Effect of Krill, Mussel and Fish Meals on Fatty Acid Profile, Carotenoid Content, Colour and Oxidation Properties of White Muscle in Arctic Charr (Salvelinus Alpinus L.)

Fish feeding and sampling

Arctic charr with initial weight 104.5±20.8 g were fed the experimental diets for 15weeks at Kälarne Aquaculture Centre, Sweden. The 12 experimental diets were fed to fish in 12 different tanks (1 m3), each containing 12 fish. Another three tanks were assigned to the control (standard reference) diet. The temperature of the water was 5-17°C during the rearing period, following the natural cycle. Feeding was stopped three days before sampling. Initial length and weight and final length and weight were recorded. Condition factor (CF) and daily growth coefficient (DGC) were calculated according to: CF=Weight (g)*100*Length-3 (cm) and

Before sampling, the fish were anesthetised by tricane methane sulphonate 222 (30 ml/l) and killed by rupture of the vertebral column. White muscle in dorsal muscle mass was dissected; the left part was used for lipid and carotenoid analysis and right part for colour analysis. The fillet was immediately packed in aluminium foil and frozen on dry ice. All samples were stored at -80°C until further analysis.

Fatty acid analysis

Lipids from white muscle and the diets were extracted following the method of Trattneret et al. [14] and methylated with boron trifluoride according to Appelqvist [15]. Thereafter, the methylated FA were analysed in a Varian CP-3800 gas chromatograph system (Agilent Technologies, Santa Clara, CA, USA) equipped with a flame ionisation detector, fused silica capillary column BPX 70, 50 m length × 0.22 mm I.D. × 0.25 µm film thickness (SGE, Austin, USA) and auto-sampler in split mode (Combi PAL AutoSampler, Varian AB, Stockholm, Sweden). Helium was used as the carrier gas (0.8 mL/min) and nitrogen as the make-up gas. The column temperature programme followed the method used by Trattner, et al. [14]. FA were identified by comparison retention time with the standard FA mixture GLC-461 (Nu-check-Prep, Inc., Elysian, Minnesota, USA) and retention time. Star Chromatography Workstation software version 5.5 (Varian AB, Sweden) was used for peak area integration. As internal standard, methyl 15-methylheptadecanoate (Larodan Fine Chemicals AB, Malmö, Sweden) was added to the methylated samples.

Total carotenoid analysis

The lipid extracts were used for carotenoid analysis. Total carotenoid content (TC) of muscle and diet was analysed by the spectrophotometric method described by Tolasa, et al. [16]. Absorbance at 480 nm was measured by spectrophotometer (UV-2401 (PC) CE, Shimadzu). The concentration of Carotenoid was calculated according to the formula: Concentration (mg/l) = Absorbance*10000/2100, Where 2100 is E (1%, 1 cm) = standard absorbance of AST solution with 1% (w/v) in spectrobath with1 cm in hexane at 480 nm and 10000 is the scale factor. Carotenoid content was expressed as mg/kg muscle (or diet). Average retention rate of total Carotenoid (RTC) was calculated as [17]: RTC (%) = 100 * 0.61 * (CF*WF-CI*WI) / (105*20*CD/12), where CF, CI is the Carotenoid content in white muscle (mg/kg muscle) in final fish, initial fish; WF, WI is the weight (g) of final fish, initial fish; CD is the carotenoid content in the diet (mg/kg); 0.61 is the average muscle fraction in whole fish (61%); 105 is feeding days; 20 is weight of diet fed to each tank per day; and 12 is number of fish per tank.

Specific Carotenoid analysis

The content of Carotenoid in white muscle was analysed using a high performance liquid chromatography (HPLC) (Bischoff Analysentechnikund Geräte GmbH, Leonberg, Germany) equipped with an Agilent 1100 series diode array detector (Agilent Technologies, Waldronn, Germany). Separation was performed under the following operating conditions: normal phase column, Alltech SI 5µ silica column, 4.6 x 250 mm (Alltech Associates Inc., Deerfield, IL); column temperature, 40°C; mobile phase, hexane: isopropanol (94:6, v:v); isocratic elution, flow rate 1.4 ml/min; injection volume, 10µl; detection wavelength, 480nm. Peaks were manually integrated using the software HP ChemStation Version 05.01. Standards including AST (Dr. Ehrenstorfer, Germany) and canthaxanthin (Fluka-Sigma, St. Louis, MO, USA) were used to identify peaks in the samples. Carotenoid content was quantified by standard curves run with different concentrations and expressed as mg/kg muscle.

Colour property analysis

White muscle was thawed on an orange board (similar colour to white muscle) at room temperature for 1h until its central temperature reached 15°C and it was soft enough for measurement by Minolta Chroma Meter CR-300 (Minolta, Osaka, Japan). The juice around the white muscle after thawing (not much juice loss was observed) was removed. The Chroma Meter was equipped with an 8 mm diameter aperture and calibrated on a white reference plate before use. Measurements were made at four points on the fillet. All measurements were performed in the colorimetric space L*, a*, b* according to the Commission Internationale de Éclairagesystem [18]. They were then transformed in the L*, C*, H(°)a*b*colorimetric space [19], in which the three-dimensional characteristics of colour appearance are the lightness attribute L*, chromatic attributes hue H(°), a*b* (arctan b*/a*) and chroma C*[(a*2+b*2)1/2].

Thiobarbituric reactive substances analysis

TBARS value was determined by a forced oxidation method, modified from Sigurgisladottir, et al. [20]. Water was added (0.5 ml) to fish white muscle (1 g) and it was heated at 90 ºC for 5 min to cause oxidation. The heated samples were cooled to room temperature and then 10% trichloroacetic acid was added and the samples were homogenised to extract malondialdehyde. The lysate was filtered and incubated with thiobarbituric acid in a water bath at 90 ºC for 45 minutes. Thereafter the samples were cooled to room temperature, centrifuged at 500 rpm for 10min and the absorbance was measured at 532 nm. TBARS values were expressed as nmol/kg muscle.

Data analysis

To study the effects of various components in the diets, three comparisons with different groups were made for each parameter. Comparison 1 involved groups R1, R4, M1, M2, K1 and K4 (to study the effect of lipid source); comparison 2 involved groupsR1, R2, R3, R4 and R5 (to study the effect of the protein source); and comparison 3 involved groups K1, K2, K3 and K4 (to study the effect of protein source). Data on FA, colour properties and oxidation were first analysed by partial least squares-discriminant analysis (PLS-DA) in SIMCA-P+13.0 (Umetrics, Umeå, Sweden). Comparisons of parameters between specific groups were performed by general linear model (GLM) analysis in SAS 9.3 (SAS Institute Inc., Cary, NC) with data subjected to arcsine square root transformation or logarithmic transformation when necessary. Comparison 1 was performed with a Bonferroni test, comparison 2 with a Scheffe test and comparison 3 with a Duncan test. Significance level was set to P<0.05 unless otherwise stated. Data presented are mean±SD. We also compared the data for each group with those for the control group. No significant difference was found.

Fish growth

No mortality occurred during rearing of the fish. Fish length increased by about 7.0 cm and weight by about 170 g, giving a CF value of approximately 1.4 after 15 weeks of feeding in the experiment (Table 3). The highest DGC (1.94) was found in group M1 and the lowest in group R1 (1.31). No pattern was found in any of the three comparisons made with fish growth parameters in PLS-DA analysis (data not shown). Statistical analysis suggested that there were no significant differences in fish growth parameters between any of the groups.

Abbreviations: RO, groups fed diets used rapeseed oil as single lipid source; FKO, groups fed diets used fish, krill and linseed oil as lipid source; MO, groups fed diets used rapeseed, linseed, fish and krill oil as lipid source. Lipid content and fatty acids In comparison 1 (see ‘Data analysis’ section), there was a clear discrimination between groups R1, M1, K1 and groups R4, M2, K4 (Figure 1a). The RO groups (R1, R4) were close to the MUFA (monounsaturated fatty acids) 18:3n-3 (α-linolenic acid, ALA), 18:2n-6 (Linoleic acid, LA), 18:3n-6, 20:3n-6, n-6 polyunsaturated fatty acid (n-6 PUFA); the MO groups (M1 and M2) were close to lipid content (LC), n-3 PUFA and DHA; and the FKO groups (K1 and K4) were close to 16:0, 14:0, 18:0, 16:1n-7, SFA (saturated fatty acids), EPA, 22:5n-3 (docosapentaenoic acid, DPA) and n-3/n-6 (Figure 1a).

Click to enlarge

No differences in LC were found, but many FA differed. Levels of 14:0, 16:0 and SFA were higher in FKO groups (K1and K4) than in RO groups (R1 and R4) and MO groups (M1 and M2) (Table 4). No differences in total MUFA were found, but the levels of 16:1n-7, 18:1n-7 and 22:1n-11 was higher in FKO groups than in MO and RO groups. The levels of 20:3n-6 and total n-6 PUFA were higher in group R1 than in groups M1 and K1. For n-3 FA, differences were found between RO groups and FKO groups. The proportions of 18:4n-3 and DPA were higher but the ALA level was lower in FKO group than in RO groups. EPA in group K1 showed a higher level than in group R1 and there were no differences in DHA and n-3 PUFA in the six groups. However, n-3/n-6 of FKO groups was higher than in the RO and MO groups. There were also some differences in 18:0, SFA, 18:1n-7, 22:1n-11, 20:3n-6, EPA and DPA between group M1 and group M2.

In comparison 2, discrimination was observed in the FA profile of four groups (Figure 1b). 20:3n-6, LA, n-6 PUFA, 18:1n-9 and MUFA were close to group R5, while 14:0, 16:0, SFA, 18:1n-7, EPA, DPA, n-3 LCPUFA and n- 3/n-6 were close to group R2. Group R1 was near ALA, 18:3n-6 and 20:1n- 9. Group R1 had similar levels of 14:0, 16:0 and SFA to groups R3 and R4 but lower levels than group R2. The levels of 20:1n-9 and 22:1n-11 in group R1were higher than in groups R2, R3 and R5. No difference in n-6 FA was observed between groups R1 and R3, R5. Nevertheless, most n-6 FA showed a higher level in group R1 than in group R2. 18:4n-3 proportion in group R1 was higher than in the other groups and the levels of EPA and DPA were lower than in group R2. In comparisons between groups R5, R2 and R3, no differences in n-3 FA were found, but levels of most n-6 FA in group 5 were higher than in group R2. No differences were found in LC.

Click to enlarge

except for 20:4n-6 (arachidonic acid, AA), which was most abundant in group K3. For n-3 FA, the levels of ALA and 18:4n-3 were higher in group K1 than in groups K2 and K3, while the levels of EPA and DHA were lower. Group K1 showed lower total n-3 PUFA level and n-3/n-6 ratio compared with groups K2 and K3. Group K4 had some differences to the other three groups in terms of FA. Its level of most FA showed intermediate values between group K2 and group K3.

Click to enlarge

Colour properties and carotenoid

In comparison 1, clear clusters were found between groups R1, M1 and K1 and between groups K4, R4 and M2, with the H value separated from a*, b*, C*, AST content (AC) and TC (Figure 2a). The Bonferroni test showed that group R1 had lower values of a*, b*, C*, TC and a higher value of H than groups M1 and K1 (Table 5). Similar results were found between groups R4, M2 and K4 (some non-significant differences). A higher RTC in group R1 than in group M1 was also observed. There were no differences between groups M1 and K1 or between groups M2 and K4. Group R4 exhibited a lower a* value and higher H value than group K4. No differences were found between groups M1 and M2. In comparison 2, group R1 was separated from the other RO groups. Group R4 was close to a* and AC, while TC differed from groups R3 and R5; no distinct separation was seen in groups R2, R3 and R5 (Figure 2b). Group R1 showed many differences from other groups, with a lower a* value than groups R2 and R3, lower b*, C* and TC value than groups R2 and R5 and higher H value than group R2. Lower a* and TC value and higher H value were observed in group R5 compared with group R2. Group R4 exhibited similar levels of these parameters to group R2. In comparison 3, group K2 was close to C*, TC and AC and different from groups K1, K3 and K4 (Figure 2c). Group K2 showed higher levels of a*, b* and C* than groups K1 and K3 and its TC level was also higher than that of group K1. Group K3 exhibited similar levels of all parameters to group K1. AST is the main carotenoid found in the fish white muscle. A low level of canthaxanthin was detected only in group K3. No significant difference in AST level was found. Most groups showed a low AC (<3.5 mg/kg) and low RTC (1.32- 5.08%).

Click to enlarge

a

Click to enlarge

b

Click to enlarge

Figure (2a, 2b, 2c): PLS-DA analysis of colour properties (a*, b*, L*, C*, H value) and carotenoid in white muscle of Arctic charr (n=6): a, b, c are biplots of data from groups R1, R4, M1, M2, K1, K4, groups R1, R2, R3, R4, R5 and groups K1, K2, K3, K4, explaining 81.1%, 78.5% and 69.0% of variation in the samples, respectively. AC, astaxanthin content; TC, total carotenoid content; RTC, retention rate of total carotenoid.

Comparison Group L* a* b* C* H CAN AST TC RTC TBARS

Comparison 1

Comparison 2

Comparison 3

Standard 45.3±3.2 7.5±2.1 14.3±2.0 16.2±2.7 62.8±4. n.d 2.42±0.85 2.99±0.84 3.28 15.45±5.22 $$
R 1 \quad 5 3. 6 ^ {\mathrm {a}} \pm 4. 0 \quad 0. 3 ^ {\mathrm {c}} \pm

Standard 45.3±3.2 7.5±2.1 14.3±2.0 16.2±2.7 62.8±4. n.d 2.42±0.85 2.99±0.84 3.28 15.45±5.22 $$ R 1 \quad 5 3. 6 ^ {\mathrm {a}} \pm 4. 0 \quad 0. 3 ^ {\mathrm {c}} \pm 1. 1 \quad 8. 4 ^ {\mathrm {b}} \pm 0. 6 \quad 8. 4 ^ {\mathrm {b}} \pm 0. 5 \quad 8 8. 0 ^ {\mathrm {a}} \pm 8. 2 \quad \mathrm {n . d .} \quad 0. 3 1 \pm 0. 1 2 \quad 0. 7 7 ^ {\mathrm {b}} \pm 0. 2 1 \quad 5. 9 8 \quad 1 2. 8 6 ^ {\mathrm {b c}} \pm 3. 3 9 $$ $$ R 4 \quad 4 7. 7 ^ {a b} \pm 3. 3 \quad 4. 0 ^ {b} \pm 1. 6 \quad 1 3. 1 ^ {a} \pm 2. 5 \quad 1 3. 7 ^ {a} \pm 2. 8 \quad 7 3. 6 ^ {b} \pm 3. 8 \quad n. d. \quad 1. 0 6 \pm 0. 2 5 \quad 2. 1 3 ^ {a} \pm 0. 4 6 \quad 2. 5 9 \quad 1 1. 6 1 ^ {c} \pm 1. 7 4 $$ $$ M 1 \quad 4 8. 2 ^ {a b} \pm 4. 0 \quad 6. 4 ^ {a b} \pm 2. 9 \quad 1 4. 5 ^ {a} \pm 3. 7 \quad 1 5. 9 ^ {a} \pm 4. 4 \quad 6 7. 4 ^ {b c} \pm 7. 8 \quad \mathrm {n . d .} \quad 1. 3 3 \pm 0. 7 2 \quad 2. 4 1 ^ {a} \pm 0. 8 1 \quad 4. 4 1 \quad 1 2. 8 0 ^ {b c} \pm 3. 6 7 $$ $$ M 2 \quad 4 6. 5 ^ {\mathrm {b}} \pm 3. 7 \quad 5. 4 ^ {\mathrm {a b}} \pm 1. 9 \quad 1 3. 9 ^ {\mathrm {a}} \pm 2. 4 \quad 1 5. 0 ^ {\mathrm {a}} \pm 2. 8 \quad 6 9. 3 ^ {\mathrm {b c}} \pm 5. 2 \quad \mathrm {n . d .} \quad 0. 8 0 \pm 0. 2 9 \quad 2. 1 3 ^ {\mathrm {a}} \pm 0. 7 1 \quad 3. 0 1 \quad 2 3. 4 8 ^ {\mathrm {a}} \pm 5. 1 5 $$ $$ \mathrm {K 1} \quad 4 9. 1 ^ {\mathrm {a b}} \pm 1. 8 \quad 7. 7 ^ {\mathrm {a}} \pm 1. 8 \quad 1 4. 7 ^ {\mathrm {a}} \pm 2. 4 \quad 1 6. 7 ^ {\mathrm {a}} \pm 2. 6 \quad 6 2. 5 ^ {\mathrm {c}} \pm 4. 2 \quad \mathrm {n . d .} \quad 2. 7 7 \pm 1. 2 2 \quad 2. 6 9 ^ {\mathrm {a}} \pm 0. 5 3 \quad 5. 0 8 \quad 1 1. 1 4 ^ {\mathrm {c}} \pm 2. 3 5 $$ $$ \begin{array}{l l l l l l l l l l} \mathrm {K 4} & 4 4. 3 ^ {\mathrm {b}} \pm 3. 3 & 8. 2 ^ {\mathrm {a}} \pm 1. 8 & 1 4. 0 ^ {\mathrm {a}} \pm 1. 2 & 1 6. 2 ^ {\mathrm {a}} \pm 1. 8 & 6 0. 1 ^ {\mathrm {c}} \pm 4. 3 & \mathrm {n . d .} & 2. 5 9 \pm 1. 0 5 & 3. 0 3 ^ {\mathrm {a}} \pm 0. 7 5 & 1. 6 7 & 1 8. 4 8 ^ {\mathrm {a b}} \pm 3. 6 7 \end{array} $$ $$ R 1 \quad 5 3. 6 ^ {a b} \pm 4. 0 \quad 0. 3 ^ {c} \pm 1. 1 \quad 8. 4 ^ {b} \pm 0. 6 \quad 8. 4 ^ {b} \pm 0. 5 \quad 8 8. 0 ^ {a} \pm 8. 2 \quad n. d. \quad 0. 3 1 \pm 0. 1 2 \quad 0. 7 7 ^ {c} \pm 0. 2 1 \quad 5. 9 8 \quad 1 2. 8 6 ^ {a b} \pm 3. 3 9 $$ $$ R 2 \quad 5 2. 2 ^ {a b} \pm 2. 8 \quad 4. 4 ^ {a} \pm 0. 7 \quad 1 2. 8 ^ {a} \pm 1. 2 \quad 1 3. 5 ^ {a} \pm 1. 3 \quad 7 1. 2 ^ {b} \pm 2. 4 \quad n. d. \quad 0. 8 3 \pm 0. 3 0 \quad 1. 7 5 ^ {a} \pm 0. 3 6 \quad 1. 3 2 \quad 1 7. 0 7 ^ {a} \pm 6. 5 9 $$ $$ \mathrm {R 3} \quad 5 3. 8 ^ {\mathrm {a}} \pm 2. 9 \quad 2. 1 ^ {\mathrm {a b}} \pm 1. 1 \quad 1 1. 9 ^ {\mathrm {a b}} \pm 1. 4 \quad 1 2. 2 ^ {\mathrm {a b}} \pm 1. 6 \quad 8 0. 6 ^ {\mathrm {a b}} \pm 4. 3 \quad \mathrm {n . d .} \quad 0. 3 9 \pm 0. 1 2 \quad 1. 4 6 ^ {\mathrm {a b}} \pm 0. 3 5 \quad 1. 9 9 \quad 1 1. 6 5 ^ {\mathrm {b}} \pm 1. 4 5 $$ $$ R 4 \quad 4 7. 7 ^ {\mathrm {b}} \pm 3. 3 \quad 4. 0 ^ {\mathrm {a b}} \pm 1. 6 \quad 1 3. 1 ^ {\mathrm {a}} \pm 2. 5 \quad 1 3. 7 ^ {\mathrm {a}} \pm 2. 8 \quad 7 3. 6 ^ {\mathrm {b}} \pm 3. 8 \quad \mathrm {n . d .} \quad 1. 0 6 \pm 0. 2 5 \quad 2. 1 3 ^ {\mathrm {a}} \pm 0. 4 6 \quad 2. 5 9 \quad 1 1. 6 1 ^ {\mathrm {b}} \pm 1. 7 4 $$ $$ R 5 \quad 5 0. 8 ^ {a b} \pm 3. 5 \quad 1. 5 ^ {b c} \pm 2. 2 \quad 1 2. 9 ^ {a} \pm 3. 7 \quad 1 3. 1 ^ {a} \pm 3. 9 \quad 8 4. 5 ^ {a} \pm 8. 2 \quad n. d. \quad 0. 3 0 \pm 0. 2 9 \quad 1. 0 6 ^ {b c} \pm 0. 4 0 \quad 2. 4 \quad 9. 7 1 ^ {b} \pm 1. 3 7 $$ $$ \mathrm {K 1} \quad 4 9. 1 ^ {\mathrm {a}} \pm 1. 8 \quad 7. 7 ^ {\mathrm {b}} \pm 1. 8 \quad 1 4. 7 ^ {\mathrm {b}} \pm 2. 4 \quad 1 6. 7 ^ {\mathrm {b}} \pm 2. 6 \quad 6 2. 5 \pm 4. 2 \quad \mathrm {n . d .} \quad 2. 7 7 \pm 1. 2 2 \quad 2. 6 9 ^ {\mathrm {b}} \pm 0. 5 3 \quad 5. 0 8 \quad 1 1. 1 4 ^ {\mathrm {c}} \pm 2. 3 5 $$ $$ \mathrm {K 2} \quad 4 8. 0 ^ {\mathrm {a}} \pm 2. 8 \quad 1 2. 5 ^ {\mathrm {a}} \pm 2. 8 \quad 1 9. 0 ^ {\mathrm {a}} \pm 2. 4 \quad 2 2. 8 ^ {\mathrm {a}} \pm 3. 4 \quad 5 7. 0 \pm 3. 4 \quad \mathrm {n . d .} \quad 3. 5 0 \pm 1. 0 4 \quad 4. 2 8 ^ {\mathrm {a}} \pm 1. 1 2 \quad 2. 3 3 \quad 2 1. 4 4 ^ {\mathrm {a}} \pm 3. 1 0 $$ $$ \mathrm {K 3} \quad 4 3. 6 ^ {\mathrm {b}} \pm 3. 1 \quad 8. 3 ^ {\mathrm {b}} \pm 1. 8 \quad 1 4. 6 ^ {\mathrm {b}} \pm 2. 0 \quad 1 6. 9 ^ {\mathrm {b}} \pm 2. 2 \quad 6 0. 5 \pm 5. 7 \quad < 0. 2 4 0 \quad 2. 6 6 \pm 1. 5 4 \quad 3. 4 9 ^ {\mathrm {a b}} \pm 0. 9 2 \quad 3. 4 9 \quad 1 5. 0 5 ^ {\mathrm {b c}} \pm 5. 1 8 $$ $$ \begin{array}{l l l l l l l l l l} \mathrm {K 4} & 4 4. 3 ^ {\mathrm {b}} \pm 3. 3 & 8. 2 ^ {\mathrm {b}} \pm 1. 8 & 1 4. 0 ^ {\mathrm {b}} \pm 1. 2 & 1 6. 2 ^ {\mathrm {b}} \pm 1. 8 & 6 0. 1 \pm 4. 3 & \mathrm {n . d .} & 2. 5 9 \pm 1. 0 5 & 3. 0 3 ^ {\mathrm {a b}} \pm 0. 7 5 & 1. 6 7 & 1 8. 4 8 ^ {\mathrm {a b}} \pm 3. 6 7 \end{array} $$ a-d Values with different letters within comparisons are significantly different (P<0.05); n.d., not detected TBARS TBARS values were analysed together with n-3 PUFA, n-6 PUFA, LC and TC in the PLS-DA model to investigate possible interactions between oxidation and these parameters. Separation was found between groups R1, M1 and K1 and between groups R4, M2 and K4 (Figure 3a). Statistical analysis showed that groups M2 and K4 had a higher TBARS value than group R4 (Table 5). Group R2 was separated from the other groups close to TBARS (Figure 3b).

Click to enlarge

The TBARS value of group R2 was higher than that of groups R3, R5 and R4. Group K2 differed from other FKO groups close to TBARS (Figure 3c). The TBARS value of group K2 was higher than that of groups K1 and K3. In plots a and b, n-3 PUFA distributed on the same side as TBARS (not close) while n-6 PUFA was on the other side. In plots b and c, TBARS was close to TC, but far from n-3 PUFA (Figure 3).

Click to enlarge

Click to enlarge

Figure (3a, 3b, 3c): PLS-DA analysis of thiobarbituric reactive substances (TBARS) value with fatty acids and carotenoid in white muscle (n=6): a, b, c are biplot of data from groups R1, R4, M1, M2, K1, K4, groups R1, R2, R3, R4, R5 and groups K1, K2, K3, K4 explaining 66.5%, 76.3% and 80.5% of variation in samples, respectively. For abbreviations, see Figures 1 and 2.