Feeding ecology of stream-dwelling Characidae ( Osteichthyes : Characiformes ) from the upper Tocantins River , Brazil

In this contribution we studied the trophic ecology of four Characidae species from the Cavalo Stream, upper Tocantins River, considering diet overlap and trophic niche breadth. The diet of the four species was composed of adult and immature insects, both autochthonous and allochthonous in origin. Autochthonous items dominated the diet of Moenkhausia dichroura (Kner, 1858), Bryconamericus sp., and Creagrutus atrisignum Myers, 1917. By contrast, allochthonous items were dominant in the diet of Astyanax bimaculatus (Linnaeus, 1758). Trophic niche breadth varied among species, with the highest value recorded for M. dichroura (0.48), followed by Bryconamericus sp. (0.39), A. bimaculatus (0.33) and C. atrisignum (0.29). Similarity analysis revealed two groups with different patterns of food preference. The first group was composed of insectivorous and the second by omnivorous species. The overlap in food items consumed by the four species studied was high. We suggest that resources are not limited in this stream and that competition might not be regulating these populations. This is one more case corroborating the general pattern registered for Tropical environments, where resource partitioning and specialization are responsible by the organization of fish communities.

Fish communities are excellent models to test theoretical concepts about resource partitioning (ARANHA et al. 1998, 2000, MOTTA & UIEDA 2004), niche breadth (CASATTI 2002) and niche overlap (DOUGLAS & MATTHEWS 1992, DOUGLAS et al. 1994, NOVAKOWSKI et al. 2008).Although stream-dwelling fish species consume a variety of food items available in their surroundings, competition does not seem to take place among them.It has been suggested that trophic specialization (VOJTECH et al. 2002, DYER et al. 2010) or resource partitioning (HYNES 1970, SCHOENER 1974, ROSS 1986, DYER et al. 2010) are responsible for this lack of competition.Several studies carried out in tropical aquatic systems have demonstrated that one food resource can be consumed by many different fish species in a community.It may also happen that many species will explore different resources (e.g., POUILLY et al. 2003, 2004, 2006, NOVAKOWSKI et al. 2008).
The coexistence of species in highly diverse communities has been explained by two main theories (GRAVEL et al. 2006).The first one, called "the niche theory", considers that a niche is composed of different elements that supply the demands of a species and absorb the impacts of that species on the environment (CHASE & LEIBOLD 2003, NOBIS & WOHLGEMUTH 2004, MIKKELSON 2005).In this context, niche differences among species are fundamental to explain the maintenance of biodiversity in different scales (LEIBOLD & MCPEEK 2006).The second theory, called "the neutral theory", claims that species diversity in a given area is the result of stochastic ecological and evolutionary factors that interact on both local and regional scales.The neutral theory renders the niche theory unnecessary for understanding patterns of species coexistence (HUBBELL 2001).
Studies conducted in different streams indicate that both theories can be used to explain the patterns of organization of many Neotropical stream-dwelling fish communities.The soaring trophic plasticity of these communities allows them to coexist while exploring variable food supplies (BALASSA et al. 2004).Resource partitioning and exploratory tactics to minimize feeding overlap, which represent alternatives to competition, are made possible by the great diversity of food items available, the great diversity of morphological adaptations, and the exploration of different microhabitats and periods of activity of species (HURLBERT 1978, CHASE & LEIBOLD 2003).
Species of Characidae are largely widespread and morphologically diversified (BRITSKI et al. 1999).The family is composed of many species (NELSON 1994), and many cases of trophic resource partitioning have been documented for them (UIEDA et al. 1997).The ictiofauna of the Cavalo stream, located at the upper Tocantins River, has 36 fish species.Of these, 10 (~28%) are Characiformes and four belong to the Characidae.The

Feeding ecology of stream-dwelling Characidae (Osteichthyes:
Characiformes) from the upper Tocantins River, Brazil The four Characidae mentioned above had been previously classified into different feeding guilds, depending on the study site and the composition of the community where they had been found.The objective of this study was to evaluate the trophic ecology of these species at the Cavalo stream.We considered interspecific variations in diet composition, diet overlap, and trophic niche breadth.

MATERIAL AND METHODS
The Cavalo Stream (14°26'12"S, 48°34'53"W) is a 3 rd order tributary of the Traíras River in the upper Tocantins Basin (14°40'S, 49°00'W and 13°10'S, 48°00'W).Samples were collected on a 150 m long stretch of the Cavalo Stream, at a site formed by rocky substrata of riffle and rapid zones interspersed by sandy-mud pools.The marginal vegetation was sparse and the canopy covered approximately 90% of the study site (MIRANDA & MAZZONI 2003).
The hydrological regimen of the upper Tocantins River Basin is well defined (CARAMASCHI et al. 2012), with the high tide season extending from October to April and the low season from May to September.Sampling was carried out at each first fortnight of April, June (low-water-level season) and December (highwater-level season) of 1996 and January (high-water-level season) of 1997.Samples were collected by electrofishing (Honda EX 1000 generator, 220V, 2-3A, CA) according to the sample design presented by MAZZONI et al. (2000).During each sampling, we tried to collect enough specimens to obtain a robust analysis.Unfortunately, however, when the water was high, the speed of the rapids hampered our sampling efforts.
After capture, fish were immediately fixed (formalin 10%) and transported to the laboratory for identification and processing.Two-hundred-thirty-one individuals were analyzed, of which 55 were C. atrisignum, 57 were A. bimaculatus, 61 were Bryconamericus sp. and 58 were M. dichroura (Table I).Voucher specimens of each species are available in the ichthyological collection of Museu Nacional do Rio de Janeiro (MNRJ -36169 to 36172).For each sampled specimen we first registered standard length (Sl, cm) and total weight (Wt, gr).We then removed its stomach after performing an abdominal incision.Stomach content analysis was performed under a stereomicroscope through the volumetric (Vo) and frequency of occur-  (1992).Food items were classified to the lowest taxonomic categories following PÉREZ (1988).The relative importance of each food item was assessed using the percentage values of the Alimentary Index (AI i -KAWAKAMI & VAZZOLER 1980) as follows: AI i = [Fo i × Vo i /⌺ (Fo i × Vo i )] × 100, where i = 1, 2, ….n food items; Fo i = frequency of occurrence (%) of a given item; Vo i = volume (%) of a given item.Differences in dietary patterns occurring during the two hydrological seasons (four months) were tested for each species using the Spearman test .When no differences were registered among the four sampling months, the data set was grouped for further analysis.Patterns of diet similarity, based on the AIi matrix of the four studied species, were assessed by a Bray Curtis cluster analysis (UPGMA distances) (MCCUNE & MEFFORD 1997).
Diet breadth was assessed by the Levin's niche breadth index (B), where B = 1/⌺p 2 j, pj is the proportion of individuals using the resource j .(KREBS 1989).The Levin's index varies from 0 (diet specialized in few prays) to 1 (generalist diet).Diet overlap was calculated using the Horn index (R o ) as follows: Ro = (⌺(pij + pik) log(pij + pik) -⌺ pij logpik)/2 log2 (KREBS 1989), where pij is the proportion of resource i compared to all resources used by species j and pik is the proportion of resource i compared to all resources used by species k.Following ZARET & RAND (1971) and JOHNSON & ARUNACHALAM (2012), overlap values that were higher than 0.6 were considered significant.

RESULTS
The four studied species were abundant in the study area and occurred in both hydrological seasons (high and low-tide seasons) and provided enough stomach-content data (at least five individuals/species) to conduct the analyses.There were no significant differences in the general feeding pattern of all species between the hydrological seasons -Speaman rank correlation coefficients "r s " ZAR (1999) were always higher than 0.9 for p < 0.05 (Table I).As a consequence, we grouped all data for the subsequent analyses (Table II).
Similarity analysis revealed two groups with different patterns of food preference.The first group was formed exclusively by invertivorous species and included M. dichroura, Bryconamericus sp., and C. atrisignum.The second group was composed of a single omnivorous species (A.bimaculatus).Nonetheless, group 1 was further subdivided into two subgroups.The diet of the species in the first group (M.dichroura and Bryconamericus sp.) contained both autochthonous and allochthonous items.By contrast, the second subgroup contained species that eat exclusively autochthonous items (C.atrisignum) (Fig. 2).We detected significant (overlap > 0.6 according to ZARET & RAND 1971) feeding overlap in three (75%) of the four pairs of species analyzed (Table III).

DISCUSSION
Species of Characidae, such as those studied in the present paper, are largely widespread through the Neotropics and Central America (REIS et al. 2003, BUCKUP et al. 2007, LIMA et al. 2007) and have been classified into different feeding guilds, for instance insectivorous (e.g., CASATTI et al. 2003, SANTOS et al. 2009, BONATO et al. 2012), herbivorous (e.g., MESCHIATTI 1995, CASSEMIRO et al. 2002) and omnivorous (e.g., CASSEMIRO et al. 2002, BRANDÃO-GONÇALVES et al. 2009, RONDINELI et al. 2011), changing feeding habits according to the environment in which they occur.Our results showed that the diet of the four species studied is largely insectivorous, based both on autochthonous and allochthonous items, and that there are no seasonal differences related to the hydrological regimen.Other studies conducted in Neotropical streams also showed that small Characidae species frequently feed on invertebrates, and include both autochthonous and allochthonous items in their diets (TANK et al. 2010, ORTAZ et al. 2011, RONDINELI et al. 2011, XIMENES et al. 2011, BONATO et al. 2012, MANNA et al. 2012).
Streams that run through dense forests have low light levels due to the closed canopy.Consequently, the input of allochthonous organic matter (ARAÚJO-LIMA et al. 1995) and terrestrial arthropods (REZENDE & MAZZONI 2003, MOULTON 2006) in these streams often outweighs the autochthonous algal production (VANNOTE et al. 1980, HENRY et al. 1994, THORP 2002, TANK et al. 2010), affecting fish diet, as well as the entire trophic web (DODDS 2007).Thus, the input of allochthonous matter (animal and vegetal) is an important determinant of the dynamics of a stream, because it provides food for both fish and micro/macro-invertebrates, which in turn are food resources for many other stream-dwelling organisms.The importance of such allochthonous matter as an energy source flowing through the food chain was demonstrated for both Temperate (FISHER & LIKENS 1973, VANNOTE et al. 1980, MCINTIRE & FLECKER 2010) and Tropical (GONÇALVES et al. 2006, MOULTON 2006, SELVA et al. 2007) streams.In conclusion, the contribution of allochthonous material, as a direct or indirect source of food for fish, has been shown to be very important in streams, as the one studied in the present work (e.g., HENRY et al. 1994, UIEDA & KIKUCHI 1995, MAZZONI & REZENDE 2003, REZENDE & MAZZONI 2003).
We observed that among the four species studied, three feed mainly on autochthonous and one consumed mainly allochthonous items.These results indicate that both allo-and autochthonous processes are operating in the maintenance of the stream-dwelling fish community.Similar results had been previously documented for other tropical systems (e.g., ALVIM & PERET 2004) and could confirm the relative importance of different sources of energy to streams in Tropical areas.Nonetheless, based on dietary overlap and similarity analysis, we registered that 50% of the analysed species pairs used similar food resources.Among the four co-existing Characidae, three species used the same food resource (autochthonous matter).
Several reports on the feeding ecology of stream-dwelling fishes have shown that an overlap in resource use does not necessarily imply that there is competition among the species involved, since temporal and spatial segregation are strategies that minimize direct confrontation in the use of resources (CASATTI 2002, KOTLER & BROWN 2007).Three species in our study (M.dichroura, C. atrisignum and Bryconamericus sp.) shifted among insect prey.They all preyed almost exclusively on autochthonous items and their diets overlapped in about 50% of the items.Moenkausia dichroura preyed mainly on larvae of Ephemeroptera (48.75), C. atrisignum preyed mainly on larvae of Chironomidae (50%) and Bryconamericus sp.preyed mainly on larvae of Trichoptera (52.4%).Such strategy could reflect the specific morphological adaptations of each species and/or their use of the microhabitat (CASATTI & CASTRO 2006, MAZZONI et al. 2010) and could explain the co-existence of these close related species.Although resource availability was not measured in the present study, we suppose that the available prey items were diverse enough to allow for the observed feeding patterns of these Characidae fish species.
Studies in Amazonian streams showed that the diet of almost all fish species that shared a food source consisted of allochthonous plants or allochthonous insects (LOWE-MCCONNELL 1999).These findings reveal that the similarity in the use of feeding resources may be a consequence of the wide availability of them and/or a consequence of their shared use.Several examples of studies in Amazonian and other river systems in the Neotropical region have shown that this alternative is valid and recurrent (e.g., SABINO & ZUANON 1998, CASATTI 2002, NOVAKOWSKI et al. 2008, ROLLA et al. 2009).Based on previ-  ous data on community persistence and the temporal maintenance of the densities of the four studied species in the Cavalo Stream (MIRANDA & MAZZONI 2009), we suggest that these species share unlimited resources, and that competition is not regulating their populations.

Figure 2 .
Figure 2. Similarity dendrogram based on diet composition of four species of fish analyzed in Cavalo stream.