Phylogenetic relationships of Chanidae ( Teleostei : Gonorynchiformes ) as impacted by Dastilbe moraesi , from the Sanfranciscana basin , Early Cretaceous of Brazil

Fossil gonorynchiform fishes range from the Lower Cretaceous to the early Miocene, and are represented by a few dozen living species. The order is currently divided into two major clades: Gonorynchoidei, which includes the families Gonorynchidae and Kneriidae, and Chanoidei, encompassing a single family, Chanidae, with a single recent species, the Indo-Pacific Chanos chanos, and several fossil taxa. Chanidae includes some poorly known taxa, such as Dastilbe moraesi, described from the Aptian (Lower Cretaceous) of the Areado Formation, Sanfranciscana basin, Brazil. This species is currently considered to be a junior synonym of the type species of its genus, Dastilbe crandalli, from Santana Formation, Aptian, northeastern Brazil. The analysis of abundant D. moraesi specimens revealed several new morphological features, many of which had previously been misinterpreted. Dastilbe moraesi was incorporated into a gonorynchiform character matrix as revised and modified for the Chanidae. We obtained a single most parsimonious tree in which D. moraesi is distinct and phylogenetically apart from D. crandalli. According our analysis, D. moraesi forms a sister pair with Chanos, a clade which is closely related to Tharrhias, all composing the tribe Chanini.


Introduction
Gonorynchiforms have a rich fossil record ranging from the earliest Cretaceous (Barriasian-Valanginian) to the earliest Miocene (Fara et al., 2010), and include several extant representatives.The order is divided into two main clades: suborder Gonorynchoidei, which includes the families Gonorynchidae and Kneriidae, and suborder Chanoidei, with a single family, Chanidae (Poyato-Ariza et al., 2010).

e180059[2]
The Chanidae, with a single extant species, the Indo-Pacific milkfish Chanos chanos, and several fossil taxa, is a well distinguished clade, diagnosed on the basis of several synapomorphies such as: presence of a large, very broad, concave-convex premaxilla with a long oral process; enlarged maxilla with a posteriorly swollen to a bulbous outline and a curved posterior border; presence of a notch in the anterodorsal border of the dentary; quadrate-mandibular articulation anteriorly displaced and located anterior to the orbit by displacement of the quadrate bone and correlated elongation of the symplectic; metapterygoid process of hyomandibular present, placed on anterior border of the bone; opercular bone expanded, at least one-third of head length; angle formed by preopercular limbs straight to acute; preopercular expansion present (except in Dastilbe); suprapreopercular bone present as a relatively large, flat bone; and the presence of unmodified neural arches anterior to dorsal fin autogenous in adults individuals, at least laterally (Poyato-Ariza et al., 2010).
Despite being a well-defined clade, Chanidae still includes some problematical taxa, in relation to having an undetermined phylogenetic position and concerning the specific taxonomic composition of some genera, such as Dastilbe (Poyato-Ariza et al., 2010).Dastilbe, in particular, presents a complicated taxonomic situation.At present, according to Poyato-Ariza et al. (2010), the real problem of the genus Dastilbe is the uncertainty of its specific composition and the purportedly variation of its characters.The taxonomy of Dastilbe is far from being solved inasmuch as its species are poorly characterized and the differences among them are said to be very slightly (Poyato-Ariza, 1996a).
The original description of Dastilbe crandalli by Jordan (1910) is little informative and quite inaccurate, and its illustrations are of poor quality, being therefore of limited utility for its distinction from the species described subsequently.Silva Santos (1947) provided no diagnostic character for Dastilbe elongatus, except for the vague mention that its vertebral count is distinct from that of Dastilbe crandalli.Davis, Martill (1999) mentioned that the different localities and preservational environments probably led Silva Santos (1947) to distinguish Dastilbe elongatus from Dastilbe crandalli.However, there is no reference to these criteria in that Silva Santos's paper (1947), direct or indirect.
Silva Santos (1955) mentioned that Dastilbe moraesi may be separated from its congeners by a combination of characters, but none of them are actually distinctive, either in separate or in combination.Taverne (1981) mentioned that all nominal species, including Dastilbe moraesi, are junior synonyms of a single species, the type of the genus, Dastilbe crandalli.However, his suggestion lacks any factual evidence since no material of Dastilbe moraesi had been examined.In revising species of Dastilbe from the Santana Formation, northeastern Brazil, Blum (1991a) argued for the existence of two sympatric species in the Crato member, northeastern Brazil, which were previously identified as Dastilbe elongatus.He proposed to apply the epithets elongatus and crandalli to the larger and smaller specimens, respectively.Nothing was said about the distinction of these species from Dastilbe moraesi.Poyato-Ariza (1996a) cast doubts on the criteria used by Blum (1991a) -differences in body size and meristicsto distinguish specimens of Dastilbe crandalli and D. elongatus from the Crato member, stating that there is a single species in that geological unit, which would be Dastilbe elongatus.In addition, Poyato-Ariza (1996a) suggested that the specimens of Dastilbe crandalli from Riacho Doce, State of Alagoas, its type locality, were not co-specific with small specimens of Dastilbe from the Crato member, differing in details of caudal skeleton.Indeed, Poyato-Ariza (1996a) was also the first author to explicitly distinguish Dastilbe moraesi from the other species of Dastilbe on the basis of peculiar traits of the mandibule, preopercle, and caudal skeleton.However, he warned that the validity of Dastilbe moraesi on the basis of these features depended on further investigation because he analyzed a single, imperfectly preserved specimen.Davis, Martill (1999) carried out a comprehensive investigation about the taxonomy of Dastilbe, concluding that all diagnostic features reported so far to distinguish Dastilbe batai, Dastilbe crandalli, and Dastilbe elongatus, i.e., size, meristics, fin positions, and details of caudal skeleton, are within the range of variation of a unique, highly plastic species, which would be Dastilbe crandalli.Davis, Martill's (1999) opinion on the validity of Dastilbe moraesi is controversial.The characters thought to be diagnostic for Dastilbe moraesi by Poyato-Ariza (1996a) were refuted by Davis, Martill (1999), who asserted that "… no criteria have been found to satisfactorily define more than one species of Dastilbe".On the other hand, Davis, Martill (1999) do not reject Dastilbe moraesi as a valid species, but note that further studies are required.Their contention may be explained by the fact that no specimen of Dastilbe 3 e180059[3] moraesi had been examined.Grande, Poyato-Ariza (1999) retained Dastilbe moraesi as a valid species, although these authors studied the same specimen examined by Poyato-Ariza (1996a).
Recently, Amaral, Brito (2005) proposed to redescribe Dastilbe from Gondwana but no data were offered regarding the validity of its species.Ribeiro et al. (2005), on the basis of newly collected material, re-examined the status of Dastilbe moraesi and supported it as valid.Dietze (2007) agreed that Dastilbe elongatus is a junior synonym of the type species but argued that Dastilbe moraesi is too poorly known to arrive at any conclusion in this respect.Finally, Brito, Amaral (2008) concluded that, due to the high plasticity of the characters, it is impossible to distinguish any nominal species from the type species.However, as pointed out by Poyato-Ariza et al. (2010), none of these studies conducted a re-examination of type material of any nominal species of this genus, which is, according to these authors, an indispensable requirement to achieve more definitive taxonomic conclusions.
Based on the analysis of primary types of Dastilbe moraesi and Dastilbe elongatus (a junior synonym of Dastilbe crandalli) (Figs. 1 and 2; see examined material), as well as abundant additional material of D. moraesi recently collected in the type locality during two field trips (in 2005 and 2011), we were able to perform a detailed morphological reanalysis of Dastilbe moraesi in order to define its status and phylogenetic position.
Recently collected specimens from the Sanfranciscana basin were prepared mechanically with needles under the stereomicroscope.Museum specimens from Araripe basin were acid and mechanically prepared, following the methods of Toombs, Rixon (1959).Photos were taken with a digital camera (Motican 2500) attached to a SMZ 800 Nikon stereomicroscope.Drawings were made using a camera lucida attached to a SMZ 800 Nikon stereomicroscope as well as from digital photographs.Anatomical nomenclature follows Poyato-Ariza et al. (2010).Standard length (SL) refers to the traditional standard length, measured from snout tip to caudal peduncle at origin of caudal fin.Total fragment length (TFL) refers to the total fossil fragment length, when it is only partially preserved.Concerning Nanaichthys, all characters were coded as in Amaral, Brito (2012: tab. 1), except when we were able to reinterpret a character based on the illustrations and photographs presented in their paper (this is particularly relevant in the reinterpretation and character coding proposed by us regarding the morphology of the premaxilla and maxilla in Nanaichthys, Fig. 3); the new characters were coded based on the information (description and illustrations) provided in Amaral, Brito (2012) (see specific comments in the character list below).Characters of the remaining terminal taxa were coded exactly as in Poyato-Ariza et al. (2010), except when indicated.
Phylogenetic analysis was performed in TNT (Goloboff et al., 2008) using the "New Technology Search" option in the default settings.Multi-state characters were treated as unordered.
The inclusion of D. moraesi in the Chanidae (Node 4) is supported by several synapomorphies, such as: presence of a large, very broad, concave-convex premaxilla with a long oral process (C.27:1); the posterior region of the maxilla very enlarged, swollen and having a bulbous outline, with curved posterior border (C.33:1); quadratemandibular articulation located anterior to orbit, with the quadrate displaced but not elongate (C.48:1); presence of an elongated symplectic (C.49:1); presence of a metapterygoid process of hyomandibula (C.52:1); presence of an expanded opercular bone, with at least one third of the head length (C.54:1); by having an preopercular expansion present on preopercle posteroventral corner and part of the posterodorsal limb (C.61:1); and the presence of autogenous hemal arch in preural centrum 2 (C.103:1).
Nanaichthys, according to our analysis, is the sistergroup of the clade formed by the remaining taxa included in Node 6, forming together the subfamily Chaninae.
Dastilbe crandalli is the sister-group of Chanini (including Tharrias, Dastilbe moraesi and Chanos) based on the presence of the anterodorsal border of the dentary slightly concave, forming a relatively high symphysis (C.35:1).
The inclusion of Dastilbe moraesi within the Chanini (including Tharrhias and Chanos) is based on the presence of exoccipitals with a posterior concave-convex border, with a projection above the basioccipital (C.5:1), a large mesethmoid with broad posterolateral wing-like expansions (C10:1), and the presence of a neural arch and spine of the first preural centrum both well developed, with the spine about half as long as preceding ones (C.92:1) (Fig. 8).The sister group relationship between Dastilbe moraesi and Chanos (Node 11) is supported by the presence of the posterolateral expansion of exoccipitals (C.4:1), and the absence of postcleithra (C.85:1).Dastilbe moraesi is phylogenetic diagnosed from the remaining Chanidae by the autapomorphic condition of its caudal-fin skeleton, characterized by the presence of only five hypurals (Fig. 9).

Discussion
Dastilbe moraesi, a small gonorynchiform fish from the Early Cretaceous of the Sanfranciscana basin in southeastern Brazil, has been considered one of the most obscure taxa previously assigned to the family Chanidae.Indeed, the lack of sufficient material was the main constraint toward reaching a more definitive conclusion regarding its phylogenetic affinities and taxonomic status.The analysis of abundant material from the type locality, Fazenda São José do Geribá, Presidente Olegário, Minas Gerais state, Brazil, filled this gap and provided vast material for observation of detailed morphological data as the quality of preservation of the material is outstanding.Amaral, Brito, 2012) was not supported herein.According to our results, Nanaichthys is more closely related to Aethalionopsis, Parachanos, Chanos, and the Brazilian fossil gonorynchiforms than it is to the Rubiesichthyinae (Gordichthys + Rubiesichthys).Some anatomical characters were either misinterpreted by those authors (e.g., numbers 30 and 32 herein) or assumed without enough evidence (e.g., numbers 62 and 75 herein).As a consequence, the complex palaeobiogeographical implications proposed by Amaral, Brito (2012), encompassing influences of the Caribbean Tethys, Mediterranean Tethys and Laurasia prior to the Aptian/Albian marine transgressions over the continental terrains of Western Gondwana, are not supported as well.
Concerning Aethalionopsis, our results are compatible with those presented by Poyato-Ariza et al. (2010), in which this genus is the sister-group of the clade formed by the remaining Chaninae, as supported in this study by the presence of an approximately straight angle formed by preopercular limbs (C.58:1).Parachanos, Dastilbe, and the clade Chanos + Tharrhias formed an apical polytomy within Chanidae in Poyato-Ariza et al. (2010) and Amaral, Brito (2012).In our analysis, Parachanos is the sister-group of the clade that includes Dastilbe crandalli +(Tharrhias +(Dastilbe moraesi + Chanos)), which (Node 9) is supported by the presence of a dentary symphysis higher than immediately posterior part of the dentary and robust (C.35:1).
Dastilbe moraesi is morphologically more similar to the living Chanos chanos than previously thought.In the present analysis, several new similarities shared between both taxa were observed, such as the very similar morphological aspect of the anterior maxillary process (Fig. 6).In both D. moraesi and Chanos chanos, for instance, the anterior portion of maxilla has characteristic processes which is very similar in position and degree of development.The number of processes in the anterior portion of the maxilla is not included as a distinct character because the condition is very difficult to be properly assessed in most fossil chanids, since the anteriormost part of the premaxillary process of the maxilla is covered by the premaxilla in lateral view, but it may constitute an additional character supporting the sister-group relationships between D. moraesi and Chanos.The same is true for the degree of posterior development of the exoccipitals (also shared by Tharrhias).
In conclusion, our results are interesting not only because it reveals that Dastilbe moraesi is a valid species, as previously proposed by Ribeiro et al. (2005), but also that it is not the sister group of D. crandalli (assuming that this species is a senior synonym of D. elongatus analyzed in this study).Significant individual variation within each species, as suggested by Davis, Martill (1999) and reiterated by Brito, Amaral (2008), was not confirmed in our study since all specimens of both D. crandalli and D. moraesi are found to be very conservative in their morphological and meristic diagnostic attributes.As for the nomenclature to reflect the topology obtained, it is necessary to erect a new genus for D. moraesi, which will be done in a forthcoming account in which the morphology of D. moraesi will be treated in greater detail.

Fig. 4
Fig. 4. a. Single most parsimonious tree (CI= 0.743; RI= 0.733; 166 steps long) depicting phylogenetic relationships of Dastilbe moraesi and other chanids.Clades 1 to 11 are supported by synapomorphies described in S2; b. highlighted vertical branch length expresses the occurrence of taxa in the fossil record.

Fig. 7 .
Fig. 7. Opercular region of Dastilbe moraesi (CPUFMT 745, 16.3 mm TFL) in lateral view, anterior to left: Synapomorphies of Chanidae observed are: opercular bone is expanded, at least one-third of the head length; and suprapreopercular bone is present as a relatively large, flat bone.
The resulting topology of our phylogenetic analysis differs significantly from previous, even recent, hypotheses of chanid relationships.We confirmed the sister-group relationships between Rubiesichthys and Gordichthys as previously proposed by Poyato-Ariza et al. (2010) (without the inclusion of Nanaichthys) and Amaral, Brito (2012) (which included Nanaichthys), a hypothesis supported by the presence of both the dorsal and ventral borders of the maxillary articular process very curved (C.31:1); the presence of a ventral metapterygoid process of hyomandibular (C.52:2); and the presence of a posterior process on the posterior border of first supraneural (C.83:1).However, the position of Nanaichthys as the sister-group of the Laurasian clade Rubiesichthys + Gordichthys