Abstract
Lactobacillus paracasei MCC1849 (LP) has the potential to modulate immune function. The purpose of this study was to investigate the impact of flavor compounds on changes in sensory characteristics during refrigerated storage of a functional LP-fermented dairy beverage. A sample containing 1 × 108 cells/mL viable LP stored at 10 °C for 0, 1, or 2 weeks was evaluated by a trained panel. The scores for ‘odor’, ‘afterflavor of fermented odor’, ‘fermented odor’, and ‘acidity’ increased significantly. Three types of gas chromatography-mass spectrometry for volatiles in the solvent extract and headspace gas, and also for hydrophilic compounds were performed. The compounds related to the change in four sensory attributes were examined by partial least squares regression analysis. A non-stored sample with a flavor composed of 13 compounds related to the attributes replicated the fermented odor characteristics of the 2-weeks-stored sample.
Introduction
There are many traditional fermented foods and beverages such as cereals, vegetables, soybeans, meat, fish, and milk products (i.e., yogurt and cheese) throughout the world (Tamang et al., 2016). Lactic acid bacteria (LAB) have been widely used for dairy products and have also been reported as functional microorganisms with immunomodulatory properties (Taverniti and Guglielmetti, 2011).
Recently, LAB with high functionality have been used in fermented milk products. For example, yogurt fermented with functional LAB has been developed. Intake of yogurt fermented with Lactobacillus delbrueckii ssp. bulgaricus OLL1073R-1 can augment natural killer cell activity and reduces the risk of catching the common cold and flu in the elderly (Makino et al., 2010). While viable LAB are used for fermented foods, some non-viable LAB have also been applied for foods because of their immunomodulatory properties. Many experiments have been conducted to compare the effects of live versus killed cells of the same probiotic strain, and the use of non-viable microbial cells or cell components was reported to influence host's immune system (Taverniti and Guglielmetti, 2011).
Lactobacillus paracasei MCC1849 (LP) was selected from many strains by Morinaga Milk Industry Co., Ltd. because of its immunostimulatory effects (Morinaga Milk Industry Co., Ltd., 2014; 2015). The oral administration of live and heat-killed LP ameliorated influenza symptoms in mice (unpublished data). In addition, the effects of non-viable LP on immune function in the elderly were investigated (Maruyama et al., 2015). LP has the potential to improve resistance to the common cold infections and maintain a desirable mood state in healthy young adults (i). Orally administered heat-killed LP enhances antigen-specific IgA production and likely affects follicular helper T cell differentiation in Peyer's patches (ii).
Non-viable LP do not ferment during storage, making it highly applicable to foods because the taste and aroma of products are not affected. In fact, it has been used in various foods such as supplements, beverages, desserts, soups, and bread in Japan. On the other hand, live non-pathogenic microorganisms such as yeast or LAB are expected to improve the microbial balance, particularly in the gastrointestinal tract. Viable LP is also expected to function in intestinal regulation in addition to its immunostimulatory effects. However, the use of viable LP in products has encountered some challenges. For example, the flavor of products was altered during storage because of product fermentation by LP. To develop a functional fermented food using LP, it is important to reveal changes in sensory characteristics and flavor compounds. If key compounds involved in sensory changes can be clarified, it may be possible to develop products with greater storage stability, by controlling the taste and aroma caused by these key compounds.
Several papers have reported changes in volatile compounds and sensory characteristics during refrigerated storage of LAB-fermented milk (yogurt) and the correlation between them (Kwak, 1995; Güler, 2007; Saint-Eve et al., 2008; Li et al., 2017). However, in the present research using a specific functional strain (LP), it is necessary to clarify in detail the changes in characteristic flavor and odor compounds due to LP fermentation, and to identify the compounds affecting the taste and aroma, leading directly and effectively to the development of products with good storage quality. We previously reported changes in odor compounds of an LP-fermented dairy beverage during two weeks of refrigerated storage (Akiyama et al., 2018).
The aim of this study was to identify flavor compounds affecting the sensory characteristics of an LP-fermented dairy beverage during refrigerated storage. Changes in sensory characteristics and flavor compounds during refrigerated storage were investigated. Flavor compounds impacting sensory characteristics were identified using partial least squares regression (PLSR) analysis, and then a model flavor consisting of these compounds was prepared. Finally, sensory evaluation of the sample containing the model flavor, which reflected the changes in the taste and aroma during storage, was carried out to confirm the flavor compounds responsible for changes in sensory characteristics during storage.
Materials and Methods
Sample preparation Sample preparation was as described by Akiyama et al. (2018). In practice, a fermented dairy beverage containing 1 × 108 cells/mL viable LP (Morinaga Milk Industry Co., Ltd.) with 0.8% milk solids-not-fat (MSNF) and 0.1% milk fat was prepared. The number of viable LP was set to 1 × 1010 cells in one commercial product (100 mL), which was the level established to modulate immune function (Morinaga Milk Industry Co., Ltd., 2014, 2015; Maruyama et al., 2015; i; ii). High-fructose starch syrup containing sugar NF55S (12.70% w/w; Showa Sangyo Co., Ltd., Tokyo, Japan), skim milk (0.89% w/w; Morinaga Milk Industry Co., Ltd.), unsalted butter (0.17% w/w; Morinaga Milk Industry Co., Ltd.), maltodextrin MAX1000 (0.09% w/w; Matsutani Chemical Industry Co., Ltd., Hyogo, Japan), citric acid (0.16% w/w; San-Ei Gen F.F.I., Inc., Osaka, Japan), monosodium fumarate (0.02% w/w; San-Ei Gen F.F.I., Inc.), soybean polysaccharide SM-1200 (0.15% w/w; San-Ei Gen F.F.I., Inc.), and high-intensity sweetener SANSWEET® SA5050 (0.002% w/w; San-Ei Gen F.F.I., Inc.) were each dissolved in water, and then blended together. After heating to 60 °C, the mixture was homogenized at 14 MPa with a homogenizer (type H3-1D INV, 200 L/H; Sanmaru Machinery Co., Ltd., Shizuoka, Japan), and then batch-pasteurized at 80 °C for 10 min. Viable LP was cultured in medium mainly containing sugar, yeast extract, and salt. Bacterial cells were harvested by centrifugation, washed using sterile water, and frozen (frozen viable LP, Morinaga Milk Industry Co., Ltd.). After the mixture was cooled to 10 °C, the frozen viable LP was added to the mix; then, it was aseptically filled into 2-L polyethylene terephthalate (PET) bottles, and stored at 10 °C for 0, 1, and 2 weeks. For GC-MS analysis, the same 0-week sample was divided to prepare two series of stored samples.
The same samples were used for the present GC-MS study and the previous GC-O study (Akiyama et al., 2018). Viable LP counts, pH, and lactic-acid acidity were measured during the two weeks of refrigerated storage (Akiyama et al., 2018).
Isolation of volatiles by solvent extraction (SE) An aliquot (200 g) of the 0-, 1-, or 2-weeks-stored sample was transferred to a 500-mL conical flask, 4-octanol (2 µg/mL in methanol, 1 mL) and 3-heptanol (10 µg/mL in methanol, 2 mL) were added as the internal standard, and dichloromethane (200 mL) was added slowly. The mixture was stirred gently with a magnetic stirrer for 1 h at room temperature, then separated using a separating funnel and dried over anhydrous sodium sulfate. The extract was concentrated in a rotary evaporator to a final volume of about 1 mL. Semi-quantification was performed using relative peak areas of each compound and internal standards.
SPME device and sampling condition for headspace (HS) volatiles A 2.0-g sample was placed in a 20-mL headspace glass vial sealed tightly with a screw cap and PTFE/silicone septum. A manual SPME holder containing a PDMS/DVB fiber (Supelco Inc., Bellefonte, USA) was inserted into the vial and exposed for 30 min at room temperature. The fiber was then transferred to the injector of the gas chromatograph and desorbed for 10 min. Thermal desorption was carried out in the splitless mode at a desorption temperature of 250 °C. The SPME fiber was left in the injection port for re-conditioning during the complete run of the GC before being exposed to the headspace of the next sample. All samples were run in triplicate. The comparison among samples was conducted with mean peak area (PA) of each compounds.
Gas chromatography-mass spectrometry (GC-MS) analysis of SE and HS volatiles GC-MS analysis of the extract was performed in triplicate on a 5973 mass selective detector (Agilent Technologies Inc., Santa Clara, USA) with fused silica capillary column DB-WAX (60 m × 0.25 mm, 0.25-µm film thickness; Agilent Technologies Inc.). The flow of the helium carrier gas was 1.6 mL/min. The oven temperature was programmed to an initial 50 °C for 2 min, then increased at 3 °C /min to 220 °C, and held at 220 °C for 75 min. The injection port was maintained at 250 °C. The inlet was operated in the split mode (split rate 20:1) and 5-µL sample was injected for SE volatiles. For HS volatiles, inlet was operated in the splitless mode.
GC-MS-after-derivatization (AD) analysis The sample was stored at −80 °C. The sample (400 µL) in a 2-mL Eppendorf tube was extracted with 1 000 µL MeOH/H2O/CHCl3 (2.5/1/1, v/v/v). Sixty microliters of 0.2 mg/mL ribitol, used as an internal standard, was added to the mixture, mixed with 5-mm diameter zirconia beads using a vortex mixer, suspended in a ball-mill (20 Hz, 1 min, room temperature) and then sonicated (5 min).
The sample was centrifuged at 16 000 g, for 3 min at 4 °C, and then 800 µ L of the supernatant was transferred to a 1.5-mL Eppendorf tube. Water (400 µL), purified using a Millipore Milli-Q system (Bedford, USA), was added and the sample was vortexed. Following centrifugation (16 000 g, for 3 min at 4 °C), 800 µL of supernatant was transferred to another 1.5-mL Eppendorf and capped. The cap was subsequently pierced and the extract was evaporated to remove methanol in a centrifuge vacuum concentrator at room temperature for approximately 2 h. After evaporation, the extract was freeze-dried in a glass bottle at room temperature overnight.
For derivatization, 100 µL of methoxyamine hydrochloride in pyridine (20 mg/mL) was added to the sample, and the mixture was incubated in a Thermomixer comfort (Eppendorf, Tokyo, Japan) at 30 °C for 90 min to induce the methoxyation reaction. A second derivatizing agent, 50 µL of N-methyl-N-(trimethylsilyl)trifluoroacetamide, was added and the mixture was incubated at 37 °C for 30 min to induce the silylation reaction. The samples were then used for GC-MS analysis. One microliter of sample was injected in split mode (25:1, v/v). For measurement, samples were analyzed by a GC-MS system (QP2100 Ultra system; Shimadzu, Kyoto, Japan) in conjunction with a silica capillary column (30 m × 0.25 mm, fused with a CP-SIL 8 CB-fused film [thickness, 0.25 µm]; Agilent Technologies Inc.) equipped with an AOC-20i/AOC-20s autosampler/injector (Shimadzu). Helium was used as the carrier gas at a flow rate of 1 mL/min. The oven temperature was programmed to an initial 80 °C for 2 min, then increased at 15 °C/min to 330 °C, and held at 330 °C for 6 min. The injection port was maintained at 230 °C. All samples were run in triplicate. The comparison among samples was conducted with mean relative peak area (RPA) normalized to the ribitol peak (included as the internal standard) of each compounds.
Identification of volatile compounds and standard compounds Volatile compounds were identified by comparing their mass spectra and linear retention indices using C6–C28 n-alkanes with those of standard compounds. The standard compounds for identification of volatile compounds were purchased from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan) or Sigma-Aldrich Japan Co., LLC (Tokyo, Japan). The preparation of 2,3-dihydro- 3,5-dihydroxy-6-methyl-4H-pyran-4-one (DDMP) was conducted as described in the literature (Kim and Baltes, 1996).
Identification of hydrophilic compounds Hydrophilic compounds were identified by comparing their mass spectra and linear retention indices using C8–C40 n-alkanes with those in libraries (the NIST library and an in-house library prepared from authentic chemical standards).
Quantification of 13 compounds affecting sensory characteristics Thirteen flavor compounds (2,3-butanedione, 2-pentanone, 2-heptanone, 3-hydroxy-2-butanone, 1-hydroxy- 2-propanone, 2-furanmethanol, DDMP, 5-dodecanolide, acetic acid, octanoic acid, benzoic acid, palmitic acid, and lactic acid) identified as greatly affecting sensory characteristics by PLSR analysis were quantified using an internal standard method. Calibration curves were constructed for the 13 compounds by SE and GC-MS analysis of the 0-week LP-fermented dairy beverage to which known amounts of standards and internal standard were added. Quantitative values of 13 compounds in both 0- and 2-week (-stored) samples were calculated from the calibration curves of each compounds and semi-quantitative values in Table 1. The composition of the 13 compounds in model flavor for sensory evaluation was calculated by the difference between the 0-week sample and the 2-week sample. GC-MS analysis was carried out in triplicate.
Table 1.
Volatile compounds and concentrations detected by SE-GC-MS analysis of LP-fermented dairy beverages stored for 2 weeks at 10 °C.
|
|
0 weeksa |
1 week |
2 weeks |
Change from 0 weeks to 2 weeks |
| No. |
Compound |
Concentration ± SDb (µg/100 g) |
TK-HSD testc |
Concentration ± SD (µg/100 g) |
TK-HSD test |
Concentration ± SD (µg/100 g) |
TK-HSD test |
| 1 |
2,3-Butanedione |
0.53 ± 0.03 |
a |
3.34 ± 0.14 |
b |
9.02 ± 0.58 |
c |
Increased |
| 2 |
2-Pentanone |
0.16 ± 0.01 |
a |
0.23 ± 0.02 |
b |
0.29 ± 0.03 |
c |
Increased |
| 3 |
2-Heptanone |
0.05 ± 0.00 |
a |
0.07 ± 0.00 |
a |
0.37 ± 0.05 |
b |
Increased |
| 4 |
3-Hydroxy-2-butanone |
1.46 ± 0.10 |
a |
19.16 ± 1.12 |
b |
58.55 ± 4.38 |
c |
Increased |
| 5 |
1-Hydroxy-2-propanone |
2.25 ± 0.10 |
a |
3.43 ± 0.19 |
b |
5.06 ± 0.47 |
c |
Increased |
| 6 |
2-Furancarboxaldehyde |
3.89 ± 0.11 |
a |
4.02 ± 0.08 |
a |
3.54 ± 0.25 |
b |
Decreased |
| 7 |
Acetic acid |
1.19 ± 0.11 |
a |
2.03 ± 0.19 |
b |
3.79 ± 0.30 |
c |
Increased |
| 8 |
Butyric acid |
8.08 ± 0.20 |
a |
8.66 ± 0.50 |
a |
8.76 ± 0.57 |
a |
- |
| 9 |
2-Furanmethanol |
1.41 ± 0.06 |
a |
1.84 ± 0.05 |
b |
1.97 ± 0.22 |
b |
Increased |
| 10 |
Hexanoic acid |
16.39 ± 0.24 |
a |
16.49 ± 0.29 |
a |
16.53 ± 0.45 |
a |
- |
| 11 |
1-(2-Furyl)-2-hydroxyethanone |
3.99 ± 0.16 |
a |
4.93 ± 0.08 |
b |
5.54 ± 0.26 |
c |
Increased |
| 12 |
Octanoic acid |
14.41 ± 0.29 |
a |
14.69 ± 0.36 |
a |
16.03 ± 0.62 |
b |
Increased |
| 13 |
5-Decanolide |
0.28 ± 0.03 |
a |
0.32 ± 0.02 |
a |
0.52 ± 0.09 |
b |
Increased |
| 14 |
2,3-Dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (DDMP) |
0.16 ± 0.01 |
a |
0.27 ± 0.02 |
b |
0.25 ± 0.03 |
b |
Increased |
| 15 |
Decanoic acid |
11.68 ± 0.63 |
a |
11.25 ± 0.97 |
a |
12.86 ± 1.66 |
a |
- |
| 16 |
9-Decenoic acid |
3.23 ± 0.14 |
a |
3.10 ± 0.22 |
a |
3.41 ± 0.36 |
a |
- |
| 17 |
Lactic acid |
2.36 ± 0.84 |
a |
27.77 ± 2.35 |
b |
94.11 ± 12.80 |
c |
Increased |
| 18 |
5-Dodecanolide |
0.21 ± 0.02 |
a |
0.30 ± 0.02 |
a |
0.79 ± 0.13 |
b |
Increased |
| 19 |
5-(Hydroxymethyl)-2-furancarboxaldehyde |
55.03 ± 2.35 |
a |
57.24 ± 3.40 |
a |
54.41 ± 4.04 |
a |
- |
| 20 |
Benzoic acid |
22.68 ± 5.69 |
a |
131.02 ± 3.30 |
b |
133.95 ± 6.26 |
b |
Increased |
| 21 |
Lauric acid |
4.19 ± 0.34 |
ab |
3.75 ± 0.36 |
a |
4.78 ± 0.88 |
b |
- |
| 22 |
Myristic acid |
3.15 ± 0.59 |
ab |
2.95 ± 0.27 |
a |
3.89 ± 0.74 |
b |
- |
| 23 |
Palmitic acid |
7.10 ± 2.46 |
a |
7.23 ± 0.70 |
a |
9.81 ± 1.42 |
b |
Increased |
a 0 weeks, 1 week, and 2 weeks: 0-weeks, 1-week, and 2-weeks stored samples, respectively.
b Concentration ± SD: Semi-quantitative values calculated by relative peak areas of each compound and internal standard. The data is mean ± standard deviation (0 weeks,
n = 3; 1 week and 2 weeks,
n = 6).
c TK-HSD test: Tukey-Kramer's honestly significant difference test. There are significant differences between mean scores in the same line followed by different letters (
p < 0.05).
Sensory evaluation of stored samples Analytical sensory evaluation was carried out for 0-, 1-, and 2-weeks samples to investigate changes in sensory characteristics during refrigerated storage of the LP-fermented dairy beverage.
Thirty-nine assessors (25 males and 14 females who worked at Morinaga Milk Industry Co., Ltd.) who had correctly answered more than 7 out of 10 questions in the first selection, using a discrimination test for 5 basic tastes (Yoshioka, 2010), were selected. In the second selection, 13 assessors (7 males and 6 females) were selected by a series of four triangle tests using 3 kinds of 2-weeks-stored fermented dairy beverages containing 0, 2 × 107, or 1 × 108 cells/mL viable LP at 0 weeks. They showed higher rates of correct answers in the tests. Thirteen assessors developed 8 sensory attributes (odor, fermented odor, impure milk odor, sweetness, acidity, smoothness, aftertaste of sweetness, and afterflavor of fermented odor) by discussing for 90 min. In this evaluation, “odor” means the strength of aroma. The assessors were trained by pre-sensory evaluation using LP fermented dairy beverages that were stored for 0 or 2 weeks, and the intensity of terms was adjusted by discussing for 90 min twice.
The sensory evaluation test was carried out in a private, exclusive room equipped with air conditioning (room temperature 23 °C, humidity 40–45%) and a deodorization device under white light. Each sample was evaluated twice by the trained 13-member panel using a 15-cm scale (score 0: weak-15: strong) for 8 attributes (Compusense five version 4.25; ChemSensMetrix Lab, Tokyo, Japan).
Sensory evaluation of 0-week sample with a model flavor The other sensory evaluation was carried out to confirm if the model flavor, which consisted of the compounds associated with changes in sensory characteristics during refrigerated storage, could simulate the taste and aroma of the 2-weeks sample.
For the sensory evaluation, the model flavor was prepared using the 13 compounds identified as prominently affecting sensory characteristics, based on the quantitative data obtained from the LP-fermented 2-weeks sample. The 13 compounds were mixed as follows: 2,3-butanedione, 36.5 mg; 2-pentanone, 0.220 mg; 2-heptanone, 0.356 mg; 3-hydroxy-2-butanone, 276 mg; 1-hydroxy-2-propanone, 40.7 mg; 2-furanmethanol, 2.82 mg; DDMP, 0.593 mg; 5-dodecanolide 1.37 mg; acetic acid, 35.9 mg; octanoic acid, 1.93 mg; benzoic acid, 206 mg; palmitic acid, 5.59 mg; and lactic acid, 976 mg. The mixture was made up to 100 g with ethanol to yield the 1 000-fold concentrations as determined in the LP-fermented dairy beverages.
The following 3 types of samples were prepared: 0-weeks sample (without the model flavor), 0-weeks sample with 0.1% of the model flavor, and 2-weeks sample. Each sample contained 1 × 108 cells/mL viable LP at 0 weeks. The addition of 0.1% of the model flavor produced the same concentration of the 13 compounds as the 2-weeks beverage. Sensory evaluation was performed as described above.
Statistical analysis For sensory evaluation and GC-MS data, a multiple comparison test using JMP software (version 8; SAS, Cary, USA) and principal component analysis (PCA) using SIMCA-P software (version 12.0; Umetrics, Umeå, Sweden) were conducted. SE-, HS-, and AD-GC-MS data of compounds that increased or decreased significantly were used for PCA.
In addition, the relationship between the flavor compounds and each of the four sensory attributes whose score changed significantly during 2-weeks storage was investigated by PLSR analysis using the SIMCA-P software. As data pretreatment, auto-scaling (transformed so that each variable had a unit variance and zero mean) were used to ensure an unbiased contribution of each X variable to the Y variable. The number of latent factors (PLS components) in the PLS model was determined by computing the optimum number at the balance between R2 and Q2 with the SIMCA-P software algorithm. After fitting the model, R2 and Q2 were calculated as the index of model quality. R2 displays the fraction of the sum of squares for the selected components. Q2 shows the fraction of the total variation of Y that can be predicted by components, as estimated by cross-validation. Model quality was evaluated by two parameters, RMSEE (Root Mean Square Error of the Estimation (the fit) for observations) for the calibration set and RMSEP (Root Mean Square Error of Prediction) for the prediction set. RMSEE and RMSEP are the standard deviation of the calibrated residuals and the predicted residuals (errors), respectively. They are computed as the square root of [S (observed value–predicted value)2/N (number of samples)] (Ochi et al., 2012a; 2012b).
Results and Discussion
Change in sensory characteristics of LP-fermented dairy beverage during refrigerated 2-weeks storage In our previous study (Akiyama et al., 2018), viable LP counts decreased by 55% at 2-weeks storage compared with those at 0 weeks, and pH and the acidity changed from 3.8 to 3.7 and from 0.24 to 0.26 (lactic acid w/w%), respectively, suggesting that the continued LP fermentation had little influence on pH and acidity during 2 weeks of refrigerated storage.
In this study, analytical sensory evaluation of 0-, 1-, and 2-weeks samples was conducted using eight sensory attribute terms to investigate changes in sensory characteristics during refrigerated storage. As shown in Fig. 1, the scores for ‘odor’, ‘fermented odor’, and ‘afterflavor of fermented odor’ significantly (p < 0.05) increased between 0 and 1 week, and the scores for ‘odor’, ‘fermented odor’, ‘afterflavor of fermented odor’, and ‘acidity’ increased between 0 and 2 weeks. On the other hand, characteristics of ‘impure milk odor’, ‘sweetness’, ‘smoothness’, and ‘aftertaste of sweetness’ did not change significantly.
In gas chromatography-olfactometry (GC-O) study of an LP-fermented dairy beverage (Akiyama et al., 2018), total odor intensities of odor compounds analyzed by GC-O became stronger with the storage period. ‘Odor’ change in sensory evaluation was considered to reflect increases in total odor intensities. In addition, fatty-metallic, coconut, and woody odors, which had strong intensities and became stronger during 2-weeks storage in the GC-O study, were considered to impact increases in ‘fermented odor’ and ‘afterflavor of fermented odor’.
Change in flavor compounds of LP-fermented dairy beverage during refrigerated 2-weeks storage In order to comprehensively investigate changes in flavor compounds during refrigerated 2-weeks storage, three types of GC-MS analysis: SE-GC-MS, HS-GC-MS, and AD-GC-MS were carried out.
Twenty-three compounds and the mean concentration of each compound, as detected by SE-GC-MS analysis, are shown in Table 1. Among the 23 compounds in the solvent extract, the concentration of 15 compounds significantly increased (p < 0.05) with storage and only 2-furancarboxaldehyde (Table 1-no. 6) decreased. Among the 15 compounds, the concentration of four compounds, 2,3-butanedione (no. 1), 3-hydroxy-2-butanone (no. 4), lactic acid (no. 17), and benzoic acid (no. 20), increased dramatically by 630%, 1 312%, 1 177%, and 578% in the 1-week sample, respectively. 2,3-Butanedione and 3-hydroxy-2-butanone, which have a buttery odor (Molimard and Spinnler, 1996), and lactic acid are known to contribute most to the flavor properties of yogurt (Rash, 1990; Ott et al., 2000). In addition, benzoic acid has been identified in plain yogurt (iii). In the sensory evaluation, the score for ‘odor’, ‘fermented odor’, and ‘afterflavor of fermented odor’ increased significantly in the 1-week sample. Properties of these four compounds were inferred to relate to ‘odor’, ‘fermented odor’, and ‘afterflavor of fermented odor’. However, 2,3-butanedione was hardly detected in 0- and 1-week samples in GC-O analyses (Akiyama et al., 2018). It was thought to be due to that concentration of 2,3-butanedione in 0- and 1-week sample extracts for GC-O analysis were below detection limit.
The HS-GC-MS analyses revealed 14 compounds (Table 2). 2-Heptanone (Table 2-no. 2), 3-hydroxy-2-butanone (no. 5) and 2-furancarboxaldehyde (no. 10) were also detected by SE-GC-MS analysis. Among the 14 compounds in the HS gas, only 3-hydroxy-2-butanone (no. 5) increased significantly in one week, and dimethyl sulfide (no. 1), 2-heptanone (no. 2), hexyl acetate (no. 4), 3-hydroxy-2-butanone (no. 5), 1-hexanol (no. 6), 2-nonanone (no. 7), and 8-nonen-2-one (no. 9) increased significantly between the first and second week. 2-Heptanone and 2-nonanone, which have a fruity odor (Friedrich and Acree, 1998; Molimard and Spinnler, 1996), increased by 1 057% and 899%, respectively, and 3-hydroxy- 2-butanone, which has a fermented (buttery) odor, also increased by 5 708% in 2 weeks. On the other hand, nonanal (no. 8) and 4-butanolide (no. 12) decreased significantly.
Table 2.
Volatile compounds detected by HS-GC-MS analysis of LP-fermented dairy beverages stored for 2 weeks at 10 °C.
|
|
0 weeksa |
1 week |
2 weeks |
Change from 0 weeks to 2 weeks |
| No. |
Compound |
Mean PA ± SDb |
TK-HSD testc |
Mean PA ± SD |
TK-HSD test |
Mean PA ± SD |
TK-HSD test |
| 1 |
Dimethyl sulfide |
1 306 ± 127 |
a |
1 211 ± 290 |
a |
2 125 ± 583 |
b |
Increased |
| 2 |
2-Heptanone |
69 157 ± 2 950 |
a |
115 601 ± 7 489 |
a |
730 653 ± 404 916 |
b |
Increased |
| 3 |
Limonene |
154 960 ± 8 774 |
a |
164 747 ± 86 994 |
a |
105 426 ± 6 558 |
a |
- |
| 4 |
Hexyl acetate |
4 163 ± 278 |
a |
7 730 ± 2 109 |
a |
26 574 ± 14 937 |
b |
Increased |
| 5 |
3-Hydroxy-2-butanone |
11 864 ± 1 338 |
a |
181 469 ± 13 482 |
b |
677 238 ± 32 432 |
c |
Increased |
| 6 |
1-Hexanol |
0 |
a |
742 ± 232 |
a |
22 110 ± 9 484 |
b |
Increased |
| 7 |
2-Nonanone |
36 636 ± 3 019 |
a |
56 141 ± 5 052 |
a |
329 537 ± 116 455 |
b |
Increased |
| 8 |
Nonanal |
39 344 ± 3 338 |
a |
17 032 ± 7 015 |
b |
11 537 ± 3 935 |
b |
Decreased |
| 9 |
8-Nonen-2-one |
0 |
a |
7 196 ± 921 |
a |
43 571 ± 20 653 |
b |
Increased |
| 10 |
2-Furancarboxaldehyde |
361 386 ± 8 323 |
ab |
398 132 ± 6 353 |
a |
239 122 ± 103 227 |
b |
- |
| 11 |
1-Octanol |
4 836 ± 873 |
a |
9 434 ± 856 |
ab |
12 508 ± 4 218 |
b |
Increased |
| 12 |
4-Butanolide |
45 814 ± 28 478 |
a |
12 007 ± 2 699 |
b |
12 993 ± 10 958 |
b |
Decreased |
| 13 |
2-Furanmethanol |
15 220 ± 1 674 |
a |
19 961 ± 1 443 |
a |
24 914 ± 7 835 |
a |
- |
| 14 |
Hexadecanal |
38 008 ± 13 790 |
ab |
59 591 ± 28 941 |
a |
19 755 ± 16 899 |
b |
- |
a 0 weeks, 1 week, and 2 weeks: 0-week, 1-week, and 2-weeks stored samples, respectively.
b Mean PA ± SD: Mean peak area ± standard deviation (0 weeks,
n = 3; 1 week and 2 weeks,
n = 6).
c TK-HSD test: Tukey-Kramer's honestly significant difference test.
There are significant differences between mean scores in the same line followed by different letters (p < 0.05).
Twenty-two hydrophilic compounds were detected and 17 compounds were identified by AD-GC-MS analysis (Table 3). Lactic acid (Table 3-no. 1), glycine (no. 8), and unknown_4 (no. 19) increased significantly during 2-weeks storage, while 1,3-dihydroxyacetone (no. 6), (E)-butenedioic acid (no. 10), and unknown_1 (no. 12) decreased significantly.
Table 3.
Compounds detected by AD-GC-MD analysis of LP-fermented dairy beverages stored for 2 weeks at 10 °C.
|
|
0 weeksa |
1 week |
2 weeks |
Change from 0 weeks to 2 weeks |
| No. |
Compound |
Mean RPA ± SDb |
TK-HSD testc |
Mean RPA ± SD |
TK-HSD test |
Mean RPA ± SD |
TK-HSD test |
| 1 |
Lactic acid |
0.09 ± 0.01 |
a |
1.53 ± 0.05 |
b |
4.18 ± 0.15 |
c |
Increased |
| 2 |
2-Hydroxyethanoic acid |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
- |
| 3 |
Ethanedioate |
0.08 ± 0.02 |
a |
0.07 ± 0.01 |
a |
0.08 ± 0.01 |
a |
- |
| 4 |
3-Hydroxybutyrate |
0.06 ± 0.01 |
a |
0.06 ± 0.01 |
a |
0.06 ± 0.00 |
a |
- |
| 5 |
Diaminomethanal |
0.12 ± 0.03 |
a |
0.15 ± 0.01 |
a |
0.15 ± 0.02 |
a |
- |
| 6 |
1,3-Dihydroxyacetone |
0.08 ± 0.01 |
a |
0.05 ± 0.00 |
b |
0.01 ± 0.00 |
c |
Decreased |
| 7 |
Phosphoric acid |
1.22 ± 0.08 |
a |
1.26 ± 0.06 |
a |
1.31 ± 0.06 |
a |
- |
| 8 |
Glycine |
0.01 ± 0.00 |
a |
0.02 ± 0.00 |
b |
0.02 ± 0.00 |
b |
Increased |
| 9 |
Butanedioic acid |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
- |
| 10 |
(E)-Butenedioic acid |
0.79 ± 0.04 |
a |
0.60 ± 0.01 |
b |
0.28 ± 0.01 |
c |
Decreased |
| 11 |
2-Hydroxybutanedioic acid |
0.02 ± 0.00 |
a |
0.01 ± 0.00 |
b |
0.02 ± 0.00 |
a |
- |
| 12 |
Unknown_1 |
0.11 ± 0.01 |
a |
0.09 ± 0.01 |
b |
0.09 ± 0.00 |
b |
Decreased |
| 13 |
(3S,4R,5R,6S)-6-Methyltetrahydro-2H-pyran-2,3,4,5-tetraol |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
0.01 ± 0.00 |
a |
- |
| 14 |
Unknown_2 |
0.07 ± 0.00 |
ab |
0.07 ± 0.01 |
a |
0.08 ± 0.00 |
b |
- |
| 15 |
2-Hydroxypropane-1,2,3-tricarboxylic acid |
11.56 ± 0.71 |
a |
10.82 ± 0.37 |
a |
11.26 ± 0.51 |
a |
- |
| 16 |
D-Lyxo-hex-2-ulopyranose |
1.01 ± 0.08 |
a |
0.97 ± 0.02 |
a |
1.04 ± 0.05 |
a |
- |
| 17 |
Unknown_3 |
11.56 ± 1.29 |
a |
11.10 ± 1.01 |
a |
10.62 ± 1.50 |
a |
- |
| 18 |
(1R,2R,3S,4S,5R,6S)-Cyclohexane-1,2,3,4,5,6-hexol |
0.05 ± 0.01 |
a |
0.05 ± 0.01 |
b |
0.05 ± 0.00 |
ab |
- |
| 19 |
Unknown_4 |
0.00 ± 0.00 |
a |
0.02 ± 0.00 |
b |
0.03 ± 0.00 |
c |
Increased |
| 20 |
beta-D-Galacto-hexopyranosyl-(1->4)-beta-D-arabino-hex-2-ulofuranose |
3.20 ± 0.18 |
a |
2.88 ± 0.10 |
b |
3.14 ± 0.14 |
a |
- |
| 21 |
2-(Hydroxymethyl)-6-[4,5,6-trihydroxy-2-(hydroxymethyl) oxan-3-yl]oxyox ane-3,4,5-triol |
3.59 ± 0.29 |
a |
3.16 ± 0.15 |
b |
3.39 ± 0.12 |
ab |
- |
| 22 |
Unknown_5 |
4.83 ± 0.30 |
a |
4.44 ± 0.28 |
a |
4.68 ± 0.18 |
a |
- |
a 0 weeks, 1 week, and 2 weeks: 0-weeks, 1-week, and 2-weeks stored samples, respectively.
b Mean RPA ± SD: Mean relative peak area ± standard deviation (0 weeks,
n = 3; 1 week and 2 weeks,
n = 6).
c TK-HSD test: Tukey-Kramer's honestly significant difference test. There are significant differences between mean scores in the same line followed by different letters (
p < 0.05).
PCA profile using GC-MS data during refrigerated 2-week storage PCA was performed to confirm whether each stored sample could be discriminated using the SE-, HS-, and AD-GC-MS data of compounds that increased or decreased significantly and which compounds contribute largely to the discrimination. As shown in Fig. 2, the constructed PCA plot explained 61.5% of the total variance, comprising 45.1% from the first principal component (PC1) and 16.4% from the second principal component (PC2). The PCA result showed that each stored sample could be discriminated using the SE-, HS-, and AD-GC-MS data, and the data were sufficiently reliable to represent the change in compound profile of the samples during 2-week storage. In addition, each sample could be especially discriminated by the PC1 axis. The profiles of samples moved from left to right along the PC1 dimension with storage.
Table 4 shows the PC1 and PC2 loadings. The compounds with high PC1 loadings largely contributed to the change in compound profile during 2-week storage. Many carbonyl compounds detected by SE-GC-MS such as 2,3-butanedione (Table 4-no. 1), 3-hydroxy-2-butanone (no. 3), 1-hydroxy-2-propanone (no. 8), 2-heptanone (no. 9), 2-pentanone (no. 11), and acids such as lactic acid (no. 5) and acetic acid (no. 6) contributed to the positive PC1 axis. 3-Hydroxy-2-butanone (no. 4) detected by HS-GC-MS also showed high positive PC1 loadings. Four compounds (Table 4-no. 2 lactic acid, no. 7 unknown_4, no. 32 1,3-dihydroxyacetone, and no. 33 (E)-butenedioic acid) detected by AD-GC-MS contributed to the PC1 axis and the other compound (no. 27 glycine) detected by AD-GC-MS contributed to the PC2 axis.
Table 4.
PC loadings by PCA applied to solvent-extraction (SE-) GC-MS, headspace (HS-) GC-MS, and after-derivatization (AD-) GC-MS data.
| No. |
Compound |
PC1 |
PC2 |
| 1 |
2,3-Butanedione |
0.99 |
−0.07 |
| 2 |
Lactic acid (AD)a |
0.99 |
−0.01 |
| 3 |
3-Hydroxy-2-butanone |
0.99 |
−0.03 |
| 4 |
3-Hydroxy-2-butanone (HS)b |
0.98 |
−0.10 |
| 5 |
Lactic acid |
0.98 |
−0.10 |
| 6 |
Acetic acid |
0.98 |
−0.02 |
| 7 |
Unknown_4 (AD) |
0.97 |
0.15 |
| 8 |
1-Hydroxy-2-propanone |
0.96 |
0.10 |
| 9 |
2-Heptanone |
0.94 |
−0.32 |
| 10 |
5-Dodecanolide |
0.92 |
−0.14 |
| 11 |
2-Pentanone |
0.92 |
0.09 |
| 12 |
1-(2-Furyl)-2-hydroxyethanone |
0.92 |
0.34 |
| 13 |
2-Nonanone (HS) |
0.87 |
−0.35 |
| 14 |
8-Nonen-2-one (HS) |
0.85 |
−0.31 |
| 15 |
5-Decanolide |
0.85 |
−0.05 |
| 16 |
1-Hexanol (HS) |
0.83 |
−0.15 |
| 17 |
Octanoic acid |
0.82 |
−0.21 |
| 18 |
2-Heptanone (HS) |
0.79 |
−0.39 |
| 19 |
2-Furanmethanol |
0.77 |
0.31 |
| 20 |
Hexyl acetate (HS) |
0.76 |
−0.36 |
| 21 |
Benzoic acid |
0.73 |
0.66 |
| 22 |
1-Octanol (HS) |
0.71 |
0.32 |
| 23 |
Palmitic acid |
0.69 |
−0.35 |
| 24 |
Dimethyl sulfide (HS) |
0.68 |
−0.17 |
| 25 |
Decanoic acid |
0.47 |
−0.46 |
| 26 |
DDMPc |
0.47 |
0.69 |
| 27 |
Glycine (AD) |
0.44 |
0.80 |
| 28 |
4-Butanolide (HS) |
−0.52 |
−0.51 |
| 29 |
Unknown_1 (AD) |
−0.60 |
−0.58 |
| 30 |
2-Furancarboxaldehyde |
−0.71 |
0.63 |
| 31 |
Nonanal (HS) |
−0.77 |
−0.47 |
| 32 |
1,3-Dihydroxyacetone (AD) |
−0.99 |
−0.01 |
| 33 |
(E)-Butenedioic acid (AD) |
−0.99 |
0.01 |
a (AD): Detected by GC-MS after derivatization.
b (HS): Detected by headspace GC-MS.
The other compounds: Deteted by solvent extracted GC-MS.
c DDMP: 2,3-Dihydro-3,5-dihydroxy-6-methyl-4
H-pyran-4-one.
PLSR analysis of relation between sensory characteristics and flavor compounds To examine the relationship between sensory characteristics and flavor compounds, PLSR analysis was conducted. PLSR models were constructed to determine the relationship between four sensory attribute terms (odor, fermented odor, acidity, afterflavor of fermented odor) that increased significantly during 2-weeks storage and the flavor compounds (SE-16 compounds, HS-10 compounds, AD-6 compounds) detected by GC-MS analyses and increased or decreased significantly. Four sensory attribute terms and these flavor compounds were used as response variables and explanatory variables for the PLSR analysis, respectively. The obtained PLSR models were interpreted using the diagnostic index of model quality (R2 and Q2) denoted goodness of fit and predictability estimated by cross-validation, respectively. Both R2 and Q2 for the four sensory attribute terms were more than 0.9, while RMSEE was less than 0.2 and RMSEP was less than 0.3 (Table 5). These results showed that reliable models for 4 sensory attributes were obtained using flavor-compound data by GC-MS analyses. In order to select important compounds, the importance of X-variables was evaluated using variable importance for projection (VIP) scores (Eriksson et al., 2006). Compounds with a VIP score > 1.0 (Table 6) were judged as important to the modeling of X-terms. Among compounds with high VIP scores, two unknown compounds (AD unknown_1, AD unknown_4) were listed.
Table 5.
Prediction modeling by partial least squares regression (PLSR) analysis.
|
Odor |
Fermented odor |
Acidity |
Afterflavor of fermented odor |
| R2 |
0.990 |
0.993 |
0.990 |
0.992 |
| Q2 |
0.980 |
0.985 |
0.990 |
0.985 |
| RMSEE |
0.172 |
0.102 |
0.064 |
0.077 |
| RMSEP |
0.202 |
0.125 |
0.079 |
0.095 |
R2: Goodness of fit
Q2: Goodness of predictability estimated by cross-validation RMSEE: Root mean square error of the estimation for observations RMSEP: Root mean square error of prediction
Table 6.
Compounds related to sensory attributes of LP-fermented dairy beverage during refrigerated 2-weeks storage and VIP scores obtained by PLS regression analysis.
Reverse engineering of changes in sensory characteristics during 2-week storage using a model flavor A model flavor, composed of 13 compounds that were selected based on the result of PLSR analysis, was developed. In Table 6, the compounds that were detected only in HS-GC-MS or AD-GC-MS analysis were eliminated as candidates for the model flavor because there was no quantitative information available for preparation of the model flavor. In addition, the effects of AD-detected compounds except for lactic acid and (E)-butenedioic acid on sensory characteristics were considered to be minimal because the individual content of the compounds and the changes during 2-week storage were small. Compounds showing decreases during 2-week storage such as 2-furancarboxaldehyde were also eliminated. Compounds with a concentration below the odor threshold, i.e., 5-decanolide (concentration: 0.28–0.52 µg/100 g, threshold (van Gemert, 2011): 10 µg/100 g) were also eliminated from the candidates. Additionally, 1-(2-furyl)-2-hydroxyethanone was eliminated from the candidates because of the lack of a supplier. Thus, 5-dodecanolide, 2-heptanone, 2-furanmethanol, and DDMP with VIP scores ≥ 1.0 in any of 4 sensory attributes were added to the model flavor. In addition, palmitic acid with a VIP score < 0.8 was used as a component of the model flavor because of the quantitative availability. The 13 compounds selected for the model flavor have been detected as fermented odors by GC-MS and GC-O analyses for yogurt in previous reports (Ott et al., 1997; Friedrich and Acree, 1998; Alewijn et al., 2003; Cheng, 2010; Routray and Mishra, 2011; Dan et al., 2017).
In order to determine the blending amount of each compound in the model flavor, and the amount to be added to the beverage sample, the concentration of each compound in the samples, which corresponded to the amount of change during 2-weeks refrigerated storage, were analyzed by a standard addition method. The results are shown in Table 7.
Table 7.
Selected 13 compounds and concentrations used in the model flavor for second sensory evaluation.
| No. |
Compound |
Concentrationa (mg/100 g) |
| 1 |
2,3-Butanedione |
36.5 |
| 2 |
2-Pentanone |
0.220 |
| 3 |
2-Heptanone |
0.356 |
| 4 |
3-Hydroxy-2-butanone |
276 |
| 5 |
1-Hydroxy-2-propanone |
40.7 |
| 6 |
2-Furanmethanol |
2.82 |
| 7 |
DDMPb |
0.593 |
| 8 |
5-Dodecanolide |
1.37 |
| 9 |
Acetic acid |
35.9 |
| 10 |
Octanoic acid |
1.93 |
| 11 |
Benzoic acid |
206 |
| 12 |
Palmitic acid |
5.59 |
| 13 |
Lactic acid |
976 |
a Concentration in the model flavor corresponding to the level of change during 2-week refrigerated storage, calculated by a standard addition method.
b DDMP: 2,3-Dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one
A second sensory evaluation was carried out to confirm if the model flavor could replicate the change in sensory characteristics during 2-week storage. In the first sensory evaluation of changes in sensory characteristics during 2-week refrigerated storage, scores of 4 sensory attributes, ‘odor’, ‘fermented odor’, ‘afterflavor of fermented odor’, and ‘acidity’, changed significantly between the 0- and 2-weeks samples (Fig. 1). However, in the second sensory evaluation using a model flavor, 3 sensory attributes, ‘odor’, ‘fermented odor’, ‘afterflavor of fermented odor’, which relate to odor, changed significantly between the 0- and 2-week samples (p < 0.05) (Fig. 3 - (a)). Thus, the evaluation results of the 2-week sample differed partially between the first and second sensory evaluations. In the first evaluation, the evaluation was carried out with the combination of 0-, 1-, and 2-weeks, samples, whereas in the second, it was evaluated with a combination of 0-weeks, 0-weeks with flavor (0 weeks+), and 2-weeks samples. We suggest that the fact that the flavor-added sample (0 weeks+) was similar to the flavor of the 2-weeks sample had some influence on the evaluation of the 2-weeks sample.
There was a significant (p < 0.05) difference in the sensory scores of ‘odor’ attribute between the 0-weeks sample and the 0-weeks sample with a model flavor (0-weeks+ sample) (Fig. 3 - (b)); however, there was no significant difference between the 2-weeks sample and the 0-weeks+ sample. For ‘fermented odor’ and ‘afterflavor of fermented odor’, the 0-weeks+ sample showed no significant differences from the 0-weeks sample, but showed a tendency to differ from the 0-weeks sample (Fig. 3 - (c) and (d)). In addition, there were no significant differences in both attributes between the 0-weeks+ sample and the 2-weeks sample.
The above result shows the possibility that the model flavor could replicate the sensory changes in 3 odor attributes. However, the model flavor could not mimic the changes in ‘acidity’. Though the model flavor contained lactic acid, the score of ‘acidity’ did not change significantly. By comprehensive GC-MS analyses, many compounds were detected other than the 13 compounds. If the flavor contained more compounds, such as increasing AD compounds and HS compounds, and amounts of the compounds such as (HS-) nonanal, which decreased during 2-weeks refrigerated storage, were adjusted, the score of “fermented odor’, ‘afterflavor of fermented odor’, and ‘acidity’ might show greater changes. Further, other compounds not contained in the model flavor might affect the sensory attributes. Potent odorants found in a GC-O study (Akiyama et al., 2018) may also be candidates.
Conclusion
In this study, comprehensive compounds analysis and sensory evaluation of LP-fermented beverages under refrigerated storage were carried out, and the flavor compounds related to the sensory changes were examined by PLSR analysis. By using the model flavor prepared based on the analysis results, changes in three of four sensory characteristics during storage could be replicated. In conclusion, we were able to confirm some of the compounds that contributed to changes in sensory characteristics during storage. Therefore, it might be possible to develop an LP-fermented dairy beverage with reduced sensory change during refrigerated storage, via the use of specific competitive compounds against the 13 compounds as controlling agent. Furthermore, in order to increase the storage stability of products, a future research subject will be to identify the unknown compounds and investigate the influence of compounds with high VIP besides the 13 compounds, for example, 2-weeks sample with compounds showing decreases during 2-weeks storage such as nonanal will be evaluated compared with 0-weeks sample.
Abbreviations:
LP
Lactobacillus paracasei MCC1849
LAB
lactic acid bacteria
PLSR
partial least squares regression
SE
solvent extraction
HS
headspace
GC-MS-AD
gas chromatography-mass spectrometry after derivatization
GC-O
gas chromatography-olfactometry
DDMP
2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one
PCA
principal component analysis
PC1
first principal component
PC2
second principal component
References
- Akiyama, M., Suzuki, T., Murakami, Y., Onishi, M., Mizota, Y., Miyaji, K., Ikeda, M., and Iwabuchi, H. (2018). Changes in odor compounds of a Lactobacillus-fermented dairy beverage during two weeks of refrigerated storage. Food Sci. Technol. Res., 24, 1129-1137.
- Alewijn, M., Sliwinski, E.L., and Wouters, J.T.M. (2003). A fast and simple method for quantitative determination of fat-derived medium and low-volatile compounds in cheese. Int. Dairy J., 13, 733-741.
- Cheng, H. (2010). Volatile flavor compounds in yogurt: A review. Crit. Rev. Food Sci. Nutr., 50, 938-950.
- Dan, T., Wang, D., Wu, S., Jin, R., Ren, W., and Sun, T. (2017). Profiles of volatile flavor compounds in milk fermented with different proportional combinations of Lactobacillus delbrueckii subsp. bulgaricus and Streptococcus thermophilus. Molecules, 22, 1633-1646.
- Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikström, C., and Wold, S. (2006). Appendix II: Statistical notes. In “Multi-and megavariate data analysis: Part I Basic principles and applications. Second revised and enlarged edition,” ed. by Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikström, C. and Wold. S. Umetrics AB, Umeå, Sweden, pp. 381-399.
- Friedrich, J.E. and Acree, T.E. (1998). Gas chromatography olfactometry (GC/O) of dairy products. Int. Dairy J., 8, 235-241.
- Güler, Z. (2007). Changes in salted yoghurt during storage. Int. J. Food Sci. Technol., 42, 235-245.
- Kim, M.-O. and Baltes, W. (1996). On the role of 2,3-dihydro-3,5- dihydroxy-6-methyl-4(H)-pyran-4-one in the Maillard reaction. J. Agric. Food Chem., 44, 1, 282-289.
- Kwak, H.S. (1995). Effect of volatile flavor compound on yogurt during refrigerated storage. Korean J. Food Sci. Technol., 27, 939-943.
- Li, C., Song, J., Kwok, L., Wang, J., Dong, Y., Yu, H., Hou, Q., Zhang, H., and Chen, Y. (2017). Influence of Lactobacillus plantarum on yogurt fermentation properties and subsequent changes during postfermentation storage. J. Dairy Sci., 100, 2512-2525.
- Makino, S., Ikegami, S., Kume, A., Horiuchi, H., Sasaki, H., and Orii, N. (2010). Reducing the risk of infection in the elderly by dietary intake of yoghurt fermented with Lactobacillus delbrueckii ssp. bulgaricus OLL1073R-1. Brit. J. Nutr., 104, 998-1006.
- Maruyama, M., Abe, R., Shimono, T., Iwabuchi, N., Abe, F., and Xiao, J.-Z. (2015). The effects of non-viable Lactobacillus on immune function in the elderly: a randomized, double-blind, placebo-controlled study. Int. J. Food Sci. and Nutr., 67, 67-73.
- Molimard, P. and Spinnler, H.E. (1996). Review: Compounds involved in the flavor of surface mold-ripened cheeses: origins and properties. J. Dairy Sci., 79, 169-184.
- Morinaga Milk Industry, Co., Ltd. (2014). Novel Lactobacillus and novel Lactobacillus-containing medicine, food, beverage and feed. Japan Tokkyo Koho, JP 5646124, Nov. 14.
- Morinaga Milk Industry, Co., Ltd. (2015). Novel Lactobacillus and novel Lactobacillus-containing medicine, food, beverage and feed. Japanese patent. Published unexamined patent application. JP WO2015/004949, Jan. 15.
- Ochi, H., Naito, H., Iwatsuki, K., Bamba, T., and Fukusaki, E. (2012a). Metabolomics-based component profiling of hard and semi-hard natural cheeses with gas chromatography/time-of-flight-mass spectrometry, and its application to sensory predictive modeling. J. Biosci. Bioeng., 113, 751-758.
- Ochi, H., Bamba, T., Naito, H., Iwatsuki, K., and Fukusaki, E. (2012b) Metabolic fingerprinting of hard and semi-hard natural cheeses using gas chromatography with flame ionization detector for practical sensory prediction modeling. J. Biosci. Bioeng., 114, 506-511.
- Ott, A., Fay, L.B., and Chaintreau, A. (1997). Determination and origin of the aroma impact compounds of yogurt flavor. J. Agric. Food Chem., 45, 850-858.
- Ott, A., Hugi, A., Baumgartner, M., and Chaintreau, A. (2000). Sensory investigation of yogurt flavor perception: Mutual influence of volatiles and acidity. J. Agric. Food Chem., 48, 441-450.
- Rash, K. (1990). Compositional elements affecting flavor of cultured dairy foods. J. Dairy Sci., 73, 3651-3656.
- Routray, W., and Mishra, H.N. (2011). Scientific and technical aspects of yogurt aroma and taste: A review. Compr. Rev. Food Sci. Food Safety, 10, 208-220.
- Saint-Eve, A., Lévy, C., Le Moigne, M., Ducruet, V., and Souchon, I. (2008). Quality changes in yogurt during storage in different packaging materials. Food Chemistry, 110, 285-293.
- Tamang, J.P., Watanabe, K., and Holzapfel, W.H. (2016). Review: Diversity of microorganisms in global fermented foods and beverages. Front. Microbiol., 7, 1-28.
- Taverniti, V. and Guglielmetti, S. (2011). The immunomodulatory properties of probiotic microorganisms beyond their viability (ghost probiotics: proposal of paraprobiotic concept). Genes Nutr., 6, 261-274.
- van Gemert, L.J. (2011). “Odour Thresholds Compilations of Odour Threshold Values in Air, Water and other Media.” Oliemans Punter & Partners BV, Utrecht.
- Yoshioka, Y. (2010). Quality assurance system of Morinaga Milk Industry Co., Ltd. The Soft Drinks Technology, 161, 151-160.
- i) https://doi.org/10.3920/BM2017.0197 (Aug. 13, 2018)
- ii) https://doi.org/10.1371/journal.pone.0199018 (Aug. 13, 2018)
- iii) https://www.vcf-online.nl/VcfHome.cfm (Jan. 8, 2019)
), and 2-week sample (
). Score: 0 (weak)-15 (strong). Tukey-Kramer's HSD tests were performed. Significant (p < 0.05) differences are observed between 0-week and 1-week samples (a), 0-weeks and 2-weeks samples (b), and 1-week and 2-weeks samples (c).
), and 2-weeks sample (
). Score: 0 (weak)-15 (strong). Tukey- Kramer's HSD tests generated the following p values:
