Effect of Fatty Acids Profile with Thermal Oxidation of Chicken Fat on Characteristic Aroma of Chicken Flavors Assessed by Gas Chromatography-Mass Spectrometry and Descriptive Sensory Analysis

Abstract

The effect of fatty acids on the characteristic flavor obtained from the Maillard reaction is of particular interest today. Chicken fats were respectively heated at 7 temperatures with an interval increment of 20°C within 60–180°C, while non-heated chicken fat was as a control. Thirteen kinds of fatty acids were identified by gas chromatography-mass spectrometr. (GC-MS). These fatty acids were divided into 3 categories including 7 saturated fatty acids (SFA), 3 monounsaturated fatty acids (MUFA), and 3 polyunsaturated fatty acids (PUFA). Fifty-two kinds of aroma compounds were detected from lipid-Maillard reaction products (MRPs). Analysis of variance (ANOVA) statistical analysis indicated the significant difference (p < 0.001) in six sensory attributes (fatty, meaty, roasty, off-flavor, fresh and overall odor) of MRPs. Partial least square regression (PLSR) was used to detect positive correlation among fatty acids, volatile compounds and sensory attributes. The result showed that 100°C sample (S-100) was correlated with overall flavor and MUFA.

Introduction

Savory meat flavors, which are commonly generated during cooking (e.g., boiling, frying, and roasting), have been a critical and desirable flavoring ingredient due to its extensive applications in the food industry. Many studies have revealed that animal fats have significant impacts on generation of characteristic meat flavors. (Calkins and Hodgen, 2007; Song et al., 2011) Many volatile chemicals, such as aldehydes, ketones, and free fatty acids derived from heated fats have been investigated. (Mottram, 1998; Song et al., 2010b) Based on those studies, it was found that moderate lipid oxidation could generate abundant desirable compounds and thus enhance food flavors, (Zamora and Hidalgo, 2011) although excessive lipid auto-oxidation often brought off-flavors. On the other hand, compounds generated from oxidative breakdown of lipids could participate in the Maillard reaction to form flavoring compounds. (Tikk et al., 2008)

Lipid decomposition and/or oxidation were considered the cause of many animals' species-specific flavors. (Shi et al., 2013) Thermal oxidation of lipid can impart desirable meat flavors in different ways. Many volatile compounds including aldehydes, ketones, nitrogen-containing compounds and sulfur-containing compounds can be produced during lipid decomposition. (Shi et al., 2013) Even more, some characteristic flavors could be influenced by fatty acids. (Lorenz et al., 2014) Free fatty acids as flavor precursors can react with amino acids to generate and improve meaty, fatty, and overall food flavors. For example, the effect of free fatty acids on the odor of pork was reported according to a reaction model. (Aaslyng and Schäfer, 2007) It was also reported that oxidized animal fat participated in Maillard reaction and produced aroma compounds. (Song et al., 2011; Zamora and Hidalgo, 2011) Previous study demonstrated that animal fat made great contribution to characteristic meat flavor and some researches even reported the effect of cooking or feeding on fatty acids profile. Our study further explored the effect of fatty acid profiles with different oxidized temperature on the formation of volatile compounds and flavor acceptance. And the correlation among fatty acid profiles, the formation of volatile compounds and flavor acceptance was also investigated. In addition, previous study reported the effect of individual fatty acids on odor compounds according to a model experiment, while the additional fatty acids could not reflect the truth of meat processing. Our study used animal fat with thermal oxidation as the source of unsaturated fatty acids to explore the formation of odor compounds, which has practical guiding significance and reference value for natural meat flavor.

Gas chromatography-mass spectrometry (GC-MS) was applied to detect fatty acids and volatile compounds. (Song et al., 2010a; Bermudez et al., 2012) In addition, the partial least square regression (PLSR) was used to explore the correlations between sensory evaluation data and instrumental/chemical data. (Eric et al., 2013) However, the study on chicken flavor by GC-MS, PLSR combined with sensory evaluation simultaneously was few reported.

Therefore, the objectives of this study were to (a) analyze the fatty acid profiles of different thermally oxidized chicken fats by GC-MS; (b) apply descriptive sensory analysis to describe and evaluate the attributes of MRPs (Maillard reaction products) generated by amino acids with thermal oxidation of chicken fat; (c) analyze the aroma compounds derived from the MRPs by SPME/GC-MS; and (d) study the correlations among the aroma compounds, fatty acid profile and sensory attributes by PLS1 and PLS2.

Materials and Methods

Materials and reagents    Chicken oil was purchased from Anhui Muyang Oil and Fats Co., Ltd. (Anhui, China) in May of 2014. Chicken breast was purchased from a local TESCO supermarket (Shanghai, China) in May of 2014. Enzyme Protamex was purchased from Novozymes (Bagsvaerd, Denmark). D-xylose, L-cysteine hydrochloride, glycine, β-alanine, thiamine hydrochloride, methanol (chromatography grade), methyl pentanoate (chromatography grade) and other chemicals were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Standard chemical 1,2-dichlorobenzene (≥98%) was purchased from Sigma-Aldrich Co. Ltd. (Shanghai, China). Trimethylsulfonium hydroxide used as a methylating reagent (0.2 mol/L methanol Soln.) was purchased from TCI (Shanghai) Development CO. Ltd. (Shang, China).

Preparation of thermally oxidized chicken fat    The chicken fat (50 g) was placed in a 250 mL 3-neck round bottom flask and heated at each of the following 7 temperatures of 60, 80, 100, 120, 140, 160 or 180°C in a thermostatic oil bath with magnetic stirring at 200 rpm. The feeding air was purged into the flask at a rate of 100 L/h. After heating for 30 minutes, the samples were immediately cooled in ice-water and stored at −18°C for further use. Each treatment was conducted in triplicate. Seven oxidized fat samples corresponding to 7 temperature treatments (i.e., C-60, C-80, C-100, C-120, C-140, C-160 or C-180) were used for fatty acid analysis and further Maillard reaction.

Preparation of fatty acid methyl esters    Chicken fat samples were converted to their fatty acid methyl esters (FAMEs) according to ISO 5509:2000. Trimethylsulfonium hydroxide (TMSH) was used as the methylation reagent. (Drechsel et al., 2003) Methyl tridecanoate (C13:0 ME) in a concentration of 0.48 mg/mL was used as an internal standard

To quantify the fatty acid, the samples were run in triplicate, and the integrated areas based on the total ion chromatograms were normalized to the areas of the internal standard and averaged. The fatty acid concentrations in the oxidized fat samples were determined by comparison with the concentration of the internal standard (Methyl tridecanoate).

Preparation of enzymatic hydrolysis-chicken breast (chicken base)    Chicken breast, which contained 12.83% of protein measured by the Kjeldahl method, was minced with a tissue-tearor and mixed with deionized water at a meat-water ratio of 1:3 in an enzyme reactor with mechanical stirring at 500 rpm, and adjusted to pH 6.5. Then the chicken breast was hydrolyzed at 50°C for 3 h using the protamex with an enzyme/substrate ratio (E/S) of 0.3% (g protamex/g chicken breast).

Preparation of MRPs    A mixture was prepared with the following ingredient: D-xylose (1.28 g), L-cysteine hydrochloride (0.85 g), glycine (0.43 g), β-alanine (0.43 g), thiamine hydrochloride (0.85 g), sodium chloride (NaCl) (2.13 g), glutamic acid monosodium salt (0.43 g), chicken protein hydrolysate (33.62 g), chicken fat (C-control) or thermally treated chicken fat (C-60, C-80, C-100, C-120, C-140, C-160 or C-180). Then its pH was adjusted to 7.0 with 4 mol/L NaOH and 1 mol/L HCl. The solution was transferred into 90 mL screw-sealed tubes and heated in a thermostatic oil bath with magnetic stirring (150 rpm) at 108°C for 60 min. The Maillard reaction products (MRPs) were named S-control, S-60, S-80, S-100, S-120, S-140, S-160 and S-180 for further analysis. Each treatment was repeated three times.

Analysis of free fatty acids (FFAs)    Fatty acid profile was analyzed by GC (Agilent Technologies 7890A) according to the ISO 5508:1990method. An aliquot of 0.3µL of the methyl esters of fatty acids (MEFA) was manually injected in a splitless mode. A HP-INNOWAX capillary column (60 m × 0.25 mm × 0.25 µm) was used to separate the fatty acids. The oven temperature was programmed at 60°C for 3 min, ramped to 100°C at the rate of 15°C/min and held for 10 min, then ramped to 195°C at a rate of 8°C/min and held for 20 min, finally ramped to 250°C at 10°C/min and held for 10 min. Fatty acids were identified by their retention times and compared with the authenticated standards. Fatty acid was quantified in percentage of the identified total fatty acids. (Amira et al., 2010)

Analysis of volatile compounds of MRPs by SPME/GC-MS    At first, 8.0 g sample was introduced in a 20 mL headspace transparent glass vial. Meanwhile, an internal standard (4 µL of 480 mg/L of 1, 2-dichlorobenzene in methanol) was added into the samples with a micropipette. Then the vial was sealed with PTFE/BYTL septum and equilibrated at 75°C for 10 min. SPME-fibre (75 µm Carboxen/polydimethylsiloxane) was used to extract volatile compounds. Thirdly, after 40 min of the SPME adsorption, the fibre was thermally desorbed in the GC injector at 250°C for 5 min. The volatile compounds were separated with an HP-INNOWAX capillary column (60 m × 0.25 mm × 0.25 µm). The oven temperature was programmed at 50°C for 3 min, ramped to 230°C at the rate of 4°C/min, and held for 10 min. The flow rate of carrier gas (helium) was 1 mL/min in a split ratio of 2:1. The MS detector (Agilent Technologies 5975C) was operated under an ionization voltage at 70 eV and emission current of 35 µA. The detector was set at a full scan range of m/z 35–450 at a rate of 4.45 scans/s. Identification and quantification of volatile compounds was referenced by literature. (Shi et al., 2013)

Sensory descriptive analysis    An evaluation panel was screened according to basic olfactory and taste discrimination tests (American Society for Testing and Materials, 1981). Twenty five candidates were recruited from the school of Perfume and Aroma Technology at Shanghai Institute of Technology (Shanghai, PR China). All the panel members were experienced with sensory evaluation. (Mitterer-Daltoé et al., 2012) Total of 8 well-trained panelists (5 females and 3 males at the age of 22 – 45) were selected to constitute a sensory panel for descriptive analysis. A 10-point interval scale (0 = none, 9 = extremely strong) was applied for evaluating sensory intensity.

The selected panel was trained according to the ASTM STP 758 (ASTM 1981) and ISO 8586-1:1993(E). Reference samples were presented as follows: The refined chicken oil was labeled in a “fatty” attribute; The defatted chicken breast (500 g, 2.0 cm of thickness) that was boiled in water for 40 min was labeled in a “meaty” attribute; The chicken breast (200 g, 1.5 cm of thickness) that was roasted on a barbecue grill for 20 min was labeled in a “roasty” attribute; The excessively oxidized chicken oil (100 g of chicken oil) that was placed in a thermotank at 35°C for 3 days to promote the oxidative rancidity was labeled in an “off-flavor” attribute; The fresh chicken breast (250 g, 1.0 cm of thickness) was labeled in a “fresh” attribute; The intensity of all aroma attributes taken together was labeled in an “overall flavor” attribute. In addition, the samples were coded with a three-digit number in a random order before the sensory evaluation to avoid a so-called order effect, and kept in a water bath at 50°C in order to avoid a temperature influence. The panelists breathed the fresh air for ten minutes after three samples were evaluated to avoid sensory fatigue. The evaluation was performed in triplicate to take an average score of three sets of parallel tests.

Statistical analysis    Data from the analyses of fatty acid profile, chemical compounds or descriptive analysis were evaluated by analysis of variance (ANOVA) using the SAS 8.2 (SAS Institute Inc., Cary.NC. USA). ANOVA with Duncan's multiple comparison test was used to check the significant difference among the samples. Analysis of PLSR was performed by the Unscrambler 9.7 (CAMO ASA, Oslo, Norway). Correlations between the fatty acid profiles of the oxidized fat and individual sensory attribute responses of MRPs were analyzed by PLS1. And the PLS2 was applied to illustrate potential connections among the fatty acid profiles, sensory attributes and flavor compounds, taking advantage of the PLS2 parameter that can handle several responses simultaneously (Tikk et al., 2008).

Results and Discussion

Effect of temperature on fatty acid profile of chicken oil    The quantitative values of fatty acids in the oxidized chicken oil is presented in Table 1. The ANOVA analysis with the Duncan's multiple comparison test showed that the sample of C-180 had the highest content of saturated fatty acid (SFA) 25 mg/g, C-140 had the high content of monounsaturated fatty acid (MUFA) 23 mg/g and polyunsaturated fatty acid (PUFA) 4.8 mg/g. In contrast, the sample of C-control (without the thermal treatment) presented the lowest contents of MUFA 5.3 mg/g and PUFA 1.5 mg/g.

Table 1. Quantitative value of fatty acids in thermally oxidized chicken fat

Fatty acid	Fatty acid concentration (mg/g)x
	C-control	C-60	C-80	C-100	C-120	C-140	C-160	C-180
C8:0	0.041 ± 0.008e	0.053 ± 0.008e	0.11 ± 0.01d	0.17 ± 0.01c	0.22 ± 0.03b	0.36 ± 0.02a	0.39 ± 0.03a	0.41 ± 0.07a
C10:0	0.032 ± 0.002f	0.049 ± 0.008f	0.088 ± 0.005e	0.12 ± 0.02d	0.17 ± 0.04c	0.23 ± 0.01b	0.24 ± 0.02b	0.29 ± 0.01a
C14:0	0.095 ± 0.010f	0.15 ± 0.02f	0.33 ± 0.07e	0.41 ± 0.03d	0.60 ± 0.03b	0.49 ± 0.03c	0.54 ± 0.03bc	1.6 ± 0.1a
C14:1 n-9	0.083 ± 0.001e	0.11 ± 0.01de	0.20 ± 0.03a	0.20 ± 0.01a	0.18 ± 0.02ab	0.16 ± 0.01bc	0.14 ± 0.02cd	0.086 ± 0.025e
C15:0	0.073 ± 0.001b	0.075 ± 0.007b	0.14 ± 0.02a	0.16 ± 0.13a	0.045 ± 0.015b	0.067 ± 0.009b	0.062 ± 0.047b	0.046 ± 0.025b
C16:0	3.0 ± 0.1h	3.9 ± 0.1g	7.9 ± 0.3f	9.5 ± 0.4e	12 ± 0.2d	15 ± 0.2c	17 ± 0.3b	21 ± 0.9a
C16:1 n-7	0.80 ± 0.05g	1.0 ± 0.1g	2.2 ± 0.1f	2.7 ± 0.2e	3.4 ± 0.3d	4.4 ± 0.2c	4.7 ± 0.1b	5.4 ± 0.2a
C16:2 n-9	0.016 ± 0.000b	0.018 ± 0.006b	0.029 ± 0.008ab	0.032 ± 0.003ab	0.039 ± 0.008a	0.035 ± 0.014a	0. 023 ± 0.005ab	0.015 ± 0.005b
C17:0	0.022 ± 0.003b	0.024 ± 0.007b	0.038 ± 0.008ab	0.47 ± 0.01a	0.050 ± 0.004a	0.057 ± 0.005a	0.050 ± 0.023a	0.036 ± 0.010ab
C18:0	0.59 ± 0.02e	0.74 ± 0.05e	1.4 ± 0.1d	1.6 ± 0.0cd	1.7 ± 0.2c	2.2 ± 0.2a	1.9 ± 0.1b	2.1 ± 0.0ab
C18:1 n-9	4.4 ± 0.1f	5.8 ± 0.1e	11 ± 0.3d	13 ± 0.5c	16 ± 0.5b	19 ± 0.8a	17 ± 0.5b	16 ± 0.7b
C18:2 n-6	1.4 ± 0.1g	1.8 ± 0.1f	2.9 ± 0.1e	3.18 ± 0.17d	3.8 ± 0.2c	4.6 ± 0.3a	4.1 ± 0.1b	3.4 ± 0.1d
C18:3 n-3	0.083 ± 0.001c	0.10 ± 0.00c	0.18 ± 0.01ab	0.19 ± 0.01ab	0.20 ± 0.03a	0.16 ± 0.01b	0.11 ± 0.02c	0.11 ± 0.02c
SFa	3.8 ± 0.1h	5.0 ± 0.1g	10 ± 0.4f	12 ± 0.4e	15 ± 0.3d	19 ± 0.4c	20 ± 0.4b	25 ± 1.0a
MUFA	5.3 ± 0.1g	6.9 ± 0.1f	14 ± 0.3e	16 ± 0.5d	20 ± 0.5c	23 ± 0.6a	22 ± 0.5b	22 ± 0.9b
PUFA	1.5 ± 0.0f	1.9 ± 0.1e	3.1 ± 0.1d	3.4 ± 0.2c	4.1 ± 0.3b	4.8 ± 0.3a	4.3 ± 0.1b	3.5 ± 0.1c

^x  Mean ± standard deviation for control chicken fat sample and seven oxidized chicken fat samples with different heated temperature. Means in the same row with different superscripts are significantly different (p < 0.001) using Duncan's multiple comparison test.

C-control was chicken fat sample without any heated process, which was as control samples. C-60, C-80, C-100, C-120, C-140, C-160 and C-180 were thermally oxidized chicken fat samples heated with 60, 80, 100, 120, 140, 160 and 180°C respectively.

SFA=∑(C8:0+C10:0+C14:0+C15:0+C16:0+C17:0+C18:0)

MUFA=∑(C14:1+C16:1+C18:1)

PUFA=∑(C16:2+C18:2+C18:3)

The predominant SFA was palmitic acid (C16:0) in all samples, which ranged from 3 mg/g to 21 mg/g, followed by C18:0 ranged from 0.59 to 2.2 mg/g. The C16:0 fatty acid in the sample of C-180 presented the highest concentration, which was caused by pyrolysis of triglycerides under high temperature.(Brunton et al., 2002) The oleic acid (C18:1) presented higher content in all samples with its total content varying from 4.4 to 19 mg/g, followed by C16:1 ranged from 0.8 to 5.4 mg/g. In addition, the C18:2, C18:3 and C16:2 fatty acids were detected. The content of C18:3fatty acid was lower than that of C18:2. Moreover, the C-180 sample had only half content of C18:2 fatty acid of the C-control. This phenomenon was ascribed to the instability of polyunsaturated fatty acids because they were more readily oxidized than other fatty acids. (Aaslyng and Schäfer, 2007)

Volatiles from the interaction of oxidized chicken oil and Millard reaction    As shown in Table 2, the main volatile compounds derived from the MRPs were identified via SPME-GC/MS. The major volatile compounds comprised 19 heterocyclic, e.g., the nitrogen- or sulfur-containing compounds, 13 aldehydes, 7 alcohols, 3 ketones and 10 carboxylic acids. The contents of these chemical compounds in the thermally treated samples were significantly different along with increasing temperature. Those short chain volatile chemicals, such as aldehydes, alcohols, ketones and acids that were mainly derived from lipid oxidation (Frankel et al., 1961), showed remarkable increases when the processing temperature increased. However, the concentrations of heterocyclic compounds (such as thiazoles, thiophenes, thiols) that were mainly formed during the Maillard reaction or lipid-Maillard reaction (the reaction of lipid-derived volatiles with Maillard reaction product) presented a significant decrease. The above-mentioned compounds made a great contribution to fatty, meaty, roasty, fresh odor (Calkins and Hodgen, 2007) and most of them were detected in cooking chicken products frequently (Tang et al., 1983).

Table 2. Volatile compounds identified in Maillard reaction products by SPME-GC/MS

Code	Volatile compounds	RIw	IDs	Characteristic aroma	concentration (ug/100gMRPs)t
					S-Control	S-60	S-80	S-100	S-120	S-140	S-160	S-180
	Heterocyclic compounds(N, O, S)
A1	4-Methylthiazole	1302	C	roasted meat	18a	5.9c	1.9d	ND	8.2b	ND	1.6e	1.9d
A2	2-Acetyl thiazole	1674	B	roast, sulfur	18a	8.0b	0.73d	ND	ND	2.8c	ND	ND
A3	5-Formyl-4-methylthiazole	1893	C	roasted meat	9.7a	2.1d	0.80e	5.4b	3.0c	0.68e	0.56e	0.85e
A4	5-(2-Hydroxyethyl)-4-methylthiazole	2350	C	meaty	2.3×103a	1.1×103d	1.5×102g	2.0×103b	1.4×103c	2.4×102f	1.5×102g	2.6×102e
A5	Thiophene, 2-pentyl-	1477	C	metallic	ND	ND	ND	ND	ND	ND	ND	5.94a
A6	3-Thiophenethiol	1629	C	cooked meat	16a	0.38c	ND	ND	0.91b	ND	ND	ND
A7	3-Thiophenecarboxaldehyde	1727	B	sulfur	18a	3.2c	0.36f	5.0b	ND	2.1d	1.4e	1.5e
A8	2-methylfuran	860	B	roasted meat	ND	12b	4.4c	ND	ND	14a	4.1c	ND
A9	2-pentyl-furan	1239	B	meaty	ND	1.9e	3.4d	ND	ND	4.3c	14b	37a
A10	Dihydro-2-methyl-3(2H)-furanone	1284	C	roasted meat	ND	ND	1.0a	ND	ND	1.1a	ND	ND
A11	2-Methyl-3-furanthiol	1349	C	meaty	33a	31b	2.1e	9.5c	ND	8.2d	1.1f	ND
A12	2-Furanmethanethiol	1456	C	meaty	18a	12b	1.4f	7.0c	4.3d	4.3d	1.8e	ND
A13	Furfural	1488	B	caramel	ND	24a	4.9a	ND	ND	7.3c	4.5d	11b
A14	Furfurol	1697	C	burnt	21a	4.0b	1.4c	ND	ND	0.61d	ND	ND
A15	5-Pentyl-2(5H)-Furanone	2118	C	caramel	ND	ND	0.63d	ND	1.5c	ND	7.0b	12a
A16	bis(2-methyl-3-furyl)disulfide	2198	C	meaty	ND	3.1c	5.2a	ND	ND	4.9b	ND	ND
A17	Pyrazine, methyl-	1286	C	cocoa, meat	28a	ND	0.49d	ND	ND	0.93c	0.94c	1.7b
A18	2-Acetyl pyrrole	2004	B	roasted	5.1a	0.76d	0.18g	1.7b	1.2c	0.31f	0.25gf	0.49e
A19	2-Pentyl Pyridine	1595	C	fat	ND	3.2b	0.80e	ND	3.3b	1.3d	1.8c	4.6a
	Aldehydes
A20	Hexanal	1089	B	fatty	14c	2.1h	3.2g	19b	12e	7.1f	12d	23a
A21	Heptanal	1194	B	oily	4.6c	ND	1.7e	ND	ND	3.8d	9.6b	22a
A22	Octanal	1301	B	floral, fatty	11d	9.1e	5.9f	24a	ND	ND	12c	23b
A23	(2E)-2-Heptanal	1343	B	oily	ND	ND	ND	ND	28a	ND	3.0b	ND
A24	Nonanal	1410	A	fatty, oily	24c	20e	16g	18f	18f	23d	70b	1.0×102a
A25	2-Octenal	1451	B	nut, fat	ND	ND	1.3e	4.6b	3.7c	2.3d	3.6c	11a
A26	Benzaldehyde	1553	B	nutty	42a	9.1d	3.5e	25b	13c	3.9e	4.0e	9.9d
A27	2-Nonenal	1558	B	fatty	ND	ND	2.2c	ND	2.4c	ND	5.5b	15a
A28	Undecanal	1625	C	fatty, oily	2.7b	ND	0.23d	ND	ND	ND	0.56c	7.3a
A29	E-2-Decenal	1672	B	nutty	ND	ND	ND	12c	ND	ND	14b	49a
A30	2,4-Nonadienal	1735	B	fatty, oily	ND	ND	ND	ND	2.43a	ND	0.44c	1.09b
A31	2-Undecenal	1778	B	nutty, almond	12a	8.6f	3.8g	14c	16b	ND	11e	44a
A32	2,4-Decadienal	1841	B	fatty, oily	49b	18e	9.1g	31c	20d	14f	31c	64a
	Alcohols
A33	1-Pentanol	1263	B	fusel oil	24c	8.0f	5.8h	18e	33b	7.0g	21d	36a
A34	1-Penten-3-ol	1325	B	mushroom	ND	ND	0.49c	ND	ND	0.38d	1.1b	4.7a
A35	1-Hexanol	1366	A	oily	ND	ND	3.0e	8.7a	ND	4.3d	4.8c	8.4b
A36	1-Octen-3-ol	1463	A	mushrooms	11a	1.1f	3.2e	4.1d	4.3d	5.8c	10b	11a
A37	1-Heptanol	1469	A	floral	14b	5.3e	6.0d	2.5g	2.4g	5.1f	10c	16a
A38	1-Octanol	1572	A	burnt	ND	7.8c	5.5d	ND	8.2c	3.5e	12b	22a
A39	Benzyl Alcohol	1904	C	flower	7.4c	2.7d	1.9e	2.9d	ND	3.0d	10b	20a
	Ketones
A40	2-Octanone	1298	C	herb	35a	11b	ND	ND	ND	ND	ND	3.5c
A41	3-Octen-2-one	1427	C	mushroom	ND	ND	1.0c	ND	ND	1.7c	4.9b	7.3a
A42	3-Nonen-2-one	1533	B	mushroom	ND	ND	ND	ND	ND	ND	1.7b	4.5a
	Carboxylic acid
A43	Acetic acid	1474	A	sour	59a	23d	4.7f	34c	46b	7.0e	3.5g	5.3f
A44	Butanoic acid	1654	A	fermented	ND	1.7b	1.0c	ND	2.4a	0.74d	0.69d	ND
A45	Pentanoic acid	1779	C	herbal	5.5a	ND	0.90b	ND	ND	ND	ND	ND
A46	Hexanoic acid	1870	A	mushroom	23a	17b	3.5f	14d	17b	3.5f	4.5e	16c
A47	Heptanoic acid	1979	A	fatty	9.2a	3.1d	1.9e	3.7c	3.7c	0.86g	1.5f	6.6b
A48	Octanoic acid	2087	A	fatty	15a	6.1b	2.1e	4.6c	3.7d	1.4f	2.1e	6.1b
A49	Nonanoic acid	2203	A	fatty	ND	6.5b	ND	ND	ND	ND	2.7c	8.2a
A50	Hexadecanoic acid	2428	A	fatty	ND	ND	30d	4.8×102a	1.5×102b	3.0e	ND	36c
A51	Benzoic acid	2509	C	flower	12a	7.8b	5.0d	ND	ND	ND	3.1e	5.8c
A52	Hexadecenoic acid	2570	C	fatty	ND	ND	6.9c	1.4×102b	4.9×102a	ND	ND	ND

ND: not detected

^w  Linear retention index calculated on an HP-INNOWAX capillary column (60 m × 0.25 mm × 0.25 µm) with a homologous series of n-alkanes (C₇ – C₃₀).

^s  ID, identification; the identification was indicated by the following: A, mass spectrum and RI agreed with that of the authentic compound ran under similar GC-MS condition; B, mass spectrum and RI agreed with literature data and the flavornet database (http://www.flavornet.org, accessed June2007), Acree, 2004 (on C20M stationary phase) and From the LRI Database on the Web (http://www.odour.org.uk/lriindex.html); C, mass spectrum compared with NIST98 and Wiley mass spectral database.

^t  The concentrations were calculated through an internal standard method, the data correspond to the mean standard deviation (average of triplicate) and the different superscripts indicated the significant difference (p < 0.001) using Duncan's multiple comparison test.

S-control was samples prepared with C-control chicken fat samples, which was as control samples. S-60, S-80, S-100, S-120, S-140, S-160 and S-180 were samples prepared from thermally oxidized chicken fat samples C-60, C-80, C-100, C-120, C-140, C-160 and C-180 respectively.

Sulfur-containing furan derivatives, as the Maillard reaction-derived compounds, presented strong meaty flavor. Thiol or sulfide group-substituted furans, e.g., 2-furanmethanethiol, 2-methyl-3- furanthiol and bis(2-methyl-3-furyl)disulfide generally presented meaty, roasty and burnt characteristics at low concentrations, but gave sulfurous and objectionable odors at high concentrations. (Mottram, 1998) 2-Acetyl pyrrole and 2-pentyl pyridine typically presented fatty and roasted aroma, which were also found in fried chicken (Jayasena et al., 2013b). Thiazoles and thiophenes were compounds derived from Maillard reaction and carbonyl groups were originated from lipid oxidation (Xu et al., 2011). Many short chain volatile chemicals, such as saturated and unsaturated aldehydes, alcohols, ketones and carboxylic acid, could be derived from lipid degradation. For example, 1-pentanol, 1-heptanol, 1-octanol, 2-penten-1-ol, 1-octen-3-ol, 2-pentylfuran and 2,4-nonadienal were identified as oxidation products of C18:1 and C18:2 fatty acids. (Aaslyng and Schäfer, 2007) Hexanal was widely considered as a marker of lipid oxidation in meat. (Shahidi and Pegg, 1994) Heptanal, octanal, 2-decenal, 2,4-nonadienal, 2-unedcenal, 2,4-decadienal, 1-octen-3-ol and pentanoic acid were proposed as chromatographic fingerprint of oxidized fat. (Song et al., 2013) In addition, 2-heptanal, nonanal, 2-octenal, 2-nonenal, 2-decenal, 2-undecenal and 2,4-decadienal were responsible for a chicken-specific flavor (Jayasena et al., 2013a). Pentanoic acid, hexanoic acid and nonanoic acid were detected in cooked turkey breast (Brunton et al., 2002). Other acids such as acetic acid, butanoic acid, heptanoic acid, octanoic acid and hexadecanoic acid were also found in lipid oxidation (Shi et al., 2013). Most acidic compounds provide sour, oily and fatty odors and have relatively low odor thresholds.

Sensory analysis of MRPs with oxidized chicken oil    The Duncan grouping in light of six sensory attributes (i.e., fatty, meaty, roasty, off-flavor, fresh and overall flavor) is shown in Table 3. Sample of S-control was produced with unprocessed chicken fat while other samples were prepared from oxidized chicken fat treated under different temperatures. ANOVA analysis indicated that the scores of each attribute among the 8 MRPs samples were significantly different (p < 0.001). As shown in Table 3, S-control received the highest scores in fatty and meaty attributes, which were given 6.58 and 7.88 points, respectively. The S-180 sample received the highest score of 6.21 points in the roasty note and 7.08 points in off-flavor, which was followed by S-160. The S-100 sample was received the highest score of 7.88 points in overall flavor, but lowest score of 1.92 points in off-flavor. Obviously, sensory attributes of MRPs treated under different heating temperatures were significantly different. Inadequate or excessive lipid oxidation did not lead to pleasant flavor. This finding is in agreement with literature (Song et al., 2010b; Tian et al., 2014).

Table 3. The mean scores of the six attributes for the eight MRPs

Samples	Mean scorey
	Fatty (***)	Meaty (***)	Roasty (***)	Off-flavor (***)	Fresh (***)	Overall flavor (***)
S-control	6.58a	7.88a	3.25c	4.00c	5.88c	4.75d
S-60	5.13b	7.58a	3.54c	3.00d	5.21d	5.25c
S-80	2.63f	5.38d	5.04b	2.96d	5.04d	5.29c
S-100	3.63de	7.04b	2.25d	1.92e	6.08c	7.88a
S-120	3.88cd	5.25d	2.04d	2.63d	4.75de	4.67d
S-140	3.08ef	6.67b	4.88b	2.38de	4.38e	6.13b
S-160	2.58f	4.88d	5.25b	5.33b	6.71b	3.79e
S-180	4.46c	6.08c	6.21a	7.08a	7.58a	2.46f

Notations *** indicate significance at p < 0.001.

^y  Mean score for each attribute within a column with different letters are significantly different (p < 0.001) using Duncan's multiple comparison test (n=24, 8 panelists with 3 replications).

S-control was samples prepared with C-control chicken fat samples, which was as control samples. S-60, S-80, S-100, S-120, S-140, S-160 and S-180 were samples prepared from thermally oxidized chicken fat samples C-60, C-80, C-100, C-120, C-140, C-160 and C-180 respectively.

Correlation among fatty acids, volatile compounds, and sensory attributes    In this study, partial least squares regression (PLSR) was performed to gain an overview of the relationships among the free fatty acids, aroma compounds and sensory features. Before regression, all variables were weighted with their own standard deviations (1/SDev) to obtain unbiased contribution of each variable. A subsequent full cross validation method was used to validate the model.

Fig.1 profiles a correlation loadings plot for the saturation degree of fatty acids and aroma compounds as the X-matrix, and sensory attributes and Maillard reaction products (MRPs) as the Y-matrix. The derived parameter PLS2 included 3 significant PCs that explained 82% of the cross-validated variances. The X-matrix explained variances of PC1=38%, PC2=32% and PC3=12%, while the Y-matrix explained variances of PC1=27%, PC2=19% and PC3=12%. The large and small circles indicated 50% and 100% explained variance, respectively. Six sensory attributes were placed between the inner and outer ellipses (r² = 0.5 and 1.0 respectively), indicating they were well explained by the PLSR model. Heterocyclic compounds such as 4-methylthiazole (A1), 2-acetylthiazole (A2), 5-formyl-4-methylthiazole (A3), 5-(2-hydroxyethyl)-4-methylthiazole (A4), 3-thiophenethiol (A6), 3-thiophenecarboxaldehyde (A7), 2-methyl-3-furanthiol (A11), 2-furanmethanethiol (A12) showed a positive correlation with the meaty and fatty flavors. These compounds were mainly derived from Maillard and Lipid-Maillard reactions. (Yang et al., 2015) In addition, 2-octanone (A40), pentanoic acid (A45), hexanoic acid (A46) and octanoic acid (A48) interpreted fatty attribute positively, which derived from lipid oxidation. Hexanal (A20), undecanal (A28), 2-undecenal (A31), 2,4-decadienal (A32), 1-pentanol (A33), 1-octen-3-ol (A36) and 1-heptanol (A37) were compounds correlated with the fresh attribute positively. Hexal (A20) and 2-undecenal (A31) had a characteristic odor of green. 1-Octen-3-ol (A36) had mushroom odor and 1-pentanol (A33), 1-heptanol (A37) presented herb odor. 2,4-Decadienal (A32) was described as a fatty, chicken flavor at 10 ppm and fruit flavor at lower concentration.(Calkins and Hodgen, 2007) Both green and mushroom, herb, fruit odors were reduced to fresh attribute. High concentration of heptanal (A21), nonanal (A24), 2-decenal (A29) and benzyl alcohol (A39) led to the off-flavor. It should be ascribed to excessive oxidation of chicken fat with high temperature. 5-Pentyl-2(5H)-furanone (A15), 2-pentylpyridine (A19), 2-octenal (A25), 2-nonenal (A27), 1-penten-3-ol (A36), 1-hexanol (A35), 1-octanol (A38) and 3-octen-2-one (A41)showed positive correlation with a roasty odor. PUFA showed poor interpretation for sensory attributes, which should be due largely to their thermal instability and decomposition. Fig. 1 showed that S-control was separated along PC2 with other samples on the up side and samples of S-160 and S-180 were separated along PC1 with other samples on the right side. S-60, S-80, S-100, S-120 and S-140 were grouped at the left-down corner. In addition, the correlation loadings plot based on PC2 and PC3 showed that S-100 correlated with overall flavor and volatile compounds of 5-pentyl-2(5H)-furanone (A15), 2-pentyl pyridine (A19), 2-octenal (A25), 1-hexanol (A35) and butanoic acid (A44). Hence the above mentioned compounds made great contribution to flavor acceptance for S-100. S-80, S-160 and S-140 showed correlation with roasty and the relative aroma compounds were furfural (A13), 1-octanol (A38) and 3-octen-2- one (A41). S-60 and S-180 presented poor sensory qualities. Fig.2 showed the correlations between fatty acid profiles (X-matrix) and volatile compounds (Y-matrix). PC1 versus PC2 were presented in the correlation loading plot. Other PCs did not present here, as little information was gained through their correlation. For X-matrix, the explained variance was PC1=64% and PC2=24%; for Y-matrix, PC1=28% and PC2=26%. Saturated fatty acids (SFA) were mainly located on the negative axle of the plot and unsaturated fatty acids (MUFA and PUFA) were located on the positive axle along PC2. C8:0, C14:0, C16:0 and C16:1n-7 showed a positive correlation with heptanal (A21), nonanal (A24), 2-octenal (A25), 2-nonenal (A27), undecanal(A28), E-2-decenal (A29), 2-undecenal (A31), 1-penten-3-ol (A34), 1-octen-3-ol (A36), 1-heptanol (A37), 1-octanol (A38), 2-pentylfuran (A9), 5-pentyl-2(5H)-furanone (A15), 3-octen-2-one (A41) and 2, 4-decadienal (A32). Most compounds were considered as lipid oxidation marker in meat. The quantitative value of unsaturated fatty acids C18:1n-9, C14:1n-9, C18:2n-6 and C18:3n-3 showed negative links with 2-acetylthiazole (A2), 3-thiophenethiol (A6), 3-thiophenecarboxaldehyde (A7), 2-methyl-3-furanthiol (A11), 2-furanmethanethiol (A12), furfurol (A14), 5-(2-hydroxyethyl)-4- methylthiazole (A4), 2-octanone (A40), pentanoic acid (A45) and octanoic acid (A48) which mainly correlated with meaty and fatty attributes.

Fig. 1.

PLSR correlation loadings plot: The model was derived from GC-MS isolated compounds, SFA, MUFA and PUFA as the X-matrix and sensory attributes, MRPs as the Y-matrix.

Fig. 2.

PLSR correlation loadings plot: The model was derived from fatty acids as the X-matrix and volatile compounds as the Y-matrix.

In conclusion, this study first demonstrated the fatty acids profile and aroma compounds affected by temperature. High heated temperature of 140°C–180°C led remarkably higher content of SFA but lower content of MUFA. Lower heated temperature of 60°C–80°C had high content of PUFA. The result of PLSR showed the positive correlations between MUFA and overall odor. According to analysis, 100°C was recommended to generate a fatty acids profile for an acceptable chicken flavor.

Acknowledgements    The research was supported in part by National Natural Science Foundation of China (21306114 and 21476140). The National Key Technology R&D Program (2011BAD23B01), Shanghai Engineering Technology Research Center of Fragrance and Flavor (12DZ2251400), Shanghai Excellent Young Teacher Foundation(yyy10070).

References

•  Aaslyng,  M. D. and  Schäfer,  A. (2007). The effect of free fatty acids on the odour of pork investigated by sensory profiling and GC-O-MS. Eur. Food Res. Technol., 226, 937-948.
•  Amira,  M. B.,  Hanene,  J. H.,  Madiha,  D.,  Imen,  B.,  Mohamed,  H., and  Abdelhamid,  C. (2010). Effects of frying on the fatty acid composition in farmed and wild gilthead sea bream (Sparus aurata). Int. J. Food Sci. Technol., 45, 113-123.
•  Bermudez,  R.,  Franco,  I.,  Franco,  D.,  Carballo,  J., and  Lorenzo,  J. M. (2012). Influence of inclusion of chestnut in the finishing diet on fatty acid profile of dry-cured ham from Celta pig breed. Meat Sci., 92, 394-9.
•  Brunton,  N. P.,  Cronin,  D. A., and  Monahan,  F. J. (2002). Volatile components associated with freshly cooked and oxidized off-flavours in turkey breast meat. Flavour Frag. J., 17, 327-334.
•  Calkins,  C. R. and  Hodgen,  J. M. (2007). A fresh look at meat flavor. Meat Sci., 77, 63-80.
•  Drechsel,  D.,  Dettmer,  K., and  Engewald,  W. (2003). Studies of thermally assisted hydrolysis and methylation-GC-MS of fatty acids and triglycerides using different reagents and injection systems. Chromatographia, 57, S283-S289.
•  Eric,  K.,  Raymond,  L. V.,  Huang,  M.,  Cheserek,  M. J.,  Hayat,  K.,  Savio,  N. D.,  Amédée,  M., and  Zhang,  X. (2013). Sensory attributes and antioxidant capacity of Maillard reaction products derived from xylose, cysteine and sunflower protein hydrolysate model system. Food Res. Inte., 54, 1437-1447.
•  Frankel,  E. N.,  Evans,  C. D.,  McConnell,  D. G., and  Jones,  E. P. (1961). Analyses of lipids and oxidation products by partition chromatography. Fatty acid hydroperoxides. J. Am. Oil Chem. Soc., 38, 134-137.
•  Jayasena,  D. D.,  Ahn,  D. U.,  Nam,  K. C., and  Jo,  C. (2013a). Factors affecting cooked chicken meat flavour: a review. World's Poultry Sci. J., 69, 515-526.
•  Jayasena,  D. D.,  Ahn,  D. U.,  Nam,  K. C., and  Jo,  C. (2013b). Flavour chemistry of chicken meat: a review. Asian-Australasian J. Anim. Sci., 26, 732-42.
•  Lorenz,  S.,  Buettner,  A.,  Ender,  K.,  Nürnberg,  G.,  Papstein,  H.-J.,  Schieberle,  P., and  Nürnberg,  K. (2014). Influence of keeping system on the fatty acid composition in the longissimus muscle of bulls and odorants formed after pressure-cooking. Eur. Food Res. Technol., 214, 112-118.
•  Mitterer-Daltoé,  M. L.,  Petry,  F. C.,  Wille,  D. F.,  Treptow,  R. O.,  Martins,  V. M. V., and  Queiroz,  M. I. (2012). Chemical and sensory characteristics of meat from Nellore and Crioulo Lageano breeds. Int. J. Food Sci. Technol., 47, 2092-2100.
•  Mottram,  D. S. (1998). Flavour formation in meat and meat products: a review. Food Chem., 62, 415-424.
•  Shahidi,  F. and  Pegg,  R. B. (1994). Hexanal as an indicator of meat flavor deterioration. J. Food Lipids, 1, 177-186.
•  Shi,  X.,  Zhang,  X.,  Song,  S.,  Tan,  C.,  Jia,  C., and  Xia,  S. (2013). Identification of characteristic flavour precursors from enzymatic hydrolysis-mild thermal oxidation tallow by descriptive sensory analysis and gas chromatography-olfactometry and partial least squares regression. J. Chromatogr. B, 913-914, 69-76.
•  Song,  S.,  Zhang,  X.,  Hayat,  K.,  Huang,  M.,  Liu,  P.,  Karangwa,  E.,  Gu,  F.,  Jia,  C.,  Xia,  S.,  Xiao,  Z., and  Niu,  Y. (2010a). Contribution of beef base to aroma characteristics of beeflike process flavour assessed by descriptive sensory analysis and gas chromatography olfactometry and partial least squares regression. J. Chromatogr. A, 1217, 7788-7799.
•  Song,  S.,  Zhang,  X.,  Hayat,  K.,  Jia,  C.,  Xia,  S.,  Zhong,  F.,  Xiao,  Z.,  Tian,  H., and  Niu,  Y. (2010b). Correlating chemical parameters of controlled oxidation tallow to gas chromatography–mass spectrometry profiles and e-nose responses using partial least squares regression analysis. Sensors Actuators B, 147, 660-668.
•  Song,  S.,  Zhang,  X.,  Hayat,  K.,  Liu,  P.,  Jia,  C.,  Xia,  S.,  Xiao,  Z.,  Tian,  H., and  Niu,  Y. (2011). Formation of the beef flavour precursors and their correlation with chemical parameters during the controlled thermal oxidation of tallow. Food Chem., 124, 203-209.
•  Song,  S.,  Zhang,  X.,  Hayat,  K.,  Xiao,  Z.,  Niu,  Y., and  Eric,  K. (2013). Coordinating fingerprint determination of solid-phase microextraction/gas chromatography-mass spectrometry and chemometric methods for quality control of oxidized tallow. J. Chromatogr. A, 1278, 145-152.
•  Tang,  J.,  Jin,  Q. Z.,  Shen,  G. H.,  Ho,  C. T., and  Chang,  S. S. (1983). Isolation and identification of volatile compounds from fried chicken. J. Agric. Food Chem., 31, 1287-1292.
•  Tian,  H.,  Zhan,  P.,  Li,  W.,  Zhang,  X.,  He,  X.,  Ma,  Y.,  Guo,  Z., and  Zhang,  D. (2014). Contribution to the aroma characteristics of mutton process flavor from oxidized suet evaluated by descriptive sensory analysis, gas chromatography, and electronic nose through partial least squares regression. Eur. J. Lipid Sci. Technol., n/a-n/a.
•  Tikk,  K.,  Haugen,  J. E.,  Andersen,  H. J., and  Aaslyng,  M. D. (2008). Monitoring of warmed-over flavour in pork using the electronic nose - correlation to sensory attributes and secondary lipid oxidation products. Meat Sci., 80, 1254-63.
•  Xu,  Y.,  Chen,  Q.,  Lei,  S.,  Wu,  P.,  Fan,  G.,  Xu,  X., and  Pan,  S. (2011). Effects of lard on the formation of volatiles from the Maillard reaction of cysteine with xylose. J. Sci. Food Agric., n/a-n/a.
•  Yang,  Z.,  Xie,  J.,  Zhang,  L.,  Du,  R.,  Cao,  C.,  Wang,  M.,  Acree,  T., and  Sun,  B. (2015). Aromatic effect of fat and oxidized fat on a meat-like model reaction system of cysteine and glucose. Flavour Fragr. J., 30, 320-329.
•  Zamora,  R. and  Hidalgo,  F. J. (2011). The Maillard reaction and lipid oxidation. Lipid Technol., 23, 59-62.