Human disease-related pathways and ARGs
The human disease-related pathways and ARGs in samples from the SHIME model were assessed at different time points, which included samples prior to antibiotic treatment (C), sequential doses (10, 100, and 1000 mg L− 1) of antibiotic treatments at seven days interval designated as V10, V100, and V1000, followed by 14 days after termination of vancomycin (NR) and 14 days after FMT (FR). The hierarchy cluster heatmap showed that gene numbers of human disease-related pathways, including drug resistance, cancer, cardiovascular diseases, immune system diseases, infectious diseases, metabolic diseases, and neurosurgery diseases were more abundant in antibiotic exposure groups than that in control group, which was especially apparent in ascending colon (Fig. 1). For instance, gene numbers of bladder cancer, renal cell carcinoma, primary immunodeficiency, type II diabetes mellitus, and prion diseases were nearly 1.3–1.5 times more enriched in the V1000 group than in control. Moreover, gene numbers of these human disease-related pathways still maintained at a higher level (about 1.2–1.3 times enriched than control) after natural recovery and FMT treatment could restore them to baseline level.
The relative abundance of ARGs such as aminoglycoside, beta-lactam, multidrug, and tetracycline resistance genes were lower in vancomycin exposure groups than that in control samples, which was also especially obvious in ascending colon (Fig. 2). For instance, log relative abundance reduction of aac6ie (aminoglycoside) was about 2.9 log units after 1000 mg L− 1 vancomycin treatment than in control, and bl2b_tem1 (beta_lactam), ermb (MLSB), and qacedelta1 (multidrug) were about 2.0, 2.3, and 1.5 log units, respectively. Besides, the ARGs such as ant3ia (aminoglycoside), bl1_ec (beta_lactam), ermf (MLSB), mexf (multidrug), tetq (tetracycline), and sul1 (sulfonamide) were not detected after vancomycin treatment. Similarly, these ARGs were unable to return to the baseline level following the natural recovery with about 2.2 to 3.8 log units lower of relative abundance. FMT would restore most of them to some extent (about 1.7 log units lower to 0.6 log units higher).
Microbiota community composit ion and diversity
In this study, the effects of vancomycin on gut microbial communities’ composition were also investigated. Based on the 16S rRNA gene sequence analysis, the most abundant taxonomic groups assigned at the phylum level were Proteobacteria (64.7–98.7%), Bacteroidetes (0.1–18.7%) and Firmicutes (0.5–12.4%) present in the ascending colon (total of 92.5–99.4%), followed by Synergistetes, Fusobacteria, and Verrucornicrobia (Fig. S2). However, the most three abundant bacterial phyla in the descending colon (total of 79.8–94.8%) were Proteobacteria (53.6–83.4%), Bacteroidetes (0.1–23.0%) and Synergistetes (2.3–10.8%). Moreover, an obvious decrease of Bacteroidetes (from 17.7–18.7–0.1%) and Firmicutes (from 5.9–12.4% to 0.5–0.9%), and increase of Proteobacteria (from 68.3–68.9% to 83.4–98.7%) were seen after 1000 mg L− 1 vancomycin treatment. The significant changed communities’ composition at the phylum level was sustained after natural recovery and returned to baseline after FMT treatment. At the genus level, the antibiotic-treated subjects were shown to be substantially overgrown by Burkholderia (from 0.2–3.2% to 2.8–17.9%) and Achromobacter (from 0.3–0.4% to 6.3–8.9%) with the dose of vancomycin, while the percentage of Pseudomonas was increased in the V10 group (from 6.9–45.5% to 17.0-53.2%) and Klebsiella increased in the V1000 group (from 15.0-39.7% to 22.3–51.4%). Similarly, following the natural recovery period, the gut microbiota was incompletely restored, and the complete recovery observed after FMT. These results were in line with human disease-related pathways and ARGs results.
Meanwhile, fecal microbiota alpha diversity was assessed. The taxon richness (Chao1 index), evenness (Simpson index), and diversity (Shannon index) are shown in Fig. S3a to c. Compared with control sample C_A, a drop in microbiota richness (Chao1), evenness (Simpson), and diversity (Shannon) was observed in sample collected from ascending colon after 1000 mg L− 1 of vancomycin treatment; however opposite change was observed in sample V10_A. As shown in Fig. S3, the decreasing of microbiota richness, evenness, and diversity caused by vancomycin treatment restored after two weeks of vancomycin discontinued (NR). Besides, the beta diversity of the microbiota communities and weighted UniFrac distance was affected by antibiotic treatment. As shown in Fig. 3, the samples that collected after vancomycin treatments were differed from the control group in both UniFrac NMDS and PCoA analyses. The beta diversity results also showed that gut microbial composition remained comparable after two weeks of vancomycin discontinued (NR) because these two samples were clustered together with vancomycin treatment groups. Meanwile, after two weeks of FMT treatment, the samples clustered to control group for both UniFrac NMDS and PCoA analyses. Moreover, weighted UniFrac distance between 1000 mg L− 1 vancomycin treatment group and control group (V1000 vs. C) were higher than that within the control group (C vs. C), and that between the recovery groups and the control group (NR vs. C, FR vs. C) were slightly higher (Fig. S4).
Bloomed opportunistic pathogens
The linear discriminant analysis effect size (LEfSe) comparison analysis between the control and antibiotic groups was shown in Fig. 4. LEfSe analysis indicated that vancomycin resulted in significant decreases in several taxa, including the members of Bacteroidetes (Bacteroides, Parabacteroides), Firmicutes (Clostridium, Dialister) and Campylobacter. The changes were accompanied by increases in the relative abundance of Proteobacteria phylum, including Achromobacter (LSD = 4.61), Pseudomonas (LSD = 3.18), and Klebsiella (LSD = 3.34), which were consistent with the gut microbial communities’ composition results.
Moreover, five opportunistic pathogens were isolated from V1000_A sample (Table S2), which included Klebsiella aerogenes (NKU-Kae), Klebsiella pneumoniae (NKU-Kpn7), Klebsiella oxytoca (NKU-Kox6), Achromobacter xylosoxidans (NKU-Axy), and Pseudomonas aeruginosa (NKU-Pae). Table S3 represents the minimal inhibitory concentration (MIC) of these opportunistic pathogens. Data suggested that all the isolated opportunistic pathogens were resistant to vancomycin with MIC256 mg L− 1, as well as beta-lactam antibiotics including ampicillin (MIC64–256 mg L− 1) and amoxicillin (MIC64–256 mg L− 1), sulfonamides (sulfamethoxazole MIC128–256 mg L− 1) and MLSB (erythromycin MIC8–64 mg L− 1). However, Klebsiella oxytoca (NKU-Kox6), Achromobacter xylosoxidans (NKU-Axy) and Pseudomonas aeruginosa (NKU-Pae) showed the more comprehensive resistant properties to several kinds of antibiotics than Klebsiella aerogenes (NKU-Kae) and Klebsiella pneumoniae (NKU-Kpn7), including kanamycin, tetracycline, and streptomycin. Table S4 showed a total of nine ARGs detecting in these opportunistic pathogens. The results showed that all the opportunistic pathogens carried qnrs and bl2be_shv2, and except NKU-Kae, all others contained aac3iia, vang, arna, rosb, and ant2ia ARGs. Therefore, their intrinsic ARGs such as vang, aac3iia, sul1, and bl2be_shv2 might be attributed to high values of MICs to vancomycin, gentamicin, sulfamethoxazole, ampicillin, and amoxicillin. As shown in Table S5, a considerable variability in serum resistance for the opportunistic pathogens was observed, which constituted an important virulence trait that allowed these opportunistic pathogens to persist in vivo. And the survival of Pseudomonas aeruginosa (NKU-Pae) strain was significantly higher (12–54 times) than other strains with a virulence level of Grade 5.
Correlation between microbial taxa and human disease-related pathways or ARGs
Figure 5 showed the results of co-occurrence patterns between microbial taxa and human disease-related pathways. It can be seen that significantly increased bacteria after vancomycin exposure such as Achromobacter and Klebsiella were positively associated with almost all of those human disease-related pathways, and the decreased bacteria were negatively associated with these pathways. Specifically, the correlation coefficients of Achromobacter with colorectal cancer, viral myocarditis and toxoplasmosis were about 0.8 (P < 0.01) and the correlation coefficient of Klebsiella oxytoca with Staphylococcus aureus infection was 0.7 (P < 0.01).
The network analysis of co-occurrence patterns between microbial taxa and ARG subtypes was shown in Fig. 6. Interestingly, a very similar pattern of results was observed that a significantly decreased the bacteria was positively associated with the most ARGs. For example, the correlation coefficients of Bacteroides fragilis with bl2e_cepa aac6ie and tet32 were about 1.0 (P < 0.001), and the correlation coefficient of Parabacteroides with ermb was about 0.8 (P < 0.01). Only two significantly increased bacterial genera were positively associated with several ARGs. For example, the strong correlations were found in Klebsiella oxytoca with yidy/mdtl and tetc (about 0.9, P < 0.01), and in Pseudomonas with ermf and tetq (about 0.8, P < 0.05).