Effects of UV-C and single- and multiple-cycle high hydrostatic pressure treatments on flavor evolution of cow milk: Gas chromatography-mass spectrometry, electronic nose, and electronic tongue analyses

ABSTRACT

The effects of Ultraviolet Light-C and single- and multiple-cycle high hydrostatic pressure treatments on the flavor evolution of cow milk were studied using solid-phase microextraction-gas chromatography-mass spectrometry, electronic nose, and electronic tongue analyses. The milk samples showed significant differences in their electronic nose profiles, but almost similar in the electronic tongue profiles. The variety and content of volatile compounds in the milk were increased after high hydrostatic pressure and Ultraviolet Light-C treatments, in which Ultraviolet Light-C treatment increased the variety and content of volatiles more than 6–8 times and 27.4–54.4%, respectively, than that of the high hydrostatic pressure treatment. Two 2.5-min cycles of high hydrostatic pressure treatment at 200 MPa resulted in the minimum amount of aliphatic hydrocarbons (25.4%) and maximum amount of aromatic hydrocarbons (31.1%); one 5-min cycle at 400 MPa and two 2.5-min cycles at 600 MPa produced the maximum amount of aliphatic hydrocarbons (45.6%) and acids (27.3%), respectively. The Ultraviolet Light-C treatment produced the maximum amount of aldehydes, aliphatic hydrocarbons, and alcohols.

KEYWORDS:

• Cow milk
• High hydrostatic pressure
• UV-C
• Flavor
• Electronic tongue
• Electronic nose
• Gas chromatography-mass spectrometry

Introduction

Flavor compounds, including volatile and non-volatile components, contribute to the taste of cow milk and is an important attribute of consumer acceptance and preference for milk and its products.^[1–3] Fresh cow milk flavor is distinct yet subtle, which can be overshadowed by off-flavor, thus reducing the sensory quality and economic value of dairy products.^[4] Some distinct volatiles have been used to assess the degree of freshness or deterioration of cow milk.^[3] The flavor properties of cow milk and its products vary depending on cow species, daily diet, environment, age, biological factors, processing, packaging material, and storage conditions.^[2,3,5–7]

Traditionally, heat treatments, particularly ultra-high temperature (UHT) processing, have been used to eliminate potential pathogens and spoilage organisms in milk. However, they produce thermally derived-off flavor compounds such as aldehydes, methyl ketones, and various sulfur compounds.^[5,8] The amino acids and peptides in milk also contribute directly to its flavor, but heat treatments may form ammonia compounds, which cause unpleasant odors.^[3]

Recently, novel non-thermal technologies such as high hydrostatic pressure (HHP) and Ultraviolet Light-C (UV-C) radiation have shown potential to replace, at least partially, traditional preservation processes for the pasteurization of food. HHP has shown the most potential to eliminate microorganisms, thus retaining the sensory and nutritional quality of dairy products.^[9] Previous study has been focused on the effect of single-cycle HHP on the flavor of dairy products. For examples, the relative concentration of 27 selected volatile compounds were increased after HHP treatment compared to an untreated sample.^[10] While high-pressure thermal processing processing increased aldehydes, ketones, furans, pyrans, and alcohols, the content of aliphatic hydrocarbons was reduced compared to non-treated human milk.^[11] HHP also lead to a significant increase in the rate of volatile compound formation,^[12] and the total alcohols, aldehydes, ketones, hydrocarbons, and sulphur compounds in cheese made from unpasteurized milkwere affected by HHP.^[13] Multiple-cycle HHP has been used to effectively inactivate microorganisms, improve food quality, and save cost compared to single-cycle HHP with the same dwelling time.^[14,15] UV-C, specifically at the wavelength range of 250–270 nm, has been used to successfully reduce the microbial load of opaque liquids, such as bovine cow milk, human milk, and dairy products, without significantly affecting their sensory quality.^[16–18] Nevertheless, the effect of multiple-cycle HHP and UV-C treatments on the flavor of cow milk and dairy products has been rarely studied.

Electronic nose (e-nose) and electronic tongue (e-tongue) are sensor technologies that simulate human smell and taste (gas and liquid sensors). E-nose is designed to detect and discriminate among complex odors using an array of sensors that non-selectively interact with odor molecules to produce signals in computer. A pattern recognition process takes place using the data in odor fingerprint for identifying, classifying, and hedonic analysis.^[19,20] E-tongues have been used to detect the sourness, bitterness, and astringency based on an array of taste sensation that combines signals from non-specific, low selective chemical sensors with partial specificity (cross-sensitivity) to differentiate components. The digital fingerprint of gustatory compounds then analyzed with appropriate pattern recognition tools for food analysis.^[21] E-tongues have been proven to be highly efficient alternatives to conventional detection methods in the field of flavor profile analysis because of their high sensitivity, easy-to-build, cost-effective, rapid analysis, and low detection limits.^[22,23] The e-noses and e-tongues have versatility in a wide range of food items, including beverages, grains, cooking oils, eggs and dairy products, meat and fish, and fresh/fresh-cut fruit and vegetables.^[19,24,25] They can classify milk during storage, estimate and predict sensory characteristics, monitor changes in probiotic fermented milk, and detect adulteration in milk.^[1,26]

Gas chromatography-mass spectrometry (GC-MS) has been used to characterize the flavor profile of dairy products and identify the milk flavor from cows fed with different diets.^[10,13,27] Although GC-MS has a high efficiency and accuracy for qualitative and quantitative analysis of volatile compounds, several shortcomings restrict the practicability.^[4] Recently, the simultaneous utilization of e-nose (or e-tongue) and other instrumental methods of analysis have been reported because it can better explain the discrepancies between flavor profiles for different milk samples. The combined analysis of e-nose, e-tongue, and GC-MS showed that the addition of Cheddar cheese to yogurt accelerated the flavor formation and increased the amount of flavor compounds.^[28] Ai et al. analyzed the differences in volatiles in fresh whole and skim milk via e-nose, solid-phase microextraction (SPME), SPME-GC-MS, and GC-olfactometry.^[2] Combined e-nose and e-tongue analyses have been used to differentiate different pasteurized milk brands and determine their shelf life.^[29] The odor profile of five commercial milk flavorings and one self-made enzyme-induced milk flavoring prepared from lipolyzed milk fat samples have been analyzed and compared by SPME-GC-MS, e-nose, and sensory analyses.^[25,30] The objectives of this study were: (1) to determine the type and content of volatiles by GC-MS and distinguish the flavor attributes using e-nose and e-tongue techniques, and (2) to compare the flavor changes of fresh cow milk after single- and multiple-cycle HHP and UV-C treatments.

Materials and methods

Materials

Fresh cow milk (protein content of 3.0% and standardized fat content of 4.0%) was collected directly from Shanghai Bright Hoistan pasture, preserved in sterile containers, and aseptically transported to the laboratory at Shanghai Jiao Tong University under refrigerated conditions at 4°C within 2 h.

HHP treatment

The fresh milk samples (50 mL) were aseptically transferred to sterile polyamide/chlorinated polypropylene (PA/CPP) bags, heat-sealed, and then processed in an HHP unit (S-IL-100-350-09-W HP Food Processors, Stansted fluid Power Ltd., UK). The single-cycle HHP treatments were carried out at 200, 400, or 600 MPa for 5 min. The multiple-cycle HHP treatments consisted of two 2.5-min cycles at 200, 400, or 600 MPa. The temperature was maintained at 25°C, the pressure increase rate was 8.3 MPa/min, and the depressurization time was <4 s. After the HHP treatments, all the samples were analyzed immediately. The untreated fresh milk samples were used as the controls.

UV-C treatment

UV-C treatments were conducted using an UV-C equipment (UVH30DC3, 30 W, Shanghai Guoda UV Equipment Co., Ltd, Shanghai, China) at 254 nm and a radiation dose of 11.8 W/m^2.[14] Fifty milliliters of fresh cow milk sample was taken in an acrylonitrile butadiene styrene (ABS) plastic dish (190 mm length, 130 mm width, and 120 mm height). The UV-C treatment time was set at 5 min. Before each test, the equipment was cleaned and sanitized by ethyl alcohol. To avoid extensive sample evaporation, a ventilator was installed on the back part of the apparatus for dissipating heat, and the temperature recorded inside the chamber was below 25°C.

GC-MS

The volatile compounds were determined in triplicate by GC-MS as described previously.^[7] Briefly, samples were extracted from milk using the SPME technique that was performed using an auto-sampler (GC Sampler 80, Agilent, Santa Clara, CA, USA). A milk sample (10 mL), 2 μL of the internal standard solution (0.1 mg/mL 2,4,6-trimethylpyridine in methanol), and 3.7 g of sodium chloride were placed in an annealed 20-mL brown glass vial with screw cap. The SPME fiber used was 50/30 μm Divinylbenzene/Carboxen (DVB/CAR) on Polydimethylsiloxane (PDMS) (Supelco, Bellefonte, PA, USA). The samples were incubated for 20 min at 50°C with 500 rpm vibration, and the fiber was exposed to the vial head space for 30 min at 50°C with 250 rpm vibration to allow the adsorption of volatiles. The adsorbed volatiles were desorbed in the GC injector port in the split-less mode at 260°C for 1 min. Finally, the split valve was opened, and the fiber was kept in the injector for 10 min for cleaning. The volatile compounds in the headspace of the instrument were analyzed using an automated GC (model: 7890A, Agilent, Santa Clara, CA, USA). The column used had dimensions of 0.25 μm and 30 m × 0.25 mm (HP-INNOWax, Agilent, Santa Clara, CA, USA), and pure helium was used as the carrier gas at a rate of 1 mL/min. A MS detector (model: 5975C, Agilent, Santa Clara, CA, USA) was used in the electron impact ionization mode at 70 eV in the mass range of 20–350 m/z. The ion source temperature was 230°C, the interface temperature was 280°C, and the detector was operated at 1847 V. The temperature was programmed in two stages. The initial temperature was kept at 40°C for 5 min and then increased at a rate of 5°C/min to 220°C. In the second stage, the temperature was increased at a rate of 20°C/min to 250°C for 7.5 min. The compounds were identified by comparing the Saturn spectra reported on National Institute of Standards and Technology (NIST) 2011 Mass Spectral Library (MainLib: 212,961 spectra and Replib: 30,932 spectra). The main molecular, and qualifier ions were selected for each compound identified. The relative retention times of the detected compounds were also determined by injecting 20 μL of 0.1 mg/mL n-alkane standard solutions (ASTM D2887-01 Calibration Mix, Restek, Pennsylvania, USA) with a split ratio of 1:200. The signals were processed using Agilent MSD Productivity ChemStation Enhanced Data Analysis software (Agilent, Santa Clara, CA, USA).

E-nose

An e-nose system (iNose, Shanghai Rui Fen Scientific Instrument Co. Ltd., Shanghai, China) with a sensor array of 14 sensors (S1-S14), signal collection system, data processing software, and headspace sampling system was used in this study. S1 corresponds to ammonia and amine; S2 corresponds to hydrogen sulfide and sulfide; S3 corresponds to hydrogen; S4 corresponds to ethyl alcohol and organic solvent; S5 corresponds to alcohols, ketones, aldehydes, and aromatic compounds; S6 corresponds to methane, methane, and natural gas; S7 corresponds to combustible gas; S8 corresponds to Volatile Organic Compounds (VOC) (usually used for air pollution detection); S9 corresponds to liquefied petroleum gas, natural gas and coal gas; S10 corresponds to liquefied gas and flammable gas; S11 corresponds to alkane, alcohol, natural gas, and smoke; S12 corresponds to alcohol, organic solvent; S13 corresponds to flue gas and cooking odor; S14 corresponds to methane and gas. Ten milliliters of milk samples was put in a 40 mL e-nose sample vial with a lid and measured by e-nose. The air was refresh into the sensor chambers and flushed over the sensors at a flow rate of 1 L/min. The data were recorded every second by the computer. The experiment lasted for 60 s, which was long enough for the sensors to stabilize, and 600 s was used for the desorption. After each testing, the e-nose was cleaned by clean air for 90 s between two repetitions and for >6 min between three different milk samples. Each analysis was performed in triplicate for each sample.

E-tongue

The e-tongue analyses were performed using a SmarTongue (Shanghai Rui Fen Scientific Instrument Co. Ltd., Shanghai, China), whose sensor array comprised six inert metal sensors (S1-S6 corresponds to Pt, Au, Pd, W, Ti, and Ag electrode, respectively) with varying sensitivities. A conventional Ag/AgCl electrode was used as reference electrode for measuring sensor potential values based on Nernst equation. The e-tongue mainly consisted of a sampler, sensor array and signal processing software. The sample (15 mL) was weighed into a beaker. The e-tongue experimental conditions were: 180 s of acquisition time, 1 s of acquisition period, sample temperature of 25 ± 1°C, pulse interval of 200 mV, sensitivity of 104, and voltage range of –1.0–1.0 V. After each testing, the sensors were rinsed with deionized water for 120 s to prevent the accumulative effect of impurities in electrode surface. Specific variables were extracted from the obtained voltammograms. All the samples were analyzed in triplicates.

Data analysis

The e-nose and e-tongue data were analyzed by principal component analysis (PCA) and linear discriminant analysis (LDA), a part of the software integrated in the e-nose and e-tongue. PCA is a powerful linear combinatorial method and unsupervised pattern recognition technique for classification by e-noses and e-tongues. It reduced the complexity of data-set from the initial n-dimensional space (n sensors) to a few dimensions and explained the variables in the data by a two-dimensional (2D) or three-dimensional (3D) loading plot that elucidated the relationships between the original variables and their effect on the system. Principle components means the eigenvectors of the data matrix and are uncorrelated. The principal components are ordered. The first one explains the highest amount of variance, followed by the second component that explains the highest residual variance, and so on.^[25,28,29] LDA is a probabilistic parametric classification technique which minimizes the variance between categories by data projection from high-dimensional space to low-dimensional space and maximizes the variance within categories.^[22,31,32] Compared to PCA, the LDA method can analyze the distribution of points in the same category and the distance between them. On the descriptions, LDA provides a classification model characterized by linear dependence of the classification scores. The information about the relation of the factors in the model analyzed was obtained by the eigenvalues of LDA. The statistical analysis of variance was performed using the SPSS 19.0 statistical data analytical software (SPSS Inc., Chicago, IL, USA). The significant differences between the means were determined by a least significant difference test at p < 0.05.

Results and discussion

Volatile components in milk

Table 1 shows the volatile profiles of the untreated and treated milk samples obtained by GC-MS. The identified compounds in all the samples were organized into the following chemical groups: aliphatic hydrocarbons, alcohols, aldehydes, ketones, acids, esters, aromatic hydrocarbons, ethers, and sulfur compounds. A total of 44 volatile compounds were detected in the untreated milk samples, accounting for 1107.93 ng/mL of the samples. As reported previously, GC-MS method identified different numbers (11 to 80) of volatiles in different milk products.^[4,10,27] Moreover, aromatic hydrocarbons (372.95 ng/mL), aliphatic hydrocarbons (271.82 ng/mL), and acids (196.2 ng/mL) were the main groups among the identified components in the untreated milk. A total of 45–47 and 53 volatile compounds were identified in the HHP and UV-C-treated samples, respectively. The contents of the volatile compounds in all the treated milk samples decreased in the order: UV-C (1 × 5 min; 2143.79 ng/mL) > 200 MPa (1 × 5 min; 1557.14 ng/mL) > 200 MPa (2 × 2.5 min; 1518.93 ng/mL) > 600 MPa (2 × 2.5 min; 1512.39 ng/mL) > 400 MPa (2 × 2.5 min; 1359.01 ng/mL) > 400 MPa (1 × 5 min; 1229.32 ng/mL) > 600 MPa (1 × 5 min; 977.76 ng/mL). Altogether, it indicated that the HHP and UV-C treatments increased the amount of volatile compounds in the cow milk samples. Our previous study has shown that HHP treatment can improve the flavor in squids and cooked rice.^[33,34]

Table 1. Comparison of the content of volatile compounds (ng mL⁻¹) in milk treated with high hydrostatic pressure and UV-C radiation.

RT		CAS	Control	200 MPa	200 MPa	400 MPa	400 MPa	600 MPa	600 MPa	UV-C
(min)	Compounds	number	(untreated)	(1 × 5 min)	(2 × 2.5 min)	(1 × 5 min)	(2 × 2.5 min)	(1 × 5 min)	(2 × 2.5 min)	(5 min)
Acids
29.098	Hexanoic acid	142-62-1	70.92 ± 4.93^a	130.44 ± 9.08^c	147.24 ± 16.71^d	61.60 ± 6.29^a	92.82 ± 3.33^b	70.56 ± 5.74^a	72.76 ± 8.44^a	68.62 ± 5.60^a
33.549	Octanoic acid	124-07-2	123.07 ± 6.91^a	181.49 ± 12.38^c	179.52 ± 11.29^c	122.88 ± 1.31^a	156.28 ± 12.52^b	146.36 ± 19.59^ab	130.13 ± 17.57^a	125.54 ± 9.19^a
35.874	Nonanoic acid	112-05-0	2.21 ± 1.17^abc	3.15 ± 0.24^c	2.42 ± 0.37^bc	1.26 ± 0.25^a	1.59 ± 0.36^ab	2.75 ± 0.44^c	3.03 ± 0.71^c	9.25 ± 0.73^d
37.817	n-Decanoic acid	334-48-5	—	58.05 ± 4.20^de	50.47 ± 8.17^cd	38.41 ± 0.50^b	47.77 ± 6.42^bc	46.82 ± 4.96^bc	60.92 ± 6.78^e	54.10 ± 4.69^cde
Alcohols
10.344	1-Decanol, 2-hexyl-	2425-77-6	14.44 ± 1.27^b	21.55 ± 1.16^c	16.70 ± 1.72^b	23.00 ± 1.37^cd	14.46 ± 0.94^b	9.51 ± 0.94^a	24.42 ± 2.21^de	26.43 ± 1.99^e
13.937	1-Dodecanol	112-53-8	9.28 ± 0.86^b	10.05 ± 1.28^b	9.33 ± 1.90^b	6.52 ± 0.85^a	10.67 ± 1.50^b	8.42 ± 0.73^ab	8.76 ± 1.56^ab	22.91 ± 2.22^c
14.759	1-Pentanol	71-41-0	18.68 ± 2.41^bcd	14.87 ± 3.70^ab	11.21 ± 0.52^a	13.82 ± 3.22^ab	15.19 ± 4.14^abc	15.75 ± 1.13^abc	21.07 ± 3.34^d	20.18 ± 1.13^cd
16.905	1-Hexanol	111-27-3	—	—	—	—	—	—	—	13.30 ± 0.98^b
18.741	2-octanol	123-96-6	6.31 ± 0.54^c	3.98 ± 0.36^b	11.68 ± 2.08^d	2.01 ± 0.11^a	2.67 ± 0.32^ab	1.79 ± 0.13^a	2.52 ± 0.30^ab	3.96 ± 0.35^b
19.272	1-tetradecanol	112-72-1	—	—	—	—	—	—	—	16.46 ± 1.34^b
20.56	1-hexanol, 2-ethyl-	104-76-7	9.83 ± 0.70^bcd	10.64 ± 4.56^cd	10.23 ± 1.00^bcd	6.87 ± 0.71^ab	6.09 ± 1.18^a	5.41 ± 0.72^a	7.81 ± 2.02^abc	12.07 ± 1.17^d
22.26	1-octanol	111-87-5	11.20 ± 1.06^a	16.28 ± 0.53^c	15.47 ± 2.60^bc	11.16 ± 0.39^a	13.85 ± 0.89^abc	11.52 ± 0.24^a	13.10 ± 2.02^ab	32.89 ± 2.61^d
32.935	n-Tridecan-1-ol	112-70-9	8.67 ± 1.20^b	7.66 ± 2.21^b	8.60 ± 3.23^b	4.47 ± 0.34^a	4.70 ± 0.72^a	4.16 ± 0.13^a	6.79 ± 0.53^ab	5.97 ± 0.81^ab
Aldehydes
2.2	1-heptene	592-76-7	—	—	—	—	—	—	—	8.38 ± 3.16^b
2.268	Hexane, 2,2,5-trimethyl-	3522-94-9	—	7.94 ± 1.76^b	—	8.46 ± 1.35^b	—	—	12.58 ± 1.54^c	10.53 ± 3.77^bc
8.297	Hexanal	66-25-1	11.19 ± 1.12^b	5.95 ± 1.15^a	5.15 ± 0.48^a	6.96 ± 1.76^ab	2.78 ± 1.48^a	6.47 ± 1.89^a	15.77 ± 1.67^c	41.35 ± 6.26^d
11.601	Heptanal	111-71-7	—	—	6.38 ± 0.89^a	—	—	—	—	60.53 ± 9.61^b
17.879	Nonanal	124-19-6	13.81 ± 0.58^a	27.38 ± 6.82^a	17.22 ± 1.99^a	12.90 ± 1.29^a	14.09 ± 0.88^a	8.22 ± 0.15^a	9.09 ± 1.80^a	537.32 ± 59.58^b
20.681	Decanal	112-31-2	7.51 ± 5.62^b	4.30 ± 2.89^ab	2.65 ± 0.20^a	2.41 ± 0.22^a	3.20 ± 0.83^ab	3.49 ± 0.35^ab	3.67 ± 1.28^ab	16.06 ± 1.26^c
Aliphatic Hydrocarbons
2.336	Heptane, 3-methyl-	589-81-1	8.67 ± 1.60^ab	21.14 ± 2.77^c	6.55 ± 1.60^a	20.87 ± 1.07^c	21.63 ± 3.74^c	10.08 ± 2.22^ab	13.93 ± 3.79^b	12.46 ± 5.16^b
2.523	Octane	111-65-9	52.56 ± 2.64^b	137.99 ± 13.61^e	28.53 ± 2.94^a	104.68 ± 6.44^c	123.12 ± 8.89^d	57.65 ± 2.07^b	58.33 ± 5.21^b	38.49 ± 2.13^a
2.88	1-Octene	111-66-0	10.73 ± 0.97^ab	35.35 ± 9.06^c	6.91 ± 5.00^a	38.91 ± 2.02^c	44.37 ± 2.88^c	20.18 ± 5.28^b	16.64 ± 5.76^b	62.80 ± 6.06^d
3.488	Ethane, 1,1-diethoxy-	105-57-7	9.63 ± 0.50^a	13.75 ± 0.48^bc	12.45 ± 4.74^abc	14.87 ± 1.22^bc	16.22 ± 1.60^c	11.74 ± 1.19^ab	11.39 ± 2.22^ab	21.29 ± 1.51^d
4.702	Heptane, 2,2,4,6,6-pentamethyl-	55162-61-3	—	—	—	—	—	—	103.98 ± 10.33^b	111.85 ± 7.22^c
4.949	Nonane, 3-methyl-	5911-04-6	44.49 ± 17.48^a	108.64 ± 5.67^bc	130.25 ± 21.28^cd	134.46 ± 9.75^d	131.22 ± 11.30^cd	61.27 ± 13.97^a	118.28 ± 10.19^bcd	95.81 ± 6.38^b
5.847	Decane	124-18-5	54.39 ± 8.73^a	114.54 ± 7.08^c	92.87 ± 13.49^b	92.23 ± 6.04^b	91.40 ± 7.68^b	47.38 ± 1.42^a	87.28 ± 10.16^b	58.80 ± 4.49^a
7.108	1-decene	872-05-9	4.24 ± 0.37^a	13.35 ± 1.26^c	9.17 ± 0.69^b	10.16 ± 0.53^b	10.11 ± 1.04^b	4.48 ± 0.25^a	4.19 ± 0.79^a	38.08 ± 3.11^d
8.041	Cyclopentane, pentyl-	3741-00-2	4.66 ± 0.35^a	7.84 ± 0.94^a	5.54 ± 1.14^a	15.05 ± 1.29^b	7.97 ± 0.76^a	7.26 ± 1.33^a	19.57 ± 3.67^c	27.07 ± 3.71^d
8.654	Undecane, 5-methyl-	1632-70-8	4.82 ± 0.30^a	6.55 ± 0.53^ab	5.11 ± 1.52^a	10.84 ± 3.07^c	4.20 ± 0.15^a	8.95 ± 2.19^bc	17.24 ± 1.12^d	16.96 ± 1.52^d
8.757	Dodecane	112-40-3	7.57 ± 1.08^a	9.32 ± 1.15^a	6.52 ± 0.85^a	21.48 ± 8.74^c	5.63 ± 2.18^a	9.63 ± 3.63^a	12.21 ± 1.24^ab	18.34 ± 2.06^bc
9.441	Nonadecane	629-92-5	5.94 ± 2.30^a	5.08 ± 1.76^a	5.98 ± 1.35^a	15.04 ± 1.28^b	7.34 ± 4.37^a	6.86 ± 4.01^a	22.79 ± 4.41^c	15.50 ± 1.16^b
9.671	Undecane, 2,6-dimethyl-	17301-23-4	2.86 ± 0.09^a	4.57 ± 1.53^ab	4.31 ± 0.76^ab	6.50 ± 1.71^d	3.25 ± 0.12^a	3.58 ± 0.35^a	6.01 ± 1.07^bc	5.91 ± 0.53^bc
9.783	Decane, 2,5-dimethyl-	17312-50-4	4.04 ± 1.29^a	8.14 ± 2.40^d	6.35 ± 1.28^abc	5.30 ± 0.59^ab	6.34 ± 2.15^abc	6.08 ± 0.32^abc	7.06 ± 0.60^bc	5.87 ± 0.20^abc
10.94	Undecane, 3-methyl-	1002-43-3	25.14 ± 3.15^b	37.63 ± 2.97^c	25.82 ± 3.25^b	34.83 ± 1.29^c	22.54 ± 2.11^b	14.58 ± 0.49^a	38.16 ± 5.30^c	27.64 ± 1.88^b
11.849	D-Limonene	5989-27-5	6.64 ± 0.25^bc	6.60 ± 2.31^bc	12.22 ± 1.32^d	5.94 ± 0.54^abc	5.44 ± 0.69^ab	4.45 ± 0.83^a	6.62 ± 1.03^bc	7.75 ± 0.43^c
12.059	Dodecane	16661-24-8	13.72 ± 0.87^bc	19.17 ± 2.99^de	10.64 ± 2.73^a	16.25 ± 1.14^cd	10.91 ± 0.62^b	7.34 ± 0.69^a	20.99 ± 2.53^e	16.65 ± 1.33^cd
13.107	1-decyne	764-93-2	7.32 ± 0.39^ab	11.21 ± 0.89^bc	13.19 ± 5.60^c	7.03 ± 1.38^ab	9.54 ± 2.60^abc	5.13 ± 0.71^a	9.19 ± 0.92^abc	19.45 ± 1.88^d
17.147	Tridecane, 3-methyl-	6418-41-3	4.40 ± 0.29^c	5.84 ± 0.98^d	4.08 ± 0.22^bc	6.43 ± 0.90^d	3.01 ± 0.61^ab	2.51 ± 0.25^a	8.07 ± 1.28^e	5.71 ± 0.25^d
Aromatic Hydrocarbons
6.931	Toluene	108-88-3	307.01 ± 31.09^c	271.84 ± 16.78^bc	452.63 ± 39.17^e	165.65 ± 8.50^a	254.79 ± 21.22^b	159.68 ± 19.60^a	353.65 ± 39.31^d	198.58 ± 16.20^a
9.543	Ethylbenzene	100-41-4	3.13 ± 0.58^a	4.10 ± 0.44^a	5.72 ± 1.06^b	7.63 ± 0.29^c	11.24 ± 1.27^d	3.22 ± 1.14^a	2.76 ± 0.40^a	7.69 ± 0.45^c
33.451	Phenol, 3-methyl-	108-39-4	—	—	—	—	—	—	—	4.07 ± 0.25^b
37.672	Phenol, 2,4-bis(1,1-dimethylethyl)-	96-76-4	62.81 ± 5.02^d	17.00 ± 4.88^ab	14.09 ± 3.08^a	20.26 ± 3.88^ab	17.86 ± 3.82^ab	27.82 ± 0.66^c	21.17 ± 2.69^b	31.96 ± 2.46^c
Esters
5.445	Butanoic acid, methyl ester	623-42-7	15.51 ± 3.77^ab	23.12 ± 6.01^c	10.14 ± 1.85^a	13.52 ± 2.87^a	11.26 ± 0.75^a	11.15 ± 4.58^a	20.55 ± 2.99^bc	14.42 ± 0.99^ab
11.748	Hexanoic acid, methyl ester	106-70-7	8.87 ± 0.53^e	7.75 ± 1.50^de	6.13 ± 1.36^bcd	4.62 ± 0.74^abc	4.34 ± 1.33^ab	4.23 ± 0.35^a	4.75 ± 0.46^abc	6.34 ± 0.90^cd
13.463	Hexanoic acid, ethyl ester	123-66-0	5.95 ± 0.29^ab	9.04 ± 2.66^c	6.59 ± 0.79^abc	5.63 ± 0.45^ab	5.75 ± 1.59^ab	4.46 ± 1.27^a	7.71 ± 1.76^bc	28.29 ± 2.21^d
19.055	Octanoic acid, ethyl ester	106-32-1	5.46 ± 1.00^a	9.39 ± 1.25^cde	8.31 ± 1.74^bc	7.08 ± 1.16^ab	11.53 ± 0.99^e	11.03 ± 0.71^de	9.60 ± 1.93^cde	8.92 ± 0.13^bcd
Ethers
1.869	Ethyl ether	60-29-7	75.99 ± 6.84^b	83.59 ± 5.52^b	66.24 ± 11.73^ab	66.40 ± 10.26^ab	72.96 ± 8.92^ab	68.73 ± 5.86^ab	56.16 ± 7.85^a	69.50 ± 15.38^ab
3.371	Ethyl acetate	141-78-6	3.16 ± 0.59^a	6.82 ± 0.92^b	7.17 ± 2.96^b	9.37 ± 1.13^bc	10.40 ± 0.66^c	7.98 ± 1.81^bc	6.73 ± 2.01^b	16.58 ± 1.50^d
Ketones
2.64	Acetone	67-64-1	24.31 ± 7.62^a	32.03 ± 1.43^ab	38.80 ± 7.69^b	27.31 ± 5.65^a	27.20 ± 4.63^a	29.89 ± 3.04^ab	37.99 ± 3.01^b	27.23 ± 1.48^a
3.58	2-Butanone	78-93-3	10.77 ± 1.55^ab	13.88 ± 0.40^b	14.67 ± 4.89^b	13.11 ± 1.33^b	14.67 ± 1.35^b	12.07 ± 1.31^b	8.00 ± 0.52^a	27.62 ± 2.43^c
16.606	Cyclopentanone, 2-methyl-	1120-72-5	—	—	—	—	—	—	—	5.46 ± 0.38^b
35.589	2H-Pyran-2-one, Tetrahydro-6-pentyl-	542-28-9	5.23 ± 0.43^abc	7.28 ± 0.88^de	6.74 ± 1.95^cde	4.44 ± 0.21^ab	5.86 ± 0.56^bcd	6.40 ± 0.95^cd	8.19 ± 0.68^e	3.99 ± 0.14^a
Sulfur compounds
29.763	Dimethyl sulfone	67-71-0	0.81 ± 0.08^abc	0.83 ± 0.07^abc	0.92 ± 0.06^c	0.71 ± 0.08^a	0.76 ± 0.05^ab	0.71 ± 0.07^a	0.75 ± 0.05^ab	0.88 ± 0.09^bc

RT: retention time (min).

Values were expressed as average ± SD (n = 3).

For the HHP-treated samples, the three main volatile compounds were aromatic hydrocarbons, aliphatic hydrocarbons, and acids. Hydrocarbons are the representative class of volatiles compounds. They can be formed by lipid autoxidation processes or by the decomposition of carotenoids and have a very limited influence on foodstuffs because of their high perception threshold.^[7] Acids are responsible for the lipolyzed flavor in milk that derived from lipolysis, proteolysis or the fermentation of lactose. Not only acids are the crucial aromas themselves in milk, they also played great role as precursors of other compounds, such as methyl ketones, alcohols, aldehydes, and esters.^[7] Both the percentages of aliphatic hydrocarbons (25.4–45.6%) and acids (18.2–27.2%, apart from 600 MPa [1 × 5 min]) were higher in the HHP-treated samples than in the untreated samples (24.5 and 17.7%, respectively); however, the percentages of aromatic hydrocarbons showed an opposite trend. Among all the HHP treatments, one 5-min cycle at 400 MPa showed the highest content of aliphatic hydrocarbons (45.6%) and the lowest content of aromatic hydrocarbons (15.7%). Samples treated by two 2.5-min cycles at 200 MPa showed the lowest content of aliphatic hydrocarbons (25.4%) and the highest content of aromatic hydrocarbons (31.1%). The highest (27.3%) and lowest (17.6%) acid contents were observed on samples after one 5-cycle and two 2.5-min cycles at 600 MPa, respectively. Moreover, the milk treated at 200 MPa (1 × 5 min) showed a higher total amount of aldehydes, alcohols, esters, and ethers than those treated at other pressures. Our previous studies also found that HHP (at 200, 400, and 600 MPa for 10 min) affected the flavor of squid samples as a higher pressure correlated with a higher concentration of aliphatic hydrocarbons.^[33] Previous research observed the same phenomenon, an increase in the aldehyde concentration after the HHP treatments (482, 586, and 620 MPa).^[10] Ultra-high pressure homogenization (UHPH)-treated milk at 200 MPa (30 and 40°C) did not affect the contents of hexanoic acid, acetone, nonanal, and 2-butanone compared to the fresh milk.^[35] These changes in the volatile compounds were dependent on the origins of fresh milk, treatment conditions, and the mechanisms involved in the formation of volatiles under HHP and UHPH treatments.^[27,35,36] In the UV-C-treated milk samples, the amounts of nine volatiles decreased in the order: aldehydes > aliphatic hydrocarbons > acids > aromatic hydrocarbons > alcohols > ethers > ketones > esters > sulfur compounds. Among the 53 identified volatile compounds, the major five volatile compounds were nonanal (537.32 ng/mL milk), toluene (198.58 ng/mL milk), octanoic acid (125.54 ng/mL milk), 2,2,4,6,6-pentamethylheptane (111.85 ng/mL milk), and 3-methylnonane (95.81 ng/mL milk). The concentrations of the aldehydes in the UV-C-treated samples were 14.8–35.6 times higher than those in the HHP-treated samples, which can be attributed to different oxygen availability and treatment conditions. Aldehydes have low olfactory perception level, and are the common oxidation products in milk resulting from the auto-oxidation of unsaturated fatty acids and spontaneous decomposition of hydroperoxides promoted by heat or light.^[10] Moreover, the amounts of alcohols, aliphatic hydrocarbons, esters, and ketones were higher in the UV-C-treated samples than in the HHP-treated samples. In particular, only five compounds (1-hexanol, 1-tetradecanol, 1-heptene, 3-methylphenol, and 2-methylcyclopentanone) were characterized in the UV-C-treated samples. Ester compounds are the important contributions to food aromas. At ambient temperatures, esters with low numbers of carbon atoms are highly volatile with the 10 times lower perception thresholds than their alcohol precursors.^[7] Alcohols are derived from many metabolic pathways, and are important volatile compounds contributing to an alcoholic, fruity, winey, sweet, and harsh note in dairy products.^[7] Ketones are the principal flavor compounds with pungent aroma in dairy products.^[7] Sulfur-containing molecules are responsible for the “cooked” off-flavor developed during milk processing.^[4] No difference was observed in relation to dimethyl sulfide in all the milk samples.

The internal factors affecting the amount of volatile compounds were the species, diet, breed, sex, and age of cows.^[27,36] The milk volatiles under different treatments and storage conditions commonly resulted from four different pathways: (1) lipid oxidation and degradation; (2) interactions between the lipids and amino acids or proteins; (3) Maillard reaction of sugars with amino acids or peptides; and (4) vitamin degradation.^[3,5,8] However, the actual mechanism of the formation of volatiles under high pressure or UV-C radiation is not fully understood and should be further investigated in the future.

Analysis of e-nose signals

An e-nose consists of an array of metal oxide sensors and has capacity of monitoring volatiles and odors in dairy products.^[21,24,29] In this study, the 14 sensors of e-nose showed different responses. Figure 1 shows the results of PCA and LDA analyses based on e-nose. Principal component 1 (PC1) accounted for 94.5% of the variance, whereas principal component 2 (PC2) accounted for the remaining 3.7% of the variance in the data matrix, indicating that the difference in various samples was mainly reflected by PC1 (Fig. 1a). Therefore, the two principal components basically represent the main distribution characteristics of the samples. Samples 3, 4, and 5 could be well separated from the other samples, and the sample information of 1, 2, 6, 7, and 8 was similar (Fig. 1). In particular, a significant discrimination was observed between sample 4 and sample 3 or 5 based on the differences in their general flavor profile of e-nose (Fig. 1b). This indicated that the e-nose could distinguish the differences among the milk flavors. A discrimination index (DI) was used to characterize the degree of difference in the samples in the e-nose analysis: the higher the value of DI, the more the differences among different samples. Figure 1 shows that the DI was 90.3%, indicating that the PCA and LDA of the overall classification were satisfactory.

Figure 1. Principal component analysis A and linear discriminant analysis B of volatile compounds of (1) untreated; (2) UV-C-treated; and (3) high hydrostatic pressure treated (200 MPa [2 × 2.5 min]; 4: 400 MPa [2 × 2.5 min]; 5: 600 MPa [2 × 2.5 min]; 6: 200 MPa [1 × 5 min]; 7: 400 MPa [1 × 5 min]; 8: 600 MPa [1 × 5min]) milk samples by e-nose.

Analysis of e-tongue signals

E-tongue analyses have been used to evaluate the acid, sweet, bitter, salty, and astringent tastes in foods and pharmaceuticals.^[28] In this study, the classification results of the flavor compounds in milk were analyzed using an e-tongue equipped with five sensors. Figure 2 shows that DI was 99.2%, indicating that the PCA and LDA analyses of the overall classification were satisfactory. Figure 2a shows that 52.5 and 29.2% of the total variation were explained by PC 1 and PC 2, respectively. They represented 81.7% of the total variance and almost reflected the overall fingerprint information of the milk samples. A significant discrimination was observed between sample 3 and the rest of the samples. The PCA could also distinguish eight milk samples. Moreover, sample 3 significantly differed from the rest of the samples. The sample information of 1 and 2 (or 7 and 8) were very similar, and unclear discriminations were observed among samples 4, 5, and 6 (Fig. 2b). Compared to PCA, the samples occupied different areas of the figure; no significant overlap was observed. Hence, LDA was effective in discriminating the samples. E-tongue and e-nose was also effective in evaluating different types of milk and other food stuffs.^[37] The e-nose and e-tongue are working cooperatively to offer complementary information about milk after different treatments.

Figure 2. Principal component analysis A and linear discriminant analysis B of (1) untreated; (2) UV-C-treated; and (3) high hydrostatic pressure treated (200 MPa [2 × 2.5 min]; 4: 400 MPa [2 × 2.5 min]; 5: 600 MPa [2 × 2.5 min]; 6: 200 MPa [1 × 5 min]; 7: 400 MPa [1 × 5 min]; 8: 600 MPa [1 × 5 min]) milk samples by e-tongue.

Conclusions

In this study, GC-MS, e-noses, and e-tongues were applied to detect the quality changes of fresh cow milks processed by UV-C radiation, single-cycle, and multiple-cycle HHP treatments. The variety and content of volatile compounds were increased after HHP and UV-C treatments. The percentages of aliphatic hydrocarbons and acids were higher in most HHP-treated samples than in the untreated samples. Among all the HHP treatments, one 5-min cycle at 400 MPa lead the highest amount of aliphatic hydrocarbons and the lowest amount of aromatic hydrocarbons. Two 2.5-min cycles at 200 MPa resulted in the lowest amount of aliphatic hydrocarbons and the highest amount of aromatic hydrocarbons. One 5-min cycle and two 2.5-min cycles at 600 MPa produced the highest and lowest amounts of acids, respectively. The amounts of aldehydes, alcohols, aliphatic hydrocarbons, esters, and ketones were higher in the UV-C-treated samples than that in the HHP-treated samples. In particular, only five compounds (1-hexanol, 1-tetradecanol, 1-heptene, 3-methylphenol, and 2-methylcyclopentanone) were characterized in the UV-C-treated samples. The simultaneous utilization of e-nose and e-tongue achieved a satisfactory classification of the milk samples evaluated in this study. E-nose and e-tongue analysis showed that the PCA and LDA analyses of classification were satisfactory. E-nose analysis indicated that multiple-cycle HHP treatments at 200, 400, and 600 MPa caused the most significant differences among all the treatments. In e-tongue analysis, LDA was more effective in discriminating the samples than PCA. The characteristic of control and UV-C-treated samples were very similar in e-tongue analysis. Also the e-tongue feature of single-cycle HHP treatments at 400 and 600 MPa were similar. This study showed that HHP and UV-C treatments have potential to accelerate the flavor formation and increase the amount of flavor compounds in milk.

Funding

This work was supported by the Shanghai Minhang District Commission of Science and Technology (2013MH088) and the State Key Laboratory of Dairy Biotechnology, Bright Dairy and Food Co. Ltd. (SKLDB2013-02).

Additional information

Funding

This work was supported by the Shanghai Minhang District Commission of Science and Technology (2013MH088) and the State Key Laboratory of Dairy Biotechnology, Bright Dairy and Food Co. Ltd. (SKLDB2013-02).