Morphological variation among populations of Hemigrammus coeruleus (Characiformes: Characidae) in a Negro River tributary, Brazilian Amazon

We explored patterns of phenotypic variation in Hemigrammus coeruleus from the Unini River basin, a blackwater river in the Brazilian Amazon. Geometric morphometrics was used to evaluate variation in body shape among populations from four tributaries (UN2-UN5). We found no evidence for sexual dimorphism in body size and shape. However, morphological differences among populations were detected as the analyses recovered significant groups corresponding to each sub-basin, with some overlap among them. The populations from UN2, UN3 and UN5 had more elongate bodies than fish from UN4. The most morphologically divergent population belonged to UN4, the tributary with the most divergent environmental conditions and the only one with seasonally-muddy waters. The morphological variation found among these populations is likely due to phenotypic plasticity or local adaptation, arising as a product of divergent ecological selection pressures among sub-basins. This work constitutes one of the first to employ a population-level geometric morphometric approach to assess phenotypic variation in Amazonian fishes. This method was able to distinguish subtle differences in body morphology, and its use with additional species can bring novel perspectives on the evaluation of general patterns of phenotypic differentiation in the Amazon.


Introduction
Causes and processes driving population divergence have been central aspects of research in evolutionary biology (Bernatchez, 2004).Intraspecific phenotypic diversification is often produced and maintained by divergent selective regimes (e.g., Langerhans et al., 2003) arising from either phenotypic plasticity or genetic divergence (Langerhans et al., 2004).Although diversification is possible with gene flow (e.g., Gaither et al., 2015;Marques et al., 2016), divergence is more likely to occur among isolated populations (Rundle, Nosil, 2005).Adaptive phenotypic plasticity is predicted to be most advantageous in a heterogeneous environment when future conditions can be reliably predicted (Sultan, Spencer, 2002;Schlichting, Smith, 2002).
Water bodies of the Amazon basin provide a variety of habitats at large (Goulding et al., 2003) and local scales (Ferreira et al., 2007), and host the most diversified freshwater ichthyofauna in the world (Lowe-McConnell, 1999).A major portion of this diversity is composed of small-sized species inhabiting small floodplain rivers and streams, such as those found in the Negro River and tributaries.These species benefit from the annual flood pulse, which can range up to 13 m between the dry and wet seasons.These floods inundate large areas of rainforest (Goulding, 1997), providing shelter, food sources, and the possibility for population admixing for these so-called flooded-forest specialists (Piggot et al., 2011).
One species commonly found in floodplain areas is Hemigrammus coeruleus Durbin 1908, a small, ornamentally appealing characid, with a red lateral band, starting from the eye and ending at the caudal peduncle.This species has a maximum total length of 5.8 cm (Lima et al. 2003), and sexual dimorphism has been reported in old-age aquarium specimens (J.Zuanon, email, jzuanon3@gmail.com,September 2012).Although the species has also been found in the Trombetas and Nhamundá river basins (Reis, 2011), over 600 Km downstream from the confluence of the Negro and Solimões rivers, the main databases on neotropical fishes define its occurrence area as the Solimões River and lower Negro River (Buckup et al., 2007;Froese, Pauly, 2016).
One of the main tributaries of the lower Negro River is the Unini River, a low-gradient black-water tributary that contains several large tributaries (sub-basins), in which H. coeruleus is common and widespread.Although most of these sub-basins are permanently black water and present similar environmental features, Solimõezinho River, the smallest tributary, exhibits seasonally muddy waters and differs from the other tributaries in some geomorphological features (Tab.1).Either directly or indirectly, these differences may generate divergent selection pressures for fish populations in each sub-basin, and we could expect the specimens from this particular tributary to present distinct phenotypes.This scenario provides a good opportunity to study morphological variation within and among populations.
Our work aimed to identify patterns of morphological variation in H. coeruleus from populations of the Unini River basin, and discuss potential processes that created them.Thus, we addressed the following questions: (1) Does morphology differ between the sexes?(2) Does body shape differ between populations of hydrologically connected areas?(3) Are there environmental differences between areas that could be related to the patterns of morphological variations found?This study represents one of the first to employ a populationlevel geometric morphometric approach to evaluate patterns of phenotypic variation within and among populations of fishes from dynamic Amazonian environments.

Material and Methods
Study area.Amazonian blackwater rivers have their origin in Tertiary sandy-sediment terrain covered by pristine rainforest.Their waters are acidic, poor in nutrients and have a high concentration of dissolved humic acids (Sioli, 1985).The Unini River is a low-gradient black-water tributary of the Negro River, located at 1°20'; 2°25' S and 61°30'; 65° W, in central Amazonia, Amazonas State, Brazil (Fig. 1).Its basin lies within three major conservation units, the Amanã Sustainable Development Reserve (Amanã SDR), Jaú National Park (Jaú NP), and the Rio Unini Extractive Reserve (Rio Unini ER).Mean annual rainfall in the area is about 2,580 mm, with the annual floods causing river level variation of about 7 m between dry and rainy seasons (Almeida, 2014).Mean temperatures range from 26° to 26.7°C (DNPM, 1992).
Field sampling.Fish were collected from several localities in four tributaries (sub-basins) of the Unini River (Fig. 1), using hand nets and beach seines, on five occasions from August 2009 through November 2010.Shortly after collection, fish were anesthetized with clove oil and fixed 3 e160152[3] in 10% formaldehyde.After 10-30 days, fish were preserved in 70% ethanol.Voucher specimens were deposited in the reference collection of Fish Ecology Lab, UFRJ, under the number DEPRJ 8372.
Sampling localities in the Preto River (UN2) were located at least 30 km upstream from its confluence with the upper Unini River.The Arara River (UN3) is located 60 km downstream from the Preto River, and fish sampling was performed at sites located at least 30 km upstream from the Arara River mouth.Moving downstream for an additional 90 km, is the Solimõezinho River (UN4), where sampling was performed from its mouth to sites 20 km upstream.The Pauini River (UN5) was the lowest sub-basin sampled.Fish were collected at least 40 km upstream from its confluence with the Unini River.
During the collection of specimens, water temperature, electrical conductivity, pH and dissolved oxygen were measured using a YSI 556 multiprobe system (Yellow Springs Instruments).Main stem width was measured in the field using a Garmin GPS 60CX and water colour was visually categorized into three types (black, clear and muddy).Solimõezinho River (UN4) was the only tributary with seasonally muddy waters (Tab.1).Structural and geomorphological features for each sub-basin were also obtained from satellite images using softwares Google Earth and ArcMap version 10.1 (Esri Inc.).Sinuosity was calculated by dividing the total extension of the river's main stem by the vectorial distance between the extremes of the main stem (Schumm, 1956).Environmental features are presented in Tab. 1.  Data collection.Geometric morphometric methods were used to estimate morphological variation for H. coeruleus.
We measured the standard length of 121 adult specimens that were then placed on grid paper, and the left side of each was photographed using a GE X5 digital camera in macro mode.Coordinates were obtained by digitization of 11 anatomical landmarks using software TpsDig (Rohlf, 2004a) in order to obtain the shape of each specimen.Landmarks were chosen based on previous studies with neotropical fishes (Langerhans, DeWitt, 2004; Gomes-Jr., Monteiro, 2008;García-Alzate et al., 2010) and are shown in Fig. 2. We used the centroid size (the square root of the summed squared distances from each landmark to the configuration centroid, which is the average of all landmarks) as a bodysize measurement in all analyses (Bookstein, 1991).The landmark configurations were superimposed by Procrustes fit (Rohlf, Slice, 1990) using MorphoJ version 1.05f (Klingenberg, 2011) in order to obtain the shape of each specimen.This method removes the effect of size by placing the centroid of the specimens at the origin of the Cartesian System (0,0).Procrustes fit also adjusts the centroid size to one (1), and rotates the configuration until the sum of squares of the distance between corresponding landmarks is the lowest possible allowed for the estimation of a mean shape (Bookstein, 1991;Monteiro, Reis, 1999;Zelditch et al., 2004).
Data analysis.Morphological variation between the sexes was assessed by dissecting 72 previously photographed specimens (25 males, 47 females), including representatives from all sub-basins.The sex ratio of collected specimens was 1:1.88 (males/females); this does not differ from the reported sex ratio for the species in the Unini River (T.Barros et al., unpublished data).To test for differences in shape between the sexes, a MANOVA/Hotelling T-square was performed using the first seven principal component axes (which account for 82.46% of total variation) in R stats package (R Development Core Team, 2015).This was done using the principal components of a PCA that used the Procrustes coordinates generated in MorphoJ version 1.05f (Klingenberg, 2011).To test if there were differences in body size related to sex, a Student T test was performed using centroid size.
To analyze morphological divergence among the Unini River sub-basins (UN2, UN3, UN4 and UN5), we used a total of 92 adult specimens (> 20.9 mm SL), 23 from each tributary.Body-size differences among populations were assessed by one-way ANOVA.Significant differences between each population were determined by a post hoc Tukey test.Body-shape variation among sub-basins was assessed by submitting the matrix of shape variables to a canonical variate analysis (CVA), using MorphoJ.A discriminant function analysis (DFA) was performed to assess differences in shape among and within sub-basins, in R MASS package (Venables, Ripley, 2002).Differences in shape among populations were visualized through the grids of deformation generated by the software TPSRegr (Rohlf, 2004b), using landmark coordinates and scores obtained from CVA.To evaluate the influence of ontogenetic allometry on shape variation, a linear regression analysis was performed in MorphoJ (Klingenberg, 2011).This regression included the Log of centroid size and the scores of principal components.All graphs were generated using the statistics program R (R Development Core Team, 2015).

Results
Fish used in the analysis for sexual dimorphism ranged from 20.4 mm to 40.2 mm SL.There was no evidence for morphological variation between the sexes (Fig. 3).The sexes do not differ statistically in the first seven principal components (accounting for 82.46% of total variation in body shape), as evidenced by MANOVA (Hotelling T2=0.036,P=0.934).Body size also did not differ between the sexes as indicated by the Student T test (P=0.55).
Fish used in the population analyses ranged from 20.9 mm to 40.2 mm SL.ANOVA revealed significant differences in body size among sites (F=12.23;P<0.0001).The largest body sizes were found in UN5, the lowermost sub-basin, and the smallest in UN2, the uppermost sub-basin (Fig. 4).Differences between populations are presented in Tab. 2.  Tributaries are ordered from upstream to downstream.
Samples were ordered by canonical variates, and the first two canonical axes depicted 81.66% of the variation among sub-basins.The results of the linear regression indicated that ontogenetic allometry accounted for only 10.88% of total variation, therefore, it was opted not to regress this variation out from the dataset.The first canonical axis (47.34%) ordinated the samples from UN4 with negative scores and samples from UN5 with positive scores.The majority of specimens from UN2 and UN3 samples presented intermediate scores.Body shape variation in the first axis was related to body depth and dorsal-fin position.Samples from UN5, with positive scores, showed more elongate bodies and dorsal fins located more posteriorly relative to samples from UN4, which exhibited negative scores (Fig. 5).The second canonical axis (34.31%) ordinated samples from UN4 and UN5 mostly with positive scores, UN3 with intermediate scores and UN2 with negative scores.The shape difference associated with the second axis was related to head size, the length of the caudal peduncle and position of the opercle endpoint.Samples from UN2 and UN3 showed a larger head, a longer caudal peduncle and a more ventrally positioned opercle than samples from UN4 and UN5 (Fig. 5).Although some degree of overlap between sub-basins was observed (especially between UN2 and UN3), two groups were strongly defined.Based on body shape, the discriminant analysis correctly classified 91.3% and 82.6% of specimens from UN4 and UN5, respectively (Tab.3).

Discussion
Differences in body morphology can be used to detect local adaptation among populations (Marques et al., 2016).However, within populations there can be variation due to sexual maturity (Reiss, Grothues, 2015) and sexual dimorphism (Gomes-Jr.,Monteiro, 2008).Thus, we investigated morphological divergence related to sexual dimorphism prior to assessing population divergence.Within the Characidae, sexually dimorphic characters have been reported in several species.Most dimorphic characters are associated with sexual structures, such as the presence of hooks in anal and pelvic fins (Menezes et al., 2009;Dala-Corte, Fialho, 2014), sexual glands (Menezes, Weitzman, 2009;Menezes et al., 2009), or even fin shape (Menezes et al., 2009), but body shape (Ortega-Lara et al., 2002;Riehl, Baensch, 2007) and body size (Mazzoni et al., 2005;Mazzoni, Petito, 1999) have also been reported to differ.For H. coeruleus, aquarium observations of old-age individuals (approximately 7 cm SL) indicate differences between males and females (J.Zuanon, email communication, jzuanon3@gmail.com,September 2012) related to body colour, and fin length and colour, which are not possible to detect with the landmarks used here, and in shape and size (females slightly bigger and slender).However, in the populations of H. coeruleus studied herein (largest individuals around 4 cm SL) there were no differences between the sexes in size and shape.Mortality

Morphology of Hemigrammus coeruleus in Amazonian rivers
Neotropical Ichthyology, 15(1): e160152, 2017 6 e160152 [6] rates in natural habitats differ dramatically from artificial environments, and individuals appear unlikely to survive long enough in the wild to reach large sizes (T.Barros et al. unpublished data), when sex differences in body shape could become apparent.
Despite limiting our analyses to only mature adult specimens we found size differences among populations.The uppermost sub-basins (UN2 and UN3) contained small-sized individuals while the largest individuals were collected in UN5 (the lowermost sampled).Since it has been shown that body and head shape change during the course of fish growth, there was some concern that phenotypic divergence was a product of differences in size between populations.Thus, we tested for differences associated with allometric growth among populations, but this accounted for only 10.88% of all variation.
Since population divergence was found to be little influenced by the size of individuals or sexual dimorphism (intrinsic factors), we believe that divergent environmental pressures acting in the Unini River tributaries are the likely cause of phenotypic differences.Even with a high degree of overlap between populations from UN2 and UN3, due to high morphological variation within each group, the analyses using landmark coordinates were able to detect morphological differences among populations resulting in significant groups corresponding to each sub-basin.Specimens from UN4 and UN5 differed from the other populations, evidenced by the two strong goups formed in the CVA, with the Solimõezinho (UN4) distinctly diverging from the continuous formed by UN2 and UN3.This pattern suggests that, in addition to two groups, there is a gradient of variation in body shape, which is expected in populations that are not isolated, such as those evaluated here.The morphological distinctiveness of the UN4 population could be due to the unique physical characteristics of this river and / or differences in predation pressure.Below we discuss these possibilities.We found major differences in geomorphological descriptors among tributaries.These could directly or indirectly influence ecological conditions experienced by fish populations.The Solimõezinho River (UN4) exhibited, by far, the fewest lakes and narrowest floodplain.Composition of ichthyofauna is highly influenced by water flow, type of environment and habitat structure (Lowe-McConnell, 1999).Minor differences were found in physical and chemical features between the sub-basins; however, only the Solimõezinho River (UN4) had muddy waters.This feature apparently did not result in changes in the chemical properties of the water, as could be expected (usually, the more particulate matter carried, the higher the electrical and pH).However, it might play an important role in visually oriented predator avoidance and foraging behaviour (Rodríguez, Lewis, 1997), since transparency drops to only a few centimetres (compared to ≥ 1.4 m in the sediment-free black-water rivers).Evidence for morphological divergences due primarily to water transparency have been reported in other studies, such as Bartels et al. (2012) for populations of Perca fluviatilis.
Predation has received much attention in morphometric studies of freshwater fishes both in nature and in the laboratory, where it has been shown to act as an important selective pressure on body shape (Langerhans et al., 2004;Gomes-Jr., Monteiro, 2008;Burns et al., 2009;Webster et al., 2011;van Rijssel, Witte, 2013, Araújo et al., 2014).Elongate bodies, such as those from UN2, UN3 and UN5, are often associated with steady-swimming and relatively good swimming performance (Langerhans, Reznick, 2010).Langerhans et al. (2003) compared specimens of the cichlid B. wawrini and the characid B. caudomaculatus from lakes and rivers and showed that in lotic environments swimming performance is positively correlated with more elongate bodies.However, it was not possible for us to test for such a correlation.Additionally, elongate bodies are predicted to improve the ability to escape from pursuit predators (Langerhans, Reznick, 2009).Conversely, deeper bodies, as the ones found in UN4, can misdirect a predator strike away from the optimal capture point, decreasing prey vulnerability in short strikes (Webb, 1986).In lowtransparency conditions, pursuit foraging behaviour by visually oriented predators could be limited to very short distances or to ambush strikes only.Elongate bodies (as in UN2, UN3 and UN5) would not bring additional advantages under these conditions.Potential short-distance pursuit predators such as Cichla species, highly abundant in most of the Unini River and its tributaries, including UN2, UN3 and UN5, were not recorded in the Solimõezinho River (UN4) (H.Lazzarotto, unpublished data), which suggests their absence or very low abundance in this tributary.
In addition to predation, local habitat has been found to influence body shape divergence in other small characids (e.g., Langerhans et al., 2003;Sidlauskas et al., 2006).The variation in body shape of H. coeruleus in the most divergent tributary (UN4) could also be a result of local specialization to its unique habitat features.This result meets the predictions that the population with the most distinct morphology should come from the tributary with the most divergent environmental conditions.Langerhans (2008), studying phenotypic patterns between flow regimes, indicated that, on the intraspecific scale, both genetic divergence and phenotypic plasticity play important roles in phenotypic differentiation.Although genetic diversification is possible with gene flow (e.g., Gaither et al., 2015;Marques et al., 2016), structuring due to ecological selection or genetic drift is more likely under reproductive isolation (Rundle, Nosil, 2005).The current study uses specimens from distant localities, but a few individuals of H. coeruleus were also observed and collected in sites between tributaries located on the Unini River main stem.This suggests that the species is distributed along the entire length of the river, which would enable gene flow and individual mixing between tributaries.Furthermore, in Amazonian rivers, major environmental variation occurs seasonally when water temperature, conductivity, transparency, habitat structure and fish densities (influencing the degree of predation pressure) dramatically change (Sioli, 1985).This set of seasonally predictable conditions, that impose a major influence on the life-cycle of fishes by creating heterogeneous conditions, greatly extending suitable habitat, and facilitating gene flow and individual movement throughout the flooded forest (Piggot et al., 2011), should theoretically favor plasticity over local specialization (Sultan, Spencer, 2002;Schlichting, Smith, 2002).Our results, however, cannot determine whether the observed differences in shape among populations is due to phenotypic plasticity, genetic differentiation, or a combination of both.Future studies using a common garden experimental approach could determine whether observed morphological differences among populations have a genetic basis.Likewise, future molecular work in H. coeruleus could assess genetic differentiation, gene flow, and population interconnectedness.
Our morphometrics approach was able to distinguish subtle differences in body morphology among populations of H. coreuleus related to environmental differences.The replication of a similar approach in other fish species, either from the Unini River basin or from the Amazon, could help elucidate general patterns of phenotypic variation in seasonal environments.Furthermore, although yet to be tested in this way, this powerful tool could be useful in detecting recent conditions that populations were subjected to, such as environmental impacts.

Fig. 4 .
Fig. 4. Box-plot showing maximum, median, minimum values of centroid size of Hemigrammus coeruleus from four tributaries of Unini River.Quartiles correspond to 25% and 75% values.Tributaries are ordered from upstream to downstream.

Fig. 5 .
Fig. 5. Axes of morphological variation for population data set.Ellipses indicate the 95% confidence interval for each population.Deformation grids represent positive and negative extremes of the observed shape variation associated with the first and second canonical axes; N=92.
Tab. 3. Confusion matrix generated by Canonical VariateAnalysis based on shape variables of populations of H. coeruleus from four tributaries of Unini River.