A total of 18 individuals of P. magdaleniatum, P. magdalenae and A. pardalis (6 for each species) were obtained in the local fish market of the Cauca riverbed in the rainy (April-June) and dry (December-January, July-August) seasons in 2013. The samples were preserved at -4 ºC and protected from light before use.

Three samples per season were analyzed in each species. Each individual was previously measured, weighed, and sliced; the edible muscle was cut into pieces of approximately 3x3 cm and exhaustively extracted with a methanol/dichloromethane mixture (1:1 v/v). The extracts were filtered and evaporated to dryness at a temperature below 40 ºC with reduced pressure and constant agitation, and the obtained dry extracts were kept refrigerated, protected from light, air, heat, and humidity until their use. Lipid content was gravimetrically obtained and classified by the categories used by some authors: very low fat (< 2% p/p), low fat (2–4% p/p), medium fat (4–8% p/p) and high fat (> 8%) (Ackman, 1990).

Fatty acid methyl esters were obtained by subjecting a quantity of extract to basic hydrolysis with NaOH 2N in methanol at 55 ºC for 20 minutes. Then, a solution of HCl methanol at 5% was added and heated in bain-marie at 55 °C for 20 minutes. The mixture was neutralized, the organic solvent evaporated, two dichloromethane/water (1:1 v/v) extractions were performed and the organic phase was taken to dryness in a rotary evaporator at a temperature below 40 °C, at reduced pressure, and constant agitation. Pyrrolidine derivatives were obtained adding 1.000 µL of the pyrrolidine and 100 µL glacial acetic acid at 100 µg of methyl esters and subjecting the mixture to boiling at 100 ºC in an open reflux for 90 minutes. The pyrrolidine derivative fraction was obtained using a water/dichloromethane mixture (1:1 v/v). The organic phase was dried and then concentrated. The reactions were monitored by thin layer chromatography using chromatographic sheets 60F₂₅₄ (Merck) as stationary phase, with hexane/ethyl acetate (3:1) as mobile phase and revealed with phosphomolybdic acid at 5% in ethanol with further heating at 100 °C. The Retention Factor (Rf) was with standard values for fatty acids, methyl esters and pyrrolidine derivatives.

Samples were injected into an Agilent 6890N gas chromatographer coupled to an Agilent 5973N mass spectrometer. For the GC-MS analysis of the methyl esters and pyrrolidine derivatives of fatty acids, an Agilent 19091J-413, HP-5 (0.32 mm x 30 m x 0.25 µm) column was used with a furnace program starting at 100 °C, 10 °C/min gradient until a peak temperature of 325 °C. The splitless mode at 220 °C, 0.38 psi pressure, 14.1 mL/min total flow, and 300 °C auxiliary detector temperature were used for the injection. Scan mode was used in the mass detector in a 30–800 uma mass interval. The injection volume was 5.0 µL, and the analysis time was 60 minutes.

The identification of the fatty acids present in the samples was made through the analysis of the mass spectra of the methyl esters and pyrrolidine derivatives, and compared with the reports in the following databases: NIST Mass Spectral Database (NIST98^®, NIST02^®, NIST5a^®), Methyl Esters of Fatty Acids – Archive of Mass Spectra and N-Acyl Pyrrolidine Derivatives of Fatty Acids – Archive of Mass Spectra (Christie, 2013). The percentage of abundance (%) of fatty acids was calculated from the area under the peak curve of the gas chromatogram of the methyl ester fractions.

Five indexes were calculated to determine the nutritional value of the three fish species, namely n-6/n-3 ratio (PUFA n-6/PUFA n-3), unsaturation index (UI = ∑UFA X number of double bonds/100), atherogenicity index (AI=[C12:0 + (4 X C14:0) + C16:0]/(MUFA + PUFA n-6 + PUFA n-3]) (Ulbricht and Southgate, 1991), thrombogenicity index (TI = (C14:0 + C16:0 + C18:0)/[(0.5 X MUFA) + (0.5 X PUFA n-6) + (3 X PUFA n-3) + (PUFA n-3/PUFA n-6)]) (Ulbricht and Southgate, 1991), and FA hypocholesterolemic/hypercholesterolemic ratio (h/H = (C18:1 n-9 + C18:2 n-6 + C20:4 n-6 + C18:3 n-3 + C20:5 n-3 + C22:5 n-3 + C22:6 n-3)/(C14:0 + C16:0) (Santos-Silva et al., 2002), where PUFA: Polyunsaturated fatty acid, MUFA: Monounsaturated fatty acid and UFA: Unsaturated fatty acid.

The effects of species (P. magdaleniatum, P. magdalenae and A. pardalis) and seasons (dry and rainy) on the global percentages of abundance of the fatty acids and lipids were assessed by multi-factor analysis of variance using a general linear model. The same procedure was used to evaluate the effect of the species, the season, and the kind of fatty acid on the abundance of each fatty acid. Furthermore, the nutritional values for each species between seasons were explored by two Principal Component Analyses using all the fatty acids found in the samples as variables and the five nutritional indexes were calculated. In the latter case, undetectable values for PUFA n-3 were established at 0.01%, based on the lowest level measured in this study. Statistical analyses were performed with Statgraphics Centurion XVI^® and Past (Hammer et al., 2001) programs.

Biometric indexes (BI = Weight in grams/Standard length in centimeters) were higher for P. magdaleniatum (both seasons) and P. magdalenae (rainy season) (Table 1). The lipid percentage was significantly lower in P. magdaleniatum compared to P. magdalenae and A. pardalis (Table 1, F_(2,17) = 6.07; p = 0.0127). In general, the muscle lipid percentages found in this study were similar to the ranges previously reported for P. magdaleniatum (0.4% -1.9%) and P. magdalenae (1.3% - 5.2%) (Perea et al., 2008) and other freshwater species (0.35% - 7.92%) (Rahman et al., 1995; Luzia et al., 2003; Rasoarahona et al., 2005; Özogul et al., 2007; Ramos-Filho et al., 2010). Comparing the rainy and dry seasons, P. magdaleniatum varied from low fat to very low fat, while P. magdalenae and A. pardalis exhibited medium fat for both seasons, according to the criteria described by Ackman (1990). However, ANOVA did not show significant differences between seasons for each species (F_{(1, 17)} = 0.14; p = 0.7172). Likewise, fatty acid percentages were similar among species (F_(2,16) = 3.65; p = 0.0554) and between seasons (F_(1,16) = 0.05; p = 0.8304).

The yields obtained in the preparation of the methyl esters derivatives were high (P. magdaleniatum: 82.0%; P. magdalenae: 90.4%; A. pardalis: 87.2%), and for the pyrrolidine derivatives they were 100% in all cases. Rf values for fatty acids (0.51), methyl esters (0.91) and pyrrolidine derivatives (0.20) were obtained using oleic acid and cholesterol patterns.

A total of 50 fatty acids were identified in P. magdaleniatum, 41 in P. magdalenae, and 32 in A. pardalis (Tables 2 and 3). The most abundant saturated fatty acids were isopalmitic acid (i-C15:0) in P. magdalenae (22.51% and 18.62%) and anteisopalmitic acid (ai-C15:0) in P. magdaleniatum (13.58% and 36.06%) and A. pardalis (10.24% and 10.45%), which were also identified in other freshwater species but in lower quantities (Rasoarahona et al., 2005), followed in abundance by C16:0 and C18:0 acids in the three species. Oleic acid (C18:1 n-9) was the most abundant monounsaturated acid in P. magdaleniatum (22.41% and 10.09%), P. magdalenae (15.82% and 11.16%), and in A. pardalis (11.37% and 6.45%), which concured with other reports on freshwater species (Özogul et al., 2007; Jabeen and Chaudhry, 2011). Linolenic acid (18:3 n-3) was not detected in P. magdaleniatum or P. magdalenae, which is in agreement with a previous report on P. magdaleniatum (Perea et al., 2008). Polyunsaturated fatty acids n-3 were not detected in A. pardalis, which may reflect preferences in the diet, mainly composed of other fish, occasionally crustacean, insects and other trophic groups (Tobías-Arias et al., 2006).

Multi-factor ANOVA showed that the fatty acids (F_{(68, 413)} = 11.08; p = 0.0000) and their interactions with the species (F_{(136, 413)} = 1.89; p = 0.0001) were significant, indicating that the abundance of fatty acids changed in the seasons, depending on the fatty acid. Moreover, such changes were different among the species. Specifically, in the dry season, P. magdaleniatum displayed a significant reduction in C14:0, C22:0 and C16:1 n-9 acids (p < 0.05; Tables 2 and 3). Likewise, in the dry season, A. pardalis showed a significant reduction in C16:0, i-C17:0, C20:0, i-C16:1 n-7 and C18:1 n-7 isomer acids and a significant increase in C18:0 (Tables 2 and 3). In contrast, P. magdalenae exhibited non-significant changes between seasons (Tables 2 and 3).

The n-6/n-3 ratios in P. magdalenae and P. magdaleniatum changed between seasons (Table 4). The best value for the n-6/n-3 fatty acid ratio (1.04) was found for P. magdalenae caught in the dry season when this species performs reproductive migration from the swamps to the river (Mojica et al., 2002). However, the n-6/n-3 ratio was superior in the rainy season when P. magdalenae returns from the river to the swamps after fulfilling its breeding cycle. In contrast, in P. magdaleniatum the best n-6/n-3 fatty acid ratio (2.72) was observed in the rainy season, which matches its breeding season (Mojica et al., 2002). In the dry season this ratio was greater than the desirable ratio for the human diet.

The n-6/n-3 ratios present in P. magdaleniatum and P. magdalenae are strongly associated with breeding times and confirm the inverted bio-ecological habits of P. magdaleniatum with respect to P. magdalenae, in the two seasons assessed (Mojica et al., 2002). This antagonistic behavior of the two species is relevant to their nutritional value as the determination of the presence of these fatty acid ratios could be considered a useful parameter to study migratory habits and breeding cycles of freshwater fish when determining the most adequate times to capture them for consumption purposes. Similar changes in n-6/n-3 ratios between seasons resulting from spawning has been described for other species (Rasoarahona, 2005; Guler et al., 2007; Inhamuns and Franco; 2008).

Since the n-6/n-3 fatty acid ratios of P. magdalenae in the dry season and P. magdaleniatum in the rainy season are included in the range considered healthy (1–4; Simopoulos, 2002), both species would represent a healthy source of PUFAs in a particular season.

The values for UI (between 0.37 and 0.63) showed a low presence of unsaturated fatty acids for the three species (Table 4). Nevertheless, low AI values obtained in the two seasons from the three studied species (0.04 – 0.70) indicated a low risk of atherogenicity. Further, the TI values (1.06 – 1.07) for P. magdalenae and A. pardalis in the dry and rainy seasons, respectively, were slightly higher than 1, but they might be considered acceptable when compared to other healthy foods (Ulbricht and Southgate, 1991). On the other hand, although the h/H values for P. magdalenae and A. pardalis in the dry and rainy seasons, respectively, were slightly lower than 1, they might be considered moderately high (0.62 – 0.80), which indicates low hypercholesterolemia risk owing to their consumption. According to these results, moderate consumption (1-2 times a week) of the three species, in any of the two hydrologic seasons, represents a low risk for cardiovascular illnesses.

Principal component analyses using all the fatty acids identified in the three species as variables showed that the first component drastically separated P. magdaleniatum in the dry season from the other species, wherein this species was characterized by the absence of n-3 PUFAs. On the other hand, the second component separated contrasting seasons for the three species (Figure 1a). Similarly, in the principal component analysis using the five nutritional indexes, the first component that explained more than 99% of variability, drastically separated P. magdaleniatum in the dry season from the other species (Figure 1b). This latter analysis also suggested that the n-6/n-3 ratio and h/H index determined the opposite behavior in P. magdaleniatum in the dry season.

The nutritional values of the wild species analyzed in this study (AI: 0.04-0.70; TI: 0.66-1.07; h/H: 0.62-31.94) differed from those found in a previous work (AI: 1.1-2.89; h/H: 0.25-1.29; Perea et al., 2008), although they were similar to the values found for freshwater fish from the Brazilian Pantanal (AI: 0.51-0.79; TI: 0.68-1.18; h/H: 1.14-1.80; Ramos-Filho et al., 2010).