Marilen Benner

GESTATIONAL ANTIBIOTICS AFFECT IMMUNITY 191 7 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Relative abundance (%) Cecal microbiota - Genus level - Akkermansia Escherichia-Shigella Candidatus_Saccharimonas Romboutsia Anaerovoracaceae;Family_XIII_UCG-001 Peptococcaceae;uncultured Peptococcus Oscillospirales;UCG-010; UCG-010 Ruminococcaceae;uncultured Ruminococcaceae;UBA1819 Ruminococcus Incertae_Sedis Harryflintia Anaerotruncus Ruminococcace Oscillospiraceae; uncultured Oscillospiraceae;UCG-005 Oscillospiraceae; UCG-003 Oscillibacter Oscillospiraceae; NK4A214_group Intestinimonas Colidextribacter Oscillospiraceae Butyricicoccaceae;UCG-009 Butyricicoccus Lachnospiraceae;uncultured [Eubacterium]_xylanophilum_group Tyzzerella Tuzzerella Roseburia Lachnospiraceae_UCG-006 Lachnospiraceae_UCG-001 Lachnospiraceae_NK4A136_group Lachnospiraceae_FCS020_group Lachnoclostridium Lachnospiraceae;g_GCA-900066575 Blautia Anaerostipes Acetatifactor Lachnospiraceae;g_A2 Lachnospiraceae Clostridium_sensu_stricto_1 Candidatus_Arthromitus Clostridia_vadinBB60_group Clostridia_UCG-014 Clostridia Lactobacillus Enterococcus Erysipelotrichaceae; uncultured Turicibacter Faecalibaculum Dubosiella Anaeroplasma Desulfovibrio Bilophila Mucispirillum Helicobacter Parabacteroides Tannerellaceae Rikenellaceae_RC9_gut_group Rikenella Alistipes Alloprevotella Muribaculaceae Odoribacter Bacteroides Enterorhabdus Bacteria Control Control Control Shannon Index Antibiotics Antibiotics Antibiotics PC2 (82%) PC1 (69%) 1 2 3 4 5 A B C D x Day -3 0 4 17 13 AB AB Figure 1. Antibiotic intervention in pregnant Balb/c mice. (A) Schematic representation of the experimental model. After randomly assigning mated mice to the treatment groups, 11 mice of the control and 8 mice of the treatment cohort were pregnant and included for further analysis. (B) 16S rRNA composition of the cecal samples per individual animal (1 sample of the antibiotic treatment cohort excluded due to read count <500). (C) Principal component analysis of 16S rRNA sequencing data on genus level. (D) Shannon Index presenting microbial diversity. individual algorithm. The output of the 8 classification algorithms each yielded a ranked list of features, representing the contribution of the 129 immune features to cohort classification. Each list was weighted based on coefficients and frequency of an individual feature contributing to classification (34) to aggregate classifiers to a single ranking. Classification algorithms were repeatedly run using the top 80% of the ensemble ranking (recursive feature selection). Each classification run was carried out 10 times, each run being subjected to 10-fold cross validation. At a global average accuracy of 90% as cutoff to ensure robust classification (Figure 2A), ensemble accuracy for each iteration of feature combinations showed optimal classification