Feeding ecology of Auchenipterichthys longimanus ( Siluriformes : Auchenipteridae ) in a riparian flooded forest of Eastern Amazonia , Brazil

Feeding habits of the midnight catfish Auchenipterichthys longimanus collected in rivers of the Caxiuanã National Forest (Eastern Amazonia, Brazil) were investigated through the different hydrological periods (dry, filing, flood and drawdown). A total of 589 specimens were collected throughout seven samplings between July 2008 and July 2009, of which 74 were young males, 177 adult males, 89 young females and 249 adult females. The diet composition (Alimentary index – Ai%) was analyzed by a non-metric multidimensional scaling (nMDS) and by the analysis of similarity (ANOSIM), which included 37 items grouped into nine categories (Aquatic insects, Other aquatic invertebrates, Arthropods fragment, Fish, Plant fragment, Seeds, Terrestrial insects, Other terrestrial invertebrates, and Terrestrial vertebrates). We also calculated the niche breadth (Levins index) and the repletion index (RI%). Differences in the diet composition between hydrological seasons were registered, primarily on diet composition between dry and flood season, but changes related with sex and maturity were not observed. The midnight catfish showed more specialists feeder habit in the flood period (March 2009) and more generalist habits in the dry season (November 2008). The amount of food eaten by A. longimanus based on repletion index (RI%), did not differ significantly from sex and maturity. However, we evidenced differences in RI% when comparing the studied months. These results provide important biological information about the trophic ecology of auchenipterids fish. In view of the higher occurrence of allochthonous items, this research also underpins the importance of riparian forests as critical environments in the maintenance and conservation of wild populations of fish in the Amazon basin.


Introduction
The midnight catfish Auchenipterichthys longimanus (Günther, 1864), belonging to the family Auchenipteridae, is a Neotropical endemic species, with wide distribution in the Amazon and Orinoco basins (Ferraris et al., 2005).Most species of this family is omnivorous, nocturnal habit and are active swimmers in the water column (Rodrigues et al., 1990;Ferraris, 2003).During the night, these fish can be seen swimming just below the water surface in search of insects that fall from the riparian vegetation (Ferraris, 2003).
Thus, the feeding habit of fishes can be highly influenced by seasonal hydrological changes in the environment (e.g.Winemiller, 1989;Wootton, 1998, Lowe-McConnell, 1987) and further studies on the dietary ecology contributed to the understanding of the relationship between the aquatic fauna and the riparian forests (Winnemiller, 1989, Layman et al., 2007) For Vidotto-Magnoni & Carvalho (2009), the analysis of fish diets is also essential to better understand the ecology and behavior of the species, as well as other aspects, namely reproduction, growth and population dynamics (Andrian & Barbieri, 1996;Ballesteros et al., 2009).Dietary ecology and feeding habits can be explored by quantifying variation in resource use, trophic niche breadth and feeding intensity (Hammerschlag et al., 2010).These aspects, in the other hand, are influenced by food availability, food diversity, competition, and abiotic factors such as water level changes.
Starting from the hypothesis that fish feeding is influenced by hydrological seasons, the aim of this study was, therefore, to assess the feeding ecology of midnight catfish A. longimanus from rivers of Caxiuanã National Forest in the Eastern Amazonia, Brazil, through the different hydrological periods (dry, filing, flood and drawdown).In this area, A. longimanus is one of the most abundant species (Montag et al., 2008), and understanding the trophic ecology of this species will bring important information on the auchenipterids biology.

Study area
The sampling were carried out in the rivers of the Caxiuanã National Forest (1º45'27.5"S51º27'33.2"W),which is located in the Eastern Amazonia, between the Xingu and Tocantins Rivers in the municipalities of Melgaço and Portel, State of Pará, Brazil (Fig. 1).
The region is described as a lowland amazon rainforest located at sea level, being influenced both by the tide and by the oscillation of Anapu River (Hida et al., 1999).However, both flows have very low intensity, with an annual fluctuation in the water level only about one meter (Hida et al., 1999).In the region it could be observed four hydrological period by the fluviometry (Fig. 2): drawdown (June-August), dry (September-November), filling (December-February) and flood (March-May).
Because of a geological event of valleys drowning of the Anapu River, the region is also defined as "ria fluvial", showing a slight floodplain even during the flood period (Costa et al., 2002).

Fish sampling
Bimonthly collections were carried out between July 2008 and July 2009 in rivers of the Caxiuanã National Forest, assessing the annual hydrological cycle (drawdown, dry, filling and flood).We used gill nets, on average, 1.5 m high and 100 m long, with three to six centimeter meshes between opposite knots.The sites where the nets were located were selected with the help of a local fisherman.After capturing, the individuals were measured (total length in cm), weighed (total mass in g), sexed and their stomachs were removed through a cut in the mid-ventral region.The stomachs were weighed (0.001g precision) and kept in 10% formalin for 24 hours, and then preserved in 70% alcohol.
The midnight catfish specimens were deposited in the ichthyology collection of the Museu Paraense Emílio Goeldi -MCT/MPEG (Pará, Brazil)

Data analysis
The specimens were grouped by sex, maturity and hydrological period.On maturity, the fish were classified as young or adults according to the estimated size at first maturity for A. longimanus obtained by Freitas (2010), which classifies as adults any male larger than 12.5 cm and female larger than 13.0 cm.
The stomach contents were identified to the most precise taxonomic category possible, using a stereomicroscope (Nikon, model SMZ-10; 40x precision) and available literature.For each item we calculated the frequency of occurrence (FO%; Hyslop, 1980) and mass percentage (M%; Hynes, 1950), which was obtained by weighing each item and then calculating its importance among the total weight of all items.These two parameters (FO% and M%) were combined into the Alimentary Index (Ai%; modified from Kawakami & Vazzoler, 1980), which finally attributes the importance of the items in the diet.The Ai% was calculated by the equation Ai% = (FO%*M%/FO% *M%)*100 and the empty stomachs were not included in the analysis.
The food items were grouped into nine categories (aquatic insects, other aquatic invertebrates, arthropods fragment, fish, plant fragment, seeds, terrestrial insects, other terrestrial invertebrates, and terrestrial vertebrates) and had their respective values of Ai% calculated.
Non-metric multidimensional scaling (nMDS) ordinations were based in a cluster analysis using Unweight Paired Group Method by Mathematical Average -UPGMA (Romesbeirg, 1984) and the Bray-Curtis similarity matrices (Clarke & Warwick, 2001).The nMDS computes the distortion, or stress, between the similarity matrix and the ordination produced in a graphic representation of the axes.
The analysis of similarities (ANOSIM) was used to test the null hypothesis (H 0 ) that there were no significant differences in the diet composition between sex, maturity and hydrological period with a significance level of 0.05.In this analysis, we calculated the R-statistic, a comparative measure of the separation degree of the samples, which will permeate the decision of acceptance or rejection of H 0 .The range of variation of R is -1 to 1, which values near zero mean acceptance of H 0 .Both nMDS and ANOSIM was performed in the statistical software PRIMER 6 (Clarke & Gorley, 2006).
The Ai% values were exploratory applied in the niche breadth (range of the diet) using the standardized Levins index (Hurlbert, 1978).This was given by the equation B , where B a = standardized niche breath; P ij = proportion of prey j in the diet of predator i and n = total number of prey consumed by a particular predator.This index ranges from 0, when a species consumed only one type of prey (specialist consumer) to 1, when a species consumed many types of prey similarly (generalist consumer).
Finally, we performed a repletion index (RI%; Zavala-Camin, 1996) using the expression modified from Santos (1979) , where W i is the sum of the food content of individual i; W t is the total weight of individual i.The repletion index was analyzed individually for sex, maturity and collection periods (months).To investigate differences in the amount of food eaten by sex and maturity, the variation of the RI% was evaluated for the t test with a significance level of 0.05.The monthly results obtained for the RI% were assessed by the nonparametric Kruskal-Wallis (H) tested at a probability level of 0.05, considering as the null hypothesis (H 0 ) that the diet of A. longimanus is independent of period.

Results
Throughout the seven campaigns, 589 stomachs of A. longimanus were analyzed, of which 74 were from young males, 177 from adult males, 89 from young females and 249 from adult females.We identified 37 food items, most of allochthonous origin (Table 1).The proportion of empty stomachs represented 15-30% between the classes and was not included in the Ai% analysis.
Among terrestrial insects, Lepidoptera were the most consumed item in the diet of A. longimanus (Ai% = 9.1%), followed by Formicidae (8.6%) and Orthoptera (3.7%).The arthropods fragments represented the terrestrial arthropods remains that were not possible to identify.Among the seeds, seeds of Virola surinamensis represented 15.2% of the diet composition of the midnight catfish even considering that this item appears only twice (January and March; see Table 1).
The stress obtained with the ordination (nMDS) was 0.02, which ensures good consistency of the results.The diet composition of A. longimanus did not show a clear pattern of separation between sex and maturity (ANOSIM, all P> 0.05).Moreover, the ANOSIM showed a significant difference in the diet related to the hydrological period (R= 0.148, P= 0.057), primarily on diet composition between dry and flood season (R= 0.545, P= 0.015, Table 2).The multivariate analysis (nMDS) was based in 50% of similarity, which basically separated the dry and flood elements of the feeding analysis (Fig. 3).
In the dry season the diet of the A. longimanus was mainly composed of plant fragment (September, Ai% = 73.2% and November, 8.2%) and terrestrial insects (Sep= 13.6% and Nov= 8.2%) while in the flood season the feeding analysis revealed seeds (March, Ai%= 78.7% and May, Ai%= 2.7%) and terrestrial insects (Mar= 10.7% and May= 66.2%) as principal items (Fig. 3).
The ANOSIM also revealed differences in the diet composition between the drawdown and dry season (R= 0.313, P= 0.034, Table 2), with terrestrial insects and arthropods fragment as primary items in the feeding of the A. longimanus in drawdown season.

Discussion
According to the diet analysis of the midnight catfish A. longimanus, this species may be considered insectivorous, due to the high consumption of terrestrial insects and arthropods fragments in most studied months (July andNovember 2008 andJanuary, May andJuly 2009).However, if we analyze each collection period, this species could be classified as an omnivorous because its opportunistic frugivorous habits in September 2008 (dry season) and March 2009 (flood), consuming plant fragment as flowers and seeds, respectively.This kind of item that falls on the water from the riparian forest seems to be of easy capturing by A. longimanus because its water surface swimmer habit (Ferraris, 2003) , which would explain the high consumption these plant parts.
Regarding the trophic ecology of auchenipterids, the feeding habits of A. longimanus are similar to those found for several species of the Family Auchenipteridae, such as Trachycorystes in the lowland region of the central Amazon basin (Manaus, State of Amazonas, Brazil) (Goulding, 1980); Tocantinsia in upper Tocantins River, Brazil (Carvalho & Kawakami, 1984); Parauchenipterus (=Trachelyopterus) in a reservoir of the Paraná basin, Brazil (Andrian & Barbieri, 1996) and in the northeastern Brazilian basins (Santos, 2005); and Auchenipterichthys, Auchenipterus, and Tatia in the basin of Anapu River (Montag, 2006).Andrian & Barbieri (1996), for example, analyzed the feeding of Parauchenipterus galeatus Several studies also highlight the importance of these food items in the diet of Neotropical fishes (Waldhoff et al., 1996;Esteves & Aranha, 1999;Abelha et al., 2001;Claro-Jr. et al., 2004;Vidotto-Magnoni & Carvalho, 2009), which substantiates the importance of riparian forests for aquatic biological communities that seek other food sources in these ecosystems (Goulding, 1980).
Despite the constant importance of animal items in the diet of the midnight catfish between the months (terrestrial insects and arthropods fragments), seasonal changes in its diet showed in this study can be mostly related to the availability of plant items.In the flood period (March/2009) the seeds of Virola surinamensis were very abundant and represent the mainly item in the diet of A. longimanus.In that month the availability of this kind of seed also revealed two situations: the lowest values of niche breadth and the highest level of food consumption.
Within the flood period the water invaded the forest and theoretically supplies more food resource for fishes (Goulding, 1980;Gomiero et al., 2008).Nevertheless, in the flood period in March/2009, the A. longimanus presented the most specialists feeding behavior.Already in May/2009 (still flood season) the low food intake may be related to the reproductive activity (Jan-May 2009) evidenced by Freitas (2010).The large consumption of seeds in March 2009 is also suggested by Freitas (2010) as an energy source for reproduction, and besides reinforcing the importance of riparian forests on the fish diet (Goulding, 1980) may also suggest an active role of the midnight catfish in seed dispersal (Mannheimer et al., 2003;Freitas, 2010).
Despite the relationship of feeding and fish reproduction is recognized in the literature (Bellesteros et al., 2009), it needs to be further investigated for A. longimanus in the Amazon rivers to improve our understanding of the diet role in the breeding of tropical fish.
In the dry season (November/2008), the feeding of A. longimanus was also strongly composed of plant fragments, flowers.However, this month presented the greatest niche breadth and the lowest level of food consumption.With the low water level, the rivers are not flooding the forest anymore and the fish that feeds on allochthonous items from the riparian forest has its food supply reduced (Goulding, 1980;Esteves & Aranha, 1999;Gomieiro et al., 2008).Summarizing, the diet composition and food intake of A. longimanus from the Caxiuanã Nacional Forest rivers appears to be influenced by extrinsic factors such as hydrological seasons even knowing that the water level variation is described as slight (Costa et al., 2002).In fact, this variation seems to be enough to create mechanisms to adapt its biological processes such as reproduction.
By the way, seasonality changes in the feeding of fishes have been frequently documented in many studies (Claro-Jr. et al., 2004, Santos, 2005;Luz-Agostinho et al., 2009;Barnes et al., 2010;Tófoli et al., 2010).Within the auchenipterids, Santos (2005) analyzed the feeding habit of Parauchenipterus galeatus (= Trachelyopterus galeatus) in the northeastern Brazil and found that the high consumption of allochthonous items, mainly terrestrial insects and fruits, varied in relation to the hydrological period.
Based on this, we can incorporate the trophic opportunism concept in the feeding analysis of A. longimanus.As considered by Gerking (1994), an opportunistic fish is that makes use of an abundant and unusual food source, as in the case studied here, flowers and seeds of forests trees.Some studies also reported significant seasonal differences in fish diet, suggesting    that the items are consumed due to greater availability, characterizing them as opportunistic feeders (Abelha & Goulart, 2004;Ceneviva-Bastos & Casatti, 2007;Rezende & Mazzoni, 2003, Esteves & Aranha, 1999) These variations are relevant because the food supply is not static, particularly in tropical biotopes, suggesting the trophic plasticity of fish, as has been classically demonstrated in the literature (e.g.Andrian et al., 1994;Abelha et al., 2001).Ringuelet (1942) has already noticed that to understand the feeding habits of a species, we must analyze all possible variations brought by the environment.The most current studies, which include trophic studies in fish, have generally taken into account some of these variations, such as hydrological period (Claro-Jr. et al., 2004;Santos, 2005;Luz-Agostinho et al., 2009), food availability (Vidotto-Magnoni & Caralho, 2009), intra and interspecific relationships (Piorski et al., 2005;Fugi et al., 2008), amongst others.
Although some studies have been addressed the trophic ecology of auchenipterids fish (Ortêncio-Filho et al., 2001;Claro-Jr. et al., 2004;Pouilly et al., 2004;Krumme et al., 2004;Santos, 2005;Luz-Agostinho et al., 2009), these studies did not totally focus at auchenipterids or cover only few species, showing there may still be much to learn about the feeding habits of these catfishes.Thus, our study provides important information to increase the biological knowledge about auchenipterids trophic ecology.These parameters are highly relevant to compose the management plan for the Caxiuanã National Forest (still not documented), besides characterizing the riparian forest environments as fundamental in the maintenance and conservation of natural populations of fish in the region.

Fig. 1 .
Fig. 1.Location of the Caxiuanã National Forest, showing the rivers where the fish were collected.Curuá River -Ferreira Penna Research Station (ECFPn), Caxiuanã River, Puraquequara River and Caquajó River.Some black spots represent more than one collection site.

Fig. 2 .
Fig. 2. Fluviometry from the rivers of the Caxiuanã National Forest in the period between July 2008 and July 2009.Data obtained from the fluviometric station of Caiçara of the National Water Agency (ANA).

Fig. 3 .
Fig. 3. (a) Graphic representation of the non-metric multidimensional scaling analysis (NMDS) of the diet of Auchenipterichthys longimanus from rivers of the Caxiuanã National Forest (PA, Brazil).Stress = 0.02.The geometric forms represent the hydrological seasons: drawdown; dry; filing; and flood.(b-d) contribution (expressed according to the circle size) of the seeds (b), terrestrial insects (c) and plant fragment (d) in the feeding analysis.

Fig. 4 .
Fig. 4. Trophic niche breadth (Levins index; B a ) of the midnight catfish Auchenipterichthys longimanus from rivers of the Caxiuanã National Forest of (PA, Brazil) related to hydrological periods.

Table 1 .
Alimentary Index (Ai%) of food items in the diet of the midnight catfish Auchenipterichthys longimanus from river of the Caxiuanã National Forest (PA, Brazil) from July 2008 to July 2009.

Table 2 .
Pairwise comparison from ANOSIM between hydrological periods for the diet composition of the midnight catfish Auchenipterichthys longimanus from rivers of the Caxiuanã National Forest (PA, Brazil) from July 2008 to July 2009.*significant values for the test.

Table 3 .
Significance values for the Mann-Whitney pairwise comparison of Repletion Index (RI%) of the midnight catfish Auchenipterichthys longimanus from rivers of the Caxiuanã National Forest(PA, Brazil).*significant values for the test.