S100A8 and S100A9 Are Important for Postnatal Development of Gut Microbiota and Immune System in Mice and Infants

BSIC ND TRANATIONAL AT S100A8 and S100A9 Are Important for Postnatal Development of Gut Microbiota and Immune System in Mice and Infants Maike Willers,* Thomas Ulas,* Lena Völlger, Thomas Vogl, Anna S. Heinemann, Sabine Pirr, Julia Pagel, Beate Fehlhaber, Olga Halle, Jennifer Schöning, Sabine Schreek, Ulrike Löber, Morgan Essex, Peter Hombach, Simon Graspeuntner, Marijana Basic, Andre Bleich, Katja Cloppenborg-Schmidt, Sven Künzel, Danny Jonigk, Jan Rupp, Gesine Hansen, Reinhold Förster, John F. Baines, Christoph Härtel, Joachim L. Schultze, Sofia K. Forslund, Johannes Roth, and Dorothee Viemann

BACKGROUND & AIMS: After birth, the immune system matures via interactions with microbes in the gut. The S100 calcium binding proteins S100A8 and S100A9, and their extracellular complex form, S100A8-A9, are found in high amounts in human breast milk. We studied levels of S100A8-A9 in fecal samples (also called fecal calprotectin) from newborns and during infancy, and their effects on development of the intestinal microbiota and mucosal immune system. METHODS: We collected stool samples (n ¼ 517) from fullterm (n ¼ 72) and preterm infants (n ¼ 49) at different timepoints over the first year of life (days 1, 3, 10, 30, 90, 180, and 360). We measured levels of S100A8-A9 by enzyme-linked immunosorbent assay and analyzed fecal microbiomes by 16S sRNA gene sequencing. We also obtained small and large intestine biopsies from 8 adults and 10 newborn infants without inflammatory bowel diseases (controls) and 8 infants with necrotizing enterocolitis and measured levels of S100A8 by immunofluorescence microscopy. Children were followed for 2.5 years and anthropometric data and medical information on infections were collected. We performed studies with newborn C57BL/6J wild-type and S100a9 -/mice (which also lack S100A8). Some mice were fed or given intraperitoneal injections of S100A8 or subcutaneous injections of Staphylococcus aureus. Blood and intestine, mesenterial and celiac lymph nodes were collected; cells and cytokines were measured by flow cytometry and studied in cell culture assays. Colon contents from mice were analyzed by culture-based microbiology assays. RESULTS: Loss of S100A8 and S100A9 in mice altered the phenotypes of colonic lamina propria macrophages, compared with wild-type mice. Intestinal tissues from neonatal S100-knockout mice had reduced levels of CX3CR1 protein, and Il10 and Tgfb1 mRNAs, compared with wild-type mice, and fewer T-regulatory cells. S100-knockout mice weighed 21% more than wild-type mice at age 8 weeks and a higher proportion developed fatal sepsis during the neonatal period. S100-knockout mice had alterations in their fecal microbiomes, with higher abundance of Enterobacteriaceae. Feeding mice S100 at birth prevented the expansion of Enterobacteriaceae, increased numbers of T-regulatory cells and levels of CX3CR1 protein and Il10 mRNA in intestine tissues, and reduced body weight and death from neonatal sepsis. Fecal samples from term infants, but not preterm infants, had significantly higher levels of S100A8-A9 during the first 3 months of life than fecal samples from adults; levels decreased to adult levels after weaning. Fecal samples from infants born by cesarean delivery had lower levels of S100A8-A9 than from infants born by vaginal delivery. S100 proteins were expressed by lamina propria macrophages in intestinal tissues from infants, at higher levels than in intestinal tissues from adults. High fecal levels of S100 proteins, from 30 days to 1 year of age, were associated with higher abundance of Actinobacteria and Bifidobacteriaceae, and lower abundance of Gammaproteobacteria-particularly opportunistic Enterobacteriaceae. A low level of S100 proteins in infants' fecal samples associated with development of sepsis and obesity by age 2 years. CONCLUSION: S100A8 and S100A9 regulate development of the intestinal microbiota and immune system in neonates. Nutritional supplementation with these proteins might aide in development of preterm infants and prevent microbiotaassociated disorders in later years. C olonizing microbes provide the newborn immune system with crucial instructing information that induces postnatal maturation of immunity. 1,2 This codevelopment ideally results in a balanced homeostasis between the host and the microbiota, 3 whereas alterations predispose individuals to inflammatory and metabolic diseases. [4][5][6][7] Specific microbiota compositions have been described to precede necrotizing enterocolitis (NEC) 8 in preterm infants and to be important direct sources of sepsis-causing bacterial translocation. 9,10 Except for fullterm birth, vaginal delivery (VD), and breastfeeding, [11][12][13][14][15][16] little is known about what host factors influence the interplay between intestinal immunity and initial gut colonization to ensure developing homeostasis.
In healthy neonates, high serum concentrations of the endogenous Toll-like receptor 4 ligands S100A8 and S100A9 induce a state of microbial hyporesponsiveness in blood monocytes, also described as "stress tolerance." [17][18][19] Physiologically, they form a heterodimer complex (S100A8-A9) known as calprotectin. 20,21 Later in life, when acutely released during inflammatory settings, S100A8-A9 exerts primarily amplifying effects and serves as biomarker of inflammation, for example, fecal calprotectin (hereafter "fS100A8-A9") in inflammatory bowel disease (IBD). 22 Breast milk contains extremely high levels of S100A8-A9 23 and fS100A8-A9 levels are high in healthy breast-fed infants. 24 After birth, blood-derived (BD) monocytes replace yolk sac-derived intestinal macrophages and differentiate into tissue-resident lamina propria macrophages (LPMPs). [25][26][27] The invasion of S100-primed neonatal monocytes together with the high local supply at the luminal site via breast milk prompted us to speculate about S100A8-A9 playing an important role for the postnatal development of intestinal immunity and microbial colonization.
We found that high levels of fS100A8-A9 in the neonatal gut induce a regulatory LPMP phenotype permissive for the expansion of regulatory T cells (Tregs), which promotes a favorable development of the gut microbiota. The clinical relevance is corroborated by strong associations of neonatal fS100A8-A9 deficiency with dysbiosis-linked diseases like NEC, sepsis, and obesity, which can experimentally be prevented by a single nutritional supply of S100A8 at birth.

Human Samples
Stool samples were collected from 2 birth cohorts of healthy term (n ¼ 72, Supplementary

Mice and Mouse Studies
Mouse strains housed under specific pathogen-free conditions were the Foxp3-eGFP reporter mouse B6.Cg-Foxp3 tm1 Mal/J for the isolation of naïve CD4 þ T cells, and the C57BL/6J wild-type (WT) mouse and the S100a9 -/mouse 28 (also deficient for S100A8 due to the dependency of the posttranscriptional stability of S100A8 on the presence of S100A9 28,29 and the failure to form heterocomplexes 20,21 ) to harvest colons and mesenterial and celiac lymph nodes. In indicated experiments, S100a9 -/pups were injected intraperitoneally (IP) or fed per os within 24 hours after birth with recombinant murine S100A8 or phosphate-buffered saline (controls). For details regarding mating and cross-fostering settings, cell isolation, immunofluorescence microscopy of tissue samples, and of murine fecal samples, see Supplementary Experimental Procedures.

Murine Model of Neonatal Sepsis
At the age of 3 days, mice were infected subcutaneously with 7 Â 10 4 colony-forming units (CFU) of Staphylococcus aureus and returned to their mothers. Mice were monitored over 80 hours for survival or killed 24 hours after infection to harvest blood for cytokine measurements and organs. Bacterial

BACKGROUND AND CONTEXT
After birth, the immune system develops via interactions with microbes in the gut. S100A8 and S100A9 (also called calprotectin) are immune-stimulatory S100 calcium binding proteins present in high amounts in breast milk.
NEW FINDINGS S100A8 and S100A9 regulate development of the intestinal microbiota and immune system in neonates.

LIMITATIONS
Further studies are needed to determine the effects of S100 proteins in humans and in newborns.

IMPACT
Nutritional supplementation with S100 proteins might aide in development of preterm infants and prevent microbiota-associated disorders during childhood. burden was determined as described previously 17 by plating homogenates of lungs and livers onto blood-agar plates and counting bacterial colonies after 18 hours of incubation at 37 C.

Cell Culture Assays
LPMPs were ex vivo incubated with lipopolysaccharide (LPS) or phosphate-buffered saline and processed for gene expression studies. LPMPs were co-cultured with naïve T cells from Foxp3-eGFP reporter mice to assess their Treg-inducing capacity. For details of cell culture conditions, flow cytometry and quantitative reverse transcriptase polymerase chain reaction, see Supplementary Experimental Procedures.

16S rRNA Gene Bacterial Profiling
16S rRNA gene profiling of human fecal samples was performed as described previously. 10 DNA was isolated using the PowerSoil DNA Isolation Kit (MOBIO, Carlsbad, Canada) with an additional 1.5-hour Protease K incubation before the first centrifugation step. From each DNA sample, the V3-V4 region of the 16S rRNA gene was amplified using primers spanning the V3-V4 hypervariable regions (Supplementary Table 4). PCR including quantification of amplicons and library preparation was performed as described elsewhere. 30

Cytokine Studies
Plasma levels of tumor necrosis factor-alpha (TNF-a) and interleukin-1 beta (IL-1b) were measured by using murine LEGENDplex assays (BioLegend, San Diego, CA). FACS Canto II flow cytometer was used for measurements and LEGENDplex Data Analysis Software v7.0 (BioLegend) for data analysis.

Data Availability
16S rRNA sequencing files were submitted to the National Center for Biotechnology Information Sequence Read Archive (www.ncbi.nlm.nih.gov/sra) and are available with BioProject accession number PRJNA514340.

Results
Abundant S100A8-A9 in the Gut of Healthy Infants To obtain a comprehensive picture of fS100A8-A9 levels in newborn infants, 517 stool samples collected from a birth cohort of 72 healthy term infants (Supplementary Table 1) and 49 preterm infants (Supplementary Table 2) were analyzed. In term infants, fS100A8-A9 levels are significantly higher during the first 3 months of life compared with normal adult values and normalize along with weaning until the end of the first year ( Figure 1A). In preterm infants, initial fS100A8-A9 levels are significantly lower than in term infants (Supplementary Figure 1A), but increase during the first month of life ( Figure 1B). In both cohorts, fS100A8-A9 levels were dependent on the mode of delivery (MOD) with higher levels after VD than cesarean section (CS) ( Figure 1C and D), particularly after primary CS (Supplementary Figure 1B). In term infants, the MOD impacts on fS100A8-A9 levels only in the first week of life, but in preterm infants until the third week of life (Supplementary Figure 1C and D). Breast milk is one important source of fS100A8-A9 in newborn infants. 23,24 To investigate whether S100A8-A9 is also produced endogenously in the neonatal gut, we used human SI and LI tissue samples of newborn and adult patients without underlying IBD. Under these physiologic conditions, S100-protein expression is detectable in LPMPs but not in intestinal epithelial cells, with the proportion of S100A8-expressing LPMPs being significantly higher in the SI ( Figure 1E) and LI ( Figure 1F) of neonates compared with adults. Thus, the gut of healthy neonates is exposed to high amounts of S100A8-A9, while deficient states are particularly observed in preterm infants delivered by CS.

High fS100A8-A9 During Infancy Supports Developing Gut Eubiosis
To assess whether physiologically high fS100A8-A9 in neonates plays a role for the development of the gut microbiome we generated 16S rRNA gene profiles of 414 stool samples from the term infant cohort. Overall diversity of gut microbiota composition significantly increases during the first year of life ( Figure 2A) and follows known = Figure 1. Postnatal abundance of S100A8-A9 in the human intestinal tract. (A-D) S100A8-A9 levels in fecal samples collected at indicated ages from a cohort of (A,C) healthy term infants (n ¼ 72) and (B, D) preterm infants (n ¼ 48). (A, B) Bars on a logarithmic scale represent means ± SEM. *P < .05, ****P < .0001 (Kruskal-Wallis test). Dotted line indicates the cutoff for normal adult fS100A8-A9 levels (50 mg/g). (C, D) Regression of fS100A8-A9 levels subgrouped according to the MOD (grayshaded 95% confidence interval). P value indicates differences between both groups (likelihood ratio tests of nested models with/without MOD as predictor beyond age). (E, F) Top, representative images of human adult and neonatal SI (E) and LI (F) tissue samples from individuals without underlying intestinal inflammation immunostained for S100A8 (red), CD68 (green), and nuclei (with 4 0 ,6-diamidino-2-phenylindole [DAPI]; blue). Bottom, total number of CD68 þ LPMPs (blue dotted bar) differentiated in S100A8 -CD68 þ LPMPs (green bar section) and S100A8 þ CD68 þ LPMPs (red bar section). Values represent means ± SEM of counts in 4 high-power fields of n ¼ 4 adult and n ¼ 6 neonatal SI samples (E) and n ¼ 4 adult and n ¼ 4 neonatal LI samples (F). *P <.05, **P <.005 (Mann-Whitney U test). dynamics in bacterial classes, 2,11,16 most prominent the increase of Actinobacteria, Bacteroidia, and Clostridia and the decrease of Bacilli and Gammaproteobacteria ( Figure 2B). To test the impact of fS100A8-A9 and MOD on longitudinal microbiome features nested model comparisons were performed using strict filter criteria for differential taxon abundances while accounting for confounding influences. The initial colonization pattern (birth to 10 days), especially that of Bacilli, Bacteroidia, Clostridia, and Gammaproteobacteria, is primarily determined by the MOD but not fS100A8-A9 (Supplementary Figure 2 and Supplementary  Table 5). However, after that period, MOD impact only remains detectable for the expansion of Clostridia. In contrast, high fS100A8-A9 levels during entire infancy (!30 days to 1 year) are linked to a specific microbiome signature correlating with a higher abundance of Actinobacteria and anticorrelating with the abundance of Gammaproteobacteria ( Figure 2C and Supplementary Table 5). At the family level, fS100A8-A9 has particular impact on the expansion of Bifidobacteriaceae, and in turn contributes to the reduction of Enterobacteriaceae in the infant's gut (Supplementary Table 5). Prediction models based on these gut microbial communities revealed significant effects on the abundance gut metabolic pathways of fS100A8-A9 after the first 10 days of life, but not of the MOD (Supplementary Table 5), with fS100A8-A9 promoting production and inhibiting . (E,F) Proportion of S100a9-/-mice with Enterobacteriaceae-positive feces at day 10 after (E) feeding or (F) IP injection of S100A8 or phosphate-buffered saline (PBS) (Ctrl) within 24 hours after birth (each n ¼ 22-25). *P <.01, **P <.005, n.s., not significant (sign tests). Figure 3). The data suggest that S100-alarmins are crucial for promoting a favorable infant gut state of abundant Actinobacteria-like Bifidobacteriaceae and growth restriction of Gammaproteobacteria including pathobiont Enterobacteriaceae, resulting in a shift toward health-promoting gut metabolic functions.

degradation of short-chain fatty acids (Supplementary
To get further mechanistic insight, we performed mouse experiments using WT and S100a9 -/mice. Dams were cohoused to ensure comparability of the mother's gut microbiota and separated on day 18 of pregnancy shortly before giving birth. In the WT and S100a9 -/offspring, no striking difference is detectable in the overall fecal bacterial count ( Figure 3A). However, while cultivable amounts of Enterobacteriaceae, mainly Escherichia coli based on its characteristic morphology on MacConkey agar, in the S100a9 -/offspring are already detectable from day 1 on, they are not observed in WT mice before weaning (day 21) ( Figure 3B).
In a cross-fostering setting, early-life overgrowth of Enterobacteriaceae is completely abrogated in S100a9 -/neonates fostered by WT mothers, while in WT neonates fostered by S100a9 -/mothers, cultivable amounts of Enterobacteriaceae get traceable already on day 10 ( Figure 3C and D). Thus, endogenous production of S100A8-A9 alone is insufficient in order to achieve full effect, which instead requires supplementation by S100-alarmins via breast milk, whereas the effect of breast milk-derived S100A8-A9 alone is full restriction of Enterobacteriaceae growth.
To exclude direct antimicrobial effects from S100alarmins by chelating Mn 2þ and Zn 2þ , 23,31,32 we supplied S100a9 -/neonates with the S100A8 homodimer that lacks binding sites for divalent metal ions. 21,32,33 A one-time feeding of S100A8 after birth is sufficient to successfully limit the subsequent expansion of Enterobacteriaceae ( Figure 3E) suggesting that the S100-priming of neonatal intestinal immunity is important for the microbiota-shaping effect, whereas the systemic administration of S100A8 at a dose strongly immunoregulatory for blood monocytes 17,18 has no significant impact on the abundance of Enterobacteriaceae ( Figure 3F).
Collectively, our findings clearly reveal that high fS100A8-A9 in infants affects the mutual relationship between the host and the gut microbiota in a beneficial manner by restricting the growth of pathobiont Gammaproteobacteria, particularly Enterobacteriaceae including E coli, and promoting the colonization with eubiont Actinobacteria including Bifidobacteriaceae.

S100-alarmins Regulate LPMPs
To explore how S100A8-A9 primes intestinal immunity, we isolated colonic LPMPs from WT and S100a9 -/mice at day defined ages (Supplementary Figure 4A and B). Morphologically, there are no differences between WT and S100a9 -/-LPMPs (Supplementary Figure 4C) or numeric alterations in the proportions or the replacement of yolk sac LPMPs by BD-LPMPs 27 ( Figure 4A and Supplementary Figure 4B). The same holds true for the proportion of chemokine (C-X3-C motif) receptor 1 (CX3CR1)-positive LPMPs ( Figure 4B). However, the mean expression of CX3CR1 is significantly lower on S100a9 -/-BD-LPMPs compared with WT BD-LPMPs during the neonatal period and adolescence ( Figure 4C and D). Next, freshly isolated LPMPs were used to capture the in vivo tuned baseline of gene expression ( Figure 4E). In S100a9 -/-LPMPs, Tnf is significantly higher expressed on day 3, but lower on day 10 and day 21 than in WT LPMPs, whereas the expression of Il1b is comparable. The microbial responsiveness of the LPMPs was studied by ex vivo LPS stimulation ( Figure 4F). On day 3, S100a9 -/-LPMPs respond with a significantly stronger expression of Tnf and Il1b, whereas from day 10 the pattern switches to a significantly reduced inducibility of these proinflammatory genes compared with WT LPMPs. During the first postnatal days, the extreme high levels of S100A8-A9 apparently tolerize the LPS response of LPMPs, whereas with increasing age while basal S100A8-A9 levels decrease its amplifier function takes effect on induction. 19,[34][35][36] In contrast, the regulatory genes Il10 and transforming growth factor-b (Tgfb1) show nearly constant basal ( Figure 4E) as well as LPS-induced ( Figure 4F) expression deficits in S100a9 -/-LPMPs compared with WT LPMPs, pointing to an important impact of S100A8-A9 on the regulatory phenotype of LPMPs. Bars represent means ± SEM. *P <.05, **P <.01, ***P <.005 (post hoc unpaired t-test) (each n ¼ 4-11). MFI, mean fluorescence intensity. (D) Upper left panels, representative images of LI tissue samples from day 10 WT and S100a9 -/mice immunostained for CX3CR1 (red), F4/80 (green), and nuclei (DAPI; blue). Scale bar, 50 mm. Lower left panels, single color stainings of the dotted zoom-out. Scale bars, 10 mm. Right, CX3CR1 expression by F4/80 þ LPMPs. Box plots show medians, interquartile ranges and total ranges of the MFI of CX3CR1 per F4/80 þ region of interest (ROI) determined in 14 randomly selected images. *P < .01 (unpaired t-test). (E, F) Expression of indicated genes in LPMPs isolated from the LI of WT and S100a9 -/mice at indicated ages. (E) Basal expression level (in vivo) and (F) ex vivo LPS-induced fold changes of expression. Plots represent means ± SEM (each n ¼ 3-6). *P < .05, **P < .01, ***P < .005 (post hoc unpaired t-test).

December 2020
Calprotectin Ensures Gut Homeostasis 2137 nodes are comparable in WT and S100a9 -/mice. However, the subsequent expansion of Tregs in the colon is significantly impaired in adolescent S100a9 -/mice compared with WT mice ( Figure 5A-D). In co-culture, Tregs are significantly better inducible from naïve T cells in the presence of WT LPMPs than S100a9 -/-LPMPs ( Figure 5E and F). Supplementation of TGF-b increases the overall yield of Tregs that nevertheless still remains lower in co-culture with S100a9 -/-LPMPs compared with WT LPMPs ( Figure 5G). The additional supplementation of IL-10, however, compensates for the deficient Treg-inducing capacity of S100a9 -/-LPMPs ( Figure 3H). These findings suggest that the S100A8-A9-induced LPMP phenotype promotes the postnatal development of Tregs in the colonic mucosa particularly by providing IL-10.

December 2020
Calprotectin Ensures Gut Homeostasis 2139 significantly higher level of CX3CR1 on LPMPs after S100A8 supplementation compared with control mice, independent of the route of administration ( Figure 6B and G). Likewise, Tregs expand significantly better after S100A8 application, particularly via the enteral route ( Figure 6C and H). S100A8 effects on the transcriptional programming were assessed on day 10 when the baseline expression of Il10 and Tgfb1 and the LPS-inducibility of Tnf, Il1b, and Il10 are impaired in S100a9 -/-LPMPs ( Figure 4E and F). S100A8 given IP has no effect on the basal expression of Il10 and Tgfb1 ( Figure 6D) and only improves the LPS-inducibility of Tnf and Il1b ( Figure 6I). However, postnatal one-time feeding of S100A8 significantly increases the basal expression level of Il10 ( Figure 6E) and enhances the responsiveness of all LPSinducible genes, that is, Tnf, Il1b and Il10 ( Figure 6J). Cross-fostering experiments ( Figure 6K-N) corroborated the importance of the enteral supply of S100-alarmins. Breast milk of WT mothers alone induces a full regulatory LPMP phenotype in S100a9 -/neonates with respect to CX3CR1 levels and Tnf, Il10, and Tgfb1 expression and proper Treg expansion comparable to the co-fostered WT offspring. Instead, comparing WT neonates fostered by S100a9 -/mothers with co-fostered WT neonates reveals that sole endogenous production of S100A8-A9 leads only to a deficient expression of Tnf and Tgfb1 and less strong Treg expansion. Collectively, these findings demonstrate that the perinatal enteral supply of S100A8 regulates the inflammatory responsiveness of LPMPs which promotes the postnatal development of Tregs in the gut.

High fS100A8-A9 During Infancy Prevents
Dysbiosis-linked Diseases NEC in preterm infants is preceded by gut dysbiosis with an early abundance of Enterobacteriaceae. 8, 10 We determined the proportion of S100A8-expressing LPMPs in intestinal tissue samples obtained from NEC patients in comparison with age-matched controls (Supplementary Table 3). In NEC samples, barely any S100A8 þ LPMPs are found compared with the high number of S100A8expressing LPMPs in neonates without underlying intestinal inflammation ( Figure 7A). The lack of enteral supply of S100A8-A9 due to the common practice of fasting in imminent NEC and the reduced endogenous production of S100A8-A9 in such patients suggest that S100-alarmins protect against NEC.
Another early-life disease associated with gut dysbiosis is neonatal LOS. 9,10 In the cohort of preterm infants, fS100A8-A9 levels preceding LOS are significantly lower than corresponding levels in infants without later sepsis ( Figure 7B). fS100A8-A9 levels below 28 mg/g are associated with a 14-fold higher LOS risk than levels above 153 mg/g (Supplementary Table 6). In a murine model of S aureus-induced infection, we tested the prevention of neonatal sepsis by enterally supplied S100A8 ( Figure 7C). One-time feeding of S100A8 at birth significantly lowers the death rate of S100a9 -/neonates from later sepsis ( Figure 7D) along with an effective restriction of the cytokine response ( Figure 7E) and bacterial burden ( Figure 7F).
In the long term, gut dysbiosis has been identified as risk factor for the development of obesity. 7,41,42 We find comparable birth weights in WT and S100a9 -/pups, however, from day 10 on a significantly stronger weight gain in S100a9 -/mice compared with WT mice ( Figure 7G). Fostering of S100a9 -/neonates by WT mothers significantly restricts the weight gain resulting in adult body weights comparable to that of co-fostered WT mice. In turn, WT mice that are fostered by S100a9 -/mothers end up with significantly higher body weights than the co-fostered WT offspring ( Figure 7H). In humans, the body mass index (BMI) is a predictive value for early childhood obesity 43 and was monitored in our term infant birth cohort. The delta BMI from birth until 2 years of age was correlated with the mean of neonatal fS100A8-A9 levels while accounting for confounding influences. There is a strong inverse relationship; children with low fS100A8-A9 levels show a significantly stronger increase in their BMI over time compared with children with high fS100A8-A9 levels whose BMI at 2 years differs little from that at birth ( Figure 7I).

Discussion
Our knowledge about what programming of neonatal immunity ensures optimal adaptation to the new extrauterine environment and what parameters determine such programming is fragmented. By detailed cellular and molecular studies in mice and 2 human birth cohorts we identified S100A8-A9 in the neonatal gut as a crucial host factor contributing to the successful co-development of gut mucosal immunity and the microbiota. We demonstrated that fS100A8-A9 levels are physiologically high in healthy term babies, particularly after VD compared with CS. Laborinduced stress is obviously the driving force for fS100A8-A9 release as levels after secondary CS (with preceding labor) are comparable to those after VD. We showed previously that breastfeeding provides infants with a large amount of S100A8-A923 and here identify LPMPs in the neonatal intestine as a second source of fS100A8-A9. The lower S100A8-A9 levels in breast milk of mothers that gave birth to a preterm infant, 23 slower increase of feeding volumes and higher CS rates are presumably the main reasons for reduced fS100A8-A9 in preterm compared with term babies. Other factors like birth weight, Apgar score, diet, and probiotics might also influence neonatal fS100A8-A9 and are currently analyzed in a prospective multicenter clinical study.
In our birth cohort, high fS100A8-A9 levels correlated with a specific gut microbiota state characterized by a strong expansion of Actinobacteria-like Bifidobacteriaceae and growth restriction of Gammaproteobacteria-like Enterobacteriaceae, implying an increase in healthbeneficial gut metabolic activity including enriching short-chain fatty acid availability, although it is important to recognize this is a prediction from taxonomic composition which future studies must validate experimentally. Bifidobacteriaceae might contribute to the restriction of Enterobacteriaceae by producing acetate. 44 But for deeper species resolution and assessment of the gut functional potential future shotgun metagenomic analyses are essential. Enterobacteriaceae include important neonatal sepsis-causing opportunistic pathogens, eg, E coli, Enterobacter, and Klebsiella species. 9,10 E coli has even been proposed as marker of gut dysbiosis associated with diseases like NEC, 8,45 childhood atopy, 6 IBD, [46][47][48] and obesity. 49 In murine neonates, endogenously produced S100A8-A9 partly but enterally supplied S100A8-A9 via breast milk completely prevented the intestinal overgrowth of Enterobacteriaceae. Herein, S100A8-A9 likely acts in synergy with immunoregulatory human milk oligosaccharides, which prevent the mucosal attachment of pathobionts like E coli 50 while encouraging the growth of Bifidobacteriaceae. 51 Our study demonstrates that S100A8-A9 in the intestine has obviously a dual role of extended relevance for the development of mucosal immunity. Directly after birth, under noninflammatory conditions, the exposure to S100A8-A9 induced microbial tolerance in LPMPs by restricting their LPS responsiveness in terms of Tnf and Il1b expression (eg, against the first wave of microbial colonization). This is similar to the S100A8-A9-mediated stress tolerization of newborn monocytes found previously. [17][18][19] After the neonatal period (in mice at >3 days), the amplifier function of S100A8-A9 19,34-36 applied by supporting the proinflammatory response of LPMPs (eg, in response to expanding microbiota and for defending possible pathogens). Particularly, however, S100A8-A9 consistently induced a regulatory LPMP phenotype with full effect of high CX3CR1 levels, and Il10 and Tgfb1 expression especially when supplied enterally, which promoted the expansion of Tregs. Our ex vivo as well as in vivo findings point to IL-10 playing a crucial role for the S100-mediated regulation of Tregs by LPMPs. The microbiome-shaping effect might amplify the expansion of Tregs by increasing the levels of commensal-derived butyrate. 52 Potential direct Treg-inducing activity of S100A8-A9 appears less relevant as its extracellular half-life is only approximately 24 hours, 35,53 whereas the Treg-promoting effect of a single feeding of S100A8 at birth was not evident before day 10. This is the first evidence that next to microbes also endogenous TLR-ligands like S100A8-A9 have crucial instructing function for the newborn intestinal immune system. The here-identified immunomodulatory effects of fS100A8-A9 explain well the resulting shaping effect on the microbiota composition. CX3CR1, IL-10, and TGF-b as well as Tregs have been demonstrated to be important factors controlling the gut microbiota composition, specifically also the expansion of Enterobacteriaceae. [54][55][56][57] Interestingly, signs of colitis in IL10R-deficient mice are absent during the first 2 weeks of life, while intestinal inflammation and macrophage dysfunction begin during the third week, concomitant with weaning. 58 Thus, the immunoregulatory effects other than IL-10 induction might be of particular relevance for the S100-mediated impact on the gut microbial ecology.
The clinical relevance of S100-programming of neonatal gut immunity is shown for NEC and LOS, frequent dysbiosisassociated diseases in preterm infants. We observed virtually no expression of S100A8-A9 in the intestine of patients with NEC. This finding is supported by a previous observation of unusually low fS100A8-A9 concentrations in the meconium of patients with later NEC. 59 In contrast, due to the physiologically increased basal levels in neonates, which moreover are developmentally controlled (GA and age) 59,60 with high interindividual variation, 61 fS100A8-A9 is no valuable biomarker in NEC like in IBD of older children and adults. 22 In our cohort, low fS100A8-A9 levels reliably predicted a high LOS risk even when controlling for GA. Experimentally, 1-time feeding of S100A8 to S100a9 -/neonates prevented fatal outcomes of later sepsis. It might appear contradictory that there was only one E coli-positive LOS case within the blood-culture positive subjects of our preterm infant cohort while Enterobacteriaceae expand at low fS100A8-A9 levels. On the other side, our study might suggest a new perspective on the linkage between gut dysbiosis and sepsis by suggesting both being separate consequences of insufficiently S100-primed neonatal immunity that predisposes to unregulated immune responses to microbial challenges. [17][18][19] The here-identified link between low neonatal fS100A8-A9 levels and a disproportionate weight gain during childhood points to a role in dysbiosis-associated long-term diseases. Skewed gut microbiota composition has been shown to predispose to the development of obesity. 7,41,42,49 In the mouse model, 1-time feeding of S100A8 after birth prevented excessive weight gain in adolescent S100a9 -/mice. Future clinical studies in humans will have to clarify whether food supplement with S100A8-A9 during infancy prevents the development of unfavorable gut microbiota signatures and therewith associated diseases to promote long-term health.
Collectively, our data demonstrate for the first time that abundant fS100A8-A9 trains intestinal mucosal immunity by regulating the inflammatory programming and cellular development during infancy, which goes along with colonization by a favorable microbiota. Intestinal S100A8-A9 deficiency increases the risk of newborn individuals to develop unfavorable gut microbiota signatures and associated diseases. Animal data suggest nutritional supplementation of S100-alarmins to high-risk neonates as promising preventive measure.

Supplementary Material
Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/ j.gastro.2020.08.019.

Study Population
Stool samples (n ¼ 517) were collected from 2 birth cohorts of healthy term and preterm infants. Term infants (n ¼ 72, Supplementary Table 1) were prospectively enrolled at the Hannover Medical School from May 2015 to June 2017. The inclusion criteria were being born healthy and appropriate for gestational age (GA) with a GA of 37 0/ 7 to 41 6/7 weeks, an Apgar score of !8 after 5 minutes, and intended breast feeding as predominant diet. GA was calculated on the basis of the mother's last menstrual period. When early ultrasound at 11 to 13 6/7 weeks of gestation using fetal crown-rump length deviated by more than 7 days, dating was done by ultrasound. After written informed consent was obtained from parents, d1 and d2 stool samples were collected by medical staff from diapers and after that by parents on d10, d30, and d90 and at 6 and 12 months. Infants from pregnancies that involved in vitro fertilization, multiple gestations, births that resulted from maternal trauma, and newborns with any congenital malformations or chromosomal anomalies, with clinical or laboratory signs of amnion infection syndrome or if the mother or infant received any peripartal antibiotic therapy were excluded. Cohort subjects were followed until 2.5 years of age. Metadata collected on maternal characteristics were age and BMI at delivery and prepregnancy BMI. Childrelated data included the BMI (birth, d30, 3m, 6m, 1y, and 2y), diet (exclusive breastfeeding, mixed feeding with formula, age at introduction of solid food) history of allergies, infections, antibiotics and inhalations during the first year of life. Information on life circumstances included the presence of siblings, smoking status of the parents, pets, and residence (county/urban).
The birth cohort of preterm infants was established between January 2012 and January 2017 and included 49 preterm infants born at a GA of 23 0/7 to 31 6/7 weeks in the Neonatal Intensive Care Unit of the Department of Pediatrics at the University of Lübeck (Supplementary  Table 2). Infants with lethal malformations or congenital anomalies of the gastrointestinal tract were excluded. LOS was defined as sepsis starting after 72 hours of life. LOS was diagnosed with the occurrence of !2 clinical signs of systemic inflammatory response and 1 laboratory sign. Clinical signs of LOS included temperature >38 C or <36.5 C, tachycardia >200/min, new onset or increased frequency of bradycardias or apneas, hyperglycemia >140 mg/dL, base excess <À10 mval/L, changed skin color, and increased oxygen need. Laboratory signs of LOS were Creactive protein >10 mg/L, platelet count <100/nL, immature/total neutrophil ratio >0.2, and white blood cell count <5/nL. Blood culture confirmed LOS was defined as clinical LOS with bacterial growth in the blood culture, which was taken before commencing antibiotic therapy. LOS occurred in n ¼ 22 cases at a mean age of 14.8 days ± 4.9, from which 11 were blood culture proven (Staphylococcus haemolyticus 5, Staphylococcus. epidermidis 4, Escherichia coli 1, Enterococcus faecalis 1). Fecal samples were collected from the preterm birth cohort on days 1 to 7, 8 to 12, 13 to 19, and 20 to 30 by medical staff from diapers or after rectal enema with 37 C NaCl 0.9% solution. Stool was immediately stored at À20 C, and transferred to À80 C within 3 days.
Human intestinal tissue samples were obtained upon written informed consent from n ¼ 8 adult patients and n ¼ 18 newborn infants and included SI and LI resections (Supplementary Table 3). Biopsies in adults were performed due to the exclusion of malignancies or IBD or in the context of ileostomy or bile duct carcinoma and served as adult controls. Intestinal tissue from newborn infants was resected due to the diagnosis of NEC (n ¼ 8) or in the context of inborn atresia, perforation, stenosis, volvulus, or ileostomy (n ¼ 10). Biopsies of the latter served as neonatal intestinal controls. The tissue was formalin-fixed and paraffin-embedded and used for immunofluorescence microscopy.
Fecal S100A8-A9 (fS100A8-A9) Fecal sample aliquots were suspended in extraction buffer adopted from Hycult Biotech's H325 Human Calprotectin enzyme-linked immunosorbent assay (ELISA) kit. Suspensions were thoroughly vortexed, filtered through a 100-mm cell strainer, and then incubated on a shaker on ice for 20 minutes. The homogenates were centrifuged for 20 minutes at 10,000g at 4 C. The upper portion of the supernatants was pipetted off, frozen, and stored at À80 C until quantification of fS100A8-A9 by an in-house ELISA as described previously. 1,2

Immunofluorescence Microscopy
Human paraffin-embedded intestinal tissues sections were deparaffinized in Roti-Histol (Roth), rehydrated and subjected to heat-induced antigen retrieval in citrate buffer. In NEC samples, resection margins unaffected from necrosis and severe inflammation were analyzed. Before antibody staining, the tissue was washed and blocked with 5% skim milk powder in tris-buffered saline (TBS) to minimize nonspecific binding. The slides were stained overnight at 4 C in the dark with mouse anti-human CD68 monoclonal antibody (mAb) (M0814, Dako, Hamburg, Germany) and rabbit anti-human S100A8 polyclonal antibody (purified by T.V.) followed by AlexaFluor488 donkey anti-mouse and AlexaFluor555 donkey anti-rabbit secondary antibodies (both Life Technologies, Darmstadt, Germany). Tissue was mounted in VECTASHIELD Antifade Mounting Medium with 4 0 ,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA). Images were acquired using the Zeiss Axioplan 2 fluorescence microscope with a Zeiss EC Plan-Neofluar 10x/0.3 objective and a Zeiss EC Plan-Neofluar 40x/0.75 Ph2 objective, the AxioCam Color camera and AxioVision 40 v4.8.2 imaging software from Zeiss (Jena, Germany). The number of CD68 þ and S100A8 þ cells in human intestinal tissues sections was determined per highpower field by manual counting performed by 2 blinded observers.
Murine colon tissue was embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek, Torrance, CA) without prior fixation, frozen on dry ice, and stored at À80 C. Cryosections (8 mm) were fixed with acetone and stored at À20 C. Sections were rehydrated in TBS supplemented with 0.1% Tween20 (TBST) and blocked with 5% rat serum. For immunostaining, the following antibodies were used: rat anti-mouse forkhead box P3 (FoxP3) mAb (FJK-16s, eBioscience, San Diego, CA), rat anti-mouse CD3 mAb (17A2, purified in the Förster laboratory, Hannover, Germany), rat anti-mouse CD4 mAb (GK1.5, purified in the Förster laboratory), rat anti-mouse F4/80 mAb (BM8; Biolegend, San Diego, CA), and mouse anti-mouse Cx3cr1 mAb (SA011F11, BioLegend). Images were acquired using an Olympus BX61 epifluorescence microscope. For the analysis of Cx3cr1 expression on F4/80 þ LPMPs, at least 5 randomly selected images acquired from proximal, mid and distal LI sections prepared from 2 WT and 2 S100a9 À/À neonatal mice (day 10) were analyzed per condition. The same exposure times were used for the acquisition of the read-out channel for Cx3r1 staining. The intensity of Cx3cr1 expression was measured by defining regions of interest (ROIs) encompassing F4/80stained areas and measuring the maximum fluorescence intensity of the Cx3cr1 signal in ROIs. For the analysis of Tregs, CD3 þ CD4 þ T cells with a FoxP3-positive nucleus were counted manually by 2 blinded observers.

Mice and Mouse Studies
The following mouse strains used in this study were housed under specific pathogen-free conditions at the Central Animal Facility at Hannover Medical School. The Foxp3-eGFP reporter mouse B6.Cg-Foxp3 tm1 Mal/J was obtained from Jackson Laboratory (Bar Harbor, ME) and used for the isolation of naïve CD4 þ T cells from adult spleens. C57BL/6J WT mice purchased from Charles River (Sulzfeld, Germany) and S100a9 À/À mice (C57BL/6 background) 3 were constantly bred and litters used randomly at indicated ages to harvest colon contents, LIs, mesenteric lymph nodes (mLN), and celiac lymph nodes (cLN). No animals needed to be excluded from the studies. WT and S100a9 À/À mothers were co-housed for at least a week before and after mating until day 18 of pregnancy and then separated for delivery. In indicated experiments, S100a9 À/À pups were supplemented within 24 hours after birth either by IP injection of 2.5 mg S100A8 in 20 mL phosphatebuffered saline (PBS) or enteral feeding of 5 mg S100A8 in 20 mL PBS. Mice supplemented with PBS alone served as controls. For cross-fostering experiments, mating of WT and S100a9 À/À mice was terminated to exchange litters within 12 hours after birth.

Isolation of LPMPs From the Murine Colon
LPMPs were isolated from harvested murine LIs using a modified protocol published elsewhere. 4 After dissection, the colon was opened longitudinally, the stool was collected and the tissue was thoroughly washed in ice-cold PBS. In day 3 and day 10 neonates, 3 to 4 LIs were pooled for one experiment. To remove intestinal epithelial cells, the tissue was cut into 0.5-cm pieces and incubated in prewarmed PBS supplemented with 2 mM EDTA for 15 minutes at 37 C with agitation. Subsequently, the tissue was repeatedly shaken, minced, and digested for up to 60 minutes at 37 C under constant agitation in RPMI medium (Lonza, Basel, Switzerland) containing 10% fetal calf serum (FCS), 1% penicillin/streptomycin, 75 mg/mL liberase, and 30 U DNase I (both Sigma). The digested tissue was filtered through a 100-mm cell strainer followed by a 70-mm cell strainer. Thereby isolated lamina propria mononuclear cells (LPMCs) were pelleted and resuspended in PBS containing 2 mM EDTA and 2% bovine serum albumin (FACS buffer) and used for cell number determination, flow cytometry studies and cell sorting. LPMPs were sorted at the central Research Facility Cell Sorting at the Hannover Medical School, either on FACSAria Fusion and FACSAria IIu (both BD Biosciences, Heidelberg, Germany) or on MoFlo XDP (Beckmann-Coulter, Krefeld, Germany). Therefore, LPMPs were stained as described below and sorted by gating on live CD45 þ CD11b þ F4/80 þ/interm cells. The purity of isolated LPMPs was >90%. For the preparation of cytospins, LPMPs were seeded on superfrost slides and centrifuged at 300g for 5 minutes. Cells were fixed in 2% paraformaldehyde (PFA) and stained with hematoxylin/eosin. Remaining LPMPs were used for gene expression studies and ex vivo stimulation assays.

Flow Cytometry
All staining panels included a Fixable Viability Dye eFluor 506 (eBioscience) to exclude dead cells as well as CD16/CD32 (2.4G2, BioLegend) for blocking purposes. Staining was performed for 30 minutes in the dark at 4 C. After staining cells were fixed with 2% PFA.

Data Acquisition and Analyses
All flow cytometry analyses were performed using a FACS Canto II flow cytometer (BD Biosciences). Data were analyzed using DIVA software v8.0.1 (BD Biosciences) and Kaluza software v2.1 (Beckman Coulter, Miami Lakes, USA).

Cell Culture Conditions
Sorted LPMPs were seeded at a concentration of 1 Â 10 6 cells/mL in RPMI medium supplemented with 10% FCS, 1% L-glutamin and 1% penicillin/streptomycin. After overnight culture LPMPs were stimulated for 4 hours with 100 ng/mL LPS. Control cells were incubated with PBS, respectively. Subsequently, LPS-stimulated cells were harvested and processed for gene expression studies.

Microbiological Analysis of Murine Fecal Samples
Colon and cecum content were collected and suspended at 1 mL PBS/g feces. Serial dilutions were plated on Luria/ Bertani agar (Roth) and incubated overnight under aerobic conditions at 37 C and 10% CO 2 to determine the numbers of CFU/g feces. For the selective growth of Enterobacteriaceae, diluted samples were plated on MacConkey agar (AppliChem, Darmstadt, Germany). Presumptive identification of E coli was made based on its characteristic colony morphology on the MacConkey agar (dry, donut shaped, dark pink in color, and surrounded with dark pink area).

Bioinformatics of 16S rRNA Gene Sequencing Data
The raw fastq files were processed using mothur version 1.40.5. 6 After generation of contigs, sequences containing ambiguous bases and sequences longer than 500 base pairs were removed. With the primer sequences from the MiSeq sequencing experiment, a custom reference alignment was generated using the SILVA reference version 132. 7 Sequences containing more than 8 homo polymers were excluded and sequences of nonbacterial origin removed. Chimeras were detected by the VSEARCH algorithm 8 as implemented in mothur. Classification of sequences was performed using the Greengenes database gg_13_8_99. 9 Sequences were clustered into operational taxonomic units (OTUs) according to their taxonomic assignment. The analysis after pre-processing in mothur was performed in R (version 3.6.1), mainly relying on the phyloseq package. 10 Following mapping of 16S reads to OTUs and binning at taxonomic class and family levels, total mapped read abundances were down sampled using the rarefaction toolkit (RTK) 11 at default settings to account for differences in sampling depth as well as to compute alpha diversity metrics. For metagenome inference amplicon sequences were filtered, quality controlled, and taxonomically assigned using the LotuS pipeline. 12 Default parameters for the small subunit ribosomal ribonucleic acid and the MiSeq platform were used, while SILVA, Greengenes and HITdb (v1.00) 13 were used incrementally for taxonomic classification. Resulting abundance tables were normalized using RTK to 95% of the minimal column sum. Metagenome functional content was inferred from the raw amplicon sequences and processed taxonomic abundances using the PICRUSt2 method (v2.3.0-b). 14 MetaCyc pathway abundances were generated using the default map files and functional databases. 15 For nested model comparisons taxon abundances and MetaCyc pathway abundances (rarefied reads) were modeled individually using mixed effect models (R lme4 package) with fS100A8-A9, MOD, and age in days as fixed effects, with child identity as a random effect so as to account for the longitudinal nature of the data. Restricted maximum likelihood (REML) approximations were disabled to ensure likelihood ratio tests for nested model comparisons are accurate. For each tested variable (fS100A8-A9, MOD, age) a model omitting this feature was constructed and the 2 models compared in a likelihood ratio test (R lmtest package). In addition, each tested variable was directly compared against each tested feature (using a Mann-Whitney U [MWU] test for MOD and a Spearman test for age, and fS100A8-A9), with nonparametric directional effect sizes calculated as Spearman rho or Cliff's delta (R orddom package). Features were considered to be strongly and significantly affected by each feature if Benjamini-Hochberg-corrected P values for both nested model and simple test fulfilled false discovery rate (FDR) <0.1, and if absolute effect size exceeded 0.2.

Statistical Analysis
Statistical tests applied for 16S rRNA gene sequencing data analysis are described previously. Group comparison of fS100A8-A9 levels and proportions of S100A8 þ LPMPs in human tissue samples was performed by applying the 2tailed MWU test. Age dependency of fS100A8-A9 levels was evaluated by running a Kruskal-Wallis test including post hoc Dunn's multiple comparison test. To test for differences between WT and S100a9 À/À mice in gene expression or in fraction of Tregs measured at different ages, a nested model test was further used building generalized linear models (R lme4 package) of measured data using a Gamma distribution error model and setting zero values to a pseudocount floor of 1-e-10. Models with age and genotype were contrasted to models containing age only to test for main effect of genotype. Models with age, genotype, and an interaction term between them were contrasted to models containing age and genotype only to test for interaction effect between age and genotype. Models were compared using the R lmtest package. To test for differences between control and supplemented mice in gene expression or in fraction of Tregs, a nested model test was further used building generalized linear models (R lme4 package) of measured data using a Gamma distribution error model and setting zero values to a pseudocount floor of 1-e-10. Models with age and genotype were contrasted to models containing age only to test for main effect of supplementation. Models with age, supplementation, and an interaction term between them were contrasted to models containing age and supplementation only to test for interaction effect between age and supplementation. Models were compared using the R lmtest package. To test for differences between WT and S100a9 À/À mice in proportion of Tregs, a nested model test was further employed building generalized linear models (R lme4 package) of proportions of Treg to CD3 þ cells using a binomial error model. Models with age and genotype were contrasted to models containing age only to test for main effect of genotype. Models with age, genotype and an interaction term between them were contrasted to models containing age and genotype only to test for interaction effect between age and genotype. Models were compared using the R lmtest package. Comparisons in coculture assays and cross-fostering settings were evaluated by two-tailed t-tests. The difference between genotypes resp. intervention groups in the proportion of Enterobacteriaceae-positive fecal samples was tested by using the sign test. The association between fS100A8-A9 levels and the occurrence of LOS was assessed by calculating odds ratios and their 95% confidence intervals and by using a nested model test building generalized linear models (R lme4 package) of measured data using a Gamma distribution error model and setting zero values to a pseudocount floor of 1-e-10. The model with fS100A8-A9 and GA was contrasted to a model containing GA only to test for main effect of fS100A8-A9. Kaplan-Meier survival curves were generated using the Mantel-Cox test. The fS100A8-A9 levels of human term and preterm infants delivered by VD and CS and the body weight of WT and S100a9 À/À mice was modeled against the fixed factor time (age). Linear models were fitted using the lm function of the standard R-package stats. Likelihood ratio tests of the full model against the control model without the fixed effect of MOD respective genotype were performed to determine the P-value of the difference between the mouse strains regarding gain of weight over time. Bivariate linear regression was used to assess how the mean of fS100A8-A9 during the first 10 days of life relates to the delta of the BMI from birth until 2 years of age of the children of the birth cohort of term human infants (Supplementary Table 1). We obtained crude and multivariable-adjusted estimates of this association. The multivariable-adjusted models included terms for maternal age and BMI at delivery, maternal prepregnancy BMI, birthweight (all continuous variables), offspring sex (boy or girl), diet (exclusive and mixed breastfeeding), antibiotics during the first year of life (yes/no), presence of siblings (yes/no), smoking status of the parents (yes/no), pets (yes/ no), and residence (county/urban). P values < .05 were judged to be significant.