Fig. 4

Bacterial community structure in time. a) Persistence and abundance of ASVs detected. Persistence was expressed as the number of subjects in which an ASV was detected whereas abundance was expressed as log-normalized number of reads assigned to that ASV (the black line represent results of the linear fitting: 95% CI [9.81, 10.38], t(1926) = 70.27, p < 0.001). Cluster 1 (reported in red) was composed of ASVs with a lower persistence and abundance than Cluster 2 (reported in blue). b) Number of edges formed (green) and destroyed (red) at each time point. Lines represent the result of two linear mixed models with subject as random intercept (Table S9). The dashed line represents the effect at the end of the isolation—it was reported only for formed edges since the number of destroyed edges was not significantly impacted by the isolation. c) Community networks at the beginning of the mission (day 1), during the simulated trip to Earth (day 302), and at the end of the mission (day 716). Selected time points were highlighted in panels d and e using a black circle. Days are reported at the top of each network whereas nodes with no edges (namely ASVs not detected at a given time point) were not reported to save space for graphical representation. d) Node-level centrality score during the mission and the follow-up. Each point represents the mean centrality score at a given sampling time whereas the two dotted lines represent the standard error on the mean. Amplicon sequence variants of cluster 1 reported a lower level of centrality in respect with those of cluster 2 at all time points (Wilcoxon rank sum test: all p-values < 0.01). e) Average abundance of ASVs assigned to clusters 1 and 2. Points and dotted lines report the mean and its error as described in panel d. At each time point ASVs of cluster 2 showed a higher abundance than those in cluster 1 (Wilcoxon rank sum test: all p-values < 0.01)