Gross anatomy and histology of the alimentary system of Characidae (Teleostei: Ostariophysi: Characiformes) and potential phylogenetic information

A compared study of the morphology of the alimentary tract and liver of seven selected species corresponding to the main clades of the Characidae family is presented herein. Three new set of characters corresponding to 1) alimentary tract gross anatomy, 2) alimentary tract histology and 3) liver gross anatomy are evaluated as potential sources of data for future phylogenetic studies of the Characidae. Not considerable interspecific variation was observed at the histological level and therefore this source is not considered to be phylogenetically informative at the taxonomic level analyzed. In contrast, liver and alimentary tract gross anatomy presented important interspecific variation while a relatively homogenous intraspecific morphology was observed. Those characters are optimized in tree-topologies from previous phylogenetic analyses and their evolution and potential relationship with ecological traits are discussed.


Introduction
The Characidae, with more than 1200 species, is the most diverse Neotropical family of fishes and the fourth most diverse worldwide. Its species are found in most freshwater bodies from northern Patagonia, Argentina, to southern USA and in a vast variety of environments reflected in their great ecological and morphological diversity (Mirande, 2010;Reis et al., 2003). This makes this family an excellent model for studying the evolution of these features. Although some works have deal with particular aspects of the anatomy of the alimentary tract of some species, an integrative, extensive, and comparative approach including representatives of the main clades within the family is still lacking. For example, Leknes (2005) studied the histology of the intestine in Gymnocorymbus ternetzi (Boulenger, 1895) and Medina Perozo et al. (2009) studied the histology of the caeca of five species of Serrasalmidae and one of Characidae. Also, a histological characterization of the alimentary tract in Paracheirodon axelrodi (Schultz, 1956) was made by Gómez-Ramírez et al. (2011) and an immunohistochemical study (serotonin, cholecystokinin, somatostatin) of the digestive tract of Oligosarcus hepsetus (Cuvier, 1829) was published by Vieira-Lopes et al. (2013). In addition, the alimentary tract of a few species belonging to other families of Characiformes have been characterized, such as the anostomids Leporinus friderici (Bloch, 1794) and L. taeniofasciatus Britski, 1997, and Leporinus obtusidens (Valenciennes, 1837) (Ulibarrie, 1984). In the most comprehensive phylogenetic analyses of the Characidae, a relatively low number of species was analyzed (about a 15%) and there are many internal clades with very low support and morphological synapomorphies (Mirande, 2010;Mirande et al., 2011Mirande et al., , 2013. Also, some 274 discordance exists between morphological approaches and the molecular based published phylogenetic analyses of the family (Calcagnotto et al., 2005;Javonillo et al., 2010;Oliveira et al., 2011). Additional morphological information is needed to better assess the phylogeny of the family and especially to provide synapomorphies to many nodes.
According to morphological approaches Markiana Eigenmann, 1903(e.g. Mirande, 2010, it is related with Astyanax Baird & Girard, 1854, and included in an informal subfamily-level group termed "Astyanax clade" by Mirande (2010). According to molecular hypotheses (e.g. Oliveira et al., 2011 ), it is included in the Stevardiinae. Interestingly, Baicere-Silva et al. (2011) found, from data of sperm and spermatogenesis, similarities between Markiana and the Stevardiinae, supporting the molecular approach and supporting the idea that additional morphological information is needed.
The aims of this work are: a) to provide a description of the gross anatomy of the alimentary tracts and liver of seven selected species of Characidae representing the most diverse clades of Characidae (Mirande, 2009(Mirande, , 2010Javonillo et al., 2010;Oliveira et al., 2011); b) to characterize the alimentary tract at the histological level of six of these species; c) to propose a set of novel characters potentially informative for future phylogenetic researches; d) to contrast new evidence with the two alternative hypotheses of Markiana nigripinnis (Perugia, 1891) phylogenetic position; e) to optimize the new characters proposed in topologies from previous phylogenetic analyses (Mirande, 2010;Oliveira et al., 2011) and f) to discuss these results with regards to their evolution and relationships with the ecology and physiology of the analyzed species.
Adult specimens of each of these species were fixed in Bouin solution after overdose with benzocaine. After 1 day, they were transferred to ethanol 70%. Appropriate actions were taken to minimize pain or discomfort of fish, and this study was conducted in accordance with international standards on animal welfare, as well as being compliant with national regulations and the "Comité Nacional de Ética en la Ciencia y la Tecnología". At least 4 individuals of both sexes of each species were dissected under a stereo microscope, except for Markiana nigripinnis, since there was only one male individual available. This species and Characidium borellii were not used for histological studies and were only examined at a gross anatomical level. Dissections consisted on a ventral opening of the abdominal cavity, body walls were removed and the esophagus was sectioned in its anterior portion in order to free the alimentary tract, liver, pancreas and gas bladder. Pictures were taken using a digital camera attached to the stereomicroscope from left right dorsal and ventral views, after and before removing the liver. The alimentary tract of four specimens of Astyanax rutilus were fully sectioned and prepared for histological analysis in order to assess intraespecific variability. Also, the alimentary tract of one male of each of the remaining species was processed for histological characterization, as follows. The alimentary tract was separated with the pancreas attached to it and dehydrated together through increasing concentrations of ethanol solution and then transferred to xylene and included in paraplast at 60ºC during 6 h. Then, it was fully sectioned with a microtome at 7µm, and mounted in microscope slides. Sections were stained with Masson Trichromic. Microphotographs were obtained with a digital camera coupled to a stereo microscope in order to study the histological features of the alimentary tract and pancreas of these species.
A series of characters were defined corresponding to three different sources: 1) alimentary tract gross anatomy, 2) liver lobes gross anatomy and 3) histology of the alimentary tract.
In addition to a general description of the structures herein examined, observations are expressed as phylogenetic characters with alternative character-states which permit a better comparison between species and to have some clue about their evolution when mapped on a phylogenetic hypothesis of the family. That mapping was done on the phylogenetic hypothesis by Mirande (2010) using TNT (Goloboff et al., 2008) and discussed with regards to its evolution and relationship with known data from ecology and physiology of these species. We could not optimize those characters in the phylogenetic hypothesis by Oliveira et al. (2011) because this author only analyzed three of the species selected in the present work (A. anisitsi, G. ternetzi, M. nigripinnis) but we tested the phylogenetic position of M. nigripinnis as a member of the Stevardiinae, as proposed by Baicere-Silva et al. (2011) and Oliveira et al. (2011) by comparing differences in tree lengths of morphology-based phylogenetic trees when the inclusion of Markiana in the Stevardiinae is forced with a constraint.
Voucher specimens are deposited at the Ichthyological collection of the Museo Argentino de Ciencias Naturales "Bernardino Rivadavia" (MACN-ict), Buenos Aires.

Results
Gross anatomy of the alimentary tract. The alimentary tract of the analyzed species consists of an esophagus, continued by an anteroventrally curved stomach followed by a twisted intestine. The first intestine portion, referred to as the pyloric intestine, presents a dorso-lateral curve (1) to the right side of the stomach where it presents a variable number of caeca among the studied species (0-8). These caeca also vary in their position with respect to the stomach (at left, right, or ventral) although its distal portion was always directed to the posterior region (Table 1). This intestine portion is followed by a posterior prolongation surpassing the stomach where it presents a second curve (2) in anteroventral direction. From this point it develops in an anterodorsal direction at the left side of the stomach and may present a third curve (3) to an anterior direction. Afterwards, it presents a curve (4) in a posterodorsal direction. From curve 4, intestine projects posteriorly surpassing curve 2, in this portion it may present a slight curve (5) ventrally followed by a slight posterior curve (6) in some species (Figs. 1-2). Intestine foldings were observed under a stereo microscope by transparency of the alimentary tract wall (e.g. Fig. 2A). Mucosa Foldings are mainly transversal whereas those of the caeca are mainly longitudinally directed.
Intra and interspecific variability of this general pattern is discussed in the "Description of phylogenetic characters" section.
Gross Anatomy of Liver lobes. The liver was found to be ventrally to the esophagus and anterior to the stomach and pyloric intestine. It presented a series of lobes developing posteriorly at the sides of the stomach and anterior intestine portions. These lobes varied in their development and disposition among species. It was observed one dorsal lobe (dotted in bluein Figs. 3-4) and one ventral lobe (dotted in red) in the left side, one dorsal lobe at the right side (green) and a small ventral lobe (purple), present in some species .

Pancreas gross anatomy and histology.
The pancreas of the studied species was observed as longitudinal cords with ramifications following the intestine and each intestinal caeca. Histologically it presented a well-developed exocrine pancreas with serous acinous cells presenting a central spherical nucleus with one nucleolus, a basophilic basal portion, and an eosinophilic apical portion with zymogen granules. Associated with this, a pancreatic duct with a cylindrical epithelium was observed.
The endocrine portion of pancreas presented a, b and d cells. It was associated with blood and lymphatic vessels, nerves, and nervous plexus. No differences were noticed among the selected species (Fig. 6D).

Histology of the alimentary tract.
Histology of the alimentary tract of selected species presented almost no differences at the analyzed level (types of epitheliums, tunic's structure, presence of unicellular and multicellular glands, and type of musculature). The general pattern observed is described below and variations to it are mentioned.
Esophagus. The esophagus presents a mucosa with many foldings, composed by a pseudostratified epithelium, with abundant mucous cells ( Fig. 5A-B). The number and size of this foldings in esophagus and also in the intestine are apparently dependent on size of the individuals rather than to species identity, with larger individuals presenting a higher number and deeper foldings. One species, Cheirodon interruptus, presents gustative papillae in the esophagus region (Fig.  5C). Beneath the epithelium, a layer of loose connective tissue is present, surrounded by a muscular tunica with an inner longitudinal and an external circular layer of striated muscle. Some longitudinal muscular fibers are observed in the loose connective tissue. The striated muscle continued into the stomach in the cardiac and fundic portions.  Stomach. The cardiac stomach is relatively short and presents a simple epithelium as a layer of secreting mucus cells, which also cover the other regions of the stomach. Beneath it,a thin layer of dense connective tissue is observed (Fig. 5E). The fundic stomach follows the cardiac stomach and presents tubular-acinous glands (Fig. 5F) in its first portion. In the following portion, ramified tubular glandsare observed surrounded by loose connective tissue and dense septa of connective tissue (Fig. 5G). The final portion of the stomach, the pyloric stomach, is characterized by a great development of the smooth muscle and lack of acinous or tubular glands (Fig. 5H). This portion presents numerous foldings of the mucosa which are accompanied by dense connective tissue. Some loose connective tissue may be observed between the dense connective tissue layer and the smooth muscle. At the end of the pyloric stomach, a pyloric valve is present separating it from the first intestinal portion, the pyloric intestine (Fig. 6A).
Intestine. The pyloric intestine portion presents a variable number of caeca which are histologically similar to the rest of the intestine (Fig. 6B-C). This is characterized by a simple columnar epithelium with a striated border and mucus secreting goblet cells. A dense connective tissue is observed beneath it (Fig. 6B) surrounded by a thin layer of smooth muscle. The mucosa presents numerous foldings in this region. The peritoneumis composed of a thin connective tissue and a cubical mesothelium.  State 0 present in Astyanax endy, Bryconamericus thomasi, Cheirodon interruptus, and Markianna nigripinnis. Also in Leporinus friderici and L. taeniofasciatus (sensu Albrecht et al., 2001). State 1 present in Astyanax rutilus and Gymnocorymbus ternetzi. State 2 present in Aphyocharax anisitsi and Characidium borellii. State 2 is characterized by a reduction in intestine relative size and curves 2, 4, and 5 ( Fig. 1) are adjacent, fused in one curve, and curves 3 and 6 are absent.

Discussion
This paper represents the first comparative full description of the anatomy of the alimentary tract of fishes of the Characidae discussed in a phylogenetic context and provides a new set of morphological characters that can be used in future phylogenetic analysis. Previous related works had focused on studying some anatomical aspects on particular species of the family, (i.e.: Leknes, 2005Leknes, , 2011Medina Perozo et al., 2009;Gómez-Ramírez et al., 2011). Herein we analyze general anatomical patterns with a taxon sampling that takes into account previous phylogenetic hypotheses for the family (Mirande, 2009(Mirande, , 2010Oliveira et al., 2011). Also, species with no evident specializations regarding feeding, for example, which could mask variations strictly correlated with phylogeny were selected. Then, taken together this provides a general framework for the study of this system in the characids and also allows us to qualitatively estimate the level of variability expected among groups and to evaluate its potential phylogenetic information by comparing it with previous phylogenetic hypotheses.
The alimentary tract of the selected species presented a relatively conserved general pattern, especially at the histological level, nevertheless considerable variations on the development of liver lobes, intestine coiling and number and disposition of intestinal caeca were observed.
At the histological level, intraspecific variation was only observed in the relative length of mucosa foldings of the alimentary tract that was also variable among the species analyzed herein, which seems to be more related to specimens size rather than to species identity, therefore not used as a character here. The only interspecific difference observed at the histological level was: taste buds in the esophagus were only present in Cheirodon interruptus among the selected species. According to Cazorla et al. (2003) and Escalante (1987), this species has predatory habits, feeding mainly on Chironomidae and Trichoptera larvae. This feature together with the presence of striated muscle in the esophagus and first stomach portion may allow this species to eliminate preys that are being swallowed and are not palatable, what could happen in case of predation on a "wrong" prey, for example in case of prey confusion. Predation is supposed to be guided by several stimuli and a "specific searching image" the predator has (Tinbergen, 1960;also reviewed in Bond, 2007). Interestingly striated muscle in this portion is also present in the other species herein examined which is probably related to elimination of large size ingested items that cannot pass through the alimentary tract. Additionally, C. interruptus would present a second control related to taste buds for avoiding ingestion of unpalatable preys.
Pyloric caeca of the studied species presented foldings mainly in the longitudinal direction, in contrast with those in the rest of the intestine, that are mainly transverse. The disposition of the pyloric foldings probably facilitates its filling and emptying in a bidirectional way. Its main function is supposed to be absorptive, in contrast of those of mammals for example, which are supposed to have a fermentative function instead. These are filled and emptied together with the first portion of the intestine, increasing the whole absorptive surface of it. In trout and cod they account for more uptake capacity than all remaining regions of the gut combined (Buddington & Diamond, 1986). This absorptive function was corroborated here histologically by the presence of a striated border present in its epithelium related to the presence of microvilli in these cells. The number of intestinal caeca and intestine length is expected to be increased in herbivorous species, intermediate in omnivorous ones, and minimized in carnivorous fish but this is not always the rule. For example, Pogoreutz & Ahnelt (2014) studied four sympatric species of Gobiidae from Indonesia and found that one herbivorous species presented a short gut, there attributed to a phylogenetic inertia due to a carnivorous plesiomorphic state for that species. In the present work, A. anisitsi and C. borellii presented a reduced intestine, qualitatively observed (Figs.1-2), and low number of intestinal caeca. Concordantly, A. anisitsi in the Pantanal feeds mainly on aquatic insects during the rainy season and terrestrial insects in the dry season (Corrêa et al., 2009) while in the upper Paraná River floodplain in Brazil, this species fed mainly on microcrustaceans (Cladocera and Copepoda), in addition to Decapoda larvae, insect larvae and detritus (Russo & Hahn, 2006). Although there is no information on C. borellii diet, other Characidium species show a carnivorous diet: C. rachovii feed on Diptera (aquatic stage) and Amphipoda (Bastos et al., 2013), and C. lanei and C. pterostictum feed mainly on insects larvae and arthropods (Aranha et al., 2000). Cheirodon interruptus also had a carnivorous diet (Cazorla et al., 2003;Escalante, 1987) but a "medium" sized intestine and number of caeca (Figs. 1-2; Table 1). In the other hand, two omnivorous Anostomidae species of the genus Leporinus, with omnivorous diet, presented 10 and 12 intestinal caeca respectively and a relatively developed intestine . Paracheirodon axelrodi, a carnivorous miniature species presented no intestinal caeca (Gómez-Ramírez et al., 2011) what may be related both to its diet and miniature reductive characteristics. Markiana nigripinnis basically ingested food from vegetable origin, consisting on roots, fruit, seeds, and fibers (Resende et al., 1998) and presents a moderate development of intestinal length and caeca. Astyanax rutilus has an omnivorous diet also, although the main items ingested by this species were terrestrial insects and micro-crustaceans and the development of intestinal caeca and intestine was also moderate. This seems to be also the case of B. thomasi and A. endy, both species with no data on their feeding habits but examination of the guts during dissections and observation of histological sections of these species indicate that they may be omnivorous, eating insects and also algae (observed in histological preparations of B. thomasi). Also it is relevant to note that those species with omnivorous diets herein mentioned seems to have a tendency to a greater development of their liver and pancreas. This means that, in general terms, herbivorous species have a greater portion of its body devoted to feeding and in consequence it may be expected from this, that miniature species are more likely to evolve in "carnivorous clades", or less likely to occur in herbivorous ones, although herbivorous miniatures do occur as in the case of Otothyris (Loricariidae) (Rosa et al., 2014), for example.
The alimentary tract of the selected species presented a distinctively general pattern. Many of the characters of gross anatomy proposed herein appear as autapomorphies of some of the selected species. In fact, a search using implicit enumeration with the selected characters resulted in 4 most parsimonious trees with L=22 (results not shown). This means an increase on tree length in about 40% when optimized in Mirande (2010) topology (L=31), reflecting the great level of homoplasy of most of these characters and the relatively scarce phylogenetic information they present at this level. Nevertheless, these types of characters could have more relevant phylogenetic information for resolving relationships of closely related species, and therefore should not be discarded. In the other hand, histology of these species seems to be greatly conserved and therefore its use as a source of characters for this family is not recommended, especially considering its laborious acquisition. In contrast, histology may be a very interesting source of characters for resolving relationships between distant phylogenetic groups.
The tree length calculated for the optimization of the present characters in the topology proposed by Mirande (2009Mirande ( , 2010, in which Markiana nigripinnis is part of the "Astyanax clade", was L=31 (Fig. 7). The optimization when this species is presented as a basal member of the Stevardiinae presented a tree length decreased on one step (L=32, under equal weights). This is because a synapomorphy of the clade composed by M. nigripinnis and A. rutilus is lost: distal margin of right dorsal lobe of liver not surpassing vertical through anterior margin of gas bladder posterior chamber (Char. 10.1). Also, an ambiguous reconstruction of Char. 7 it is observed (left ventral lobe submedially located (state 1) present in M. nigripinnis and paralleled in A. anisitsi, A. rutilus and G. ternetzi). The information presented here therefore is supporting a closer relationship of M. nigripinnis with "Astyanax clade" than with Stevardiinae, although only supported by one synapomorphy. Considering also the very few species analyzed herein and the need of the evaluation of the general congruence of this information altogether with information of other sources (osteological, molecular, etc.) in a phylogenetic framework, we consider that the present results are not conclusive with respect to the phylogenetic position of Markiana.
The node containing Cheirodon interruptus and Aphyocharax anisitsi presented as a synapomorphy character 8 (0>1), left ventral lobe less developed and distal portion not reaching curve 2 of the intestine, paralleled in A. rutilus. This group is congruent with the node 195 of Mirande (2009Mirande ( , 2010 including the subfamilies Aphyocharacinae, Aphyoditeinae, and Cheirodontinae. The state 2 of character 2 (Curve 4 of alimentary tract reaching the posterior most region of stomach or not reaching the stomach), present in A. anisitsi and C. borelli may be a restriction due to the body shape of these species with an elongated cylindrical body and carnivorous feeding habits. The great differences in Gymnocorymbus ternetzi gastrointestinal tract morphology with respect to the other studied species is probably related to restrictions due to its particular body shape, laterally compressed and with a great relative body height. Then, these cases may be indicating that general body shape is probably restricting the alimentary tract development at a certain level at least. However, it can be also said that both the anatomy of alimentary tract as the general form of body are, at least at some point, influenced by phylogeny and those characters should not be discarded from phylogenetic analyses.
Characters 4, 5 and 6, present autapomorphies for 2, 3 and 3 species respectively, which may represent synapomorphies in future analyses including closest species of these taxa. Also character 9 presented and ambiguous reconstruction of the characters states in the internal nodes which may be clarified in future phylogenetic analyses including more taxa.
Finally, we highlight the needing of studies integrating data from ecology, compared anatomy, behavior, and physiology in phylogenetic contexts for a better understanding of these features and their interactions in an evolutionary frame. Therefore, we consider that this work could be the basis for further investigations on the relationships between the anatomy of the alimentary system, its physiology, the diet and ecology of these species and their feeding behaviors, in a phylogenetic framework.