Core Fucosylation of Maternal Milk N-Glycan Evokes B Cell Activation by Selectively Promoting the l-Fucose Metabolism of Gut Bifidobacterium spp. and Lactobacillus spp.

This study provides novel evidence for the critical role of maternal milk protein glycosylation in shaping early-life gut microbiota and promoting B cell activation of neonates. The special core-fucosylated oligosaccharides might be promising prebiotics for the personalized nutrition of infants.

. The total number of samples from secretor mothers was 43. The mode of delivery was vaginal for 100% of the mothers, and the mode of feeding was breastfeeding for 100% of the infants. Maternal secretor status was found to affect the gut microbiota composition of breast-fed infants. We examined the maternal secretor status and eliminated samples of nonsecretor mothers (n ϭ 13) and their infants (Table S1). The remaining secretors were grouped into the G and A groups according to their Fut8 gene status (SNP at rs10483776). b BMI, body mass index.
the N-glycans in these milk samples were heavily core fucosylated. The high core fucosylation level of milk protein was further demonstrated by lectin blotting with Aspergillus oryzae lectin (AOL), which preferentially recognizes core fucosylation on N-glycans (33). Major milk glycoproteins such as LF, IgA, and IgG are all highly core fucosylated (see Fig. S1A in the supplemental material). Interestingly, we found that the core fucosylation levels of milk proteins varied significantly among individual mothers (Fig. S1B). A previous study showed that the plasma fucosylation levels were highly correlated with SNPs in the human Fut8 gene (34), in which a G/A polymorphism of rs10483776 on chromosome 14 significantly affected plasma fucosylation levels of females. By DNA sequencing, the A¡G polymorphism of rs10483776 was found in 15 of the 56 Chinese mothers providing samples ( Fig. 1C; see Table S1 in the supplemental material). Moreover, the Fut8 enzyme activity of the cells in milk samples was increased in mothers within the "G" group (Fig. 1D). Importantly, the core fucosylation level of milk proteins from G group mothers was markedly higher than that of the "A" group mothers during different lactation stages (day 6, P ϭ 0.0004; day 42, P ϭ 0.0236; and day 120, P ϭ 0.0027), while the protein concentrations of milk samples within the two groups were similar ( Core fucosylation level of maternal milk N-glycan affects the gut microbiome of infants. Because the core fucosylation levels of milk N-glycan were significantly different among individual mothers, we investigated whether this variation affects the gut microbiota composition of their breast-fed infants. By high-throughout sequencing of gut bacterial genes, we analyzed the intestinal microbiome of infants at days 6, 42, and 120 after birth (Table 1). Because maternal Fut2 gene status was found to affect HMO abundance and the gut microbiota composition of breast-fed infants, we examined the maternal secretor status and eliminated samples of nonsecretor mothers and their infants (Table 1; Table S1). Of the remaining samples, a comparison of the total and major HMOs between the A and G groups resulted in no significant difference as detected by a newly developed liquid chromatography-mass spectrometry (LC-MS) system ( Fig. S3 [details in Materials and Methods]).
The 16S rRNA sequencing of fecal samples from 6-day-old infants obtained a total of 1,968,299 high-quality filtered reads, which were assembled into 82,012 effective tags per infant. After systematic analysis, we found that the alpha diversity of infants' gut microbiota reflected by Chao1 and the Shannon index showed no significant difference between the corresponding groups (Table S2 [all P values are Ͼ0.05). The dominant bacterial phyla in the gut of 6-day-old infants fed by G group mothers were Actinobacteria (47.89% Ϯ 15.45%) and Proteobacteria (31.57% Ϯ 16.71%), which constituted more than 79.46% of the total bacteria ( Fig. 2A). In contrast, the gut microbiota of infants fed by A group mothers was dominated by Proteobacteria (46.37% Ϯ 14.20%) and Firmicutes (33.73% Ϯ 15.43%), with a lower abundance of Actinobacteria (14.55% Ϯ 6.78%; P ϭ 0.0438 compared with G group). By day 42, the relative abundances of Actinobacteria (37.56% Ϯ 9.87%) and Proteobacteria (27.20% Ϯ 7.21%) still occupied the major proportion of total gut microbiota in infants fed by high-G-group mothers, and the average abundance of Bacteriodetes in these infants was 20.86% Ϯ 8.22%. In contrast, the abundances of Actinobacteria (25.27% Ϯ 3.15%) and Bacteriodetes (12.50% Ϯ 7.29%) were much lower in gut of infants fed by A group mothers (P ϭ 0.0285 and P ϭ 0.0327). By day 120, the relative abundance of Firmicutes had become the major phylum in infants of both groups (47.20% Ϯ 8.51% in the G group and 42.89% Ϯ 6.73% in the A group; P Ͼ 0.05), the abundance of Actinobacteria in the G group is 21.30% Ϯ 3.43%, which was no longer higher than that in the A group (25.35% Ϯ 5.57%; P Ͼ 0.05).
At the genus level (Fig. 2B), 6-day-old infants fed by mothers with higher corefucosylated milk (G group) harbored a higher abundance of Bifidobacterium (46.68% Ϯ 5.24%), whereas the relative abundance of Bifidobacterium (14.29% Ϯ 3.61%) was significantly lower (P ϭ 0.0028) in the gut of A group infants. By days 42 and 120, Bifidobacterium was still more abundant in infants fed by G group mothers than in those fed by A group mothers (day 42, 31.46% Ϯ 3.47% versus 23.97% Ϯ 2.96%, P ϭ 0.0462; day 120, 30.59% Ϯ 3.65% versus 25.36% Ϯ 5.83%, P Ͼ 0.05). In contrast, the abundance of Lactobacillus spp. in gut of infants was much lower compared than that of Bifidobacterium spp. during early lactation (day 6 to day 42); Lactobacillus occupied no more than 1.00% of the total bacteria. However, the proportion of this genus was largely increased by day 120, and we observed a significant difference between the G and A groups (43.19% Ϯ 7.89% versus The fecal microbiota profiles of infants fed by mothers with high (G group)-or low (A group)-core-fucosylated milk N-glycan at the phylum level based on 16S rRNA gene sequencing. (B) The fecal microbiota profiles of infants at genus level. (C) Characteristics of infant gut microbiota at days 6, 42, and 120 postpartum, illustrated by principal-coordinate analysis (PCoA) clustering analyses. Data from individuals (points) were clustered, and the centers of gravity were computed for each class. (D) The linear discriminant analysis effect size (LefSe) was adopted to identify the bacterial groups that showed significant differences in abundance between the high (G) and low (A) groups at day 6 postpartum. (E) Use of the Kyoto Encyclopedia of Genes and Genomes (KEGG) to evaluate the gut microbial functions between the high and low groups at day 42 postpartum. In order to screen the marker genes with significant differences between groups, the difference function between different groups was first detected by rank sum test and the dimension reduction was evaluated by LDA (linear discriminant analysis). The length of the histogram represents the influence of the difference function. (F) The Carbohydrate-Active enZYmes Database (CAZy) was used to evaluate the gut microbial functions of infants within the high and low groups at day 42 postpartum. GHs, glycoside hydrolases; GTs, glycosyl transferases; PLs, polysaccharide lyases; CEs, carbohydrate esterases; AAs, auxiliary activities; CBMs, carbohydrate-binding modules.
Core Fucosylation of Milk N-Glycan Evokes B Cell ® 14.21% Ϯ 3.71%; P ϭ 0.0256). The major Bifidobacterium and Lactobacillus spp. detected in gut of the infants include Bifidobacterium pseudocatenulatum, Bifidobacterium bifidum, Bifidobacterium breve, Lactobacillus gasseri, Lactobacillus mucosae, Lactobacillus casei, and Lactobacillus fermentum, etc.: the relative abundances of most of these species were much higher in G group infants than those in A group infants as detected at days 6, 42, and 120 postbirth (see Fig. S4 in the supplemental material).
By using principal-coordinate analysis (PCoA), we decomposed all the OTU data into two main factors that explained 67.72% (day 6), 70.41% (day 42), and 37.71% (day 120) of the variance (Fig. 2C). All the fecal samples of infants mainly clustered into the two groups were highly correlated with the maternal Fut8 gene phenotype (day 6, P ϭ 0.0026, day 42, P ϭ 0.04937, and day 120, P ϭ 0.0385, as analyzed by weighted UniFrac t test). The linear discriminant analysis (LDA) effect size (LEfSe) was further adopted to identify the bacterial groups that showed significant differences in abundance between groups. As shown in Fig. 2D, the phylum of Actinobacteria in infants fed by G group mothers at day 6 was significantly (LDA score ϭ 5.00) more abundant than that for the A group, which mainly contained B. pseudocatenulatum (LDA score ϭ 4.48). The genus Ammoniphilus and species of L. casei were also significantly more abundant in infants within the G group (LDA scores ϭ 4.02 and 4.32, respectively). In contrast, the key phylotypes (significantly more abundant microbial groups) detected in infants fed by mothers with low-core-fucosylated breast milk (A group) was the class of Clostridia (LDA score ϭ 4.49), which mainly contained a rich abundance of Tyzzerella (LDA score ϭ 4.00), a genus belonging to the family of Lachnospiraceae under the order of Clostridiales. In addition, Corynebacterium tuberculostearicum, part of the normal skin flora (35), was also abundant in this group (LDA score ϭ 4.11). Ammoniphilus and Corynebacterium are rarely found in human gut; thus, our observation may due to environmental and ethnic effects on our study cohort. No key phylotypes were detected between infants of the two groups by day 42. However, as the lactation duration increased, the Lactobacillus spp. (LDA score ϭ 4.95) became new biomarkers in the gut of infants fed by mothers with high-core-fucosylated milk, and the genera Pseudomonas (LDA score ϭ 4.00) and undefined Clostridiales (LDA score ϭ 4.47) were significantly more abundant in infants fed by mothers with low-core-fucosylated milk N-glycans.
Using the Kyoto Encyclopedia of Genes and Genomes (KEGG) and the Carbohydrate-Active enZYmes Database (CAZy), we further evaluated the gut microbial functions between the groups to identify the major enzymes/pathways involved in the metabolism of fucosylated milk N-glycans. Fifteen KEGG modules were preferentially enriched in feces of infants collected by day 42 postbirth (Fig. 2E; n ϭ 11 in the G group, and n ϭ 12 in the A group): in the G group, 12 were identified as ␤-galactosidase (K12308), putative ABC transport system permease protein (K10119), long-chain acyl coenzyme A (acyl-CoA) synthetase (K01897), implicating the upregulation of glycan hydrolysis, transportation, and biosynthesis of energy in gut microbes. The CAZy database analysis revealed that infants fed by G group mothers had enriched genes associated with glycoside hydrolases (GHs) and glycosyl transferases (GTs) (Fig. 2F).
Fut8 ؉/؊ maternal mouse-fed neonates have distinct gut microbiota. To examine the effects of milk N-glycan core fucosylation on offspring gut microbiota and health development, female mice with a heterozygous Fut8 gene (Fut8 ϩ/Ϫ ) were adopted as maternal models with low-core-fucosylated milk N-glycan compared with wild-type (Fut8 ϩ/ϩ ) mice, and their effects on Fut8 ϩ/ϩ offspring were evaluated (Fig. 3A). As expected, the core fucosylation level of N-glycans in the breast milk of Fut8 ϩ/Ϫ mice was reduced compared with that in Fut8 ϩ/ϩ mice (Fig. 3B). The activity of Fut8 in Fut8 ϩ/Ϫ mice was about half that of Fut8 ϩ/ϩ mice (Fig. 3C). After lactation for 3 weeks, the body weight of offspring mice fed by Fut8 ϩ/Ϫ mice was similar to those of offspring fed by Fut8 ϩ/ϩ mice ( Fig. 3D; P Ͼ 0.05, n ϭ 3). However, alterations in the gut microbiota of the offspring fed by Fut8 ϩ/Ϫ mice were observed. As shown in Fig. 3E, the PCoA revealed a distinct clustering of microbiota composition between Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ mouse-fed offspring, suggesting a significant difference in their beta diversity (P ϭ 0.0363, weighted UniFrac t test). The major bacterial phyla in offspring fed by Fut8 ϩ/ϩ mice were Bacteroidetes (63.13% Ϯ 5.85%), Firmicutes (33.21% Ϯ 5.32%), and Proteobacteria (3.16% Ϯ 0.38%) (Fig. 3F). Notably, the Bacteroidetes group (especially Bacteroides acidifaciens), which can ferment a wide range of sugar derivatives (36), was significantly affected by the reduction of core-fucosylated N-glycan in the milk of Fut8 ϩ/Ϫ mice (33.30% Ϯ 6.89%, Female mice with a heterozygous Fut8 gene (Fut8 ϩ/Ϫ ) were adopted as maternal models with low-core-fucosylated milk N-glycan compared with wild-type mice (Fut8 ϩ/ϩ ), and their effects on Fut8 ϩ/ϩ offspring were evaluated. (B) The breast milk proteins of mice were analyzed by SDS-PAGE and AOL blotting (1:8,000). The Coomassie brilliant blue (CBB) staining of gels shows comparable amounts of whole-protein lysates in each sample. (C) Fut8 enzyme activities in the cells of mammary gland from maternal mice by detected by HPLC. (D) Photos of the neonatal mice fed by Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ mothers after 3 weeks of lactation, and their body weights during lactation were compared. Data are reported as mean Ϯ SEM. n.s., no significant differences between groups were detected. (E) The fecal microbiota profiles of Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring at the phylum level based on 16S rRNA gene sequencing. (F) The PCoA revealed distinct clustering of microbiota composition between Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring. (G) The relative abundance of fecal microbial groups of Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring at the genus level.
Fut8 ؉/؊ maternal mouse-fed neonates showed lower proportion of splenic CD19 ؉ CD69 ؉ B lymphocytes and attenuated humoral immune response. Since aberrant neonatal microbiota composition can elicit abnormal immune responses, we further investigated the population of lymphocytes in spleen and thymus of offspring mice. A significantly reduction in the total lymphocyte numbers in the spleen and thymus of Fut8 ϩ/Ϫ mouse-fed offspring were detected compared with those fed by Fut8 ϩ/ϩ mice (see Fig. S6 in the supplemental material [P ϭ 0.0076 and P ϭ 0.0009; n ϭ 3]). Flow cytometry analysis revealed that the frequencies of B cells (CD19 ϩ ) and macrophages (F4/80 ϩ ) were significantly reduced in Fut8 ϩ/Ϫ mouse-fed offspring (P ϭ 0.0454 and P ϭ 0.0233; n ϭ 3), whereas those of TER119 ϩ cells were increased ( Fig. 4A [P ϭ 0.0439; n ϭ 3]), The proportions of CD4 ϩ T cells, CD8 ϩ T cells, Gr-1 ϩ cells, CD11C ϩ cells, and NK cells (DX5 ϩ ) in these mice were comparable with control. As B cells play important roles in humoral immunity by secreting antibodies, we therefore immunized the neonates after breastfeeding for 3 weeks by Fut8 ϩ/Ϫ or Fut8 ϩ/ϩ mice (Fig. 4B) and compared the serum IgG levels of them to each other. As shown in Fig. 4C, significantly lower level of serum IgG in Fut8 ϩ/Ϫ mouse-fed offspring (P ϭ 0.0421; n ϭ 3) was found compared with Fut8 ϩ/ϩ mouse-fed offspring post-ovalbumin (OVA) immunization, which was in accordance with the lower proportion of B cells observed ( Fig. 4D [P ϭ 0.0168; n ϭ 3]). However, when we treated the offspring mice with a minimally effective dose of mixed antibiotics (mixAbx), which effectively decreases gut microbiota without affecting B cell survival in vitro (Fig. 4E) (11), no difference in splenic B cell proportions was observed between Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ mouse-fed offspring ( Fig. 4F [P Ͼ 0.05]). These results suggested that the regulatory role of core-fucosylated milk N-glycan on humoral immunity of neonates was gut microbiota dependent.
Metabolites of L-fucose by Lactobacillus and Bifidobacterium spp. evoked B cell activation in vitro. Previous studies (37,38) showed that the metabolism of L-fucose in Bifidobacterium and Lactobacillus spp. harboring a L-fucose operon resulted in the production of 1,2-propanediol and L-lactate. Indeed, when we cultured B. pseudocatenulatum CGMCC1.5001, L. casei ATCC 334, and L. gasseri ATCC 33323 strains with L-fucose, we detected the growth of these strains and the production of 1,2propanediol and lactate (Fig. 5A). Next, we asked whether the metabolites of L-fucose can exert effects on B lymphocytes in the Peyer's patches (PPs) of mice. To test this, we isolated total lymphocytes from the PPs of mice and incubated them with 1,2propanediol and lactate. The results showed that the frequencies of B cells (CD19 ϩ ) and activated B cells (CD69 ϩ CD19 ϩ ) (39) in PPs were markedly increased following stimulation with these metabolites (Fig. 5B [all P values are Ͻ0.0001; n ϭ 3]).
The recognition of antigen by the B cell receptor (BCR) complex on the surface of B cells triggers signaling cascades via spleen tyrosine kinase (Syk) that ultimately lead to B cell activation and development (40). To further elucidate the role of L-fucose metabolites in B cell activation, we incubated 1,2-propanediol and lactate with 3-83 B cells (expressing IgG2a-BCR recognizing p31), and assessed the phosphorylation level of Syk. Compared with 3-83 cells, the phosphorylation levels of Syk, as can be induced by p31, were upregulated upon treatment with L. gasseri supernatant and cell lysates . Consistent with the change in B cell activation among PP cells, the L-fucose metabolites of L. gasseri enhanced the phosphorylation of Syk at a dose of 10 g/ml. However, the incubation of 3-83 B cells with 1,2-propanediol or lactate did not result in the increased phosphorylation of the p42 and p44 isoforms of mitogen-activated protein kinase (MAPK) (Erk1 and Erk2, respectively) (Fig. 5D). This . Two weeks later, mice were immunized with 50 g of OVA by subcutaneous injection. Mice sera were collected before and 14 days post-OVA immunization. (C) Comparison of the levels of serum IgG in Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring after immunization. The concentrations of IgG in the sera of mice (n ϭ 3/group) were measured by enzyme-linked immunosorbent assay (ELISA) using mouse MAb isotyping reagents. Data are shown as mean Ϯ SEM (n ϭ 3; *, P Ͻ 0.05). (D) Flow cytometry analysis of the CD3 ϩ and CD19 ϩ cell proportions in spleen of Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring after immunization. Data are shown as mean Ϯ SEM (n ϭ 3; *, P Ͻ 0.05). (E) Experimental design. Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring were orally administered with a minimally effective dose of antibiotics (mixAbx: ampicillin, neomycin, and metronidazole at 40 mg/liter and vancomycin at 20 mg/liter) from the 10th day to the 21st day after birth. (F) Flow cytometry analysis of the CD3 ϩ and CD19 ϩ cell proportions in spleen of Fut8 ϩ/ϩ and Fut8 ϩ/Ϫ maternal mouse-fed offspring after immunization. Data are shown as mean Ϯ SEM (n ϭ 3). n.s., no significant differences were detected between groups. Core Fucosylation of Milk N-Glycan Evokes B Cell ® suggests that these metabolites activate B cells through BCR signaling pathway, but not MAPK pathway. Furthermore, we adopted a new Biacore strategy to analyze the molecular interactions between L-fucose metabolites and BCR. 1,2-Propanediol and lactate were found to interact directly with BCR protein at a concentration range of 1.37 to ϳ12.3 M (Fig. 5E), which further supported our hypothesis that the L-fucose metabolites of Bifidobacterium and Lactobacillus spp. induce B cell proliferation and activation through the BCR-mediated signaling pathway. Core-fucosylated oligosaccharides promoted the growth of Bifidobacterium and Lactobacillus spp. Lactobacillus and Bifidobacterium spp. were more abundant in high-core-fucosylated milk-fed infants and Fut8 ϩ/ϩ mouse-fed offspring, suggesting that these bacteria might preferentially use core-fucosylated milk N-glycans. Fuc-␣1,6-GlcNAc-GlcNAc and Fuc-␣1,6-GlcNAc are basic fucosyl structures on milk N-glycan (Fig. 6A). Therefore, we chemically synthesized Fuc-␣1,6-GlcNAc-GlcNAc and Fuc-␣1,6-GlcNAc (41) (Fig. 6B and C), and cultured several Bifidobacterium and Lactobacillus strains (see Table S3 in the supplemental material) in liquid minimal medium supplemented with Fuc-␣1,6-GlcNAc and Fuc-␣1,6-GlcNAc-GlcNAc as the sole carbon source. Interestingly, most bacteria grew in the Fuc-␣1,6-GlcNAc-GlcNAc-supplemented medium (Fig. 6D), but only some of the Lactobacillus spp. (L. casei ATCC 334 and L. gasseri ATCC 33323) survived in Fuc-␣1,6-GlcNAc-supplemented medium (Fig. 6E). These results suggest that Lactobacillus strains have superior ability to hydrolyze the ␣1,6linkage-bond fucose.

DISCUSSION
Human milk glycans provide a broad range of carbon sources for gut microbes in infants. In addition to HMOs, there are numerous N-glycans in human milk proteins, Core Fucosylation of Milk N-Glycan Evokes B Cell ® which may impact the gut microbial composition of infants. Core fucosylation of N-glycan is the most common posttranscriptional modification of proteins in mammals. To our knowledge, this is the first study to systematically investigate the correlation between maternal milk glycosylation and the gut microbiome of infants.
Fucosylated glycans have marked effects on human health through manipulation of gut microbiota (42,43). Many studies have found that the status of mothers' secretor gene Fut2 or Lewis blood-type-related gene Fut3 could influence the ␣1,2/3or ␣1,3/4-fucosylation levels of HMOs and in turn affect the gut microbiota of their breast-fed infants (15). Although the fucosylation of N-glycans in milk is also modified by Fut2 and Fut3, but according to the glycoforms of human milk proteins detected by multiple reaction monitoring (32), core fucosylation is the major fucosylation form of human milk N-glycans, this was also proved by our mass spectrometry detection in milk samples of Chinese mothers. By lectin blotting of milk protein, we found that the Fut8 gene SNP (A¡G mutation at rs10483776) of the enrolled Chinese mothers strongly affected the fucosylation level of their milk N-glycans. On the other hand, when measuring HMOs in these milk samples, we did not find any difference in the abundances of total and major HMOs between groups A and G, which suggested that the A¡G mutation in the Fut8 gene did not affect the fucosylation levels of HMOs. It is not a surprise that the fucosylation of HMOs is mainly catalyzed by Fut2 and Fut3, while the ␣1,6-fucosylated glycotopes on milk glycoproteins (catalyzed by Fut8) were found to be absent on HMOs (44,45): thus, the differences in the gut microbiomes observed in infants fed by the G and A groups of mothers were attributed to the different core fucosylation levels of milk N-glycans.
In infants fed by mothers with high-fucosylated milk N-glycans, the Bifidobacteria, mainly including B. pseudocatenulatum, B. bifidum, and B. breve, were dominant during early lactation, which markedly contributed to the high proportion of Actinobacteria in infants that possess the ability to use fucosylated glycans (46). These Bifidobacterium spp. can be shared by mothers and their offspring as a source of probiotics (47,48). Furthermore, they have antagonistic activities on Gram-negative enteric pathogens (49) and can improve the inflammatory status of insulin-resistant obese children (50). Other more abundant microbial groups in infants fed by G group mothers, especially during late lactation, included L. casei, L. mucosae, and L. gasseri, which are usually isolated from infant feces (51,52). L. casei possesses a strong ability to ferment milk glycans by the expression of various GHs and gene clusters (53,54). This strain also shows many beneficial effects on infants such as growth promotion (55) and inhibition of pathogens (56,57). L. mucosae strains have been shown to decrease epithelial permeability and improve epithelial barrier function. The presence of this organism provides competitive exclusion against many pathogenic organisms and help with the development of new probiotic food products (58). L. gasseri was found attenuated allergic airway inflammation through PPAR␥ (peroxisome proliferator-activated receptor gamma) activation in dendritic cells (59). On the contrary, the gut of infants fed by mothers with lowfucosylated-milk N-glycans were dominated by the order Clostridiales (at day 6) or the genera of undefined Clostridiales and Pseudomonas (at day 120). Pseudomonas can be an oral and enteric pathogen found dominant in preterm infants (60); overgrowth of these bacteria may increase the risk of infection under immunocompromised conditions (61). Clostridiales are in high prevalence in the first 2 weeks of life (62): this was unfortunately known for a few pathogenic species that include Clostridium botulinum, Clostridium perfringens, Clostridium tetani, and Clostridium difficile in the family Clostridiaceae (63). However, in our study, Clostridiales were largely overlooked because of the difficulties to culture the organisms in vitro. Further understanding of the selective nutritional requirement that favor the growth of these bacteria would be helpful to study their effects on intestinal maturation and health outcomes in infants.
Mouse models can complement human studies and provide further insights into how core fucosylation of milk N-glycans influences the gut microbiota composition and the health outcome of breast-fed infants. Interestingly, a greater abundance of Lactobacillus spp. (mainly L. gasseri and L. reuteri), Bacteroides spp., and Bifidobacterium spp. was detected in the neonates fed by Fut8 ϩ/ϩ mice compared with those fed by Fut8 ϩ/Ϫ mice. Bacteroides spp. are glycan consumers (36): thus, our results also suggested their ability to utilize fucosylated milk N-glycans. In contrast, the Fut8 ϩ/Ϫ mouse-fed neonates had more abundance of members of the Lachnospiraceae NK4A136 group and Akkermansia spp. than those fed by Fut8 ϩ/ϩ mice. Lachnospiraceae abundance in gut of infants was found to be significantly associated with higher body mass index (BMI) and with increased odds of being overweight or obesity (64). However, the presence of the Lachnospiraceae NK4A136 group was negatively correlated with intestinal inflammation (65). Thus, the functions of this group of bacteria deserve further investigation. Akkermansia, especially the species Akkermansia muciniphila, is known as an intestinal mucin-degrading bacterium (66). Although positive correlations were observed between fucosylated HMOs and A. muciniphila (67), this genus was also found to be significantly enriched in infants with eczema (68). It was speculated that higher abundance of A. muciniphila in eczematous infants might reduce the integrity of intestinal barrier function and therefore increase the risk of developing eczema. Based on the above information, our results found in the neonatal mice suggested an alteration of gut microbiota toward a probably unhealthy pattern in response to changes in core fucosylation levels of maternal milk N-glycans.
As the aberrant neonatal microbiota composition can elicit abnormal immune responses, we therefore further investigated the major immune organs of the neonates fed by Fut8 ϩ/ϩ or Fut8 ϩ/Ϫ mice. Indeed, a significantly smaller population of total lymphocytes in the thymus and spleen of neonates fed by Fut8 ϩ/Ϫ mice was detected compared with those fed by Fut8 ϩ/ϩ mice. These neonates also exhibited a significantly lower proportion of active B cells in their spleen, resulted in attenuated IgG production after OVA stimulation. Of concern, since microbiota suppression by MixAbx treatment in the offspring mice hindered the development of B cells, regardless of whether the offspring were fed by Fut8 ϩ/ϩ or Fut8 ϩ/Ϫ mice, it is reasonable to think that the core-fucosylated milk N-glycan may regulate neonatal B cell responses through gut microbiota modulation.
Kim et al. (11) reported that gut microbiota-derived SCFAs activated B cells by increasing acetyl-CoA and regulating metabolic sensors to increase oxidative phosphorylation, glycolysis, and fatty acid synthesis, which produce energy and factors required for antibody production, but how B cells interact or import these factors remain unclear. Interestingly, in our study, stimulation of 1,2-propanediol and lactate, the L-fucose metabolites of B. pseudocatenulatum, L. casei, and L. gasseri, to mouse PP-derived lymphocytes resulted in a marked elevation of B cell activation. These suggested a novel way of probiotic bacteria in immune regulation other than through the production of SCFAs. BCR-mediated immune responses to antigen (Ag) stimulation regulate several biological functions, including B cell activation and differentiation. To further study the mechanisms, the 3-83 B cell line was adopted. We found that 1,2-propanediol and lactate could initiate the BCR signaling of 3-83 B cells, which plays an important role in the maturation and survival of B cell lineages, and the consequential humoral immune responses (69). This explains the phenomena we observed in Fut8 ϩ/Ϫ mouse-fed neonates, where the relatively reduced B cell proportion resulted in the downregulation of humoral immunity post-OVA stimulation. Our study thus highlighted a novel mechanism whereby L-fucose metabolites, mainly 1,2-propanediol (37,38), interacted directly with BCR molecules to promote B cell activation. Further molecular studies regarding the interaction between 1,2-propanediol and BCR may contribute to the elucidation of the underlying mechanisms.
To elucidate how core fucose on milk N-glycan selectively promotes the growth of Bifidobacterium and Lactobacillus spp., Fuc-␣1,6-GlcNAc and Fuc-␣1,6-GlcNAc-GlcNAc were synthesized chemically. Fucosylated oligosaccharide is an important core structure that forms part of human mucosal and milk glyco-complexes (70)(71)(72). The distinctive ␣1,6 linkage of fucose on N-GlcNAc may contribute to its selective effects. Previous studies showed that Lactobacillus casei fermented the N-GlcNAc moiety of Fuc-␣1,6-GlcNAc and excreted L-fucose (73). It harbors a novel ␣-L-fucosidase (AflC) gene which specifically hydrolyzes natural ␣1,6-linked fucosyl-oligosaccharides in vitro (74). This might explain why L. casei is more abundant in infants fed by mothers with high-corefucosylated milk N-glycan. However, compared with oligosaccharides, the structure of protein N-glycans is more complicated. Bacterium-derived fucosidases were reported to have very low activity in decomposing the ␣-1,6-bond core fucose (75). Fuc-␣1,6-GlcNAc can only be hydrolyzed by ␣-L-fucosidases with the assistance of endo-Nacetylglucosidases (Endo), which hydrolyzes the N-glycans between two adjacent N-GlcNAc. Some bifidobacteria express Endo-like enzymes, which allows them to become dominant in the intestines of infants by using milk N-glycans (18). Endos mainly belong to the GH18 and GH85 families; they are also found in the genomes of other gut microbes, such as multiple species of Bacteroides (accession no. WP_048696603.1; GI no. 880967032), Enterococcus faecalis (accession no.: AAO82555.1; GI no. 29344798), and Lactobacillus spp. (accession no. YP_004888239.1; GI no. 1061453). Indeed, we found that GH18 genes were upregulated in the gut of infants fed by mothers with high-core-fucosylated-milk N-glycans (Fig. 2E). Thus, use of synthetic Fuc-␣1,6-GlcNAc and Fuc-␣1,6-GlcNAc-GlcNAc by L. casei and L. gasseri suggests the coexpression of Endo and AflC-like enzymes in their genomes and their superiority when competing with other gut microbes. However, when Fuc-␣1,6-GlcNAc was incubated with Bifidobacterium spp., no obvious growth-promoting effects were detected, although ␣-L-fucosidase genes were also found in the genomes of Bifidobacterium spp., but with poor activity for hydrolyzing the ␣1,6-linkage-bond fucose (75). Therefore, the substantially greater proportion of bifidobacteria in infants fed by mothers with highcore-fucosylated milk N-glycan may due to excreted L-fucose by Lactobacillus spp. This suggests a synergism between Lactobacillus and Bifidobacterium in digesting milk N-glycans in the gut of infants. However, more studies are needed to define in detail the mechanism how Lactobacillus and Bifidobacterium specifically consume corefucosylated N-glycan and impact immune development of infants.
Conclusions. Infants fed by mothers with higher-core-fucosylated milk N-glycans (G group) harbored greater abundance of Bifidobacterium spp. and Lactobacillus spp. and reduced abundance of Clostridiales and Pseudomonas during lactation. Compared with Fut8 ϩ/ϩ mouse-fed neonates, the neonates fed by Fut8 ϩ/Ϫ mice were characterized by reduced abundance of Lactobacillus spp., Bacteriodes spp., and Bifidobacterium spp. and increased abundance of members of the Lachnospiraceae NK4A136 group and Akkermansia spp. in their gut: these neonates also exhibited a lower proportion of active B cells in spleen. In vitro study showed that 1,2-propanediol and lactate, the metabolites of L-fucose produced by Lactobacillus and Bifidobacterium spp., could evoke B cell activation through the BCR-mediated signaling pathway. The chemically synthesized core-fucosylated oligosaccharides showed the ability to promote the growth of Lactobacillus and Bifidobacterium spp., and therefore could be considered a promising prebiotic for the personalized nutrition of infants (Fig. 7).

MATERIALS AND METHODS
Subjects and sample collection. The study was approved by the ethical committees of Dalian Medical University, Dalian, China. A subset of 56 infant/mother dyads from the First Affiliated Hospital of Jinzhou Medical University was selected. Written informed consent was obtained from the parents before enrollment. Subjects were enrolled at approximately 34 weeks of gestation and asked to fill out detailed health history questionnaires. To eliminate the interference of delivery mode and feeding pattern on gut microbiome of infants, only mothers who gave birth by vaginal delivery and exclusively breast-fed were enrolled ( Table 1). The milk and fecal samples were collected as described before (15) and immediately stored at Ϫ20°C until they were transported to the laboratory on dry ice, where they were stored at Ϫ80°C prior to use. The milk and feces samples were collected by days 6, 42, and 120 postpartum. Normally 3 days postpartum, the mothers start to lactate, and by day 6, their babies had been fed for at least 3 days; in addition, by that time the mothers have enough milk for sampling. Day 42 is after puerperium, which is the period of adjustment after delivery when the anatomic and physiological changes of pregnancy are reversed and the body returns to the normal nonpregnant state; normally Chinese mothers are asked to go back to the hospital for routine checkup. In addition, according to a previous study (76), from week 4 to week 8 after birth, some microbial groups in the gut of infants significantly increase in proportion: day 42 (6 weeks postpartum) is in the middle of this period. We collected the samples by day 120 postpartum to detect the long-term development of gut microbiota in infants and also to ensure the infants were exclusively breastfeeding, because many of the infants were introduced to formula or complementary foods after 4 months postpartum. All of the infants consumed breast milk only, and infants who received antibiotics, probiotics, or formula powder because of diseases or lack of breast milk were excluded during the study period.
Detection of milk N-glycan and HMOs by mass spectrometer. The milk N-glycan detection was carried out using an LTQ-XL linear ion trap electrospray ionization mass spectrometer (ESI-MS) coupled with a high-performance liquid chromatography (HPLC) system (Thermo Scientific, USA), as described previously (77). The HMOs were detected by a newly developed method based on a zwitterionic LC matrix with a mixed-mode action of hydrophilic interaction with cation exchange for cleanup and separation of HMOs (13,78). The neutral sugars were eluted, and acidic HMOs were resolved and identified. Particularly, we used a different column material (named ASP; 150 mm by 2.1-mm inside diameter [i.d.]) to allow acidic sugars to be eluted first, and focused on detailed separation and neutral HMOs. HMO identification and quantification were performed via Agilent Mass Hunter Qualitative Analysis software (version B.03.01) (detailed in Text S1 in the supplemental material).
SNPs. DNA was extracted from breast milk using the Qiagen DNeasy blood and tissue kit (Qiagen, Venlo, The Netherlands). Genomic DNA purified from each mother's breast milk was amplified with primers Fut8-F (5=-TAT AAA GGC ACA GAA ACA GAC A) and Fut8-R (5=-TTG ATG GTG GCT CCA TTG CC), which produces a 337-bp amplicon containing the mutated rs10483776 SNP (A¡G) allele of the Fut8 gene. Successful amplification was confirmed by gel electrophoresis, and the PCR products were sent to the sequencing company (Sangon Biotech) for DNA sequence detection.
FUT8 enzyme activity assay. Cells isolated from human milk and mouse mammary gland were suspended in 20 l lysis buffer containing 10 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 1% Triton X-100. The cell lysate was assayed for Fut8 activity by HPLC, as described previously (23).
Western blot and lectin blot analysis. Each milk sample was defatted via centrifugation at 8,000 rpm for 10 min. After that, 1 l of skim milk was taken and dissolved in 30 l protein loading buffer (250 mM Tris-HCl [pH 6.8], 0.5% bromophenol blue, 50% glycerol, 10% SDS, 5% ␤-mercaptoethanol), and then 10 l of proteins was subjected to SDS-PAGE. After SDS-PAGE, the proteins were transferred to polyvinylidene difluoride (PVDF) membranes for immunoblotting or lectin blotting after incubation with the appropriate primary antibodies (Abs for human LF, IgG, and IgA were purchased from Abcom) or biotin-conjugated AOL (Seikagaku, Tokyo, Japan) (23).
Fecal DNA extraction, 16S rRNA, and metagenomic sequencing. The microbial genome DNA from fecal samples of infants and mice was extracted using the E.Z.N.A. stool DNA kit (Omega Bio-tek, Inc.) Core Fucosylation of Milk N-Glycan Evokes B Cell ® according to the manufacturer's instructions. The library construction, qualification, and sequencing were done by Illumina HiSeq (Novogene Bioinformatics Technology Co., Ltd., Beijing, China) (79); for details regarding 16S rRNA and metagenomic sequencing, please see Text S1.
Mice. Heterozygous Fut8 ϩ/Ϫ mice on the ICR background were maintained in a room illuminated for 12 h (08:00 to 20:00 h) and kept at 24 Ϯ 1°C with free access to food and water in the specific-pathogenfree laboratory animal facility of Dalian Medical University. All of the animal experiments were conducted according to the Guide for the Care and Use of Laboratory Animals (NIH publication no. 8023 [80]). Each Fut8 ϩ/Ϫ and Fut8 ϩ/ϩ maternal mouse nurtured three female neonates (n ϭ 3).
Growth of strains in the presence of synthetic oligosaccharides. The standard strains used in this study were obtained from China General Microbiological Culture Collection Center (CGMCC). They were routinely grown at 37°C on MRS medium supplemented with 0.1% L-cysteine and under anaerobic conditions. The strains were diluted to an optical density at 546 nm (OD 546 ) of 0.1 in 1 ml of sugar-free MRS basal medium (73) containing 10 mM Fuc-␣1,6-GlcNAc-GlcNAc or Fuc-␣1,6-GlcNAc. Growth was monitored by measuring the OD 546 . Three technical replicates were performed for each strain.
Molecular interaction detection by Biacore. BCR molecules were purified according to the method described in reference 41. The interactions between BCR and L-fucose metabolites were performed at room temperature using a BIAcoreT100 system with CM5 chips (GE Healthcare). An HBS-EP buffer consisting of 150 mM NaCl, 10 mM HEPES (pH 7.4), and 0.005% (vol/vol) Tween 20 was used as running buffer. The blank channel of the chip served as the negative control. To measure the interaction between L-fucose metabolites and BCR, 1 g/ml BCR protein was immobilized on the chip. Gradient concentrations of lactate and 1,2-propanediol were then flowed over the chip surface. The binding kinetics were analyzed with the software BIA Evaluation version 4.1 using a 1:1 Langmuir-binding model (81).
Statistical analysis. The comparisons between two groups were performed using an unpaired Student's t test with Welch's correction by Graph Pad Prism version 5 (Graph Pad Software, La Jolla, CA). The data are shown as mean Ϯ standard error of the mean (SEM). P values of Ͻ0.05 are considered statistically significant (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001). For gut microbiota analysis, the Chao1 index and the Shannon index at the genus level were calculated with QIIME (version 1.7.0). The abundance and diversity of the OTUs (beta diversity) were examined using principal-coordinate analysis (PCoA) with unweighted UniFrac analysis in R software. LEfSe was used with the Kruskal-Wallis rank sum test to detect features with significantly different abundances between assigned taxa, and linear discriminant analysis (LDA) was performed to estimate the effect size of each feature. Bacterial groups with an LDA score of м4.00 were shown as significantly abundant within the indicated group. Data correlating to differential abundances of KEGG modules and CAZy enzymes were tested by Wilcoxon's rank sum test, and P values were corrected for multiple testing by the Benjamin and Hochberg method. The bacterial genes with an LDA score of м2.5 were shown as significantly abundant in the indicated group. Statistical analysis of the quantitative multiple-group comparisons was performed using one-way analysis of variance (ANOVA [and nonparametric test]), followed by Tukey's test (to compare all pairs of columns) with the assistance of GraphPad Prism 5. When analyzing the growth of mice or bacteria, two-way ANOVA were performed with the assistance of GraphPad Prism 5.
Data availability. The sequencing data were deposited in NCBI SRA under the accession number PRJNA527059.
TEXT S1, DOCX file, 0.1 MB.   All the authors declare they have no competing interests.