Longitudinal analysis of the rectal microbiota
All samples were sequenced to a depth > 10 Mio non-human paired reads. The average paired read number was 19 Mio (11-24 Mio). In total, 1144 microbial species were found in the rectal swabs. The majority of the microbial species are bacteria, only 2 species belong to the Eukaryota Kingdom (Candida albicans and Candida tropicalis), and 5 species are Archae, with only 2 members classified to the species level (Methanobrevibacter smithii & Methanosphaera stadtmanae). Only 52 species had an average abundance greater than 0.5% in the overall population, and the 18 most abundant species (relative abundance > 1%) found in the samples were Finegoldia magna (7.72%), Anaerococcus obesiensis (5.23%), Escherichia coli (5.20%), Enterococcus faecalis (2.80%), Fenollaria timonensis (2.71%), Peptoniphilus lacrimalis (2.39%), Peptoniphilus harei (1.95%), Enterococcus faecium (1.81%), Peptoniphilus grossensis (1.63%), Levyella massiliensis (1.60%), Ezakiella coagulans (1.28%), Latilactobacillus sakei (1.15%), Porphyromonas bennonis (1.13%), Akkermansia muciniphila (1.09%), Lactobacillus gasseri (1.05%), Prevotella buccalis (1.05%), Phocaeicola vulgatus (1.02%), and Anaerococcus prevotii (1.00%) (Figure 1A). Co-occurrence analysis revealed that 28 species co-occurred positively (Rho > 0.5 and p-value<0.01), and Akkermansia muciniphila was negatively correlated with the relative abundances of Finegoldia magna, Anaerococcus obesiensis and Peptoniphilus lacrimalis (Supplementary Figure 1). We did not observe a significant change in the α-diversity (Shannon index) between either the two time points (ANOVA with linear mixed effect model with patient ID as a random effect: p-value=0.38, Figure 1B) or the carriage of an MDRO (Wilcoxon test, p-value=0.27). We observed a high β-diversity (Morisita-Horn index) within patients between the time points, and the acquisition or loss of an MDRO seemed to increase the β-diversity without reaching significance (Wilcoxon test, p-value=0.22). We also observed a slight impact of the time difference between the time points on the β-diversity, but the difference did not reach significance (Supplementary Figure 2).
Culture of MDROs
From the 26 patients included, 6 MDROs from 6 patients were isolated four 3GCREB strains (Escherichia coli (n=2, P11 & P28), Klebsiella aerogenes (n=1, P20), and Klebsiella pneumoniae (n=1, P24)) and two vancomycin-resistant Enterococcus faecium strains (P04 & P10). Three patients acquired an MDRO during the study period (P10; 2-year interval, P20 and P24; 1-year interval), and three patients lost the MDRO during the study period (P11; 2-year interval, P04 and P28; 1-year interval). None of the patients were colonized at the two sampling time points.
Longitudinal analysis of the resistome
In total, 71 genes associated with resistance to cephalosporin, methicillin or vancomycin were found in the population, and most of them were prevalent in samples where no MDRO was isolated. The most prevalent beta-lactamase genes are associated with Bacteroides species such as cfxA-like and cfxA (n=32, 55%; n=14, 24%), cblA (n=20, 34%) and cepA (n=14, 24%), as well as genes involved in beta-lactamase PC1 (blaZ) expression (blaZ: n=11, 19%; blaR: n=12, 21%; blaI: n=19, 33%), which are often associated with gram-positive bacteria such as Bacillus, Enterococcus and Staphylococcus species. The genes from the blaEC family associated mostly with E. coli and Shigella species were also highly prevalent (blaEC-5: n=13,22%; blaEC-15: n=7,12%; blaEC-8: n=6,10%; blaEC-18: n=5,9%; blaEC-19: n=3,5%). Finally, several beta-lactamase genes that are transferable among Enterobacteriaceae were also detected (blaTEM-1: n=13, 22%; blaTEM-150: n=1, 2%; blaTEM-2: n=1, 2%; blaTEM-57: n=1, 2%; blaCTX-M-1: n=3, 5%; blaCTX-M-14: n=2, 3%; blaSHV-1: n=3, 5%; blaSHV-187: n=1, 2%). The mec cassette was found at least partially in 6 samples (10%), the van cassette was found in 37 samples (71%), and the operons vanB and vanD were the most prevalent (Figure 2A).
Interestingly, the ESBL genes or van cassette from the MDROs isolated were found in every corresponding MDRO-carrying metagenome, with a certain limitation for the van cassette, which was found only partially in patient 10. However, none of the genes were found in the second metagenomes of those patients, indicating that the genetic determinant of resistance was acquired at the time of sampling and was not detectable before or after the positive culture of the isolate (Figure 2A). We did not observe any impact on the acquisition of culturable MDROs or the amount of AMR genes (Resistome: p-value=0.7787), but we observed a decrease in the amount of typable plasmid at timepoint T2 (Inc-type: p-value = 0.0089). We also observed a positive correlation between the resistome and the amount of typable plasmid, indicating that the majority of the AMR genes were carried on plasmids (Pearson correlation: R²=0.47, p-value < 0.001). We did not observe a significant impact of sex, age or antibiotic usage at the time of sampling or in the last 6 months on the resistome.
Strain level comparison
Of the 52 metagenomes, 45 contained E. coli, but only 34 samples had enough markers in the metagenome to perform strain analysis for comparison to the isolated strains. We observed good concordance between the strains isolated from P11 at timepoint 1 and the corresponding strain in the metagenome (0 mutations over 42190 sites), while the strain found at timepoint 2 was more phylogenetically distant (> 1000 mutations). We also observed full concordance of the AMR genes between the isolated strain and the corresponding metagenome (Rel. abundance: 4.2%). For Patient 28, we observed a different pattern in which the isolated strain was closer to timepoint 2 (73 mutations/41702 sites) than to timepoint 1 (144 mutations/41230 sites) when it was originally isolated. We also observed good concordance between the antimicrobial resistance genes of the isolated strains and the corresponding metagenome (Rel. abundance: 1.3%), where only two genes (blaEC-8 and blaTEM-2) were missing from the metagenome. Only one patient had closely related E. coli strains (P30: 2/41601 mutations) between the time points. These two strains are the only ones closely related to a level corresponding to the previously published mutation rate of 6.5*10^-7 [20]. We also detected closely related strains between P18 and P17 at timepoint 1 (11/42972 mutations) (Figure 3).
We found E. faecium in 12 metagenomes with a relative abundance ranging from 0.001% to 32.9%, but we never found E. faecium in two consecutive timepoints within a patient except for P24. We observed a close genetic relationship between the P4-T1 and P10-T2 strains from the metagenome (11/256599), with a mutation rate of 4.3 * 10-5, which is related to the 25 published mutations per year per genome [21]; however, surprisingly, the strains from the metagenome to their isolated counterparts were phylogenetically distant (P4: 126/262138; P10: 716/256987). We identified the van cassette in every metagenome where a VRE was isolated as well as in samples P24-T1 and P24-T2, where the relative abundance of E. faecium was relatively high (13.9% and 16.9%, respectively) (Figure 4). No growth of vancomycin-resistant organisms was observed at T1, but we did observe the growth of E. gallinarum on selective agar at T2. The relative abundance of E. gallinarum in the metagenome was 0% at time T1 and 0.008% at time T2.
We detected K. aerogenes in 4 metagenomes, but the relative abundance was less than 0.1% in all the metagenomes except P20-T2 (11.1%), which is the sample with a positive culture. Due to the low prevalence, we could not perform a strain analysis. However, we observed good concordance between the AMR profile of the P20-T2 metagenome and that of the isolated strain, and P20-T1 did not contain K. aerogenes (Suppl. Figure 2A). We observed the same pattern as for K. pneumoniae; only 8 metagenomes were positive (rel abundance ranging from 0.006 to 3.3%), and the sample with a culture positive for MDR K. pneumoniae P24-T2 had one of the highest abundances (1.0%), but we did not find K. pneumoniae in P24-T1. The AMR profiles of the isolates from P24-T2 were fully covered within the metagenome P24-T2 (Suppl. Figure 2B).