Kefir grains were purchased from Beinuo Biological Corporation (Shanghai, China). Skim milk powder (3.4 g pro/100 g) were purchased from brightdairy Dairy Co., Ltd. (Shanghai, China). Alcalase (P4860, >2.4 U/g), pepsin (P7000, >250 units/mg powder), pancreatin (P1750, 4 × USP specifications), ACE (from rabbit lung), Hippuryl-histidiyl-leucine (HHL; substrate for ACE) were purchased from the Sigma-Aldrich Chemical Co. (Shanghai, China). Acetonitrile (HPLC grade) and trifluoroacetic acid (TFA, HPLC grade) were obtained from Fisher Scientific Inc. (Hudson, NH, USA). All the other reagents were of analytical grade.

The purchased kefir grains were sub-cultured in skim milk according to the method of Garofalo et al. (2015) with slight modifications. Kefir grains were first inoculated into the sterile skim milk (95°C, 30 min) in the concentration of 12.5% (w/v), and then incubated for 3 months at 28°C. Grains were transferred two times a week.

The diversity of bacterial and fungal in kefir grains were determined. The DNA of bacterial and fungal were extracted by Ezup Column Bacteria Genomic DNA Purification Kit (Sangon Biotech, Shanghai). Full-length of bacterial 16S rRNA genes were amplified with the universal primers 27F (AGAGTTTGATCCTGGCTCAG) and 1492R (TACGGCTACCT- TGTTACGACTT). 5.8S rDNA genes for fungal was amplified with the primers ITS 1 (CTTGGTCA TTTAGAGGAAGTAA) and ITS 4 (TCCTCCGCTTATTGATATGC). A set of 10-base barcodes for every DNA sample was added to the forward and reverse PCR primers. The PCR was run on a mastercycler gradient (Bio-radT100, USA). The total reaction volume was 25 μL, containing 12.5 μL 2 × Taq PCR MasterMix, 1 μL Primer (5 μM), 2 μL template DNA, and 9.5 μL ddH₂O. The amplification program was as follows: 95°C for 5 min; 35 cycles of 95°C for 1 min, 55°C for 1 min, 72°C for 90 s, and with a final extension at 72°C for another10 min. The PCR products were purified using a GeneJET Gel Extraction Kit (Thermo Scientific, USA). 16S rRNA, ITS 1 and ITS 4 genes were sequenced by PacBio Sequel and Illumina Miseq platform, respectively. After that, the SMRT Link (version 7.0) and Illumina Analysis Pipeline (version 2.6) were used to conduct the image analysis, base calling and error estimation of bacterial and fungal raw data.

The milk fermentation process was conducted according to the previous study (Solieri et al., 2015). Briefly, kefir grains were first inoculated in MRS broth for activation at 37°C. Then the microorganisms were rinsed three times using 50 mmol Trise-HCl buffer (pH 6.5) and re-suspended in 11% (w/w) skimmed bovine milk. The number of microorganisms was adjusted to 1 × 10⁸ ~ 6 × 10⁸ CFU/mL. After that, 2% (v/v) of suspension containing kefir grains was inoculated to skimmed bovine milk prepared by ultra-high temperature sterilization technology (UHT). Fermentation was carried out for 24 h at 37°C at 10 rpm. Samples at 6, 12, 18 and 24 h were gathered for the following analysis.

The consumption of casein during the fermentation was determined using a RP-HPLC method with gradient elution (Bonfatti et al., 2008). The samples were performed on a Shimadzu LC-15C HPLC system equipped with a column (ZORBAX SB-C8, 4.6 mm i.d. × 250 mm, 5 μm, Agilent Technologies, USA). Solvent A was 0.1% (v/v) TFA in water, and solvent B was 0.1% (v/v) TFA in acetonitrile. Separations were carried out with the following program: 33–35% solvent B from 0 to 5 min, 35–37% solvent B from 5 to 9 min, 37–40% solvent B from 9 to 18 min, 40–41% solvent B from 18 to 22 min, 41% solvent B from 22 to 27 min, 41–43% solvent B from 27 to 28 min, 43–45% solvent B from 28 to 36 min, 45–33% solvent B from 36 to 37 min. The flow rate was 0.5 mL/min, column temperature was 45°C, and the peaks were detected at 214 nm.

The molecular weight distribution of peptides in fermented samples in the present study was determined using size exclusion chromatography according to our previous study (Wang and Li, 2017). Briefly, the pH of samples was first adjusted to 4.6 with 10% TCA, centrifuged (12,000 × g) at 4°C for 10 min and the supernatant was collected. Peptide molecular distribution of each sample was determined by HPLC equipped with a TSK gel G2000 SWXL column. Five reference substances including aprotinin (6,512 Da), bacitracin (1,423 Da), WPWW (674 Da), NCS (322 Da) and Gly-Sar (146 Da) were used to fit the calibration curve. The peptide contents in different samples were determined by using Pierce™ Quantitative Colorimetric Peptide Assay (Catalog number 23275).

The fermented samples were applied to sequential GI digestion according to our previous study (Wang and Li, 2017). Briefly, fermented samples were incubated with pepsin (1:50, w/w, enzyme/sample, pH 2.0) for 2 h at 37°C, and then the pH was adjusted to 7.5 with 1 M NaOH. The gastric digest was further hydrolyzed by pancreatin (1:50, enzyme/gastric digest, w/w) at 37°C for another 4 h. The simulated pancreatic digestion was terminated by heating solution in boiling water for 10 min.

The ACE inhibitory activity was determined by measuring the amount of hippuric acid (HA) generated from HHL according to previous studies with a few modifications (Abubakar et al., 1998; Lan et al., 2015). ACE was dissolved in 0.2 M borate buffer (pH 8.3) containing 1.0 M NaCl at the concentration of 0.5 U/mL, and the HHL were also dissolved in the same buffer at the concentration of 12.5 mM. Then 75 μL ACE solution and 50 μL HHL solution were mixed and incubated for 60 min at 37°C. The reaction was terminated by adding 150 μL 1 M HCl. The absorbance of HA was monitored at 228 nm with a spectrophotometer after ethyl acetate extraction.

The peptidomics analysis of the fermented milk was performed on a qTOF mass spectrometer Q-TOF-2 (Waters/Micromass, Manchester, UK) equipped with a nanoESI source. The TOF analyzer was calibrated using Glu-fib (Sigma-Aldrich, Shanghai, China). Before MS analysis, peptide mixtures were desalted using C18 Zip-Tip pre-packed micro-columns (Millipore, Bedford, MA, USA). The mobile phase A was 3% acetonitrile solution with 0.1% formic acid and B was 95% acetonitrile solution with 0.1% formic acid. The gradient elution was performed as follows: 0% B for 1 min, linearly 90% B in 60 min, and 90% B for 10 min.

For peptide sequencing, the MS/MS data for each peptide were processed by the MassLynx Maximum Entropy 3 (MaxEnt 3), and converted into PKL files, which were suitable for further analysis. The nano-ESI fragment ion peak lists, generated as PKL files, were analyzed by searching sequence databases with Mascot (Matrix Science, London, UK). Additionally, automated de novo sequencing combined with database searching was performed on the qTOF-MS/MS data using the MassLynx (version 4.1, Waters). Peptides identified were further investigated in related to the BIOPEP database¹ for the ACE inhibitory activity.

The cleavage specificity of CEPs toward the milk protein was determined according to the previous study (Schechter, 2012). The P1 and P1’ subsites were designated as the amino acid residues in the N-terminal direction and C-terminal direction, respectively. P1 subsite will interact with the S1 subsite in CEPs and P1’ will interact with the S1’ subsite in CEPs. Therefore, the P1-P1’ peptide bond was the hydrolyzed bond. The quantitatively analysis for each specific amino acid was calculated, and the calculation process was as following (Solieri et al., 2018):

For amino acid A in position n (P1 or P1’ subsite), the cleavage probability of P1-P1’ peptide bond will be:

Where the Total_An is the total amino acid A cleaved in position n, and the Total_Apro is the total amino acid A in protein. Therefore, the mean cleavage probability of the protein is:

The coefficient Kn was defined as the ratio of Pn and P n ¯ , which showed the positive or negative influence of the amino acid A in the P1-P1’ bond:

When Kn>0 indicated the positive influence of the amino acid A in the P1-P1’ bond, and Kn<0 indicated a negative influence in the bond.

All data are presented as mean ± standard deviation (SD) for three replicates. One way ANOVA and Tukey analyses were performed to determine differences between samples, using the SPSS 19.0 (SPSS Inc., Chicago, IL, USA).

From the taxonomic point of view, 99.96% of the identified bacteria was firmicutes, and Lactobacillus was the predominant genus (99.95%). This result was in accordance with that of a previous study which reported that the most abundant bacterial population in kefir grains was Lactobacillus (Garofalo et al., 2015). The other identified bacterial genera in kefir grains included Pseudomonas, Acetobacter, Delftia, Bacillus, Lactococcus, Acinetobacter, and their abundances were lower than 0.5% (Figure 1A). At the species level, Lactobacillus kefir and Lactobacillus casei were the predominant species which the total abundances were more than 99%. It is reported that the strong biofilm formation capacity of Lactobacillus kefir is a key factor to maintaining the shape and size of kefir grains (Garofalo et al., 2015; Wang et al., 2021). Besides, another 5 species with low abundance were also detected, and the Lactobacillus helveticus was the most abundant one. Lactobacillus helveticus is important for the flavor of fermented milk product because it will produce the nutty flavors and prevent the bitter taste (Griffiths and Tellez, 2013). Previous study also reported that the fermented bovine milk by Lactobacillus helveticus contained ACE inhibitory peptides which could help to decrease the blood pressure (Aihara et al., 2005). As for the fungal diversity, Ascomycota was the most abundant phyla (over 92%). Several previous studies about the microflora of kefir grains from different regions also found that Ascomycota was the predominant phyla (Dertli and Con, 2017; Wang et al., 2021). At the genus level, Saccharomyces and Kluyveromyces were the most abundant fungal (54.76 and 41.55%, respectively), followed the Kazachstania which accounted for 2.33% (Figure 1B). This result was in agreement with previous studies about the fungal composition of kefir grains (Marsh et al., 2013; Gut et al., 2019). Additionally, Trichoderma, Penicillium, Epicocum, Sarcinomyces, Chaetomium, and Myceliophthora were also detected with the abundance lower than 1%.

About 80% of the bovine milk is casein which can be subdivided into β-casein, αs1-casein, αs2-casein and κ-casein. In the fermented process, the consumption of these four casein fractions by kefir grains was determined. As shown in Figure 2, all the hydrolyzed casein fractions increased with the extension of fermentation time. The hydrolyzed contents of β- and αs1-casein were always higher than that of αs2 and κ-casein (p < 0.05). This may be related to the abundance of β- and αs1-casein. It is reported that the β-, αs1-, αs2- and κ-casein occurs in the approximate proportions 4:4:1:1 in milk (Visser et al., 1991). At 6 h, the hydrolysis degrees of four casein fractions were very low, 1.13 mg/mL for β-casein, 1.05 mg/mL for αs1-casein, 0.33 mg/mL for αs2-casein and 0.17 mg/mL for κ-casein. However, at the end of fermentation of 24 h, a 6.8–12.6 fold increase were observed. The difference in hydrolysis degree may be caused by the preference of CEPs from the kefir grains, which would be discussed in the following sections.

In the present study, the molecular weight of generated peptides was divided into four fractions by size exclusion chromatography, i.e., > 5,000 Da, 3,000–5,000 Da, 1,000–3,000 Da, and < 1,000 Da. As shown in Figure 3 the total content of peptides increased at first and then decreased. The maximum concentration of peptides (9.29 mg/mL) was obtained at 18 h, and then decreased to 8.49 mg/mL at 24 h. In addition, the peptides with low molecular weights increased with the extension of fermentation time. In prior period of fermentation (6 h and 12 h), the content of high molecular weight peptides (> 3,000 Da) was more than 70%, and this proportion decreased to 56.3% for 18 h and 40.8% for 24 h, respectively.

The full set of peptides generated in the fermentation process was analyzed by mass spectrometry. A total of 122 milk peptides were released by the CEPs of kefir grains at the end of fermentation (Supplementary Tables S1–S4). According to the protein origin of these identified peptides, β-casein was the most preferred substrate of kefir grains. In particular, 81 peptides were derived from β-casein, which accounted for 66.4% of the total identified peptides. Followed by the κ-casein-derived peptides and αs1-casein-derived peptides, which were 19 peptides (15.6%) and 17 peptides (13.9%), respectively. For the αs2-casein, only 5 peptides (4.1%) were detected in the fermentation system indicating the poor capacity of CEPs to hydrolyze αs2-casein. The significant difference between peptides derived from different casein fractions may be related to the CEPs of microorganism in kefir grains, especially the Lactococcus lactis. CEPs can be classified according to the degradation patterns toward β-, αs1-, αs2- and κ-casein (Kunji et al., 1996). Generally, two kinds of CEPs have been described, P_I-type and P_III-type. The primary substrate of P_I-type is β-casein, and it can also hydrolyze the κ-casein to a lesser extent. For P_III-type CEPs, they can degrade β-, αs1-, and κ-casein equally (Visser et al., 1986; Pritchard and Coolbear, 1993). In addition, an intermediate proteases named P_I-/P_III-type has also been identified, which can cleave β-casein as with P_I-type and α-, and κ-casein to a lesser extent (Exterkate et al., 1993; Sadat-Mekmene et al., 2011a; Villegas et al., 2015). In the present study, the kefir grains exhibited a predominant hydrolyzation activity toward β-casein, and a lower proteolytic activity toward αs1-casein. These results indicated that the CEPs activity of kefir grains in this study may be the P_I-/P_III-type.

The identified peptides in the fermented products were searched against the milk bioactive peptide database and BIOPEP database (Minkiewicz et al., 2008; Nielsen et al., 2017) for finding peptides matched with the known ACE inhibitory peptides. As shown in Table 2, 14 peptides (11.5% of the total identified peptides) were designated with the ACE inhibitory activity. Among these peptides, 12 peptides were derived from β-casein mainly distributed in the region of f47-90 and f132-208, and the other two peptides were from αs1-casein. Four peptides of YPFPGPIPN, LHLPLP, KVLPVPQ and RPKHPIKHQ have been proven to reduce the blood pressure in spontaneously hypertensive rats (Saito et al., 2000; Fuglsang et al., 2003; Quirós et al., 2007; Garcia-Tejedor et al., 2015; Martini et al., 2020). The other 10 peptides have been proven the in vitro antihypertensive activity.

Usually, the resistance to the GI protease is a prerequisite for bioactive peptides to exert their physiological effects (Wang and Li, 2017). In the present study, the ACE inhibitory activities of the fermented milk at different time points were determined before and after the simulated GI digestion (Figure 5). The inhibitory activity increased with the extension of fermentation time, suggesting the continuous generation of ACE inhibitory peptides. At 24 h of fermentation, the inhibitory activities for before and after GI digestion were 73.5 and 46.3%, respectively. The activity retention rate was about 63%, which indicated that these ACE inhibitory peptides possessed well digestion stability. It is notable that, the ACE inhibitory activity at 6 h increased after simulated GI digestion. The degree of hydrolysis and peptide content were low at the beginning of fermentation, and the protease in GI tract might exert positive effect on the peptide releasing. In previous studies, peptides containing Pro in their sequences were proven to be resistant to the digestive proteases (Tagliazucchi et al., 2016; Ma et al., 2021). As shown in Table 2, all the 14 identified ACE inhibitory peptides were Pro-containing peptides and the number of Pro ranged from 1 to 4.