Formation and Alterations of the Potentially Harmful Maillard Reaction Products during the Production and Storage of Brown Fermented Milk

To improve the quality and safety of brown fermented milk (BFM), the formation and alterations of potentially harmful Maillard reaction products (MRPs), including 3-deoxyglucosone (3-DG), methylglyoxal (MGO), 5-(hydroxymethyl)-2-furfural (HMF), acrylamide and flavour components were investigated during the browning, fermentation and commercial storage. MRPs were shown to be produced mainly during the browning stage. The levels of different substances varied during the fermentation and commercial storage stage. The proportion and type of carboxylic acids in the flavour components significantly increased during the fermentation stage. Browning index of milk during the browning stage was shown to be positively associated with the 3-DG (Pearson’s r = 0.9632), MGO (Pearson’s r = 0.9915), HMF (Pearson’s r = 0.9772), and acrylamide (Pearson’s r = 0.7910) levels and the total percentage of the flavour components from four different categories (Pearson’s r = 0.7407). Changes in physicochemical properties of BFM during production not only contribute to predict the formation of potentially unhealthy MRPs, but also Lactobacillus species used for the fermentation should be carefully selected to improve the quality of this product.


Introduction
Brown fermented milk (BFM) is a fermented beverage made from the skimmed milk powder and reducing sugar, which is produced using Maillard browning and Lactobacillus casei fermentation [1]. BFM has both effects on human health and a unique colour and flavour, and is, therefore, popular with consumers in China, Japan, Southeast Asia, Europe and other regions [1]. Lactobacillus casei has been widely used in other dairy products such as cheese and ice cream [2,3]. Studies have shown that Lactobacillus casei have positive effect on consumer's health, such as modulates immune function and improves constipation [4,5].
Colour and flavour of many food items, such as bread, coffee, chocolate and cakes, are based on the Maillard reaction during food manufacturing [6,7]. The browning process, which takes advantage of the Maillard reactions between the amino acids and reducing sugars in the reconstituted milk, is crucial for the formation of the BFM colour and flavour. However, the Maillard reaction has been reported to generate a variety of potentially unhealthy Maillard reaction products (MRPs), such as α-dicarbonyl compounds (α-DCs) [8,9], 5-(hydroxymethyl)-2-furfural (HMF) [10] and acrylamide [11,12]. Additionally, Lactobacillus fermentation produces various volatile flavour components, such as carbonyl compounds and organic acids in yogurt [13]. However, little is known about the potentially harmful MRPs, flavour constituents and colour components generated during the manufacturing and storage of BFM. Therefore, it is necessary to investigate the formation and changes of the MRPs during the processing of BMF.
α-DCs are highly reactive intermediates in the Maillard reaction, which can convert amino acid and protein modifications to advanced glycation end products (AGEs). AGEs have been shown to reduce the nutritional quality of foods and to be involved in the development of several diseases, such as neurodegenerative diseases, diabetes and others [14][15][16]. HMF is generated through the Amadori rearrangement in the Maillard reaction or by the direct dehydration of sugars, and this compound was demonstrated to have carcinogenic effects on animals and human cells [17,18]. Furthermore, acrylamide is a neurotoxic, genotoxic and potentially carcinogenic substance, generated in the Maillard reaction between asparagine and the reducing sugars during thermal processing [19].
Microorganisms may affect the MRPs during fermentation [20,21]. Methylglyoxal (MGO), glyoxal and diacetyl were reported to be produced during the Lactobacillus growth [22]. Lactobacillus-based fermentation leads to the reduction in acrylamide formation in mixed rye bread [23]. Additionally, yeasts were shown to reduce the formation of HMF and acrylamide in foods [20,24], and the relationships between different MRPs during the storage of commodities have been determined, including the conversion of 3-deoxyglucosone (3-DG) to HMF in honey or beer [9,25].
Therefore, the aim of this study was to investigate the formation and alterations of 3-DG, MGO, HMF, acrylamide and flavour components during the manufacturing and commercial storage of BFM, in order to help improve the quality and safety of BFM and to assist in developing a regulatory strategy for this food item.

Formation and Alterations of the MRPs in BFM
The formation of 3-DG, MGO, HMF and acrylamide during the browning, fermentation and storage of BFM is presented in Table 1. During the browning, both of 3-DG and MGO levels significantly increased with time, and 3-DG concentration reached a maximum of 82.13 ± 0.64 mg·L −1 at 180 min after the initiation of the reaction, showing that these compounds are produced during the BFM browning. Additionally, 3-DG content was approximately 20-fold higher compared with that of MGO, indicating that the α-DCs with C6 skeleton are formed at higher levels during browning. Our results are in agreement with the reports examining the browning of dairy products, biscuits, and other foods [26][27][28]. HMF levels increased from 0.16 ± 0.03 to 0.53 ± 0.02 mg·L −1 with time, which supports previous studies analysing biscuit baking [29]. However, we observed a decreased rate of acrylamide formation during the BFM browning. Previous studies demonstrated that acrylamide is rapidly formed during the processing at high temperatures accompanied by low moisture, such as frying or baking [30][31][32][33], which indicates that the relatively low temperatures and high moisture content may be primarily responsible for low acrylamide formation rates.
During fermentation, we observed a decrease in 3-DG concentration with time, which was shown to be significantly different. Similar results were reported by [34], who demonstrated that 3-DG does not generate browning colour products, such as melanoidins, but instead preferentially generates the cleavage products under acidic conditions. Therefore, 3-DG participates in complex reactions and its concentration decreased during BMF fermentation. However, MGO and HMF levels increase with the fermentation time, while that of acrylamide does not change significantly. MGO was reported to be formed during Lactobacillus fermentation [22], and it may be produced in the Maillard reaction between the reducing sugars and amino compounds during BFM fermentation [27,35,36]. HMF levels increase due the acidic conditions induced by Lactobacillus fermentation, which accelerate the conversion of lactose into glucose and galactose, and lead to the rapid formation of HMF through Maillard reactions [12]. An additional source of HMF during fermentation is the dehydration and cyclisation of 3-DG formed during the browning stage [34]. These results suggest that the Lactobacillus fermentation plays an important role in the alterations in potentially harmful MRPs during BFM processing.  During the commercial BFM storage (4 • C, 21 days), 3-DG levels increased during the first five days and decreased afterwards, which was comparable to its previously reported changes in a probiotic drink, but its levels in other dairy products were shown to be lower [8,27]. MGO contents increased gradually for 10 days, and remained at the same levels afterwards. Therefore, α-DC levels were shown to be unstable at the early stages of the commercial BFM storage, while stabilising later. No changes in the HMF levels were observed, which was consistent with the levels determined in the infant milk stored at 20 • C, but not in that stored at 37 • C, in which the HMF levels were shown to increase [37]. Therefore, low storage temperatures may inhibit the formation of HMF in BFM. Acrylamide contents in the stored BFM were below the LOQ, mainly due to its low levels formed during the browning and fermentation, and the dilution effect during the final blending, indicating that this compound does not represent an important health risk factor for the BFM use.

Formation and Alterations in the Flavour Components of BFM
Total levels of flavour components were identified during the browning (180 min) and fermentation (72 h) of BFM. Browning stage flavours components mainly included carboxylic acids, followed by aldehydes, alcohols and ketones ( Table 2). The proportion of these compounds in the total flavour components gradually increased with the browning time ( Figure 1A). Flavour components such as acetaldehyde, 2H-pyran-2-one, 2-furanmethanol and furfural are produced during the browning stage and they provide the characteristic flavour of a product [6,38]. The number of flavour components identified in the fermentation stage was higher than that in the browning stage, of which the carboxylic acid species were particularly numerous, such as acetic and butanoic acids. The total proportion of carboxylic acids, aldehydes, alcohols, and ketones gradually increased during fermentation ( Figure 1B). These results suggest that the carboxylic acids, such as acetic acid and butanoic acid, represent the most important flavour components in BFM, which was previously demonstrated in a study analysing fermented yogurt [13]. Moreover, the characteristic flavour components formed through the Maillard reactions were shown to exist during the fermentation process as well. The number of flavour components identified in the fermentation stage was higher than that in the browning stage, of which the carboxylic acid species were particularly numerous, such as acetic and butanoic acids. The total proportion of carboxylic acids, aldehydes, alcohols, and ketones gradually increased during fermentation ( Figure 1B). These results suggest that the carboxylic acids, such as acetic acid and butanoic acid, represent the most important flavour components in BFM, which was previously demonstrated in a study analysing fermented yogurt [13]. Moreover, the characteristic flavour components formed through the Maillard reactions were shown to exist during the fermentation process as well.

BFM Colour Alterations
The values of the colour parameters obtained during BFM browning and fermentation are presented in Table 3 (a*, b*, L*, ΔE* and BI). During browning stage, a* value increased from −2.16 ± 0.15 to 4.19 ± 0.06, while b* increased from 10.32 ± 0.08 to 16.65 ± 0.05 with the browning time, indicating that reconstituted milk lost greenness and blueness and became more red and yellow, respectively, during browning. The L* value, indicating the lightness of the sample, decreased from 88.79 ± 0.2 to 80.41 ± 0.04 with the browning time, indicating that the samples became darker in colour. Due to these results, the overall ΔE* values and BIs increased as well. The browning of the milk is primarily due to the Maillard browning of lysine and reducing sugars in the browned milk [39]. No significant changes in the colour of BFM were observed during the fermentation stage and commercial storage, most likely because of the lack of formation of coloured compounds.

BFM Colour Alterations
The values of the colour parameters obtained during BFM browning and fermentation are presented in Table 3 (a*, b*, L*, ∆E* and BI). During browning stage, a* value increased from −2.16 ± 0.15 to 4.19 ± 0.06, while b* increased from 10.32 ± 0.08 to 16.65 ± 0.05 with the browning time, indicating that reconstituted milk lost greenness and blueness and became more red and yellow, respectively, during browning. The L* value, indicating the lightness of the sample, decreased from 88.79 ± 0.2 to 80.41 ± 0.04 with the browning time, indicating that the samples became darker in colour. Due to these results, the overall ∆E* values and BIs increased as well. The browning of the milk is primarily due to the Maillard browning of lysine and reducing sugars in the browned milk [39]. No significant changes in the colour of BFM were observed during the fermentation stage and commercial storage, most likely because of the lack of formation of coloured compounds.

Relationships between BIs, MRP Levels and Flavour Component Levels during the Browning of BFM
Colour is an important indicator of browning, and it is usually used to determine the finalisation of the browning process. Maillard reactions are directly involved in the browning of BFM. BIs, MRP and flavour component proportions change with the temperature during Maillard reaction, as shown in Figure 2A-C.
together with that between the BIs and the total percentage of flavour components from four described categories (Pearson's r = 0.7407), indicating a strong correlation (|r| > 0.95) between the BIs and 3-DG, MGO and HMF levels during the browning of BFM. However, the correlations observed between the BIs and acrylamide and flavour components were weaker, indicating that the degree of browning can directly reflect the levels of potentially harmful MRPs during BFM browning. This may help improve the monitoring of MRP formation during the BFM manufacturing. Previous studies support the conclusion that the MRP levels correlate with the degree of browning in foods [6,40]. However, no correlation has been observed between MRP levels and BI changes during fermentation and storage of BFM, further confirming that the colour formed during the browning does not change during fermentation and storage.
The results of these MRPs, flavour and colour changes during BFM manufacturing and commercial storage provide a fundamental basis for improving the quality and safety of BFM, ensuring a stable, sustainable and healthy development of the dairy industry. Positive correlations between the BIs and 3-DG levels (Pearson's r = 0.9632), MGO (Pearson's r = 0.9915), HMF (Pearson's r = 0.9772) and acrylamide levels (Pearson's r = 0.7910) were observed, together with that between the BIs and the total percentage of flavour components from four described categories (Pearson's r = 0.7407), indicating a strong correlation (|r| > 0.95) between the BIs and 3-DG, MGO and HMF levels during the browning of BFM. However, the correlations observed between the BIs and acrylamide and flavour components were weaker, indicating that the degree of browning can directly reflect the levels of potentially harmful MRPs during BFM browning. This may help improve the monitoring of MRP formation during the BFM manufacturing. Previous studies support the conclusion that the MRP levels correlate with the degree of browning in foods [6,40]. However, no correlation has been observed between MRP levels and BI changes during fermentation and storage of BFM, further confirming that the colour formed during the browning does not change during fermentation and storage.
The results of these MRPs, flavour and colour changes during BFM manufacturing and commercial storage provide a fundamental basis for improving the quality and safety of BFM, ensuring a stable, sustainable and healthy development of the dairy industry.

Browning Stage
Skimmed milk powder was reconstituted to skim milk samples (120 g·kg −1 ) in water at 40 • C, and then glucose (40 g·kg −1 ) was added. The mixture was gently stirred for 30 min and heated to 65 • C. The sample was mixed homogenously at 20 MPa and browned at 95 • C for 180 min, which was followed by decreasing the temperature to 37 • C with cold water.

Fermentation Stage
The browned sample was inoculated with the L. casei 01 strain at 2.5 × 10 6 CFU mL −1 , and incubated at 37 • C until titration acidity reached more than 200 (approximately 72 h). Afterward, the fermented curd was broken and cooled rapidly to 4 • C. Finally, the fermented milk, syrup and flavouring were mixed and homogenised at 25 MPa and stored in a refrigerator at 4-7 • C.
BFM preparation trials were performed in triplicate, and the mean values were reported.

Quantification of 3-DG and MGO
Preparation of 3-DG and MGO was performed as previously described [41], with minor modifications. Briefly, 1 mL of sample and 50 µL of OPD (60 mg·mL −1 ) were added to a capped brown bottle. The mixture was incubated in water at 60 • C and stirred at 50 rpm. Following a 30-min incubation, the mixture was purified and concentrated using preconditioned Oasis HLB SPE cartridges. The samples were centrifuged at 10,000× g for 5 min and filtered through 0.22 µm nylon for LC/MS/MS analysis.
The derivatised and purified samples were analysed using an Agilent 1200 Series HPLC system coupled with Agilent 6410 triple-quadrupole MS (Agilent, Waldbronn, Germany) with electrospray ion source in positive ion mode and MRM mode. Nitrogen was used as the nebuliser (pressure, 40 psi), drying (flow rate, 10 L min −1 at 350 • C) and collision gas. The capillary voltage was 4 kV, while the