Quantification and classification of fatty acids in marine fish oil from Southeast Sulawesi using gas chromatography and chemometrics

Materials

Twenty species of marine fish were collected from different locations around the Southeast Sulawesi coast, Indonesia (Fig. 1). Each location consists of 5 kg/species with a weight range of 300–500 g to minimize variation possibilities and preserve the homogeneity of fish samples. All samples were collected fresh and caught within a period of 0–24 hours. The fish samples were eviscerated, filleted, packed flake ice in a cooling box, and sent to the Farmacy Laboratory of Halu Oleo University. The filleted fish samples were dried at a temperature of 80°C–90°C (2 × 24 hours) using an oven (Stuart Scientific) according to the methods described by Alinafiah et al. [17] with slight modification, powdered (Philips blender), and then stored at 4°C for future analysis. n-Hexane, heptane, ethanol, and all reagents used for analysis were of analytical grade and were supplied by E. Merk Darmstadt, Germany.

Figure 1. Location of sample collection. [Click here to view]

Fish oils extraction

Marine fish oils were extracted using Soxhlet methods according to the methods described by Rincón-Cervera et al. [6] with slight modifications. Powdered fish (50 g) were placed in a Soxhlet apparatus and extracted with 750 ml n-hexane for 2 hours at a temperature of 80°C. After extraction was complete, n-hexane was evaporated at 50°C using a rotary evaporator (Stuart). The oils were collected, weighed, and stored at 4°C.

Oils purification

Crude fish oils were purified using bentonite at a ratio of 1% (w/w) of oils weight, centrifuged (Boeco Germany) for 5 minutes, filtered, and stored in a dark bottle in a freezer for subsequent analysis [20].

FAs analysis

The FA analysis of marine fish oils was carried out according to Irnawati et al. [21]. FAs were derived from FAs methyl esters (FAMEs). FAMEs of each oil sample (0.4 g) were prepared using an alkali catalyzed (1.5 ml of methanolic potassium hydroxide) and in a crew-capped glass test tube. The mixtures were heated at 60°C for 10 minutes, cooled at room temperature, and then added 2 ml of boron trifluoride-methanol. The sample mixtures were heated again under the same conditions; when cooled, 1 ml of heptane was added with 1 ml of sodium chloride and shacked. Finally, the upper heptane layer was collected for FA analysis using an Agilent GC (7890B) equipped flame ionization detector and an HP-88 capillary column (100 m × 0.3 mm × 0.2 μm). The temperature program was 5 minutes at 100°C, heating until 240°C at a rate of 4°C/minute, holt time at 15 minutes. The temperature of the injector was 260°C with a split ratio of 10:1, and injection volume was 1 μl. The detector temperature was 260°C. Helium was used as carrier gas (1.8 ml/minute). Peak identification was carried out by comparing the retention time from CRM47885 FAME Mix standards (Supelco) (Fig. 2).

Figure 2. Chromatogram of FAMEs mix standards (CRM47885) using a GC (Agilent, 7890B) with flame ionization detector. [Click here to view]

Classification of marine fish oil

Pattern recognition analysis unsupervised techniques, namely principal component analysis (PCA) and supervised techniques, namely partial least-square-discriminant analysis (PLS-DA), sparse variant PLS-DA (sPLS-DA), and orthogonal PLS-DA (OPLS-DA) were used in this study. The concentration of FA compositions was used as a variable during the classification of marine fish oil from different origins.

FAs analysis of marine fish oil

In this study, FAs from marine fish oils extracted from Epinephelus sp. and Lutjanus sp. were measured using GC-FID. The sampling of marine fish was done from several locations in South East Sulawesi province as illustrated in Figure 1. Different locations affected the compositions of FAs in each type of fish oil due to the difference in environment which is associated with the difference in nutrition sources. Table 1 shows the result of FAs analysis using the GC-FID technique from Epinephelus sp. and Table 2 from Lutjanus sp. fish oils. Total saturated FAs (SFA), monounsaturated (MUFA), and polyunsaturated (PUFA) percentages for both ranged from 45.59% to 61.42%, 20.32% to 39.30%, and 12.97% to 24.56%, respectively. The important FAs identified in most samples were myristic/C14:0 (1.89%–8.49%), palmitate/C16:0 (24.30%–33.68%), heptadecanoic/C17:0 (1.49%–2.87%), stearate/C18:0 (5.90%–13.38%), tricosanoate/C23:0 (0.12%–8.66%), lignocerate/C24:0 (1.80%–4.99%), palmitoleic/C16:1 (4.05%–8.11%), oleate/C18:1c (9.76%–22.58%), nervonate/C24:1 (0.13%–5.77%), linolelaidate/C18:2t (0.11%–5.67%), linoleate/C18:2c (0.19%–5.36%), and cis-4,7,10,13,16,19-docosahexaenoate, DHA/C22:6 (4.31%–21.84%).

Table 1. FA composition measured using a GC-FID from Epinephelus sp. fish oils. [Click here to view]

Table 2. FA composition measured using a GC-FID from Lutjanus sp. fish oils. [Click here to view]

Palmitic acid (C16:0) and stearic acid (C18:0) were the dominant SFA, contributing 44.46%–54.60% and 12.94%–23.11% of the total SFA of lipids for all samples, respectively. Palmitoleic acid (C16:1) and oleic acid (C18:1c) were found to be dominant in MUFA, accounting for 19.22%–26.11% and 46.32%–57.41% of total MUFA of lipids, respectively. Similarly, several researchers reported that palmitic acid (C16:0) is a dominant SFA and palmitoleic acid (C16:1) and oleic acid (C18:1c) are major MUFA [7,22]. While the most represented PUFA were linolelaidate (C18:2t), linoleate (C18:2c), and 4,7,10,13,16,19-docosahexaenoate, DHA (C22:6), accounting for 0.96%–34.22%, 0.84%–30.75%, and 26.72%–72.32% of total PUFA of lipids, respectively.

Classification of marine fish oil

PCA using variables of FA composition was performed to differentiate fish oils from Epinephelus sp. and Lutjanus sp. Figure 3 indicates the score plot between two fish oils of Epinephelus sp. and Lutjanus sp. with principal component 1 (PC1) of 50.1% and PC2 of 19.8%. The samples were almost overlapped each other showing a high similarity of their variables (FA composition) between the two groups observed through PCA. The reduction of original variables into several PCs was done in PCA to reduce the variables to be more effective for sample differentiation [23]. However, because PCA is an unsupervised technique that is usually used for initial sample grouping, it is required to apply supervised pattern recognition techniques such as PLS-DA to obtain better discrimination between two groups of fish oil samples. Supervised pattern recognition techniques such as PLS-DA are often used to evaluate the PCA result due to their capability for better classification between samples. Figure 4A shows that Epinephelus sp. fish oil could be better discriminated from Lutjanus sp. fish oil using PLS-DA compared to the PCA result using FA composition as the variables with PC1 of 31.8% and PC2 of 30.5%. It indicated that the latent variables used in creating the PLS-DA model could maximize the variation to obtain better discrimination between two fish oils. Epinephelus sp. fish oil samples from different locations appeared in a tight cluster compared to Lutjanus sp. fish oil samples. It showed more variability of FA compositions within Lutjanus sp. fish oils. Using PLS-DA, only several samples of Lutjanus sp. fish oils overlapped with samples from Epinephelus sp. fish oils, indicating the good performance of PLS-DA to discriminate between group samples. Through variable importance in projection (VIP) analysis in the PLS-DA model, variables of FAs that have a high contribution in discriminating fish oils from Epinephelus sp. and Lutjanus sp. could be identified as shown in Figure 4B. The FAs of tricosanoate (C23:0), trans-9-elaidate (C18:1t), linolelaidate (C18:2t), cis-11,14,17-eicosatrienoate (C20:3c), palmitate (C16:0), stearate (C18:0), cis-9-oleate (C18:1c), and myristate (C14:0) had important roles in discriminating Epinephelus sp. fish oils and Lutjanus sp. fish oils because they had VIP value larger than 1.0 [24,25]. FAs of tricosanoate (C23:0), linolelaidate (C18:2t), palmitate (C16:0), and stearate (C18:0), were significantly high in Epinephelus sp. fish oils whereas FAs of trans-9-elaidate (C18:1t), cis-11,14,17-eicosatrienoate (C20:3c), cis-9-oleate (C18:1c), and myristate (C14:0) were found to be in high levels in Lujtanus sp. fish oils.

Figure 3. Score plot of PCA. [Click here to view]

The extension of PLS-DA analysis such as sPLS-DA and OPLS-DA was tried to classify Epinephelus sp. fish oils and Lutjanus sp. fish oils. Both sPLS-DA (Fig. 5A) and OPLS-DA (Fig. 5B) demonstrated better results for discriminating Epinephelus sp. fish oils and Lutjanus sp. fish oils. Better cluster separation were observed both in sPLS-DA and OPLS-DA between two groups of fish oils indicating better discrimination between the two groups. In sPLS-DA, the step of variable selection and classification was performed in one step simultaneously, meanwhile in OPLS-DA the variables were subjected to an orthogonal process. As a consequence, they often resulted in clearer discrimination and classification [26]. The variables which significantly contributed to sample discrimination in sPLS-DA were investigated through the first loading of sPLS-DA (Fig. 5C). The FAs of trans-9-elaidate (C18:1t), cis-11,14,17-eicosatrienoate (C20:3c), cis-11-eicosenoate (C20:1c), pentadecanoate (C15:0), and myristate (C14:0) were found to be high in Lutjanus sp. fish oils. Meanwhile, FAs of linolelaidate (C18:2t), tricosanoate (C23:0), heptadecanoate (C17:0), stearate (C18:0), and lignocerate (C24:0) were high in Epinephelus sp. fish oils.

Figure 4. Score plot of (A) PLSDA and (B) analysis of VIPs value. [Click here to view]

Figure 5. Score plot of (A) sPLSDA, (B) score plot of OPLS-DA, and (C) identification of important variables in sPLS-DA through loadings 1. [Click here to view]

Figure 6. Heatmap analysis. [Click here to view]

Heatmap analysis (Fig. 6) for fish oil samples using all variables used could be used to observe the distribution of FAs in all fish oil samples extracted from various Epinephelus sp. and Lutjanus sp. A high amount of nervonate (C24:1c) and linolenate (C18:3c) was found only in Epinephelus merra from Bau-Bau. cis-5,8,11,14-eicosatetraenoate (C20:4c) was extremely high in E. merra from Buton. Meanwhile, laurate (C12:0), myristoleic (C14:1c), and linoleate (C18:2c) were significantly high only in Epinephelus coioides from Muna, Lutjanus malabaricus from Bau-Bau, and L. malabaricus from Kolaka, respectively. Heatmap analysis is very useful to identify the distribution pattern of compounds in certain samples which also could be used to monitor selected compounds that become targets for such authentication and quality control purposes.