The part of the study involving human material was carried out in accordance with the recommendations of the Human Ethics Committee at Huddinge University Hospital, Stockholm, and the parents provided their informed verbal consent. No written documentation of the participants’ informed approval was required, which was agreed to by the Human Ethics Committee and was according to the regulations at the time of the initiation of the study.

The murine study was approved by and carried out in accordance with the recommendations of the Regional Animal Research Ethical Board, Stockholm, Sweden.

The fecal samples were collected from children participating in a prospective birth cohort of children with different AH (16). From a subgroup of these children, fecal samples were analyzed for the presence of lactobacilli in feces during early life (14). The samples were collected between 2 and 8 weeks after birth and the bacterial content screened. Techniques for collection and microbiota analysis of the human infant fecal samples have been reported previously (13, 14). The children were subdivided based on history of atopic disease in the immediate family [AH or no allergic heredity (NH)], and on the presence of lactobacilli type I in feces at the time of, or close to sampling (L). For the human microbiota-associated mouse model, individual infant fecal samples were randomly selected from three different groups of children, previously associated with a low- or high-allergy risk (14): the no allergic heredity, lactobacilli+ (NH_L) group (low risk) (n = 7), allergic heredity, lactobacilli+ (AH_L) group (low risk) (n = 6), and AH group (high risk) (n = 6), pooled within each group and used as the initial inoculum. In the NH_L and AH_L groups, a minority (1/7 and 1/6, respectively) of children became allergic. All (6/6) children providing samples for the AH pooled material developed allergy.

Work with axenic and conventionalized C57BL/6 mice was carried out at the Core Facility for Germfree Research at the Karolinska Institutet, Stockholm, Sweden. Mice were kept in sterile isolators and were fed autoclaved feed (R36, Lantmannen, Sweden) and water ad libitum. Mice were colonized with human infant microbiota at 6 weeks of age. Infant fecal samples were diluted in sterile phosphate-buffered saline (PBS) (0.3 g/ml) and brought to a homogenous solution. The mice were then orally gavaged with 0.1 ml of fecal PBS solution. After 6 weeks, mice were bred. A booster gavage was administered to the females on gestational days 15–17. The offspring were sacrificed at 7–8 weeks of age. The fat to lean mass body composition was measured by magnetic resonance imaging (MRI) technique (EchoMRI-700/100 Whole Body Composition Analyzer, Echo Medical Systems).

Spleens and Peyer’s patches (PP) were removed and passed through a cell strainer (90 µm, Falcon) to obtain a single cell suspension. Splenocytes were treated with red blood cell lysing buffer (Sigma-Aldrich), washed and resuspended in complete medium: RPMI 1640 with FCS (10%, Gibco) and penicillin–streptomycin (100 U/ml and 100 µg/ml, respectively, HyClone). Cells from PP were cultured in complete medium but with additional gentamicin (50 µg/ml, Gibco) to prevent bacterial contamination. After isolation, cells from spleen and PP (3–5 × 10⁶ cells/ml) were either left unstimulated or polyclonally stimulated with αCD3 (1 µg/ml, Biolegend) and αCD28 (5 µg/ml, BioLegend) for 24 h. Cell culture supernatants were harvested and stored at −80°C.

Colonic lamina propria (cLP) leukocytes were isolated as previously described (17). In brief, colon tissue was incubated with EDTA (5 mM) to eliminate epithelial cells, followed by digestion with collagenase 8 (0.5 mg/ml, Sigma Aldrich) and DNase 1 (40 µg/ml, Roche). The released cells were layered on a 30/40/70% Percoll gradient (GE Healthcare).

Cells from PP, colon, and spleen were treated with Ultra-LEAF anti-mouse CD16/32 (Clone 93, BioLegend) and subsequently stained with antibodies against surface markers: CD45-A488 (30-F11), CD3ε-BV421 (145-2C11), CD4-APC/Cy7 (GK1.5), CD25-PE/Cy7 (PC61), and CD304 (Neuropilin-1)-PE (3E12) (all from BioLegend). Intracellular staining for transcription factors was performed by fixing and permeabilizing cells using True nuclear buffer kit (BioLegend) and subsequent blocking with normal mouse serum and staining with anti-FOXP3-APC (150D, BioLegend) and RORγt-PerCP-Cy5.5 (Q31-378, BD Biosciences). Intracellular staining of cLP leukocytes was performed using the Foxp3 Staining Buffer Set (eBioscience). T-cell populations were characterized based on forward and side scatter properties and CD45 expression. For B-cell analysis, single cells isolated from PP and spleen were stained using the following antibodies: CD45-A488 (30-F11), B220-APC/Cy7 (RA3-6B2), CD21/CD35-APC (7E9), IgM- PerCP/Cy5.5 (RMM1), IgA-biotin (RMA-1), and BV421-streptavidin (all from BioLegend).

To distinguish live cells LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Invitrogen) was used. After staining, cells were acquired on an FACSVerse flow cytometer and data analyzed using FlowJo software.

Cytokines were measured in cell culture supernatants using a mouse Luminex panel (Affymetrix eBioscience) according to the manufacturer’s instructions. The levels of IgG in serum and IgA in serum and small intestinal content were measured by enzyme-linked immunosorbent assay (ELISA) with commercially available ELISA kits (Mabtech).

Sections of distal small intestine and mid-colon were snap-frozen at necropsy. Tissues were disrupted using a homogenizer (MP Biomedicals) and total RNA was isolated using the RNEasy mini kit (Qiagen). cDNA synthesis was performed with the Superscript III system (Invitrogen) and messenger RNA (mRNA) expression of housekeeping gene Hprt (F: 5′-CCCAGCGTCGTGATTAGC-3′, R: 5′-GGAATAAACACTTTTTCCAAATCC-3′), Reg3g and Retnlb (QuantiTect Primer assays, Qiagen) was measured by SYBRgreen qPCR using KAPA SYBR FAST Universal qPCR kit (KAPA Biosystems) and a LightCycler 480 (Roche Life Science). Results were calculated using the 2^−ΔC(t) method.

Distal small intestine was embedded in O.C.T. compound (VWR chemicals) and cut in 10 µm sections. Sections were stained with 1° antibodies mouse anti-E-cadherin (1.25 µg/ml, clone 36/E-Cadherin, BD Biosciences) and rabbit anti-Ki67 (1/200, clone SP6, Abcam) and 2° antibodies goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 (4 µg/ml, both from Abcam) and DAPI (1/2,000, Cat No. D1306 Invitrogen). Sections were mounted with ProLong Gold antifade reagent (Molecular Probes). Images were acquired on an LSM 780 confocal microscope (Zeiss) using the 20× magnification objective.

The non-parametric Kruskal–Wallis test with Dunn’s multiple comparisons test was performed. GraphPad Prism software, version 7, was used. Data are displayed with median values or alternatively with mean ± SD. Results were considered to be statistically significant if p < 0.05.

In order to investigate the links between early-life microbial exposure of the intestinal tract and subsequent immune system phenotype, the effects of different types of human microbiota were tested by transplanting infant microbiota into GF mice and examining the offspring of these mice. We used fecal samples collected from children aged 2–8 weeks who were part of the previously described cohort (14, 16). These children were all healthy and born full-term. To avoid factors known to affect the composition of the microbiota, the children had all been vaginally delivered, were exclusively breastfed and had not been treated with antibiotics during the time of sampling. Microbiota NH_L and AH_L were pooled stools sampled from children with NH or AH, respectively, and with a known presence of group I lactobacilli (L) (Figure 1A). These children were at low risk of exhibiting childhood allergy. Microbiota AH was pooled stool from children with AH but with no trace of group I lactobacilli (Figure 1A). These children all developed childhood allergy. GF mice were fed the differently sourced pooled infant stools by oral gavage. The immune responses in the offspring of the formerly GF mice were subsequently studied when the animals had reached 7–8 weeks of age (Figure 1B). As part of the general characterization, mice were weighed at euthanasia (Figure S1A in Supplementary Material), and although the AH mice as a group displayed significantly lower weight than the AH_L mouse group, no differences in proportions of fat to lean mass as measured by MRI technique were observed between groups (Figure S1B in Supplementary Material).

The PP are major sites for the induction of intestinal immune responses (21). Initial characterization of the PP revealed no differences in total cell numbers (Figure 2A) and all groups exhibited similar proportions of B-cells and total T-cells (Figures S2A,B in Supplementary Material). The AH_L mice had reduced proportions of CD4⁺ T-cells (Figure 2B), but no differences in the proportion of FOXP3⁺CD4⁺ T-cells were detected between the groups (Figure 2C). However, when investigating the potential development of effector T-cells in the FOXP3^–CD4⁺ T-cell population, a distinct increase of RORγt⁺ T-cells in the AH group was observed (Figure 2D), as well as significantly increased production of IL-17A by PP cells stimulated with αCD3 and αCD28 (Figure 2E). The complete gating strategy is displayed in (Figure S3 in Supplementary Material). Together these findings strongly indicated the induction of intestinal Th17-cells in animals exposed to the AH microbiota.

We further investigated CD4⁺ T-cell responses in the colon, for which the complete gating strategy is displayed in Figure S4 in Supplementary Material. Initial characterization of the cLP T-cells showed that the total CD4⁺ T-cell population was increased in the AH group (Figure 3A), although the proportion of FOXP3⁺CD4⁺ T-cells was slightly reduced in this group (Figure 3B). When examining the FOXP3⁺ cell population further, the AH cohort emerged with a distinct shift to RORγt-expressing cells (Figure 3C), indicative of regulatory T-cells (Treg) that had been induced in the periphery after microbial exposure (22–24). NRP1-expressing cells, regarded as thymic-derived Treg (25), constituted a smaller fraction of the total FOXP3⁺CD4⁺ T-cell population in these mice (Figure 3C). Analysis of the FOXP3⁻ compartment of CD4⁺ T-cells revealed that similarly to the PP, RORγt expression was markedly increased in the AH group compared with both the NH_L and the AH_L groups (Figure 3D). Overall, the AH microbiota provided robust cues for CD4⁺ T-cell accumulation and RORγt expression in the gut.

We further examined systemic T-cell populations and responses by isolating cells from the spleen. The AH mice displayed a significantly increased percentage of CD3⁺ T-cells compared with the NH_L mice (Figure S5A in Supplementary Material), but the increased population was not due to accumulation of CD4⁺ T-cells (Figure 4A). Furthermore, no differences in the percentage FOXP3⁺ T-cells (Figure 4B), or in αCD3/αCD28-induced secretion of IL-10 from splenocytes (Figure 4C) were observed. In order to investigate general Th1 and Th2 responses, splenocyte production of IFNγ and IL-4 following stimulation with αCD3/αCD28 was examined, but no significant differences were noted between groups (Figure S5B in Supplementary Material).

In line with the gut compartments, the proportion of RORγt-expressing CD4⁺ T-cells was significantly increased in the AH group (Figure 4D). Also, splenocytes isolated from AH mice produced significantly more IL-17A following αCD3/αCD28 stimulation (Figure 4E), revealing increased Th17 responses. In addition, IL-22, which can be produced by Th17-cells (26), was increased in stimulated splenocytes from the AH group (Figure 4E). Production of innate cytokines IL-6 and IL-1β, which are needed for development of functional Th17-cells (27, 28), was similar in the experimental groups (Figure S5C in Supplementary Material). The level of IL-23, which drives IL-22 expression by Th17 (29), was slightly increased in the AH group (Figure S5C in Supplementary Material).

B-cell responses, such as IgE production, are also affected by microbial context (30), which could have an impact on later life establishment of allergy. Levels of serum IgG and serum IgA were significantly elevated in the AH group (Figure 4F), indicating that increased B-cell class switching had occurred. Percentages of total B-cells, IgM⁺ B-cells and IgA⁺ B-cells were equivalent between groups (Figures S6A–C in Supplementary Material). In addition, no differences in the proportion of marginal zone B-cells, which are known producers of natural antibodies directed at the commensal microbiota (31), could be detected (Figure S6D in Supplementary Material). The complete gating strategy for B-cells is displayed in (Figure S7 in Supplementary Material). The elevated serum antibody levels show that AH mice displayed altered functional B-cell properties; however, this was not reflected in changes in the splenic cellular B-cell compartment.

Considering that the AH mice exhibited both increased intestinal and systemic Th17 responses, this could indicate that they harbored more reactive or pathogenic microbiota. The intestinal mucosa was therefore examined for alterations in barrier properties or the presence of obvious inflammation. Antimicrobial peptides (AMP) play an important role in protecting the intestinal epithelium. The C-type lectin REGIIIγ is known to exist in high concentrations in the small intestine, where it is produced by IEC and contributes to maintaining the barrier (32). All experimental groups displayed equivalent Reg3g mRNA expression levels in small intestinal tissue (Figure 5A), excluding differential effects of particular microbes on a major AMP. In the colon tissue, only low levels of Reg3g could be detected (Figure 5B). Further, no differences in expression of the AMP Retnlb, which is constitutively expressed by the colonic epithelium (33), could be observed (Figure 5B).

Altered epithelial cell turnover could be a sign of acute infection with intestinal pathogens or of chronic inflammation of the gut. IEC proliferation was determined in small intestinal tissue by immunofluorescence staining of Ki67. Actively proliferating IEC were similarly localized in the crypts in all experimental groups (Figure 5C), indicating that the Th17-skewed response observed both locally and systemically, was not secondary to intestinal disease. Furthermore, myeloid immune cell frequencies in PP were similar between the experimental groups, as was the composition of intraepithelial lymphocytes in small intestinal tissue (data not shown). However, secretory IgA, which is induced during colonization with intestinal bacteria (34), was significantly elevated in the AH group compared with the NH_L animals (Figure 5D), again indicating that AH microbiota was associated with more robust mucosal immune responses.

Although the cecum and colon harbor the highest loads of bacteria in the intestinal tract, the small intestine is an important site for the induction of gut immune responses. In order to obtain a comprehensive view of gut microbiota composition in the experimental groups, content from small intestine, cecum and colon were collected from individual animals at dissection and their microbiome composition analyzed separately using 16S rRNA sequencing. The major phyla detected throughout the intestinal tract were Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria (Figures 6A–C), which was in line with a previous study of human gastrointestinal microbiota (35).

Overall, the small intestine displayed a distinct microbial profile compared with the cecal and colonic compartments, which were similar within the respective groups. Bifidobacteriaceae, a group of bacteria known to metabolize human milk oligosaccharides (36), were ubiquitous, although the relative abundance was higher in NH_L mice compared with AH mice in all three sampled sites (Figures 6A–C). In the small intestine, the Erysipelotrichaceae were a characteristic of the NH_L group (Figure 6A) and could only be detected at very low abundance in the AH animals (Table S1 in Supplementary Material). Similarly, the Porphyromonadaceae were typical of the AH group (Figure 6A) but were only present at low-relative abundance in NH_L mice (Table S1 in Supplementary Material). AH_L animals differed markedly from both NH_L and AH animals in all three sampled sites. In the small intestine, the Enterobacteriaceae dominated (Figure 6A) and AH_L mice exhibited a lower diversity index at this site, while diversity remained high in both the NH_L and AH mice at all sampled loci (Figure 6D).

In both the NH_L and AH cohorts, the cecal and colonic microbiota were dominated by Bacteroidetes, more specifically by Bacteroidaceae in the NH_L mice and by Porphyromonadaceae in AH mice (Figures 6B,C). Firmicutes constituted the second most abundant phylum in the lower intestinal tract and were mainly comprised of Erysipelotrichaceae and Clostridiaceae in the NH_L mice (Figures 6B,C). Although AH mice harbored Clostridiaceae in the colon, the relative abundance was low and two other members of Clostridiales, Lachnospiraceae, and Peptostreptococcaceae were the most prevalent Firmicutes in both cecum and colon of this group (Figures 6B,C). The AH_L group exhibited traits associated with both NH_L and AH animals, in that both Erysipelotrichaceae and Porphyromonadaceae were present (Figures 6B,C). Contrary to both the NH_L and AH cohorts, only smaller proportions of Firmicutes could be detected in AH_L mice and members of Clostridiales were largely absent (Figures 6A–C; Table S1 in Supplementary Material). Non-metric multidimensional scaling analysis of microbiota revealed that the dissimilarities within the groups were particularly small in the AH_L group (Figure 6E).

The human commensal fungus C. albicans has been shown to induce Th17 responses via activation of the innate immune system (37). Therefore, the presence of fungi was examined in the experimental groups by PCR screening of DNA extracted from cecal content using general fungal ITS primers. The results revealed fungal presence in the AH group of mice as evidenced by the positive signal for this 18S ITS (Figure S8 in Supplementary Material). Sequencing of the ITS region identified reads that were confirmed to be C. albicans in the AH mice (data deposited in the NCBI BioProject with accession number PRJNA382773).