Label-Free Liquid Chromatography–Mass Spectrometry Quantitation of Relative N- and O-Glycan Concentrations in Human Milk in Japan

Human milk is abundant in carbohydrates and includes human milk oligosaccharides (HMOs) and N/O-glycans conjugated to proteins. HMO compositions and concentrations vary in individuals according to the maternal secretor status based on the fucosyltransferase 2 genotype; however, the profile of N/O-glycans remains uninvestigated because of the analytical complexity. Herein, we applied a label-free chromatography–mass spectrometry (LC–MS) technique to elucidate the variation in the composition and concentration of N/O-glycans in human milk. We used label-free LC–MS to relatively quantify 16 N-glycans and 12 O-glycans in 200 samples of Japanese human milk (1–2 months postpartum) and applied high performance anion exchange chromatography with pulsed amperometric detection to absolutely quantify the concentrations of 11 representative HMOs. Cluster analysis of the quantitative data revealed that O-glycans and several HMOs were classified according to the presence or absence of fucose linked to galactose while N-glycans were classified into a different group from O-glycans and HMOs. O-glycans and HMOs with fucose linked to galactose were more abundant in human milk from secretor mothers than from nonsecretor mothers. Thus, secretor status influenced the composition and concentration of HMOs and O-glycans but not those of N-glycans in human milk.


Introduction
Human milk provides optimal nutrition for infants because of the bioactive substances present in addition to basic nutrients such as protein, fat, minerals, and vitamins.Human milk is abundant in carbohydrates, and these contain approximately 6% lactose and 0.5-2% human milk oligosaccharides (HMOs) in mature milk, which exceeds the concentration of total milk protein.Infants can digest lactose into glucose and galactose that are then absorbed in the small intestine as a vital infant energy source but cannot digest HMOs.HMOs instead have many biological functions, such as preventing pathogens from invading the intestinal epithelium [1][2][3], stimulating the growth of beneficial intestinal bacteria (e.g., Bifidobacterium, Bacteroides, and Lactobacillus) [4,5], developing the infant brain [6], and regulating infant immune systems [7].HMOs are complex glycans of which >150 have so far been identified and structurally characterized; however, not every woman synthesizes the same set of HMOs.The HMO composition depends on the expression of α1,2-fucosyltransferase (FUT2), which is encoded by the Se gene [8].Individuals with an active Se locus are classified as secretors.The milk of secretor women is abundant in 2 ′ -fucosyllactose (2 ′ -FL), lacto-N-fucopentaose-I (LNFP I), and other α1,2-fucosylated HMOs, although that of nonsecretor women that lack a functional FUT2 enzyme does not contain α1,2-fucosylated HMO [9].Additionally, other factors such as the lactation period, ethnic group, and nutritional and environmental aspects can affect the concentration and composition of HMOs [10][11][12][13].
Other than HMOs, carbohydrates in human milk exist in glycoconjugates, principally glycoproteins.The majority of human milk proteins (up to 70%) are heavily glycosylated [14] and include lactoferrin, α-lactalbumin, secretory immunoglobulin A, κ-casein, and several proteins of the milk fat globule membrane.The glycan moieties of these glycoproteins contribute to protein folding, biological recognition, and protection of proteins from digestion and have also been recently shown to protect infants against pathogens [15] and serve as substrates for bifidobacterial growth in the infant gut [16].Glycoprotein glycans are covalently attached to proteins primarily through Nand O-linked glycosylation, with the former occurring on asparagine residues and the latter occurring on serine or threonine residues.N-glycans are classified as high-mannose, complex, and hybrid types depending on their monosaccharide composition and branching.The O-glycans comprise eight O-glycan core types, and the nonreducing ends of N/O-glycans are frequently fucosylated and sialylated.Thus, N/O-glycans are highly structurally diverse.The structural diversity of N/O-glycans is presumed to vary among individuals, similar to the variation seen among HMOs, but the details are unknown.Human milk has been analyzed for the Nglycans [15,[17][18][19][20][21][22][23], O-glycans [24], or both N/O-glycans [25][26][27] present.Although human milk was provided by 4-30 women for these reports, no information concerning individual differences or the distribution of N/O-glycans was available because the individual milk samples were pooled before analysis.Therefore, to investigate individual differences in N/O-glycans, it is necessary to quantify N/O-glycans in each milk sample from a larger number of individuals.Furthermore, no studies concerning N/O-glycans in the milk of Japanese women have yet been conducted.
Liquid chromatography (LC)-mass spectrometry (MS) and LC-fluorescence detection (FLD) are valuable techniques in the identification and quantitation of N/O-glycans.Analysis of N/O-glycans using LC-MS or LC-FLD requires two pretreatment steps: the release of N/O-glycan from glycoproteins and the derivatization of released N/O-glycans using a fluorescent reagent [28].Performing these pretreatments on each of a large number of human milk samples is complex and time-consuming and consequently limits the sample size.Therefore, a simple and rapid pretreatment method is required.Previous reports showed that peptide-N-glycosidase F (PNGase F) could release N-glycans from glycoprotein for quantitative analysis via LC-MS without derivatization [17,18].For O-glycans, Kameyama et al. developed eliminative oximation that enables β-elimination of O-glycans from glycoproteins using hydroxylamine and 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) to suppress O-glycan degradation [29].The O-glycans released as a result of this procedure were analyzed using LC-FLD after labeling with 2-aminobenzoate and must also be analyzed using LC-MS without labeling because the N-acetylgalactosamine (GalNAc) at the reducing end of the O-glycan immediately formed an oxime upon release from the glycoprotein; the subsequent molecules are anticipated to be more ionized by electrospray ionization than native oligosaccharides.If the released N/O-glycans can be quantitatively analyzed via LC-MS without derivatization, the N/O-glycans in multiple specimens can be simply and rapidly analyzed.
Therefore, the purpose of this study was to elucidate the variation in the composition and concentration of N/O-glycans associated with those of HMOs in human milk using a label-free LC-MS technique.We relatively quantified N/O-glycans using label-free LC-MS methods and absolutely quantified the concentrations of representative HMOs via high performance anion exchange chromatography with pulsed amperometric detection (HPAE-PAD) in human milk samples from 200 healthy women in a cohort study in Japan (The Japanese Human Milk Study) [30].We performed a cluster analysis using quantitative data of N/O-glycans and HMOs.Our study demonstrates that a difference in O-glycan composition and concentration but not in those of N-glycan is present in individuals based on their secretor status.We examined the feasibility of using relative quantitation via label-free LC-MS methods with bovine fetuin as a model glycoprotein and slightly modified reported methods to analyze N-glycans [17,31].Briefly, detergent-denatured fetuin was digested using PNGase F, and the released N-glycans were analyzed via LC-MS and LC-MS/MS without further labeling.Consequently, 28 complex-type N-glycans were detected via LC-MS, and their structures were determined via MS/MS analysis and were consistent with those in a previous report [32].The LC-MS measurements showed a high linearity from 10 to 100 µg/mL fetuin (correlation coefficient, r > 0.995, Figure S1).Thus, the label-free LC-MS method was applicable to the relative quantitation of N-glycans.

Results and Discussion
To simplify the O-glycan analysis, we verified whether chemically released O-glycans could be detected and relatively quantified as oximated derivatives without the labeling reaction via LC-MS.O-glycan was released from fetuin using the releasing reagent containing NH 2 OH and DBU, and the oximated O-glycans were analyzed via LC-MS without further labeling.Three signals of m/z 690.25, 981.34, and 1346.48 were observed in the LC-MS chromatogram, and their structures were analyzed via LC-MS/MS.In the MS/MS spectrum at m/z 1346.48 (Figure S2C), a peak of oximated HexNAc at the reducing end (m/z 237.09) was observed, resulting from the sequential fragmentation of the glycan corresponding to the elimination of two Neu5Ac (−291 Da), two Hex (−162 Da), and one HexNAc (−203 Da).This MS/MS spectrum indicated a disialylated core2 Oglycan (Neu5Acα2,3Galβ1,3(Neu5Acα2,3Galβ1,4GlcNAcβ1,6)GalNAc) with oximation at the reducing end of GalNAc.The MS/MS spectra at m/z 690.25 and 981.34 also indicated oximated core1 monosialylated and disialylated O-glycans, respectively.These O-glycans were consistent with reported O-glycans of fetuin [29], indicating that the structural analysis of oximated O-glycans was feasible.The dependence of the LC-MS area of each oximated O-glycan on the fetuin concentration was then confirmed and showed high linearity (10-100 µg/mL fetuin; correlation coefficient, r > 0.988, Figure S2D).Thus, the label-free LC-MS analysis of oximated O-glycans was applicable for the relative quantitation of O-glycans.

Relative Quantitation of N/O-Glycans in Human Milk via Label-Free LC-MS
The relative quantitation of N/O-glycans in human milk using the label-free LC-MS methods was evaluated using the dilution assay with sequential diluted pooled milk.The relative MS area of 73 N-glycans and 33 O-glycans had an excellent linear relationship (r > 0.8) with the dilution rate of the pooled milk samples, demonstrating that the label-free LC-MS could be used for the relative quantitation of most of the N/O-glycans detected in this study (Tables 1 and 2).

Relative Quantitation of N/O-Glycans in 200 Samples of Japanese Human Milk via Label-Free LC-MS
Using the label-free LC-MS method, we quantified 16 N-glycans and 12 O-glycans with proposed structures except N06 in milk samples from 200 Japanese women.The quantifica-  1 and 2. Color symbols are as follows: green circle, mannose; yellow square, GalNAc; yellow circle, Gal; blue square, GlcNAc; purple diamond, Neu5Ac; and red triangle, Fuc.

Relative Quantitation of N/O-Glycans in 200 Samples of Japanese Human Milk via Label-Free LC-MS
Using the label-free LC-MS method, we quantified 16 N-glycans and 12 O-glycans with proposed structures except N06 in milk samples from 200 Japanese women.The quantification was performed by assessing the LC-MS peak area for each glycan.The abundance of each N/O-glycan was corrected using an internal standard (Table S1).To facilitate the comparison of N/O-glycan distributions among individuals, the corrected MS areas were normalized, as detailed in Table S2 and displayed in Figure S3.The median value for the concentration of each glycan was lower than the respective mean value, indicating that the distribution of each glycan concentration was nonsymmetric and positively skewed.For most glycans, some samples had a 5-20-fold higher concentration than the mean value.The coefficient of variation (CV) of each glycan concentration was 77-252% for N-glycan and 74-175% for O-glycan in the 200 individual samples of human milk.
The absolute concentrations of the 11 representative HMOs in the same 200 samples of human milk were measured using HPAE-PAD to compare the concentration distribution of N/O-glycans with that of HMOs (Tables S1 and S2).Each concentration range was consistent with previously reported values [34,35].The CV of each HMO was 43-114% (Figure S3).Comparing the CVs of N/O-glycans with those of HMOs, the concentrations of both N/O-glycans in human milk had a more significant distribution among individuals than those of HMOs.
The concentration of HMOs can vary significantly among Individuals.In this study, the N/O-glycans varied more significantly in concentration than the HMOs did.Although this large variance may be due to the variance in protein concentration in human milk since N/O-glycans are conjugated to proteins, we believe that the significant variance in N/O-glycan concentration is independent of the variance in protein concentration because the CV of the protein concentration in the 200 samples of human milk was 19.8%, which was significantly smaller than that of N/O-glycans.Factors, such as the amount of sugar nucleotide substrates and the activities of glycosyltransferases and glycosylhydrolases, influence the concentration of N/O-glycans [36].Further studies are, therefore, needed to elucidate the cause of the higher variance in N/O-glycans.

Comparison of N/O-Glycans and HMOs between Secretor Status
The relative concentration of N/O-glycan and HMOs was compared between secretor and nonsecretor milk (Figure 4, Table S3).All milk samples were categorized as secretor milk (2′-FL > 50 mg/L) or nonsecretor milk (2′-FL < 50 mg/L) according to the 2′-FL concentration [37].Of the 200 human milk samples, 161 (80.5%) were secretor milk and 39 (19.5%) were nonsecretor milk.Among the N-glycans and HMOs classified in group A, the concentration of four N-glycans (N12, N28, N42, and N58) and LNnT were significantly higher in secretor milk than that in nonsecretor milk, whereas the concentration of two N-glycans (N31, N36) and LNT was significantly lower than that in secretor milk.The concentration of all O-glycans and HMOs classified in group B, where Fuc is linked to Gal, was higher in secretor milk than in nonsecretor milk.Among the O-glycans and HMOs classified in group C, the concentration of five O-glycans (O8, O9, O24, O26, and O27) and three HMOs (3-FL, LNFP-II, and LNDFH-II) was higher in nonsecretor milk than that in secretor milk.None of the O-glycans or HMOs classified in group C had higher concentrations in secretor milk.All N-glycans were classified into group A and appeared to be subgrouped depending on the presence or absence of sialic acid in their structures, suggesting that the profile pattern of N-glycans in human milk was related to activity of sialyltransferase.O-glycans and fucosylated HMOs were classified into group B or C depending on whether Fuc was linked to Gal.The concentration of HMOs classified into group B and C is known to considerably differ between secretor and nonsecretor milk [9], whereas that in group A is less sensitive to secretor status, implying that classification into these three groups is dependent on the maternal secretor status.Therefore, O-glycans may have been classified into group B or C according to secretor status.These results suggested that the concentration of some O-glycans, as well as HMOs, is associated with secretor status.

Comparison of N/O-Glycans and HMOs between Secretor Status
The relative concentration of N/O-glycan and HMOs was compared between secretor and nonsecretor milk (Figure 4, Table S3).All milk samples were categorized as secretor milk (2 ′ -FL > 50 mg/L) or nonsecretor milk (2 ′ -FL < 50 mg/L) according to the 2 ′ -FL concentration [37].Of the 200 human milk samples, 161 (80.5%) were secretor milk and 39 (19.5%) were nonsecretor milk.Among the N-glycans and HMOs classified in group A, the concentration of four N-glycans (N12, N28, N42, and N58) and LNnT were significantly higher in secretor milk than that in nonsecretor milk, whereas the concentration of two N-glycans (N31, N36) and LNT was significantly lower than that in secretor milk.The concentration of all O-glycans and HMOs classified in group B, where Fuc is linked to Gal, was higher in secretor milk than in nonsecretor milk.Among the O-glycans and HMOs classified in group C, the concentration of five O-glycans (O8, O9, O24, O26, and O27) and three HMOs (3-FL, LNFP-II, and LNDFH-II) was higher in nonsecretor milk than that in secretor milk.None of the O-glycans or HMOs classified in group C had higher concentrations in secretor milk.All the O-glycans that were more abundant in secretor milk possessed Fuc linked to Gal and belonged to group B. Although HMOs with Fuc linked to Gal via FUT2 (2′-FL, LDFT, LNFP-I, and LNDFH-II) were rarely found in nonsecretor milk, O-glycans with Fuc linked to Gal such as O1, O20, and O28 were present in nonsecretor milk, albeit at low concentrations.Several of these O-glycans may be derived from blood glycoprotein.Alternatively, fucosyltransferase 1 (FUT1) could be involved in synthesizing these O-glycans in human milk [38][39][40].Conversely, all O-glycans that were more abundant in nonsecretor milk did not possess Fuc linked to Gal and belonged to group C. For N-glycans, the secretor status was unrelated to the presence or absence of Fuc in their structure.The findings indicate that secretor status has a notable impact not only on the composition and concentration of HMOs but also on those of O-glycans, instead of N-glycans.
In our results, the concentration of O-glycans and HMOs was significantly different between secretor and nonsecretor.However, although the concentration of some N-glycans was different between secretor and nonsecretor, no specific association between the structures of N-glycans and secretor status was observed.Contrary to our observations, previous reports have emphasized the impact of secretor status on N-glycans [21][22][23], highlighting a higher abundance of fucosylated and sialylated glycans in secretor milk.All the O-glycans that were more abundant in secretor milk possessed Fuc linked to Gal and belonged to group B. Although HMOs with Fuc linked to Gal via FUT2 (2 ′ -FL, LDFT, LNFP-I, and LNDFH-II) were rarely found in nonsecretor milk, O-glycans with Fuc linked to Gal such as O1, O20, and O28 were present in nonsecretor milk, albeit at low concentrations.Several of these O-glycans may be derived from blood glycoprotein.Alternatively, fucosyltransferase 1 (FUT1) could be involved in synthesizing these Oglycans in human milk [38][39][40].Conversely, all O-glycans that were more abundant in nonsecretor milk did not possess Fuc linked to Gal and belonged to group C. For N-glycans, the secretor status was unrelated to the presence or absence of Fuc in their structure.The findings indicate that secretor status has a notable impact not only on the composition and concentration of HMOs but also on those of O-glycans, instead of N-glycans.
In our results, the concentration of O-glycans and HMOs was significantly different between secretor and nonsecretor.However, although the concentration of some N-glycans was different between secretor and nonsecretor, no specific association between the structures of N-glycans and secretor status was observed.Contrary to our observations, previous reports have emphasized the impact of secretor status on N-glycans [21][22][23], highlighting a higher abundance of fucosylated and sialylated glycans in secretor milk.The discrepancy with previous reports may be because the N-glycans analyzed in this study did not include N-glycans with Fuc linked to Gal.Furthermore, the concentration of O-glycans have been reported to be higher in secretor milk [24], which is not entirely consistent with our results.This discrepancy may be due to differences in the ethnicity of the human milk donor used in this study or difference in sample size compared to previous reports.Secretor status affects infant health and development through changes in glycan composition [41,42].Differences in glycan structure have different effects on the bifidobacterial growth [19,22,43], infant gut microbiota [44], and immune response [45].Further studies are needed to elucidate the specific effects of glycan profiles in human milk on infants.This study may provide a basis for future research on the functional implications of such differences in glycan profiles and their potential impact on infants.

Human Milk
Human milk samples were provided by participants in the Japanese Human Milk Study registered in the University Hospital Medical Information Network Clinical Trials Registry (UMIN ID 000015494, 2014) [30].A total of 1210 lactating women were recruited; 83 did not provide valid responses, and 5 met the exclusion criteria, i.e., the women or their partners were not Japanese.Thus, 1122 women were included in our cohort study, of whom 930 provided sufficient (>50 mL) human milk at 1-2 months postpartum.This study used 200 human milk samples randomly selected from milk samples at 1-2 months postpartum provided by 930 participants.The milk samples were pooled by mixing equal volumes of the selected 200 milk samples and were stored at −80 • C.

Reagents
The EZ-Glyco O-glycan Prep kit was purchased from Sumitomo Bakelite (Tokyo, Japan).PNGase F (including protein denaturing buffer, NP-40, and glycobuffer) was purchased from New England Biolabs (Ipswich, MA, USA).Vivaspin 500 concentrators were purchased from Sartorius (Göttingen, Germany).Water and acetonitrile for LC-MS were purchased from Kanto Kagaku (Tokyo, Japan).Pure-grade HMO standards were purchased from Carbosynth Ltd. (Compton, UK).Other reagents were purchased from standard vendors.

Preparation of N-Glycans for Label-Free LC-MS Analysis
We combined and slightly modified previously reported methods [17,31] to enable label-free LC-MS analyses of N-glycans in a large number of human milk samples.Briefly, 3 mol/L sodium borohydride solution (10 µL) was added to a fetuin solution (20 µL) or human milk (20 µL) and left at ambient temperature for 30 min.Acetic acid (10 µL) was added to neutralize the mixture prior to incubation for 30 min at ambient temperature.Subsequently, the protein-denaturing buffer (10 µL) was added and the solution was heated at 100 • C for 10 min.Then, 1 mol/L ammonium bicarbonate solution (10 µL) and 123 mmol/L iodoacetamide (10 µL) were added, and the solution was incubated at ambient temperature for 1 h in the dark.Following the addition of NP-40 (10 µL), the reaction mixture was incubated with PNGase F (3 units) at 37 • C for 16 h.Subsequently, 450 nmol/L maltohexaose solution (125.5 µL), as an internal standard for LC-MS measurement, was added and centrifuged at 15,000× g for 10 min using a 10,000 molecular weight cut-off ultrafiltration membrane (Vivaspin 500) to remove proteins.The permeate was diluted twice using acetonitrile for LC-MS measurement.HMO standards.All milk samples were categorized as secretor milk (2 ′ -FL > 50 mg/L) or nonsecretor milk (2 ′ -FL < 50 mg/L) according to the 2 ′ -FL concentration [37].

Statistics
For the relative quantitation of N/O-glycans in 200 human milk samples, the corrected LC-MS areas for each N/O-glycan were normalized to a mean of 1.For hierarchical cluster analysis, the corrected LC-MS areas were standardized to a mean of 0 and a standard deviation of 1.All statistical analyses were performed in R version 3.6.2and RStudio program (version 2023.03.0,Integrated Development for R. RStudio, PBC, Boston, MA, USA).

Ethics
The study was conducted according to the recommendations of the Ethical Guidelines for Clinical Research (Ministry of Health, Labor, and Welfare, Japan) and was approved by the Ethics Committee of the Fukuda Internal Medicine Clinic (IRB No. 20140621-03).Written informed consent was obtained from participants in accordance with the Declaration of Helsinki.

Conclusions
Herein, we used label-free LC-MS methods to relatively quantify N/O-glycans in 200 samples of human milk at 1-2 months postpartum.This quantitative analysis of N/O-glycans in human milk demonstrated that secretor status not only influences the composition and concentration of HMOs but also those of O-glycans rather than those of N-glycans in human milk.We analyzed 200 individual human milk samples, whereas previous reports analyzed pooled milk samples or <10 individual milk samples.To our knowledge, this is the first comprehensive study of N/O-glycans in a large number (200) of human milk samples and provides important insights into the understanding of N/Oglycans in human milk.The relationship between the concentration of N/O-glycans in human milk and infant health and growth is particularly interesting.These questions will be addressed in future studies as information on infant health and development was also collected in this human milk cohort study.

2 . 1 .
Feasibility Study of Relative Quantitation of N/O-Glycans via Label-Free LC-MS

*
Correlation coefficient (r) between the relative LC-MS area of each glycan and the dilution rate of the pooled milk samples in the dilution assay.

Figure 1 .
Figure 1.Modification features of N-glycans (A) and O-glycans (B) detected in human milk via labelfree LC-MS.

Figure 1 .
Figure 1.Modification features of N-glycans (A) and O-glycans (B) detected in human milk via label-free LC-MS.

16 Figure 2 .
Figure 2. Structures of N/O-glycans and HMOs.The structures of 17 N-glycans and 12 O-glycans were characterized via LC-MS/MS analysis of composite human milk samples.The letters under each structure correspond to numbers in Tables1 and 2. Color symbols are as follows: green circle, mannose; yellow square, GalNAc; yellow circle, Gal; blue square, GlcNAc; purple diamond, Neu5Ac; and red triangle, Fuc.

Figure 2 .
Figure 2. Structures of N/O-glycans and HMOs.The structures of 17 N-glycans and 12 O-glycans were characterized via LC-MS/MS analysis of composite human milk samples.The letters under each structure correspond to numbers in Tables1 and 2. Color symbols are as follows: green circle, mannose; yellow square, GalNAc; yellow circle, Gal; blue square, GlcNAc; purple diamond, Neu5Ac; and red triangle, Fuc.

Figure 3 .
Figure 3. Heatmap of the relative concentration of N-and O-glycans and HMOs (columns) in 200 samples of human milk (rows) with dendrogram.The relative concentration of each glycan was standardized to a mean of 0 and a standard deviation of 1. Colors range from blue (minimum relative concentration) to red (maximum relative concentration) and white represents the mean value.The letters below the heatmap correspond to each glycan in Figure 2. The lower letters represent the main groups (A, B, C) and subgroups (A1, A2, A3) classified by the dendrogram.

Figure 3 .
Figure 3. Heatmap of the relative concentration of Nand O-glycans and HMOs (columns) in 200 samples of human milk (rows) with dendrogram.The relative concentration of each glycan was standardized to a mean of 0 and a standard deviation of 1. Colors range from blue (minimum relative concentration) to red (maximum relative concentration) and white represents the mean value.The letters below the heatmap correspond to each glycan in Figure 2. The lower letters represent the main groups (A, B, C) and subgroups (A1, A2, A3) classified by the dendrogram.

Table 1 .
N-glycans detected in human milk.
* Correlation coefficient (r) between the relative LC-MS area of each glycan and the dilution rate of the pooled milk samples in the dilution assay.Int.J. Mol.Sci.2024, 25, x FOR PEER REVIEW 5 of 16

Table 2 .
O-glycans detected in human milk.

Table 2 .
O-glycans detected in human milk.
* Correlation coefficient (r) between the relative LC-MS area of each glycan and the dilution rate of the pooled milk samples in the dilution assay.