a Principal component analysis of metabolite composition assessed by untargeted metabolomics in supernatants collected at 6 and 12 hpi. Arrows (I, II, III) indicate shifts in the metabolic profile. b Differentially increased or decreased metabolites in comparison of the different conditions shown in proportional Venn diagrams. Data summarized from n = 5 at 6 and 12 hpi. c Hierarchical clustering based on Euclidean distance of relative metabolite abundance in the supernatants at 6 hpi. d Growth of L. rhamnosus in KBM, on IECs, in transwells physically separated from IECs, or in supernatants of IECs (spent) assessed by counting CFUs on MRS agar. Data are shown as the mean and standard deviation (SD) of n = 3 biological replicates. e In silico prediction of L. rhamnosus biomass formation in KBM or supernatants of IECs. f Growth of L. rhamnosus assessed by CFUs after 48 h incubation in KBM supplemented with single metabolites or combinations of metabolites. Data of n = 3–6, KBM = 17, IEC spend n = 8 biological replicates are shown as the mean and SD. Data were tested for significance using a t test (two tailed, one-sample) against growth in KBM, * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001, **** = p ≤ 0.0001 (CA + GGA + CARN n = 6, * = p 0.0410; CA + CARN n = 5, **** = p < 0.0001; CA + GGA n = 3, *** = p 0.0003; CA + OB n = 3, ** = p 0.0041; CA n = 5, *** = p 0.0003; CA + OV n = 3, ** = p 0.0089, CARN n = 4, * = p 0.0218; GGA + CARN n = 3, *** = p 0.0001, CA + NAG n = 3, **** = p < 0.0001). Source data are provided as a Source Data file. g Comparison of metabolic pathway activity levels between different conditions as indicated. Relative pathway change was determined by identifying the number of pathway-specific reactions for which feasible flux ranges differ according to flux variability analysis.