Introduction

Species specializing in swift water, often referred to as rheophilic (Lincoln et al., 1985), have special morphological, physiological, and behavioral adaptations to extreme environmental conditions, and rapids-adapted fishes worldwide possess convergent suits of phenotypes (Stewart & Roberts, 1976; Lujan & Conway, 2015). For example, fishes from widely divergent evolutionary lineages have reduced swimbladders and elongate bodies that reduce buoyancy and drag, respectively, when resting on substrates in fast-flowing water (Conway et al., 2012; Birindelli & Britski, 2013). Many benthic fishes possess broad pectoral and pelvic fins that function like hydrofoils, as well as mouths capable of suctioning against solid substrates, to assist in holding position in fast water (Conway et al., 2012; Leitão et al., 2015). Given the strong selective pressure for convergent morphologies and behavior in these habitats, rapids in tropical rivers provide a unique opportunity to understand mechanisms promoting the stable coexistence of diverse fish assemblages.

Evidence that fishes use resource-partitioning to limit interspecific competition is widespread (Gatz, 1981; Ross, 1986; Lujan et al., 2012; Montaña et al., 2014), and several features of rapids may increase the need for coexisting species to segregate trophic niches. Rapids create physical barriers to longitudinal dispersal for many aquatic organisms (Junk & Soares, 2001). Rheophilic species inhabiting isolated rapids also can have limited dispersal among local populations when these are separated by long stretches of slower moving water (Carvalho et al., 2007). This dispersal limitation may increase competition for food resources, as well as the importance of niche partitioning in facilitating coexistence of rheophilic fishes. Competition may be further intensified during low-water periods if high-quality food resources become less available (Lowe-McConnell, 1987). Because many rapids-adapted fishes have limited dispersal and exploit food resources associated with swift-water habitats, niche partitioning of food resources or morphological features that impact foraging behavior may be necessary to promote stable coexistence (Vitorino Júnior et al., 2016).

In the Neotropics, many rapids-adapted fishes exploit autochthonous food resources, including crabs, snails, aquatic insect larvae, sponges, and aquatic plants (Horeau et al., 1998; Moreira & Zuanon, 2002; Pagezy & Jégu, 2002; Zuanon & Sazima, 2002). Anostomid and loricariid fishes inhabiting rapids of Neotropical rivers encompass diverse phenotypes (Zuluaga-Gómez et al., 2016), but many species within these families seem to have relatively narrow trophic niches reflecting herbivory or detritivory. The family Serrasalmidae is well known for carnivorous piranhas (Chakrabarty & Fink, 2011), but several species of the family are herbivorous, some of which inhabit rapids nearly exclusively. The family, as a whole, encompasses diverse feeding strategies, including carnivory, insectivory, omnivory, herbivory, frugivory, and lepidophagy (Goulding, 1980; Sazima, 1983; Leite & Jégu, 1990; Sazima & Machado, 1990; Loubens & Panfili, 1997; Santos et al., 1997; Correa et al., 2007; Loubens & Panfili, 2001; Trindade & Juca-Chagas, 2008; Correa & Winemiller, 2014; Correa et al., 2014, 2016). Some serrasalmid species are habitat generalists, but others are restricted to specific habitat types. For example, the tambaqui [Colossoma macropomum (Cuvier 1816)] and pacu (Piaractus spp.) inhabit lowland river and floodplain habitats where they feed on fruits and seeds (Goulding, 1980; Jégu & Keith, 1999) while species of the genera Myleus, Mylesinus, Tometes, Ossubtus, and some species of Myloplus are restricted to rapids habitats (Jégu et al., 1989; Jégu & Santos, 2002; Pagezy & Jégu, 2002; Jégu et al., 2002, 2003; Jégu & Zuanon, 2005; Andrade et al., 2016a, b).

Both juveniles and adults of the rheophilic species Myleus setiger Müller & Troschel, 1844, Tometes kranponhah Andrade, Jégu & Giarrizzo, 2016, and Ossubtus xinguense Jégu, 1992 are commonly found in fast-flowing waters of rapids, often in close association with rocks covered by aquatic macrophytes of the family Podostemaceae (Andrade et al., 2013). Myleus setiger is the most widely distributed of the three species, occurring in several major Amazon tributaries that drain the Guiana and Brazilian shields, whereas T. kranponhah and O. xinguense are endemic to the Xingu Basin (Andrade et al., 2016a). Myleus setiger and T. kranponhah are not listed as species of conservation concern due their wide distribution and local abundance within the Xingu Basin (Andrade et al., 2016a). Ossubtus xinguense is considered highly threatened (Jégu & Zuanon, 2005; Andrade et al., 2016c) and is listed as vulnerable in Brazil (National Red List, 2016). None of these species have been assessed by the International Union for the Conservation of Nature (IUCN, 2018). Despite differences in their listing within Brazil, both T. kranponhah and O. xinguense may suffer harmful and irreversible effects due to recent hydrologic alterations in the Xingu River caused by the construction of Belo Monte hydropower dam, the third largest hydroelectric dam complex in the world (Andrade et al., 2016a, c). Changes in seasonal flow dynamics combined with anticipated impacts of gold mining in the region (Tófoli et al., 2017) will likely impact niche relationships of the Xingu’s endemic diversity, making it necessary to understand the mechanisms facilitating coexistence of rapids-adapted species for conservation efforts in the region.

Here we test whether three sympatric serrasalmids (M. setiger, T. kranponhah, and O. xinguense) from the lower Xingu River partition trophic niche space during the annual low-water period. All three species are known to be herbivorous, with diets dominated by aquatic macrophytes (Jégu & Santos, 2002; Jégu & Zuanon, 2005; Andrade et al., 2015, 2016a). Coexistence of these three rapids-adapted serrasalmids might be facilitated by differences in how they exploit microhabitats, food resources, or both (e.g., Dias & Fialho, 2011; Mouchet et al., 2013; Burress, 2014; Gracan et al., 2016). We evaluated both inter- and intraspecific niche variation in diet, stable isotope ratios, and functional morphology related to feeding. We predicted that intraspecific variation would be most strongly associated with ontogenetic diet shifts, with some species exploiting mostly macroinvertebrates during the juvenile stage and shifting to a diet comprised mostly of aquatic macrophytes during the adult stage. We further predict that interspecific trophic niche overlap would be lower for adults feeding on different plants or plant parts, and that morphological traits reflecting how these fishes use microhabitats and acquire food would differ according to species and life stage.

Materials and methods

Study site

The study was conducted in the lower portion of Xingu River Basin, which is characterized by numerous rapids within the network of channels comprising the Xingu, Iriri, and Bacajá rivers (Fig. 1). Collections were made along a 270-km stretch of river from Cachoeira Grande (Big Falls) on the Iriri River (3°51′10″S 52°43′40″W) to the downstream end of the Volta Grande (Big Bend reach of the lower Xingu) near Belo Monte (3°03′57″S 51°49′35″W) where, in 2016, the third largest hydroelectric power plant complex in the world was completed (Belo Monte Dam). In addition, the Bacajá River was sampled from its confluence with the Xingu River (3°45′26″S 51°34′57″W) to a location approximately 40-km upstream (Fig. 1).

Fig. 1
figure 1

Xingu River Basin showing the study area prior to flow modification made in 2016 due to the construction of Belo Monte Dam and the sampling sites

Full size image

Species characteristics

Myleus setiger and O. xinguense (hereafter ‘Myleus’ and ‘Ossubtus,’ respectively) reach standard lengths of 27 cm and 23 cm, respectively. Tometes kranponhah (hereafter ‘Tometes’) is generally larger, reaching standard lengths of 37 cm. All three species have functional traits specialized for herbivory and feeding in rapids, such as incisiform teeth that effectively cut leaves, and laterally compressed bodies that reduce hydraulic drag in fast water (Meunier et al., 2004). All three species undergo ontogenetic transitions in the length of the gastrointestinal tract, with adults having relatively longer guts (Jégu et al., 1989, 2002). This size allometry is assumed to be associated with a shift from generalist feeding to a diet dominated by epilithic macrophytes, including riverweeds (Podostemaceae) and bryophytes (mosses and liverworts).

Sample collection

Fishes were collected during low-water periods when the three serrasalmid species are restricted to patchily distributed rapids. Sampling was performed during June–July 2012, September–October 2013, October 2014, and October 2015, prior to flow modification by the Belo Monte Hydroelectric Complex. Except Myleus, which was not sampled in 2012, Ossubtus and Tometes were sampled every year. Fishes were captured from rapids using gillnets placed in relatively deep, slow-flowing areas, and castnets thrown in shallower, faster-flowing sections of rapids in or near epilithic macrophyte beds. To test whether species partition niches between different ontogenetic stages, we first classified individuals as juveniles (< 100 mm SL for Myleus and Ossubtus; < 130 mm SL for Tometes) or adults (≥ 100 mm SL for Myleus and Ossubtus; ≥ 130 mm SL for Tometes; Table 1), based on size at first maturation (Jégu & Santos, 2002; Andrade et al., 2016a, c). For stable isotopes analysis, a sample of muscle tissue was removed following euthanasia and prior to preservation of the specimen in 10% formalin. Morphological and dietary data were obtained from specimens after storage in 70% ethanol.

Table 1 Number of individuals from multiple sites of the Xingu River basin analyzed for each method in this study, and classified according to the ontogeny pre-established into juveniles (J) and adults (A)
Full size table

Morphological traits

Sixteen morphological features were measured according to methods in Gatz (1979) and Winemiller (1991): standard length (SL), body depth, body width, head depth, eye position, head length, caudal-peduncle length, caudal-peduncle depth, caudal-peduncle width, pectoral-fin length, pectoral-fin width, eye diameter, mouth width, mouth orientation, gastrointestinal length, and swimbladder length (Table 2). We also measured the length of the nasal chamber of the skull, i.e., the space that houses the olfactory bulb, in the longitudinal dimension (Andrade et al., 2016c). Linear measures were made on preserved specimens using digital callipers (precision 0.1 mm). Morphological traits were selected based on their functional roles in feeding or swimming performance and use of microhabitat (Table 1). For example, body size (indexed by SL) influences feeding ecology (Verwaijen et al., 2002; Montaña & Winemiller, 2013), relative body depth influences lateral turning (Zuluaga-Gómez et al., 2016), and relative intestine length is associated with diet (Wagner et al., 2009).

Table 2 Morphological traits with their respective formula and ecological explanation
Full size table

Diet

Each specimen examined for dietary analysis was measured (SL mm) before dissection and removal of the stomach. Empty stomachs were recorded but did not contribute to sample sizes reported for dietary analyses. Food items from stomachs were classified according to eight categories: 1—leaves and flowers of terrestrial plants; 2—seeds (mainly from riparian plants); 3—Podostemaceae (mainly leaves of this aquatic macrophyte); 4—periphyton (benthic algae and associated microfauna and biofilm); 5—detritus (particulate organic material); 6—sediments (mainly silt and sand); 7—fish fins and scales; and 8—aquatic macroinvertebrates (mainly Chironomidae and Simuliidae). The wet weight of each prey item from each stomach was determined with a digital electronic balance and recorded to the nearest 0.0001 g.

Stable isotopes

For isotopic analysis, 2 g of muscle tissue was taken from the dorsal flank region of fish specimens using a clean scalpel. Most tissue samples were stored frozen (N = 108), but a few were preserved in salt (N = 11), which has been shown to have negligible influence on carbon and nitrogen isotope ratios of fish muscle tissue (Arrington & Winemiller, 2002). Tissue samples were dried to constant weight in an air-circulating oven at 60°C, pulverized to a fine homogeneous powder, weighed, and packed into tin capsules according to methods described in Zeug & Winemiller (2008). The samples were analyzed for carbon and nitrogen stable isotope ratios at the Center for Stable Isotopes, University of New Mexico, using a Costech ECS 4010 Elemental Analyzer coupled to a ThermoFisher Scientific Delta V Advantage mass spectrometer via a CONFLO IV interface. Isotope ratios are reported using the delta (δ) notation relative to carbon from Pee Dee Belemnite as the standard for carbon and atmospheric molecular nitrogen as a standard for nitrogen. Routine analysis of laboratory standards indicated measurement error was less than 0.1‰ for both δ13C and δ15N.

Data analysis

Morphological functional space

To describe ecomorphological trends and to ordinate species and life stages, Principal Components Analysis (PCA) was performed using the correlation matrix calculated from log-transformed morphological data for species and ontogenetic stages (juveniles, adults). PCA was performed using the ‘factoextra’ package (Kassambara & Mundt, 2016) in R version 3.2.3 (R Development Core Team, 2017). Significance of morphological functional groups (i.e., species and life stages) was tested using Permutational Multivariate Analysis of Variance (PERMANOVA) made with 9999 permutations based on a Bray–Curtis distance matrix using the package ‘vegan’ 2.4-5 (Oksanen et al., 2017) in R.

Trophic niche

The dietary importance of food categories was estimated using the alimentary index: \({\text{Ai}} = {{{\text{Fi}} \times {\text{Wi}}} \mathord{\left/ {\vphantom {{{\text{Fi}} \times {\text{Wi}}} {\sum\nolimits_{i = 1}^{n} {\left( {Fi \times Wi} \right)} }}} \right. \kern-0pt} {\sum\nolimits_{i = 1}^{n} {\left( {Fi \times Wi} \right)} }}\), where Fi is the relative frequency of occurrence of prey category i and Wi is the relative weight of prey category i (Kawakami & Vazzoler, 1980). Dietary similarity was calculated as Bray–Curtis similarity based on Ai values; similarity of food category consumption was based on Whittaker’s index of association (Clarke & Gorley, 2015). Both similarity indexes consider relative abundances, but Whittaker’s index is more sensitive to differences in distributions of abundances than Bray–Curtis (Legendre & Legendre, 2012). The level of specialization between life stages was inferred using Levin’s measure of niche breadth (Krebs, 1999): \(B = 1/ \sum p_{j}^{2}\), where pj is the proportion of individuals found using resource j. Niche partitioning among species and between life stages were assessed using Pianka’s niche overlap index: \(O_{\text{jk}} = {{\mathop \sum \nolimits_{1}^{n} p_{\text{ij}} p_{\text{ik}} } \mathord{\left/ {\vphantom {{\mathop \sum \nolimits_{1}^{n} p_{\text{ij}} p_{\text{ik}} } {\sqrt {\mathop \sum \nolimits_{1}^{n} p_{{{\text{ij}}^{2} }} \mathop \sum \limits_{1}^{n} p_{{{\text{ik}}^{2} }} } }}} \right. \kern-0pt} {\sqrt {\mathop \sum \nolimits_{1}^{n} p_{{{\text{ij}}^{2} }} \mathop \sum \nolimits_{1}^{n} p_{{{\text{ik}}^{2} }} } }}\), which measures the niche overlap between species j and k, where pij is the proportion of the ith resource to the species j, pik is the proportion of ith resource to the species k, and n is the number of resource categories (Pianka, 1973). This metric ranges from zero (no overlap) to 1 (perfect overlap). The significance of niche overlap among groups (species and life stages) was tested by comparison with a null model based on the RA3 algorithm in the package EcoSimR v0.1.0 of Gotelli & Ellison (2013) with 9999 Monte Carlo randomizations. The RA3 algorithm maintains the niche breadth of each group and randomizes entries in the resource matrix by assuming all resources are used equally by all groups. To reveal dietary patterns among life stages and species, we performed Principal Coordinate Analysis (PCoA) using Ai values; significance of between-group dietary variation was tested using Permutational Multivariate Analysis of Variance (PERMANOVA), using 9999 permutations based on a Bray–Curtis dissimilarity matrix and the package ‘vegan’ 2.4-5 (Oksanen et al., 2017) in R.

Isotopic niche

Prior to numerical analysis, isotopic ratios of samples from different years, sites and preservation methods (i.e., frozen vs. salt) for each species and life stage were tested for potential isotopic differences using the Kruskal–Wallis test due to heteroscedasticity of the data. Given no significant differences, samples were retained in a merged dataset. Interspecific and ontogenetic partitioning of isotopic space was evaluated using the package Stable Isotope Bayesian Ellipses in R (SIBER) version 2.0.2, which estimates isotopic spaces and their overlap for groups (Jackson & Britton, 2014). The isotopic space occupied by each life stage of each species was estimated using sample-size-corrected standard ellipse areas (SEAC ‰2) (Jackson et al., 2011). This analysis infers that relationships based on isotopic space reflect trophic niche relationships (Layman et al., 2007). Lipid concentrations were lower than 5% (C:N ratios < 3.5 for aquatic animals); therefore, samples were analyzed without lipid correction (Skinner et al., 2016).

Trophic positions were estimated using the method proposed by Vanderklift & Ponsard (2003). Trophic position was calculated as TP = [(δ15Nfish − δ15Nsnail)/2.54] + 2, where δ15Nfish was the average δ15N for a particular life stage and species; δ15Nsnail was the average δ15N of a common primary consumer; 2.54 is the average enrichment in δ15N per trophic level (Vanderklift & Ponsard, 2003), and 2 corresponds to the trophic level of the primary consumer. For the primary consumer, we used the soft tissues of Doryssa starksi (Baker, 1913), a freshwater snail common in the rapids. To avoid bias in TP outcomes inherent to different places or periods (Jepsen & Winemiller, 2007; Zaia Alves et al., 2017), Doryssa starksi were collected from the same rapids during the fish surveys, such that spatial or temporal variation in isotopic ratios should not bias TP estimates (Jepsen & Winemiller, 2007; Zaia Alves et al., 2017),

Results

Morphological traits

PCA showed that the six groups (three species, each with two life stages) occupied separate regions of morphological trait space (PERMANOVA, Pseudo-F = 190.2; P < 0.001). Juvenile and adult Ossubtus showed the greatest separation from other groups (Fig. 2). The first two principal components together modeled 57.6% of total morphological variation among life stages and species (Table 3). PC1 (37.6% of variance) identified a gradient contrasting fishes with relatively deep bodies, narrow caudal peduncles, small eyes, wide and subterminal mouths, small olfactory chambers and long gastrointestinal tracts versus fishes with the opposite suite of attributes. PC1 therefore involved both habitat-use and trophic associated traits, and separated adult Ossubtus from remaining groups, especially juvenile and adult Myleus that both have long gastrointestinal tracts and wide mouths. PC2 (20.0% of variance) identified a different gradient contrasting fishes with relatively small and shallow bodies, high pectoral fin aspect ratios, supraterminal mouth orientation, and long heads versus fishes with the opposite combination of traits (Table 3). PC2 separated juvenile Ossubtus, a group with a relatively high aspect ratio of the pectoral fin and a longer head, from the remaining serrasalmids (Fig. 2).

Fig. 2
figure 2

Plot of principal components analysis axes 1 and 2 derived from analysis of 14 ecomorphological traits of juvenile (dashed lines) and adult (solid lines) Myleus (red ellipses), Ossubtus (green ellipses), and Tometes (blue ellipses) collected during the low-water period in lower Xingu. Ellipses represent 95% confidence intervals (individual data points not shown for clarity). High variable loadings on axes are indicated by vectors: APC aspect ratio of pectoral fin, GIT relative gastrointestinal tract length, HL relative head length, MOR mouth orientation, OFO relative olfactory fossae opening, RBD relative body depth, SL standard length, WCP relative width of caudal peduncle, WMO relative width of the mouth

Full size image
Table 3 Principal components analysis dominant axis (PC 1, 2) scores derived from 14 functional morphological traits of three rapids-adapted
Full size table

Diet

According to the alimentary index, juvenile and adult Myleus fed mostly on allochthonous plants (Ai = 51.8%, and Ai = 38.9%, respectively). Juvenile Ossubtus fed primarily on aquatic macroinvertebrates (Ai = 29.8%), whereas adult Ossubtus consumed Podostomaceae (Ai = 76.9%). Juvenile Tometes fed mostly on allochthonous plants (Ai = 26.5%), whereas the diet of adult Tometes was dominated by Podostomaceae (Ai = 62.2%; Supplementary Table S1). Whittaker’s index of association revealed that the diet of adult Ossubtus, which was dominated by Podostemaceae (Appendix 1), was most differentiated from diets of other groups (Fig. 3). Juvenile Ossubtus fed mostly on macroinvertebrates (Appendix 1), but grouped with juvenile Tometes that fed mostly on allochthonous plant material (Fig. 3).

Fig. 3
figure 3

Alimentary index (Ai%) of juvenile (J) and adult (A) Myleus, Ossubtus, and Tometes collected during the low-water period in lower Xingu. Cluster diagram shows the similarity of food category consumption based on Whittaker’s index of association

Full size image

The first two PCoA axes explained 66.7% of the total variance in diet (Fig. 4). High scores on PCo1 were associated with consumption of terrestrial plants, aquatic macroinvertebrates, periphyton, sediments, and seeds. High scores on PCo2 were associated with greater consumption of Podostemaceae, fins and scales (Table 4). PCoA revealed two groups, one corresponding to adult Ossubtus, and another formed by juveniles and adults of the other two species plus juvenile Ossubtus (Fig. 4). Juvenile Ossubtus overlapped with adults of Myleus and Tometes, whereas juveniles of Myleus and Tometes were different (Fig. 4). PERMANOVA analysis confirmed significant dietary differentiation among the three species when both stages were combined (Pseudo-F = 35.09; P < 0.001), between the two ontogenetic stages within species (Pseudo-F = 13.21; P < 0.001), and among the six combinations of species and ontogenetic stages (Pseudo-F = 14.69; P < 0.001).

Fig. 4
figure 4

Dietary niche of juvenile (dashed lines) and adult (solid lines) Myleus (red ellipses), Ossubtus (green ellipses), and Tometes (blue ellipses) collected in low-water period in lower Xingu River basin. Principal coordinate analysis, axis 1 and 2 derived from analysis of diet composition, AP allochthonous plants, DT detritus, FS fins and scales, MC macroinvertebrates, PD Podostemaceae, PP Periphyton, SD seeds, ST sediments. Ellipses represent 95% confidence intervals of juveniles (individual data points not shown for clarity)

Full size image
Table 4 Principal coordinates (PCo) scores derived from eight prey items consumed by the three rapids-dwelling serrasalmids from the lower Xingu during the dry season
Full size table

The largest ontogenetic shift in diet breath was observed for Ossubtus. Diet breadth of this species declined with body size (juveniles’ B = 6.58; adults’ B = 1.78). Myleus and Tometes had less variable diet breadth, with Myleus displaying lower values for juveniles (B = 4.08) than adults (B = 4.84) and Tometes displaying greater values for juveniles (B = 6.02) than adults (B = 2.64) (Supplementary Table S2).

Dietary overlap between groups was variable, ranging from 0.23 (between juvenile Myleus and adult Ossubtus) to 0.99 (between adults of Ossubtus and Tometes). Myleus and Ossubtus had low interspecific dietary overlap for both juveniles (Oij = 0.49) and adults (Oij = 0.29), whereas dietary overlap between Ossubtus and Tometes was high for juveniles (Oij = 0.83) and adults (Oij = 0.99) (Supplementary Table S2). Based on comparison with null model estimates, observed pairwise dietary niche overlap among all serrasalmid species and life stages was significantly higher than expected at random [observed mean value (0.60) > simulated value (0.51); P < 0.05].

Isotopic patterns

Stable isotope signatures of samples from different survey periods and sites were not significantly different for juvenile and adult Myleus (carbon: H = 0.98, df = 2, P = 0.61; nitrogen: H = 1.80, df = 2, P = 0.41), juvenile and adult Tometes (carbon: H = 5.42, df = 4, P = 0.25; nitrogen: H = 4.78, df = 4, P = 0.31), and adult Ossubtus (carbon: H = 5.97, df = 4, P = 0.20; nitrogen: H = 7.59, df = 4, P = 0.11). Juvenile Ossubtus, in contrast, showed significant differences for carbon (H = 5.40, df = 1, P < 0.05) between samples collected in September 2013 (two salted samples) and October 2015 (seven frozen samples). This difference for carbon was likely associated with inter-annual variation in isotopic ratios of food sources rather than preservation method; however, the two samples preserved in salt were excluded from subsequent analyses. Nitrogen isotopic ratios did not differ significantly among juvenile Ossubtus captured during different periods (H = 1.67, df = 1, P = 0.20). Carbon signatures had large overlap among species and ontogenetic stages (Table 5). In contrast, δ15N was more differentiated among species and life stages (Fig. 5). δ15N of juveniles and adults of Myleus and Tometes varied relatively little, whereas the range of values for Ossubtus differed considerably between juveniles and adults.

Table 5 Range, mean, and standard deviations (SD) of carbon (δ13C) and nitrogen (δ15N), calculated trophic position (TP) according to Vanderklift & Ponsard (2003), and corrected standard ellipsis areas (SEAc) values for the three serrasalmid species from Xingu River basin
Full size table
Fig. 5
figure 5

Isotopic niches of juvenile (dashed lines) and adult (solid lines) Myleus (red ellipses), Ossubtus (green ellipses), and Tometes (blue ellipses) collected from rapids during the low-water period in lower Xingu. Standard ellipse areas based on axis 1 and 2 scores were estimated using Stable Isotope Bayesian Ellipses in R (Jackson et al., 2011). Ellipses represent 95% confidence intervals of juveniles (individual data points not shown for clarity)

Full size image

Juvenile and adult Tometes and Myleus had relatively large overlap in isotopic space, representing 42.9% of total ellipse area for Tometes and 45.7% of total ellipses area for Myleus. Juvenile Tometes occupied a broader isotopic space (SEAC = 3.99‰) than adults (SEAC = 1.71‰) (Table 5). Myleus occupied a relatively small isotopic space for both juveniles (SEAC = 1.83‰) and adults (SEAC = 2.52‰). Juvenile and adult Ossubtus had no isotopic overlap, and juveniles occupied a smaller isotopic space than adults (SEAC; 2.05‰ versus 4.91‰, respectively; Table 5). Both juvenile and adult Ossubtus did not overlap with any group (Fig. 5). Based on δ15N values, and assuming all else being equal with regard to isotopic signatures of basal resource of food chains supporting these consumers, trophic positions ranged from 2.88 for adult Tometes to 3.54 for adult Ossubtus (Table 5). Juvenile and adult Tometes had significantly different trophic positions (H = 4.34, df = 5, P = 0.03), with TP = 2.91 and 2.88, respectively. Juvenile and adult Ossubtus also had significantly different trophic positions (H = 7.09, df = 5, P < 0.001), with juveniles having the lowest TP (2.97), and adults having the highest among all species and life stages (TP = 3.54). Myleus had an intermediate position (juvenile TP = 3.15, adult TP = 3.24) without a significant difference between life stages (H = 0.94, df = 5, P = 0.35) (Table 5).

Discussion

The herbivorous serrasalmids that coexist in rapids of the lower Xingu during the low-water period were well separated along gradients of morphological traits associated with feeding and swimming performance. Varying degrees of morphological differentiation were observed between juveniles and adults of each species. Morphological and trophic differentiation was greatest between Ossubtus and the other two species, and Ossubtus also revealed greatest ontogenetic differentiation, which may partially reflect differences in microhabitat use. Ossubtus seems to be more strictly confined to rapids habitats than Myleus or Tometes (Andrade et al., 2016c). Both juvenile and adult Ossubtus possess a subinferior mouth, as opposed to the terminal mouth position of Myleus and Tometes, as well as a shallower (more streamlined) body that should further reduce drag in fast-flowing water. Juvenile Ossubtus have higher values for the aspect ratio of the pectoral fin and lower values of body depth when compared with adults, suggesting that juveniles are particularly well adapted to inhabit fast-flowing water (Andrade et al., 2016c). The relatively narrow caudal peduncle of juvenile Ossubtus may further increase their swimming efficiency in fast water. Myleus and Tometes have relatively greater body depths and narrower caudal peduncles, which are less efficient for swimming in fast currents, but greatly enhance maneuverability. This body shape would be advantageous in rapids complexes that contain areas with slower water velocities where foraging can involve lateral movements without displacement from hydraulic resistance.

There were some discrepancies between dietary and isotopic patterns. Dietary overlap between life stages was high for both Myleus and Tometes. In contrast, Ossubtus had low dietary overlap between juveniles and adults, and the isotopic space occupied by juveniles also was different than the space occupied by adults. For Myleus and Tometes, the isotopic spaces of juveniles and adults overlapped extensively. If we assume that δ15N accurately reflects vertical trophic position, Tometes in both life stages and juvenile Ossubtus occupied the lowest trophic positions. This inference contrasts with our dietary analysis that indicated Tometes and juvenile Ossubtus consumed the largest fractions of macroinvertebrates and therefore should occupy higher trophic positions. Based on δ15N, Myleus occupied an intermediate trophic position and adult Ossubtus occupied the highest position among these three species, corroborating the expectation based on its shorter gut length. Conversely, based on dietary analysis, adult Ossubtus consumed mostly aquatic macrophytes, and therefore should occupy a low trophic position. This helps to understand the dietary difference among these species, but not variation in trophic position. Trophic position generally is negatively correlated with gut length in fishes (Wagner et al., 2009). Among frugivorous serrasalmids, those species with relatively shorter guts had higher trophic positions (Correa & Winemiller, 2014). Adult Ossubtus have a shorter relative gut length and higher estimated trophic position than Myleus and Tometes, as well as conspecific juveniles.

Differences in trophic position inferred from isotopic and dietary analysis could be due to variation in trophic fractionation values between species or ontogenetic stages. Some herbivores have been shown to have trophic fractionation values for δ15N higher than 2.54, the value used for our study (Caut et al., 2009; German & Miles, 2010). This would result in an overestimate of trophic position for strictly herbivorous species and may explain the high trophic position found for adult Ossubtus despite the large amount of macrophytes found in their stomachs. Discrepancies between dietary and isotopic data also could result from temporal variation in δ15N of aquatic macrophytes. Stable isotope ratios of muscle tissue reflect assimilation of material consumed over a timescale of several weeks to months (Vander Zanden et al., 2015), whereas dietary analysis represents a snapshot of food resources consumed by fish minutes before capture. The isotopic turnover rate is faster for white muscle tissue than other tissues such as scales or bones; however, complete muscle turnover can take up to three months (Busst & Britton, 2017). That means that the isotopic signature in muscle tissues of the evaluated fish species in our study should reflect the diet several weeks before the fish’s capture. Hydraulic conditions of local habitats also might affect isotopic fractionation associated with plant physiological processes (Correa & Winemiller, 2014), such as material exchanges at the cell-water boundary. Thus, we cannot rule out that isotopic ratios of rapids-dwelling herbivores might reflect food resources consumed from a different location or perhaps even from the same location but under different flow conditions. Faster growth rates of juveniles (Vander Zanden et al., 2015) should result in faster isotopic turnover of juvenile tissues and better reflection of contemporary local conditions. Consequently, the isotopic ratios of adult Ossubtus might reflect assimilation of food consumed a few months prior when aquatic macrophytes had different isotopic ratios or alternative food resources were exploited. However, Zuluaga-Gómez et al. (2016), studying Xingu fishes, inferred that phytomicrobenthos was the most important basal source supporting biomass of rapids-dwelling serrasalmids. Zuluaga-Gómez et al. (2016) only studied adult fishes, and because phytomicrobenthos samples probably contain some combination of benthic algae and microfauna (Zeug & Winemiller, 2008), this source could partially account for the relatively high δ15N of adult Ossubtus.

Despite the lack of congruence between dietary and isotopic results, both datasets revealed clear separation between Ossubtus and the other two serrasalmids, especially among adults. All three species feed within rapids, but not all food resources are autochthonous in origin, and terrestrial plant material and terrestrial arthropods can drift through these habitats. Myleus and Tometes displayed large dietary overlap, but relatively low overlap in isotopic space. Compared to Ossubtus, juveniles and adults of these species are trophic generalists with similar morphologies and diets. Ossubtus diet breadth was greater for juveniles than adults, but the isotopic niche space occupied by juveniles was smaller than that occupied by adults. In contrast, dietary and isotopic patterns were congruent for Myleus and Tometes, with adults and juveniles overlapping extensively in both dietary and isotopic space. Adult Tometes feed heavily on Podostemaceae, but do not appear to target this resource to the same degree as Ossubtus. Juvenile Ossubtus consumed mostly aquatic macroinvertebrates and Podostomaceae, whereas adults Ossubtus apparently avoid consuming aquatic invertebrates that use these plants as habitat.

Our findings suggest that food resource partitioning is not a major mechanism for coexistence of herbivorous serrasalmid fishes inhabiting rapids during the annual low-water period in the lower Xingu River. Some resources, such as Podostomaceae and aquatic insect larvae, are abundant in rapids during the low-water period. Interspecific dietary differences were most associated with the greater importance of Podostomaceae for adult Ossubtus and Tometes when compared to both juvenile and adult Myleus that consumed more allochthonous plants and relatively little Podostomaceae (Supplementary Table S1). Similarly, juvenile Ossubtus and Tometes, which coexist in rapids within macrophyte beds or behind rocks that provide hydraulic refugia (Andrade et al., 2016a, c), had broad diets, whereas juvenile Ossubtus fed mostly on aquatic macroinvertebrates (Supplementary Table S1). In addition, juvenile Tometes fed more heavily on allochthonous plants than juveniles of the other two species. Future research should examine how seasonal variation in resource quantity and quality influences the trophic ecology and habitat use patterns of these fishes. High discharge during the wet season creates harsh hydraulic conditions in rapids even for rheophilic species, and likely changes availability of various food resources and microhabitats that provide refuge from swift currents. Among the three species studied here, Ossubtus appears to be most specialized for living and feeding in rapids. High dietary overlap between Myleus and Tometes does not preclude niche partitioning along alternative niche dimensions, such as the exploitation of resources in different microhabitats either temporally, or spatially. Indeed, previous work on fish assemblages within rapids of the Xingu found patterns of functional diversity suggesting many species were adapted to exploit similar resources in a variety of microhabitats (Fitzgerald et al., 2017).

Interspecific morphological differences also seem to reflect variation in foraging behavior. The smaller eyes and larger olfactory chamber of Ossubtus suggest that vision may be less important than olfaction when searching for food. In teleost fishes, a larger olfactory chamber often is associated with a greater number of olfactory folds that enhance the sense of smell (Døving et al., 1977). Considering that the Xingu is a clearwater river with very low turbidity and good visibility, this characteristic of Ossubtus may be associated with nocturnal foraging, whereas Myleus and Tometes probably forage diurnally. Indeed, most specimens of Ossubtus were collected at night, while Myleus and Tometes were most frequently captured during daylight hours. These diel differences in foraging behavior may enhance niche segregation among species.

Microhabitat use by fishes has been shown to vary in response to natural flood pulses (Kluender et al., 2017), and hydrological modification has been found to shift basal resources supporting aquatic food webs and altering stable isotope ratios (Delong & Thoms, 2016). River impoundment results in longer food chains within the reservoir, which corresponds to more trophic transfers between basal resources and top predators (Hoeinghaus et al., 2008). Operation of the Belo Monte Hydroelectric Complex is impacting the hydrology and habitats in the lower Xingu River, and threatens populations of fishes and other aquatic organisms, especially those adapted to live in rapids (Sabaj Pérez, 2015; Fitzgerald et al., 2018). Ossubtus and Tometes are particularly vulnerable because they are endemic to the basin (Andrade et al., 2016c; Winemiller et al., 2016). Hydroelectric dams have been shown to impact the ecology of rapids in other South American rivers (Horeau et al., 1998), and it is likely that the ecology and population dynamics of serrasalmids and other rapids-dwelling species of the Xingu will change over the next several years. Future studies should analyze samples collected over an entire annual cycle as well as the same season across multiple years. More research is needed to examine spatial and temporal variation in the isotopic ratios of potential food resources and aquatic consumer taxa. In addition, laboratory experiments are needed to improve understanding of isotopic turnover rates of tissues of consumers and resources in rapids as well as fluvial habitats generally. In conclusion, herbivorous serrasalmids from rapids of the Lower Xingu River revealed ontogenetic dietary shifts and high trophic niche overlap during the dry season, suggesting that food resources may not be limiting or else these species feed within different microhabitats. In addition to providing some of the first ecological information from this poorly studied system, this study reinforces the need for approaches that analyze multiple data sources and spatiotemporal scales.