considered to be a single bone and the vertebrae incorporated into the Weberian Apparatus were counted as four elements.

Results
In an attempt to incorporate all the intermediate specimens between traditional Astyanax and Psalidodon into the previous classification of Garavello et al. (1997), we temporarily identified all specimens with one or two teeth in the outer premaxillary row on at least one of the sides of the mouth as "Psalidodon sp." morphotype.This and the other two morphotypes ("Astyanax sp.F" and "P.gymnodontus") were compared herein with the aim of confirming whether they are different or belong to a same species.The principal components analysis indicated that the morphotypes characters totally overlapped, as shown in Fig. 2 and Table 1.
A very tenuous separation between both Psalidodon morphotypes and the "Astyanax sp.F" morphotype is suggested in axis 2 of Fig. 2. The measurements that contributed to this axis were interorbital distance, the distance between the snout and isthmus, and head length, as shown in Table 1.These metrics had calculated ratios against other measurements presented in Table 2, which also overlapped.
Even when the differences between the previously defined morphotypes were maximized by using the canonical discriminant analysis, all measurements also showed overlapping characters, with only a slight separation observed between both Psalidodon morphotypes and the "Astyanax sp.F" morphotype on axis 1 (Fig. 3).
An extensive list of the most important pairs of variables suggested by this analysis also shows that the ratios of these variables overlapped (Table 2).Meristic data of the morphotypes are presented in Table 3 and show that ranges overlapped except for counts of premaxillary teeth in the outer row, which were previously chosen arbitrarily for separating the intermediary specimens between the traditional Psalidodon and Astyanax.
Based on the meristic analyses, no consistent difference was found between the three morphotypes analyzed since there were overlaps of all data, except for the premaxillary Fig. 1.Landmarks (letters) determined for delimiting trusses, and measurements (numbers) between them (adapted from Fowler, 1948).Descriptions of measurements are provided in Table 1.teeth of the outer row.In addition, despite some different averages, all the calculated ratios were similar among the morphotypes, including those suggested by the principal components and canonical discriminant analyses.Based on these data, the three morphotypes are herein considered to belong to the same species, which presents a high buccal polymorphism that had not been described previously for Astyanax species.Accordingly, the dental and soft buccal tissue characters that were thought to be unique to Psalidodon are rather the extreme end of the variation observed within a polymorphic species that was otherwise not distinguishable from species currently assigned to the genus Astyanax, and a new nomenclatural combination, Astyanax gymnodontus, is proposed herein.The genus Psalidodon is considered to be a junior synonym of Astyanax, the oldest genus, and a redescription of the species is provided below.Garavello et al., 1997: 74 (description;Segredo Reservoir).
Diagnosis.Astyanax gymnodontus can be distinguished from all other congeners at the rio Iguaçu basin except for A. varzeae by the combination of the following characters: thirdinfraorbital margin separated from preopercle by broad skin area as wide as 1/5 to 1/3 of its depth (vs.adjacent or separated by a narrower skin area); teeth with a cutting edge forming a straight line or a convex arc in frontal view (vs.concave arc); premaxillary teeth with five (rarely seven) cusps, with the median being the largest and the other cusps abruptly decreasing in size to the sides (vs.commonly seven, rarely four, five or nine cusps, gradually decreasing laterally); transversally located black humeral blotch (vs.longitudinally located); 21 to 24 anal-fin rays and 37 to 41 lateral-line scales (vs.18 to 22 and 34 to 38, respectively, in A. totae); and no hooks on pelvic or anal fins.Astyanax varzeae is the congener most similar to A. gymnodontus in the basin, but the latter can be diagnosed by having a deeper body, 2.4 to 3 times in the SL (vs. 3 or 3.5 in A. varzeae), and a longer snout, 3.2 to 4.7 times in the HL (vs.4.3 to 6.1).
Description.Body elongated and laterally compressed, body deeper at dorsal-fin origin.Predorsal and prepelvic regions of body rounded transversely.Dorsal profile of snout convex, straight or slightly concave from eye to supraoccipital process.Dorsal profile of body gently convex from supraoccipital process to dorsal-fin end, straight or slightly convex from this point to adipose-fin base, straight or slightly concave from this point to caudal fin.Ventral profile of body convex, between anterior tip of jaw and pelvic-fin origin, straight from this point to anal-fin origin.Dorsal-fin origin slightly forward or at vertical through pelvic-fin origin, its distal margin straight.Pectoral-fin origin at vertical through posterior edge of opercle, its margin rounded, when adpressed, reaching or almost reaching pelvic-fin origin.Pelvic-fin origin at about vertical through dorsal-fin origin, its margin slightly convex, when adpressed not reaching anal-fin origin.Anal-fin origin just behind vertical through end of dorsal-fin base, its distal margin slightly concave.Caudal fin with symmetric lobes, or inferior slightly longer, and scales only at base.Dorsal-fin rays ii or iii,8 to 10 (mean = 9.4); pectoral-fin rays i,11 to 14 (mean = 12.5); pelvic-fin rays i,7 or 8 (mean = 7.5); anal-fin rays iii,19 to 22 (mean = 20.8).Lateral line complete, with 37 to 41 perforated scales, generally 39 or 40; 6 to 8 rows of scales above lateral line, generally 7, and 5 or 6 below.Third-infraorbital margin separated from preopercle by a broad skin area, as wide as 1/5 to 1/3 of its depth.Dentary larger teeth 5 to 7, with three cusps, in anterior row, and smaller 4 to 7, generally 5 or 6 in posterior row.Premaxillary teeth 4 to 7, generally 5 or 6, with five cusps, in inner row, 1 to 4 with three cusps regular or irregularly arranged in outer row, when present.Outer row commonly not bilaterally symmetric.Maxillary teeth 1 to 4 with three cusps.All teeth with largest median cusp, abruptly decreasing in size up to smallest more laterally placed.Cutting edge of teeth with cusps in a straight Sexual dimorphism.Secondary sexual characters were not found in any of the examined specimens.
Distribution.Astyanax gymnodontus is only known to occur in the rio Iguaçu basin in Southern Brazil (Fig. 6).Eigenmann (1911) described Psalidodon gymnodontus based on three specimens from Porto União, rio Iguaçu basin, which represent the extreme end of the variation displayed by Astyanax gymnodontus.Redescribing P. gymnodontus including all morphotypes would not allow an effective differentiation from the Astyanax species, since one of the morphotypes is completely coincident with the features of that genus.Therefore, considering the buccal polymorphism found in that species, we suggest that Psalidodon should be considered to be a junior synonym of Astyanax.Preliminary molecular analyses carried out by A. Prioli (pers.comm.)corroborates that the three morphotypes likely represent a single species, since they are much more similar (within an acceptable range for populations of a same species) than with the remaining Astyanax species from the rio Iguaçu basin.

Discussion
Polymorphism in trophic structures is still poorly known among teleosts.Smith & Skúlason (1996) published an extensive revision on resource polymorphism on fishes, amphibians and birds, mentioning several examples in freshwater fishes of North and Central America, Europe and Africa.However, the only case they cited in the Neotropics was the trophic polymorphism that was well described for parodontid trans-Andean genus Saccodon, published by Roberts (1974).In characiforms, aside from Saccodon species, Lucena (2007) described a polymorphism in oral features of some species of the Roeboides species belonging to the R. affinis species-group.Astyanax gymnodontus represents the third case of buccal polymorphism in characiforms.
According to Hahn et al. (1997), this species has a relatively broad trophic spectrum at the Segredo Reservoir, composed of vegetal and animal materials.Hahn et al. (1997) also observed the morphotypes "Astyanax sp.F" and "Psalidodon sp.", which possess an outer row of premaxillary teeth, and feed mainly on vegetal materials, while "P.gymnodontus" feeds mainly on terrestrial coleoptera.This suggests that the polymorphism allows that species to explore further feeding habitats (Soulé & Stewart, 1970).Hence, the buccal polymorphism revealed here is also attributed to trophic factors.The continuous replacing of teeth rows during the life cycle could allow the transformation from one morphotype to another.Actually, a few specimens with a single series of teeth on one side of the premaxilla and two rows on the other side were observed, but not preserved.Nevertheless, no relationship between these transformations and an environmental variable was found by comparing their occurrence with available environmental data.Further studies focusing on trophic anatomy and behavior, as well as environmental experiments examining trophic resource abundance would improve our understanding of this feature.
The rio Iguaçu ichthyofauna is well known for its high endemism and consequently for being very distinct in comparison to the remaining rio Paraná fish fauna because of the very high waterfalls situated upstream from its mouth into the rio Paraná, which is an effective barrier to the dispersal of fish species (Garavello et al., 1997).There are many genera or even families that are widely distributed in the rio Paraná basin, including several large migratory fish species which are absent in the rio Iguaçu basin (Agostinho et al., 1997a).Astyanax is the most diversified fish genus in the rio Iguaçu basin.In addition, Astyanax species are the most abundant fishes in this basin, comprising almost 70% of fish species in a study carried out in Segredo Reservoir, and almost 80% of the total fish captured in the Foz do Areia Reservoir (Agostinho et al., 1997b).In this situation, it would be very beneficial ecologically for one congener to enlarge its feeding habitat to avoid competition.Severi & Cordeiro (1994); and dot = type-locality of Psalidodon gymnodontus.Some symbols represent more than one lot or locality.

Table 1 .
Variable loadings for the principal component analysis of the three morphotypes of Astyanax gymnodontus.Truss measurements (numbers) are indicated with landmarks (letters) shown according to Fig. 1.The following abbreviations were used: maxil.comm.: maxillary commissure; supraoc.process: supraoccipital process.The modular higher eigenvectors of principal component 2 are shown in bold (see text for details).

Fig. 2 .
Fig. 2. Plots of scores obtained by principal componentsanalysis of all measurements from the three morphotypes.

Fig. 6 .
Fig. 6.Map of Southern Brazil and adjoining countries showing the geographic distribution of Astyanax gymnodontus in the rio Iguaçu basin (rectangle in focus).Bold lines are national borders.Squares = examined material; triangles = distribution showed bySeveri & Cordeiro (1994); and dot = type-locality of Psalidodon gymnodontus.Some symbols represent more than one lot or locality.

Table 2 .
The ranges of standard length (SL) and ratios suggested by canonical discriminant analysis are shown for the three morphotypes of Astyanax gymnodontus.Abbreviations of truss measurements are shown according to Table1, and traditional measurements are: BOD, body depth; CPD, caudal-peduncle depth; CPL, caudal-peduncle length; DFL, dorsal-fin length; HL, head length; IOD, interorbital distance; JW, jaw width; OD, orbital diameter; PDL, predorsal length; and SNL, snout length.Bolded ratios correspond to those responsible for a faint separation between morphotypes in axis 2 of the principal components analysis shown in Fig.2.
Fig. 3. Plots of scores obtained by canonical discriminant analysis of all measurements from the three morphotypes.

Table 3 .
Counts from 30 specimens of the 3 morphotypes of Astyanax gymnodontus, with the arbitrary difference in the premaxillary teeth of the outer row shown in bold.