The concentrations of IgG and IgM in cord blood increased with gestational age, where the IgG concentration in cord blood from full-term neonates with a gestational age of at least 37 weeks reached between 10 and 20 mg/ml, which is within the range of IgG concentrations in maternal blood (Figure 1A, Supplementary Table 1). Direct comparison of matched cord and maternal blood samples showed no correlation in respect to IgG concentration. Notably, high IgG concentrations in maternal blood were mainly associated with lower concentration in matched cord blood samples (Figure 1B).

IgM concentration in cord blood also increased with gestational age (Figure 1C), although at term IgM concentration remained significantly lower than the concentration of about 1 mg/ml measured in maternal blood (Figure 1D, Supplementary Table 1). In contrast to IgG, the levels of IgM in neonatal cord blood were consistently lower than the serum IgM concentration in the matched maternal blood samples (Figure 1D).

To compare the antigenic repertoire of carbohydrate-specific antibodies in matched cord blood and maternal blood samples, we focused on neonates delivered at full term to maintain consistency in IgM and IgG levels. The sample population investigated comprised 26 pairs of matched cord blood and maternal blood samples. IgG concentrations in this group of samples averaged 14.6 ± 4.7 mg/ml for cord blood samples and 13.7 ± 5.0 mg/ml for maternal blood samples. IgM concentrations were 0.6 ± 0.1 mg/ml in cord blood samples and 1.0 ± 0.3 mg/ml in maternal blood samples.

Carbohydrate antigen specificity of IgG and IgM in matched cord blood and maternal blood samples was investigated using an array displaying 220 different human oligosaccharides (37). The oligosaccharide structures tested represented common carbohydrate epitopes found on human cells and occurring on bacterial glycoconjugates expressed by intestinal commensals (11, 38). Most oligosaccharides displayed on the array were characterized by their mass composition consisting of hexose (H), N-acetyl-hexosamine (N), fucose (F) and the sialic acid N-acetylneuraminic acid (S). Antibody binding to oligosaccharides was calculated as the mean fluorescence intensity of four replicates for each oligosaccharide.

The comparative analysis of 26 maternal blood samples revealed a large inter-individual variability in the range and intensity of oligosaccharide binding by IgG across samples (Figure 2A). Whereas isolated samples showed a strong reactivity toward oligosaccharides (samples MB13, MB21 on Figure 2A), others showed only a minor reactivity to few oligosaccharides (samples MB4, MB9 on Figure 2A). The difference in the average binding intensity to oligosaccharides exceeded a 100-fold between the least and the most reactive maternal blood samples (Figure 2B). The reactivity of IgG to oligosaccharides in cord blood samples was on average 10-fold lower than in the maternal blood samples, whereas a large inter-individual difference was also evident between the samples tested (Figure 2A).

Given that the placental transfer of antibodies is restricted to the IgG class, the amount and reactivity of IgM detected in cord blood samples reflect the endogenous production by the fetal immune system (25). In contrast to IgG, cord blood IgM largely did not react with the oligosaccharide antigens on the array (Figure 2A). Maternal blood IgM showed a strong reactivity to oligosaccharides in a similar way to the signals recorded for IgG (Figure 2A). Individual recognition patterns and signal intensities for IgG and IgM of the same serum strongly overlapped (Supplementary Figure 1). Correlation analysis confirmed that repertoires of carbohydrate-specific IgG and IgM correlate significantly for each maternal blood serum (Supplementary Table 1), clearly showing the lack of cord blood IgM reactivity toward oligosaccharides.

The partition of oligosaccharides into fucosylated, sialylated and undecorated structures did not reveal any major differences in reactivity between cord blood and maternal blood IgG. When focusing on non-fucosylated and non-sialylated oligosaccharides, some cord blood samples contained only marginal levels of IgG recognizing such undecorated oligosaccharides, whereas the corresponding maternal blood samples displayed strong reactivity toward this group of oligosaccharides (Figure 3A). Similar differences were visible for the abundance of IgG reactive toward individual groups of fucosylated oligosaccharides, where IgG reactivity was absent or extremely low in cord blood samples despite a high IgG reactivity in the matched maternal samples (Figure 3B). The same trend was observed for di-sialylated oligosaccharides, regardless of the fucosylation status (Figure 3C). Focusing on single glycan structures, we selected the glycans with the highest average IgG reactivity in the cord blood group. Two glycans displayed a mean RFU higher than 1,000 for the cord blood group and additional strong IgG reactivity for 24 out of 26 maternal blood samples (Figures 3D,E). Interestingly, strong maternal IgG reactivity did not always lead to high levels of IgG binding in matched cord blood, as illustrated by a subgroup of samples (Figures 3D,E). For six maternal samples with a comparable and strong IgG reactivity to structure S1, only one matched cord blood sample displayed the same strength in IgG binding to the same glycan. Compared with the matched maternal samples, the other five cord blood samples resulted in weaker IgG binding, still leading to strong recognition of structure S1 in one case and weak responses for three samples. Interestingly, despite the high levels of IgG reactivity observed in the matched maternal blood, one cord blood sample did not show any IgG binding to the same glycan (Figure 3D). A similar pattern was observed for the second glycan structure S2 (Figure 3E). The discrepancy in the IgG reactivity between cord blood and matching maternal blood samples pointed to a selective transfer of maternal carbohydrate-specific IgG to the fetal circulation.

The direct comparison of IgG reactivity profiles toward oligosaccharide antigens in the 26 matched pairs of cord blood and maternal blood samples confirmed the general lack of correlation in antigen specificity. Only six pairs displayed visible overlaps in IgG binding patterns (Figure 4). Correlation analysis of normalized IgG reactivities demonstrated that carbohydrate-specific IgG profiles of maternal and cord blood strongly correlated (r > 0.67) only for two of 26 pairs (CB2/MB2, CB22/MB22), whereas three pairs showed a moderate correlation (r = 0.36–0.67) and 21 pairs did not or only weakly correlate (39) (Table 1).

Limited transfer of carbohydrate-specific antibodies may be explained by differences in IgG subclasses specifically recognizing carbohydrate antigens. IgG2 is the major subclass of IgG induced to carbohydrate antigens (40, 41), even though other subclasses are involved in anti-carbohydrate responses (42). As the placental transport of IgG2 is less efficient compared with the transport of other IgG subclasses (43), limited transfer of carbohydrate-specific antibodies may be linked to limited transfer of the IgG2 subclass. Testing IgG2 concentrations in cord blood and maternal blood, we determined the IgG2 transfer rate for the 26 pairs by calculating the proportion of IgG2 in cord blood to IgG2 in matched maternal blood (44). The mean IgG2 transfer rate of 95.5 ± 51.5% was lower than total IgG transfer rate of 120.5 ± 52.5% (Supplementary Figure 2).

Gut bacteria play an important role in the emergence of carbohydrate-specific antibodies (3, 45). Maternal IgG significantly contribute to the protection of the neonate from bacterial infection through placental transfer to the fetal circulation (15–18). To determine whether the selective placental transfer of IgG detected for carbohydrate antigens also affected the recognition of commensal bacteria by cord blood IgG, we compared the reactivity of matched cord blood and maternal blood samples to representative species of commensal Bacteroides bacteria. The phylum of Bacteroidetes is one of the major phyla of the human gut microbiota including a variety of species, where the Bacteroides are known for their high diversity in surface glycoconjugates (11, 46). All matched cord blood and maternal samples tested showed similar global IgG reactivities to B. intestinalis, B. thetaiotaomicron, and B. vulgatus (Figures 5A–E), where mean IgG reactivities for maternal and cord blood were on the same level (Figures 5A,C,E). Also the patterns of reactivity between matched pairs of cord blood and maternal blood samples overlapped as assessed by flow cytometry histograms (Figures 5B,D,F). The similarity in antibody reactivity toward the Bacteroides species clearly showed that maternal IgG targeting bacterial antigens were efficiently transferred to the fetal circulation. By contrast, IgM reactivity to these three Bacteroides species was absent in cord blood samples, whereas elevated reactivities were measured in maternal blood samples (Figures 6A,C,E). The comparison of individual matching cord blood and maternal blood pairs clearly underlined the lack of IgM transfer from the maternal circulation to the fetus (Figures 6B,D,F). The absence of IgM specific to the Bacteroides species tested in cord blood also indicated that the fetal immune system was not primed to these commensals prior to delivery, thus supporting the idea that prenatal exposure to gut microbes is insignificant and insufficient to stimulate the production of specific antibodies. To further document the similarity in antigen recognition between matched neonatal and maternal blood derived IgG, the reactivity of selected matched samples toward Bacteroides cell lysates was investigated using Western blotting. This analysis confirmed the strong inter-individual variability in antigen recognition between maternal blood samples, as shown by the different patterns of bands recognized by individual samples (Figure 7). The analysis also demonstrated the similarity in antigen recognition between matched cord blood and maternal blood samples, thereby underlining the placental transfer of maternal IgG to commensal bacteria into the fetal circulation.

Goat anti-human IgG A488 and goat anti-human IgM A647 were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). Anti-human IgG DyLight488 and anti-human IgM DyLight488 were obtained from Abcam (Cambridge, MA). Anti-human IgG HRP was from Promega (Fitchburg, WI). Bacteroides vulgatus (DSM 1447), Bacteroides intestinalis (DSM 17393), and Bacteroides thetaiotaomicron (DSM 2079) were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany).

The study was approved by the cantonal Ethics Committee of Zurich (KEK Nr. 2019-00150). All experiments were performed in accordance with relevant guidelines and regulations. Informed consent was obtained from all participants and/or their legal guardians. Leftover maternal and cord blood samples were obtained from the University hospital Zurich, Switzerland. The study population comprised 30 full-term and 12 pre-term neonates. Out of these 42 samples, matched maternal blood was available for 26 full-term and 5 pre-term cord blood samples. Blood was centrifuged at 3,000 x g for 15 min at 4°C and sera were frozen at −20°C. IgG and IgM concentrations were determined by ELISA [IgG human ELISA Kit from Abnova (KA3817), IgM human ELISA Kit from Abnova (KA1855), Taipeh, China] according to the manufacturer's protocols. IgG transfer was calculated as proportion of the IgG concentration in cord blood to the IgG concentration in maternal blood ([IgG_cord/IgG_maternal]^*100), according to (44).

The human milk shotgun glycan microarray (version 223) has been previously described (37) and was provided by the National Center for Functional Glycomics (NCFG), BIDMC, Harvard University. The shotgun glycan microarray slides were stored at −20°C, and before use, the arrays were allowed to come to room temperature in a desiccator. Individual slides were loaded with 8-well format ProPlate Chambers (Grace Bio-Labs, OR, USA) and hydrated with TSM-T [20 mM Tris-HCl pH 7.4, 150 mM NaCl, 2 mM CaCl₂, 2 mM MgCl₂ (TSM) with 0.05% Tween]. To test antibody reactivity of individual serum samples, arrays were incubated with serum diluted to 500 μg/ml IgG in binding buffer (TSM-T with 1% BSA) for 1 h at room temperature. After incubation, slides were washed four times with TSM-T and for times with TSM. For detection of IgG, slides were incubated for 1 h at room temperature with 5 μg/ml Alexa488-labeled goat anti-human IgG in binding buffer. After washing as described above, the slides were incubated for 1 h at room temperature with 5 μg/ml Alexa647 goat anti-human IgM and washed as described with a final wash of four times with deionized water. After spin-drying, arrays were scanned at 647 nm and 488 nm using Innoscan 1100 AL (Innopsys, Carbonne, France). A mask representing the layout of the array was fit to each image using MAPIX analysis software version 8.5.0 (Innopsys, Carbonne, France) and the mean fluorescence intensity of each spot was calculated. The background signal was subtracted for each spot, the mean of four replicates was calculated and negative values were set to 1. For further analysis log (RFU) was used. Microarray production and analyses were performed according to the MIRAGE guidelines (68) (Supplementary Material).

Bacteroides species were cultivated anaerobically in rubber-sealed hungate tubes at 37°C in Peptone Yeast Glucose medium (DSMZ, medium no. 104 with 1.25 mg/ml glucose). At OD₆₀₀ of 1.0 cells were harvested in aliquots of 0.5 ml. Cell pellets were received by centrifugation at 13,000 x g for 1 min, and phosphate-buffered saline (PBS) washed pellets were either stored at −20°C for cell lysates or fixed with 2% paraformaldehyde in PBS for 15 min.

Fixed bacteria were treated with diluted serum (1:100 for IgG, 1:10 for IgM) in PBS for 1 h at room temperature, washed twice using PBS, stained with anti-human IgG DyLight488 or anti-human IgM DyLight488 for 1 h at room temperature and washed twice with PBS. The fluorescence signal was recorded by a FACSCanto II Flow cytometer (BD Biosciences, Franklin Lakes, NJ) and analysis was performed with the FlowJo software (BD Bioscience).

Bacterial cell lysates were prepared from bacterial pellets, resulting from 0.5 ml culture at OD₆₀₀ of 1.0, resuspended in 200 μl SDS buffer (2% β-mercaptoethanol, 2% SDS, 10% glycerol in 50 mM Tris-HCl, bromophenol blue, pH 6.8), boiled at 99°C for 15 min. Fifteen microliter bacterial cell lysates were run on 14% acrylamide gels at 80 V for 2 h and transferred onto PVDF at 250 mA for 1 h. Membranes were blocked overnight at 4°C with 5% BSA in PBS-T (PBS with 0.1% Tween20). Membranes were incubated with diluted serum in PBS-T (1:500) for 2 h at room temperature, washed four times for 5 min using PBS-T, incubated with anti-human IgG-HRP (1:5,000) in PBS-T for 1 h at room temperature, washed four times for 5 min with PBS-T and detected using SuperSignal® West Pico chemiluminescence substrate (Thermo Fisher Scientific, Waltham, MA) and a Fujifilm LAS-4000 luminescence image analyzer (GE Healthcare, Chicago, IL). Matched cord and maternal blood samples were used on the same membrane, which was cut after transfer, and detection of membrane pieces incubated with matched samples was performed simultaneously.

Statistical analysis was performed using GraphPad Prism (GraphPad software, San Diego, CA). For the comparison of two groups of paired samples (cord blood vs. maternal blood, IgG total vs. IgG2), either a paired student's t-test for Gauss-distributed or a Wilcoxon matched pairs signed rank test for non-Gauss distributed samples was used. For unpaired groups (pre-term vs. full-term), an unpaired t-test with Welch's correction was used. Results with p-value ≤ 0.05 were considered significantly different and marked with asterisks. For quantile normalization and correlation analysis a software tool run on R (version V3.6.1) using the “limma” package (69) with the functions cor.test() and normalizeQuantiles() was used.