Diet-morphology relationship in the stream-dwelling characid Deuterodon stigmaturu s ( Gomes , 1947 ) ( Characiformes : Characidae ) is partially conditioned by ontogenetic development

We tested whether interindividual variations in diet composition within a population of Deuterodon stigmaturus can be explained by morphological differences between individuals, and whether diet-morphology relationships are dependent on the ontogenetic development. We analyzed diet of 75 specimens sampled in a coastal stream of Southern Brazil. Variation in stomach content was summarized with a Principal Coordinate Analysis (PCoA). The retained PCoA axes were tested as response to standard length (SL), and to values of intestine length (IL) and mouth length (ML) independent of body size, using linear mixed-effects models (LMM). The most consumed food items by D. stigmaturus were filamentous algae (41%), terrestrial plants (20.3%), detritus (12%), and aquatic invertebrates (8.8%). The LMMs showed that SL was positively related to consumption of terrestrial plants, whereas IL independent of SL was negatively related to aquatic invertebrates and positively related to filamentous algae. When body sized was held constant, ML was not related to diet variation. Interindividual diet differences conditioned to body size suggest that individuals shift their trophic niche and function in the ecosystem along the ontogenetic development. Relationships between intestine length and diet composition suggest interindividual differences in foraging ability and digestibility of distinct food items.

A number of studies have documented that several characid species can alter their diets according to food availability (e.g.Kramer & Bryant, 1995b;Bojsen, 2005;Casatti & Castro, 2006;Uieda & Pinto, 2011), suggesting that characins may comprise one of the groups with highest feeding plasticity among freshwater fishes.For example, characids may present intraspecific variation in the ingestion of food items according to longitudinal location within a given stream (Manna et al., 2012).Also, characins can alter their diets in response to riparian deforestation (Bojsen, 2005).In addition to diet variation between populations, great differences were evidenced in stomach content of characid individuals within a given population, such as those owing to seasonal availability of food (Vitule et al., 2008;Uieda & Pinto, 2011), or from ontogenetic shifts (Drewe et al., 2004;Vitule et al., 2008;Manna et al., 2012).
Among the mechanisms that can lead to interindividual differences in diet of a species, some studies have suggested that morphological divergences between individuals that occur along the ontogenetic development play a relevant role (Kramer & Bryant, 1995a;Novakowski et al., 2004;Ward-Campbell & Beamish, 2005 ;Zhao et al., 2014).For example, Novakowski et al. (2004) demonstrated that three species of characins of the Roeboides Günther shift their feeding behavior and diet composition by increasing the amount of scale-eating along the ontogenetic development, and these changes match the migration of teeth to the exterior of the mouth.Drewe et al. (2004) found that Brycon guatemalensis Regan, 1908 shifts its diet from insectivorous to frugivorous, concomitant with an increase in intestine length, along the ontogenetic development.
Intestine length is an important trait associated with diet composition of fish species (Kramer & Bryant, 1995a, b).Species that ingest low nutritional and more difficult to digest food items usually have longer digestive tracts (Kramer & Bryant, 1995b).Therefore, detritivorous are expected to have longer intestines than herbivorous, which in turn are expected to have longer intestines than omnivorous and insectivorous species (Kramer & Bryant, 1995b;Davis et al., 2013).In addition, interindividual differences in intestine length within a population have been suggested to be related to the type of ingested food (German & Horn, 2006;Wagner et al., 2009;Scharnweber et al., 2011).Longer intestines are expected to more efficiently digest vegetal matter, detritus and less nutritional food resources (Kramer & Bryant, 1995b), which may potentially have an effect on individual fitness.Furthermore, size of intestine can be constrained by body size and thus larger individuals are expected to have longer intestines, which may also influence detection of changes in dietary composition along the ontogenetic development (Kramer & Bryant, 1995a;Drewe et al., 2004;Vitule et al., 2008).Despite the several well known potential sources of variation in feeding composition of fish individuals, it is still not fully understood how small-scale interindividual variations in morphological characteristics constrain feeding behavior and dietary composition of fishes.
Deuterodon Eigenmann is a genus of small-sized characins that inhabit stream ecosystems in southern and southeastern Brazil (Lucena & Lucena, 2002).Deuterodon species are known by the atrophy of their upper lip on the sides of the head leading to exposed teeth in the maxilla and premaxilla (Lucena & Lucena, 2002).These exposed teeth are used by Deuterodon species to graze the periphyton layer of streambeds and to browse leaves and submerged plant fragments (Sabino & Castro, 1990), though other items as aquatic insects can be commonly found in their diets (Vitule et al., 2008).Therefore, although the mouth of Deuterodon species suggests a specialized diet, studies evidence that species can present omnivorous and opportunistic behavior, depending on food availability (Sabino & Castro, 1990;Vitule et al., 2008), which is an expected characteristic for characin species without an apparent specialization in mouth morphology.Also, Vitule et al. (2008) indicated that diet of Deuterodon shifts along the ontogenetic development concomitant with an increase in the intestine length.
Regarding interindividual variation in diet and in morphology of characins, we aimed to investigate the feeding strategy of Deuterodon stigmaturus (Gomes, 1947), testing whether variations in stomach content is explained by interindividual differences in morphology, and whether morphology-diet relationships are related to ontogenetic development.We specifically aimed to answer the following questions: 1) Does diet composition of individuals with longer intestines include a higher proportion of low-protein and indigestible food items, such as detritus, vascular plants and algae?2) Is mouth length (measured as length of maxilla and premaxilla bones with exposed teeth) -hypothesized to be a Deuterodon feeding specialization -positively related to the ingestion of specific food such as periphyton? 3) Are shifts in diet composition and morphology conditioned to ontogenetic development (measured as standard length)?

Material and Methods
Study area and fish sampling.The study was carried out in the Paraíso stream, a second order tributary of the Mampituba River in the Southern Brazil coastal region (29°23'55"S 49°55'01"W, 16 m a.s.l.; Fig. 1).Mampituba is a mountain river that lies at the Atlantic forest region.Regional climate has well-defined seasons with a mean annual precipitation of 1,409 mm and mean monthly temperature of 18.3°C (Maluf, 2000).The sampled site comprised a 120-m long stream segment, which includes riffle-pool sequences with rocky bed and low turbidity.The channel of the sample site has about 10-m width, with 20-30 cm depth in the riffles and 80-100 cm depth in the pools.Local canopy shading is reduced due to sparse riparian vegetation.Sampling site was chosen because previous investigation has shown that the sampled site holds a great abundance of D. stigmaturus (Dala-Corte, personal observation).
We collected specimens of D. stigmaturus during five different months using a combination of sampling methods, which were both applied for riffle and pool habitats.Total number of individuals caught was 75; with a mean of 15, minimum of 10 and maximum of 18 individuals per month: August (n = 18) and September (n = 10) of 2010, and March (n = 16), May (n = 17) and July (n = 14) of 2011.The sampling methods included seine (5 mm mesh), gillnets (15 mm mesh) and cast nets (12 mm mesh).All captured fishes were anesthetized with eugenol (clove oil) before fixation with 10% formalin.In addition, in the same site where we sampled fishes, we also performed underwater observations of the feeding behavior of D. stigmaturus using method ad libitum (Sabino, 1999), in order to interpret the results in the light of the species' behavior.We carried out these observations in five additional field expeditions during the summer (December 2010 to February 2011).
Morphological traits.We used intestine length, mouth length and body size as morphological traits to predict diet variation because of previous information indicating importance of these variables to diet of Deuterodon and other fish species.Intestine length may constrain digestion of different food items (Kramer & Bryant, 1995b); mouth length may be important because of previous underwater observation of the feeding behavior of Deuterodon species (Sabino & Castro, 1990); body size is widely recognized by its importance to determine prey size and also to influence morphology and size of other morphological traits (Gatz, 1979).
To answer the questions about the effects of intraspecific variation in morphology on the dietary composition of D. stigmaturus, we first measured standard length, intestine length, and mouth length of 75 individuals.Standard length (SL) -used to infer ontogenetic development -was measured from the tip of the snout to the posterior end of the last vertebra, i.e., excluding the caudal fin.Intestine length (IL) was measured from the joint with stomach to the most distal part of the intestine (ending at urogenital opening), with the intestine fully extended.Mouth length (ML) was the length of maxillary and premaxillary bones with exposed teeth on the right side of mouth.All measurements were taken using a digital calliper with precision of 0.01 mm.
Standard length was highly correlated to IL (r = 0.95) and to ML (r = 0.99).Hence, in order to obtain values of IL and ML that are independent of SL (i.e.independent of ontogenetic development), we regressed IL and ML against SL and recorded their residuals.In other words, we calculated values of IL (IL residuals ) and ML (ML residuals ) that are not explained by body size, reducing their correlations with SL.The correlations of SL values with IL residuals (r < 0.001), and with ML residuals (r < 0.001) were almost zero, allowing us to interpret their independent effects on diet.We carried out these analyses using R (R Development Core Team, 2015).Stomach content.We analyzed a total of 75 individuals of D. stigmaturus, ranging from 29.11 to 101.35 mm SL (Table S1, supplementary material).This size range included individuals within the full range of sexual maturity according to the size at first maturity (L 50 = 63 mm SL) and also to the size at which 100% of individuals are mature (L 100 = 80 mm SL), provided by Dala-Corte & Fialho (2014).Macroscopic items found in the stomachs were quantified using an adaptation of the point method (Hynes, 1950;Hyslop, 1980).This adaptation consisted in quantifying the volume of each food item in a stereomicroscope using a graph paper under a Petri dish.Stomach contents were compressed to 1 mm height in order to allow calculating the volume in each cell of the graph paper (mm 3 ).This method allowed us to calculate the relative volume of each food item in the stomachs.
A large volume of items in the stomachs was too small to identify macroscopically.Then, we used a microscopic analysis to more accurately identify the items that compose the ingested detritus (i.e.particulate matter).Sub-samples of detritus found in each stomach were taken and 10 randomly selected fields in a transparent graph paper of 1-mm 2 were analyzed using a 100-fold microscopic increase.The number of times in which each microscopic food item was found was recorded in detritus analysis to calculate frequency of occurrence.Data analysis.To visualize possible patterns regarding the food items ingested by D. stigmaturus individuals, a matrix with Hellinger-transformed (Legendre & Gallagher, 2001) food items volume (i.e.dietary composition) was submitted to a Principal Coordinate Analysis (PCoA), using Bray-Curtis dissimilarity.We used the package BiodiversityR to test which axes explained significant percentage of variance based on the broken-stick criterion (Kindt & Coe, 2005).We thus retained the four first Principal Coordinate (PCo) axes because each one summarized a significant proportion of diet variation based on broken-stick criterion.Subsequently, to test whether intraspecific variation in stomach content differs due to changes in mouth morphology and intestine length, as well as standard length, we fitted liner mixedeffects models using the scores of the retained PCo as response variables and SL, IL residuals and ML residuals as fixed effect predictors, in a total of nine different models.Liner mixed-effects modeling allows model fitting using random effects as blocking variable to control for their effects on response variables (Bates, 2010).Thus, in each one of our fitted models, we used the five sampling months as random effect variable to control for this source of variation on individual's diet.Response variables (PCo) were submitted to logarithmic transformation to meet assumptions of normality and homogeneity of variance of model residuals.We interpreted the validity of models based on five criteria for model evaluation: (i) standardized beta coefficient (Std-beta), which is standardized mean slope value of the fixed effect; (ii) Z-value; which is the standard score or the number of standard deviations that the observation is above the mean; (iii) R 2 (fixed-effect), which consisted in the variation explained exclusively by the fixed effect variable; (iv) R 2 (full model), which is the summed variation explained by both fixed and random effects; and (v) P-value, which is the fixed effect significance.Calculation of the R 2 for the linear mixed-effects models were based on Nakagawa & Schielzeth (2013) and was carried out with the package MuMIn (Bartoń, 2015) for R environment (R Development Core Team, 2015).The PCoA and the linear mixed-effects models were carried out using the packages vegan (Oksanen et al., 2013) and lme4 (Bates et al., 2015), respectively, for R environment (R Development Core Team, 2015).

Discussion
We asked three questions in order to better understand the drivers of interindividual variation in diet composition of D. stigmaturus.Regarding the first question, when controlling for body size differences, we found that individuals with longer intestines ingested less aquatic invertebrates and a higher proportion of filamentous algae.However, diet of individuals with longer intestine did not shift completely to low-protein and indigestible food items.Regarding the second question, we did not find evidence that the length of maxilla and premaxilla bones with exposed teeth influence diet specialization of individuals, such as feeding on periphyton.We observe, however, allometric growth in mouth length in relation to standard length.This finding relates to the third question, as we found that standard length (used to infer ontogenetic development) also influenced diet composition, but standard length by itself did not completely explain dietmorphology relationships.When controlling for body size, our results revealed that variation in the intestine length between individuals led to two distinct trends: i) individuals with shorter intestines ingested more aquatic invertebrates in detriment of the other food items; and ii) individuals with longer intestines consumed more algae and detritus, but less terrestrial plants.We discuss these two trends in the next two paragraphs, respectively.
Studies have documented that some characid species may exhibit shifts in the intestine length coupled with reduction in the consumption of animal origin items along the ontogenetic development (e.g.Kramer & Bryant, 1995a,b;Drewe et al., 2004;Vitule et al., 2008).Differing from these studies, however, we did not find evidence that ingestion of animal origin items increases along the ontogenetic development.Rather, we found that these items were related to intestine length independent of ontogenetic development.The probable explanation for this difference is that intestine length did not show a pronounced allometric growth for D. stigmaturus along ontogenetic development, as it is expected to occur for several fish species (Kramer & Bryant, 1995a).We noted that in the linear regression between standard length (SL) and intestine length (IL), which we used to obtain residuals of intestine length, the relationship between IL and SL was almost linear, i.e. isometric growth.We therefore speculate that the extent of shifts that occur in diet composition along individual growth of a given species or a given population will be directly related to the degree of allometry between intestine and body size, i.e. when intestine grows disproportionally longer than the body does along the ontogenetic development.
The finding that D. stigmaturus individuals with longer intestines consumed more algae and less terrestrial plants when body size was held constant is counterintuitive, because we found the opposite relationship in the models with body size (SL).Even though we have controlled for body size in the intestine length values, raw intestine length values were highly correlated to body size (r = 0. 95), showing that in fact larger individuals are expected to have longer intestines.We did not find information in literature on differences in digestibility between algae and terrestrial plants for fishes.In addition, digestion efficiency may depend on taxonomic group, e.g.characids may be highly different from loricariids.Experimental manipulations of these food items and studies on digestion efficiency will greatly contribute to understand relationships between intestine length and diet variation for different species.
The opposite trend regarding algae ingestion between body size and intestine length independent of body size may be explained by other factors than intestine size by itself.For example, larger body size allows exploration of different habitats.Also, other morphological traits that shift concomitant with body growing, such as mouth morphology, might influence the ability of D. stigmaturus individuals to feed on different types of food items.In this sense, the combination of morphological characteristics that yields higher fitness for fish individuals can be conditioned by body size and thus may shift along the ontogenetic development.
We asked whether mouth length (ML, measured as the exposed teeth area at mouth sides) could explain diet variation between individuals.We expected this would be an important trait for the species feeding because of previous information about Deuterodon feeding behavior (Sabino & Castro, 1990), and also based on our underwater observations on foraging behavior of this species.However, when body size was held constant, our results did not provide support that small inter-individual difference in ML lead to variation in diet composition.However, ML by itself (raw values) was highly correlated to body size (r = 0.99).Hence, we cannot rule out that mouth morphology conditioned by ontogenetic development has influence on individuals feeding.For example, it is possible that a longer area with exposed teeth in large-sized adults confers an advantage to browse terrestrial plants or pick up plant fragments in the flowing water in relation to juveniles with small body size, which may explain the changes in diet composition along the ontogenetic development.
We observed that a great proportion of variation in diet composition of D. stigmaturus individuals was not explained by the morphological measures used by us, even though models included sample months as variable to control for temporal variation in food availability.This indicates that other sources of variability, such as feeding plasticity independent of the measured traits, are important factors influencing feeding of this species.This result is not surprising given previous studies that demonstrated the great feeding plasticity of fishes in relation to resource availability, mainly for characid species (Amundsen et al., 2001;Abelha et al., 2006;Uieda & Pinto, 2011;Manna et al., 2012).For Deuterodon, Vitule et al. (2008) also found a great variability in the diet of D. langei individuals.Additionally, our results are based on only a picture of the individual's stomach content and we did not evaluate past selective pressures for intestine size, such as those that are expected to occur in the early phases of the ontogenetic development.Regardless these other sources of variation, we found evidences supporting that small interindividual morphological variations play an important role constraining at least part of the diet and feeding behavior within a fish population.
In conclusion, our findings suggest that part of interindividual variations in diet composition within a given stream fish population may be constrained by morphological changes that occur along the ontogenetic development and also by morphological variations between individuals that are independent of ontogenetic development.These results corroborate previous suggestion that fishes may shift their functional and trophic niches along the ontogenetic development (Zhao et al., 2014).Furthermore, results indicate that a fish population can exhibit small-scale interindividual feeding variation due to divergences in the individual's morphology independent of ontogenetic development, which will influence the capacity of individuals to ingest and assimilate specific food items.These findings shed new light on the understanding of small-scale interindividual feeding variation within fish populations in stream ecosystems.

Fig. 1 .
Fig. 1.Location of the sampling site of Deuterodon stigmaturus (black dot) in the Paraíso stream, a tributary of the Mampituba River in Southern Brazil coastal region, between Rio Grande do Sul (RS) and Santa Catarina (SC) states.

Fig. 2 .
Fig. 2. Principal Coordinate Analysis (PCoA) showing PCo1 and PCo2 based on Bray-Curtis dissimilarities of diet composition of 75 individuals of the stream-dwelling characid Deuterodon stigmaturus.Food items name abbreviations are listed in Table3.

Fig. 3 .
Fig. 3. Principal Coordinate Analysis (PCoA) showing PCo3 and PCo4 based on Bray-Curtis dissimilarities of diet composition of 75 individuals of the stream-dwelling characid Deuterodon stigmaturus.Food items name abbreviations are listed in Table3.

Fig. 4 .
Fig. 4. Relationships between diet composition of Deuterodon stigmaturus summarized by the PCo1 of a Principal Coordinate Analysis (PCoA) and the standard length (SL) of individuals.Higher values of PCo1 indicate a greater proportion of terrestrial plants (Terrplant) and lower values indicate a greater proportion of filamentous algae (Filalgae) ingested by individuals.Sampling months (dashed lines) were used as random effect variable and SL was used as a fixed effect (solid line) variable to fit a linear mixed-effects model.

Fig. 5 .
Fig. 5. Relationships between diet composition of Deuterodon stigmaturus summarized by the PCo1 of a Principal Coordinate Analysis (PCoA) and intestine length independent of standard length (IL residuals ).Higher PCo1 values indicate a greater proportion of terrestrial plants (Terrplant) and lower values indicate a greater proportion of filamentous algae (Filalgae).Sampling months (dashed lines) were used as random effect variable and IL residuals was used as a fixed effect (solid line) variable to fit a linear mixed-effects model.

Fig. 6 .
Fig. 6.Relationships between diet composition of Deuterodon stigmaturus summarized by the PCo3 of a Principal Coordinate Analysis (PCoA) and intestine length independent of standard length (IL residuals ).Higher PCo3 values indicate a greater proportion of aquatic invertebrates ingested by individuals.Sampling months (dashed lines) were used as random effect variable and IL residuals was used as a fixed effect (solid line) variable to fit a linear mixedeffects model.

Table 1 .
Mean relative volume and correlation coefficients between food items and the Principal Coordinates (PCo) of a PCoA based on diet composition of the stream-dwelling characid fish Deuterodon stigmaturus.Boldface highlights the highest correlations.

Table 2 .
Mean relative volume of microscopic food items found in the stream-dwelling characid fish Deuterodon stigmaturus.Microscopic items were analyzed from the macroscopic food items previously classified as detritus.

Table 3 .
Twelve models fitted with linear mixed-effects modeling to predict diet variation (PCo1, PCo2, PCo3 and PCo4) of Deuterodon stigmaturus individuals based on morphological traits (fixed effect variables).Sampling months were used as random effect variable to control for temporal effects on diet variation.N = number of individuals; Df = degrees of freedom; Std-beta = Standardized beta coefficient (effect size); SE = Standard Error; Z-value = standard score or number of standard deviations the observation is above the mean; R 2 (fixed-effect) = variation explained exclusively by fixed effect variable; R 2 (full model) = variation explained by both fixed and random effect variables; P-value = model significance.Boldface highlights models with significant fixed-effect.