The role of Lithodoras dorsalis (Siluriformes: Doradidae) as seed disperser in Eastern Amazon

Ichthyochory is an important process responsible for the high diversity of plant species in tropical flooded forests. Thus, this study aimed to investigate the role of a catfish species, Lithodoras dorsalis, as seed disperser in the flooded forests at the Amazon River mouth, Brazil. Analyzing the stomach contents of 371 individuals of Lithodoras dorsalis, the Germination Potential (GP%) and Germination Speed Index (GSI) of seeds that were removed intact were investigated. This allowed us to evaluate the germination performance of two important species of plants in Amazonia, Euterpe oleracea (Açaí) and Montrichardia linifera (Aninga), after passage through the digestive tract of this catfish species. Given that digestion by L. dorsalis reduced the germination viability of M. linifera and that seeds were often destroyed during consumption, we suggest that L. dorsalis may have a limited role as seed disperser of M. linifera and instead mostly act as seed predator. However, for the species E. oleracea, L. dorsalis was a potential disperser, since the performance of germination of these seeds was improved after digestion. In addition, the number of seeds consumed was directly proportional to the catfish’s body size, reinforcing the role of doradids as potential seed dispersers in tropical forests.


Introduction
Seed dispersal is crucial for the maintenance of plant species, establishing patterns and limits that are critical for the population dynamics of angiosperms (Horn et al., 2011;Beckman, Rogers, 2013). The dispersal process has consequences for spatial distribution, gene flow, population genetics, intra and inter specific interactions and evolution (Seidler, Plotkin, 2006;Wall, Beck, 2011;Vargas, Stevenson, 2013). Several plant species have seeds adapted for dispersal by animals (zoochory) (Fenner, 1985;Margalef, 1991;Beckman, Rogers, 2013). Generally, these plants have attractive and nutritious fruits with seeds that can survive the passage through the digestive tract of dispersers (Fenner, 1985;Margalef, 1991).
The animal efficiency on plant dispersal depends on the amount of dispersed seeds (number of visits to the plant and number of dispersed seeds per visit) and dispersal quality (seed treatment in the animal's stomach and the site of deposition of the seed) (Herrera, Jordano, 1981;Schupp, 1993;Schupp et al., 2010). Seed dispersal is a mutualism and foraging by frugivores directly influences the spatial distribution of plants in the environment (Fenner, 2000;Silvertown, Antonovics, 2001). Thus, plant seed dispersal mechanisms are probably the result of natural selection that increases the chances of spreading their seeds (Fenner, 1985;Ferreira et al., 2010). Until a few years ago, fishes had little relevance in studies about frugivory and dispersion, being this process classically linked to birds and mammals (Correa et al., 2007;Correa et al., 2015b). However, fishes were probably the first vertebrate dispersers, playing an important role on the dispersal of the first angiosperms, approximately 70 million years ago in South America (Correa et al., 2015a). In tropical ecosystems, ichthyochory contributes to the maintenance of high diversity of plant species in flooded forests and wetlands (Gottsberger, 1978;Goulding, 1980;Horn, 1997;Pilati et al., 1999;Correa et al., 2007;Anderson et al., 2009). Many plant species in these habitats fruit during the flood season and they are adaptated to hydrochory or ichthyochory (Goulding, 1980;Correa, Winemiller, 2014;Correa et al., 2015b). Thus, in such environments, understanding the relationship between the fish fauna and riparian forests can serve as a basis to understand the ecological role played by fishes (Godinho, Godinho, 2003).
There are approximately 150 species of frugivore fishes in South America, which during the flood period enter the floodplain and feed on fruits and seeds dispersing at least 566 plant species (Goulding, 1980;Junk et al., 1997;Maia, Chalco, 2002;Correa et al., 2015a). They represent more than 60% of all frugivorous fishes in the world (Horn et al., 2011). These fishes have some adaptations that facilitate their activity as frugivores and seed dispersers, such molar like-teeth, long intestines and enzymes associated with carbohydrates digestion (Goulding, 1980;Drewe et al., 2004;Correa et al., 2007Correa et al., , 2015b. In addition to the adaptations cited above, some studies showed that ichthyochory effectiveness increases with the size of the fish, because of the high consumption of fruits, including the big ones, their extensive movement patterns and the long seed-retention time of large individuals (Galetti et al., 2008;Anderson et al., 2011;Correa et al., 2015a).
The rock-bacu Lithodoras dorsalis (Valenciennes, 1840) is one of the largest thorny catfish (Siluriformes, Doradidae), reaching at least 90 cm of standard length and 12 kg in weight (Goulding, 1980). This species is exploited by commercial and local fisheries in the Amazon mouth and occurs in northern South America in the Amazon estuary and neighboring coastal areas of French Guiana (Sabaj, Ferraris, 2003). Lithodoras dorsalis eat fruits and seeds and possible play a role as seed disperser (Goulding, 1980;Kubitzki, Ziburski, 1994;Barbosa et al., 2015). There are few studies focusing on seed dispersal by catfishes when compared to other frugivorous fish groups such as characids (Brycon spp.) and serrasalmids (Colossoma macropomum, Mylossoma duriventre and Piaractus spp.) (Horn, 1997;Galetti et al., 2008;Anderson et al., 2009;Anderson et al., 2011;Correa et al., 2015a). Few studies showed that seeds consumed by catfishes remain viable and/or had increased germination performance after passage through the digestive tract (Kubitzki, Ziburski, 1994;Stevaux et al., 1994;Pilati et al., 1999). The paucity of seed dispersal studies focusing on catfishes is due to the difficulty of capturing a large number of specimens to draw solid conclusions.
Due to the importance of doradids in seed dispersal and the high consumption of fruits by rock-bacu, the aim of this study was to answer the following questions: (i) What is the frequency and the proportion of fruits relative to other food items consumed by L. dorsalis? As shown for Barbosa et al. (2015), we expect L. dorsalis' diet to consist mainly of fruits; (ii) What is the relationship between the size of individuals and fruit consumption? We expect that with increasing length there is an increase in the amount of consumed fruits, positively affecting the potential seed dispersal activity; (iii) Are the seeds consumed by rock-bacu destroyed during digestion? If L. dorsalis is a disperser, we expect that the seeds are not destroyed during digestion; (iv) What is the effect of digestion on the germination rate of intact seeds consumed by L. dorsalis? Considering the possibility that the species is a disperser, we expect an increase in germination performance of intact seeds.

Material and Methods
The role of Lithodoras dorsalis as a potential seed disperser was investigated using two abundant plant species in the Amazon estuary floodplain, listed below: Euterpe oleracea (Mart.). Euterpe oleracea Mart. is a palm of the Arecaceae family, which can reach up to 25 m tall. Known popularly as "açaí", it is native to the Brazilian Amazon and occurs in the states of Pará, Amazonas, Maranhão and Amapá in areas with flooded soils and floodplains ("várzea" and "igapó") (Lorenzi et al., 1996). The species has glabrous fruits (without hairs, trichomes or similar structures on the outer surface), globular and smooth surface, reaching up in average 2.14 g, 15.1 mm of diameter and 13.2 mm of length (Paula, 1975;Almeida et al., 2011). The pericarp is purple and its pulp has a high calorific value (247 kcal in 100 g of dry fruit), measuring around 1 mm (Lorenzi et al., 2006). Due to these characteristics, the acai berry is part of the diet of Amazon human populations (Paula, 1975;Brondízio et al., 2003).

Montrichardia linifera (Arruda) Schott.
Montrichardia linifera (Arruda) Schott, popularly known as "aninga", is an amphibious herbaceous from Araceae family, reaching 4-6 m in height. This species can occur in several types of habitat, being found onshore or in water-saturated soils (Amarante et al., 2011). This macrophyte is distributed in the tropics (Mayo et al., 1997) and has considerable importance in the formation of riverbanks and white water streams, forming extensive clonal populations by sprouting from underground and submerged stems (Amarante et al., 2011). Seed length ranges from 33-51 mm (mean 39.53 mm) and width ranges from 23-33 mm (mean: 28.81 mm), weighing around 1.5 to 6.0 g (Amarante et al., 2011). M. linifera serves as food for fish, reptiles and aquatic mammals, due to the high caloric value of their fruits (355.12 kcal during the rainy season and 346.4 in the dry season) (Amarante et al., 2011). The aninga also has economic value for Amazon populations, since its floating stem can be used to build rafts for timber transport through rivers (Lins, Oliveira, 1994;Amarante et al., 2011).

Study area.
Fishing was conducted in the municipality of Abaetetuba, in the confluence of the Tocantins River and Pará River, in Pará State, Brazil (1°41'13,6"S; 48°52'48,8"W) (Fig. 1). The region's hydrological regime is characteristic of tropical tidal rivers with a daily, rapid and broad flood (reaching up to 4m) which also causes a reversal in the direction of the current (Welcomme, 1979;Barthem, Schwassmann, 1994;Hida et al., 1999).  The region's vegetation is defined as tidal floodplain vegetation with ombrophilous, broadleaved species, merged with palm trees like buriti tree (Mauritia flexuosa) and açaí berry tree (E. oleracea). The fruiting period of E. oleracea is from July to November (Guimarães et al., 2004), while M. flexuosa fruits from March to August (Sampaio, Carrazza, 2012) and M. linifera fructifies all year round.
The local climate is classified as Af following the Köppen-Geiger classification, corresponding to the typical conditions of tropical rainforest ecosystems (Peel et al., 2007). Annual precipitation is approximately 2500 mm, but during the period of our study the pluviosity was 3044 mm (mean = 253.7 ± 196.2) (Agência Nacional de Águas (ANA), 2016). The rainy season lasts from January to May (precipitation mean = 463.38 ± 87.42), and the dry season from June to December (precipitation mean = 103.87 ± 51.33) (Machado, 2008;ANA, 2016). Mean temperature is 27°C, ranging from 20°C to 35°C over the course of the year. Relative humidity is high, around 85%, varying normally between 81% and 90% (Machado, 2008).
Data sampling. Lithodoras dorsalis were sampled monthly over a full annual cycle, between July 2010 and June 2011. Specimens were collected in the main river channels and in the streams of Sirituba Island in Abaetetuba, Pará, Brazil (Fig. 1).
Weir fishing nets of aproximately 10 m in lenght, 3 m in height and a between-knots mesh size of 3-6 cm, were used to capture specimens. Nets were set at dusk (17:00-19:00) (Fig. 2a), depending on the tide, and removed at dawn (05:00-07:00). This period was adopted to ensure the capture of individuals moving from the smaller rivers to the main channel during the low tide (Fig. 2b). Specimens were removed from the water using seine nets (5 m x 1 m) and hand-or dip-nets (Figs 2c-d).

Fig. 2.
Sampling method applied to capture Lithodoras dorsalis at the Amazon River mouth, Brazil, from July 2010 to June 2011. a. Net set at the bottom during the dusk and with low tide to allow the fish entrance in small rivers; b. Net lifted during the next morning (approximately, 12 hours after setting at the bottom), blocking the exit of fishes that came during the night; c and d. Fish sample using seine nets and hand-or dip-nets. All captured specimens were analised in the laboratory at the Federal Institute of Pará (Instituto Federal do Pará -IFPA), where they were sacrificed with a lethal dose of anaesthetic. Then, a ventral-longitudinal incision was made from the urogenital opening to the head for the removal of the stomach, which was weighted (g) and emptied for the collection of its contents. The contents were weighed separately (g) and then sorted in a Petri dish. After the removal of the stomach, intact seeds were washed in running water and storaged in paper bags containing a substrate of vermiculite (MgFeAl) 3 (AlSi) 4 O 10 (OH) 2 4H 2 O (Oliveira et al., 1996).
Only E. oleracea and M. linifera reached sufficient number of seeds for analysis of germination. We analized 270 intact seeds of E. oleracea and 60 intact seeds of M. linifera. The lower number of analyzed seeds for M. linifera is because many of them were found destroyed. Only intact seeds of E. oleracea were found among digestive contents.
Following evisceration, the specimens were fixed in 10% formaldehyde, conserved in 70% alcohol, and then incorporated to the ichthyological collection of the Goeldi Museum (Museu Paraense Emílio Goeldi -MCT/MPEG), under the following catalog numbers : MPEG 19134;MPEG 19202;MPEG 19203;MPEG 19610;MPEG 19611;MPEG 21668-MPEG 21681. Germination experiments. Euterpe oleracea seeds were planted in containers containing nutrient-rich soil to a depth of 2 cm, with germinative aperture facing up (Silva et al., 2007;Gama et al., 2010). M. linifera seeds were planted in moistened vermiculite to a depth of 1 cm, also with germinative aperture facing up.
We tested the germination of E. oleracea under three treatments: Treatment A -seeds from the stomach of specimens of L. dorsalis; Treatment B -seeds from the forest with epicarp and mesocarp removed manually and; Treatment C (Control)seeds from the forest with epicarp and mesocarp intact (Samuels, Levey, 2005). For each treatment, 90 seeds were used. To M. linifera, only two germination treatments were tested: Treatment A -seeds from L. dorsalis' stomach; Treatment C (Control) -seeds from the forest, since these seeds have no external structures (like epicarp and mesocarp) to prevent germination. For this species, 30 seeds were used in each treatment.
In all treatments, seeds were considered germinated after the appearance of the radicle. The experiment was considered to be over when no futher germination occurred for seven consecutive days (Labouriau, 1983;Maia et al., 2007; Ministério da Agricultura, Pecuária e Abastecimento, 2009). In our study, the experiment lasted 42 days for E. oleracea and 29 for M. linifera.

Statistical analyses. The Frequency of Occurrence (FOi%)
was calculated for all the fruits and seeds found in L. dorsalis' stomachs. The aim was to define which plant species would be used in the germination test. The Frequency of Occurrence (FOi%) was calculated by the following formula (Hyslop, 1980): FOi% = (N i / N sto ) * 100, where: FOi% = frequency of occurrence of the item i; N i = number of stomachs where the item i was present; N sto = total number of stomachs analyzed.
To test the influence of the catfish's size in the intensity of consumption of fruits and seeds, a Simple Linear Regression was used relating the weight of fruits and seeds in the stomach with the body size of the specimens collected.
The Alimentary Index (AI i %) (modified from Kawakami, Vazzoler, 1980) of fruit and seeds consumed by individuals with different body sizes was calculated. The goal was to investigate if the importance of fruit and seeds changes as the catfish's body size increases. This index was calculated by AI i % = (FOi% * W% / ∑ FOi% * W%) * 100, where: AI i % represents the food item i alimentary index, FO i % the item i frequency of ocurrence and W% the relative weight of item i. Empty stomachs were not accounted for the analysis.
The Germination Potential (GP%) for E. oleracea and M. linifera was calculated to assess the effect of digestion in seeds consumed by L. dorsalis (viability or not). The calculation was based on the formula proposed by Labouriau, Valadares (1976): GP% = (GS / SA) * 100, where: GS is the number of germinated seeds and SA is the total number of seeds in the sample. The comparison of germination potential between treatments was performed using the chi-square (χ 2 ) test with 5% significance. Before the test, we checked chisquare assumptions and in cases that they did not fulfill, we did a description of the Germination Potential.
The Germination Speed Index (GSI) was performed to test the effect of the passage through the digestive tract of catfishes on the seeds' germination rate. This index was determined by registering the germination frequencies at intervals of 24 hours and it was calculated by the formula adapted from Maguire (1962): GSI = Σ (G i %/ N i ), where: GSI is the sum of the ratio between the percentage of germinated seeds at time i (G i %) within given time i (N i ).

Results
During the 12 months of the study period, 371 specimens of L. dorsalis were captured, of which 268 (74.93%) presented some plant species (fruits or seeds) in their stomachs. Fruits and seeds (intact and masticated) of seven different plant species were recorded and E. oleracea, M. flexuosa and M. linifera were the most common (Tab. 1). Linear regression analysis showed that L. dorsalis have a potencial role in seed dispersal. There was a positive relationship between L. dorsalis' body size and the total amount of fruits and seeds consumed by the specimens (R 2 = 0.35; p < 0.001) (Fig. 3).   Fig. 3. Relationship between body size (cm) and the general consumption of fruits and seeds (g) by Lithodoras dorsalis at the Amazon River mouth, Brazil, from July 2010 to June 2011.

Tab. 1. Frequency of Occurrence (FOi%) of fruits and seeds (intact and masticated) recorded in the diet of
Euterpe oleracea, pulp and seeds (mainly intact), seems to be more important in the diet of L. dorsalis' for individuals with intermediate size between 13.6 and 24.7 cm (Tab. 2). M. flexuosa, mainly pulp remains, was more important for smaller catfishes (<13.6 cm SL). M. linifera, pulp and seeds (mainly mastigated), was more consumed by large individuals (> 24.7 cm SL).

Tab. 2.
Alimentary Index (AI i %) of the main plant species consumed by Lithodoras dorsalis in each length class at the Amazon River mouth, Brazil, from July 2010 to June 2011. Only Euterpe oleracea fulfilled all assumptions of chisquare. For this species, the effect of digestion by L. dorsalis was similar between Treatment A (seeds from the stomach of specimens of L. dorsalis) and Treatment B (manual removal of epicarp and mesocarp) (χ 2 = 1.966; df = 1; p = 0.160). However, none of the seeds that had not passed through the digestive tract and which epicarp and mesocarp were still attached to the seed (treatment C) germinated (Tab. 3).

Euterpe oleracea
For Montrichardia linifera, the Germination Potential of seeds from forest plants was numerically diferent from that of seeds taken from the stomachs of L. dorsalis.

Tab. 3. Germination Potential (GP%) of Euterpe oleracea's
and Montrichardia linifera's intact seeds sampled at the Amazon River mouth, Brazil, from July 2010 to June 2011. Treatment A -seeds from the stomach of specimens of Lithodoras dorsalis; Treatment B -seeds from the forest with epicarp and mesocarp removed manually; Treatment C (Control) -seeds from the forest with epicarp and mesocarp intact. The seeds of E. oleracea from L. dorsalis' stomachs (Treatment A) germinated on the 15 th day (± 0.577), two days before seeds from the forest, with epicarp and mesocarp removed manually (Treatment B), which germinated on the 17 th day (± 0.707).
The Euperpe oleracea's Germination Speed Index was higher for seeds from the stomach of L. dorsalis, which amounted to 50% of germination on the 21 th day and 100% of germination on the 30 th day. However, for Treatment B, only 10% of the seeds germinated until day 21 (Fig. 4). The patterns of Germination Speed Index for M. linifera were similar to those of E. oleracea. The seeds removed from the fish's stomachs (Treatment A) germinated one day before seeds from Treatment C. Although only 33.3% of M. linifera's seeds in Treatment A germinated, all of these germinated by 19th day while 78% of seeds from Treatment C germinated in 22 days (Fig. 5).

Discussion
Lithodoras dorsalis, as well as other species of the Doradidae family, is essentially frugivorous (Ringuelet et al., 1967;Stevaux et al., 1994;Santos et al., 2004;Santos et al., 2006;Barbosa et al., 2015). In the Amazon mouth, E. oleracea, M. linifera and M. flexuosa were the most consumed fruit species and contributed the highest proportional weight to L. dorsalis diet. The presence of intact seeds of three plant species in fish digestive tract can be a positive indicator that this animal is a potential disperser (Gottsberger, 1978). In general, consumption by frugivores could increase germination success via scarification (Stevaux et al., 1994;Pilati et al., 1999). In Neotropical regions, some studies have shown that the passage of seeds through the digestive tract of fish increased germination success between 33% and 100% (Pollux, 2011).
Besides increasing germination performance, disperser's body zise, regardless of the species, can also influence the seed dispersal effectiveness (Schupp et al., 2010;Wotton, Kelly, 2012;Correa et al., 2015b;Li et al., 2016). Individuals of different sizes can generally differ in their diets due to morphological limitations during periods of development (Abelha et al., 2001;Dala Corte et al., 2016). In our study, there was a dominance of young specimens (8 to 33cm) because the Amazon River mouth is known as a nursery or growth area for L. dorsalis, a migatory fish that reproduce in upstream areas of Amazonas River (Goulding, 1980). Despite the presence of only young specimens, the results obtained from the study of L. dorsalis' feeding corroborated the cited above: the bigger the individuals, the bigger the possibility of obtaining fruits and seeds (Barbosa et al., 2015).
As the body size increases, specimens can expand their foraging ability, using a wider spectrum of resources (Abelha et al., 2001), including larger fruits and seeds like aninga M. linifera. In our study, smaller fishes fed mainly on pieces of fruits (mainly the pulp) such as buriti M. flexuosa. With an increased body size, catfishes started to consume other fruits such as açaí E. oleracea and aninga M. linifera, and also increased the abundance and diversity of items consumed per individual, e.g. crustaceans (see Barbosa et al., 2015 to access the complete diet of L. dorsalis). This higher fruits and seeds intake is evidence that L. dorsalis can be a disperser species, since one of the factors that influence the efficiency of a disperser is the number of seeds consumed per visit to the plant (Schupp, 1993;Schupp et al., 2010).
Ichthiochory allows new seedlings to develop distant from the source plant, enhancing the distribution of these species throughout the floodplain (Steavaux et al., 1994). A single seed can be consumed by many species of fish, increasing the probability of the plant species of reaching new areas and connect plant populations in fragmented landscapes (Goulding, 1980;Schupp, 1993;Nathan, Muller-Landau, 2000;Levey et al., 2005;Piedade et al., 2006;Seidler, Plotkin 2006;Schupp et al., 2010). The Amazon mouth is full of islands (Goulding et al., 2003;Machado, 2008) and seeds of plant species consumed by L. dorsalis can possibly reach several new sites that would not be attained without the aid of this potential disperser.
Even if many fishes consume a single plant species, dispersal mechanisms can cause interdependence as the result of natural selection, which selects traits that increase the survival chances of the seeds (Fenner, 1985). This may be the case of L. dorsalis and E. oleracea since, even though it feeds on several plant species, the rock-bacu led to an improvement in germination performance of seeds of açaí. Due to this possible interdependence relationship, seed dispersal is an important process for dispersers and plants, where both sides are favored: the disperser, for having a high-energy food source, e.g. açaí with 247 kcal in 100 g of dry fruit (Lorenzi et al., 2006), and the plant, for having the opportunity to perpetuate its species and colonize new environments (Beckman, Rogers, 2013;Correa et al., 2015b). The dispersion sets the initial spatial distribution of plants, determines the group of interacting species and contributes to the community structure and the maintenance of species diversity (Seidler, Plotkin, 2006;Anderson et al., 2009).
Lithodoras dorsalis speeded germination but reduced germination success for M. linifera. This means that, despite the increased germination performance, a very low number of progeny plants were generated (Schupp, 1993) and thus, this catfish cannot be considered a disperser of this specific plant species. The opposite occurred for açaí E. oleracea seeds, where the digestive process accelerated germination, resulting on a high germination percentage. The natural removal of the epicarp and the mesocarp of this plant by L. dorsalis may have enable the breaking of seed dormancy and its subsequent germination, generating several offsprings (around 85%) (Nascimento, Silva, 2005 In view of the potential role as a seed disperser for L. dorsalis, we concluded that the species has a diet composed mainly of fruits. The species is a potential disperser of açaí E. oleracea, widening the germination performance of the seeds that passed through the animal's digestive tract, when compared to seeds taken from parent trees. Also, with an increasing in body size, the seeds consumed by the rockbacu also increase, favoring its possible role as a disperser. Finally, there is the importance of riparian forests to fish populations and vice versa due to its interdependence relationship, where each one plays a key role: as food source and as seed disperser.