Breast milk collection was approved by the ethics committee of the Northeast Agricultural University. Eighteen women volunteers from each ethnic group (a total of 36) were recruited from the maternity care center in Yanbian Autonomous Prefecture, Jilin Province, and informed consent was obtained from all donors. All mothers received similar diets, according to the results of the questionnaire. The breastmilk samples were collected from mothers with full-term babies in the lactation stage from 14 to 28 days post-partum. The age of mothers ranged from 20 to 35 and feed babies with breastmilk. The parity of most mothers is 1 or 2. At a specific time between 9 a.m. and 11 a.m., milk was collected using an automatic breast pump. The first few drops were discarded and the collected milk samples were immediately shipped to the laboratory on dry ice. The samples were stored at −80°C until analysis. Six samples were randomly classified into one group, and three biological duplications were obtained.

Proteins were extracted from human milk as previously described (14). First, human milk samples were centrifuged to obtain two layers. The skimmed layer was further separated by adjusting the pH to 4.6, where the casein fraction could be precipitated. The casein precipitate was collected, washed using precooled acetone, and then dried in a fume hood, and dissolved using 4% SDS. Bicinchoninic acid was used to quantitatively analyze casein extracted from the samples.

Enzymatic hydrolysis of protein samples was conducted using FASP according to a previous study (15), which seems to be the most effective approach for biological sample aliquots owing to the low miscleavage rate (16). To each protein sample (100 μg) 8 M urea was added to obtain a final volume of 200 μL, and then DL-dithiothreitol was added until a final concentration of 10 mM; this mixture was incubated at 56°C for 30 min to destroy the intramolecular and intermolecular disulfide bonds. Next, iodoacetamide (IAA) at a concentration of 50 mM was used as an alkylation reagent for cysteine and histidine to ensure complete denaturation and reduction of protein samples, and the reaction mix was incubated in the dark for 40 min. The samples were transferred to a 10 K ultrafiltration tube, centrifuged at 12, 000 × g at room temperature, and the filtrate was discarded. Subsequently, 400 μL of urea (8 M) was centrifuged thrice at 12,000 × g at room temperature. Ammonium bicarbonate solution (50 mM, 200 μL) was introduced into an ultrafiltration tube and centrifuged at 12,000 × g at room temperature, and the filtrate was discarded. This step was repeated three times. Trypsin was placed in an ultrafiltration tube at a sample/enzyme ratio of 50/1 (mass ratio), and enzymolysis was performed at 37°C for 16 h. After enzymolysis, the digesta was centrifuged at 4°C at 12000 × g, and the filtrate was collected and lyophilized.

High performance liquid chromatography (HPLC) separation of the enzymatically hydrolyzed samples was conducted using an HPLC system (Agilent 1100, USA) equipped with Waters X Bridge C18 (5 μm, 4.6 mm × 250 mm, 120 Å). This HPLC was preparative and was used to divide the peptides into 10 fractions based on polarity. The mobile phase was composed of A (98% ddH₂O, 2% acetonitrile, pH 10) and B (98% acetonitrile, 2% ddH₂O, pH10). The 60-min chromatographic gradient for separation was as follows: 0–5 min/97% A/0.4 mL/min, and then the flow rate was kept at 0.7 mL/min and the mobile phase change was 5–5.10 min/97% A, 5.10–10 min/95% A, 10–35 min/82% A, 35–45 min/66% A, 45–58 min/5% A, 58–60 min/97% A. Samples were collected every 1 minute.

The 10 fractions of peptide samples were mixed into 3 mixtures and then subjected to HPLC-MS (Thermo Scientific EASY-nLC, USA). The 60 min gradient for separation was: 0 min/3% B, 0–2 min/8% B, 2–46 min/28% B, 46–55 min/50% B, 55–56 min/100% B, 56–60 min/100% B. The flow rate was maintained at 0.6 mL/min. The MS parameters were as follows: (1) MSn1: scan range 375–1500 m/z; maximum injection time (MIT) 50 ms; AGC target 400,000; orbitrap resolution 120,000. (2) MSn2: MIT 22 ms; AGC target 50,000; orbitrap resolution 15,000; collision energy 30%.

Proteins in the milk casein fraction were qualitatively obtained using DDA. To enhance the signal of low-abundance proteins in the MS spectra, the digested peptides from each sample were mixed. Tryptic digestion was detected by LC-MS/MS using a Thermo Scientific EASY-nLC mass spectrometer. The proteins were searched using Proteome Discoverer 2.1.0182 (Thermo Fisher Scientific, Rockford, IL, USA) with Sequest HT. The data bank used was UniProt TaxId:9606. Enzyme: trypsin; miss cleavage: 2; peptide mass tolerance: ± 10 ppm; fragment mass tolerance: ± 0.02 Da; peptide false discovery rate (FDR): less than 1%.

The samples were then subjected to HPLC-MS (Thermo Scientific Orbitrap Fusion Lumos, USA). The gradient for separation and flow rate was consistent with those of the DDA described above. The DIA MS parameters were as follows: (1) MSn1: scan range 350–1,500 m/z; MIT 50 ms; AGC target 400,000; Orbitrap resolution 60,000. (2) MSn2: scan range 200–2,000 m/z; MIT 54 ms; AGC target 300,000; orbitrap resolution 30,000; collision energy 33%.

DIA was used for quantitative analysis of proteins using Skyline software (Department of Genome Sciences, University of Washington, Ave. NE, Seattle, WA). The relative quantitation of proteins was performed using the normalized spectral count method. The parameters were set as follows: sub-ion M/Z: larger than parent ion and last ion-3; maximum number of sub-ions: 5; minimum number of sub-ions: 2; sub-ion extraction window: 5 min; dotp: greater than or equal to 0.6.

The significance level, expressed as p values, was estimated using the t-test. DEPs were obtained using standards of fold change (FC) > 1.5 or < 0.67 and p < 0.05. Benjamin-Hochberg analysis was employed to obtain the corrected p values (q value) for GO and KEGG enrichment analysis.

All identified proteins were analyzed for annotation in terms of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses using DAVID Bioinformatics Resources 6.8 (http://david.abcc.ncifcrf.gov/home.jsp). Proteins differentially expressed between the two groups were subjected to GO and KEGG enrichment analyses. Protein-protein interaction (PPI) networks between differentially expressed proteins for each comparison were examined using the STRING online database (http://string-db.org) with a minimum required interaction score of 0.4, and figures were drawn using Cytoscape (Version 3.7.1). The Cytohubba plug-in was used to analyze the top 10 hub proteins using the MCC method.

Proteomic investigation of the human milk casein fraction by qualitative DDA mode (567 proteins) and quantitative DIA technology revealed a total of 535 proteins (Supplementary Table 1 listed the detailed information). Among all the identified proteins, major caseins, such as β-casein (W5RWE1), κ-casein (P07498), and α_s1-casein (E9PDQ1), were observed, with β-casein having the highest relative abundance, followed by κ-casein and α_s1-casein. β-casein consists of 226 amino acids and is highly phosphorylated (10). Recently, β-casein has been developed as a natural nanocarrier to encapsulate and deliver hydrophobic nutrients to enhance their bioavailability (17). κ-casein has a highly glycosylated C-terminus and exhibits both antibacterial and prebiotic effects. In vitro studies have shown that κ-casein can inhibit the binding of Helicobacter pylori to human gastric mucosa while boosting the proliferation of Bifidobacterium infantis and Lactobacillus bifidus (18). α_s1-Casein contains cysteine, which can form disulfide links with κ-casein, and the low degree of phosphorylation benefits the infant immune system (19). As shown in Supplementary Table 1, lysozyme (P61626), the main whey protein, was also detected in the casein fraction. As indicated in a recent study, approximately 75% of lysozyme in human milk is naturally bound to casein, and this association does not affect the antibacterial activity of lysozyme (7).

To reveal the potential physiological function of casein, all identified and relatively quantified proteins were annotated using the GO, KEGG, and EggNOG databases. GO is a standardized system for gene function classification. Figure 1 shows that 528 proteins were annotated with 52 terms, which were further clustered into 24, 17, and 11 terms in categories of biological process (BP), cellular component (CC), and molecular function (MF), respectively. In each category, the terms with the greatest number of proteins were cellular process (481 proteins, 91.1%), cell, cell part (497 proteins, 94.13%), and binding (452 proteins, 85.61%) for BP, CC, and MF, respectively. The most prevalent molecular function of the binding activity has been reported in other studies (11). From the perspective of protein types involved in GO function annotation, it was found that the most functional protein involved in the greatest number of GO function items was the amyloid-beta A4 protein (A0A140VJC8). This protein acts as a cell-surface receptor and has roles in axon growth, neuronal adhesion, and axon formation on the surface of neurons (20).

KEGG results can be interpreted from the following five categories: cellular processes, environmental information processing, genetic information processing, metabolism, and organismal systems. Figure 2 shows that only 106 proteins were found to be related to 23 secondary branches, which fell into 3, 2, 2, 12, and 4 pathways for the aforementioned five categories, respectively. Within each branch, the specific pathways with the greatest number of proteins were transport and catabolism (12, 11.32%), signal transduction (9, 8.49%), folding, sorting, and degradation (15, 14.15%), carbohydrate metabolism (39, 36.79%), and the endocrine system (15, 14.15%). It was also found that aldehyde dehydrogenase (NAD(P) +) (A0A1B0GW77) was involved in most KEGG pathways, up to 14.

Functional classification of all the identified proteins was performed using EggNOG. The annotation results (Figure 3) indicated that, besides unknown functions (S), ‘post-translational modification, protein turnover, chaperone’ (D) were the most enriched functional class, with a total of 102 proteins, followed by ‘translation, ribosomal structure, biogenesis’ (J), and ‘intracellular trafficking, secretion, vesicular transport’ (U) with 60 and 52 proteins, respectively.

As shown in Table 1, 39 proteins, representing approximately 7.3% of the total proteins, were significantly differentially (p < 0.05) expressed between the two groups. The 39 DEPs between Korean and Han may be related to the differences in ethnicity caused by genetic variation. Han ethics group is the group with the largest population of about 1.2 billion, and there are 55 ethnic minorities. Korean ethics group in the northeast of China has obvious characteristics and has a population of about 2 million. In this study, we studied the protein differences between the milk samples of Han and Korean ethics groups in the northeast of China (Jilin Province). As far as we know, this is the first study regarding the investigation of protein differences in human milks between the Han and Korean ethics groups. However, differences in human milk serum from different ethics groups which are distributed in Yunnan (Han and Bai ethnicity), Gansu (Han and Tibetan ethnicity), Xinjiang (Uygur ethnicity), and Inner Mongolia (Mongolian ethnicity) were studied in a previous study (21) and the authors found that 34 proteins significantly differed with geography and ethnicity. Our results were similar with this study. Although the women from different groups had a similar diet during the first month after giving birth to a child since they all live in the care center. However, there was a large difference in the traditional diet for the two ethics groups. Therefore, traditional diet may be another reason for the differences in proteins between these two groups. Additionally, the life style of people in Han and Korean ethics groups was different in terms of culture, behavioral habits, socioeconomic and environmental settings, which may be other influencing factors for the differences (39). Eight immunity-related proteins (PIGR, IGK, immunoglobulin J chain, immunoglobulin alpha-2 heavy chain, and four immunoglobulin fragments) were included in the DEPs, indicating their possible role in the critical phase of immunological immaturity of the newborn. It can be concluded that the abundance of immune-related proteins in milk produced by mothers depends on the mother’s life circumstances, which may be related with ethnicity and diet (21). PIGR is the polymeric Ig receptor when IgA and IgM are transported across epithelial cells into milk, and the polypeptide of the joining (J) chain provides antibodies with the capacity to bind PIGR (22). The five differentially expressed enzymes, tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, cathepsin S, mannosyl-oligosaccharide 1,2-alpha-mannosidase IA, asparaginyl endopeptidase, and aldehyde dehydrogenase NAD (+), indicated a possible difference in human milk casein fractions in the processes of glycolipid metabolism, protein degradation, carbohydrate hydrolysis, and the removal of excess aldehydes in the body.

The volcano plot (Figure 4) revealed that 10 DEPs were upregulated (red color) and 29 were downregulated (blue color) among the differentially expressed proteins. Figure 5 shows the hierarchical clustering of 39 DEPs with three biological replicates in each group, which displays the upregulated and downregulated proteins. Hierarchical clustering can be used to group proteins with similar structures, which helps investigate protein functions. For example, nucleobindin-1 (A8K7Q1) and nucleobindin-2 (P80303) are divided into one group, and research has shown that they are homologous multidomain calcium and DNA binding proteins (23) and have multiple functions, including regulation of inflammation and bone formation (24).

Among the upregulated proteins, complement component C9 (A0A024R035) had the highest difference (the highest FC value) between the two groups with a log2FC of 2.52, P = 0.0013. As a single-chain glycoprotein, C9 is the final product of the complement cascade, which is part of the membrane attack complex (MAC). C9 can adhere to the cell surface, thus destroying the integrity of the cell membrane and resulting in osmotic lysis and cell death (25). Among the downregulated proteins, serotransferrin (P02787), an iron-binding glycoprotein, had the lowest log2FC value of −2.43, p = 0.0045. In addition to regulating the iron content of biological fluids, serotransferrin has been associated with a variety of diseases such as atransferrinemia and cardiovascular diseases. The properties of serotransferrin and its receptor can be exploited to deliver drugs specifically into the brain and cancer cells (26).

Enrichment analysis revealed that 125 of the 545 GO terms in which DEPs participated were significantly enriched (q < 0.05). The top 30 enriched GO terms are displayed in the bubble chart in Figure 6. The functions with the most proteins were response to stimulus (7 proteins, GO:0050896), extracellular organelle (8, GO:0043230), and peptidase activity, acting on L-amino acid peptides (3, GO:0070011) and peptidase activity (3, GO:0008233) for BP, CC, and MF categories, respectively. This means most of the DEPs were related with these functions and the human milk from Korean and Han exhibited different functions toward infants due to the differences in the proteins. These differences may be due to genetic background, lifestyle, and environmental factors. Blood microparticles (4, GO: 0072562) were the most significant GO terms, with a rich factor of 0.235 and a minimum q value of 2.58E-06. Among all the significantly enriched terms, negative regulation of the empirical cell apoptotic process (2, GO: 1904036) had the highest rich factor of 0.33. Figure 7 shows the DEPs involved in the above GO terms in the form of a chord diagram; it is obvious that negative regulation of the empirical cell apoptotic process included the fibrinogen alpha chain and fibrinogen gamma chain. As an abundant plasma glycoprotein, fibrinogen consists of two copies of chains: Aα, Bβ, and γ (27). As a critical part of hemostasis, water-soluble fibrinogen forms a network of fibrin fibers with the participation of thrombin and coagulates into blood clots (28). Besides participating in blood coagulation, fibrinogen is involved in inflammation, cell migration, and tumorigenesis (27). Moreover, the immunoglobulin J chain and the two immunoglobulin fragments are both involved in the response to the stimulus. C3, C9, and serotransferrin are present in blood microparticles. There are two enzymes in the chord diagram, cathepsin S and asparaginyl endopeptidase, both of which participate in peptidase activity, and the latter also participates in extracellular organelle and peptidase activity, acting on L-amino acid peptides.

For KEGG enrichment analysis, 35 of the 113 KEGG pathways were significantly enriched (q < 0.05); the first 20 pathways are displayed in Figure 8. Having a rich factor of 0.103, Staphylococcus aureus infection (hsa05150) was the most significant pathway with a minimum q value of 1.03E-10. In terms of rich factors, African trypanosomiasis (5, hsa05143) had a maximum value of 0.135. As for the pathway with the largest number of proteins, S. aureus infection and complement and coagulation cascades (0.089, 1.39E-10, hsa04610) involved seven proteins.

As showed above, complement and coagulation cascades pathway possessed the largest number of proteins, meaning that the human milk samples between the two groups may exhibit different effects on infants in term of improving immune system. As discussed in Section 3.3, eight immunity-related proteins were included in the DEPs, which may be the reason. Composed of more than 30 proteins, the complement system participates in both innate and adaptive immunity (29). As a vital part of multiple pathways of the immune system, the complement system not only participates in host defense but also involves many physiological systems, such as the coagulation cascade (30). The complement system can be activated through three different pathways-the classical, alternative, and lectin pathways—and the 3 pathways merge at C3 (31). The complement cascade reaction rapidly mobilizes complement system proteins in response to infection or tissue damage and ends with the formation of MAC and C5a (30). MAC has been extensively studied for its ability to be inserted into cell membranes to induce cell lysis (30).

The coagulation cascade starts with primary hemostasis and then releases coagulation factors for secondary hemostasis. At this time, internal and external pathways activate factor X. Finally, activated factor X cleaves prothrombin into thrombin, converting fibrinogen into fibrin, and initiating fibrin polymerization at the site of injury (32). As mentioned above, parts of the complement system (C3, C9), fibrinogen (fibrinogen alpha and gamma chain) with coagulation ability, and plasminogen involved in wound healing are involved in this pathway. The other proteins involved in this pathway are α-2-macroglobulin and CD59 glycoprotein.

Staphylococcus aureus has attracted much attention as an important pathogen in infant infections. Newborns are likely to be infected with S. aureus through the birth canal, breastfeeding, and contact with people and the surrounding environment, thus increasing the possibility of short-term morbidity and mortality and adverse long-term outcomes (33). Based on the above results, S. aureus infection was the most significantly enriched in both ethnic groups and possessed the greatest number of DEPs. It is thus, worth exploring and its pathway map is shown in Figure 9.

Clumping factor B (ClfB) with the ability to bind to fibrinogen can adhere to human cytokeratin 10. S. aureus on entry into the human body stimulates three different pathways in the complement system to produce chemoattractants that can cause neutrophils to move to the infected area. In response to the immune response, S. aureus also has multiple mechanisms that interfere with complement through the inactivation or isolation of key components. First, cell wall-associated protein A (SpA) binds to IgG Fc fragments to prevent phagocytosis and classical pathway complement fixation (34). SpA also acts as a B-cell superantigen through interactions with the heavy chain variable part of the Fab fragments. Sbi domains III and IV interact with complement factor C3 to interfere with all three complement pathways (35). Additionally, staphylokinase can activate plasminogen (PLG), thereby cleaving the surface-bound C3b and IgG and reducing the phagocytosis of neutrophils. Moreover, clumping factor A (ClfA) can bind to fibrinogen, resulting in the enhanced degradation of C3b. Finally, the secreted superantigen-like protein (SSL7) binds to IgA, thus hindering phagocytosis (34).

Two immunoglobulin fragments (A0A5C2GLT5 and A0A7S5EXD1), immunoglobulin alpha-2 heavy chain, plasminogen, fibrinogen gamma chain, C3, and rheumatoid factor RF-ET6 fragment (A2J1N5) are the proteins involved in S. aureus infection. As the precursor of plasmin, plasminogen is encoded by the PLG gene on chromosome band 6q26 (36) and is composed of five kringle-like domains containing “lysine-binding sites” and a C-terminal domain homologous to other trypsin-like proteases (37). In addition to its significant fibrinolysis, plasminogen triggers other enzymatic cascades, including complement. Moreover, it affects immune and inflammatory processes by binding to specific cell surface receptors (plasminogen, an enigmatic zymogen).

All DEPs formed a network (Figure 10) with 23 nodes and 47 edges (combined score > 0.4), indicating that 23 DEPs (19 downregulated and 4 upregulated) interacted with one another through 47 interactions. The hub proteins in the network were assessed using the MCC method, and the following proteins were ranked as the top 10: fibrinogen alpha chain (1,442), C9 (1,442), plasminogen (1,441), serotransferrin (1,441), fibrinogen gamma chain (1,440), cDNA FLJ35730 fis, highly similar to α-1-antichymotrypsin (B3KS79, 1440), C3 (724), α-2-macroglobulin (721), CD59 glycoprotein (5), and PIGR (4).

As mentioned earlier, the fibrinogen alpha and gamma chains were involved in the GO term of negative regulation of the empirical cell apoptotic process, which had the largest rich factor. As part of MAC (38), C9 was the most upregulated protein, whereas serotransferrin had the lowest FC value. The polymeric immunoglobulin receptor was the protein with the highest abundance in DEPs. As discussed in the KEGG enrichment analysis, the complement and coagulation cascades are the pathways with the most DEPs, and it can be observed that seven out of the ten hub proteins, namely, fibrinogen alpha chain, gamma chain, C9, plasminogen, C3, alpha-2-macroglobulin, and CD59 glycoprotein were obtained from this pathway. Additionally, fibrinogen gamma chain, plasminogen, and C3 were enriched in Staphylococcus aureus infection, which had the lowest significance level and the most proteins as discussed in the enrichment analysis.