Microbial exposure during early human development primes fetal immune cells

1The human fetal immune system is known to develop early during gestation with significant diversity in its immune repertoire and apparent sensitivity toward external antigens, thus indicating a much more competent and diversified system than previously thought. We have shown previously that fetal dendritic cells (DCs) promote T cell-mediated immune suppression during early fetal development, indicating an active and responsive immune architecture during the 2nd trimester of gestation. However, the factors responsible for such in utero priming and their effect on the immune system of the developing fetus remain elusive.

2Here, using a mass cytometry-based approach to study tissue-specific human fetal T cells across multiple organs, we identified cytotoxic T cells and most notably effector memory T cells (Tems) across the 2nd trimester of gestation, in addition to regulatory T cells (Tregs). Tregs are known to be essential for the regulation of commensal microbe-specific effector and memory responses during homeostasis. Several studies have recently detected microbe traces in human placenta and fetal samples. However, active microbial presence in utero is still a topic of constant debate, and novel sensitive approaches are required to understand the complexities of human gestation. Thus, we explored the presence of microbes in fetal tissues and their potential role in priming and activation of memory T cells during fetal development. Using 16S rRNA gene sequencing approach, we detected a low but consistent microbial signal across several fetal tissues in the 2nd trimester of gestation with a significant increase in bacterial diversity compared to PBS controls. Sequencing results were validated by aerobic and anaerobic culturing of fetal tissues, followed by isolation and identification of several live bacterial strains. To visualize the bacteria in fetal tissues, we performed scanning electron microscopy (SEM) and RNA-in situ hybridization (RNA-ISH) and identified differentially located bacteria within the fetal intestinal lumen. Notably, fetal-isolated bacterial strains induced memory T cell activation and expansion of fetal lymph node T cells, thereby indicating a previous encounter and priming of fetal immune system with bacterial antigens. Altogether, our observations suggest a sparse microbial presence in fetal tissues and their potential priming of the fetal immune system, thereby leaving the footprints of microbial memory during early development. These findings have broad implications for our understanding of human development and the establishment of a competent and diverse immunity before birth.

3To study the fetal immune system and its microbial components, we established a careful protocol for collection and processing of fetal tissues (see STAR Methods for detailed collection and processing protocol). The study is divided into four stages: T cell profiling by mass cytometry (CyTOF), 16S rRNA gene sequencing for bacterial identification, culture-based sequencing of live fetal bacteria, and in vitro T cell proliferation assay to demonstrate microbial immune recognition (Figure S1A). The samples collected were split between different arms of the study depending on the sample conditions and resource availability (Table S1).

4We first profiled T cells from fetal lung, skin, gut, thymus, liver, mesenteric lymph nodes (mLN), and blood by using state-of-the-art CyTOF approach as recently described (antibody list in the Key resources table). Fetal spleens were not sufficiently large enough to use for both T cell analysis and bacterial DNA sequencing; hence, they were excluded from T cell profiling. Unsupervised uniform manifold approximation and projection (UMAP) analysis of T cells from fetal (13–23 weeks estimated gestational age [EGA]) and adult samples revealed that fetal, adult, and cord blood clustered together at one end of the UMAP, whereas the fetal and adult organs formed two distinctly separate clusters (Figure 1A). This suggests organ-specific priming of T cells during the 2nd trimester of gestation. A small fraction of non-T cells were also present, which were removed from further analysis. We next separated all blood samples from tissues followed by independent UMAP analysis for both datasets. The unsupervised clustering of all blood T cells further showed no distinct separation among fetal, adult, and cord blood samples based on our panel of markers (Figure 1B). However, the tissues consistently showed distinct separation between fetal and adult organs in UMAP analysis (Figure 1C).

5We identified fetal- and adult-specific CD4+ and CD8+ T cells based on the marker projections as well as CD4+FOXP3+CD25+CTLA4+ Tregs, which formed distinct clusters for both fetal and adult organs (Figures 1D and 1E). The tight clustering of Treg populations from a range of fetal tissues indicates that Tregs have a conserved phenotype across organs. We observed a remarkable enrichment of Tregs in fetal tissues in comparison to adult organs (Figure 1F, Treg input gating strategy and reference plots in Figures S1B and S1C). Interestingly, fetal tissue Tregs were highly enriched for PD1+Ki67+ expression (Figure S1D), indicative of chronic TCR activation, active proliferation, and immune tolerance. We also observed a consistent presence of CD8+Tbet+GB+ (granzyme B) cytotoxic T lymphocytes (CTL) in fetal tissues that were significantly enriched in fetal skin, lung, and liver (reference plots and quantification in Figures S1E and S1F).

6Functional immune cells with potential antimicrobial activity and effector/memory phenotype have been demonstrated in fetal tissues. Thus, we probed our data for the expression of antimicrobial cytokines in CD4+ and CD8+ T cells (input gating strategy in Figure S1G). Notably, significant expression of tumor necrosis factor (TNF+) and interferon (IFN)γ+ T cells were observed across fetal tissues that were consistent among donors and tissue types (Figures 1G and S1H). We next explored the fetal T cell memory populations (CD4+/CD8+ and CD45RO+ T cells) across all 2nd trimester fetal tissues. Interestingly, we observed a significant proportion of CD4+ and CD8+ memory T cells in fetal tissues, especially in thymus (Figures 1H, S1I, and S1J). Fetal organs like gut, mLN, lung, skin, and liver also had significantly high levels of memory T cells. We also observed the presence of TNF+ and IFNγ+ effector Tems in fetal tissues (Figures 1I and S1J). Importantly, the effector memory phenotype was observed in all fetal organs except thymus despite having significant memory T cells (Figures 1H and 1I). Together, these data indicate the presence of antigenic stimuli that drive the generation of activated T cells within the fetal organs.

7Recent studies have explored the presence of microbes or microbial DNA in the placenta, amniotic fluid, meconium, fetal lung, and fetal intestine. However, the presence of live microbes in human fetus still remains debated. Our search for the potential stimuli at the source of the diverse immune response observed in fetal tissues led us to explore the presence of microbes in fetal tissues. Previously, we have shown that the human fetus has an active and responsive immune architecture during the 2nd trimester of gestation; hence this period was chosen for microbial investigation. We profiled multiple fetal organs, including gut, skin, lung, thymus, spleen, and placenta for the presence of bacterial signals using high-throughput 16S rRNA gene sequencing (sample information in Table S1 and primer sequences in Table S2). Given the high vulnerability for contaminations in such studies, and the new benchmarks introduced to account for reagent-associated contaminating microbes, we devised a stringent array of controls associated with each step in the workflow, ranging from tissue procurement to the sequencing run (Figures S2A and S2B). The controls taken in this study can be divided into four major categories: operator controls (hand-swabs taken from operators who have handled, dissected, and further processed the fetal samples); environment controls (swabs taken from all the surfaces that fetal tissues may have been exposed like work bench, PCR hood, etc.); PBS controls (fetus-matched buffers for each donor, and have undergone the same laboratory treatments as the fetal samples, from collection to sequencing); and reagent controls (including the molecular reagents used during DNA extraction and sequencing process such as elution buffer, McrBc reagent, etc.), which allowed us to monitor the contaminations present in extraction kits and accompanying reagents. All these controls allowed us to detect the contaminations introduced during organ collection and processing as well as by the reagents used in the process. We have also taken great care to reduce experimental bias and cross-contaminations by processing all samples in isolation, in a low-contaminant, controlled environment.

8Using the 16S rRNA gene sequencing data, we compared the overall microbial diversity between different fetal organs and their matched negative controls using weighted UniFrac distance matrices. The results of the β-diversity (diversity between fetal samples) were visualized by a principal coordinate analysis (PCoA) (Figure 2A). While negative control samples (PBS) mostly clustered on one end of the PCoA plot, fetal skin and placenta samples clustered on the opposite end, with few residing in-between. Fetal gut and lung samples were mostly dispersed between PBS and skin clusters, suggesting varying degrees of microbial diversity across tissues. PCoA plot annotated by early (<16 weeks) and late (≥16 weeks) fetal EGA across samples projected an unbiased generalized distribution across EGA, with no significant difference in early versus late gestational samples, when compared with matched PBS for bacterial diversity (Figures S2C and S2D). Most importantly, BactQuant qPCR analysis across samples showed a significant increase in bacterial DNA in fetal tissues in comparison to the controls (PBS, reagents, and environmental controls) (Figure 2B). CT (cycle threshold) value represents the number of cycles required to amplify a threshold DNA signal, thereby lower CT value corresponds to higher content of starting DNA being amplified. Notably, this trend was consistent across tissues, with a significant increase in the microbial DNA signal in fetal organs when compared to PBS (Figure 2C) as an external control and fetal thymus as an internal control (Figure S2E). Consistent with this, a significant subset of fetal tissues displayed a within-sample microbial diversity (α-diversity) that was higher than the negative controls (Figures 2D, S2F, and S2G). Moreover, fetal tissue samples were significantly more dissimilar to their matched PBS controls (β-diversity) than the PBS samples were to each other (Figures 2E and S2H). The later result, in particular, was also reproduced in an independent set of fetal samples (Figures 2E and S2H), analyzed at a separate facility (Weizmann Institute of Science [WIS]) using dedicated experimental and analytical protocols tailored for low biomass microbiome studies. Interestingly, we also noted a gradual increase in bacterial diversity in fetal tissues from external or surface-exposed organs in comparison to internal tissues (Figures 2F, S2I and S2J; 16S rRNA gene sequencing sample description and extended operational taxonomic units [OTU] information, in Table S3, sequencing reads for samples at separate facility [WIS] are in Table S4). These observations also corroborate with our findings of TNF+ T cells (Figures 1G and 1I) and cytotoxic CD8+ T cells (Figures S1F, S1H, and S1J) being more abundant within fetal skin than other organs. Together, these findings indicate the presence of an above-background microbial signal in at least some of the analyzed fetal tissues, albeit in very low abundance.

9Despite having an extensive array of controls, we are aware that contaminations may still occur; hence we attempted to compare the bacterial taxa present in control samples to those in fetal tissues. We analyzed our 16S rRNA gene sequencing data based on its operational taxonomic classification using the SILVA (v.123) database. We identified the specific OTU associated with bacterial genera present in each sample (Figure 2G). Showing the average abundance of dominant bacterial taxa identified in PBS versus those in fetal tissues, our analysis highlighted the enrichment of 7 bacterial genera specifically enriched in fetal samples (denoted as “fetal-enriched taxa” in green), including Flavobacterium, Lactobacillus, Staphylococcus, Afipia, Pseudomonas, Bradyrhizobium, and Brevundimonas. We also identified 7 bacterial genera enriched in PBS and other controls (operator, reagent, and environmental controls; denoted as “PBS-enriched taxa” in orange) including Ralstonia, Brevibacterium, Burkholderia, Azospira, Nitrosomonadaceae, Sphingomonadaceae, and Micrococcus (Figure 2G; Table S3 showing the OTU information for each sample). Interestingly, each one of the fetal-enriched genera showed a heterogeneous distribution among tissues, suggesting a true signal coming from specific organs. On the contrary, the PBS-enriched genera showed a more homogeneous distribution across samples in both controls and tissues indicative of a systemic contamination. In addition, our comparative analysis of bacterial enrichment in fetal gut samples in comparison to fetal spleen (or thymus) versus PBS control showed similar pattern with ∼22 significantly enriched taxa in fetal gut (Figures 2H and S2K). These observations suggest significant enrichment of bacterial signals in fetal organs, more notably in fetal gut.

10To determine if these DNA signatures were predictive of the presence of live bacteria, we inoculated tissue biopsies of freshly harvested fetal organs (from a separate cohort of 8 fetuses D1–D8 shown in Tables S1 and S3) in pre-reduced anaerobic blood-supplemented basal media under anaerobic conditions and plated the cultures under either aerobic or anaerobic conditions (Figure 3A; see STAR Methods for details). Several negative controls were included in each of these experiments, including PBS samples used at the hospital to clean surgical equipment during fetal dissection and PBS samples used in the laboratory to clean tools used during biopsy. None of these associated controls produced detectable colonies for any of the 8 distinct fetuses (Figures 3B, S3A, and S3B). In contrast, most of the cultured fetal tissues (6 out of 8 fetal donors) produced microbial colonies under both aerobic and anaerobic conditions (Figure S3B). Independent cultures and plating experiments produced reproducible numbers of colonies (Figure S3C). Moreover, we observed that plates carrying lung isolates appeared more oxidized, potentially reflecting a higher hemolytic ability of lung-associated bacteria. We also observed that the bacterial colony morphologies and sizes varied among different tissues suggesting a bacterial heterogeneity across organs. The absence of microbial growth from the negative controls and the apparent phenotypic differences in bacterial colonies across tissues are against the possibility of a common contamination by a single or few exogenous bacterial species. We next isolated and purified several colonies per inoculate and identified the bacterial genus via colony PCR and standard 16S sequencing (Figure 3C). Consistent with our high-throughput 16S sequencing results, a higher number of unique bacterial genera were identified in placenta or skin than in lung or gut samples and were further reduced in the spleen and thymus, consistent with a decreasing gradient of microbial diversity from external or surface-exposed fetal organs to internal organs (Figures 3C, S3B, and S3C). A large fraction of the isolated species is reported as facultative anaerobes or obligate anaerobes (Figures 3C, S3B, and S3C; Table S5). This is an interesting observation, because the oxygen supply to the fetus is reported to be dependent on the partial oxygen pressure gradient between maternal blood and fetal organs, which is low in early gestation and increases gradually along the gestational timeline. We next surveyed the prevalence of bacterial genera across donors and fetal organs (Figure 3D). Although Gardnerella, Lactobacillus, and Staphylococcus were isolated from most of the fetal organs, other commonly isolated genera included Streptococcus, Enterococcus, and Prevotella. Importantly, we observed several bacterial genera, which were both strict anaerobes and are not previously reported as vaginal or skin commensals, such as Bifidobacterium spp., Collinsella aerofaciens, or Clostridium hathewayi (re-classified as Hungatella Hathewayi) (; Table S6 shows number of tissues positive for the given bacteria and number of colonies isolated for the given bacteria across 8 fetal donors). Interestingly, many of these cultivatable bacteria were also found in our 16S rRNA gene sequencing data (Figure 2G), where Lactobacillus and Staphylococcus were among the top fetal-enriched taxa whereas Micrococcus and Burkholderia were among the top PBS-enriched taxa.

11To address the possibility of contamination during culturing, we also sequenced a section of the tissue biopsies before culturing from 7 of the 8 donors, using high-throughput 16S rRNA gene sequencing. While matching this pre-culture sequencing data with the culture-based sequencing results from the same biopsies, we successfully identified 24 out of 27 distinct bacterial genera (by their respective OTU) (Figure S3D) that were enriched in both datasets (Figure 3D, orange stars represent matching 24 genera present in both pre- and post-culture sequencing data). Notably, Lactobacillus, Staphylococcus, Streptococcus, Enterococcus, Bifidobacterium, Prevotella, and Finegoldia were among the ones identified by both pre- and post-culture sequencing, hence denoting a true signal from fetal tissues, and are not introduced due to culturing (Table S3 contains the extended sample and OTU information for the given 7 fetal donors used in Figure S3D). We also observed that a majority of these bacteria were either absent or very low in PBS samples when compared to the fetal organs. In contrast, Micrococcus, a potential contaminant that was consistently present in all samples including control PBS in our data, was recently reported to be present in fetal intestine in utero. Similarly, Streptococcus that we isolated in our culture-based method has also been previously associated with human placental tissues, although it is speculated that it might be as a result of a potential infection. Therefore, the notion of true fetal-specific bacteria still remains open and demands further investigation. Taken together, our results reveal a sparse microbial presence in certain fetal tissues, some of which can be propagated by anaerobic culturing. However, their abundance in specific fetal tissues and their prevalence across gestational timeline still remains debatable and an area of active exploration. Neverthelss, these results indicate a transient presence of microbial entities within fetal tissues, which may remain in a constant flux, hence reflective of a snapshot in time.

12To investigate the direct presence of bacteria in fetal tissues, we performed SEM on freshly isolated fetal intestine (n = 4 independent fetal gut samples), specifically the region around the ileum, shown to have the most diverse and rich microbiota within the intestinal tract. The environmental contamination was minimized by performing sample isolation and processing under sterile conditions as discussed above (Figures S2A and S2B), and the dissected regions were fixed immediately with 2.5% glutaraldehyde. Interestingly, we observed bacteria-like morphological structures, resembling bacterial cocci, clustered together within limited regions across fetal gut at the 14th weeks of gestation (n = 3) (Figure 4A). Such bacterial cocci were observed with an average size of 1–2 μm and were specifically nested within mucin-like structures present along the intestinal lumen, which were speckled across the ileum. Three of the four samples were from 14-weeks EGA fetuses and showed a similar bacterial localization pattern, whereas the fourth sample, belonging to a 10-weeks EGA fetus, did not show any significant bacterial localization across the ileum (Figure 4A). Most regions did not show any bacterial localization, and notably, these areas were also devoid of mucin-like structures (Figure 4A). Multiple scans across different regions showed similar localization patterns and morphologies in all three 14-weeks EGA fetuses, with discrete bacterial structures closely packed within meconium-associated mucin-like threads (Figure S4A). For one of the 14-weeks samples, we also scanned the upper (fore-gut) and lower (hind-gut) regions and observed similar bacterial entities albeit at a lower frequency compared to the ileum (Figure S4B). These observations revealed discrete clusters of bacterial structures present in fetal gut, consistent with the bacterial cocci-specific morphological embedding within meconium-associated mucin-like structures. We also performed SEM on the culture isolates from fetal gut and lung, wherein we observed distinct cocci-like bacteria, further suggesting that these fetal gut-associated bacteria are alive and can be propagated further (Figures 4B and S4C).

13To further complement these morphological observations, we performed RNA-ISH using a eubacteria-specific 16S rRNA probe. Interestingly, we observed distinct localization of bacterial RNA, tightly packed alongside intestinal epithelium (Epcam+) within the fetal gut lumen (Figure 4C, H&E staining image of fetal gut cross-section in Figure S4D). Notably, these bacteria were not observed on the stromal side of the intestine and were specifically present within the luminal cavity. The bacterial rRNA signal was absent in corresponding fetal liver and mesenteric lymph node (mLN) samples (Figures S4E and S4F). Quantification of the bacterial signal (with respect to DAPI staining) showed significant enrichment of eubacterial 16S rRNA in fetal gut as compared to fetal liver and mLN (n = 3) (Figures 4D and 4E). Taken together, these observations demonstrate direct presence of bacteria within the fetal gut lumen during the 2nd trimester of gestation.

14Our data demonstrate the presence of specific bacterial signals across fetal organs, specifically the fetal gut, although it is not clear what the physiological significance of these microbes could be. The effect of microbe-derived signals on the priming and maturation of the postnatal immune system in general is quite well established. Hence, we hypothesized that fetal microbial antigens could be one of the active forces behind the diverse T cell responses observed in fetal tissues. We thus proceeded to explore their potential cross-talk with the fetal immune system. Commensal microbes are known to elicit tissue-specific T cell memory responses, which get activated upon any potential breach in mucosal barrier. In the case of the gut, it is proposed that the microbe-specific memory T cells reside in the gut-associated mLN. In adult mice, microbial antigen-laden DCs are known to migrate to mLN thereby inducing microbe-specific T cell expansion. We have previously shown that human fetal DCs migrate to the fetal mLN as early as the 16th weeks of gestation and thus hypothesized that these migratory DCs may expose the T cells in fetal mLN to microbial antigens and prime them against microbial entities within fetal tissues. To test if fetal T cells could be recalled by DC presenting fetal bacteria-derived antigens, we recapitulated a minimalistic fetal environment in vitro with an autologous T cell expansion and activation assay (Figure 5A). We first sorted HLA-DR+ DCs and T cells from the same fetal mLN (Figure 5B). T cells were purified by negative selection to minimize activation, followed by labeling with a fluorescent dye CellTrace violet (CTV), prior to bacterial exposure. Sorted DCs were incubated with heat-killed fetal Lactobacilli or fetal Staphylococci for 8 h prior to coculture with T cells. These two bacteria were chosen because they were consistently present in fetal tissues in both high-throughput 16S rRNA gene sequencing and culture-based methods. Bacteria primed DCs were then presented to fetal T cells and their expansion was monitored post 6 days of incubation. For negative control, un-primed DCs (DCs alone, no bacteria) were incubated with T cells and CD3/28 Dynabeads were used for positive control. Fetal Staphylococci antigen-exposed T cells showed an average 70% (±15%) memory expansion, whereas the lactobacilli-exposed T cells showed ∼55% (±13%) memory expansion, in comparison to unprimed DC control (Figure 5C, input gating strategy and representative flow cytometry plots in Figures S5A–S5E). Upon exposure with fetal bacteria primed DCs, we observed a significant increase in total T cell count, demonstrating the expansion of T cells post bacterial priming (Figure 5D). Notably, the magnitude of T cell expansion with Staphylococci-primed DCs was consistently higher than with Lactobacilli, suggesting a bacteria specific heterogeneity in fetal T cell responses. Positive control (CD3/28 Dynabeads) showed more than ∼95% expansion throughout. Assessment of TNF and IFNγ production post 6 day bacterial exposure in culture supernatants revealed that both TNF and IFNγ were abundantly produced and released upon fetal Staphylococci exposure (Figures S5F and S5G). Interestingly, we observed a significant increase in CD45RO expression post expansion, indicating a potential memory phenotype against microbial antigens, whereas only <12% of memory T cells (CD45RO+) were present at day 0 prior to exposure (Figures 5E and 5F). Interestingly, there was a significant increase in the activation and tissue residency marker CD69 (∼50% versus 1% on day 0) post-expansion, suggesting the early stages of memory activation upon bacterial exposure (Figures 5G and 5H).

15Next, we aimed to investigate if the effect on T cell activation observed post bacterial exposure is specific to antigen presentation. We fixed fetal DCs post bacterial priming with 1% paraformaldehyde (PFA), prior to T cell presentation, to ensure that the responses obtained reflected only the processing events that occurred due to antigen exposure (Figure 6A). We selected fetal isolated Staphylococci for this priming because it was specifically enriched in the fetal gut (Figures 2G and 2H) and showed significant T cell activation in the previous assay (Figure 5). Interestingly, we observed similar fetal T cell expansion profile with both fixed and non-fixed DCs; however, the extent of proliferation was lower in fixed DCs as measured by absolute T cell count post 6 days of exposure (Figures 6B and 6C). The number of T cells surviving and proliferating with live DCs was higher than fixed DCs, although both were significantly higher than the un-primed DC control (Figure 6C). This observation suggests the effect of bacterial exposure on fetal T cell expansion is contributed by both antigen presentation as well as cytokine-mediated signals. However, antigen presentation alone is sufficient to drive the proliferation of a significant number of T cells in response to bacterial signals. These expanded T cells were CD45RO+ and a significant proportion was CD69+, suggestive of an activated memory response toward fetal bacteria (Figures 6D and 6E, representative flow cytometry plots in Figures S6A–S6C). To further validate these observations with an independent method, we also used protein transport inhibitors (PTI) (combination of monensin and brefeldin A) to block cytokine production post bacterial priming of DCs (Figure S6D). This approach also showed similar T cell expansion and activation response as in the case of PFA-induced fixation (Figures S6E–S6H).

16Taken together, these observations suggest that fetal mLN T cells exhibit an activated memory response toward microbial antigens encountered in utero. Further studies are required to explore the antigenic specificities of these T cell-microbial interactions in human fetal organs and the synergy with a non-antigen specific (cytokine-mediated) response toward the extent of proliferation.

17The events occurring during fetal gestation are essential for the overall development and growth of the individual. Key factors transferred from the mother to the fetus during this process establish the cornerstones of early life immunity and life-long tolerance. Various studies have recently suggested that certain antigens as well as bacterial entities may cross the placental barrier and make their way to fetal organs. A recent report from our group also demonstrated that human fetal mast cells are primed and sensitized during gestation through maternal immunoglobulin E (IgE) that cross the placental barrier. Interestingly, a recent report on plant development showed partial inheritance of plant microbiome by vertical transmission through seedling. All these observations open a new avenue in developmental immune-priming wherein maternal-derived microbial entities may condition the progeny toward future antigenic exposures in life.

18Our findings demonstrate that healthy human fetal tissues (in the 2nd trimester of gestation) contain effector Tems, a sparse biomass of bacteria, and an active memory T cell response toward fetal bacteria. We also demonstrate direct spatial localization of bacterial entities, localized within the lumen of developing fetal gut, during the 2nd trimester of gestation. These data are in line with recent studies that indicated antigenic priming of the fetal immune system begins during gestation. However, our study demonstrates for the first time that such microbial presence primes the fetal immune system, thereby putting early microbial memory in the context of fetal immune priming, a concept not explored before in fetal immunity. It will be interesting to explore the precise nature of these microbial antigen-specific circulatory and tissue-resident immune cells in human fetal organs, and their potential role in imparting selective defense against pathogenic microbes in neonatal and adult life. Taken together, these findings have wider implications in understanding the key factors involved in fetal immune system development and priming in utero, which may set the basis for life-long health and immunity of the organism.

19Although we have taken particular care to minimize the risk of contaminations in this study, we cannot completely exclude potential confounders due to environmental factors. In particular, fetal skin and placenta samples will have been exposed to vaginal microbiota during delivery or by the operator’s skin. However, the existence of a spatially diverse microbial signal in fetal tissues, their ability to culture-expand anaerobically, and the presence of microbial antigen-specific memory T cell activation, are difficult to reconcile with either systematic biases or random noise. Collectively, our data suggest a low but consistent presence of microbes in at least some of the healthy human fetuses in the 2nd trimester of gestation. Recent reports of the presence of microbes in fetal lung and gut provide independent assertions to our findings. However, this is an area of constant exploration and evidence from various spheres of technological advances is needed to even partially understand the dynamics of fetal development and immune priming.

20This study provides multiple evidence for the presence of live microbes in human fetal tissues, however, there are several outstanding limitations to be addressed in future studies. First, the precise source of bacteria in fetal tissues still remains unknown. More clinical studies with access to pertinent maternal samples are required to determine the source of origin and the trajectory of bacterial entry to fetal organs. Second, although we have undertaken a wide array of controls in this study, contamination from maternal organs, especially from the lower genital tract at delivery, cannot be formally ruled out. Sequencing the maternal uterine microbiome, possibly at the time of delivery, and the fetal microbes may address this issue. We acknowledge that the species/genera specificities for fetal organs is limited in this study, and more detailed metagenomics approaches will be required to precisely determine the key bacterial species present in fetal tissues and their niche preferences.

21Last, the antigenic specificities of fetal T cell-microbial interactions and its synergy with a non-specific antigen response (cytokine-mediated) needs further exploration as we cannot formally rule out the possible contribution of a super-antigen-based activation or a pan-bacterial response against fetal bacteria. Further studies are required to address these possibilities in the future.

22Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact:

23Florent Ginhoux (florent_ginhoux@immunol.a-star.edu.sg)

24This study did not generate new unique reagents.

25All the raw sequencing data reported in this paper is available on GEO and the accession number is GSE169306 (Human fetal immunity and microbial priming during early development). The authors declare that all data supporting the findings of this study are available within the manuscript and/or its Supplementary Files or available on request.

26Human fetal and adult tissues, cord blood, adult blood and amniotic fluid, were obtained in accordance from Singapore SingHealth and National Health Care Group Research Ethics Committees. All women gave written consent to the use of fetal tissues according to internationally recognized guidelines. Review of the guidance on the research use of fetuses and fetal material. CM 762). All fetal tissues (placenta, lung, thymus, spleen, gut, mesenteric lymph nodes, and skin) were obtained from 2nd trimester (12-22 weeks EGA) elective pregnancy terminations carried out for socio-psychological reasons. All fetuses were considered structurally normal on ultrasound examination prior to termination and by gross morphological examination following termination. Fetal tissues from 2nd trimester of gestation were used for this study (Table S1). The study of fetal blood was approved by the Hôpital Erasme ethics committee (ULB). All participants (or mothers, in the case of fetal samples) gave written informed consent. We obtained fetal blood (between 23 and 37 weeks EGA) as described previously.

27Mid-trimester terminations were medically induced by the use of prostaglandins, and the fetus was delivered through the birth canal. Fetal organs were collected under sterile conditions in a tissue culture hood. Aseptic equipment was used for collecting fetal tissues. Fetal skin was collected first. Due to the risk of contamination from vaginal microbes, great care was taken to minimize the risk of contamination of internal organs. Fetal skin across the abdomen was cut and pinned back to prevent it from touching internal organs. Fresh aseptic instruments were then used to remove internal organs. Fetal lungs were then removed, followed by the thymus, spleen, liver and gastrointestinal tract. The mesenteric lymph nodes were removed with the gastrointestinal tract. If a placenta was obtained, it was processed last. Tissues were processed as described below for suspension mass cytometry assays and microbiome studies. When tissues were used for both arms of the study, samples were divided using aseptic equipment. Sterile equipment was rinsed in PBS and this PBS was used as an environmental control, along with unused PBS and environment air swabs.

28Fetal organ dissection in detail:

29Prior to start of the harvest procedure, the entire bio-safety cabinet was thoroughly disinfected.

30Order of Harvest- Skin, Liver, Gut, Spleen, Thymus, Lungs and Placenta

31Separate (new and sterilized) dissection tools were used for dissection of each organ to avoid instrument associated contaminations as well as cross-tissue contaminations.Skin: Starting the dissection of skin, using a scissor, an incision was made into the skin (back, chest and abdomen) and carefully separated the skin from the muscle, followed by isolation and transfer to buffer/media.Liver: Abdominal cavity was entered, and the liver carefully detached from the diaphragm and the anterior abdominal wall, and placed in sterile buffer/media.Intestines: The omentum and mesenteric was dissected, and intestines snipped near the rectum and at the gastric-duodenal junction, rinsed with PBS transferred to sterile buffer/media.Spleen: The spleen was dissected from its splenic portal attachments, placed in sterile buffer/media.Thymus: A cut is made along the sternum to the neck to access the thoracic cavity. The thymus can be identified as fatty tissue attached to the upper mediastinum, where it is isolated with a pair of sterile forceps, and placed in sterile buffer/media.Lungs: Using the pointed/fine-tip forceps, a pinch on the bronchus was exerted and a scalpel was used to remove the lung from the thoracic cavity and transferred to a sterile buffer/media.Placenta: one cm3 of placental tissues was obtained near the cord insertion and transferred to sterile buffer/media. This biopsy includes both fetal and decidual surfaces of the placenta.

32Skin: Starting the dissection of skin, using a scissor, an incision was made into the skin (back, chest and abdomen) and carefully separated the skin from the muscle, followed by isolation and transfer to buffer/media.

33Liver: Abdominal cavity was entered, and the liver carefully detached from the diaphragm and the anterior abdominal wall, and placed in sterile buffer/media.

34Intestines: The omentum and mesenteric was dissected, and intestines snipped near the rectum and at the gastric-duodenal junction, rinsed with PBS transferred to sterile buffer/media.

35Spleen: The spleen was dissected from its splenic portal attachments, placed in sterile buffer/media.

36Thymus: A cut is made along the sternum to the neck to access the thoracic cavity. The thymus can be identified as fatty tissue attached to the upper mediastinum, where it is isolated with a pair of sterile forceps, and placed in sterile buffer/media.

37Lungs: Using the pointed/fine-tip forceps, a pinch on the bronchus was exerted and a scalpel was used to remove the lung from the thoracic cavity and transferred to a sterile buffer/media.

38Placenta: one cm3 of placental tissues was obtained near the cord insertion and transferred to sterile buffer/media. This biopsy includes both fetal and decidual surfaces of the placenta.

39Fetal organs were mechanically dispersed and incubated with 0.2 mg/ml collagenase (TypeIV; Sigma Aldrich) and DNase I (20,000U/ml, Roche) in RPMI with 10% FCS for 1 hour in a 6 well plate. Viability was typically 80%–90% measured by DAPI exclusion (Partec). Fetal gut was initially cut longitudinally through the center, washed extensively in PBS until all inner content (meconium) was removed and the PBS was clear, before mechanical dispersion and digestion as above, for 1h. Adult lung specimens were obtained from peritumoral tissue. Adult lung specimens were prepared as described previously. T cells were enriched isolated to 90% purity from adult and fetal organs by negative selection using T cell enrichment kits (Miltenyi Biotec) and separated on an AutoMacs following manufacturer’s instructions.

40T cells were stained as previously described and then analyzed by CyTOF. Briefly, T cells were stimulated with phorbol 12-myristate 13-acetate (PMA) (150 ng/ml, Sigma-Aldrich) and ionomycin (750 ng/ml, Sigma-Aldrich) for 5h, blocked with secretory inhibitors, brefeldin A (1:1000, eBioscience) and monesin (1:1000, Biolegend) after one hour, for the last 4 hours, in 10% v/v human serum, 1% v/v PSG, RPMI at 37°C, 5% CO2. The cells were then washed and stained with cell viability dye cisplatin (200 μM, Sigma-Aldrich) for 5 min at room temperature. The cells were re-washed and each individual sample was barcoded with a unique combination of anti-CD45 conjugated with either heavy metal 89, 115, 141 or 175, as previously described for 25 min on ice. The barcoded cells were washed and stained with a surface antibody cocktail in 4% v/v heat-inactivated FBS, 2mM EDTA, 0.05% w/v sodium azide in pH 7.4 PBS for 30 min on ice. The cells were again washed and re-suspended in fixation/permeabilization buffer (1:3, eBioscience) for 45 min on ice. The permeabilized cells were then stained with an intra-antibody cocktail (1:10, permeabilisation buffer, eBioscience) for 45 min on ice, before washing and staining with a DNA intercalator Ir-191/193 (1:2000 in 1.6% w/v paraformaldehye, Fluidigm) overnight at 4°C or for 20 min on ice. Finally, the cells were washed and re-suspended with EQTM Four Element Calibration beads (1:10, Fluidigm) in ultra-pure distilled water at 1x106 cells/ml. The cell mixture was then loaded and acquired on a Helios mass cytometer (Fluidigm) calibrated with CyTOF Tunning solution (Fluidigm). The output FCS files were randomized and normalized with the EQTM Four Element Calibration beads against the entire run, as per the manufacturer’s recommendations. Key Resources Table contains all the antibodies with specifications used in CyTOF experiments.

41The automated analysis was performed by the UMAP algorithm as described previously. FlowJo software and R package were used for data analysis.

42DNA from fetal tissues was extracted according to the protocol as previously described, except that the volume of Breaking Buffer and phenol/chloroform/IAA used was 400 μL. Fetal microbiome reagent and environmental controls were concurrently performed during the DNA extraction process (Table S2 contains all the primers used, if not already mentioned in the methods).

43The qPCR was performed according to the protocol as previously described, with slight modifications. Briefly, the PCR reactions were run in 10uL reactions and were composed of the universal 16S BactQuant primers (5′-CCTACGGGDGGCWGCA-3′ and 5′-GGACTACHVGGGTMTCTAATC-3′) at 1.8uM, probe (6FAM-5′-CAGCAGCCGCGGTA-3′-MGBNFQ) at 225 nM, TaqMan Universal PCR MasterMix x2, water and undiluted sample DNA (3uL). The amplification was performed on a 7900HT Real Time PCR System (Applied Biosystems) with cycling conditions of 50°C for 3 min, 95°C for 2 min, and 40 cycles of 95°C for 15 s and 60°C for 1min. Cycle threshold value (Ct value) for each 16 S qPCR reaction were obtained using a manual Ct threshold of 0.02 and manual baseline between cycles 1-15 in the Sequence Detection Systems v2.4 software (Applied Biosystems). The qPCR for each sample was run as technical duplicates or more depending on sample availability, and the values were averaged to give the final Ct value used in data analysis.

44Genomic DNA isolated from fetal tissues and control samples were treated with McrBC restriction enzyme (New England BioLabs) according to manufacturer’s instructions, to remove host DNA and to enrich for microbial DNA. The treated DNA samples were purified using 0.8X ratio of AMPure XP beads (Beckman Coulter). The McrBC-treated, purified gDNA was quantified using Quant-iT TM Picogreen® dsDNA Assay kit (Invitrogen). For amplification of the 16S rRNA gene-variable regions three rounds of amplification were performed using 100ng of McrBC-treated gDNA with LongAmp Taq DNA polymerase (New England Biolabs) according to manufacturer’s instructions. The first step aimed at enriching microbial sequences was performed using primers that amplify V2-V6 regions (Table S2). PCR cycling parameters consisted of initial denaturation step for 30 s at 94°C, followed by 12 cycles of 15 s at 94°C, 30 s at 45°C, and 1min at 65°C with a final extension for 10 min at 65°C. Subsequently a nested PCR was performed targeting V4-V5 regions using the enriched V2-V6 amplicon as template. PCR Cycling parameters for amplifying V4-V5 regions were similar to V2-V6 PCR cycling profile, except that the extension was for 30 s at 65°C. The final amplification step involved addition of barcodes for identifying, de-multiplexing individual samples and Illumina adaptor sequences to the V4-V5 enriched product using communal primers. Communal PCR amplification consisted of initial denaturation step for 30 s at 94°C, followed by 30 cycles of 15 s at 94°C, 30 s at 45°C, and 1 min at 65°C with a final extension for 10 min at 65°C. Equimolar concentrations of the prepared libraries were pooled and electrophoresed using 1% agarose gel. Pooled amplicon libraries were gel-purified using the Qiaquick Gel Extraction Kit (QIAGEN). Integrity of gel-purified libraries were validated using the High Sensitivity DNA kit (Agilent Technologies). The libraries were quantified using KAPA Library Quantification Kit (Kapa Biosystems) to ascertain the loading concentration and sequenced on a HiSeq 2500 System (Illumina) operated in Rapid Run Mode to generate 2 × 250 bp paired-end reads.

45Samples from Fetus ID W1 to W10 (10 fetuses- Table S1) were processed at WIS. DNA from fetal tissues was isolated according to the protocol as previously described, designed to extract bacterial DNA. Frozen tissues were extracted with the UltraClean Tissue & Cells DNA Isolation kit (MoBio - #12334) according to the manufacturer’s instructions. A 30 minute proteinase K (15 μl) digestion was done, and bead-beating was performed before (10 minutes) and after (5 minutes) digestion. A heating step at 90 °C for 45 min was added after proteinase K digestion and then cooled at room temperature for 10 minutes. Lysate underwent bead beating with 200 μL 0.1mm zirconia/silica beads (Biospec - #11079101z) for 10 minutes at full speed. Ethanol was added (100%, 200 μl) to the tubes followed by a short vortex mix and the whole mixture (together with beads) was loaded onto the columns in two steps. Wash and elution steps were done according to manufacturer’s instructions. Pre-heated (70°C) elution buffer (100ul) was added to the columns. Columns were incubated at 70°C for 4min before DNA was eluted. All negative controls were processed according to the exact same protocols.

46Five regions of the 16S rRNA gene were amplified using 100ng DNA as an input and a set of 10 multiplexed primers (0.2 μM each primer), 0.2mM dNTPs (Larova GmbH) and 0.02U/ul of Phusion Hot Start II DNA Polymerase (Thermo Scientific #F549). Amplification was done with an initial heating step of 98 °C for 2 min, 30 cycles of 10 s at 98 °C, 15 s at 62 °C, 35 s at 72 °C followed by a final elongation step of 5 min at 72 °C. Primers used are F1-TGGCGAACGGGTGAGTAA, F2-ACTCCTACGGGAGGCAGC, F3-GTGTAGCGGTGRAATGCG,F4-GGAGCATGTGGWTTAATTCGA, F5-GGAGGAAGGTGGGGATGAC,R1-AGACGTGTGCTCTTCCGATCTCCGTGTCTCAGTCCCARTG,R2- AGACGTGTGCTCTTCCGATCTGTATTACCGCGGCTGCTG,R3- AGACGTGTGCTCTTCCGATCTCCCGTCAATTCMTTTGAGTT,R4- AGACGTGTGCTCTTCCGATCTCGTTGCGGGACTTAACCC,R5- AGACGTGTGCTCTTCCGATCTAAGGCCCGGGAACGTATT,

47F4-GGAGCATGTGGWTTAATTCGA, F5-GGAGGAAGGTGGGGATGAC,

48R1-AGACGTGTGCTCTTCCGATCTCCGTGTCTCAGTCCCARTG,

49R2- AGACGTGTGCTCTTCCGATCTGTATTACCGCGGCTGCTG,

50R3- AGACGTGTGCTCTTCCGATCTCCCGTCAATTCMTTTGAGTT,

51R4- AGACGTGTGCTCTTCCGATCTCGTTGCGGGACTTAACCC,

52R5- AGACGTGTGCTCTTCCGATCTAAGGCCCGGGAACGTATT,

53Barcodes and Illumina adaptors were then added to the amplicon with a second PCR reaction with 5 forward primers (0.2 μM each primer):FF1-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTTGGCGAACGGGTGAGTAA,FF2-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTACTCCTACGGGAGGCAGC,FF3-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGTGTAGCGGTG RAATGCG,FF4- AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGGAGCATGTG GWTTAATTCGA,FF5- AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGGAGGAAGGT GGGGATGAC) and one 8 nucleotide barcode-specific reverse primer (0.4 μM, RR5- CAAGCAGAAGACGGCATACGAGAT-NNNNNNNN-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT).

54FF1-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTTGGCGAACGGGTGAGTAA,

55FF2-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTACTCCTACGGGAGGCAGC,

56FF3-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGTGTAGCGGTG RAATGCG,

57FF4- AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGGAGCATGTG GWTTAATTCGA,

58FF5- AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTGGAGGAAGGT GGGGATGAC) and one 8 nucleotide barcode-specific reverse primer (0.4 μM, RR5- CAAGCAGAAGACGGCATACGAGAT-NNNNNNNN-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT).

59The amplicon was diluted into the reaction (10-fold) and amplified with 6 cycles of 10 s at 98 °C, 15 s at 64 °C, 25 s at 72 °C’. Amplicons were then combined into sub-libraries (40-50 amplicons) and each library was purified using Qiaquick PCR purification kit (QIAGEN #28104) according to the manufacturer’s instructions. Multiple sub-libraries were then combined into the final library (100-430 amplicons) and further purified from primer dimers using Agencourt AMPure XP (Beckman Coulter #A63881) at a volume ratio of 1:0.85 (library: beads). The library (7pM) was supplemented with 15% PhiX (8pM) and sequenced on NextSeq 500 mid output (paired end 2x150) sequencers.

60Sequencing data was pre-processed and mapped to the reference database as described. Operational taxonomic units (OTUs) occurring only once (singletons) were removed from the entire dataset. Samples with fewer than 500 reads were removed. Alpha- (Shannon and Chao1 indices) and beta-diversity analyses (Bray-Curtis and weighted UniFrac distances) were performed in QIIME (v1.8.0) on OTU tables that had been rarefied to 1,000 reads per sample for 10 times (Table S3). Principal Coordinate Analysis (PCoA) plots were visualized with Emperor. The three-dimensional coordinates of the PCoA clusters from emperor results were then extracted to be plotted by python matplotlib module (v3.3.3). Only fetal and PBS derived samples were selected for the graph. Permutational Multivariate Analysis of Variance (PERMANOVA) test was done using the adonis function of vegan (v2.5.7) package in R. Table S3 contains all the samples and OTU information.

61Reads were demultiplexed per sample and filtered if either a) Phred score was less than 30 in more than 25% of nucleotides or b) more than three nucleotides had a Phred score of less than 10, or c) a read contained one or more ambiguous base calls. Unique reads with low counts were discarded, and reads were matched to each of the five amplified regions based on the primers’ sequences. The SMURF package was applied to combine read counts from the five regions into a coherent profiling result solving a maximum likelihood problem. The Greengenes database (May 2013 version) was used as reference. To allow adding WIS results to SIgN analysis pipeline, OTU picking requiring a minimum similarity of 99% was performed over SMURF’s results. Table S4 contains all the OTU information for samples processed and sequenced at WIS.

62To explore differentially enriched bacterial taxa (OTUs at the genus level) in fetal gut as compared to PBS and Spleen or Thymus (Figures 2H and S2K), normalized read counts information of the OTUs were retrieved from Table S3. Log2 Fold Change (LOG2FC) of the OTUs were calculated in R (v. 4.0.3) statistical programs. The quadrant plots were made using the plot function. Each data points on the plot were sized according to the relative percentage abundance of the OTUs in Gut as compared to PBS and spleen or PBS and thymus. LOG2FC ≥ 2 was considered as the cut-off for OTUs which were differentially enriched in Gut whereas a Relative Percentage Abundance (RPA) of 3% is used as filtering criteria.

63LOG2FC was calculate using the following formula.Where Ai is the normalized read count for of the ith OTU in Gut and Bi is the read count for the corresponding ith OTU in PBS/thymus.

64RPA for each OTU in gut is given by = Counts of samples where reads exist for Gut = Counts of samples where reads exist for Thymus or Spleen = Counts of samples where reads exist for PBS

65Fresh fetal tissue biopsies were taken aseptically and inoculated into pre-reduced anaerobic basal broth supplemented with 5% defibrillated horse blood, incubated anaerobically at 37°C for 48h, and then plated on anaerobic basal agar supplemented with 5% horse blood under aerobic or anaerobic conditions. Single bacterial isolates were streak-purified and then identified via standard 16S PCR and Sanger sequencing as follows. Briefly, single colonies were resuspended in 50 μL of Tris-EDTA; 1 μL of the suspension was then subjected to 1 min denaturation at 98°C, followed by 30 cycles of PCR (10 s at 98°C, 30 s at 56°C, 30 s at 72°C) and a final extension of 10 min at 72°C, using Q5® High-Fidelity DNA polymerase (NEB) and V4-V5 primers listed in Table S2. PCR products were then sequenced using V4-F as the primer, yielding sequences ≥ 330 nucleotides in length. Sequence ends were then trimmed using DNALaser Gene Seqman Pro 15 (Quality stringency: high), followed by manual inspection and further trimming to exclude regions with overlapping peaks. Chromatograms not yielding unique sequences were discarded and not processed further. Trimmed sequences were then searched against the NCBI 16S ribosomal RNA sequences (Bacteria and Archaea) database using Megablast (optimized for highly similar sequences) using default parameters. Species names of all top-scoring hits were reported (Table S5). Sequence identity was generally ≥ 97% for most of the species identified. Table S6 shows number of tissues positive for the given bacteria, and number of colonies isolated for the given bacteria, across 8 fetal donors.

66We aligned all 1194 sampled bacterial taxa (genus level, based on the OTU information in Table S3) as identified by high-throughput 16S rRNA sequencing, according to their abundance in PBS controls. We selected top bacterial genera that were specifically present in any of the fetal tissues and top ones that were specifically enriched in PBS controls. A curated pandas Python library version v0.25.3 was used to extract and align the OTU reads. The bacterial taxa which showed zero signal intensity across all sample types were removed. For each sample type and bacterial identity, the mean signal intensity was calculated and plotted using the matplotlib (v.3.2.1) and seaborn (v.0.9.0) Python libraries. Bacterial genera where signal intensity in PBS was higher or equal to those in fetal samples were depicted as potential contaminants (PBS enriched taxa). Table S3 shows the OTU information for each sample. For Figure S3D, we extracted the 16S rRNA gene sequencing data for 7 fetal donors (out of 8) which were simultaneously processed for live culture-based sequencing as well. We then extracted the signal intensities for those bacterial taxa which were identified by culture-based sequencing as shown in Figure 3D. Mean signal intensities were plotted according to their abundance in fetal tissues, along with those in PBS controls (Table S3 contains the sample and OTU information for the given 7 fetal donors used in Figure S3D).

67Fetal gut samples (14 weeks EGA) were processed for Scanning Electron Microscopy imaging under sterile conditions. Briefly, the region around mid-gut was selected for scanning and were cut open longitudinally to expose the luminal parts by using a binocular dissecting microscope and pinned down to a silicone mat in four corners. The samples (in vitro bacterial samples and fetal gut samples) were immediately fixed in 2.5% glutaraldehyde overnight at 4°C. Phosphate buffer solution (1X PBS) was used to wash the overnight fixed samples (2 changes, 20 minutes each) and the samples were kept in PBS. For bacterial cultures, bacteria isolated from 18 weeks EGA fetal lungs and gut (OD600 = 0.1) were co-incubated with nitrocellulose membrane in Luria-Bertani broth overnight at 37°C to allow bacteria to attach the membranes, prior to fixation. The samples were post-fixed in 1% osmium tetroxide and dehydrated by using ascending concentrations of ethanol. After which, the samples were critical-point-dried, stuck on carbon stubs and coated with 25 nm of gold. The samples were imaged on a Field Emission Electron Microscope (EM) at a voltage of 20 kV at Electron Microscopy Unit at the National University of Singapore.

68Formalin-fixed paraffin-embedded (FFPE) slides were prepared from FFPE fetal tissue blocks. The tissue samples with a thickness of 2 microns were mounted onto glass slides and stored at room temperature. Slides were stained using the RNAscope® Multiplex Fluorescent Reagent Kit v2 Assay Kit (Advanced Cell Diagnostics) following manufacturer’s protocol. The slides were baked in a dry oven for 1 hour at 60°C. Following steps included deparaffinizing the slides using fresh xylene (2 washes for 5 minutes each) and 100% ethanol (2 washes for 2 minutes each). The slides were then treated with RNAscope® Hydrogen Peroxide for 10 minutes at room temperature. Meanwhile, 200 mL of RNAscope® 1X Target Retrieval Reagent was poured in a slide holder and kept in a beaker filled with water on a hot plate to boil. The slides were washed and put in the slide holder for 15 minutes at 99°C for target retrieval. A hydrophobic barrier was drawn around the tissue using ImmEdge hydrophobic barrier pen. Few drops of RNAscope® Protease Plus was added onto the tissues and incubated for 30 minutes at 40°C. RNAscope probes from ACDbio was used in this experiment; EB-16 s-rRNA (Cat# 464461) in C1 channel and EPCAM (Cat# 310281) in C3 channel. The probe mix was added onto the slide and probe hybridization was performed for 2 hours at 40°C. The slides were then stored in 5X SSC overnight and amplification steps were performed the following day.

69Opal 520 and Opal 690 dyes (Perkin Elmer, Cat# NEL821001KT) were selected for EB-16 s-rRNA and EPCAM probes respectively. The Opal dye dilution optimized was 1:500 for all samples except 1:250 dilution for liver samples (EPCAM probe only). The slides were mounted using a drop of ProLong Diamond Antifade Mountant (Thermofisher, Cat#: P36970). The slides were imaged by Vectra® Polaris Automated Quantitative Pathology Imaging System. The images were edited using inForm® image analysis software and color intensity for EB-16 s-rRNA was set to 150% (green) while EPCAM was set to 120% (white). The whole slide scans of Hematoxylin and eosin stained slides were imaged using Phenochart image analysis software.

70For image quantification, Imagej and Cell Profiler softwares were used. Using Imagej software, the images were split into single channel greyscale images. These greyscale images were added into Cell Profiler software and analyzed using a published pipeline. The threshold correction factor in the pipeline was in the range of 1-4 for EB-16 s-rRNA and 1.7-2.5 for EPCAM.

71DC and T cells from fetal Mesenteric Lymph Node (mLN) were isolated by fluorescence activated cell sorting (FACS) using BD FACSAria IV (BD Biosciences). T cells were isolated by negative selection, gating for lineage negative (lineage- CD19/20/56) and HLA-DR negative cells, with upto ∼80 percent purity. HLA-DR+ and CD14- cells were selected for DC isolation. Fetal bacteria (Lactobacilli and Salmonella) were isolated and cultured as indicated above, in 5% defibrillated horse blood at 37°C for 48h. Culture stocks were made and snap frozen at −80°C. At the day of experiment, frozen vials were thawed and bacteria were washed with sterile PBS followed by heat inactivation at 65°C for 20 minutes, with intermittent vortex mixing. Flow sorted DC were plated in a 96 well plate at the density of 1000 cells per well in the DC specific media: IMDM, 10% human serum AB, 1% v/v sodium pyruvate, 2mM glutamine, and 200ng/ml human FLT3 ligand (cat no. 130-096-479, Milteny Biotec). Fetal bacteria (heat-killed) were added to DC with an MOI of 1:10 (10,000 bacteria per 1000 DC) and incubated for 8 hours at 37°C. T cells were labeled with CellTrace Violet (CTV) dye and a small portion was analyzed by flow cytometry for dye incorporation and cell viability. DC were gently washed with sterile PBS and CTV labeled T cells were added to each well in 1:10 ratio of DC:T cells (10,000 T cells per 1000 DC), and incubated for 6 days in T cell media: IMDM, 10% human serum AB, 1% v/v sodium pyruvate, 2mM glutamine, 1% Pen-Step solution (antibiotic). Post 6 days culture, T cells were harvested and stained with fluorescent antibodies followed by flow-cytometry analysis using BD-LSR-V (BD biosciences) and data were analyzed using FlowJo software.

72For PFA induced DC fixation: DC were first infected with fetal bacteria for 8 hours, followed by fixation using 1% PFA for 10 minutes. The PFA was then removed after 10 mins and the reaction was quenched with 200mM glycine. Cells were washed twice with sterile PBS and RPMI respectively. Rest of the process of T cell addition and incubation proceeds as explained earlier.

73For PTI induced blocking of cytokine release: Post bacterial priming of DC for 8 hours, cells were incubated with a combination of Brefeldin A and Monensin (Protein Transport Inhibitors) for 2 hours. The cells were then washed twice with sterile PBS and RPMI respectively. Rest of the process of T cell addition and incubation proceeds as explained earlier.

74Statistical analysis was performed in GraphPad Prism 7 using the specific tests indicated in the respective figure legends. The number of samples used for each analysis (n) is indicated in the respective figure or legend.