Skip to main content
NCBI home page
Food Chemistry: X logoLink to Food Chemistry: X
. 2026 Feb 21;34:103681. doi: 10.1016/j.fochx.2026.103681

Multi-omics analysis of functional component networks in human, porcine, and ruminant milk: Insights for precise design of infant formula and milk replacers based on bovine milk

Rui Chen a,1, Jiamei Liang a, Yuanyuan Zheng a, Yinggang Sun a, Yanzhi Wu a, Yanqin Meng b, Qiu Zhang a, Qian Li a, Xiaowei Wang a, Yuan Pan a, Siyi Li a, Ran Duan a, Qingyu Yuan a, Qiying Zhan a, Zeying Wang a,⁎
PMCID: PMC12950474  PMID: 41777635

Abstract

This integrated multi-omics study (metabolomics, lipidomics, proteomics) compared the compositions of bovine, caprine, ovine, porcine, and human milk. Human milk is defined by a unique microenvironment where complex human milk oligosaccharides and immune regulating small molecules act together. Components like Lnnt and Lnfp iii form interactive networks, which help modulate both gut microbiota and the immune system. Porcine milk contributes to neural development and immune function. It is rich in polar phospholipids and polyunsaturated fatty acids such as docosahexaenoic acid, important for neural structure and function. Porcine milk also contains specifically high levels of immunoglobulins, including IgM and IgG. In contrast, ruminant milks like bovine and caprine milk show a distinct composition, with high levels of certain functional proteins and active metabolites, such as xanthine dehydrogenase, osteopontin, and allantoin. These findings provide new insights for the precise development of better bovine milk based infant formula and animal milk replacers.

Keywords: Bovine milk, Porcine milk, Infant formula, Milk replacer, Immunoglobulin

Highlights

  • •

    Multi-omics analysis shows each milk contains unique, highly abundant nutrients.

  • •

    Human and porcine milk contain more nutrients that promote cognitive function.

  • •

    Porcine milk has the highest abundance of lactotransferrin, IgM, and IgG.

  • •

    Ruminant milks have a similar phospholipid profile and are monoglyceride-rich.

1. Introduction

As an essential nutritional source for mammalian offspring, breast milk possesses a unique and complex composition. It is rich in proteins, lipids, and metabolites, playing a crucial role in infant growth, immune regulation, and health maintenance. Simultaneously, due to the global shortage of human milk supply, bovine milk-based infant formula (IF) has become a vital alternative for infant nutrition. Furthermore, in modern animal husbandry, early weaning of offspring such as lambs, calves, and piglets, followed by feeding with milk replacer (MR), has become a common practice.

However, cow's milk allergy (CMA) remains a common issue when infants consume formula, affecting approximately 3.8% of young children, primarily due to the presence of α-S1-casein. The development of extensively hydrolyzed formulas has enabled the safe consumption of bovine milk products for allergic infants (Lin & Yanjun, 2024), significantly broadening the application of such formulas. Soy milk, as a nutritious functional food, is also used in cases of CMA (Martin et al., 2016; Tasdemir & Goelge, 2023). Nevertheless, significant immunological differences persist between breastfed and IF-fed infants, with the latter showing evidence of a Th2 bias and overactivation of immune cells (Salinas et al., 2025). This disparity may stem from the unique bioactive components in human milk. Furthermore, most studies have found that lambs fed conventional MR show lower nutrient digestibility, feed efficiency, daily weight gain, and a higher incidence of diarrhea compared to those fed natural milk. This is likely related to impaired intestinal antioxidant capacity and underdeveloped microbiota caused by MR feeding (Esfiokhi et al., 2024). There remains considerable room for optimization in conventional MR formulations.

When formulating IF or MR, it is essential to further optimize the composition based on differences in milk from various species. Recent evidence indicates significant similarities in the milk fat globule membrane (MFGM) composition between human and bovine milk (Manoni et al., 2020). However, differences exist in the expression of some functional proteins. For instance, the highly expressed GLYCAM1 in bovine milk may play a key role in immune defense and cell adhesion (Chen et al., 2025). There are notable differences in the average size of milk fat globules (MFGs) among human, caprine, bovine, and ovine milk (Thum et al., 2023), and the size of MFGs determines their digestion rate in the body. Compared to bovine milk, caprine milk contains smaller fat globules and a higher concentration of medium-chain fatty acids (MCFAs), thereby improving digestibility (Donovan, 2019). Furthermore, human milk contains abundant polyunsaturated fatty acids (PUFAs), particularly linoleic acid (LA), arachidonic acid (AA), and docosahexaenoic acid (DHA), which have been proven to offer significant health benefits. Human milk also has the highest oligosaccharide content compared to bovine and ovine milk (Zhong et al., 2024). Comparative data from Mitina et al. show that among the species studied, porcine milk has the highest content of long-chain unsaturated fatty acids (LCUFAs), exceeding the levels detected in human and bovine milk (Mitina et al., 2020).

In this study, an integrated multi-omics strategy was used to use metabolomics based on UHPLC-Q-TOF mass spectrometry, lipidomics based on UPLC-orbitrap-MS, and 4D label-free proteomics. We hypothesize that these milks possess distinct molecular profiles that underlie their species-specific biological functions. This study is the first to elucidate the fine composition differences and potential functional connections between these five mammalian milks from the perspective of systems biology. Our findings establish a novel scientific foundation for developing next-generation precision nutrition formulas, which are designed to optimize molecular functionality through tailored modifications of bovine milk.

2. Material and methods

2.1. Sample collection

In this study, milk samples were acquired from five distinct mammalian species under controlled conditions. Holstein bovine milk (HST) samples were collected from the Huishan Dairy Group. Small Tail Han ovine milk (OAS) samples were obtained from farms in Inner Mongolia and Lingle, while Liaoning Cashmere caprine milk (LCG) samples were sourced from the Liaoyang National Core Sheep Breeding Farm. Chinese human milk (CHP) samples were provided by the Shenyang Maternal and Child Care Center, and Large White porcine milk (YXP) samples were gathered from the Yangxiang Pig Group. All samples were collected during autumn at two weeks postpartum, with healthy animals in their second lactation cycle selected for milking. Following standardized protocols, each 10 mL aliquot was immediately stored at −20 °C for short-term preservation (≤14 days) prior to dry ice transportation. For extended storage, milk samples were maintained at −80 °C with a maximum duration of four months. The final dataset included six replicates for metabolomics, six for lipidomics, and three for proteomics analyses across the five milk types.

Ethical Statements: No interaction between the animals and the research staff occurred. All human milk samples were obtained through informed consent and voluntary donation by the donors themselves. The experimental protocols involving biological subjects in this research were reviewed and approved by the Institutional Ethics Committee for Biomedical Research at Shenyang Agricultural University (Ethical Approval ID: 202106015).

2.2. Experimental instruments and reagents

Metabolomics: AB Triple TOF 6600 Mass Spectrometer (AB SCIEX), Agilent 1290 Infinity Ultra High Performance Liquid Chromatograph (Agilent 1290 Infinity UHPLC), Refrigerated High-Speed Centrifuge (Eppendorf 5430R), Chromatographic Column: Waters ACQUITY UPLC BEH Amide Column, 1.7 μm, 2.1 mm × 100 mm, Acetonitrile (Merck, 1499230–935), Ammonium Acetate (Sigma-Aldrich, 70221), Methanol (Fisher, A456–4), Ammonia Solution (Fisher, A470–500);

Lipidomics: Q-Exactive Plus Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific), Nexera LC-30 A Ultra High Performance Liquid Chromatograph (SHIMADZU UHPLC), Refrigerated High-Speed Centrifuge (Eppendorf 5430R), Chromatographic Column: Waters ACQUITY UPLC CSH C18 Column, 1.7 μm, 2.1 mm × 100 mm, Acetonitrile (Thermo Fisher), Isopropanol (Thermo Fisher), Methanol (Thermo Fisher), internal standards (SPLASH® LIPIDOMIX MASS SPRC STANDARD, AVANTI, 330707-1EA): Ceramides (Cer), Lysophosphatidylcholine (LPC), Phosphatidylcholine (PC), Lysophosphatidylethanolamine (LPE), Phosphatidylethanolamine (PE), Phosphatidylinositol (PI), Phosphatidylserine (PS), Phosphatidic acid (PA), Phosphatidylglycerol (PG), sphingomyelin (SM), Cholesterol Ester, monoglyceride (MG), Diglyceride (DG), Triglyceride (TG);

Proteomics: Easy nLC nanoflow HPLC system, Trap Column: Thermo Scientific EASY-Spray Column, 2 cm × 100 μm, 5 μm C18, timsTOF Pro Ion Mobility-Quadrupole-Time of Flight Mass Spectrometer, Oasis® MCX μElution Plate (30 μm), Formic Acid, Acetonitrile, 5× Loading Buffer, UA Buffer, IAA Alkylation Buffer.

2.3. Metabolomics

2.3.1. Sample pretreatment

Samples were thawed at 4 °C and quantitatively supplemented with 100 μL of acetonitrile-water (1:1, v/v). After vortex mixing, the homogenate was ultrasonicated at low temperature for 30 min using a dedicated extraction system. Following centrifugation (14,000 ×g, 20 min, 4 °C), the supernatant was lyophilized under vacuum. For LC-MS analysis, the dried extract was reconstituted in 100 μL of the initial solvent system, then purified via secondary centrifugation (14,000 ×g, 15 min, 4 °C) to obtain high-purity analytes.

2.3.2. Chromatography-mass spectrometry analysis

Chromatographic analysis was conducted on an Agilent 1290 Infinity LC ULTRA UHPLC system with a HILIC column thermostated at 25 °C. The mobile phase consisted of solvent A (aqueous solution of 25 mM ammonium acetate and 25 mM ammonia) and solvent B (acetonitrile), delivered at 0.5 mL/min with an injection volume of 2 μL. A multi-step gradient elution was executed as follows: initial isocratic condition at 95% B for 0.5 min, followed by a linear decrease to 65% B over 6.5 min (0.5–7 min), a rapid transition to 40% B in 1 min (7–8 min), and maintained at 40% B for 1 min (8–9 min). The system was then re-equilibrated to 95% B within 0.1 min (9–9.1 min) and stabilized for 3 min (9.1–12 min). Samples were consistently stored at 4 °C in the autosampler, with analytical sequences randomized and interspersed with quality control (QC) samples to monitor instrument performance throughout the analysis.

Mass spectrometric detection was conducted using an AB SCIEX Triple TOF 6600 instrument operating in positive and negative electrospray ionization (ESI) modes. Key ESI parameters included nebulizer gas (Gas1) and auxiliary gas (Gas2) set to 60 psi, curtain gas (CUR) at 30 psi, ion source temperature maintained at 600 °C, and spray voltage adjusted to ±5500 V. High-resolution mass spectrometric data were acquired in full-scan mode across m/z 60–1000 with a 0.20 s/spectrum dwell time, coupled with auto-triggered MS/MS fragmentation (m/z 25–1000) targeting the top 10 precursor ions per acquisition cycle at 0.05 s/spectrum. Dynamic collision energy (35 ± 15 eV) and declustering potential (±60 V) were systematically optimized to maximize ion fragmentation efficiency. A dynamic exclusion window of 4 Da was applied to prevent redundant fragmentation of dominant ions.

2.3.3. Data processing and analysis

In this study, no internal standards were added during the untargeted metabolomics analysis. The primary reason is that the biological samples involved contain a wide variety of metabolites with significantly diverse physicochemical properties. A single or a small set of internal standards cannot achieve precise correction for the signal response and separation behavior of all metabolites across the full spectrum. Forced application might instead introduce additional bias. Throughout the experiment, QC samples were injected at regular intervals. The reliability of the detection data was ensured by evaluating instrument stability, which was achieved by calculating the relative standard deviation of metabolite peak areas in the QC samples. First, the QC-SVR method was employed to correct for batch effects, thereby eliminating technical variations such as signal drift and retention time shifts. Subsequently, the data were normalized using the total peak area (sum normalization) to mitigate global intensity differences caused by variations in sample loading amounts and matrix effects. The normalized dataset was then subjected to multivariate statistical analysis using the ropls package in R. Pareto-scaled principal component analysis was applied to visualize the overall clustering of samples. Finally, orthogonal partial least squares-discriminant analysis was used to identify differentially enriched metabolites between comparison groups. Metabolite annotation was conducted using a custom internal chemical database (Shanghai Applied Protein Technology) through validation across multiple dimensions: retention time alignment, exact mass accuracy (mass error < 10 ppm), MS/MS spectral matching, and collision energy consistency. All identifications were systematically revalidated through manual inspection to meet Level 2 confidence criteria per metabolomics reporting standards. Volcano plots were generated on the Omicshare bioinformatics cloud platform (https://www.omicshare.com) using thresholds of |Log2FC| > 1 and P < 0.05. Principal component analysis (PCA) plots, hierarchical clustering heatmaps, correlation heatmaps, and KEGG pathway diagrams were created using R packages, while network diagrams were constructed with Cytoscape 3.6.0.

2.4. Lipidomics

2.4.1. Sample pretreatment

Samples were spiked with a internal standards (SPLASH® LIPIDOMIX MASS SPRC STANDARD,AVANTI,330,707-1EA) and homogenized with 200 μL water and 240 μL methanol. Following the addition of 800 μL methyl-tert-butyl ether and vortex mixing, 240 μL pre-chilled methanol was incorporated with subsequent vortex agitation. After 20-min ice-water bath sonication and 30-min room-temperature incubation, phase separation was achieved via centrifugation (14,000 ×g, 15 min, 10 °C). The organic phase was collected, dried under nitrogen stream, and reconstituted in 200 μL of 90% isopropanol/acetonitrile. Post-vortex mixing, 90 μL aliquots underwent secondary centrifugation (14,000 ×g, 15 min, 10 °C), yielding analytical-grade supernatants.

2.4.2. Chromatographic-mass spectrometric analysis

2.4.2.1. Chromatographic condition

Chromatographic separation was performed on a UHPLC Nexera LC-30 A system equipped with a temperature-controlled C18 column (45 °C). The mobile phase comprised solvent A (acetonitrile-water, 6:4, v/v) and solvent B (acetonitrile-isopropanol, 1:9, v/v), delivered at a flow rate of 300 μL/min. The gradient protocol initiated with 30% B isocratic elution (0–2 min), followed by a linear increase to 100% B over 23 min (2–25 min), and concluded with a 10-min column re-equilibration phase at 30% B (25–35 min) to ensure system stability and reproducibility. To ensure analytical stability, the autosampler maintained samples at 10 °C, and injection sequences were randomized to reduce signal variability.

2.4.2.2. Mass spectrometer condition

Dual-polarity ESI was utilized for analytical detection, enabling comprehensive metabolite profiling in both positive and negative ion modes. Following UHPLC separation, the samples were analyzed with a Q Exactive series mass spectrometer (Thermo Scientific™). The ESI source conditions were set as follows: Heater Temp at 300 °C, Sheath Gas Flow Rate at 45 arb, Aux Gas Flow Rate at 15 arb, Sweep Gas Flow Rate at 1 arb, Spray Voltage at 3.0 kV, Capillary Temp at 350 °C, and S-Lens RF Level at 50%. High-resolution mass spectrometry was performed with an MS1 scan range of 200–1800 m/z, employing a data-dependent acquisition strategy that triggered 10 HCD-based MS2 scans following each full scan. The resolution was set to 70,000 (MS1) and 17,500 (MS2) at m/z 200.

2.4.3. Data processing and analysis

Lipidomics was performed using an internal standard method combined with single-point quantification. Raw data were processed with LipidSearch software, enabling automated peak alignment, feature extraction, and lipid identification and relative quantification based on MS/MS spectra, directly yielding qualitative and relative quantitative results for lipids. After obtaining the relative quantitative data, the subsequent preprocessing and statistical analysis steps were consistent with those used in untargeted metabolomics. Specifically, batch effects were first corrected using the QC-SVR method. This was followed by total peak area normalization to eliminate global intensity differences, and Pareto scaling to prepare the data for multivariate statistical analysis. Orthogonal partial least squares-discriminant analysis was then applied to the preprocessed dataset to identify differentially abundant lipids between groups. Differentially abundant lipids were screened using thresholds of Log₂FC > 1 and P < 0.05. In addition, lipid composition analysis, hierarchical clustering, correlation analysis, and pathway enrichment analysis were conducted. All statistical plots (PCA scores plots, heatmaps, volcano plots, and pathway maps) were generated using the same tools and platforms as those employed in the metabolomics analysis.

2.5. Proteomics

2.5.1. Protein extraction

Milk samples were subjected to centrifugation at a force of 1006.2 ×g for 15 min. After this process, the fat layer was carefully removed. The resulting skim milk was centrifuged at 11,180 ×g for 1 h. Proteins were extracted from the whey using a urea lysis buffer (UA). Following this, the proteins were resuspended and sonicated for 40 min using a 30-s on and 30-s off cycle. After sonication, the samples were centrifuged at 17,468.75 ×g for 20 min at 21 °C, and the supernatant was collected. Protein concentration was quantified via BCA assay prior to electrophoretic analysis. For SDS-PAGE, aliquots containing 20 μg protein were combined with 5× loading buffer, denatured by boiling (5 min), and separated electrophoretically on 12.5% resolving gels under constant current (14 mA, 90 min). Resolved protein bands were visualized through Coomassie Brilliant Blue R-250 staining.

2.5.2. Enzymatic hydrolysis in solution

Protein samples (200 μg) were denatured in 8 M urea buffer (UA) for 1 h at 25 °C. After adding 100 mM DTT, the mixture was agitated (40.2 ×g, 1 min) and incubated for 1 h. Subsequently, 50 mM iodoacetamide (IARA in UA) was introduced, followed by agitation (40.2 ×g, 1 min) and 30-min dark incubation. The urea concentration was reduced to 2 M by dilution with 100 mM ammonium bicarbonate. Trypsin (4 μg in 25 mM NH₄HCO₃; 40 μL total volume) was added, and the solution was agitated (40.2 ×g, 1 min) prior to 37 °C digestion for 16–18 h. The digest was acidified with TFA (pH ∼3), desalted via Oasis® MCX μElution Plate, and lyophilized. Purified peptides were reconstituted in 40 μL of 0.1% formic acid and quantified by A280 spectrophotometry.

2.5.3. Mass spectrometry on machine and searching database

Peptide separation was performed on an Easy nLC nanoflow HPLC system utilizing a binary mobile phase composed of 0.1% formic acid in water (Solution A) and 0.1% formic acid in 84% acetonitrile (Solution B). The column was pre-equilibrated with 95% of Solution A. The samples were loaded onto a Thermo Scientific EASY column (2 cm × 100 μm, 5 μm-C18) and then separated on a Thermo Scientific EASY column (75 μm × 100 mm, 3 μm-C18) at a flow rate of 300 nL/min. Chromatographic separation employed a multi-step gradient elution protocol: 0–50 min (0% → 35% B), 50–55 min (35% → 100% B), and 55–60 min (isocratic at 100% B). Peptides were ionized via timsTOF Pro mass spectrometry in positive ion mode over a 60-min runtime (m/z 100–1700) using the parallel accumulation-serial fragmentation (PASEF) method, with 10 PASEF scans per cycle and a 1.17-s acquisition window. Raw data (.D format) were processed through Peaks software for database searching, enabling both identification and label-free quantification.

2.5.4. Data processing and analysis

Protein identification and label-free quantification were performed using MaxQuant software (version 1.5.3.17). The acquired MS/MS spectra were searched against the following UniProt reference proteome databases: Bos mutus (wild yak) [72004]; Bos taurus(Bovine) [9913] & Bubalus bubalis (Domestic water buffalo) [89462]; Bos taurus (Bovine) [9913]; Camelus dromedarius (Dromedary) [9838]; Capra hircus (Goat) [9925]; Equus [9789]; Homo sapiens (Human) [9606]; Ovis aries (Sheep) [9940]; and Sus scrofa (Pig) [9823]. The false discovery rate (FDR) for both peptide and protein identification was set to 0.01. For protein quantification, razor and unique peptides were utilized for intensity calculation. The quantitative protein profiles were subsequently evaluated for inter-sample reproducibility. Differential expression analysis across species was conducted using appropriate statistical methods. Subcellular localization predictions were generated using CELLO (http://cello.life.nctu.edu.tw/). Protein-protein interaction (PPI) networks were constructed via the STRING database (http://string-db.org/), followed by network interaction analysis in Cytoscape 3.6.0.

2.6. Multi-omics integrated analysis

Differentially expressed metabolites and lipids were screened using thresholds of OPLS-DA VIP > 1 and P < 0.05. For proteomics, the analysis primarily focused on differentially expressed proteins. For the lipid-protein, lipid-metabolite, and protein-metabolite pairs, Pearson correlation coefficients were calculated among all differential molecules. Pairs with r > 0.8 were retained. The R package corrplot (v0.84) was used to generate correlation matrix heatmaps, and the R package pheatmap (v1.0.12) was used to create hierarchical clustering heatmaps. All significant correlation pairs were imported into Cytoscape software to construct and visualize association networks. For protein-metabolite functional pathway integration, KEGG pathway enrichment analysis was performed separately for differential proteins and differential metabolites. The R package VennDiagram (v1.6.20) was used to generate Venn diagrams to identify the top 10 pathways with the most shared involvement between the two omics datasets. These pathways were ranked in descending order based on the total number of differential proteins and metabolites within each pathway. The results were visualized using a bar plot generated with the R package ggplot2 (v3.3.6).

3. Results

3.1. Immune-associated molecules revealed by metabolomics

Metabolomics analysis identified 1498 metabolites with significantly differential abundance (OPLS-DA VIP > 1, P < 0.05), comprising 882 in positive ion mode and 616 in negative ion mode. The top 20 differential metabolites and metabolites of interest are listed in Supplementary Table 1. Principal component analysis confirmed distinct separation of metabolic profiles among groups (Fig. 1C), and subclass analysis showed that lipids and lipid-like molecules, organic acids and derivatives, and organoheterocyclic compounds were the most abundant categories (Fig. 1A), collectively indicating an active metabolic state in milk. Notably, volcano plot results (FC > 1.5 or FC < 0.67, P < 0.05) revealed that the number of downregulated metabolites was higher in HST compared to other species (Fig. 1B), preliminarily suggesting that milk from other species may contain a richer array of potential functional components.

Fig. 1.

Fig. 1

Open in a new tab

Metabolomics results. (Including both positive and negative ion modes) (A) Number of metabolites by subcategory; (B) Volcano diagrams of differentially expressed metabolites; (C) Principal component analysis diagrams; (D) Hierarchical clustering heat maps; (E) Correlation heat maps; (F) Metabolite network interaction maps; (G) KEGG bubble chart.

To explore the association of these components with immunomodulation, we performed functional enrichment analysis of the differential metabolites. Pathway enrichment results indicated that differential metabolites were significantly enriched in ABC transporters and biosynthesis of amino acids pathways (Fig. 1G), with the latter closely related to the proliferation and function of immune cells. More importantly, metabolite cluster analysis and correlation analysis revealed a high positive correlation among pyridoxal phosphate and N-acetylneuraminic acid (Neu5Ac), which were abundant in YXP, and Lnnt, which was highly expressed in CHP. Additionally, 5-aminovaleric acid betaine (5-AVAB) and echinocystic acid (EA) showed significantly higher abundance and exhibited clear clustering trends in HST and OAS samples (Fig. 1D-E). This suggests that these metabolites may act synergistically in YXP and CHP. Further network analysis (Fig. 1F) found that in negative ion mode, Lnt and linoleic acid, which were abundant in CHP and YXP, were positively correlated, but showed a significant negative correlation with allantoin, which was enriched in ruminant milk. Furthermore, Probucol was specifically enriched in CHP, taurine was highly enriched in LCG, and the important sialylated human milk oligosaccharide (HMO) 3′-sialyllactose (3’-SL) was present at higher levels in YXP. These species-specific metabolite association patterns collectively constitute a foundation for a potential immunomodulatory microenvironment.

3.2. Species-specific differences in bioactive lipid signaling molecules

Through lipidomics analysis, this study revealed significant differences in the composition of bioactive lipid signaling molecules among milk from different species. A total of 359 significantly differential lipid molecules were identified (OPLS-DA VIP > 1, P < 0.05). It is particularly noteworthy that YXP exhibited the highest content of various specific lipid molecules, especially very long-chain polyunsaturated fatty acids (VLCPUFAs) such as CerG2GNAc1(d41:2), Hex1Cer(t41:2), PG(38:2), and Hex2Cer(m40:3), which were significantly elevated compared to other milk samples (Supplementary Table 2). Cluster analysis showed that samples from the same species clustered closely together, and metabolic profiles were clearly separated between species, confirming the species specificity of lipid composition (Fig. 2E). Ruminant milk had a higher total lipid content than CHP and YXP (Fig. 2C). Over 90% of the lipids in ruminant milk were glycerolipids (GL). However, the content of glycerophospholipids (GP) was much lower in ruminant milk (5.711%–7.199%) compared to YXP (63.419%) and CHP (19.466%) (Fig. 2A). Specifically, YXP was rich in polar phospholipids (PLs) such as PE, PG, and PC, primarily composed of long-chain fatty acids and very long-chain fatty acids. These lipids are core components of cell membranes and important signaling precursors. Among key bioactive lipids, various n-3 and n-6 series polyunsaturated fatty acids (PUFAs), crucial for neural development and immune regulation, showed significant species-specific distributions (Fig. 2B). YXP and CHP demonstrated notable advantages in fatty acid composition. The relative abundances of DHA, eicosapentaenoic acid (EPA), linoleic acid (LA), and arachidonic acid (AA) were significantly higher in YXP than in the other species. Likewise, CHP exhibited significantly higher abundances of DHA and AA compared to ruminant milk.

Fig. 2.

Fig. 2

Open in a new tab

Lipidomics results. (A) Pie chart of lipid subcategory contents; (B) Bar charts of unsaturated fatty acid content; (C) Bar chart of lipid molecular content; (D) Differential lipid volcano diagram; (E) Hierarchical clustering heat map; (F) Correlation heat map; (G) Positive and negative ion mode lipid network interaction maps.

Differential comparison volcano plot analysis (FC > 1.5 or FC < 0.67, P < 0.05) further confirmed the distinct lipid profiles among species (Fig. 2D). Compared to HST, CHP exhibited a strong trend of lipid downregulation, while OAS, LCG, and YXP showed upregulation trends. In particular, the comparison between HST and YXP revealed the most extensive changes in lipid diversity, suggesting fundamental differences in their lipid signaling environments. Correlation and network analyses uncovered potential synergistic and antagonistic relationships in lipid signaling (Fig. 2F-G). Lipids within the same functional subclass (e.g., PE, PI, PC, PG) generally showed positive correlations, indicating possible co-regulation. Notably, PC(18:1_20:4) was strongly positively correlated with almost all measured lipids, suggesting it may occupy a central node in the lipid metabolic network. In negative ion mode, the PG subclass showed the strongest associations with other lipids, with PE(18:3e_22:5) and PG(40:6e) each being strongly correlated with as many as 17 and 16 other lipids, respectively. This indicates that these phospholipids containing PUFAs may act as key signaling hubs. In summary, YXP and CHP is rich in PLs and PUFA precursors related to membrane structure and cellular signaling, constituting a unique potential reservoir of lipid signaling molecules. In contrast, ruminant milk is mainly composed of TG that serve a storage function, and it has a distinct fatty acid profile.

3.3. Proteomics uncovers immune protein profiles

The number of proteins identified in milk from different species in this study is shown in Fig. 3A. The results indicate that 727 proteins were identified in HST, 747 in OAS, 1047 in LCG, 1193 in CHP, and 478 in YXP. Notably, CHP had the highest total number of identified proteins, while YXP had the fewest. The number of proteins identified in HST was similar to that in OAS. A detailed list of key proteins is provided in Supplementary Table 3. Quantitative proteomic analysis revealed that xanthine dehydrogenase (XDH), lactadherin, and osteopontin (OPN) were predominantly enriched in HST. Lactotransferrin (LTF), immunoglobulin M (IgM), and immunoglobulin G (IgG) were most abundant in YXP. Lipoprotein lipase (LPL) was most abundant in CHP. Meanwhile, the protein-protein interaction network in Fig. 3D integrates the top 100 most abundant proteins from each species's milk and the key proteins investigated in this study. As shown, ALB occupies a central position in the network and exhibits the most extensive interactions with other proteins. The number and proportional distribution of proteins across different subcellular locations are presented in the pie chart in Fig. 3C. In all species analyzed, the majority of proteins were located extracellularly, indicating that functional molecules, including most immune related substances, primarily operate in the extracellular environment. A Venn diagram (Fig. 3B) was used to analyze the overlap of milk proteins among the three samples from each species. The results showed minimal variation in protein numbers, demonstrating high experimental reproducibility.

Fig. 3.

Fig. 3

Open in a new tab

Proteomics results. (A) Protein quantity bar chart; (B) Venn diagrams; (C) Subcellular localization analysis diagrams; (D) PPI diagrams.

3.4. Integrated functional analysis

Our analysis identified the top ten KEGG pathways most strongly associated with the detected proteins and metabolites, and the results were visualized accordingly. Fig. 6D-E present the shared pathway enrichment of differential metabolites and proteins of interest from the two omics datasets, with the proteome represented in blue and the metabolome in orange. The neuroactive ligand-receptor interaction pathway was commonly enriched in HST, CHP, and OAS samples; the galactose metabolism pathway was significantly enriched in LCG and YXP samples; additionally, the glycerolipid metabolism pathway showed enrichment in HST samples. Hierarchical clustering heatmaps and correlation heatmaps (Fig. 4A-B, 5A-B, 6A-B) illustrate the relationships among key proteins, metabolites, and lipids. In the Pearson correlation analysis, a continuous color scale indicates the strength and direction of correlations: red represents a positive value (r > 0), blue represents a negative value (r < 0), with darker colors denoting increasing correlation strength. Pairwise network analysis was used to investigate the interactions among milk metabolites, lipids, and proteins across different animal species (Fig. 4C, 5C, 6C). A color gradient from red (high connectivity) to yellow and then to green (low connectivity) represents the level of connectivity. The correlation analysis revealed specific interaction patterns between functional components in the different milk samples. In HST, PS(16:2e_22:6) showed a positive correlation with EA, but negative correlations with OPN and taurine. In OAS, LTF was positively correlated with allantoin. In LCG, LTF was positively correlated with PE(18:3e_22:5). In CHP, LPL was positively correlated with EA and also showed a highly significant correlation with DG(36:2e). In YXP, Lnnt exhibited a highly significant positive correlation with TG(6:0_6:0_18:3).

Fig. 6.

Fig. 6

Open in a new tab

Results of combined metabolomics and proteomics analysis. (A) Hierarchical clustering heat maps; (B) Correlation heat maps; (C) Metabolite-protein network interaction maps; (D) Metabolite-protein co-participation pathway Venn diagrams; (E) Metabolite-protein co-participation pathway number bar charts.

Fig. 4.

Fig. 4

Open in a new tab

Results of combined lipidomics and proteomics analysis. (A) Hierarchical clustering heat maps; (B) Correlation heat maps;(C) Lipid-protein network interaction maps.

Fig. 5.

Fig. 5

Open in a new tab

Combined analysis results of metabolomics and lipidomics. (A) Hierarchical clustering heat maps; (B) Correlation heat maps; (C)Metabolite-lipid network interaction maps.

4. Discussion

This study focuses on analyzing the functional component profiles of mature milk from different species, and on developing IF and MR based on bovine mature milk. It is important to note that mature milk and colostrum differ fundamentally in their composition and functional priorities. For example, mature milk is typically richer in components associated with long-term nutrition supply and specific physiological regulation (Ji et al., 2026). In this context, the present study systematically reveals the significant differences and unique advantages in key functional components of mature milk across species, and provides an in-depth exploration of the complex functional networks and underlying mechanisms formed by these components.

As key components of IF, the milk-derived oligosaccharides Lnt and Lnnt were found to be  most abundant in CHP, followed by YXP. They can resist degradation by stomach acid, reaching the intestine intact to exert multiple physiological functions. Lnt can be metabolized by microorganisms into short-chain fatty acids, creating a competitive advantage for beneficial bacteria such as bifidobacteria and lactobacilli. At the same time, it can directly bind to corresponding receptors of viruses and fungi, inhibiting their adhesion to the intestinal epithelium, and regulate immune cell function to maintain intestinal homeostasis (Hu et al., 2023). An animal study further confirmed that administering Lnnt to newborn mice significantly promotes colon development. By postnatal day 21, it clearly stimulates the proliferation and differentiation of intestinal stem cells and enhances the expression of specific markers for goblet cells, enterocytes, Paneth cells, and enteroendocrine cells (Liao et al., 2025). Moreover, Lnfp iii, the fucosylated derivative of Lnt and Lnnt, showed  higher abundance in CHP. It promotes immune system maturation and modulates gut microbiota by stimulating Th2 responses. Epidemiological studies confirm that higher concentrations of Lnfp iii in breast milk are associated with a lower risk of CMA in infants (Seppo et al., 2017; Yang, Zhu, et al., 2024).

Sialyllactoses (SLs) are another important class of prebiotic compounds. YXP was identified as a rich source of both 3’-SL and 6′-sialyllactose (6’-SL). In the body, SLs are broken down into Neu5Ac and lactose. Neu5Ac can enter bifidobacterial cells via specific transporters, drive their metabolic pathways, thereby promoting the growth of beneficial bacteria and inhibiting the intestinal adhesion of pathogens such as Escherichia coli and Clostridium difficile. In addition, they can also exert anti-inflammatory effects by regulating the function of immune cells, which can help alleviate colon inflammation and restore intestinal homeostasis (Zhu et al., 2023). Neu5Ac is a key substance for neural development, capable of enhancing learning ability and potentially alleviating neurological issues related to hydrocephalus (Wang, 2012). In this study, Neu5Ac was found to be abundant in YXP, CHP, and HST. It shown that supplementation with 6’-SL during lactation promotes long-term cognitive development more effectively than supplementation with Neu5Ac alone (Oliveros et al., 2018), suggesting that the intact SL structure holds unique value for neural development.

Lipid composition is a core reflection of functional differences in milk. In HST, OAS, and LCG, the relative abundances of TG and DG were similar, with GLs constituting over 90% of milk fat. The higher TG levels in ruminant milk are primarily attributed to acetic acid and butyric acid produced by rumen microbial fermentation, which directly promote TG biosynthesis. MG was also present at higher levels in ruminant milk such as HST. MG is not only a natural food emulsifier, but recent research has shown that supplementing sows with glycerol monolaurate enhances intestinal barrier function and suppresses inflammation in piglets (Zhao et al., 2023). Regarding phospholipids, the contents of PC, PE, and SM detected in CHP could reach up to four times those in HST. The relative abundance of PLs in YXP was significantly higher than in other milk samples, indicating its notable potential in supporting cognitive function. PLs enriched in the MFGM help improve intestinal barrier integrity and functional maturity in infants. MFGM-mimetic structures formulated with human milk PLs and SM can improve infant neurodevelopment and increase hippocampal SM levels (Wang, Qiao, et al., 2025). The contents of PLs and SM in LCG and OAS were slightly higher than in HST. These substances, as key components of MFGM, may be involved in regulating the surface area and volume of MFGs, thereby regulating digestibility.

In the differential analysis of lipid composition, YXP exhibited a distinct lipidomic profile, with the content of several specific lipid molecules possessing potential biological importance being significantly higher than in other milk samples. Notably, lipids containing VLCPUFAs, such as CerG2GNAc1(d41:2), Hex1Cer(t41:2), Hex2Cer(m40:3), and PG(38:2), were particularly enriched in YXP. VLCPUFAs are key components for cell membrane construction, signal transduction, and endocrine regulation. High concentrations of VLCPUFAs may help prevent aging-related cognitive decline and are closely associated with early brain development (Baliyan et al., 2023). Regarding n-3 and n-6 PUFAs, the DHA content in YXP and CHP was significantly higher than in HST, OAS, and LCG. DHA is crucial for brain and visual development and plays a role in regulating neuroinflammation and neurotransmitter activity. The proportion of EPA in YXP (1.149%) was also significantly higher than in other samples. EPA can inhibit pro-inflammatory factors, promote brain and visual development, and may positively influence energy metabolism and lipid balance (Xiong et al., 2024). AA was most abundant in YXP, followed by CHP. AA is a major fatty acid in the placenta and neural cells, supporting infant neurocognitive development. n-3 and n-6 fatty acids share key enzymes in their metabolism, and excessively high dietary n-6 fatty acids can competitively inhibit the conversion of n-3 fatty acids to EPA and DHA (Janssen & Kiliaan, 2014). Therefore, the higher levels of DHA and EPA in YXP and CHP, combined with their high AA content, suggest a more balanced fatty acid profile that may be more conducive to neural development. An animal study confirmed that providing female rats with a high n-3 PUFA diet before and during pregnancy significantly increased the content of DHA-rich phospholipids in fetal brains and promoted the accumulation of functional lipids related to neurotransmission and memory (Okagu et al., 2025).

Quantitative proteomic analysis showed that XDH, lactadherin, and OPN were significantly enriched in HST. XDH is a core regulator of lipid droplet formation and fatty acid oxidation. Its reversible conversion through a redox-dependent mechanism in mammals serves as a key mediator for mammary gland adaptation to metabolic demands during gestation and lactation (Mather, 2000). Lactadherin plays an important role in intestinal barrier repair, anti-inflammatory responses, and resistance to rotavirus infection by promoting macrophage clearance of apoptotic neutrophils (Sato et al., 2023). OPN, presented at the highest level in HST, is a glycosylated phosphoprotein that plays a key regulatory role in infant neurobehavioral development, partly by enhancing the myelination process (Aksan et al., 2021). LTF was most abundant in YXP, with its content in CHP being approximately four times that in HST. LTF possesses antibacterial, anticancer, and immunomodulatory activities. Recent research also indicates that bovine LTF can inhibit Plasmodium infection by binding and degrading heme, showing potential as a novel antimalarial agent (Obayashi et al., 2024). Regarding immunoglobulins, although immunoglobulin A is the most abundant immunoglobulin in animal milk, this study found that IgM and IgG were present at higher levels in YXP, suggesting they may have unique functions in immune transfer and protection. The high level of IgG in porcine milk indicates the possible presence of an efficient, neonatal Fc receptor-mediated active transport mechanism in the mammary gland (Jiang et al., 2016). The high content of IgM may be transported to mucosal surfaces via the polymeric immunoglobulin receptor, providing rapid early immune defense for newborn piglets (Oskam et al., 2023).

Regarding key metabolites, the most enriched substances in HST samples included EA and allantoin. EA can significantly improve allergic inflammation by inhibiting immune cell infiltration and collagen deposition (Kang et al., 2022). Allantoin combats inflammation and oxidative stress by inhibiting the NF-κB pathway and activating the AKT/Nrf2/HO-1 axis, while also modulating the gut microbiota (Yang, Sun, et al., 2024). Probucol was relatively abundant in CHP and HST. This compound can promote the recovery of learning and memory ability in aging mice by increasing dendritic density and branching, showing potential value in treating age-related diseases (Xie et al., 2021). 5-AVAB showed the highest abundance in OAS samples. It is known for its antioxidant, anti-inflammatory, and anti-cancer properties, and dietary intake can induce apoptosis in colon cancer cells (D'Onofrio et al., 2021; Salzano et al., 2021). Furthermore, taurine, which was most abundant in LCG, is a key component supporting infant growth and development. It promotes fat absorption, prevents retinal degeneration, and recent studies show that taurine supplementation can restore Nr1d1 expression, enhance epithelial barrier integrity, and reduce the production of pro-inflammatory cytokines (Bakshi et al., 2023; Wang, Zhou, et al., 2025). Pyridoxal phosphate was most abundant in YXP. It can play a role in treating neurodegenerative diseases such as Parkinson's disease by activating GOT1 to inhibit ferroptosis (Min et al., 2025). Its significant correlation with Neu5Ac further suggests a potential synergistic effect in neuronal development. At the level of lipid metabolism, we discovered deeper inter-component associations. LPL, present at the highest level in CHP, showed a positive correlation with the specific diglyceride DG(36:2e), suggesting its influence on lipid accumulation through promoting DG generation. A study indicates that dietary DHA supplementation enhances LPL activity, promoting the hydrolysis of TG in very low density lipoprotein and chylomicrons (Klingel et al., 2019), while succinate supplementation can simultaneously increase both DHA and LPL content in muscle (Cao et al., 2024), reflecting the complex associations among metabolites, enzymes, and functional fatty acids.

It should be noted that the untargeted metabolomics analysis in this study is limited by the completeness of the database annotations. Future research needs to integrate species-specific databases and in vivo/in vitro functional models to conduct more in-depth functional validation and mechanistic studies on the key active components identified, in order to precisely elucidate their nutritional effects.

5. Conclusion

This study employed multi-omics technologies to systematically analyze the functional component profiles of bovine, ovine, caprine, human, and porcine milk, revealing that milk from different species has complementary functional priorities in supporting infant development. Specifically, bovine milk-based formulas require targeted fortification with neurodevelopment-associated lipids, especially polar phospholipids such as sphingomyelin and phosphatidylethanolamine, as well as long-chain polyunsaturated fatty acids like DHA and EPA, to mimic the lipid features of porcine and human milk. Concurrently, it is essential to enhance sialylated and fucosylated oligosaccharides, such as 3’-SL, 6’-SL, and Lnfp iii, to emulate the core functions of human milk in promoting gut health and immune regulation. Furthermore, the addition of key functional proteins like osteopontin and lactadherin can be considered. On the other hand, the ratio of n-6 to n-3 polyunsaturated fatty acids in triglycerides needs to be reviewed and optimized to prevent excessively high levels of n-6 fatty acids, such as LA, from inhibiting the conversion to beneficial components like DHA. Our findings indicate that human milk serves as the gold standard for constructing an immunomodulatory and gut-microbiota axis. Porcine milk, rich in neuroactive lipids, sialylated oligosaccharides, and high levels of immunoglobulins, provides an excellent reference for neurodevelopment and immune support. Meanwhile, the high abundance of specific bioactive components in bovine and caprine milk, such as xanthine dehydrogenase, osteopontin, echinocystic acid, and allantoin, offers important clues for developing formulations with targeted metabolic regulatory functions, including anti-inflammatory and antioxidant properties. In summary, utilizing the design principles of the immune microenvironment from human milk, the neurodevelopmental and immune architecture from porcine milk, and integrating the unique bioactive substances from ruminant milk, will provide a research foundation for developing the next generation of infant formula, animal milk replacers, and functionally targeted nutritional products.

CRediT authorship contribution statement

Rui Chen: Writing – original draft, Formal analysis, Data curation. Jiamei Liang: Conceptualization. Yuanyuan Zheng: Methodology. Yinggang Sun: Software. Yanzhi Wu: Validation. Yanqin Meng: Supervision. Qiu Zhang: Visualization. Qian Li: Data curation. Xiaowei Wang: Conceptualization. Yuan Pan: Formal analysis. Siyi Li: Methodology. Ran Duan: Software. Qingyu Yuan: Visualization. Qiying Zhan: Validation. Zeying Wang: Writing – review & editing, Resources, Funding acquisition.

Funding

This study received financial support from the National Key Research and Development Program of China (2023YFD1300200).

Declaration of competing interest

The study was conducted in the absence of any business or financial relationship that might be interpreted as a potential conflict of interest.

Acknowledgement

We thank Dr. Zeying Wang for his guidance and support for this study.

Footnotes

Appendix A

Supplementary data to this article can be found online at https://doi.org/10.1016/j.fochx.2026.103681.

Appendix A. Supplementary data

Supplementary material

mmc1.docx (1.3MB, docx)

Data availability

Data will be made available on request.

References

  1. Aksan A., Erdal I., Yalcin S.S., Stein J., Samur G. Osteopontin levels in human milk are related to maternal nutrition and infant health and growth. Nutrients. 2021;13(8) doi: 10.3390/nu13082670. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bakshi S., Paswan V.K., Yadav S.P., Bhinchhar B.K., Kharkwal S., Rose H.…Bunkar D.S. A comprehensive review on infant formula: Nutritional and functional constituents, recent trends in processing and its impact on infants’ gut microbiota. Frontiers in Nutrition. 2023;10 doi: 10.3389/fnut.2023.1194679. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baliyan S., Calvo M.V., Piquera D., Montero O., Visioli F., Venero C., Fontecha J. Milk fat globule membrane concentrate as a nutritional supplement prevents age-related cognitive decline in old rats: A lipidomic study of synaptosomes. Food Research International. 2023;163 doi: 10.1016/j.foodres.2022.112163. [DOI] [PubMed] [Google Scholar]
  4. Cao M., Xie N., Zhang J., Jiang M., Huang F., Dong L.…Tian J. Dietary supplementation with succinic acid improves growth performance and flesh quality of adult Nile tilapia (Oreochromis niloticus) fed a high-carbohydrate diet. Animal Nutrition. 2024;18:390–407. doi: 10.1016/j.aninu.2024.04.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Chen X., Yu H., Bai X., Wu C., Yang M. Proteomics and phosphoproteomics differences between human and cow milk fat globule membrane: An innovative guide for infant formula. Food Chemistry. 2025;472 doi: 10.1016/j.foodchem.2025.142890. [DOI] [PubMed] [Google Scholar]
  6. D’Onofrio N., Martino E., Mele L., Colloca A., Maione M., Cautela D.…Balestrieri M.L. Colorectal cancer apoptosis induced by dietary δ-valerobetaine involves PINK1/Parkin dependent-mitophagy and SIRT3. International Journal of Molecular Sciences. 2021;22(15) doi: 10.3390/ijms22158117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Donovan S.M. Human Milk proteins: Composition and physiological significance. Nestle Nutrition Institute Workshop Series. 2019;90:93–101. doi: 10.1159/000490298. [DOI] [PubMed] [Google Scholar]
  8. Esfiokhi S.M., Norouzian M.A., Abadi M.S., Ahvanooei M.R. Effects of probiotic-supplemented milk replacer on growth, blood biochemistry, fermentation, digestibility, and carcass traits in lambs. Veterinary and Animal Science. 2024;25 doi: 10.1016/j.vas.2024.100368. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Hu M., Miao M., Li K., Luan Q., Sun G., Zhang T. Human milk oligosaccharide lacto-N-tetraose: Physiological functions and synthesis methods. Carbohydrate Polymers. 2023;316 doi: 10.1016/j.carbpol.2023.121067. [DOI] [PubMed] [Google Scholar]
  10. Janssen C.I., Kiliaan A.J. Long-chain polyunsaturated fatty acids (LCPUFA) from genesis to senescence: The influence of LCPUFA on neural development, aging, and neurodegeneration. Progress in Lipid Research. 2014;53:1–17. doi: 10.1016/j.plipres.2013.10.002. [DOI] [PubMed] [Google Scholar]
  11. Ji Z., Han R., Zhao Y., Meng L., Yang Y., Wang J., Zheng N. Comparative analysis of the fatty acid profiles and milk fat globule membrane proteome in human and small ruminant colostrum and mature milk using data-independent acquisition for targeted infant nutrition improvement. Journal of Dairy Science. 2026;109(1):165–182. doi: 10.3168/jds.2025-27203. [DOI] [PubMed] [Google Scholar]
  12. Jiang X., Hu J., Thirumalai D., Zhang X. Immunoglobulin transporting receptors are potential targets for the immunity enhancement and generation of mammary gland bioreactor. Frontiers in Immunology. 2016;7:214. doi: 10.3389/fimmu.2016.00214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Kang Y.M., Kim H.M., Lee M., An H.J. Effects of echinocystic acid on atopic dermatitis and allergic inflammation of the skin and lungs. Phytomedicine. 2022;104 doi: 10.1016/j.phymed.2022.154211. [DOI] [PubMed] [Google Scholar]
  14. Klingel S.L., Metherel A.H., Irfan M., Rajna A., Chabowski A., Bazinet R.P., Mutch D.M. EPA and DHA have divergent effects on serum triglycerides and lipogenesis, but similar effects on lipoprotein lipase activity: A randomized controlled trial. The American Journal of Clinical Nutrition. 2019;110(6):1502–1509. doi: 10.1093/ajcn/nqz234. [DOI] [PubMed] [Google Scholar]
  15. Liao J., Wang M., Li H., Li T., Deng Z., Li J.…Zhang B. Human milk oligosaccharide LNnT promotes intestinal epithelial growth and maturation during the early life of infant mice. Journal of Agricultural and Food Chemistry. 2025;73(11):6678–6690. doi: 10.1021/acs.jafc.4c10055. [DOI] [PubMed] [Google Scholar]
  16. Lin M., Yanjun C. Research progress on the mechanism of probiotics regulating cow milk allergy in early childhood and its application in hypoallergenic infant formula. Frontiers in Nutrition. 2024;11 doi: 10.3389/fnut.2024.1254979. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Manoni M., Di Lorenzo C., Ottoboni M., Tretola M., Pinotti L. Comparative proteomics of Milk fat globule membrane (MFGM) proteome across species and lactation stages and the potentials of MFGM fractions in infant formula preparation. Foods. 2020;9(9) doi: 10.3390/foods9091251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Martin C.R., Ling P.R., Blackburn G.L. Review of infant feeding: Key features of breast milk and infant formula. Nutrients. 2016;8(5) doi: 10.3390/nu8050279. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Mather I.H. A review and proposed nomenclature for major proteins of the milk-fat globule membrane. Journal of Dairy Science. 2000;83(2):203–247. doi: 10.3168/jds.S0022-0302(00)74870-3. [DOI] [PubMed] [Google Scholar]
  20. Min X., Du Z., Wei J., Yuan Z., She Y., Jin X.…He X. Pyridoxal phosphate inhibits alpha-synuclein-induced ferroptosis by activating GOT1 to enhance the methionine salvage pathway in Parkinson’s disease. Experimental Neurology. 2025;393 doi: 10.1016/j.expneurol.2025.115411. [DOI] [PubMed] [Google Scholar]
  21. Mitina A., Mazin P., Vanyushkina A., Anikanov N., Mair W., Guo S., Khaitovich P. Lipidome analysis of milk composition in humans, monkeys, bovids, and pigs. BMC Evolutionary Biology. 2020;20(1):70. doi: 10.1186/s12862-020-01637-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Obayashi M., Kimura M., Haraguchi A., Gotanda M., Kitagawa T., Matsuno M.…Miyata T. Bovine lactoferrin inhibits plasmodium berghei growth by binding to heme. Scientific Reports. 2024;14(1) doi: 10.1038/s41598-024-70840-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Okagu I.U., Akerele O.A., Fillier T., Pham T.H., Thomas R., Wilson K.A., Cheema S.K. Maternal omega-3 polyunsaturated fatty acids improved levels of DHA-enriched phosphatidylethanolamines and enriched lipid clustering in the neuronal membranes of C57BL/6 mice fetal brains during gestation. The Journal of Nutritional Biochemistry. 2025;140 doi: 10.1016/j.jnutbio.2025.109891. [DOI] [PubMed] [Google Scholar]
  24. Oliveros E., Vázquez E., Barranco A., Ramírez M., Gruart A., Delgado-García J.M.…Martín M.J. Sialic acid and sialylated oligosaccharide supplementation during lactation improves learning and memory in rats. Nutrients. 2018;10(10) doi: 10.3390/nu10101519. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Oskam N., den Boer M.A., Lukassen M.V., Ooijevaar-de Heer P., Veth T.S., van Mierlo G.…Heck A.J.R. CD5L is a canonical component of circulatory IgM. Proceedings of the National Academy of Sciences of the United States of America. 2023;120(50) doi: 10.1073/pnas.2311265120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Salinas M.L., Mulakala B.K., Davidson L.A., Cai J.J., Donovan S.M., Chapkin R.S., Yeruva L. Single-cell transcriptomics reveals that human milk feeding shapes neonatal immune cell interleukin signaling pathways in a nonrandomized clinical trial. The American Journal of Clinical Nutrition. 2025;122(1):196–207. doi: 10.1016/j.ajcnut.2025.04.024. [DOI] [PubMed] [Google Scholar]
  27. Salzano A., Neglia G., D’Onofrio N., Balestrieri M.L., Limone A., Cotticelli A.…Campanile G. Green feed increases antioxidant and antineoplastic activity of buffalo milk: A globally significant livestock. Food Chemistry. 2021;344 doi: 10.1016/j.foodchem.2020.128669. [DOI] [PubMed] [Google Scholar]
  28. Sato K., Jinno S., Nakamura Y., Eto S., Inagaki M. Absolute quantification of bovine lactadherin to screen the anti-rotavirus activity of dairy ingredients. Journal of Dairy Science. 2023;106(4):2261–2270. doi: 10.3168/jds.2022-22401. [DOI] [PubMed] [Google Scholar]
  29. Seppo A.E., Autran C.A., Bode L., Järvinen K.M. Human milk oligosaccharides and development of cow’s milk allergy in infants. The Journal of Allergy and Clinical Immunology. 2017;139(2):708–711.e705. doi: 10.1016/j.jaci.2016.08.031. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Tasdemir Y., Goelge E. Production of phenylalanine-reduced soymilk for phenylketonuria patients. Quality Assurance & Safety of Crops and Food. 2023;15(4):179–189. doi: 10.15586/qas.v15i4.1396. [DOI] [Google Scholar]
  31. Thum C., Roy N.C., Everett D.W., McNabb W.C. Variation in milk fat globule size and composition: A source of bioactives for human health. Critical Reviews in Food Science and Nutrition. 2023;63(1):87–113. doi: 10.1080/10408398.2021.1944049. [DOI] [PubMed] [Google Scholar]
  32. Wang B. Molecular mechanism underlying sialic acid as an essential nutrient for brain development and cognition. Advances in Nutrition. 2012;3(3):465s–472s. doi: 10.3945/an.112.001875. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Wang S., Qiao F., He J., Fu J., Liu T., Yi H.…Lin K. Phospholipid remodeling and tri-layer membrane reconstruction mediate cognitive effects of humanized Milk fat globules in neonatal rats. Advanced Science (Weinheim) 2025 doi: 10.1002/advs.202507926. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Wang Z., Zhou L., Zheng Y., Zhong X., Huang R., Sun W.…Li W. Nuclear receptor Nr1d1 links sleep deprivation to intestinal homeostasis via microbiota-derived taurine. Journal of Translational Medicine. 2025;23(1) doi: 10.1186/s12967-025-07089-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Xie Y., Song A., Zhu Y., Jiang A., Peng W., Zhang C., Meng X. Effects and mechanisms of probucol on aging-related hippocampus-dependent cognitive impairment. Biomedicine & Pharmacotherapy. 2021;144 doi: 10.1016/j.biopha.2021.112266. [DOI] [PubMed] [Google Scholar]
  36. Xiong Y., Li X., Liu J., Luo P., Zhang H., Zhou H.…Zhou L. Omega-3 PUFAs slow organ aging through promoting energy metabolism. Pharmacological Research. 2024;208 doi: 10.1016/j.phrs.2024.107384. [DOI] [PubMed] [Google Scholar]
  37. Yang L., Zhu Y., Meng J., Zhang W., Mu W. Recent progress in fucosylated derivatives of lacto-N-tetraose and lacto-N-neotetraose. Critical Reviews in Food Science and Nutrition. 2024;64(28):10384–10396. doi: 10.1080/10408398.2023.2224431. [DOI] [PubMed] [Google Scholar]
  38. Yang S., Sun X., Liu D., Zhang Y., Gao X., He J.…He D. Allantoin ameliorates dopaminergic neuronal damage in MPTP-induced Parkinson’s disease mice via regulating oxidative damage, inflammation, and gut microbiota disorder. Food & Function. 2024;15(18):9390–9408. doi: 10.1039/d4fo02167c. [DOI] [PubMed] [Google Scholar]
  39. Zhao H., Tian M., Xiong L., Lin T., Zhang S., Yue X.…Guan W. Maternal supplementation with glycerol monolaurate improves the intestinal health of suckling piglets by inhibiting the NF-κB/MAPK pathways and improving oxidative stability. Food & Function. 2023;14(7):3290–3303. doi: 10.1039/d3fo00068k. [DOI] [PubMed] [Google Scholar]
  40. Zhong P., Yang Y., Han T., Huang W., Liu Y., Gong G.…Wang Z. Comparative analysis of free and glycoconjugates oligosaccharide content in Milk from different species. Journal of Agricultural and Food Chemistry. 2024;72(1):670–678. doi: 10.1021/acs.jafc.3c06317. [DOI] [PubMed] [Google Scholar]
  41. Zhu Y., Zhang J., Zhang W., Mu W. Recent progress on health effects and biosynthesis of two key sialylated human milk oligosaccharides, 3′-sialyllactose and 6′-sialyllactose. Biotechnology Advances. 2023;62 doi: 10.1016/j.biotechadv.2022.108058. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material

mmc1.docx (1.3MB, docx)

Data Availability Statement

Data will be made available on request.

Articles from Food Chemistry: X are provided here courtesy of Elsevier

RESOURCES