Original papers
Availability of Detection Frequency Method Using Three-Port Gas Chromatography-Olfactometry for Rapid Comparison of Whole Milk Powders
2014 年 20 巻 4 号 p. 809-814
詳細
2014 年 20 巻 4 号 p. 809-814
The purpose of this study was to investigate the common components and characteristic compounds of whole milk powders (WMPs) produced in three different regions of Asia and Oceania. The volatile components were isolated from seven WMPs produced in Japan, New Zealand, and China. Forty-one aroma-active compounds were detected in these volatile components using gas chromatography-olfactometry by the detection frequency method. The binary data converted from the detection frequency values was applied to the principal component analysis (PCA) and hierarchical cluster analysis (HCA). The HCA result showed three clusters corresponding to the production regions. Based on the PCA, approximately 70% of the total variance of WMPs was explained by the PC1 and PC2 scores. The PC1 scores can be used to examine the property of WMPs based on the aroma of dairy products and the PC2 scores indicated the property of the production region based on the scattering plot of each WMP. Based on the PC2 loading values of aroma-active compounds, it was revealed that, depending on the production region, the difference between the whole aromas of different WMPs was caused not by the characteristic compound, but by the balance of aroma impact compounds that commonly occur in WMPs.
Because of its powdered form, whole milk powder (WMP) is easy to handle and transport, and has good storage stability. It is not only consumed in its production region but is also distributed worldwide as a food commodity. Between 2009 and 2012, WMP production increased in Asia and the Oceania region, particularly in China and New Zealand (i). During this period, China was the world's largest consumer of WMP and New Zealand was the world's largest supplier; meanwhile, Japan's WMP production showed a gradual decreasing tendency (ii).
The flavor of WMP is an important factor that determines its deliciousness and quality. In order to control WMP's quality, factors leading to flavor change and formation of off-flavor during the storage period have been studied.
Flavor changes in WMPs with butylated hydroxyanisole/butylated hydroxytoluene and those in stored WMPs were detected by sensory evaluation (Hall and Lingnert, 1984), and their related volatile compounds were analyzed by headspace gas chromatography (GC) (Hall et al., 1985a). Additionally, the correlation between the flavor properties of stored WMP and the aroma compounds was reported (Hall and Lingnert, 1985b). The hexanal contents of dried dairy products such as WMPs were measured using GC after performing simple steam distillation, and the correlation between the peroxide value and the hexanal content was determined (Ulberth and Roubicek, 1995). The volatile compounds of WMPs produced throughout the year were analyzed first by using an electric nose (E-nose) and solid phase micro extraction GC (SPME-GC), and then by applying multivariate analysis to the data. As a result, WMPs manufactured in different seasons were classified based on some guiding components extracted from the data thus obtained (Biolatto et al., 2007). Meanwhile, the flavor and flavor stability of skim milk powder and WMPs were examined by the GC-Olfactometry (GC-O) intensity method (Whetstine and Drake, 2007). Although it is said that the flavor of dairy products differs based on the climate and feed of a region, and each producing region imparts a characteristic flavor to its WMP, so far no report has analyzed WMPs based on these criteria.
Creating GC-O profiles for food samples, such as WMPs produced in different regions, and classifying the samples by multivariate analysis of their profiles are time consuming and effort intensive. The GC-O detection frequency method has many advantages such as repeatability, short analysis time, and reducing the training needs of panelists (Debonneville et al., 2002; Pollien et al., 1997).
The purpose of this study was to identify the common components and characteristic compounds of WMPs produced in three regions of Asia and Oceania: Japan, New Zealand, and China, by applying multivariate analysis to the results of the GC-O detection frequency method.
Materials WMPs produced in Japan (■1 and ■2) were purchased from a domestic market of the country, those produced in New Zealand (◆1 and ◆2) were obtained from a reliable commercial source, and WMPs produced in China (●1, ●2 and ●3) were purchased from one of its domestic markets. All WMPs were packed in aluminum-polyethylene laminated bags sealed and stored at −20°C until use. In conducting sensory analysis, WMPs were used by the best-before date.
Preparing aroma concentrates To separate the volatile components from WMPs, the solvent-assisted flavor evaporation (SAFE) method was used (Engel et al., 1999). The SAFE method was employed in this study because of its usefulness for the careful isolation of volatiles from oil-rich samples. Each WMP (36 g) was dissolved in distilled water, mixed well, and distilled under the following conditions: water bath temperature, 50°C; vacuum pressure, 0.1 mPa; distillation time, 1 h; and the traps were cooled with liquid nitrogen. An internal standard compound (2-octanol, 10 µg) was added to each distillate and extracted thrice with 100 mL of dichloromethane. Each extract was dried and the solvent was reduced to approximately 5 mL using a rotary evaporator under 550 mmHg and subsequently concentrated under a nitrogen stream to almost 100 µL. By adjusting the weight of each WMP for aroma extraction and the final volume of each aroma concentrate, the evaluated samples were prepared under standardized conditions for GC-O.
Gas chromatography-mass spectrometry In order to identify the volatile components of WMPs, an Agilent Model 6890N gas chromatograph coupled to an Agilent Model 5973N series mass selective detector (MSD) was used. A DB-WAX column (60 m × 0.25 mm i.d.) with a film thickness of 0.25 µm was used. The oven temperature was programmed from 80 to 210°C at a rate of 3°C/min. The injector temperature was 250°C, and the carrier gas flow rate (helium) was 1 mL/min. The 1.0 µL injection volume was applied in splitless mode. The MSD conditions were as follows: capillary direct interface temperature, 220°C; ionization voltage, 70 eV (EI); mass range, 33 – 300 amu; and ion source temperature, 150°C.
Three-port GC-O system The three-port GC-O system was comprised of an Agilent Model 6890 GC equipped with a flame ionization detector (FID) and connected to a box heater with three sniffing ports (Fig. 1). Using this instrument, three sniffers can simultaneously sniff odorants separated on a GC column. A DBWAX column (30 m × 0.53 mm i.d.) with a film thickness of 1.00 µm was used. The oven temperature was programmed from 40 to 210°C at a rate of 5°C/min for all runs. The aroma concentrate (2.0 µL) was injected into the injector at 250°C in splitless mode. The most appropriate carrier gas flow rate (helium) was calculated for the entire GC-O system to be 6.3 mL/min. The FID conditions were as follows: temperature, 250°C; hydrogen flow, 40 mL/min; air flow, 450 mL/min; and makeup gas flow, nitrogen, 25 mL/min. Approximately one-twentieth of the column flow was diverted to the FID (Fig. 1, Crosspiece 1). The remaining flow was directed to Crosspiece 2.
Three-port GC-O system.
The flow entering Crosspiece 2 was divided into three equal parts (2 mL/min) toward three thermostated lines held at 300°C. These lines were made of fused silica capillary tubing of identical size (0.59 m, 0.15 mm i.d.). The box heater and the tube connected to the GC were maintained at 280°C. At the end of the tube, a glass sniffing port was attached, which was replaced with a new one.
Panelists The number of panelists for the sniffing experiments was 13, who were recruited from the laboratory of Ogawa & Co., Ltd. The panelists had completed more than 60 hours of a training course in a variety of aspects of sensory analysis (recognition, description and discrimination tests), and had previously experienced other GC-O sessions. To avoid physical and mental fatigue, each panelist participated in only one GC-O session per half-day, and the sniffing time was limited to 60 min per session.
Detection frequency method A total of nine values were gathered per sample in performing three sniffing analyses by three different panelists, who recorded the time when the aroma concentrate was eluted and the aroma description. Detection frequency was summed based on the aroma quality detected at the same time by each panel. For converting the frequency data to binary data when the summed frequency was six or more, the converted binary data was assigned to a value of 1, and the rest was assigned to 0. The matrix thus converted was used in the next step.
Multivariate analysis Principal component analysis (PCA) and hierarchical cluster analysis (HCA) were performed using SPSS ver. 11.0J.
Seven samples of aroma concentrates, two produced in Japan (samples ■1 and ■2), two in New Zealand (samples ◆1 and ◆2) and three in China (samples ●1, ●2, and ●3) reproduced well the original WMP's sensory properties. Samples ■1 and ■2 had a mild dairy flavor with a pleasant milk-like aroma. Samples ◆1 and ◆2 had a sweet and cheesy dairy flavor with a slightly green or hay-like nuance. Samples ●1, ●2, and ●3 had a buttery, oily, and cheesy flavor, with a fatty and fermented nuance. These aroma concentrates were used for the GC-O and gas chromatography-mass spectrometry (GC-MS) analyses.
In the present study, the GC-O detection frequency method was employed using the improved instrument, allowing three panelists to simultaneously sniff (Fig. 1). The suitable number of panelists for the detection frequency method has been reported as eight to ten people (Pollien et al., 1997). Debonneville et al. (2002) concluded that the multisniffing device was also suitable for the GC-surface nasal impact frequency (SNIF) method, another detection frequency method. The number of panelists who identified the odor by GC-O was counted, and the odor descriptions were recorded. Forty-one aroma-active compounds were perceived from all seven WMPs. Among these, 20 compounds were identified by comparing their retention indices, mass spectra, and odor descriptions with those of authentic compounds. Sixteen compounds were tentatively identified by comparing their retention indices and odor descriptions with those of the authentic compounds. Each compound shown in Table 1 has been reported from dairy products. The compounds that showed high detection frequency in this study corresponded partly to those of Drake et al. (2007), which were found in American WMP.
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/20_809_4.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09) |
When a GC-O dilution method, such as CHARM (Combined Hedonic Aroma Response Measurement) and AEDA (Aroma Extract Dilution Analysis), is used in the research of key aroma compounds, a dilution series of an extract is prepared per one sample, usually between 8 and 12 in total, and GC-O is applied to each diluted extract. Also, two or more rounds of sniffing are required for reliable detection. On the other hand, when the GC-O detection frequency method is used, GC-O is carried out one time per one sample without dilution. In the present study, using the three-port GC-O system permitted simultaneous data acquisition of three GC-O data, resulting in a shorter GC run time than the conventional dilution method.
Pollien et al. (1997) described that when the NIF method is used as the detection frequency method, a frequency difference of 30% would generally indicate a significant concentration difference. This observation is consistent with the fact that the Weber ratio of a typical discrimination threshold of odor is 30% (Kirk et al., 2007). Therefore, it was considered that if the panelists could detect over 30% of the aroma concentrates, the components could be detected. Additionally, it was thought that the compound could be recognized as the apparent impact compound for the whole aroma of the WMP if panelists could recognize over 60% of the aroma concentrates. Based on this idea, among the detection data shown in Table 1, detection values of 6 or more were converted to 1 and detection values of 5 or less were converted to 0. Multivariate analysis was applied to this matrix.
Based on the results of PCA, the total variables were aggregated to 53.3% of the first principal component (PC1) and 16.5% of the second principal component (PC2) (Table 2). Because the PC1 and PC2 scores explained approximately 70% of the total variance of WMPs and the eigenvalue of PC3 was below 1 (Table 2), this PCA result was employed to examine the relations between the aroma impact compounds and the WMP's sensory properties. Although the PC1 scores explained approximately 53% of the total variance, WMPs were not classified based on individual production regions. Therefore, the PC1 scores can be used to explain the property of WMPs based on the aroma of dairy products. In contrast, the PC2 scores indicated the property of the production region based on the scattering plot of each WMP (Fig. 2, left). The HCA result with Ward's method also showed three clusters of each production region (Fig. 3).
| PC | Eigenvalue | Propotion of PC (%) | Cumulative variance (%) |
|---|---|---|---|
| 1 | 3.7 | 53.3 | 53.3 |
| 2 | 1.2 | 16.5 | 69.8 |
| 3 | 0.7 | 10.5 | 80.4 |
| 4 | 0.6 | 8.2 | 88.6 |
| 5 | 0.4 | 6.3 | 94.9 |
| 6 | 0.2 | 2.8 | 97.8 |
| 7 | 0.2 | 2.2 | 100.0 |
Principle Component Analysis (PCA) of Whole Milk Powder Aroma. The left diagram shows a scattered plot of PCA scores, reflecting the distribution of each sample. The right diagram shows a scattered plot of loadings, reflecting the distributions in the left diagram. The underlined numbers in the right diagram are the same as those in Table 1. Letter “a” means compounds No. 1 – 6, 8, 11, 16, 19, 21, 25, 29 – 31, 33, 35, 38, 39, and 41 in Table 1; “b” means compounds No. 22 and 26 in Table 1, and “c” means compounds No. 18 and 23 in Table 1.
Dendrogram of Seven WMP Samples.
Seventeen compounds (Compounds 7, 9, 10, 12, 13, 14, 15, 17, 20, 24, 27, 28, 32, 34, 36, 37, and 40), for which the absolute values of PC1 or PC2 loading were more than 0.99, were shown to have a significant effect on the plot of each WMP (Table 3). Among these, compounds 7, 9, 14, 17, 20, 24, 28, 32, and 36, for which the PC1 loading values were more than 0.99, are well-known aroma compounds of dairy products. In particular, 2-acetyl-1-pyrroline (9) is known to have a popcorn-like and nutty sweet odor, and 5-decanolide (32) is known to have a creamy and milky odor. Each compound is well known as an aroma compound of milk. Methional (14) is a sulfur-containing compound with a characteristic cooked potato or meaty odor, and is known to contribute to the whole aroma of WMPs. Based on the PC2 loading values, these effective 17 compounds can be divided into three groups: those with 1 or more, those from 1 to −1, and those with −1 or less. Group 1 comprised compounds 9, 10, 12, 27, 32, and 36; Group 2 compounds 14, 17, and 20; and Group 3 compounds 7, 13, 15, 24, 28, 34, 37, and 40 (Fig. 2, right).
| loading | ||
|---|---|---|
| No. compounds | PCI | PC2 |
| 7 1-octen-3-one | 1.0692 | − 1.5845 |
| 9 2-acetyl-1-pyrroline | 2.5642 | 1.7874 |
| 10 unknown | − 0.2888 | 1.2850 |
| 12 (E)-2-octenal | − 0.3698 | 2.1093 |
| 13 1-octen-3-ol | 0.0607 | − 1.1368 |
| 14 methional | 2.1727 | − 0.1159 |
| 15 (E)-2-nonenal | 0.5236 | − 1.6602 |
| 17 butanoic acid | 0.9962 | − 0.5813 |
| 20 3-methylnonan-2,4-dione | 1.3344 | − 0.2840 |
| 24 hexanoic acid | 1.0692 | − 1.5845 |
| 27 4,5-epoxy-(E)-2-decenal | 0.6320 | 1.7530 |
| 28 4-hydroxy-2,5-dimethyl-3(2H)-furanone (furaneol) | 1.7002 | − 1.1948 |
| 32 5-decanolide | 2.5642 | 1.7874 |
| 34 decanoic acid | 0.5236 | − 1.6602 |
| 36 4-dodecanolide | 1.0219 | 2.2328 |
| 37 (Z)-6-dodecen-4-olide | 0.0607 | − 1.1368 |
| 40 dodecanoic acid | 0.0673 | − 1.2282 |
| a | − 0.7613 | 0.2060 |
| b | − 0.3956 | − 0.7048 |
| c | 0.1579 | − 0.7494 |
The numbers above are the same as that in Table 1 and letters “a,” “b,” and “c” shown in Fig. 2.
Samples ■1 and ■2 had a mild dairy flavor with a pleasant milk-like aroma. Each positive position of samples ■1 and ■2 on the PC2 scores was affected by 2-acetyl-1-pyrroline (9), 5-decanolide (32), (E)-2-octenal (12), 4,5-epoxy-(E)-2-decenal (27), 4-dodecanolide (36), and unknown (10) of Group 1. 2-Acetyl1-pyrroline (9) and 5-decanolide (32), the constituent compounds of Group 1, were also considered as compounds showing properties of dairy products based on the PC1 loadings. A fatty, green, and nutty odor that (E)-2-octenal possesses and a metallic odor that 4,5-epoxy-(E)-2-decenal possesses were known as factors indicating butter aroma (Schieberle et al., 1993). 4-Dodecanolide possesses a fruity and peach-like odor; it has also been reported as one of the aroma components of ultrahigh temperature (UHT) milk (Coulibaly and Jeon, 1992). Compound 10 was unknown, but its contribution to the WMP aroma was inferred from the fact that it had a butter-like odor. Each of these compounds was generally recognized as an aroma component of dairy products. Therefore, samples ■1 and ■2 (produced in Japan) were proposed to have a mild dairy aroma without an unpleasant sensation.
Samples ◆1 and ◆2 had a sweet and cheesy dairy flavor with a slight green or hay-like nuance. Their positions on the PC2 scores were affected by the Group-2 compounds. Methional (14), butanoic acid (17), and 3-methylnona-2,4-dione (20), the constituent compounds of Group 2, were also considered as compounds indicating the properties of dairy products based on the PC1 loadings. The green and hay-like odor of 3-methylpentane-2,4-dione (20) contributed to the characteristic grassy odor of samples ◆1 and ◆2. Similarly, the sweet odor with a meaty nuance that methional (14) possesses and the cheesy odor of butanoic acid (17) contributed to the characteristic odor of samples ◆1 and ◆2, respectively. Based on the high PC1 loading, 2-acetyl-1-pyrroline (9), furaneol (28), and 5-decanolide (32) were suggested as contributors of the sweet and milky odor of samples ◆1 and ◆2, even though they were classified to other groups.
Samples ●1, ●2, and ●3 had a buttery, oily, and cheesy flavor with a fatty and fermented nuance. Their positions on the PC2 scores were affected by the Group-3 compounds. 1-Octen-3-one (7), hexanoic acid (24), and furaneol (28), present in Group-3 compounds, showed the features of dairy products based on the PC1 loading. Acids such as hexanoic acid (24), decanoic acid (34), and dodecanoic acid (40), and (E)-2-nonenal (15) possess a fatty, waxy, and sweaty odor. Both 1-octen-3-one (7) and 1-octen-3-ol (13) possess a deteriorated fat odor with a mushroom nuance. Such compounds in Group 3 probably contributed to the odor perceived in samples ●1, ●2, and ●3.
The seven WMPs, produced in three different regions, were classified into each region by the following experimental steps. Step 1: obtain the sniffing data by using the three-port GC-O based on the detection frequency method; step 2: convert the data obtained in step 1 to binary data; and step 3: apply the binary data obtained in step 2 to multivariate analysis. It was shown that using the three-port GC-O for sniffing is useful for shortening the experiment time, and that the multivariate analysis is useful for considering individual characteristics that different types of foods possess.
In conclusion, it was found that, depending on the production region, the difference between the whole aromas of different WMPs was caused not by the characteristic compound, but by the balance of aroma impact compounds that commonly occur in WMPs.
