Marilen Benner

CHAPTER 7 190 RESULTS Microbial disruption does not result in pregnancy complications in mice Gestational antibiotic intervention of metronidazole, neomycin and polymyxin was previously shown to influence immunity of the offspring without observed pregnancy complications (31). Following this established treatment regime (Figure 1A), we assessed effective maternal microbial modification. 16S rRNA analysis of the gut showed a distinct microbiota composition in antibiotic treated mice compared to the control cohort (Figure 1B-C). In total 39 different genera showed a significant change in relative abundance as well when both groups were compared using ANCOM pipeline (Supplementary Table 1). In addition, a significant decrease of total alpha diversity represented by Shannon index diversity was observed in the antibiotic treated group (p-value = 0.0001616, Figure 1D). As major microbiota metabolites, SCFAs are involved in regulating intestinal integrity and intestinal immunity (32). Antibiotic treatment did not affect SCFA levels in maternal cecum, as no significant differences were observed regarding levels of acetic acid, propionic acid and butyric acid (Supplementary Figure S1A). Iso-butyric acid, valeric acid, and iso-valeric acid were not quantifiable in the cecal content of all sampled mice. To investigate a possible direct microbial effect of treatment on the prenatal environment, the 16S rRNA profiles of placental tissue were analyzed as well. Although some samples had reads as revealed by Qiime2 and DADA2 analysis, these represented mostly unspecific contamination, probably due to the high concentration of eukaryotic DNA in the samples. Therefore, based on our methods, we could not identify the presence of a specific bacterial community in the placentas analyzed. Pregnancy outcome was assessed as the number of pregnancies, intact and resorbed fetuses. No statistically significant differences were observed between groups. Eleven pregnant mice of the control group, and 8 of the antibiotic-treated group had a mean litter size of 7.8 (range 3-11, mean resorption rate 1) and 9.4 (range 3-11, mean resorption rate 0.8), respectively (Supplementary Figure S1B+C). No clinical symptoms in the antibiotic- treated group were observed that would suggest adverse effects on maternal health and as a consequence, pregnancy. Building on this intervention model, we then proceeded to analyze the consequences of antibiotic treatment on the maternal immune system. Antibiotic treatment associated with shift in immune parameters differentiating from control cohort Maternal immunity was assessed by flow cytometry of placenta, spleen, ILN, MLN and peritoneal cavity lavage fluid (PCLF), mRNA of intestines and placenta, and cytokine levels of maternal serum, amniotic fluid and splenocytes. A total of 129 different parameters were analyzed (Supplementary Table 2). To detect the classifying immune alterations occurring upon antibiotic treatment, a previously validated machine learning ensemble classification strategy was used (33, 34) suited for robust feature selection in the given low sample size setting. This method combines 8 classification algorithms, which compensates for possible biases inherent to the