Network analysis of the proteome and peptidome sheds light on human milk as a biological system | Scientific Reports

Scientific Reports volume 14, Article number: 7569 (2024) Cite this article

1684 Accesses

1 Citations

Metrics details

Proteins and peptides found in human milk have bioactive potential to benefit the newborn and support healthy development. Research has been carried out on the health benefits of proteins and peptides, but many questions still need to be answered about the nature of these components, how they are formed, and how they end up in the milk. This study explored and elucidated the complexity of the human milk proteome and peptidome. Proteins and peptides were analyzed with non-targeted nanoLC-Orbitrap-MS/MS in a selection of 297 milk samples from the CHILD Cohort Study. Protein and peptide abundances were determined, and a network was inferred using Gaussian graphical modeling (GGM), allowing an investigation of direct associations. This study showed that signatures of (1) specific mechanisms of transport of different groups of proteins, (2) proteolytic degradation by proteases and aminopeptidases, and (3) coagulation and complement activation are present in human milk. These results show the value of an integrated approach in evaluating large-scale omics data sets and provide valuable information for studies that aim to associate protein or peptide profiles from biofluids such as milk with specific physiological characteristics.

Proteins in human milk have a wide variety of biological functions, ranging from nutrition to immune modulation1. Their synthesis can occur in the mammary epithelial cells (MECs), which is the case for the major milk proteins, such as the caseins and \(\alpha \)-lactalbumin (LALBA, also known as ALA)2. Other proteins are believed to be synthesized in other parts of the body and are subsequently transferred towards and through the MECs3. Shared location of synthesis, shared mechanism of transfer, or functioning in the same biological pathways can result in interdependencies between proteins4.

Parts of the amino acid sequence of proteins can, once detached from the original sequence, exert a completely different biological and biochemical activity. This detachment can occur during proteolytic degradation, resulting in peptides and free amino acids. In human milk, proteolytic degradation starts already when milk is secreted into the alveolar lumen5 and is due to proteolytic systems comprising proteases, protease activators, and protease inhibitors6. Active proteases, such as plasmin (PLG) and kallikrein, hydrolyze peptide bonds between amino acids in the protein sequence, disrupting the protein’s primary structure7.

It is known that peptides play a considerable role in many cellular processes in the body, for example, acting as hormones, cytokines, or growth factors8,9. Nevertheless, the role of peptides in human milk is not entirely understood yet. Some peptides can exert specific bioactivities, such as immunomodulatory, antimicrobial, antioxidative, or angiotensin-converting enzyme (ACE) inhibitory effects10,11. These bioactivities could be beneficial for protecting the mammary gland against infection but also have health benefits for the breastfed infant12. Although proteolytic degradation in the digestive system could result in the breakdown of bioactive peptides, specific peptide sequences might be protected against, or resistant to, further proteolytic degradation. In addition, new bioactive peptides may be formed upon enzymatic digestion in the infants’ gastrointestinal tract from either intact proteins or larger peptides11.

To date, several studies have investigated the human milk peptidome from a mechanistic perspective13,14, focusing on cleavage patterns and protease specificity. Although this has provided valuable insights into the human milk peptidome, much still needs to be discovered. Since peptides are a product of larger peptides or proteins, and since the proteolytic systems themselves are part of the proteome, it is expected that relationships exist between the proteome and peptidome. Analysis of these relationships in an integrated approach is an important step in increasing knowledge about the proteolytic activity in human milk.

This study aimed to investigate associations among proteins, among peptides, and between proteins and peptides in human milk. Proteomics and peptidomics profiles from 297 human milk samples, obtained using LC-MS/MS, were subjected to network analysis using Gaussian graphic modeling (GGM), and observed associations were discussed. The rationale behind this approach is that the associations observed in the GGM network can provide information about the biological function of the proteins and peptides and how they are formed or end up in the milk15,16. The resulting pairwise partial correlations enable a distinction between indirect and direct associations by adjusting for the contribution of all remaining variables17. The importance of studying human milk as a unique biological system has recently been emphasized by Christian et al. in a comprehensive perspective18. Human milk is the sole source of nutrition for infants, in a vulnerable period that is critical for development and health. Yet, little is currently known about the functionality of nutritional, biological, and immunological pathways of its components driving growth, development, and health in early life18. This research thereby provides an alternative way to integrate and interpret large-scale multi-omics data sets and shows the added value of studying associations in the human milk protein and peptide profile.

The LC-MS/MS analysis resulted in the identification of 1690 proteins and 9192 peptides originating from 48 precursor proteins.

After filtering the data on the requirement of identification in more than half of the samples, 480 proteins (Supplementary Table S1) and 1455 peptides (Supplementary Table S2) remained, with the peptides still originating from 48 precursor proteins (Supplementary Table S3). The relative contribution of the precursor proteins to the peptidome showed that the majority of the peptides originated from \(\beta \)-casein (38.5%), polymeric immunoglobulin receptor (PIGR) (10.5%), and butyrophilin subfamily 1 member A1 (BTN1A1, also known as BTN) (8.5%), a similar pattern as found in previous studies19,20.

A protein and peptide association network was inferred by the generation of GGMs. Edges were drawn in the network if partial correlations were significant (local fdr < 0.1). This resulted in an initial network containing 16,961 edges connecting 1895 nodes. This network summarizes the web of associations and interactions (between 448 protein and 1447 peptides) inferred from the abundance profiles of proteins and peptides. Molecular functions of proteins are often carried out through interaction with other proteins21. Similarly, peptides are essential players in human physiology, originating from, and interacting with, a diverse range of proteins and actively participating in numerous cellular functions22,23. For these reasons, the observed associations between proteins and peptides can be viewed as proxies for the study of underlying molecular mechanisms. Only 16,961 edges (corresponding to 0.9% of all possible edges between proteins and peptides) were found to be statistically significant, indicating a rather sparse and disconnected network as often observed in protein-protein interaction studies24.

Network representation of associations between proteins and peptides. The network is constructed by calculation of Gaussian graphical models (GGMs) and subsequent clustering with the Leiden community detection algorithm. Purple nodes represent proteins, and orange nodes represent peptides. The thickness of the edges is proportional to the partial correlation coefficients from the GGMs. Clusters with a significant overrepresentation of Gene Ontology (GO) annotation have a label in bold and are indicated with *. A high-resolution version of this figure, which includes labeling of the individual nodes, is provided as Supplementary Fig. S1.

Using the Leiden community detection algorithm (Fig. 1, Supplementary Fig. S1), 119 clusters were found containing proteins/peptides sharing similar patterns of association: 35 included both proteins and peptides, whereas 11 clusters only comprised proteins, and 73 clusters only comprised peptides. The biological relevance of these clusters is discussed in the next sections. Most connections (95.4%) were observed among features from the same data set, that is, cross-associations among proteins and among peptides.

GO annotation of proteins was used to investigate the overrepresentation of annotations in clusters of associated proteins. An overview of the 12 clusters for which the proteins showed significant overrepresentation of annotations can be found in Table 1. We present a discussion of a selection of these clusters.

A significant overrepresentation was found for Cluster 1 of proteins annotated with pentameric immunoglobulin M (IgM) complex (adjusted p = 0.016). Nevertheless, it is important to note that annotation for molecular function was only available for 6 out of the 44 proteins in this cluster. This is because most of these proteins are variable regions of immunoglobulins (Igs), for which GO annotation is unavailable. Besides the variable regions, this cluster also comprises the heavy constant regions of IgM and IgA, as well as the Ig J chain, which links multimeric IgA and IgM. Therefore, the associations between heavy chains, light chains, J chain, and variable regions indicate their physical relation as substructures of antibodies. In addition, the associations between IgA and IgM also indicate their common origin, since IgA and IgM in milk are produced mainly in the plasma cells in the mammary tissue25. Subsequent transepithelial transport of these proteins mediated by PIGR results in their secretion in milk26. Among the determinants of the human milk immunoglobulins are, e.g., maternal vaccination, smoking, psychological stress, and maternal or infant infection27.

Cluster 4 shows a significant overrepresentation (p = \(1.18\times 10^{-7}\)) of proteins commonly located in blood microparticles, which are microvesicles found in the blood. The cluster comprises 35 proteins, of which 17 are annotated as a component of blood microparticles. Among these are, for example, serum albumin (ALB), the major milk protease PLG, and protease inhibitors (4 serine protease inhibitors (SERPINs) and an inter-\(\alpha \)-trypsin inhibitors (ITIs)). Immunoglobulin G (IgG) is present in Cluster 4 as well. It is known that IgG in milk mainly originates from blood serum25, and the association with the blood proteins in Cluster 4 supports this.

It is generally assumed that PLG, which has an important role in blood coagulation, is blood-derived and transported into the milk from the systemic circulation28. In addition, a recent study has shown that one of the SERPINs in this cluster, SERPINA1 (also referred to as \(\alpha \)1-antitrypsin (A1AT)), is synthesized in the liver and enters the milk via direct transmission from the systemic circulation3. Considering the overrepresentation of proteins typically found in blood with functional characteristics in blood coagulation, it can be hypothesized that this cluster represents proteins originating from the systemic circulation, all being passively transported via a transcellular or paracellular pathway through the mammary epithelium. Furthermore, changes in the abundance of these proteins indicate a change in the relative importance of this transport. In reviewing the blood-milk-barrier (BMB) in cows, Wellnitz et al. refers to three possible mechanisms causing an increase in transport through the BMB29. First, the barrier can be impaired during inflammation when changes take place in the epithelial barrier. Second, the BMB can be impaired during lactation by oxytocin release or the physical force of suckling. Third, prolonged milk stasis is also known to impair tight junction permeability. An increase of the proteins in Cluster 4 might therefore indicate an impaired BMB due to one of the mechanisms described by Wellnitz et al.29.

Proteins that are known to be part of the milk fat globule membrane (MFGM) were found in Cluster 630. Within the epithelial cell, milk fat globules (MFGs) are surrounded by a single-layer membrane that comprises proteins, such as lactadherin (MFGE8), cluster of differentiation 36 (CD36), mucin 1 (MUC1), and BTN1A1. These proteins are believed to support the MFG in moving towards and binding to the apical plasma membrane31, which forms the outer bilayer of the MFGM after secretion. The clustering of the typical MFGM proteins (Cluster 6) and the GO overrepresentation of the bounding membrane of organelle as a cellular component (p = \(6.40\times 10^{-6}\)) confirms that these proteins are related to the membrane and thus have a common origin. It is known that one of the determinants of the fat content of human milk is the produced milk volume32. Fat content itself is related with MFG size, where a higher fat content results in larger MFGs33. These larger MFGs are easier to disrupt, resulting in an increased amount of dissociated MFGM and consequently MFGM proteins in the milk.

Cluster 7 shows an overrepresentation of proteins located in the extracellular space (p = \(7.52\times 10^{-3}\)). Among the proteins in this cluster are the major milk proteins, including both caseins and whey proteins, such as LALBA, \(\beta \)-casein, \(\kappa \)-casein, and lactoferrin (LTF, also known as LF). It is known that these proteins are synthesized in the mammary gland2, a process that is regulated by lactogenic hormones (insulin, prolactin, and glucocorticoids), and amino acids34. Therefore, considering the strong associations between these proteins, it can be assumed that their expression is related to activation of the translation machinery and can be distinguished from the other proteins. Interestingly, this Cluster 7 does not include \(\alpha \)S1-casein. Unlike the other casein subunits, \(\alpha \)S1-casein does not decrease over lactation like \(\beta \)-casein and \(\kappa \)-casein35. In addition, it was found that this protein is not uniquely expressed in the mammary gland but also in monocytes36. The abundance of this protein in milk might therefore be more dependent on other factors than hormonal regulation.

Cluster 13 shows an overrepresentation of lipoprotein particles as the cellular location of the proteins (p = \(2.42\times 10^{-9}\)). The cluster comprises 6 different apolipoproteins, including apolipoproteins A1, A2, A4, B100, D, and E. It was shown in a study with mice that lipoprotein particles could be transferred from serum, deliver cholesterol in the MEC, and be secreted into the milk37. Considering the overrepresentation of apolipoproteins, the associations found in Cluster 13 might be an indicator of this mechanism. To date, no research has been done to investigate the cause of variation of these proteins in milk. However, it might reflect the extent of cholesterol transport in the mammary gland, which is known to be related with lactation stage and maternal diet38,39.

Cluster 14 comprises, amongst others, 3 cathepsins (B, D, and Z), progranulin (GRN), N-acetylglucosamine-6-sulfatase (GNS), and legumain (LGMN). These are proteins typically found in the lysosomal lumen40, which is also revealed by the GO overrepresentation of the lysosome as a cellular component for this cluster (p = \(1.57\times 10^{-6}\)). In addition, the strong associations observed between these proteins suggest that these proteins are released into the alveolar lumen through a common mechanism such as lysosomal exocytosis41. Lysosomes play an important role in cellular homeostasis, development, and aging. This role therefore points to lactation stage being the driving factor for the abundance of these proteins in the milk, which is in line with Watson et al., who points to the importance of lysosomal enzymes in mammary gland involution42.

The proteins ezrin (EZR), radixin (RDX), and moesin (MSN) form together the ERM protein family. These ERM proteins can bind with the Na(+)/H(+) exchange regulatory cofactor (SLC9A3R1, also known as NHERF1), and it is known that both ERM and SLC9A3R1 can act as a crosslinker between the actin cytoskeleton and cell membranes by interaction with the intracellular domain of the apical membrane protein podocalyxin (PODXL)43,44. Together, these proteins play an essential role in tissue integrity45. The association of these proteins in Cluster 26 suggests a loss of apical membrane from the MECs. Surprisingly, these proteins do not cluster with the typical MFGM proteins (Cluster 6), even though the outer bilayer of the MFGM is formed from the apical membrane. This suggests that the apical membrane found in human milk does not originate only from the MFGM. Explanations for this can be ongoing cell renewal of the MECs, apoptosis, or even the frozen storage of the samples, which results in damaging cells present in the milk and, consequently, a release of parts of the apical membrane. A study carried out by Qu et al. shows that frozen storage results in increased levels of, amongst others, EZR, MSN, and SLC9A3R1 in milk46.

Cluster 110 shows an overrepresentation of ribosomal constituents. Very little is known about why ribosomal proteins are present in milk. They might originate from exosomes, apoptosis of MECs, or intact or damaged cells present in the milk. Nevertheless, their association shows that their levels in milk depend on similar driving factors and possibly share the same secretion mechanism or origin.

Overall, results indicate that the abundance of the majority of the proteins in human milk depends primarily on the pathway of entering the milk.

Our results show that clusters with peptides often comprise peptide ladders, differing only a few amino acids from the neighboring peptides (Table 2, Supplementary Table S4). The peptide ladders, which were found in peptide clusters, are presumably formed by aminopeptidases, which cleave a single amino acid off a peptide sequence (exoproteolysis). Several proteins with aminopeptidase activity were identified in the proteomics data of this study, which are in order of average abundance: dipeptidyl peptidase 2 (DPP7), cytosol aminopeptidase (LAP3), aminopeptidase B (RNPEP), aminopeptidase N (ANPEP), and leukotriene A-4 hydrolase (LTA4H). Although not all these aminopeptidases have been identified before in human milk, the activity of aminopeptidases in human milk has been evidenced47. The strong association observed between peptides of a peptide ladder suggests that this type of proteolytic degradation occurs in an abundance-dependent manner where the formation of a peptide depends on the abundance of its precursor.

Before cleavage of larger peptides by aminopeptidases is possible, initial proteolytic degradation of proteins must occur (endoproteolysis). It has been suggested that endogenous proteases, especially PLG, carry out such proteolysis13. PLG is a highly specific protease that hydrolyzes the peptide bond between lysine (K) or arginine (R) in the P1 position and any other amino acid in the P1’ position. This specificity matches with several outer C-terminal or N-terminal positions of the peptide ladders observed in the clusters (Supplementary Table S4). Further allocation of proteases to endoproteolytic cleavage sites remains speculative due to the overlapping specificity of proteases and the presence of less specific proteases. Nevertheless, the observed associations reveal signatures of endoproteolytic and exoproteolytic degradation of proteins and direct future studies in further investigation of the role of these two mechanisms in shaping the human milk peptidome.

We found that from all identified proteins with potential protease activity (n = 25), only 9 appeared in a cluster with peptides. The most probable explanation for the lack of strong associations between proteases and peptide clusters is the fact that the abundance of a protease is not necessarily equal to or related to its proteolytic activity in the natural milk environment. This can be due to, for example, the protease being present in the zymogen or inactive state, the pH of the milk, or the inhibition of proteases through protease inhibitors. Although most of the observed associations are between molecular features of the same type, that is, among proteins and among peptides, several interesting associations were found between proteins and peptides and will be discussed.

Network representation (circular layout) of a selection of clusters. The network is constructed by calculation of Gaussian graphical models (GGMs) and subsequent clustering with the Leiden community detection algorithm. Purple nodes represent proteins with their respective gene names, and orange nodes represent peptides with the gene name of the precursor protein and the respective sequence range. The thickness of the edges is proportional to the partial correlation coefficients from the GGMs. Selected clusters show associations between proteins and peptides and the selection of clusters is made from Fig. 1 with corresponding cluster labels.

We found that the fibrinogen chains that make up the fibrinogen complex (\(\alpha \) (FGA), \(\beta \) (FGB), and \(\gamma \) (FGG)), associated strongly with fibrinogen peptides (Fig. 2 Cluster 12). Fibrinogen is a protein complex synthesized in the liver, which plays a central role in blood coagulation. This coagulation is activated when fibrinopeptides are cleaved off enzymatically from both FGA and FGB by thrombin, resulting in the formation of fibrin and fibrin clots48. The fibrinopeptide of FGA was identified in the peptide data, suggesting that fibrinogen chains present in human milk can occur in the activated form, that is, as fibrin. To prevent clot formation, fibrin is degraded by PLG49, a process referred to as fibrinolysis. It is known that active PLG is present in human milk50, and several of the degradation products formed during fibrinolysis can be observed in Cluster 12 in Fig. 251. Furthermore, from all cleavage sites of the fibrinogen peptides, 50% matches the specificity of plasmin. It can be noted that FGA is more degraded, with 28 identified peptides, whereas FGB and FGG have 1 and 3 identified peptides, respectively. This matches with the fact that the FGA chain is cleaved first in the degradation of fibrin52. Additionally, \(\alpha \)-2-macroglobulin (A2M), a protease inhibitor that is known to regulate the degradation of fibrin by inhibition of PLG, also appears in Cluster 12. These findings indicate the presence and association of several components and degradation products of blood coagulation in human milk. The origin of these proteins and peptides remains an open question. One explanation might be that they are blood-derived and indirectly end up in the milk through, for example, damage of the mammary epithelial barrier. Nevertheless, it is more probable that they are part of the standard human milk composition since FGA was identified in 296 of the 297 samples. This also agrees with a study by Green et al., who investigated PLG-deficient mice and suggested that an accumulation of fibrin in the mammary gland could block mammary ducts and ultimately induce involution53. Our observation of positive associations between fibrinogen chains and their degradation products suggests that if more fibrinogen is present in the milk, more degradation occurs. From this, it can be hypothesized that the fibrinolysis pathway in milk is present to prevent blocked ducts and, therefore, to maintain lactation. Further research is required to identify the determinants of the presence of these proteins and peptides in human milk.

Cluster 25 comprises, among others, parathyroid hormone-related protein (PTHLH, also known as PTHRP) and 16 of its peptides. It has been suggested that PTHLH is involved in regulating calcium transport through the mammary gland54. After synthesis, PTHLH is degraded into three secretory forms, ranging from sequence position 37–72, 74–130, 143–175, respectively (signal peptide is included in the numbering of the sequence positions)55. It can be noted from Cluster 25 in Fig. 2 that peptides derived from all three secretory forms of PTHLH were identified and associated with the precursor protein. Although the functions and determinants of the different secretory forms of PTHLH in human milk are not known yet, our results show that they are all present in the secretory form in the milk and that their abundance depends on the abundance of intact PTHLH.

Cluster 51 (Fig. 2) shows the association between peptides from complement component 4 (C4) and the intact C4 isotypes C4A and C4B. These proteins are part of the complement system, a set of proteins, enzymes, and receptors in the blood that plays a key role in the innate immune system’s defense against pathogens. Several other proteins from this complex were identified in the proteomics data, including C3, C7, C9, plasma protease C1 inhibitor (SERPING1), and complement factor H (CFH). The presence of complement proteins in human milk has been evidenced before56 and can boost the protective mechanisms of the infants’ mucosae57. A recent study by Xu et al., showed in mice that the presence of complement components in milk plays an important role in modification of the gut microbiota and subsequently lysis of bacterial cells58. Nevertheless, it remains unknown what determines the abundance of these proteins and peptides in human milk, and further research is needed to examine this further. The identification of C4 in the current study (sequence coverage = 84%) covers regions between sequence positions 23 and 1682, whereas the total length of C4 is 1744 amino acids. This provides evidence for the presence of intact C4 in human milk. C4 can participate in the classical and lectin complement pathways and is cleaved into fragments upon activation59. In the peptide fraction, all but one of the C4 peptides that were identified originate from a specific region (between positions 1337 and 1449), which is the C-terminal part of the C4b fragment (position 757–1446). This C-terminal region of C4b is cleaved off in the formation of the C4d fragment (position 957–1336). Together, this shows that the identified C4 peptides are byproducts of the activation cascade of C451. The association of these peptides with intact C4 suggests that the presence of activated fragments of C4 in human milk depends on the abundance of intact C4.

Fibroblast growth factor-binding protein 1 (FGFBP1) is a protein that can bind fibroblast growth factors (FGFs), a family of cell signaling proteins, and release them from the extracellular matrix. All identified FGFBP1 peptides (n = 6) originate from the N-terminal region of the protein (between positions 24 and 51), which is also covered in the identification of the intact protein. Of the cleavage sites of the peptides, 15 out of 24 have lysine in position P1, suggesting that PLG is responsible for most of the cleavages. The strong association between the peptides and their intact protein (Fig. 2 Cluster 63) suggests a specific proteolytic degradation unrelated to the degradation of other proteins in milk. Such degradation might be related to the role of FGFBP1 in protecting FGF against degradation60, but this remains speculative since no previous studies were found on proteolytic degradation of FGFBP1.

Overall, the protein-peptide associations revealed several mechanisms of specific proteolytic degradation that take place in human milk. Specifically, degradation of fibrin(ogen), PTHLH, complement C4, and FGFBP1 was associated with the abundance of their precursor proteins. The degree of proteolysis of these proteins differs from the proteolytic degradation of the most abundant precursor proteins in milk.

Human milk samples from a selection of 300 mother-child dyads from the CHILD Cohort Study were used. The selection of these samples was made based on the allergy status of the mother and the infant, including equal numbers of different combinations of mother-child allergy statuses61. The information on allergy status was used in a previous investigation where specifically the relation between allergy status and the milk proteome was investigated61. In the current study the allergy status was not used in the data analysis. During the data analysis, 3 samples were omitted as outliers due to their distinct peptide profile. These samples showed a total peptide abundance several magnitudes higher than the average, possibly due to the occurrence of mastitis. The reported results concern therefore 297 samples.

Samples were analyzed in randomized order, with a technical replicate added randomly to every 7 injections. In addition, technical replicates were added as a control for technical variation and were prepared from a pooled human milk sample from the Dutch Human Milk Bank (Amsterdam, The Netherlands).

The CHILD Cohort Study is a Canadian national population-based cohort (https://www.childstudy.ca) in which information was collected over time from parents and their infants62. Pregnant mothers were recruited from the general population from Vancouver, Edmonton, Manitoba, and Toronto. Local Human Research Ethics Boards approved the study protocols, and the study was carried out following the Declaration of Helsinki. All parents provided written informed consent at the time of enrollment in the study. Milk samples were collected according to the CHILD protocol63. In short, foremilk and hindmilk samples expressed prior to and after feeding the infant were collected from several feedings during a day and were pooled to minimize within-feed variation and diurnal variation. Samples were collected between 6 and 35 weeks post-partum [median = 15.6 weeks, interquartile range (IQR) = 4.6]. Samples were stored at 4 \(^{\circ }\)C in the home refrigerator and, within 24 hours, picked up and transported on ice to the CHILD laboratory. There, samples were aliquoted and stored until further analysis at − 80 \(^{\circ }\)C. Further transport of the samples was done on dry ice. Temporal storage of the samples at 4 \(^{\circ }\)C allowed for post-expression protein degradation. However, post-synthesis protein degradation takes place within the mammary gland as well as before and after expression5. Milk with proteins partially degraded post-synthesis therefore reflects what the infant consumes. Furthermore, a study by Howland et al., showed that differences among individual mothers can still be detected after temporal storage at 4 \(^{\circ }\)C64.

The proteomics data used in this manuscript has been described in a previous manuscript61. In the current study, a stricter filtering of the data on missing values was applied. To aid the reader, we provide here a brief description of the sample preparation and analysis.

Skimmed milk was obtained by centrifugation at 10,000g and 4 \(^{\circ }\)C for 30 min. Then, skimmed milk was again centrifuged at 1000g and 4 \(^{\circ }\)C for 10 min to remove any remaining lipids. Finally, skimmed milk samples were prepared with filter-aided sample preparation for protein analysis as described before20.

Trypsin digested proteins were analyzed with LC-MS/MS as described before, with minor adjustments65.

The Andromeda search engine of the MaxQuant software v1.6.17.0 was used to analyze the raw LC-MS/MS data66. A database with protein sequences was created by an initial MaxQuant run using the full human proteome (downloaded from UniProtKB on 20-01-2021, n = 194,237)67. Protein identifiers obtained as identification from this initial run were used to create a human milk database for a second run (n = 24,175), in which also a cow milk protein (n = 1006) and an allergen protein database (n = 721) were added68. A full description of how these databases were obtained can be found in Dekker et al.68.

In MaxQuant, digestion specificity was set to Trypsin/P, with maximally 2 missed cleavages. A fixed propionamide modification was set for cysteines, and variable modifications for acetylation of the peptide N-term, deamidation of the side chains of asparagine and glutamine, and oxidation of methionine, with a maximum of 5 modifications per peptide were set. A leading protein was selected for each identified protein group as described elsewhere68. A false discovery rate of 1% was used at both peptide and protein levels. The first search had 20 ppm peptide tolerance, the main search 4.5 ppm tolerance, and the MS/MS fragment mass tolerance was 20 ppm. Label-free quantification (LFQ) was used to obtain protein abundances. Gene names were used to abbreviate protein names where appropriate.

Skimmed milk samples were prepared for peptide analysis as previously described20. In short, proteins were removed using precipitation. For this, an equal volume of 200 g/L trichloroacetic acid in milli-Q water was added, followed by centrifugation at 3000g for 10 min at 4 \(^{\circ }\)C. From the supernatant that was obtained, 50 \(\upmu \)L was cleaned up using solid phase extraction (SPE) on C18+ Stage tip columns (prepared in-house), as previously described19,69. Finally, eluted peptides were reconstituted in 50 \(\upmu \)L of 1 mL/L formic acid in water.

Peptides were analyzed with LC-MS/MS, using the same method as for the protein analysis described above. For the peptidomics analysis, 4 \(\upmu \)L of peptide solution was loaded onto the column.

The raw LC-MS/MS data files from the peptide analysis were processed similarly to the proteomics data. Differences were the digestion specificity which was set to unspecific without fixed cysteine modification and with variable modifications for acetylation of the protein N-term, deamidation of the side chains of asparagine and glutamine, and oxidation of methionine, with a maximum of 5 modifications per peptide. The sequence database which was created for the processing of the proteomics data containing human milk, cow milk, and allergen proteins, was used (as described above). Peptide length was set to a minimum of 8 and a maximum of 25 amino acids, to achieve the best compromise between computational time and complete identification. It was shown in a study by Dingess et al., that the majority of the peptides endogenously present in human milk are covered by this range in peptide length70. Raw intensities were used for further data analysis.

Statistical analysis and visualizations were, unless specified differently, carried out using R version 4.0.171.

MaxQuant proteinGroups (proteomics) or peptides (peptidomics) result files were filtered so that common contaminants were removed and only proteins and peptides that were identified in more than half (>150) of the samples were retained. In this way, a selection of the most prevalent and abundant proteins and peptides was used for further data analysis. Outliers (n = 3) were removed and in the remaining data, missing values were imputed using the GSimp package for R with default parameters72, which implements a Gibbs sampler-based algorithm to impute missing values with the assumption that missing values are not at random (MNAR) and left censored.

To investigate associations within and between the datasets, network analysis was applied to a combined data matrix comprising proteins (n = 456) and peptides (n = 1455) in 297 samples.

To build the network, partial correlations were estimated using Gaussian graphical modeling (GGM). The GGMs were built with a shrinkage-based regularization approach, which estimates the partial correlation coefficients in a pairwise manner. To build the GGMs, the ggm.estimator.pcor function from the GeneNet package for R was used71,73.

Partial correlation coefficients \(\rho \)ij describe the pairwise correlation between protein or peptide Xi and Xj after accounting for their correlation with all other proteins and peptides. This approach accounts for confounders and covariates, indirect associations often present in omics data sets, and enabled the study of direct associations among proteins, among peptides, and between proteins and peptides.

In the inference of the network, only significant edges were used. To determine the significance of the edges, the built-in empirical Bayes local false discovery rate (fdr) statistic was used74. Edges were considered significant if the probability of their “presence” was larger than 0.9 (which is equal to a local fdr < 0.1).

Adjacency matrices with partial correlations from the GGMs were visualized in networks using Cytoscape v3.9.175. In the GGM network, proteins and peptides are presented as nodes, and GGM-estimated partial correlations are the

edges between the nodes. Subsequent clustering of networks was performed using the Leiden algorithm76, through the clusterMaker2 plugin for Cytoscape77. For this clustering, Constant Potts Model was used as a quality function with a resolution parameter of 10-3, \(\beta \) value 0.01, and 1000 iterations. Clusters comprising more than 3 nodes were retained for further investigation.

To determine whether protein clusters were overrepresented with specific gene ontology (GO) annotations, the GORILLA tool (Gene Ontology enRIchment anaLysis and visuaLizAtion tool) (http://cbl-gorilla.cs.technion.ac.il/)78 was used. The two-list mode was used, with all identified proteins as the background set. p values were corrected with the Benjamini–Hochberg method79. An adjusted p value < 0.05 was considered significant.

The mass spectrometry proteomics and peptidomics data have been deposited to the ProteomeXchange Consortium via the PRIDE80 partner repository with the data set identifiers PXD034806 and PXD036477. Sample metadata can be made available upon request. Requests can be submitted via email to [email protected].

Donovan, S. M. Human milk proteins: Composition and physiological significance. In Nestle Nutrition Institute Workshop Series, Vol. 90, 93–101. https://doi.org/10.1159/000490298 (2019).

Vilotte, J. L. et al. Genetics and biosynthesis of milk proteins. In Advanced Dairy Chemistry, Vol. 1A: Proteins: Basic Aspects (McSweeney, P. L. H. & Fox, P. F., eds.), fourth edn, 431–461. https://doi.org/10.1007/978-1-4614-4714-6_14 (2013).

Jager, S. et al. Proteoform profiles reveal that alpha-1-antitrypsin in human serum and milk is derived from a common source. Front. Mol. Biosci. 9, 1–10. https://doi.org/10.3389/fmolb.2022.858856 (2022).

Article CAS Google Scholar

Vella, D., Zoppis, I., Mauri, G., Mauri, P. & Di Silvestre, D. From protein-protein interactions to protein co-expression networks: A new perspective to evaluate large-scale proteomic data. EURASIP J. Bioinf. Syst. Biol. 2017, 6. https://doi.org/10.1186/s13637-017-0059-z (2017).

Article CAS Google Scholar

Nielsen, S. D., Beverly, R. L. & Dallas, D. C. Milk proteins are predigested within the human mammary gland. J. Mammary Gland Biol. Neoplasia 22, 251–261. https://doi.org/10.1007/s10911-018-9388-0 (2017).

Article PubMed Google Scholar

Dallas, D. C., Murray, N. M. & Gan, J. Proteolytic systems in milk: Perspectives on the evolutionary function within the mammary gland and the infant. J. Mammary Gland Biol. Neoplasia 20, 133–147. https://doi.org/10.1007/s10911-015-9334-3 (2015).

Article PubMed PubMed Central Google Scholar

Fox, P. Enzymology of milk and dairy products: Overview. In Kelly, A. L. & Larsen, L. B. (eds.) Agents of Change: Enzymes in Milk and Dairy Products, 1–10. https://doi.org/10.1007/978-3-030-55482-8_1 (2021).

Schulte, I., Tammen, H., Selle, H. & Schulz-Knappe, P. Peptides in body fluids and tissues as markers of disease. Expert Rev. Mol. Diagn. 5, 145–157. https://doi.org/10.1586/14737159.5.2.145 (2005).

Article CAS PubMed Google Scholar

Foreman, R. E., George, A. L., Reimann, F., Gribble, F. M. & Kay, R. G. Peptidomics: A review of clinical applications and methodologies. J. Proteome Res. 20, 3782–3797. https://doi.org/10.1021/acs.jproteome.1c00295 (2021).

Article CAS PubMed Google Scholar

Nielsen, S. D., Beverly, R. L., Underwood, M. A. & Dallas, D. C. Release of functional peptides from mother’s milk and fortifier proteins in the premature infant stomach. PLoS One 13, e0208204. https://doi.org/10.1371/journal.pone.0208204 (2018).

Article CAS PubMed PubMed Central Google Scholar

Dallas, D. C. et al. A peptidomic analysis of human milk digestion in the infant stomach reveals protein-specific degradation patterns. J. Nutr. 144, 815–820. https://doi.org/10.3945/jn.113.185793 (2014).

Article CAS PubMed PubMed Central Google Scholar

Wada, Y. & Lönnerdal, B. Bioactive peptides derived from human milk proteins: An update. Curr. Opin. Clin. Nutr. Metab. Care 23, 217–222. https://doi.org/10.1097/MCO.0000000000000642 (2020).

Article CAS PubMed Google Scholar

Guerrero, A. et al. Mechanistic peptidomics: Factors that dictate specificity in the formation of endogenous peptides in human milk. Mol. Cell. Proteom. 13, 3343–3351. https://doi.org/10.1074/mcp.M113.036194 (2014).

Article CAS Google Scholar

Khaldi, N. et al. Predicting the important enzymes in human breast milk digestion. J. Agric. Food Chem. 62, 7225–7232. https://doi.org/10.1021/jf405601e (2014).

Article CAS PubMed PubMed Central Google Scholar

Altenbuchinger, M., Weihs, A., Quackenbush, J., Grabe, H. J. & Zacharias, H. U. Gaussian and mixed graphical models as (multi-)omics data analysis tools. Biochim. Biophys. Acta Gene Regul. Mech. 1863, 194418. https://doi.org/10.1016/j.bbagrm.2019.194418 (2020).

Article CAS PubMed Google Scholar

Hayashi, N. et al. Multiple biomarkers of sepsis identified by novel time-lapse proteomics of patient serum. PLoS One 14, 1–25. https://doi.org/10.1371/journal.pone.0222403 (2019).

Article CAS Google Scholar

Krumsiek, J., Suhre, K., Illig, T., Adamski, J. & Theis, F. J. Gaussian graphical modeling reconstructs pathway reactions from high-throughput metabolomics data. BMC Syst. Biol. 5, 21. https://doi.org/10.1186/1752-0509-5-21 (2011).

Article CAS PubMed PubMed Central Google Scholar

Christian, P. et al. The need to study human milk as a biological system. Am. J. Clin. Nutr. 113, 1063–1072. https://doi.org/10.1093/ajcn/nqab075 (2021).

Article CAS PubMed PubMed Central Google Scholar

Dingess, K. A. et al. Human milk peptides differentiate between the preterm and term infant and across varying lactational stages. Food Funct. 8, 3769–3782. https://doi.org/10.1039/c7fo00539c (2017).

Article CAS PubMed Google Scholar

Dekker, P. M., Boeren, S., Van Goudoever, J. B., Vervoort, J. J. & Hettinga, K. A. Exploring human milk dynamics: Interindividual variation in milk proteome, peptidome, and metabolome. J. Proteome Res. 21, 1002–1016. https://doi.org/10.1021/acs.jproteome.1c00879 (2021).

Article CAS Google Scholar

Richards, A. L., Eckhardt, M. & Krogan, N. J. Mass spectrometry-based protein–protein interaction networks for the study of human diseases. Mol. Syst. Biol. 17, 1–18. https://doi.org/10.15252/msb.20188792 (2021).

Article CAS Google Scholar

Stanfield, R. L. & Wilson, I. A. Protein–peptide interactions. Curr. Opin. Struct. Biol. 5, 103–113. https://doi.org/10.1016/0959-440X(95)80015-S (1995).

Article CAS PubMed Google Scholar

Lei, Y. et al. A deep-learning framework for multi-level peptide–protein interaction prediction. Nat. Commun. 12, 5465. https://doi.org/10.1038/s41467-021-25772-4 (2021).

Article ADS CAS PubMed PubMed Central Google Scholar

Liu, G. et al. Identifying protein complexes with clear module structure using pairwise constraints in protein interaction networks. Front. Genet. 12, 20. https://doi.org/10.3389/fgene.2021.664786 (2021).

Article CAS Google Scholar

Hurley, W. L. & Theil, P. K. Perspectives on immunoglobulins in colostrum and milk. Nutrients 3, 442–474. https://doi.org/10.3390/nu3040442 (2011).

Article CAS PubMed PubMed Central Google Scholar

Atyeo, C. & Alter, G. The multifaceted roles of breast milk antibodies. Cell 184, 1486–1499. https://doi.org/10.1016/j.cell.2021.02.031 (2021).

Article CAS PubMed Google Scholar

Rio-Aige, K. et al. The breast milk immunoglobulinome. Nutrients 13, 1810. https://doi.org/10.3390/nu13061810 (2021).

Article CAS PubMed PubMed Central Google Scholar

Kelly, A. L., O’Flaherty, F. & Fox, P. F. Indigenous proteolytic enzymes in milk: A brief overview of the present state of knowledge. Int. Dairy J. 16, 563–572. https://doi.org/10.1016/j.idairyj.2005.10.019 (2006).

Article CAS Google Scholar

Wellnitz, O. & Bruckmaier, R. M. Invited review: The role of the blood-milk barrier and its manipulation for the efficacy of the mammary immune response and milk production. J. Dairy Sci. 104, 6376–6388. https://doi.org/10.3168/jds.2020-20029 (2021).

Article CAS PubMed Google Scholar

Vanderghem, C. et al. Study on the susceptibility of the bovine milk fat globule membrane proteins to enzymatic hydrolysis and organization of some of the proteins. Int. Dairy J. 21, 312–318. https://doi.org/10.1016/j.idairyj.2010.12.006 (2011).

Article CAS Google Scholar

Monks, J. et al. Xanthine oxidoreductase mediates membrane docking of milk-fat droplets but is not essential for apocrine lipid secretion. J. Physiol. 594, 5899–5921. https://doi.org/10.1113/JP272390 (2016).

Article CAS PubMed PubMed Central Google Scholar

Nommsen, L., Lovelady, C., Heinig, M., Lönnerdal, B. & Dewey, K. Determinants of energy, protein, lipid, and lactose concentrations in human milk during the first 12 mo of lactation: The DARLING Study. Am. J. Clin. Nutr. 53, 457–465. https://doi.org/10.1093/ajcn/53.2.457 (1991).

Article CAS PubMed Google Scholar

Duan, B. et al. Correlations of fat content in human milk with fat droplet size and phospholipid species. Molecules 26, 1596. https://doi.org/10.3390/molecules26061596 (2021).

Article CAS PubMed PubMed Central Google Scholar

Rhoads, R. E. & Grudzien-Nogalska, E. Translational regulation of milk protein synthesis at secretory activation. J. Mammary Gland Biol. Neoplasia 12, 283–292. https://doi.org/10.1007/s10911-007-9058-0 (2007).

Article PubMed Google Scholar

Liao, Y. et al. Absolute quantification of human milk caseins and the whey/casein ratio during the first year of lactation. J. Proteome Res. 16, 4113–4121. https://doi.org/10.1021/acs.jproteome.7b00486 (2017).

Article CAS PubMed Google Scholar

Vordenbäumen, S. et al. Casein \(\alpha \) s1 is expressed by human monocytes and upregulates the production of GM-CSF via p38 MAPK. J. Immunol. 186, 592–601. https://doi.org/10.4049/jimmunol.1001461 (2011).

Article CAS PubMed Google Scholar

Monks, J. et al. A lipoprotein-containing particle is transferred from the serum across the mammary epithelium into the milk of lactating mice. J. Lipid Res. 42, 686–696. https://doi.org/10.1016/s0022-2275(20)31630-8 (2001).

Article CAS PubMed Google Scholar

Yang, Z. et al. Human milk cholesterol is associated with lactation stage and maternal plasma cholesterol in Chinese populations. Pediatr. Res. 91, 970–976. https://doi.org/10.1038/s41390-021-01440-7 (2022).

Article CAS PubMed Google Scholar

Zhang, N. et al. Temporal changes of phospholipids fatty acids and cholesterol in breast milk and relationship with diet. Eur. J. Lipid Sci. Technol. 122, 1900187. https://doi.org/10.1002/ejlt.201900187 (2020).

Article CAS Google Scholar

Lübke, T., Lobel, P. & Sleat, D. E. Proteomics of the lysosome. Biochim. Biophys. Acta Mol. Cell Res. 1793, 625–635. https://doi.org/10.1016/j.bbamcr.2008.09.018 (2009).

Article CAS Google Scholar

Tancini, B. et al. Lysosomal exocytosis: The extracellular role of an intracellular organelle. Membranes 10, 1–21. https://doi.org/10.3390/membranes10120406 (2020).

Article CAS Google Scholar

Watson, C. J. & Kreuzaler, P. A. The role of cathepsins in involution and breast cancer. J. Mammary Gland Biol. Neoplasia 14, 171–179. https://doi.org/10.1007/s10911-009-9126-8 (2009).

Article PubMed Google Scholar

Le Tran, N., Wang, Y. & Nie, G. Podocalyxin in normal tissue and epithelial cancer. Cancers 13, 2863. https://doi.org/10.3390/cancers13122863 (2021).

Article CAS PubMed PubMed Central Google Scholar

Jiang, L. et al. CLIC proteins, ezrin, radixin, moesin and the coupling of membranes to the actin cytoskeleton: A smoking gun?. Biochim. Biophysi. Acta Biomembranes 643–657, 2014. https://doi.org/10.1016/j.bbamem.2013.05.025 (1838).

Article CAS Google Scholar

Horrillo, A., Porras, G., Ayuso, M. S. & González-Manchón, C. Loss of endothelial barrier integrity in mice with conditional ablation of podocalyxin (Podxl) in endothelial cells. Eur. J. Cell Biol. 95, 265–276. https://doi.org/10.1016/j.ejcb.2016.04.006 (2016).

Article CAS PubMed Google Scholar

Qu, J. et al. Changes in bioactive proteins and serum proteome of human milk under different frozen storage. Food Chem. 352, 129436. https://doi.org/10.1016/j.foodchem.2021.129436 (2021).

Article CAS PubMed Google Scholar

Demers-Mathieu, V., Nielsen, S. D., Underwood, M. A., Borghese, R. & Dallas, D. C. Analysis of milk from mothers who delivered prematurely reveals few changes in proteases and protease inhibitors across gestational age at birth and infant postnatal age. J. Nutr. 147, 1152–1159. https://doi.org/10.3945/jn.116.244798 (2017).

Article CAS PubMed PubMed Central Google Scholar

Scheraga, H. A. The thrombin-fibrinogen interaction. Biophys. Chem. 112, 117–130. https://doi.org/10.1016/j.bpc.2004.07.011 (2004).

Article CAS PubMed Google Scholar

Horan, J. T. & Francis, C. W. Fibrin degradation products, fibrin monomer and soluble fibrin in disseminated intravascular coagulation. Semin. Thromb. Hemost. 27, 657–666. https://doi.org/10.1055/s-2001-18870 (2001).

Article CAS PubMed Google Scholar

Demers-Mathieu, V., Underwood, M. A. & Dallas, D. C. Premature delivery impacts the concentration of plasminogen activators and a plasminogen activator inhibitor and the plasmin activity in human milk. Front. Pediatr.https://doi.org/10.3389/fped.2022.917179 (2022).

Article PubMed PubMed Central Google Scholar

Koomen, J. M. et al. Direct tandem mass spectrometry reveals limitations in protein profiling experiments for plasma biomarker discovery. J. Proteome Res. 4, 972–981. https://doi.org/10.1021/pr050046x (2005).

Article CAS PubMed Google Scholar

Kirschbaum, N. E. & Budzynski, A. Z. A unique proteolytic fragment of human fibrinogen containing the A\(\alpha \) COOH-terminal domain of the native molecule. J. Biol. Chem. 265, 13669–13676. https://doi.org/10.1016/s0021-9258(18)77401-2 (1990).

Article CAS PubMed Google Scholar

Green, K. A., Nielsen, B. S., Castellino, F. J., Rømer, J. & Lund, L. R. Lack of plasminogen leads to milk stasis and premature mammary gland involution during lactation. Dev. Biol. 299, 164–175. https://doi.org/10.1016/j.ydbio.2006.07.021 (2006).

Article CAS PubMed Google Scholar

Seki, K. et al. Parathyroid-hormone-related protein in human milk and its relation to milk calcium. Gynecol. Obstet. Invest. 44, 102–106. https://doi.org/10.1159/000291496 (1997).

Article CAS PubMed Google Scholar

Plawner, L. L., Philbrick, W. M., Burtis, W. J., Broadus, A. E. & Stewart, A. F. Cell type-specific secretion of parathyroid hormone-related protein via the regulated versus the constitutive secretory pathway. J. Biol. Chem. 270, 14078–14084. https://doi.org/10.1074/jbc.270.23.14078 (1995).

Article CAS PubMed Google Scholar

Ogundele, M. O. Role and significance of the complement system in mucosal immunity: Particular reference to the human breast milk complement. Immunol. Cell Biol. 79, 1–10. https://doi.org/10.1046/j.1440-1711.2001.00976.x (2001).

Article CAS PubMed Google Scholar

Noel, G. et al. Human breast milk enhances intestinal mucosal barrier function and innate immunity in a healthy pediatric human enteroid model. Front. Cell Dev. Biol. 9, 1–15. https://doi.org/10.3389/fcell.2021.685171 (2021).

Article ADS Google Scholar

Xu, D. et al. Complement in breast milk modifies offspring gut microbiota to promote infant health. Cell 187, 750-763.e20. https://doi.org/10.1016/j.cell.2023.12.019 (2024).

Article CAS PubMed Google Scholar

Wang, H. & Liu, M. Complement C4, infections, and autoimmune diseases. Front. Immunol. 12, 1–15. https://doi.org/10.3389/fimmu.2021.694928 (2021).

Article CAS Google Scholar

Huang, W. et al. Sox12, a direct target of FoxQ1, promotes hepatocellular carcinoma metastasis through up-regulating Twist1 and FGFBP1. Hepatology 61, 1920–1933. https://doi.org/10.1002/hep.27756 (2015).

Article CAS PubMed Google Scholar

Dekker, P. M. et al. The human milk proteome and allergy of mother and child: Exploring associations with protein abundances and protein network connectivity. Front. Immunol. 13, 25 (2022).

Article Google Scholar

Subbarao, P. et al. The Canadian Healthy Infant Longitudinal Development (CHILD) study: Examining developmental origins of allergy and asthma. Thorax 70, 998–1000. https://doi.org/10.1136/thoraxjnl-2015-207246 (2015).

Article PubMed Google Scholar

Moraes, T. J. et al. The Canadian healthy infant longitudinal development birth cohort study: Biological samples and biobanking. Paediatr. Perinat. Epidemiol. 29, 84–92. https://doi.org/10.1111/ppe.12161 (2015).

Article CAS PubMed Google Scholar

Howland, V. et al. Impact of storage conditions on the breast milk peptidome. Nutrients 12, 2733. https://doi.org/10.3390/nu12092733 (2020).

Article CAS PubMed PubMed Central Google Scholar

Liu, Y., de Groot, A., Boeren, S., Abee, T. & Smid, E. J. Lactococcus lactis mutants obtained from laboratory evolution showed elevated vitamin K2 content and enhanced resistance to oxidative stress. Front. Microbiol.https://doi.org/10.3389/fmicb.2021.746770 (2021).

Article PubMed PubMed Central Google Scholar

Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 26, 1367–1372. https://doi.org/10.1038/nbt.1511 (2008).

Article CAS PubMed Google Scholar

Bateman, A. et al. UniProt: The universal protein knowledgebase in 2021. Nucleic Acids Res. 49, D480–D489. https://doi.org/10.1093/nar/gkaa1100 (2021).

Article ADS CAS Google Scholar

Dekker, P. M. et al. Maternal allergy and the presence of nonhuman proteinaceous molecules in human milk. Nutrients 12, 1169. https://doi.org/10.3390/nu12041169 (2020).

Article CAS PubMed PubMed Central Google Scholar

Lu, J. et al. Filter-aided sample preparation with dimethyl labeling to identify and quantify milk fat globule membrane proteins. J. Proteom. 75, 34–43. https://doi.org/10.1016/j.jprot.2011.07.031 (2011).

Article CAS Google Scholar

Dingess, K. A., van den Toorn, H. W., Mank, M., Stahl, B. & Heck, A. J. Toward an efficient workflow for the analysis of the human milk peptidome. Anal. Bioanal. Chem. 411, 1351–1363. https://doi.org/10.1007/s00216-018-01566-4 (2019).

Article CAS PubMed PubMed Central Google Scholar

Development Team Core. R. A language and environment for statistical computing (2020).

Wei, R. et al. GSimp: A Gibbs sampler based left-censored missing value imputation approach for metabolomics studies. PLoS Comput. Biol. 14, e1005973. https://doi.org/10.1371/journal.pcbi.1005973 (2018).

Article CAS PubMed PubMed Central Google Scholar

Schäfer, J., Opgen-Rhein, R. & Strimmer, K. Reverse engineering genetic networks using the GeneNet package. R News 6, 50–53 (2006).

Google Scholar

Efron, B. Large-scale simultaneous hypothesis testing: The choice of a null hypothesis. J. Am. Stat. Assoc. 99, 96–104. https://doi.org/10.1198/016214504000000089 (2004).

Article MathSciNet Google Scholar

Shannon, P. et al. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504. https://doi.org/10.1101/gr.1239303 (2003).

Article CAS PubMed PubMed Central Google Scholar

Traag, V. A., Waltman, L. & van Eck, N. J. From Louvain to Leiden: Guaranteeing well-connected communities. Sci. Rep. 9, 1–12. https://doi.org/10.1038/s41598-019-41695-z (2019).

Article CAS Google Scholar

Morris, J. H. et al. ClusterMaker: A multi-algorithm clustering plugin for Cytoscape. BMC Bioinform. 12, 1–14. https://doi.org/10.1186/1471-2105-12-436 (2011).

Article CAS Google Scholar

Eden, E., Navon, R., Steinfeld, I., Lipson, D. & Yakhini, Z. GOrilla: A tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinform. 10, 48. https://doi.org/10.1186/1471-2105-10-48 (2009).

Article Google Scholar

Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B (Methodol.) 57, 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x (1995).

Article MathSciNet Google Scholar

Vizcaíno, J. A. et al. 2016 update of the PRIDE database and its related tools. Nucleic Acids Res. 44, D447–D456. https://doi.org/10.1093/nar/gkv1145 (2016).

Article CAS PubMed Google Scholar

Download references

The authors dedicate the paper to the bright memory of Jacques J.M. Vervoort, a friend and recognized scientist, who passed away in July 2021. He contributed substantially to the conception and design of this study. The authors acknowledge the funding provided by the Netherlands Organization for Scientific Research (NWO-TTW) (project number 15299). The authors thank Bassel Dawod for his help with the samples and the method section and thank the CHILD Cohort Study (CHILD) participant families for their dedication and commitment to advancing health research. In addition, we are grateful to everyone without whom this study could not have been completed, including all members and staff of the CHILD Cohort Study. These include research staff, administrative staff, volunteers, lab technicians, statisticians, and clinical staff at the following institutions: McMaster University, University of Manitoba, University of Alberta, University of Toronto, and the University of British Columbia.

Food Quality and Design Group, Wageningen University and Research, Wageningen, 6708 WE, The Netherlands

Pieter M. Dekker & Kasper A. Hettinga

Laboratory of Biochemistry, Wageningen University and Research, Wageningen, 6708 WE, The Netherlands

Pieter M. Dekker & Sjef Boeren

Laboratory of Systems and Synthetic Biology, Wageningen University and Research, Wageningen, 6708 WE, The Netherlands

Edoardo Saccenti

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

K.A.H. and P.M.D. conceptualization; S.B. and P.M.D. investigation; E.S. and P.M.D. formal analysis; P.M.D. writing—original draft; K.A.H., E.S. and S.B. writing—review and editing; K.A.H. project administration; K.A.H. and E.S. supervision. All authors reviewed the manuscript.

Correspondence to Kasper A. Hettinga.

The authors declare no competing interests.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

Dekker, P.M., Boeren, S., Saccenti, E. et al. Network analysis of the proteome and peptidome sheds light on human milk as a biological system. Sci Rep 14, 7569 (2024). https://doi.org/10.1038/s41598-024-58127-2

Download citation

Received: 27 September 2023

Accepted: 26 March 2024

Published: 30 March 2024

DOI: https://doi.org/10.1038/s41598-024-58127-2

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative