Study population
We enrolled 144 of the 312 residents of the two nursing home. Ten (6.9%) of them were confirmed to carry ESBL-E. Microbiological analysis identified seven Escherichia coli, two Enterobacter cloacae and one Klebsiella pneumoniae. The relative abundance of these species was not significantly different between carriers and non-carriers (Mann-Whitney U test, ESBL-E carriers versus non-carriers for E. coli, E. cloacae and K. pneumoniae relative abundances, all p values > 0.05) (Supplementary figure S1). There were no significant differences in baseline characteristics between carriers and non-carrier subjects, including antibiotic exposure (Table 1). Prevalence of ESBL-E carriage were not different between the two nursing homes (Fisher’s exact test, p value > 0,05). Among ESBL-E carriers, four were exposed to broad spectrum penicillin during 7 days. As all residents were housed in two nursing homes from the same University hospital, diet, hygiene and care was performed to a standardized practice by the same medical and nursing staff and meals were prepared in the same facility.
Diversity of the gut microbiome is lower in ESBL-E carriers.
We obtained the mean 308,696 sequences per samples, providing enough depth to reach species-level (16).
Comparison of within-samples species diversity (alpha-diversity), using Chao1 and unique observed species indexes, showed a less diverse gut microbiome in ESBL-E carriers (Mann-Whitney U test, ESBL-E carriers versus non-carriers, Chao1 and unique observed species, p value < 0.001 and p value < 0.01, respectively, Fig. 1). Principal-coordinates analysis (PCoA) of Bray Curtis distances revealed that carriers and non-carriers also harbored a slight but significantly distinct gut microbial compositions. The R value, close to 0 suggests the existence of an important overlap area between the 2 groups (analysis of similarities, ANOSIM, R=0.177, p value=0.048, Supplementary figure S2). Carriers and non-carriers were not clustered according to their functional architecture (ANOSIM, R=0.1, p value=0.15), but we observed a trend in increased functional biodiversity in carriers (2 129 865 sequences/samples, Mann-Whitney U test, ESBL-E carriers versus non-carriers, Chao1 and unique observed species, p value < 0.01 and p value < 0.175, respectively, Supplementary figure S3). Importantly, we did not find any significant association between other covariables and interindividual microbial distance, including sex, age, length of stay, comorbidities and medications.
ESBL-E carriers gut microbiome is enriched in Bacteroides and Prevotella.
To identify a taxonomic signature associated with ESBL-E carriage, we compared bacterial communities of carriers and non-carriers at genus and species levels using DESeq2 with bootstrap iterations. The great majority of genera overrepresented in the gut microbiome of ESBL-E carriers mapped to Bacteroidales spp and Clostridiales spp orders (respectively 4 and 2 of 9). In particular, samples from carriers showed a more than 2 mean log2fold higher proportion of Megasphaera spp, Alloprevotella spp (FDR corrected p value < 0.10, in 70% or more bootstrapping iterations). Less frequently, Prevotella spp and Bacteroides spp appeared increased in ESBL-E carriers (FDR corrected p value < 0.10, in 60% or more bootstrapping iterations). The representation of Pseudomonas spp and Johnsonella spp genera appeared strongly decreased with a mean log2fold change below -4 (FDR corrected p value < 0.10, in 70% and 60% or more bootstrapping iterations respectively) (Fig. 2A, Supplementary table S1).
More precisely, bacterial species composing the community in ESBL-E carriers were significantly enriched in 8 species. Bifidobacterium animalis showed a strongly increased abundance of 6 mean log2fold change and three Prevotella species were increased over 2 mean log2fold change (P. corporis, P. timonensis and P sp. P4-76). Species belonging to Bacteroides spp genera represented a large proportion of the overrepresented individual bacteria in these residents (B. faecichinchillae, B. coprophilus, B. ovatus). The lower proportion of Johnsonella spp genera in ESBL-E carriers was mostly represented by a lower abundance of J. ignava and Pseudomonas spp genera by Pseudomonas aeruginosa (FDR corrected p value < 0.10, in at least 70% and 60% or more bootstrapping iterations respectively). Three individual bacterial species showed a strongly decreased abundance below -3 mean log2fold change, namely Clostridium hylemona and Collinsella tanakaei (FDR corrected p value < 0.10, in at least 70% or more bootstrapping iterations) and less frequently Bifidobacterium adolescentis (FDR corrected p value < 0.10, in at least 60% or more bootstrapping iterations) (Fig. 2B, Supplementary table S2).
As many parameters may influence the composition of the gut microbiome, we used the MaAslin pipeline to test for association between baseline characteristics and taxonomical composition of the gut microbiome of all residents. We did not find any significant association between age, sex, length of stay, medications, comorbidities and taxonomical composition of the gut microbiome.
The gut microbiome of ESBL-E carriers is enriched in carbohydrate metabolism pathway genes.
The analysis of functional modules harbored by the gut microbiome showed robust differences in relative abundance of 8 KEGG modules between ESBL-E carriers and non-carriers (FDR corrected p value < 0.10, in 70% or more bootstrapping iterations) (Fig. 3, Supplementary table S3). All functional pathways identified were overrepresented in ESBL-E carriers and were involved in carbohydrate metabolism, amino-acid and branched-chain amino acid metabolism and environmental information processing. Among them, KEGG modules involved in cysteine biosynthesis (M00338), glutamate transport (M00233), sulfonate transport (M00436), N-acetylglucosamine biosynthesis (M00267 and M00019) and isoleucine biosynthesis (M00570) were the most frequently significant. We identified a total of 50 KEGG orthologies (KOs) with modified relative abundance between carriers and non-carriers, which were all overrepresented in ESBL-E carriers gut microbiome (FDR corrected p value < 0.10 in 70% or more bootstrapping iterations) (Supplementary figure S4 and supplementary table S4). Finally, to further investigate carbohydrate metabolism, we searched for a signature in CAZymes composition. Nine CAZymes had a modified relative abundance between carriers and non-carriers, all overrepresented in ESBL-E carriers gut microbiome (FDR corrected p value < 0.10 in 70% or more bootstrapping iterations) (Supplementary figure and supplementary table S5).