Human milk IgA promotes normal immune development by limiting Th17-inducing Erysipelatoclostridium ramosum in the infant gut

Significance

The infant gut microbiota plays a vital role in immune development. Secretory IgA (SIgA), the most abundant antibody in the gut, is a key modulator of microbiota composition throughout life. Maternal milk is the primary source of intestinal SIgA for the developing infant, but the ability of maternal milk SIgA to guide microbe-mediated immune imprinting remains poorly understood. Here, we characterize the microbiota reactivity of human milk SIgA, demonstrating relationships between maternal SIgA and infant gut microbiota composition. This leads to the identification and characterization of E. ramosum, a microbe that is controlled by milk SIgA and plays an important role in immune development worthy of further consideration in studies of allergy prevention at the maternal–infant interface.

Abstract

The gut microbiota is highly dynamic during the first year of life and plays a crucial role in immune development. Breastfeeding is known to support infant health, but the contributions of the numerous breastmilk components to gut microbiota and immune maturation remain unclear. Secretory IgA (SIgA), the most abundant antibody in human milk, is a key modulator of gut microbiota composition. We have shown previously that mouse milk SIgA protects against asthma by limiting segmented filamentous bacteria in mice. The present study uncovers a similar mechanism in humans. Using human milk from the CHILD Cohort Study, we define a relationship between human milk SIgA and infant gut microbiota composition. This leads to the identification of Erysipelatoclostridium ramosum as a key player in immune development, which is controlled by milk SIgA. Cell culture modeling demonstrates that SIgA restricts the capacity of E. ramosum to adhere to the intestinal epithelium and to induce Th17 responses, which are implicated in allergic and other chronic diseases.

Secretory IgA is the most abundant antibody in the human gut and can both limit and promote gut microbiota members, playing a dual role in controlling microbiota composition (1, 2). The first year of life represents an important developmental window during which the highly dynamic gut microbiota trains the maturing immune system (3, 4). During this period, infants are not yet able to produce adult titers of Secretory IgA (SIgA) and obtain this antibody almost exclusively through breastfeeding (5). The SIgA in human milk is produced by plasma cells that originate in the mother’s gut and is thought to target the maternal gut microbiota (6). Thus, human milk SIgA likely plays a role in mother-to-infant gut microbiota transfer.

The World Health Organization recommends exclusive breastfeeding for the first 6 mo of life (7). Breastfeeding is known to protect against later development of numerous diseases, including allergies, asthma, inflammatory bowel disease (IBD), necrotizing enterocolitis, and infections (8–11). Breastfeeding also promotes healthy gut microbiota maturation (12–14). The nutritive and immune components of milk limit premature diversification and expansion of certain bacteria, controlling the rate of host–microbe interactions over the early-life period (15–17). However, the relative contributions of the numerous components of human milk to its health-promoting properties are not fully characterized. SIgA is present at 1 to 8 g/L in human milk, making up the largest proportion of immune proteins passed from mother to infant. SIgA in human milk provides passive immunity to protect against infection and diseases of preterm infancy (18–20). However, the ability of human milk SIgA to protect against chronic immune-mediated diseases associated with dysbiosis in the early-life gut microbiota has not been established.

We found previously that milk-derived SIgA limits the Th17 cell-inducing genus, Candidatus Arthromitis, also known as Segmented Filamentous Bacteria (SFB), in the murine neonatal gut (21). In the absence of maternal SIgA, SFB expansion led to elevated intestinal Th17 responses that persisted into adulthood and contributed to more severe inflammatory responses to a mouse model of asthma. Th17 cells are involved in maintaining barrier integrity, but also contribute to diseases such as IBD, asthma, and infection in humans, contributing vitally to both homeostasis and inflammation (22–25). In mice, SFB has been identified as a key driver of Th17 development. Mouse models of IBD and asthma are exacerbated by SFB, and SFB can act alone to rescue Th17 responses in germ-free mice (26–30). The dynamics and immunological effects of this potent pathobiont have been extensively studied in mice, but a similar microbe in humans has not been identified.

In the present study, we explore a potentially similar mechanism in humans using human milk samples and longitudinal infant health and microbiota data from the CHILD Cohort Study, along with human cell culture models. We ultimately describe a relationship between human milk SIgA and infant gut microbiota composition, and uncover a vital role for human milk SIgA in limiting Th17 responses in the infant gut by targeting Erysipelatoclostridium ramosum, a known keystone member of the infant gut and potential mediator of allergy susceptibility (31).

Results

Characterization of the Human Milk SIgA Repertoire.

The use of human milk from participants in the CHILD Cohort Study enabled the comparison of milk SIgA levels and binding patterns to infant gut microbiota, age, and immune markers. Since atopic/allergic disease outcomes (defined under the Materials and Methods section) were of interest, 300 participants were chosen using case–control matching based on infant atopy status. Atopic cases were rigorously defined and required both objective evidence of allergic inflammation as documented by IgE-mediated skin test positivity to food or inhalant allergens together with a clinical diagnosis of allergic disease. Cases and controls were matched by study site, birth mode, smoke exposure, maternal atopy status (defined under Materials and Methods), and mother’s self-identified race and ethnicity (Table 1). Matching was successful for all variables excluding maternal atopy status, which is to be expected due to the heritable component of atopic disease.

Table 1.

CHILD sample characteristics

Matching variables	Case (N = 100)	Control (N = 200)	P
Study site			<0.001
Edmonton	18	17
Toronto	37	19.5
Vancouver	32	25.5
Winnipeg	12	38
Infant antibiotic use (1st y)	19	21.5	0.61
Delivery mode			0.3
Vaginal	71	76.5
C-section	29	23.5
Prenatal smoke exposure	3	5.5	0.33
Mom atopy (yes)	71	48.2	0.0002
Mom age (yrs)	Avg 33.8	Avg 33.5	0.76
Mom race (White)	58	66	0.17

Percent representation within case (atopic) and control (nonatopic) groups is listed for each variable used for case–control matching. P represents the P-value resulting from a t test or comparing the two groups.

Metagenomic data from stool samples of 3-mo-old infants born to 300 participants were used to define a panel of species common to the infant gut (Table 2). Milk SIgA was assessed for binding against available strains (SI Appendix, Table S1) of each of the most abundant and prevalent bacterial species of the infant gut through a set of bacteria-binding SIgA ELISAs (Fig. 1 A and B). Total SIgA was also quantified for each sample (mean = 1.042 g/L, SD = 0.684). As described previously (32), human milk SIgA-targeting varied greatly between mothers, ranging 30-fold, illustrating the high level of heterogeneity in both SIgA quantity and specificity (Fig. 1B). Although SIgA-targeting of one species was generally associated with SIgA-targeting of another species and with total SIgA levels, correlations between SIgA-binding variables were not equal across species (SI Appendix, Fig. S1A). However, some general patterns emerged: Bacteroidaceae and Bifidobacteriaceae species were generally targeted at the lowest level by milk-derived SIgA. Both of these families are thought to be beneficial when abundant in early life (12) and tend to be untargeted by SIgA in adult feces (33). Enterobacteriaceae species, which typically decline in relative abundance over the first month of life, were targeted at the highest level by milk SIgA. This is also in agreement with studies of adult feces and may hint that SIgA plays a role in the decline of this family over the course of infancy (34). Across the cohort, the binding capacities of human milk SIgA mirror those observed previously in adult fecal SIgA, in support of the intestinal origin of mammary gland B cells (35).

Table 2.

Panel of bacteria tested in bacteria-binding SIgA ELISA

Family	Species	Prevalence (/288)	% relative abundance (range)
Bifidobacteriaceae	B. longum	162	44.3 (99.7)
	B. breve	94	42.1 (98.9)
	B. bifidum	66	39.9 (92.9)
Enterobacteriaceae	E. coli	201	17.8 (99.7)
	K. pneumoniae	77	10.9 (72.5)
	K. variicola	66	6.04 (49.8)
Bacteroidaceae	B. fragilis	57	18.7 (55.6)
	B. uniformis	56	6.76 (76.3)
Lachnospiraceae	R. gnavus	73	7.58 (50.9)
Enterococcaceae	E. faecalis	181	2.50 (50.3)
Erysipelotrichaceae	E. ramosum	80	3.39 (30.9)
Clostridiaceae	C. perfringens	45	0.383 (2.13)
Streptococcaceae	S. salivarius	148	0.571 (23.8)
	S. parasanguinis	146	0.140 (5.6)
Tannerellaceae	P. distasonis	59	3.98 (29.7)

Among the top 10 most abundant families in the infant fecal metagenome, species with prevalence >50/288 were chosen. The family Clostridiaceae did not contain any species with prevalence >50/300, so the most prevalent species within this family was chosen. Average relative abundance among infants positive for each bacteria, along with the range in relative abundance among those infants, are also displayed. The sample of 288 infants corresponds to the 300 tested milk samples from the CHILD study cohort.

Fig. 1.

Characterization of the human milk SIgA repertoire. (A) Schematic of SIgA ELISAs performed on 300 human milk samples from the CHILD Cohort Study to quantify total and bacteria-binding SIgA against a panel of the most common and abundant bacterial members of the infant gut microbiota. (B) Quantity of SIgA (in μg/mL) in each milk sample that binds to each species listed in Table 2 measured by bacteria-binding ELISA. Boxplots show the interquartile range (IQR) with the horizontal line indicating the median. Whiskers extend to values within 1.5 × IQR. Individual data points are overlaid with jitter to display sample-level variation. (C) Associations between bacterial relative abundance in the infant stool at 3 mo and either total SIgA or bacteria-binding SIgA in milk determined by Maaslin2 analysis. Size and color of circles represent the size of the correlation coefficient and the direction of the correlation, respectively. Stars indicate significant associations (FDR < 0.1).

Total SIgA quantities in milk samples did not differ between atopic and nonatopic mothers or mothers of atopic or nonatopic infants (SI Appendix, Fig. S1 B and C). However, total milk SIgA was significantly negatively associated (P < 0.05) with levels of several immune markers in the infant serum at 1 y of age, including cytokines known to drive neutrophil and Th17 responses [CCL20, IL8 (CXCL8), IL-12b] and leukocyte infiltration (MCP.3, MCP.1, MCP.2, SLAMF1) (SI Appendix, Fig. S1D). This led to the hypothesis that, as we observed in mice, human milk SIgA targets and limits Th17-inducing microbes in the infant intestine, with long-term consequences for systemic immunity.

The Human Milk SIgA Repertoire Is Associated with Infant Gut Microbiota Composition.

Total and bacteria-targeting SIgA levels were next compared to metagenomic relative abundance in the 3-mo-old infant for all species listed in Table 2 using Maaslin2, an R package designed to assess associations between clinical variables and metagenomic data using generalized linear models (36) (Fig. 1C). Five species (Bacteroides fragilis, Bacteroides uniformis, Bifidobacterium bifidum, Bifidobacterium breve, E. ramosum) displayed strong associations (coef > 0.3) with the level of species-specific SIgA in milk, whereas only B. breve displayed strong association with total SIgA, illustrating that SIgA binding patterns are more indicative of infant gut microbiota composition than total SIgA levels in milk. Most associations were positive, where for a given microbe, higher SIgA binding capacity in milk was associated with higher relative abundance in the infant. While this may be partially explained by SIgA’s capacity to promote certain bacteria in the gut through binding, it is also likely that SIgA binding patterns in milk occur due to the overlap in microbiota composition between mother and infant. Overall, this analysis shows that maternal milk SIgA binds to all common members of the infant gut microbiota at varying levels, and supports the importance of assessing specificity, in addition to titer, in understanding the character and function of human milk SIgA.

Of those displaying significant associations with SIgA-targeting in milk, E. ramosum was the only species in the infant gut displaying a significant negative association (FDR < 0.1) with maternal milk SIgA-targeting, suggesting that SIgA may limit this species in the infant gut (Fig. 1C). This strong association with SIgA, along with the overall high levels of E. ramosum binding SIgA human milk (Fig. 1B), led us to investigate E. ramosum further as a potentially important microbe in infant immune development.

E. ramosum Colonization Increases with Age in the Infant Gut.

The exact age of infants at the time of the scheduled “3 mo” fecal collection ranged from 2.8 to 7.5 mo (mean = 3.70, SD = 1.06). This range allowed us to study temporal dynamics of microbiota maturation. Notably, of all species investigated, E. ramosum was the top age-associated species in the infant gut, with higher E. ramosum levels found in older infants (Fig. 2A). E. ramosum was also significantly higher in 1-y-old infants than 3-mo-old infants across the entire CHILD cohort, regardless of allergy status (n = 2,848) (Fig. 2B). A positive association between this bacterium and allergic disease has been well established previously in a meta-analysis of data from multiple early-life cohorts (31), but in the CHILD study, E. ramosum levels trended lower in feces of allergic infants than in that of control infants at the 3 mo time point (P = 0.065) and eventually caught up and reached similar levels by 1 y, indicating a potentially complex relationship between E. ramosum dynamics and allergy susceptibility.

Fig. 2.

E. ramosum is elevated in older infants and nonbreastfed infants. (A) Associations between species relative abundance and the exact age of infant at the time of fecal sampling assessed by Maaslin2. Significant associations (FDR < 0.05) are dark. N = 288 infants from the CHILD Cohort Study. (B) Relative abundance of the E. ramosum in feces from 2848 infants in the CHILD Study at ~3 mo and ~1 y of age grouped by allergy status (Control or Case). E. ramosum prevalence in each group were as follows: 3 mo control: 29.4%, 1 y control: 70.0%, 3 mo allergic: 23.8%, 1 y allergic: 70.1%. (C) Relative abundance of E. ramosum in feces from 2,848 infants at ~3 mo of age grouped based on breastfeeding status at 3 mo of age. Prevalence in each group were as follows: exclusive: 26.6%, partial: 32.0%, none: 38.5%. Boxplots show the IQR with the horizontal line indicating the median. Whiskers extend to values within 1.5 × IQR. Individual data points are overlaid with jitter to display sample-level variation. Statistical significance was assessed by Mann–Whitney U tests in (B) and (C). Maaslin2 was used to assess significance in (A).

E. ramosum levels were also significantly lower in exclusively breastfed infants compared to nonbreastfed or partially breastfed infants, in agreement with the negative relationship between milk SIgA and E. ramosum colonization, and supporting a weaning-associated bloom of this microbe in early life, as observed with SFB (Fig. 2C). Whether E. ramosum-specific SIgA in milk is solely responsible for this phenomenon cannot be fully confirmed with the presented data. However, the uniquely strong and significant negative relationship between E. ramosum colonization and specific SIgA (Fig. 1C) points to bacteria-binding SIgA as a key contributor to the breastfeeding-associated differences in E. ramosum colonization.

Numerous studies have found that altered gut microbiota maturation is associated with later development of allergies (4, 37, 38), and E. ramosum has been identified previously as a top age-discriminatory taxon in early life, increasing after 3 mo of age in healthy infants (12, 39). E. ramosum levels are also elevated in formula-fed infants (12), and a lack of SIgA-targeting of the family Erysipelotrichaceae is associated with later development of allergies (40), further implicating human milk SIgA in controlling the colonization of this microbe in the infant gut. In a very recent meta-analysis of early-life gut microbiota studies, E. ramosum was identified as a keystone member of the infant gut and a top allergy-associated microbe, across the globe (31). We identify associations between E. ramosum and milk SIgA, age, breastfeeding status, and allergic outcomes, which corroborate associations observed in previous studies.

SIgA from Human Milk Targets and Limits E. ramosum.

E. ramosum is highly SIgA-targeted in studies of adult and infant feces (33, 34, 40). It also produces an IgA protease, suggesting an evolutionary pressure to evade SIgA binding (41). One mechanism by which SIgA controls the microbiota is by preventing attachment or adherence of bacteria to the intestinal epithelium (42). We used human-derived intestinal epithelial cells to address mechanistically whether E. ramosum adherence is limited in the intestine by SIgA.

Immunohistochemical visualization confirmed that E. ramosum adheres to human absorptive intestinal epithelial (Caco-2/TC7) cells. Even after numerous washes in the staining process, bacteria are seen in direct contact with cell membranes (Fig. 3A). E. ramosum also induced cytoskeletal rearrangement, resulting in actin reorganization and accumulation at the cell membrane (Fig. 3A). Upon incubation with SIgA purified from the 20 human milk samples with the greatest E. ramosum SIgA binding capacity, E. ramosum adherence to TC7 cells was significantly reduced (Fig. 3B). Heat-denatured SIgA failed to limit adherence, illustrating that SIgA structure is essential to its capacity to prevent adherence. SIgA purified from the 20 human milk samples with the lowest E. ramosum binding capacity also failed to limit adherence, implicating bacteria specificity in SIgA’s effects. B. breve, which does not adhere to TC7 cells, was employed as a negative control.

Fig. 3.

E. ramosum adheres to the human intestinal epithelium and adherence is blocked by SIgA purified from human milk. (A) Sample confocal microscopy images of TC7 cells infected with E. ramosum labeled by IHC with antibody against Gram-positive Lipoteichoic acid (green), DAPI (blue), and fluorescently labeled Phalloidin to stain actin (red). The red scale bar indicates 10 μm. Orange boxes surround adherent bacteria. (B) Quantification of bacteria that adhere to confluent TC7 cells in the presence of milk-derived SIgA purified from samples with the highest E. ramosum binding capacity (+HB SIgA), absence of SIgA (−SIgA), heat denatured HB SIgA (+DN SIgA), or SIgA purified from human milk with the lowest E. ramosum SIgA binding capacity (LB SIgA). Total SIgA concentration was 5 μg/mL for all SIgA conditions. Nonadherent B. breve was used as a negative control. Statistical significance was assessed by the Mann–Whitney U test in (B) to compare -SIgA to and all other values. ***P < 0.001.

In contrast, E. ramosum adhered at high levels to mucous-producing HT29-MTX cells, regardless of SIgA exposure, which may be due to the glycoprotein availability in mucous facilitating binding (SI Appendix, Fig. S2A). E. ramosum failed to adhere to cultured murine intestinal cells (CMT-93) (SI Appendix, Fig. S2B), suggesting a difference in interaction between human and mouse epithelial cells with this bacterium. E. ramosum was not highly sensitive to oxygen, compared to obligate and strict anaerobes (43, 44) (SI Appendix, Fig. S2G), and adhered to epithelial cells in both aerobic and anaerobic coculturing conditions (SI Appendix, Fig. S2H), confirming that the potential confounding effects of oxygen exposure in this assay do not substantially affect E. ramosum viability or behavior. Together, these data demonstrate that E. ramosum adheres to human intestinal epithelial cells and alters cell morphology, and that milk SIgA limits this direct bacteria–host interaction.

E. ramosum Induces Th17-Inducing Pathways in the Human Gut Epithelium.

The epithelial response of TC7 cells infected with E. ramosum was characterized by RNA-sequencing (RNA-seq). The expression of several inflammatory cytokines was significantly upregulated upon infection (Fig. 4A). Heat-killed E. ramosum failed to upregulate any of the inflammatory genes highlighted in Fig. 4A (P-value cut-off = 0.1). KEGG pathways that were significantly enriched by E. ramosum infection included Rheumatoid arthritis, Legionellosis, and IL-17 signaling, all of which involve Th17-mediated inflammation (Fig. 4B). The immune response was rescued by SIgA, as preincubation of E. ramosum with SIgA purified from human milk dampened the expression Cxcl8 and Cxcl3, two of the top Th17-inducing genes upregulated by E. ramosum according to RNA-seq analysis, in TC7 cells (Fig. 4 C and D). Although E. ramosum adhered to mucous-producing HT29-MTX cells, infection did not upregulate Cxcl8 or Cxcl3 expression, possibly due to impaired direct host–microbe interaction in the presence of a protective mucous layer (SI Appendix, Fig. S2 C and D). E. ramosum also failed to upregulate Th17-related cytokines in murine CMT-93 cells (SI Appendix, Fig. S2 E and F). In summary, E. ramosum induces pathways involved in Th17-mediated inflammation in human intestinal epithelial cells specifically, and human milk SIgA impairs this effect.

Fig. 4.

E. ramosum induces Th17-related inflammatory pathways in TC7 cells, and human milk SIgA rescues this response. (A) Volcano plot of gene expression changes in E. ramosum-infected TC7 cells compared to uninfected TC7 cells assessed by RNA-Seq analysis. Vertical lines indicate fold-change cut-off of 1.5. Horizontal lines indicate a P-value cut-off of 0.05. Upregulated genes involved in inflammation and immune response are highlighted in boxes. (B) KEGG pathways enriched by E. ramosum infection of TC7 cells assessed by RNA-Seq analysis. (C and D) Cxcl8 (C) and Cxcl3 (D) expression in TC7 cells infected with heat-killed E. ramosum (HK), E. ramosum (ECR), or E. ramosum preincubated with SIgA purified from human milk (ECR+IgA). Expression levels are represented as fold change compared to the housekeeping gene Btm (beta-2-microglobulin) relative to uninfected cells (CON). Statistical significance was determined by Kruskal–Wallis tests for (C) and (D). ns = not significant, *P < 0.05, **P < 0.01.

To determine whether bacterial adherence alone is sufficient to induce a Th17 response, we quantified the epithelial immune response to another adherent bacterium. We identified B. fragilis, which we show to be positively associated with milk SIgA (Fig. 1C), has been shown previously to induce Treg responses in vivo (45), and which adheres to TC7 cells at a similar concentration to E. ramosum (SI Appendix, Fig. S3A). In contrast to E. ramosum, B. fragilis failed to upregulate Cxcl8 or Cxcl3, indicating that microbial adherence alone is not sufficient to induce a Th17 response (SI Appendix, Fig. S3 B and C). Whether E. ramosum secretes a stimulatory factor or induces a response through specific adhesion machinery is yet to be determined. Genome annotation using Kbase (46) led to the identification of a fibronectin/fibronectin binding protein encoded by the E. ramosum genome as a likely mediator of bacterial adhesion. This gene is not present in the genome of B. fragilis, which is known to utilize surface glycans and SIgA to adhere to host cells (47). Fibronectin-binding proteins are present in numerous gram-positive bacteria and are required for virulence of some pathogens (48). Whether this protein is responsible for E. ramosum-induced Th17 responses specifically requires further investigation. However, this protein is likely involved in host interaction and immune responses to this bacterium.

To further elucidate the pathway from epithelial E. ramosum infection to Th17-mediated inflammation, human-derived monocytic dendritic cells (moDCs) cells were utilized in a sequential infection model. Supernatant from E. ramosum–infected TC7 cells, containing both bacteria and epithelial cytokines, drove increased differentiation of IL23+ DCs, CD11b+CD103+ DCs, MHCII+ DCs, all of which act to skew Th17 differentiation (49, 50) (Fig. 5 A–C). Expression of Il6, which encodes a cytokine that, together with Il23 drives Th17 differentiation, was also significantly upregulated (Fig. 5D). Expression of Tgfβ, which can trigger nonpathogenic or regulatory Th17 cell differentiation, was not induced, confirming that E. ramosum drives inflammatory rather than regulatory Th17 responses (51)(Fig. 5E). Expression of Il12 and Il33, Th1 and Th2-inducing cytokines, respectively, was not induced by E. ramosum (Fig. 5 F and G). Il12 expression was instead slightly downregulated by E. ramosum, supporting the skewing of moDCs toward Th17 pathways specifically (52). Epithelial cytokines in addition to live bacteria were required to induce this strong immune response, since supernatant of TC7 cells infected with heat-killed E. ramosum failed to upregulate Il6 and led to impaired moDC differentiation. Finally, to confirm that E. ramosum-induced moDCs drive Th17 responses, we exposed naïve human peripheral blood-derived CD4+ T cells to infected moDCs. E. ramosum induced expression of Il17, the primary cytokine produced by Th17 cells (Fig. 5H). Heat-killed E. ramosum and IL-6 alone failed to induce this response. These data demonstrate that the intestinal epithelial response to live and adherent E. ramosum drives a strong inflammatory Th17 response.

Fig. 5.

The epithelial E. ramosum response skews moDCs toward Th17-inducing phenotypes. (A–C) Flow cytometric quantification of (A) IL23+ cells. (B) CD11b+CD103+ cells, or (C) MHCII+ cells 72 h after exposure to supernatant of uninfected TC7 cells (CON), TC7 cells infected with heat-killed E. ramosum (HK), or TC7 cells infected with E. ramosum (ECR). (D–G) Il6 (D), Tgfβ (E), Il12 (F), and Il33 (G) expression in moDCs cells exposed to supernatant from TC7 cells infected with heat-killed E. ramosum (HK) or E. ramosum (ECR). Expression levels are represented as fold change compared to the housekeeping gene Btm (beta-2-microglobulin) relative to uninfected cells (CON). (H) Il17 expression in naïve human CD4+ T cells exposed to IL-2, anti-CD3, and moDCs analyzed in (A–G) with or without IL-6. Statistical significance assessed by the Mann–Whitney U test for (A–G). ns = not significant. **P < 0.01.

In contrast to our observations in human cells, E. ramosum failed to adhere to and did not upregulate Th17-related gene expression in mouse-derived CMT-93 cells (SI Appendix, Fig. S2). E. ramosum was also tested in vivo. Germ-free mice (GF) were infected with E. ramosum and Th17 cells were quantified in the small intestine and colon (SI Appendix, Fig. S4A). B. fragilis was administered to GF mice to serve as a negative control (SI Appendix, Fig. S4). Colonization was confirmed by qPCR using generalized 16s primers (SI Appendix, Fig. S4B). Infected mice did not display a Th17 response (SI Appendix, Fig. S4C). Since germ-free mice have deficient immune systems and display altered infection responses, we hypothesized that the presence of a microbiota may be necessary for E. ramosum to induce Th17 responses. To test this, SPF mice were pretreated with Vancomycin for 2 wk before oral gavage with E. ramosum (SI Appendix, Fig. S4D). Despite stable luminal colonization across the infected mice (SI Appendix, Fig. S4E), a significant Th17 response was not induced (SI Appendix, Fig. S4F). E. ramosum also failed to induce a general or E. ramosum-specific SIgA response in either in vivo experiment, whereas B. fragilis increased total fecal IgA levels in mice (SI Appendix, Fig. S4 G and H), further supporting limited contact between E. ramosum and host cells in the murine gut. E. ramosum thus behaves differently in the presence of murine cells, limiting the feasibility of an in vivo model.

Discussion

It is well established that human milk promotes long-term health and protects against several chronic diseases, as it is evolutionarily suited to meet the needs of the developing infant (13, 14). Understanding the mechanisms for this phenomenon is essential for designing human milk alternatives and promoting healthy milk production that optimally benefits the infant. In the present study, we explore the effects of one component of human milk, SIgA, in promoting infant immune health by modulating the infant gut microbiota.

Characterization of the SIgA levels and binding patterns of 300 mothers enrolled in the CHILD Cohort Study revealed that the general binding patterns of human milk align with those observed in human feces. Human milk SIgA targets Enterobacteriaceae species at high levels, in agreement with other work (32), whereas Bacteroidaceae species are not highly targeted by milk SIgA. Access to infant gut microbiota data from corresponding infants allowed us to directly compare human milk SIgA-binding to infant gut microbiota composition. This led to the observation that human milk SIgA binding patterns reflect infant gut microbiota composition: Most bacteria displayed positive associations with SIgA-binding, which were stronger than associations with total SIgA in human milk. Bifidobacteriaceae and Bacteroidaceae species in particular displayed strong positive associations with milk SIgA binding. This indicates that human milk SIgA-binding patterns are more predictive of offspring gut microbiota composition than SIgA quantity overall and may hint at a capacity of human milk SIgA to promote members of these families linked to beneficial effects during infant development (4, 53, 54).

Due to the lack of maternal microbiota data available through the CHILD Cohort Study, the relationship between the maternal microbiota and human milk SIgA could not be assessed. Future studies characterizing both the maternal and infant gut microbiota, along with human milk SIgA-binding capacities, would confirm the role of SIgA in the transfer of microbes between mother and infant. Breastmilk was also only collected at one time point in this study, limiting the comparison of SIgA to microbiota composition at various timepoints in early life. However, based on recent results from Johnson-Hence et al. (32), we would not expect any major shifts in microbiota reactivity of breastmilk SIgA over the course of the breastfeeding period. Another limitation in our approach was the lack of available infant stool from which to isolate bacteria of interest. In our work, milk SIgA was tested against bacteria grown from previously obtained isolates, as described in SI Appendix, Table S1. Although some SIgA is strain-specific, most intestinal SIgA is broadly reactive against conserved epitopes (1, 2). We do not expect that strain-specific differences would affect our results drastically, but future studies directly testing SIgA against infant stool isolates of bacteria would confirm the relationships described here.

Our lab has previously shown that milk SIgA limits colonization of SFB in the neonatal mouse gut, dampening Th17 responses and protecting against severe, Th17-skewed asthma later in life (21). In this study, we found that SIgA levels in maternal milk were inversely correlated with several cytokines involved in Th17 response regulation in infant serum at 1 y. The infants in this study were breastfed only for at least the first 3 mo of life, so immune marker data were collected long after the introduction of solid food and, in many cases, the cessation of breastfeeding. The Th17-skewed response in the serum of 1 y old infants who received less SIgA at 3 mo old suggest that the immune effects of human milk SIgA persist after the breastfeeding period, in agreement with our findings in mice. Within this study, we were not able to identify any associations between immune biomarkers and E. ramosum levels in infants or SIgA binding variables due to a low sample size. Analysis of serum immune biomarkers in a larger subset of infants, including nonbreastfed infants, would be useful in defining the role of E. ramosum in SIgA-associated early-life Th17 responses.

Collectively, this led us to hypothesize that human milk SIgA targets and limits Th17-inducing bacteria in the human infant gut. E. ramosum, the only tested species to display a significant negative correlation with SIgA binding in milk, emerged as a candidate, warranting further investigation into its dynamics and immunological effects.

E. ramosum has been identified previously as an age-discriminatory and keystone taxon in the infant gut microbiota, increasing in abundance over the first year of life (12, 31, 55). In infants enrolled in the CHILD Cohort Study, we found that E. ramosum expands significantly over the first year. E. ramosum levels were also significantly higher in nonbreastfed infants at 3 mo of age. The gradual progression of gut microbiota composition over the course of infancy has been shown previously to protect against asthma, and to be promoted by breastmilk (56). SIgA-mediated control of E. ramosum may be one mechanism for this phenomenon.

SFB displays similar dynamics in mice, expanding shortly after weaning in the absence of milk SIgA. A loss of SIgA-mediated control of SFB led to increased susceptibility to the HDM model of asthma (21), but others have shown that SFB can also limit infection and protect against autoimmune disease (29, 57), illustrating the importance of a balanced and well-timed SFB bloom in the early-life gut. In a recent meta-analysis of early life gut microbiota studies around the world, E. ramosum was found to be the top microbe associated with allergic disease outcomes later in life, regardless of geography, supporting its important and conserved role in immune development (31). This study noted E. ramosum as a potential biomarker for allergic disease, warranting further study. Interestingly, within the CHILD Cohort, E. ramosum levels were higher in control than allergic infants at 3 mo, a pattern that was lost by 1 y. This contradicts previous findings but hints at a potentially complicated relationship between E. ramosum and allergies that may not be fully explained by microbiota sequencing at only 2 timepoints in early life. As described for SFB, this study and others show that premature expansion as well as a lack of E. ramosum characterize the dysbiotic infant gut microbiota and associate with allergic and chronic immune-mediated disease outcomes. Balance and timing seem to be key to the effects of both E. ramosum and SFB in early life. Although these observations support a role for this E. ramosum in infant immune health and allergy development, the direct effect of this microbe on disease susceptibility can only be postulated based on the associations observed thus far.

To dive deeper into the host effects of E. ramosum, we employed human intestinal epithelial and immune cell culture models. We found that SIgA targets and limits epithelial adherence of E. ramosum, supporting the direct negative effect of human milk SIgA on colonization. RNA-seq analysis revealed that E. ramosum also potently induces inflammatory Th17-associated responses in human intestinal epithelial cells, and SIgA dampened this effect. SFB similarly induces Th17 responses through adherence to the intestinal epithelium of mice, an effect that is limited by SIgA (58, 59). By employing moDCs and naïve CD4+ T cells in a sequential infection model, we were able to show that live E. ramosum infection of the epithelium leads to differentiation of DCs which induce Th17 responses, and this was not mediated by the general proinflammatory cytokine, IL-6, alone. Our observation that B. fragilis, a bacterium that also adheres to TC7 cells, did not induce the same cytokine response suggests that epithelial adherence alone is not sufficient to promote Th17 differentiation. The factor that uniquely equips E. ramosum with the capacity to induce Th17 responses was not identified in this study. This comparison adds a layer of complexity to the well-studied relationship between bacterial adherence and Th17 responses (58), illustrating that adherence alone does not induce Th17-related responses in all cases, and potentially hinting at additional important shared features of adherent and Th17-inducing bacteria, which should be investigated further.

The use of epithelial cells grown in culture does not fully recapitulate the human intestine in structure, complexity, or nutrient and oxygen concentrations, limiting our ability to assess the degree to which this microbe adheres in the human gut. However, the consistent adherence phenotype observed across conditions, along with the conserved inverse relationship between SIgA and colonization in both the human cohort and in vitro analyses, support that E. ramosum is able to adhere to and interact with the human intestine. Animal experiments have enabled a mechanistic understanding of microbiota contributions to development and mouse models of allergies and IBD (51, 58, 60). Unfortunately, E. ramosum did not colonize or affect mouse cells or a mouse host, regardless of the presence of other microbes, limiting our ability to confirm the role of this species in early-life Th17 responses and asthma development. According to existing literature, E. ramosum causes bacteremia in immunocompromised patients (61, 62), evades SIgA binding in feces of both IBD patients and infants who develop allergies (33, 40), and produces an IgA protease (41), supporting the close interaction between this species and the immune system, and its capacity to contribute to inflammatory Th17 responses even in the complex human host. Although the link between SIgA, E. ramosum adherence, and Th17 responses could not be fully verified in vivo, the clinical associations observed in this and prior studies, together with our human epithelial and immune cell infection models, clearly demonstrate that E. ramosum is regulated by and interacts closely with the immune system.

Altogether, the data presented here point to some parallels between E. ramosum in humans and SFB in mice. Both microbes are limited by milk, expand slowly in early life, drive Th17 responses, and associate with allergic and chronic immune-mediated disease in complex ways. However, E. ramosum is phylogenetically distant from SFB, diverging at the class level (63). E. ramosum also differs substantially from SFB in growth requirements and aerotolerance (59). Although adherent, it is not determined whether E. ramosum attaches to cells using a holdfast structure. Additionally, the induction of Th17 responses by E. ramosum could not be modeled in a murine host. Thus, we do not propose that E. ramosum itself is a biological equivalent to SFB. However, its dynamics, associations with disease, interaction with SIgA, and in vitro effects on human cells closely recapitulate the colonization patterns and immunological effects of SFB in mice. Since SFB is known to be a vital contributor to both homeostasis and disease in mouse models, E. ramosum should be studied further as a likely key effector in infant immune development and Th17-related disease. Without direct in vivo evidence of E. ramosum’s effects, however, the parallel between SFB and this bacterium remains speculative.

Overall, this study characterizes human milk SIgA, identifies associations between SIgA and both infant gut microbiota and immune features, and illustrates an important role for human milk derived SIgA in limiting E. ramosum, a likely key player in immune development and susceptibility to allergic disease.

Materials and Methods

CHILD Study Sample and Data Acquisition.

This study utilizes existing data obtained through the CHILD Cohort Study, a longitudinal birth cohort including 3,624 participating mothers from 4 centers across Canada (64). The present study included a detailed analysis of a subset of 300 mother–infant pairs. Only mothers still breastfeeding (exclusively or not) for at least the first 3 mo of life were selected for breastmilk analysis. Infant fecal metagenomic data, health outcomes, and breastfeeding status were also obtained for 2,848 participating mother–infant dyads. Detailed information related to sample selection, data acquisition, and breastmilk collection are available in SI Appendix, Supporting Text.

Total SIgA ELISA.

For quantification of total SIgA in milk samples, high-binding ELISA plates (Greiner 650061) were coated with 100 μL/well of anti-human IgA (Sigma-Aldrich I1261) diluted 1:5,000 in carbonate-bicarbonate buffer (pH 9.6) overnight at 4 °C. Delipidated human milk samples were diluted 1:10,000 in Phosphate Buffered Saline (PBS) + 1% Bovine Serum Albumin (BSA) (Sigma-Aldrich A3294). Plates were washed 3× with wash buffer (PBS + 0.5% Tween 20) and blocked for 2 h at room temperature with PBS + 1% BSA. Plates were washed once, and samples were added in duplicate to wells. Human secretory IgA purified from colostrum was used to establish a standard curve (Bio-Rad PHP133). Samples and standards were incubated for 2 h at room temperature. Plates were washed 3× with wash buffer and 100 μL of anti-human IgA HRP (ab97215) (diluted 1:5,000 in PBS + 1% BSA) was added to each well before incubation for 1 h at room temperature in the dark. Plates were washed 3× with wash buffer and 100 μL of BD TMB substrate solution (BD 555214) was added to each well for detection of HRP. After 15 min, 2N Sulfuric Acid was used to stop the reaction and OD was measured at 450 nm. Diluted milk samples were added to uncoated wells as negative controls on each plate.

For quantification of total SIgA in feces, fecal pellets were incubated at room temperature in 1 mL of PBS for 30 min and then homogenized using a pipette. Homogenized fecal material was then centrifuged for 5 min at 8,000 g and fecal supernatant was collected and diluted 1:100 in PBS. The same assay procedure as above was carried out on diluted fecal supernatant. Anti-mouse IgA (Abcam ab97231) was used to coat the plate, and a mouse IgA ELISA standard (Thermo Fisher 39-50450-65) was used to establish the standard curve.

Bacteria-Binding SIgA ELISA.

For quantification of bacteria-binding SIgA, each bacterium listed in Table 2 (obtained as human fecal isolates or from DSMZ or ATCC) was cultured anaerobically overnight using a rich medium made up of 50% Fastidious Anaerobe Broth (NEOGEN 700004634) and 50% Reinforced Clostridial Media (BD 218081). Strain details are listed in SI Appendix, Table S1. Bacteria were diluted to an OD of 1, washed twice with PBS, fixed for 20 min at room temperature using 1% Paraformaldehyde in PBS, and resuspended in sodium acetate buffer (pH 5.5) for coating of high-binding ELISA plates (Greiner 650061). Standard wells were coated with anti-human IgA as above. Milk samples were diluted 1:200 in PBS + 1% BSA and the rest of the ELISA protocol was carried out as in the Total SIgA ELISA. Bacteria-coated wells with no milk added, and uncoated wells with milk added were used to establish background and as negative controls.

For quantification of mouse fecal E. ramosum binding SIgA, E. ramosum was used to coat an ELISA plate as described above. Mouse fecal supernatant was added to the plate, undiluted. Anti-mouse IgA (Abcam ab97231) was used to coat the plate, and a mouse IgA ELISA standard (Thermo Fisher 39-50450-65) was used to establish the standard curve. The assay was carried out with the same remaining reagents and parameters as above.

SIgA Purification from Milk.

The 20 milk samples with the highest or lowest E. ramosum binding SIgA levels were identified through ELISAs. 500uL of each sample was pooled to obtain 10 mL of high E. ramosum binding milk or 10 mL of low E. ramosum binding milk. The milk was centrifuged for 30 min at 2,000 g twice to remove the fat layer. SIgA was purified from the pooled milk using affinity chromatography as described previously (65). Briefly, CaptureSelect IgA-XL Affinity Matrix beads (Thermo 2943972050) were packed in a column and washed with PBS. Delipidated milk was added to the beads and beads were allowed to settle before draining. This was repeated until all milk had been run-through the column. 0.1 M Glycine (pH 3) was used to denature and elute the SIgA from the beads, and 0.5 M Tris-base (pH 8) was used to neutralize the collected elutions and restore protein structure. Neutralized, eluted SIgA was quantified by Total SIgA ELISA. The purified SIgA samples were stored at a total SIgA concentration of 0.5 g/L. The high E. ramosum binding sample contained 9.72 mg/L of E. ramosum binding SIgA. The low E. ramosum binding sample contained 3.07 mg/L of E. ramosum binding SIgA.

Defining Associations between Milk SIgA Binding and Infant Microbiome and Biomarker Data.

ELISA results were compiled and analyzed in R. Associations between SIgA binding variables in milk, or infant exact age at time of fecal collection, and species relative abundance in infant feces were determined using the Maaslin2 package (1.12.0) with the following parameters: minimum feature prevalence of 0.1, no normalization, log transformation. The exact age of infants was always included as a “fixed effect” and site (city in which participant was located) was treated as a “random effect” in all comparisons.

Epithelial Cell Culture and E. ramosum Infection.

TC7 cells (human absorptive colonic epithelial cells) (Sigma-Aldrich SCC209), CMT-93 cells (mouse colonic epithelial cells) (ATCC CCL-223), or HT29-MTX cells (human mucous-producing colonic epithelial cells (European Collection of Authenticated Cell Cultures Cat#12040401) were grown in DMEM+++ [DMEM, 10% v/v heat-inactivated fetal bovine serum (FBS), 1% v/v nonessential amino acids, and 1% v/v Glutamax] in T-75 flasks at 37 °C in 5% CO₂. Cells were split using 0.25% Trypsin when they reached 80 to 90% confluency (every 3 to 4 d) at a 1:6-1:10 dilution. Between passages 2 and 10, cells were seeded in 24-well plates at 50,000 cells per well in DMEM+++ for adherence assays and qPCR/RNA-seq experiments. Cells were seeded in 8-well chamber slides (Ibidi 80826) at a density of 10,000 cells per well for visualization of E. ramosum adherence. Cells were allowed to reach confluency for 3 d before infection for TC7 and CMT-93 cells, or 7 d to allow for mucous layer production for HT29-MTX cells.

E. ramosum was grown anaerobically as described above. Bacteria were then washed 1× with PBS and resuspended in DMEM+++ aerobically. For preincubation with SIgA, bacteria were resuspended in DMEM+++ with 5 mg/L high- or low-E. ramosum binding SIgA purified from pooled human milk and incubated for 2 h at 37 °C and 5% CO₂ before exposure to cells. For preincubation with heat-denatured SIgA, DMEM+++ with 5 mg/L SIgA purified from human milk was boiled for 10 min, quickly placed on ice for 10 min, and then used to resuspend bacteria before incubation. TC7, CMT-93, or HT29-MTX cells in 24-well plates were washed once with PBS before adding the resuspended E. ramosum (20,000 CFU per well). TC7 cells in 8-well chamber slides were washed once with PBS before adding resuspended E. ramosum (8,000 CFU/well). Cells were incubated at 37 °C and 5% CO₂ for the periods indicated below.

Adherence Assays.

For adherence assays, epithelial cells were exposed to bacteria in DMEM+++ with or without SIgA for 2 h at 37 °C and 5% CO₂ to allow for adherence to occur. The supernatant, containing nonadherent bacteria, was brought into an anaerobic chamber, serially diluted, and plated anaerobically using 50% FAB 50% RCM plates, which were previously kept anaerobic for 24 h for CFU counting. The cells were then washed vigorously 3 times with PBS to remove any remaining nonadherent bacteria. 0.25% Trypsin was used to detach epithelial cells and adherent bacteria, which were serially diluted and plated anaerobically for CFU counting of adherent bacteria as described.

For anaerobic adherence, the same procedure was carried out in an anaerobic chamber. Media were kept anaerobic for 48 h before the assay, bacteria were grown and resuspended in DMEM+++ anaerobically, TC7 cells were brought into the anaerobic chamber for bacteria exposure, and dilutions and plating were carried out anaerobically.

Oxygen Tolerance Test.

Bacteria were cultured anaerobically as described above. Bacteria were diluted from an overnight culture to 3.5 × 10⁶ CFU/mL and resuspended in DMEM+++. Bacteria were counted by serial dilution and CFU plating anaerobically, and then placed in an aerobic, 5% CO₂ incubator at 37 °C for 6 h, before serial dilution and CFU plating to quantify viable bacteria after 6 h of oxygen exposure. CFU counts were used to quantify % viability.

Visualization of E. ramosum Infection.

For microscopy, TC7 cells in 8-well chamber slides were exposed to bacteria or DMEM+++ containing no bacteria for 2 h. Cells were washed three times with PBS and then fixed with 4% paraformaldehyde for 20 min at room temperature. Cells were washed three times with PBS and then stored in PBS at 4 °C until staining. Cells were washed two times with PBS and incubated with 300 μL of 0.1% Triton in PBS for 5 min at room temperature. Cells were washed 3× and blocked with 300 μL of PBS containing 5% normal goat serum and 1% BSA for 30 min at room temperature. Block buffer was removed and 200 μL of mouse anti-lipoteichoic acid (1:50) (Invitrogen MA1-7402) and rabbit anti-Occludin (1:250) was added. Cells were incubated with primary antibodies overnight at 4 °C. Cells were washed twice and stained with Goat anti-mouse 488 (1:200), Goat anti-rabbit 568 (1:200), Phalloidin 647 (1:40), and DAPI (5 μg/mL) for 45 min at room temperature, shielded from light. Cells were washed 3×, mounted with a Prolong Diamond antifade reagent, and visualized under a Leica Stellaris 5 Confocal microscope using a 63× objective. Images were processed with LAS X Lighting Expert.

TC7 Cells RNA-seq.

For RNA-seq, TC7 cells were infected for 6 h with bacteria as described above. The supernatant was then removed, and the cells were washed three times with PBS. Subsequently, 200 μL of RNAProtect Tissue Reagent (Qiagen 76104) plus DMEM (5:1 v/v) was added to each well. A cell scraper was used to dislodge cells. RNA was extracted using the RNeay mini kit (Qiagen 74104). For RNA-seq, the Agilent 4200 TapeStation was used for quality control (QC), and samples were prepared using the standard protocol for the Illumina Stranded mRNA Library Prep kit. Sequencing was performed with the Illumina NextSeq2000 with Paired End 59 bp × 59 bp reads. Resulting data were demultiplexed using the Illumina BCL Convert and sequences were aligned to the Homo sapiens (hg38) reference sequence using the DAGEN RNA app on Basespace Sequence Hub. FastQC and STAR were used for data processing and alignment, and the DESeq2 package was used in R for differential abundance analysis. The EnhancedVolcano package was used in R for generating volcano plots.

For KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment analysis, the functional enrichment analysis tool in the STRING database (https://string-db.org/) was used. Enrichment scores, P-values, and FDR-corrected values were extracted and visualized in R.

Cytokine RT-qPCRs.

For RT-qPCRs, RNA was extracted from TC7 cells, CMT-93 cells, HT29-MTX cells, or moDCs as described above for RNA-seq. RNA was converted to cDNA using the QuantiTect Reverse Transcription Kit (Qiagen 205311). RT-qPCRs were performed using the QuantiNova SYBR Green PCR Kit (Qiagen 208052) and the primer pairs listed in SI Appendix, Table S2.

Genome Annotation Using Kbase.

To identify protein-coding genes within bacterial genomes, the RefSeq genomes for E. ramosum (NZ_CP036346) and B. fragilis (NZ_CP069563) were downloaded using KBase (46). The genomes were then annotated using RASTtk (v1.073) to define the list of protein-coding genes. The resulting SEED function lists were assessed visually to identify genes related to bacterial virulence and disease within each genome. Under the SEED function “Adherence,” only E. ramosum displayed a hit. B. fragilis did not display any protein-coding genes related to this function.

Monocytic Dendritic Cell Culture.

THP-1 cells (peripheral blood monocytes) (ATCC TIB-202) were cultured in rich RPMI media made up of RPMI 1640 (Thermo Fisher 11875093) plus 10% v/v FBS and 1% v/v Gultamax. Cells were split once they reached 8 × 10⁵ cells/mL to a concentration of 1 × 10⁵ cells/mL in T-75 flasks. Between passages 2 and 10, 1 × 10⁵ cells were seeded in a T-25 flask in 5 mL of rich RPMI media with the addition of 100 ng/mL GM-CSF (Gibco 300-03-5UG) and 100 ng/mL IL-4 (Gibco 200-04-1MG) for monocytic dendritic cell (moDC) differentiation. Media and cytokines were replaced once every 3 d and cells were used between days 2 and 7.

moDC Cell Sequential Infection.

For supernatant exposure experiments, moDCs were seeded at 50,000 cells/well in 24-well plates in rich RPMI media. Supernatant from TC7 infection experiments (6-h infection, described above) was added in a 1:1 (v/v) ratio to dendritic cells. Cells were left for 48 to 72 h at 5% CO2 and 37 °C to allow for moDC maturation. Cells were then collected and resuspended in RNAprotect Tissue reagent for qPCR as above or washed twice in PBS and analyzed by flow cytometry.

moDC Flow Cytometry.

For flow cytometric analysis, moDCs were stained with FITC anti-human CD86 (Biolegend 374203), APC anti-human MHCII (Biolegend 326811), PacBlue anti-human CD11b (Biolegend 982602), and PE-Cyanine7 anti-human CD103 (Thermo Fisher 25-1038-42) diluted 1:100 in FACS buffer [1× HBSS (Gibco 14175095) plus 2 mM EDTA, 10 mM HEPES (Thermo Scientific J60111.AK), 5 mM FBS] for 30 min at 4 °C. Cells were then washed twice in FACS buffer and permeabilized using the BD Cytofix/Cytoperm kit (BD AB_2869008). Cells were resuspended in PE anti-human IL-23 (Biolegend 501806) diluted 1:50 in BD Perm/Wash buffer for 30 min at 4 °C for intracellular staining. Cells were then washed twice and resuspended in FACS buffer for flow cytometry using a CytoFLEX L Analyzer. Compensation and visualization were performed using FlowJo (version 10.7.1).

T Cell Differentiation Assay.

Human Peripheral Blood CD4+ T Cells (Lonza 2W-200) were thawed in Iscove’s Modified Dulbecco’s Medium (IMDM) (Gibco 12440079) with 5% FBS, centrifuged at 300 × g for 10 min, and resuspended in IMDM + FBS with 5 ng/mL IL2 (PeproTech 200-02) and 150 ng/mL anti-CD3 (Invitrogen 16-0037-81) to induce proliferation. T cells were then counted and added to a 24-well plate at ~80,000 cells per well. moDCs from the sequential infection protocol above were added to the T cells. Half of the contents of each well of moDCs were added to a single well of T cells. IL-6 (PeproTech 200-06) was added at 100 ng/mL to some wells. moDCs were cultured with T cells for 5 d at 5% CO₂ and 37 °C to allow for moDC maturation. T cells were collected, resuspended in RNAprotect Tissue reagent, and characterized by qPCR as described above.

Germ-Free Mouse Experiment.

4-week-old male GF C57BL/6 mice were gavaged with 10⁹ CFU of B. fragilis or E. ramosum grown anaerobically and resuspended in PBS, or PBS alone. Mice were killed 10 d after gavage and intestines were collected and intestinal Lamina Propria Lymphocytes (LPLs) were analyzed by flow cytometry.

SPF Mouse Experiments.

4-wk-old SPF C57BL/6 mice of the heterozygous Pigr+/− (MMRRC 30988) strain we have used previously (21) were treated with Vancomycin in drinking water (0.5 g/L) for 2 wk. Mice were switched to antibiotic-free water for 24 h before oral gavage with 10⁹ CFU of E. ramosum grown anaerobically and suspended in PBS, or PBS. Mice were killed 10 d after gavage and intestinal LPLs were analyzed by flow cytometry. Feces was also collected at the experimental endpoint.

Mouse Ethics Statement.

Animal experiments were performed in accordance with guidelines from the Canadian Council on Animal Care and the University of British Columbia (UBC) Animal Care Committee. Experiments were approved by the committee under the animal protocol #A21-0286.

E. ramosum RT-qPCR in Feces.

DNA was extracted from mouse fecal pellets using the QIAmp PowerFecal Pro DNA QIAcube Kit (Qiagen 51826). RT-qPCR was performed on the DNA samples using the method described above and the following primer pairs: Universal 16s (F:ACTCCTACGGGAGGCAGCAGT, R:ATTACCGCGGCTGCTGGC) (21), E. ramosum 16s (F:AGGATGGACTTATGGCGCATT, R: TACCGTCACTCGGCTACCAT) (66).

Lamina Propria Lymphocyte Flow Cytometry.

LPLs were isolated according to a protocol described previously from the small intestine and colon (67). Flow cytometric quantification of T cell subsets was carried out as performed previously (21), and as described above for moDCs. The following antibodies were used for flow cytometry on mouse T cells: anti-mouse CD45-A700 (Invitrogen 56-0451-82), anti-mouse CD3-APC eFluor 780 (eBioscience 47-0032-82), anti-mouse CD4 eFluor 506 (Invitrogen 69-0042-82), anti-mouse RORγt APC 17-6891-82), anti-mouse Tbet-1 eFluor 450 (Invitrogen 48-5825-82), anti-mouse Foxp3 PE (Invitrogen 12-5773-82).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

Raw sequencing files obtained in RNA-sequencing used in the publication been uploaded to NCBI Sequence Read Archives at the BioProject accession code: PRJNA1279607 (68). The human participant data analyzed in this study are subject to restrictions: data will be made available upon request from the CHILD Cohort Study, and information can be found at https://childstudy.ca/for-researchers/data-access/. Requests to access these datasets should be directed to child@mcmaster.ca. The shotgun metagenomic data on infant feces used in this study have been made available in the NCBI database under BioProject accession code PRJNA838575 (69). All R code used in the publication can be found at https://github.com/kdonald-code/eramosum-pnas-code (70). All remaining study data are included in the article and/or supporting information.