Research Article |
Corresponding author: Nelson Colihueque ( ncolih@ulagos.cl ) Academic editor: Devin Bloom
© 2017 Nelson Colihueque, Olga Corrales, Miguel Yáñez.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Colihueque N, Corrales O, Yáñez M (2017) Morphological analysis of Trichomycterus areolatus Valenciennes, 1846 from southern Chilean rivers using a truss-based system (Siluriformes, Trichomycteridae). ZooKeys 695: 135-152. https://doi.org/10.3897/zookeys.695.13360
|
Trichomycterus areolatus Valenciennes, 1846 is a small endemic catfish inhabiting the Andean river basins of Chile. In this study, the morphological variability of three T. areolatus populations, collected in two river basins from southern Chile, was assessed with multivariate analyses, including principal component analysis (PCA) and discriminant function analysis (DFA). It is hypothesized that populations must segregate morphologically from each other based on the river basin that they were sampled from, since each basin presents relatively particular hydrological characteristics. Significant morphological differences among the three populations were found with PCA (ANOSIM test, r = 0.552, p < 0.0001) and DFA (Wilks’s λ = 0.036, p < 0.01). PCA accounted for a total variation of 56.16% by the first two principal components. The first Principal Component (PC1) and PC2 explained 34.72 and 21.44% of the total variation, respectively. The scatter-plot of the first two discriminant functions (DF1 on DF2) also validated the existence of three different populations. In group classification using DFA, 93.3% of the specimens were correctly-classified into their original populations. Of the total of 22 transformed truss measurements, 17 exhibited highly significant (p < 0.01) differences among populations. The data support the existence of T. areolatus morphological variation across different rivers in southern Chile, likely reflecting the geographic isolation underlying population structure of the species.
Morphological variability, morphometry, multivariate analysis, Trichomycterus areolatus , truss-based system
Almost all species display morphological variation within and among populations in response to environmental and genetic factors, or as a consequence of behavioral and physiological differences (
Trichomycterus genus is an interesting group of catfishes for the studying morphological variation because most species have a wide distribution in different habitats across a broad altitudinal and latitudinal range in South America. Trichomycterus belongs to the family Trichomycteridae, which is native to southern Central and South America (
In Chile, Trichomycterus is represented by five endemic species (
Morphological studies of T. areolatus, which have focused mainly on northern and central Chilean populations, indicated intraspecific variation for some morphological characters. Variation has been documented in the bones of caudal skeleton (
Morphometric variation between populations can provide a basis for population differentiation, which is an important tool for evaluating population structure and identifying discrete groups (
In this study we determined the level of morphological variation in three T. areolatus populations collected from different river basins located in southern Chile. Given that these basins present particular hydrological conditions, we hypothesized that populations differ or segregate morphologically from each other, based on the river basin that they were sampled from. In order to address these objectives, we examined several specimens of each population based on 22 morphometric distance characters, using a truss network that covers the body shape dimensions in a relatively homogeneous way. This ensured a significant amount of information about the shape of individuals was gathered. This dataset was then subject to multivariate analyses to evaluate the degree of population separation, and to identify the body regions that experienced shape variations.
Specimens were collected from the Tijeral (TIJ) (n = 22) (40°37'S; 73°02'W) and Huilma (HUI) (n = 32) (40°43'S; 73°13'W) Rivers in the Bueno River basin, Province of Osorno, 10th Region; and from the Biobío River (BIO) (n = 50) (37°11'S; 72°47'W) in the Biobío River basin, Province of Biobío, 8th Region (Figure
Twenty-four morphometric characters were analyzed, including two conventional characters, total length and standard length, and 22 distance characters derived from a truss network constructed by interconnecting eleven landmarks representing the basic shape of the fish (Figure
Madj = M (LS ∙ LO-1)b
where Madj is the size adjusted measurement, M is the original measurement, LS is the overall mean of the standard length (SL) for all fish from all samples in each analysis and LO = is the SL of the fish. Parameter b was estimated for each character from the observed data as the slope of the regression of log M on log LO.
Position of the anatomical landmarks used to measure the size of 22 morphological characters on Trichomycterus areolatus based on a truss network. Definition of each character and its classification according to body shape dimension covered by them was as follows: a Head length, 1–2 = ventral tip of the operculum to tip of the head, 2–4 = tip of the head to posterior margin of the head b Head depth, 1–4 = ventral tip of the operculum to posterior margin of the head, 3–4 = base of the pectoral fin to posterior margin of the head c Anterior body length, 1–3 = ventral tip of the operculum to base of the pectoral fin, 3–5 = prepelvic length; 4–5 = posterior margin of the head to base of the pelvic fin; 3–6 = base of the pectoral fin to anterior base of the dorsal fin; 4–6 = predorsal length d Middle body depth, 5–6 = base of the pelvic fin to anterior base of the dorsal fin, 5–8 = base of the pelvic fin to posterior base of the dorsal fin, 6–7 = anterior base of the dorsal fin to anterior base of the anal fin, 7–8 = anterior base of the anal fin to posterior base of the dorsal fin e Middle body length, 5–7 = base of the pelvic fin to anterior base of the anal fin, 6–8 = dorsal fin base length, 7–9 = anal fin base length f Posterior body length, 7–10 = anterior base of the anal fin to dorsal posterior margin of the caudal peduncle, 9–10 = posterior base of the anal fin to dorsal posterior margin of the caudal peduncle g Peduncle depth, 8–9 = anterior caudal peduncle depth, 10–11 = posterior caudal peduncle depth h Peduncle length, 8–10 = dorsal caudal peduncle length, 9–11 = ventral caudal peduncle length.
Prior to statistical analysis the variables were analysed for conformance to assumptions regarding normal distribution and homogeneity of variance using the Kolmogorov–Smirnov (K-S) and Levene´s tests, respectively. Differences among populations were tested with an analysis of variance (ANOVA), either with parametric (one-way ANOVA) or non-parametric (Brown–Forsythe test) tests, using each character as a response variable. The Brown–Forsythe test was applied, given that some variables presented a heterogeneous variance among groups. The transformed data were subjected to principal component analysis (PCA) and discriminant function analysis (DFA) to evaluate any phenotypic differences among populations. Individual scores from PCA were used to construct a scatterplot to reveal the specimen groupings. The eigenvectors and eigenvalues were obtained from the PCA correlation matrix, which allowed the largest part of the variance of original variables to be reduced to a small number of components. The analysis evaluates the relationships among populations according to their proximity in the space defined by the components. Thus, plotting the component scores of specimens can reveal natural groupings, without a priori knowledge of such groupings. The significance of the separation among groups was determined using an analysis of similarity (ANOSIM) test. This test is a generalization of the univariate ANOVA and it has the property to consider all variables during the calculation of similarity among populations based on the Euclidean distance matrix. In this test, r-values range from 0 to 1, where 0 indicates no separation of groups and 1 corresponds to complete discrimination between groups (Clark 1993). Only 22 truss measurements were included in the PCA analyses. The number of principal components useful for this analysis was determined by using the Parallel Analysis (PA) based on the retaining of PCA eigenvalues from the data greater than PA eigenvalues from the corresponding random data (
Table
Morphometric data for 24 characters of three Trichomycterus areolatus populations from southern Chile.
Character | Tijeral River (Mean ± SD) (n = 22) |
Huilma River (Mean ± SD) (n = 32) |
Biobío River (Mean ± SD) (n = 50) |
ANOVA (Exact p-value) |
---|---|---|---|---|
Total length, TL (cm) | 7.02 ± 1.64 | 6.66 ± 0.92 |
6.16 ± 1.24 | 0.021* |
Standard length, SL (cm) | 6.27 ± 1.53 | 5.66 ± 0.75 | 5.51 ± 1.17 | 0.036* |
Truss measurements (cm) | ||||
1–2 | 1.04 ± 0.29 | 0.96 ± 0.12 | 0.58 ± 0.10 | <0.001***(§) |
1–3 | 0.25 ± 0.12 | 0.17 ± 0.08 | 0.46 ± 0.08 | <0.001*** |
1–4 | 0.63 ± 0.16 | 0.59 ± 0.11 | 0.49 ± 0.08 | <0.001***(§) |
2–4 | 0.80 ± 0.19 | 0.74 ± 0.14 | 0.84 ± 0.14 | <0.001***(§) |
3–4 | 0.73 ± 0.18 | 0.63 ± 0.13 | 0.53 ± 0.09 | <0.001***(§) |
3–5 | 1.82 ± 0.71 | 1.80 ± 0.30 | 1.94 ± 0.39 | 0.024 *(§) |
4–5 | 2.51 ± 0.56 | 2.29 ± 0.35 | 2.20 ± 0.44 | 0.858 NS |
3–6 | 2.57 ± 0.69 | 2.32 ± 0.37 | 2.48 ± 0.53 | <0.001*** |
4–6 | 2.89 ± 0.70 | 2.60 ± 0.42 | 2.65 ± 0.54 | <0.001*** |
5–6 | 0.96 ± 0.28 | 0.77 ± 0.15 | 0.96 ± 0.26 | <0.001*** |
5–7 | 0.99 ± 0.39 | 0.74 ± 0.30 | 0.93 ± 0.23 | 0.004 **(§) |
6–7 | 0.94 ± 0.24 | 0.81 ± 0.16 | 0.77 ± 0.17 | 0.007 ** |
5–8 | 1.56 ± 0.48 | 1.48 ± 0.29 | 1.36 ± 0.29 | 0.053NS |
6–8 | 0.85 ± 0.27 | 0.89 ± 0.21 | 0.61 ± 0.13 | <0.001***(§) |
7–8 | 0.75 ± 0.26 | 0.69 ± 0.16 | 0.66 ± 0.12 | 0.929 NS(§) |
7–9 | 0.33 ± 0.15 | 0.36 ± 0.12 | 0.41 ± 0.10 | <0.001***(§) |
8–9 | 0.53 ± 0.19 | 0.42 ± 0.08 | 0.49 ± 0.10 | <0.001***(§) |
7–10 | 1.86 ± 0.53 | 1.63 ± 0.24 | 1.84 ± 0.42 | <0.001*** |
8–10 | 1.30 ± 0.39 | 1.0 ± 0.19 | 1.44 ± 0.32 | <0.001*** |
9–10 | 1.40 ± 0.38 | 1.15 ± 0.17 | 1.36 ± 0.29 | <0.001*** |
9–11 | 1.22 ± 0.36 | 0.98 ± 0.20 | 1.25 ± 0.28 | <0.001*** |
10–11 | 0.57 ± 0.16 | 0.48 ± 0.09 | 0.48 ± 0.10 | 0.040 * |
The PCA based on 22 truss measurements retained two components according to PA, explaining 56.16% of the total variance. The first (PC1) and second (PC2) principal components accounted for 34.73 and 21.44% of the total variance, respectively (Table
Component loadings of the first two principal components derived from PCA based on the correlation matrix of 22 truss measurements of Trichomycterus areolatus populations from southern Chile. Characters of greater contribution on each component are in bold.
Component | |||
---|---|---|---|
PC1 | PC2 | ||
Eigenvalue | 7.640 | 4.716 | |
Explained variance (%) | 34.728 | 21.438 | |
Cumulative variance (%) | 34.728 | 56.166 | |
Character | Body shape dimension | ||
1–2 | Head length | -0.177 | -0.341 |
1–3 | Anterior body length | 0.324 | 0.046 |
1–4 | Head depth | -0.148 | -0.282 |
2–4 | Head length | 0.209 | 0.134 |
3–4 | Head depth | -0.126 | -0.307 |
3–5 | Anterior body length | 0.250 | -0.130 |
4–5 | Anterior body length | -0.062 | 0.031 |
3–6 | Anterior body length | 0.253 | 0.039 |
4–6 | Anterior body length | 0.159 | -0.002 |
5–6 | Middle body depth | 0.237 | -0.083 |
5–7 | Middle body length | 0.222 | -0.170 |
6–7 | Middle body depth | 0.066 | -0.293 |
5–8 | Middle body depth | 0.060 | -0.382 |
6–8 | Middle body length | -0.109 | -0.381 |
7–8 | Middle body depth | 0.096 | -0.346 |
7–9 | Middle body length | 0.226 | -0.139 |
8–9 | Peduncle depth | 0.239 | -0.180 |
7–10 | Posterior body length | 0.259 | -0.036 |
8–10 | Peduncle length | 0.324 | 0.080 |
9–10 | Posterior body length | 0.314 | 0.036 |
9–11 | Peduncle length | 0.310 | 0.045 |
10–11 | Peduncle depth | 0.127 | -0.274 |
The DFA based on 17 truss measurements with highly significant (p < 0.01) differences among populations produced two discriminant functions. The first (DF1) and second (DF2) discriminant functions explained 95.7% and 4.3% of the total variance, respectively, together accounting for 100% of the morphological variation. This result was supported by the high canonical correlations among the discriminant functions and groups, which had values of 0.969 y 0.641 for the first and second functions, respectively. Furthermore, both DF1 and DF2 generated statistically significant differences among the groups (Wilks’s λ = 0.036, χ2 = 309.803, d.f. = 34, p < 0.01). This result indicated significant morphological differences among the three populations. The DF1 vs DF2 scatter-plot revealed a clear separation among the point clouds for the three populations (Figure
Discriminant function analysis showed 93.3% correct classification of individuals into their original populations, and the cross-validation test produced comparable results (92.3%) (Table
Structure matrix coefficients that show the intra-group correlations between each of the characters and the discriminant functions. Characters of greater contribution in each discriminant function are in bold.
Character | Body shape dimension | Function | |
---|---|---|---|
DF1 | DF2 | ||
1–2 | Head length | -0.428 | 0.353 |
1–3 | Anterior body length | 0.390 | 0.142 |
6–8 | Middle body length | -0.227 | -0.173 |
2–4 | Head length | 0.220 | -0.152 |
3–6 | Anterior body length | 0.190 | -0.071 |
1–4 | Head depth | -0.149 | 0.121 |
7–10 | Posterior body length | 0.139 | 0.102 |
4–6 | Anterior body length | 0.121 | 0.014 |
9–10 | Posterior body length | 0.234 | 0.378 |
8–9 | Peduncle depth | 0.116 | 0.376 |
8–10 | Peduncle length | 0.263 | 0.372 |
3–4 | Head depth | -0.183 | 0.356 |
5–7 | Middle body length | 0.069 | 0.332 |
9–11 | Peduncle length | 0.217 | 0.311 |
7–9 | Middle body length | 0.109 | -0.290 |
6–7 | Middle body depth | -0.054 | 0.285 |
5–6 | Middle body depth | 0.121 | 0.223 |
Percentage of Trichomycterus areolatus specimens from populations of southern Chile correctly classified into their original group and after cross-validation.
Group | Population | Tijeral River | Huilma River | Biobío River | Total |
---|---|---|---|---|---|
Original (%)† | Tijeral River | 81.8 | 18.2 | 0.0 | 100 |
Huilma River | 9.4 | 90.6 | 0.0 | 100 | |
Biobío River | 0.0 | 0.0 | 100 | 100 | |
Cross-validated (%)‡ | Tijeral River | 77.3 | 18.2 | 4.5 | 100 |
Huilma River | 6.3 | 90.6 | 3.1 | 100 | |
Biobío River | 0.0 | 0.0 | 100 | 100 |
The results of this study obtained from the truss-based morphometrics indicated that T. areolatus from southern Chile showed significant phenotypic heterogeneity among the populations. Both multivariate analyses using PCA and DFA suggested three distinct phenotypic populations of T. areolatus. The segregation among populations was confirmed by PCA and DFA scatter-plot based on scores for each sample that showed non-, or slight, overlapping clusters of points for each population. This result was also supported by the percentage of correctly-classified individuals, where 93.3% of individuals showed correctly classification into their respective groups by DFA, indicating low intermingling among the populations. Populations of the Tijeral plus Huilma Rivers versus the Biobío River showed non-overlapping, possibly due to the large distances among drainages. This was not the case in the morphological parameters of populations of the Tijeral River versus the Huilma River that showed some overlapping between populations, which may be attributed to the small geographic distances between them given that both belong to the same drainage in southern Chile (Bueno River basin) and probably share more similar environmental conditions. Our results are similar to those of
The PCA confirmed that the variation in morphological measurements of the study populations of T. areolatus involved several characters related to the head region, body depth, and caudal peduncle. For example PC1, which was the most important component contributing to separation between populations (34.72% of the total explained variance), presented twelve characters with strong loading on this component, associated with body length and depth. For their part, PC2 that was also an important component (21.44% of the total explained variance), showed characters mainly associated to body depth variation in the head region, including the caudal peduncle region. Thus, the morphological variation registered in T. areolatus populations by PCA revealed changes on several body regions and/or dimensions.
Understanding the origin of morphological differences between populations of T. areolatus is challenging. This is because fish morphology is a complex phenotype that is determined by genetics and environment factors, and the interaction between them (
The morphometric variations observed among different populations of T. areolatus in the present study might also be associated with phenotypic plasticity in response to the different environmental factors of various habitats. Included within these factors are the hydrological characteristics of rivers, given that available data reveal an important influence of some hydrological parameters on body shape variation in fishes. For example, several studies have shown that water velocity can modify body shape in various fish species, such as salmonids (
The truss-based morphometrics represent a system of morphometric measurements that enhance discrimination between group, based on a systematic detection of body shape differences in both diagonal and horizontal and vertical directions (
In conclusion, the findings of this study based on truss-based morphometrics indicated significant morphological variations among three T. areolatus populations from southern Chile. Thus, these results suggest an underlying population structure of the species.
The suggestions and constructive comments of all those who helped to improve the final version of this manuscript are gratefully acknowledged. The authors are very grateful to the people who helped during the field works. We also wish to thank to Dr. Juan Francisco Gavilán for providing collaboration in this study.