Amoxicillin exposure significantly increases the population and virulence of antibiotic-resistant Klebsiella pathogen in the simulated human intestinal microbiota

Background: The extensive use and misuse of antibiotics have caused the emergence and spread of antibiotic resistance among bacterial pathogens, which is threatening to this idyllic world. The human intestinal microbiota is significant to health, while its composition and stability are easily perturbed by antibiotic exposure. However, the association between antibiotic exposure and the abnormal bloom of opportunistic pathogens in human gut, and its health implications are not fully understood. Therefore, this study investigated the influence of amoxicillin (AMX), one of the most prescribed antibiotics in human life, on human intestinal microbiota using the simulator of the human intestinal microbial ecosystem (SHIME). Results: Our results suggest that AMX exposure could cause the substantial effect on human intestinal microbiota in different colon regions, with most significant alteration in ascending colon. Intriguingly, AMX significantly increased the abundance of antibiotic-resistant Klebsiella pathogen in gut microbiota, resulting in the enrichment of human disease-related pathways. By isolation of Klebsiella strains from SHIME before and after exposure to AMX, we also observed that the virulence phenotypes of these pathogens including biofilm formation, serum resistance, and Galleria mellonella infection were significantly increased after AMX administration. The genotype analysis from whole genome sequencing indicated that the enhanced virulence phenotypes were plausibly caused by the mutations of virulence factors relevant to fimbriae, lipopolysaccharide, capsule, and siderophores. More importantly, the effects from AMX administration could not be fully recovered after two-week drug discontinuance, showing the potential long-term adverse influence of this antibiotic. Conclusions: This study for the first time demonstrates that AMX used for the treatment of bacterial infections may cause adverse effects on human health via significantly increasing population and virulence of intestinal Klebsiella opportunistic pathogen. Given that Moreover, to the virulence of Klebsiella including biofilm formation, serum resistance, and G. mellonella infection. Genome sequence comparison analysis of strain NKU_Kleb8 (isolated from SHIME before AMX exposure) with the strain NKU_KlebA1 (isolated from SHIME after AMX exposure) and NKU_Kleb8A7 (evolved from NKU_Kleb8 after AMX exposure) revealed that the increased virulence may be attributed to the mutations of four major virulence factors relevant to fimbriae, LPS, capsule, and siderophores. Our results may open up new perspectives for assessing the direct effects of AMX on the intestinal microbiota—a key “organ” in individual health. Further analysis of the increased and evolved Klebsiella in vivo should be conducted for better understanding the effects of AMX on human health. phylogenetic tree establishment, Results are expressed as mean values and standard deviations. The statistical analysis was performed with SPSS 17.0 software. The T-tests were conducted to compare the difference between groups and all statistical tests were two-tailed. Statistical significance was set at three different levels (* p < 0.05, ** p < 0.01, and *** p < 0.001). Spearman test for correlation analysis between gene numbers of human disease-related pathways and the Klebsiella abundance were performed in R with the vegan package.

inherent resistance to β-lactam already poses a significant therapeutic challenge for Klebsiella infections, the AMX-induced virulence increasing phenomenon found in this study undoubtedly further increases its health risk.

Background
The human intestine constitutes an especially large and diverse microbial habitat that is colonized mainly by bacteria but also by archaea, viruses, and eukaryotes [1]. Previous studies have manifested that the human intestinal microbiota comprises about 150 times more genes than the human genome, making it become 'another' genome of human beings [2,3]. Moreover, the human intestinal microbiota has been demonstrated to provide numerous important functions for human health including nutrient supply by fermentation of indigestible dietary polysaccharides, synthesis of essential amino acids and vitamins, modulating the immune function, protection from infection of pathogens, etc. [4,5].
Antibiotic therapy has saved millions of human lives since its discovery in the 1940s, however indiscriminately killing or preventing the growth of both pathogenic and commensal bacteria is its major drawback. In recent years, the adverse influence of antibiotics on human health has become a public concern since it could cause the dysbiosis of human intestinal microbiota, resulting multiple human diseases [6,7].
Otherwise, antibiotics may also promote the evolution, proliferation, and maintenance of antibiotic resistant bacteria (ARB) and antibiotic resistance genes (ARGs) in the human gut [8,9]. In general, both in vivo and in vitro experiments were widely used in examination of antibiotic influence on gut microbiota [10][11][12][13][14][15]. In vivo experiments would reflect the actual influences of antibiotics on human body, while the detected fecal samples are generally stood for the distal intestinal microbiota that could not reveal the impacts of antibiotics on different gut regions [16,17]. Instead, in vitro simulator with different gut regions can reproduce human gut environment in the absence of host cells, thus allow to examine the effects of antibiotics on different regions of human intestinal microbiota without the disturbances of neurohumoral regulation, the individual differences, dietary habits, and physiological status. However, most classic in vitro models such as single-stage chemostats and semi-continuous fermenters could neither integrate the entire gastrointestinal tract nor maintain microbiome stability over an extended timeframe, and the results obtained generally need the confirmation of in vivo experiments [18]. As a widely used multi-compartment simulator, the simulator of the human intestinal microbial ecosystem (SHIME) has been used to study microbiota interactions in gut and the effects of toxic compounds on gut microbiota [19]. To date using SHIME model to study the influences of antibiotics on human gut microbiota is still rare.
Amoxicillin (AMX), an inexpensive but broad-spectrum oral penicillin-type β-lactam antibiotic, is one of the most prescribed medicine in human life. It kills bacteria by interfering with the synthesis of the bacterial cell wall peptidoglycan [20,21]. Over the last several years, extensive studies have provided valuable insights into the effects of AMX on human intestinal microbiota [22][23][24][25]. However, the real impacts caused by AMX on human gut flora is still controversy. In addition, the association between AMX exposure and the abnormal bloom of opportunistic pathogens in human gut, and associated health implications are not fully understood. Therefore, this study investigates the influence of AMX on human intestinal microbiota in SHIME model. Considering the reasonable dosage of AMX for adult human study is about 750 to 1500 mg/day and only half volume of adult gut can be simulated in used SHIME model, 600 mg/day of AMX was used to examine its impacts on human gut microbiota and associated functional pathways [22,24]. Moreover, we isolated Klebsiella strains from SHIME before and after exposure to AMX, and detected their virulence phenotypes including biofilm formation, serum resistance, and Galleria mellonella infection. Our data suggest that AMX exposure caused an adverse impact on the gut microbiota with significantly increased antibiotic-resistant Klebsiella and human disease-related pathways. Furthermore, exposure to AMX significantly increased the virulence of antibiotic-resistant Klebsiella, which may be attributed to the mutations of fimbriae, lipopolysaccharide, capsule, and siderophores virulence factors. To the best of our knowledge, this study for the first time demonstrates that AMX used for the treatment of bacterial infections may cause adverse effects on human health via significantly increased population and virulence of intestinal Klebsiella opportunistic pathogen.

AMX exposure significantly increased Klebsiella abundance in simulated gut microbiota
To reveal the effects of AMX treatment on the gut bacterial community, AMX was administrated into SHIME model in Fig. 1A. Before the exposure, the system was initially operated in the absence of AMX for three weeks, then 600 mg/day of AMX was injected into the system for another week, followed by the last two weeks of operation after the drug discontinuance. Fecal samples at 14, 21, 24, 28, 35 and 42 days from the ascending, transverse and descending colon vessels were collected for 16S rRNA gene amplicon sequencing. Specifically, these samples represent samples before AMX exposure, after 600 mg/day of AMX treatment, and after termination of AMX (Fig. 1A). The bacteria compositions at the genus and phylum levels at different sampling points were shown in Fig. 1B and Fig. S1, respectively. In terms of the taxonomic assignment at the phylum level, Bacteroidetes, Proteobacteria, and Firmicutes were the three most dominant bacterial phyla in both ascending (accounting for total percent of 91.7%-98.0%) and transverse colon (77.5%-88.8%), while three most abundant phyla in the descending colon were Proteobacteria, Bacteroidetes, and Synergistetes (81.8%-91.1%) (Fig. S1). Regarding genus level, Bacteroides, Klebsiella, Pseudomonas, and Pyramidobacter were predominant in all three colon regions (Fig. 1B). Linear discriminant analysis effect size (LEfSe) results indicated that AMX exposure caused obvious decreases in Bifidobacterium, Phascolarctobacterium and Parabacteroides (Fig. 1C), in accompanying with significant increases in Klebsiella (opportunistic pathogen, LSD = 5.26) and Bacteroides uniformis (probiotics, LSD = 4.75). As Klebsiella genus is an important antibiotic-resistant pathogen in human gut and its unnormal bloom is often associated with multiple diseases, thus we put more attentions on the increased Klebsiella abundance in this study (Fig. 1D).
Furthermore, the significant (p < 0.05) enrichment of Klebsiella was observed for both ascending (increased from 1.6% to 32.9%) and transverse colons (increased from 1.5% to 14.0%). Although the abundance of Klebsiella exhibited a partially reduction after the discontinuance of AMX for 2 weeks, the Klebsiella abundance was still obviously higher compared to AMX-free controls (Fig. 1D). For descending colon, a similar higher Klebsiella abundance was also observed although no statistically significant difference between AMX-exposed treatments and AMX-free controls.

Positive correlations between gene numbers of human disease-related pathways and the Klebsiella abundance
The functional composition of bacterial communities was predicted by PICRUST, gene numbers of human disease-related function pathways in bacterial communities in three regions of the colon at different sampling time were obtained in the heatmap (Fig. 2). The results suggested that with the time increased, gene numbers of human disease-related pathways including cancer, drug resistance, infectious diseases, metabolic diseases and neurosurgery diseases exhibited firstly increased then marginally decreased. Regardless of the colon region, AMX exposure significantly enhanced gene numbers of the above pathways, with the greatest enhancement effect in the ascending colon. For example, compared to C-A2 (AMX-free treatment sample) that collected from the ascending colon before AMX administration, gene numbers of the drug resistance pathways were 3. 4-4.4 times enriched in the AMX-A2 (AMX exposed treatment sample) that collected after AMX administration for seven days. Similarly, gene numbers of other pathways including bladder cancer, African trypanosomiasis, pertussis and Alzheimer's disease were nearly 4-5.6 times enriched in the AMX-A2 sample than C-A2. Moreover, these human diseaserelated pathway gene numbers were still maintained at relatively higher levels (about 1.7-2.4 times enriched than AMX-free controls) after the discontinuance of AMX for two weeks.
To explore the potential factors responsible for the changed gene numbers of human disease-related pathways during the incubation period, pearson correlation analyses were performed between gene numbers of pathways and the abundance of Klebsiella. As shown

Molecular mechanisms for Klebsiella's contribution to human diseases
In order to further explore the potential molecular mechanisms for the contribution of Klebsiella to human diseases, we isolated Klebsiella strains from AMX-exposed sample AMX-A2 (the most significant effect on human diseases happened in this sample). Using a Klebsiella-specific culture medium, a total of 8 Klebsiella strains were isolated, within which NKU_KlebA1, NKU_KlebA2, NKU_KlebA4, and NKU_KlebA5 possessed all 5 detected ARGs related to drug resistance pathways (bl2b_tem1 and bl2be_shv2 related to β-lactam resistance; arna and rosb related to CAMP resistance; and vang related to vancomycin resistance) (Table S1). Considering NKU_KlebA1 from AMX-exposed treatment has a close genetic distance with the Klebsiella strains from AMX-free control in the 16S rRNA genebased phylogenetic tree (Fig. 4D)

AMX exposure significantly increased the virulence of Klebsiella and its molecular mechanisms
To investigate the AMX's influence on virulence of Klebsiella, 8 Klebsiella strains were also isolated from AMX-free control sample C-A2 (i.e., before AMX exposure) for comparison.
Virulence phenotypes including biofilm formation, serum resistance and animal toxicology of all the 16 Klebsiella isolates were evaluated by crystal violet assay, serum bactericidal assay, and G. mellonella infection assay, respectively. The results demonstrated that Klebsiella strains isolated from AMX exposed sample had stronger biofilm formation (p < 0.001, Fig. 4A), serum resistance (p< 0.01, Fig. 4B), and G. mellonella inhibition capacities (p < 0.05, Fig. 4C) than those from AMX-free sample, suggesting AMX exposure dramatically increased the virulence of Klebsiella .
The phylogenetic tree of all isolated Klebsiella species based on 16S rRNA gene identification revealed that the Klebsiella NKU_Kleb8 (isolated from AMX-free sample C-A2) and NKU_KlebA1 (isolated from AMX exposed sample AMX-A2) were within a close genetic distance (Fig. 4D). Further, the whole-genome analysis of strains NKU_KlebA1 and NKU_Kleb8 also implied that these two strains shared only 8 single nucleotide polymorphisms (SNPs) and 471 insertion-deletions (InDels) (as detailed in supplemental Excel S3A). Furthermore, multilocus sequence typing (MLST) confirmed these two strains both belong to ST22, and average nucleotide identity (ANI) analysis demonstrated the nucleotide identity of these two strains were 99.97% (Fig. S2). Therefore, these results collectively suggested that NKU_KlebA1 might be the evolved strain from NKU_Kleb8 exposed to AMX. Using methods described above, we also found NKU_KlebA1 had a substantially higher level of virulence than NKU_Kleb8 in the term of biofilm formation, serum resistance and G. mellonella infection ( Fig. 4E-G).
To further confirm that AMX exposure could significantly increase the virulence of Klebsiella, a strain named NKU_Kleb8A7 was evolved from strain NKU_Kleb8 exposure to 600 mg/day AMX for seven days, and results of virulence shown in Fig. 4E-G also manifested a significantly higher virulence of NKU_Kleb8A7 than NKU_Kleb8 (p < 0.01). We  Table 1. In detail, capsule virulence factor cpsI appeared InDel sites in NKU_KlebA1, cpsO and wcbB were new emerged virulence factors in NKU_Kleb8A7, and all the others appeared SNP sites in NKU_Kleb8A7. For fimbriae virulence factors, lpfC, pilB , and pilG had genomic mutations in both evolved strains, and most virulence factors appeared InDel sites in NKU_KlebA1 but appeared SNP sites in NKU_Kleb8A7. LPS virulence factor gtrB mutated in both strains, and all the others were new emerged genes or appeared SNP sites in NKU_Kleb8A7. Mutations in fepA, EntD and iutA were also observed in both strains, and all the other siderophores virulence factors were new emerged genes or appeared SNP sites in NKU_Kleb8A7. Owing to their importance in protein function and disease causing, the potential unintended consequence of these mutations needs a special concern.

Enrichment of Klebsiella and human disease-related pathways in intestinal microbiota after AMX exposure
Using designed SHIME model, the impact of AMX on the human intestinal microbiota was investigated by high-throughput sequencing approach in this study. Bacteriodetes, Proteobacteria, Synergistetes, and Firmicutes were identified as four predominant phyla in simulated gut microbiota. Similar bacterial communities have been reported in some fecal samples and SHIME models in previous studies [26][27][28]. In addition, the inconsistent bacterial phyla were identified in three different colon regions, providing further evidence that colon regions play important roles in shaping the structure of gut microbiota [17].
The observed enrichment of bacteria in Klebsiella after exposure to AMX was consistent with previous findings that β-lactam antibiotics, vancomycin, and sulfonamide antibiotics exposure could increase the Klebsiella population [10][11][12][13]. To our best knowledge, there is currently just one study noticed that AMX increased the abundance of Klebsiella in human intestinal microbiota, while unfortunately no further mechanistic explanation and accompanying health consequences were provided [23]. Although Klebsiella genus is commensal organisms in the human gut, fecal carriage of extended-spectrum β-lactamaseproducing K. pneumoniae had been found as a significant risk factor for infection as early as 20 years ago [29]. Moreover, increased researches have also confirmed that human gut was a major reservoir for Klebsiella, and the bloom of those intestinal Klebsiella may cause the various infections in different body parts of human-being [30,31]. For example, pneumonia, Crohn's disease, colitis, cystitis, wound infections, and liver abscess have been found to relate to the bloom of Klebsiella in the human gut [11,32]. Antibioticresistant Klebsiella isolated in this study belong to sequence type 22 (ST22), a rare sequence type described in two previous studies [33,34]. Above studies confirmed Klebsiella (ST22) could cause pneumonia and display multidrug resistance.
As with abundance of Klebsiella, human disease-related pathways in different colon regions exhibited similar change trends as the function of time. More importantly, significant positive correlations between human disease-related pathways and Klebsiella suggested that the Klebsiella bloom caused by AMX might contribute to the increase of these pathways. It is known that frequent use of AMX in medicine may cause the development and spread of ARB and ARGs in human gut microbiome [25,35,36]. As expected, β-lactam resistance pathways increased after β-lactam antibiotic AMX exposure and its nice positive correlation with Klebsiella abundance was also consistent with the fact that Klebsiella spp are intrinsic resistance to β-lactam antibiotics via producing βlactamase ( Fig. 2) [37]. Unexpected increases of vancomycin and CAMP resistance pathways with the increasing of Klebsiella indicated the potential co-selection on different antibiotic resistance from AMX [38]. In the present study, human disease-related pathways like cancer, infectious diseases, metabolic diseases, and neurosurgery diseases were all strongly associated with the Klebsiella abundance, which is also supported by some previous studies [39][40][41][42][43]. Taken cancer and infectious disease as examples, Klebsiella spp.
has been considered as one of the most concerning pathogens frequently involved in infections of cancer patients and led to high mortality [40,44]. Infectious diseases such as urinary tract infection, pyelonephritis, urosepsis, wound infection, cytomegalovirus disease, and pneumonia had been extensively reported to associate with the Klebsiella [11,32]. KEGG pathways with human disease annotation by whole-genome analysis shown in this paper confirmed Klebsiella's plausible contribution to human diseases, and many researchers have also revealed that Klebsiella possesses intrinsic virulence factor, antibiotic resistance, pathogen host interactions and secretion system, which determine their negative effects on human health [45][46][47].

AMX effects were colon region-different and persistent
In this study, Klebsiella was found to be more significantly enriched in ascending colon than the other two colon-regions (Fig. 1), which was in keeping with previous reports for the changes of intestinal microbiota in exposure to the antibiotic mixtures and other toxic compounds like chlorpyrifos and arsenic [17,28,48]. These "SHIME-compartment" specific effects could be due to the inconsistent compounds biodegradation, gut microbiome community, and pH in different colon regions [17,28,48]. Meanwhile, the largest increase of predicted human disease-related pathways was also shown in ascending colon after AMX exposure (Fig. 2), which clarified that the primary effect observed at the level of the microbiota could also be identified at the genomic and metabolic level, as reported in other researches [49,50]. Furthermore, these AMX effects were still evident for at least two weeks after the AMX discontinuance although the resilient reductions of Klebsiella and human disease-related pathways were observed. A recent research revealed that AMX mediated opportunistic pathogen Escherichia/Shigella blooms persisted 42 days after the interruption of antibiotic therapy [22]. Similar phenomenon was also reported by Zwittink and colleagues, who also discovered that 5 days treatment with AMX and ceftazidime combinations allowed Enterococcus to thrive and remain dominant for up to two weeks and caused Bifidobacterium abundance to remain decreased till postnatal week six after antibiotic treatment discontinuation [51]. Therefore, our results demonstrated the serious negative side-effect of AMX could be persistent and colon region-different, which should be considered as an important aspect of the risk assessment for AMX prescription.

AMX dramatically increased the virulence of Klebsiella by mutations of virulence factors
Our finding that seven days of 600 mg/day AMX acclimated NKU_Kleb8A7 had significantly higher virulence than NKU_Kleb8 supported the hypothesis that AMX exposure could significantly increase the virulence of Klebsiella, which has not been reported yet.
However, antibiotics chloramphenicol and erythromycin had been found to increase the virulence of Acinetobacter baumannii by increasing the production of the K locus exopolysaccharide (one important component of biofilm) [52]. Treatment with ciprofloxacin could enhance the expression of virulence genes of opportunistic pathogen Salmonella typhimurium, resulting in the increased virulence [53]. These results indicated that the significantly increased virulence of Klebsiella revealed in our research might also be attributed to the improved virulence genes expression. It was mentioned earlier that the inherent resistance to β-lactam already poses a significant therapeutic challenge for Klebsiella infections, the AMX-induced virulence increasing phenomenon found in this study undoubtedly further increase its health risk [54][55][56].
K. pneumoniae employs four major classes of virulence factors (fimbriae, lipopolysaccharide (LPS), capsule, and siderophores) to protect itself from the inhibition of host immune [57]. Fimbriae virulence factors are important mediators for the adhesion and biofilm formation of K. pneumoniae. In this study, fimbriae virulence factors of pilB , pilG, and fimD were all mutated in both evolved strains NKU_KlebA1 and NKU_Kleb8A7 which may substantially increase the biofilm formation ability of K. pneumoniae, resulting in increasing their virulence [58][59][60][61]. LPS and capsule virulence factors have been suggested as the main deterrent to complement-mediated lysis [62,63]. O antigen is the outermost subunit of LPS and has important roles in protecting against complement. Lack of O-antigen may render K. pneumoniae sensitive to complement-mediated killing in the bloodstream, thus making the strain less virulent [64]. GtrB encodes bactoprenol glucosyl transferase, which was necessary to express O-antigen in Shigella flexneri, and its mutation in both evolved strains may attribute to their increasing serum resistance ability [65]. Moreover, the new emerged O-antigen related genes such as wzt, wzx, manB, and manC in NKU_Kleb8A7 putatively improve O-antigen's deterrent to complement-mediated lysis. Cps gene cluster is involved in the production of the capsule polymer, and acapsular strains are easier detected by opsonophagocytosis and more likely killed in serum via the alternative complement pathway, thus the mutated cpsB and cpsI may be the important mediators of increasing serum resistance [66,67]. Siderophores are necessary virulence factors for K. pneumoniae to acquire iron from the host in order to survive and propagate during infection, and the production of more than one siderophore help avoiding neutralization of one siderophore by the host [68]. Enterobactin is the primary siderophore used by K. pneumoniae, fepA is required for its biosynthesis and entD is necessary for its transport, the mutations of these two genes in both evolved strains may substantially increased their affinity for iron [69]. Similaly, iutA that encods aerobactin was also mutated in both evolved strains, which maybe a very important factor for G. mellonella infection [70]. Moreover, the new emerged yersiniabactin synthesis and secretion genes such as irp1, irp2, fyuA, and ybtT in NKU_Kleb8A7, as well as salmochelin formation related gene iroN and iroE putatively attributed to its stronger infection ability to G. mellonella [71,72].
Several other virulence factors were recently identified as being important for virulence of K. pneumoniae. It's well-known that many human pathogens use type II secretion system to secrete toxins and enzymes to damage specific tissues, resulting in cell damage and disease [73,74]. Virulence factor gspG, which encodes the putative type II secretion protein, was confirmed to be mutated after AMX exposure in both evolved strains of NKU_KlebA1 and NKU_Kleb8A7. As one important subassembly of type II secretion system, pseudopilin gspG family was reported to drive folded proteins across the outer membrane in a piston-like manner. The expression of gspG from lacZP in E. coli had been reported to produce gspG-specific antiserum recognized protein [75]. The C terminus of protein G represented a permissive site which is important for its function, the mutations shown in this study would affect its combination with proteins secreted and antibody in serum, which may then cause the increasing survival of Klebsiella when exposure to serum or G. mellonella infection [76]. Pathogenic species also use type III secretion systems to inject toxins into attacking immune cells, the mutated genes such as vscN and exsA may upregulate expression of type III secretion systems in NKU_Kleb8A7 and then improve its survival and affection ability [77,78]. ABC iron transport system is also identified important for K. pneumoniae virulence, which is involved in the acquisition of iron [79]. This kind of virulence factor genes fbpC and hitC were confirmed to mutate after AMX exposure in both evolved strains, which may enhance their infection ability in G. mellonella. However, further studies are needed to verify whether the mutations of the above virulence factors is related to the increased virulence of K. pneumoniae observed in this study.

Conclusions
Exposure to AMX increased the abundances of Klebsiella and human disease-related pathways in simulated human gut microbiota and the most significant enhancement effect was observed for ascending colon. Meanwhile, the significant increase in Klebsiella population could persistent for at least two weeks after AMX discontinuance. The increased predicted functional pathways were positively associated with Klebsiella and whole-genome analysis also confirmed Klebsiella strains indeed carry these human disease-related KEGG pathways. Moreover, exposure to AMX dramatically enhanced the virulence of Klebsiella including biofilm formation, serum resistance, and G. mellonella infection. Genome sequence comparison analysis of strain NKU_Kleb8 (isolated from SHIME before AMX exposure) with the strain NKU_KlebA1 (isolated from SHIME after AMX exposure) and NKU_Kleb8A7 (evolved from NKU_Kleb8 after AMX exposure) revealed that the increased virulence may be attributed to the mutations of four major virulence factors

SHIME model start up and samples collection
The constructed SHIME was formed by five double-jacketed reactors simulating the stomach, small intestine, ascending colon, transverse colon, and descending colon, respectively (Fig. 1A). The last three reactors were inoculated with a mixture of fecal microbiota from a healthy adult volunteer, who did not suffer from gastrointestinal diseases or take antibiotics in the past six months [28,80]. During the first two weeks of the experiment, AMX-free nutritional medium was added to the reactors to stabilize the microbial community. After this period, the SHIME was sequentially exposed to AMX-free nutritional medium for another one week and followed by nutritional medium + 600 mg/day AMX for one week, and AMX-free nutritional medium for two weeks. The details of the SHIME system and the startup process are summarized in the Supporting Material.
Liquid samples (mixtures of fecal microbiota with SHIME feed) were collected from simulated ascending colon, transverse colon and descending colon vessels at six time points of 14, 21, 24, 28, 35, and 42 day, as detailed in Fig. 1A. Based on the situation of AMX exposure, these samples can be classified into three groups. Specifically, the first group of samples were collected after stabilization of SHIME setup for two weeks (C-A1, C-T1, and C-D1 from ascending colon, transverse colon and descending colon, respectively) and before the AMX administration (i.e., C-A2, C-T2, and C-D2). The second group sampled after the administration of AMX for three days (AMX-A1, AMX-T1, and AMX-D1) and seven days (AMX-A2, AMX-T2, and AMX-D2). Finally, the third group of samples were collected after the discontinuance of AMX for one week (R-A1, R-T1, and R-D1) and two weeks (R-A2, R-T2 and R-D2). Samples were stored at −80 °C for further analyses. Raw Illumina fastq files were demultiplexed, quality-filtered, and analyzed using Quantitative Insights Into Microbial Ecology (QIIME) [81]. Sequences that were shorter than 55 bp, contained primer mismatches, ambiguous bases or uncorrectable barcodes, were removed. 16S rRNA gene sequences were assigned to operational taxonomic units (OTUs) using UCLUST with a threshold of 97% pairwise identity, and then classified taxonomically using the Ribosomal Database Project (RDP) classifier 2.0.1 [82]. LEfSe analysis was served to determine bacterial taxa that significantly differed between the control and AMX exposure group using the Galaxy application tool (http://huttenhower.sph.harvard.edu/galaxy/) with a linear discriminant analysis, where the cutoff score is of 3.0 and a p of < 0.05 for statistical significance [84]. These analyses were conducted by Guangzhou Gene Denovo Co., Ltd (Guangzhou, China).

Qualitative PCR, Biofilm formation, serum killing and G. mellonella infection assay of Klebsiella isolated strains
A total of 16 Klebsiella strains were isolated from ascending colon samples by a Klebsiellaspecific MacConkey-inositol-adonitol-carbenicillin medium, as described by Gao et al [85].
Among them, eight were isolated from sample C-A2 (before AMX administration) and the others were from AMX-A2 (after AMX exposure). DNA was extracted from Klebsiella strains using the Wizard Genomic DNA Purification Kit (Promega, USA) according to manufacturer's protocols. 16S rRNA gene identification conducted by BGI Tech Solutions Beijing Liuhe Co., Ltd (Beijing, China) also confirmed the isolated strains were Klebsiella.
The phylogenetic tree of these Klebsiella strains was conducted by MEGA X [86]. 5 ARGs (bl2b_tem1, bl2be_shv2, arna , rosb , and vang ) related to drug resistance pathways from the isolated strains were confirmed using Qualitative PCR on a Biometra T100 gradient (Biometra) with the primers listed in Table S3 [87].
Biofilms formation were assayed using the standard crystal violet quantification assay described by O'Toole and associates [88]. Serum-killing assay was performed based on the method described by Rosen and associates [89]. G. mellonella infection assay was conducted as described by McLaughlin and associates [90]. Details of phylogenetic tree establishment, Qualitative PCR, biofilm formation, serum killing and G. mellonella infection assay procedures are described in the Supporting Material. SNP mutations between NKU_Kleb8A7 and NKU_KlebA1 with NKU_Kleb8 were obtained by sequence alignment using the software package MUMmer version 3.23 [92]. After the alignment of LASTZ (http://www.bx.psu.edu/~rsharris/lastz/), the best match results (less than 10 bp) were extracted by using axtBest to obtain the preliminary InDel results [93].

Genome sequencing and analysis of Klebsiella
The 150 bp (3×SD) from upstream and downstream of the reference sequence InDel sites were aligned and validated with the sample sequencing reads by BWA software [94]. After filtering, the reliable InDel sites were obtained. These analyses were conducted by BioMarker Technology Co., Ltd (Beijing, China).

Data analysis
Results are expressed as mean values and standard deviations. The statistical analysis was performed with SPSS 17.0 software. The T-tests were conducted to compare the difference between groups and all statistical tests were two-tailed. Statistical significance was set at three different levels (*p < 0.05, **p < 0.01, and ***p < 0.001). Spearman test for correlation analysis between gene numbers of human disease-related pathways and the Klebsiella abundance were performed in R with the vegan package.

Availability of data and materials
The datasets and scripts developed and generated in this manuscript are included within the manuscript and its supporting files.

Consent for publication
Not applicable.

Authors' contributions
LL carried out the laboratory work, analyzed the data, and wrote the manuscript. ZC and QW revised the manuscript and provided financial support. JS, WB, SW, YG, HL, HQ, SZ and JY carried out the laboratory work. YX provided financial support. DM and YL guided the laboratory work and revised this manuscript, and provided financial support. All authors read and approved the final manuscript.

Ethics approval and consent to participate
The study was approved by the Biomedical Ethics Committees of Nankai University. The participant has given written informed consent to understand the study purpose, procedures, risks, benefits, and rights.   Table S1.
Antibiotic resistance genes (ARGs) related to drug resistance pathways detecting in 8 Klebsiella strains from AMX-exposed sample. bl2b_tem1 and bl2be_shv2 related to βlactam resistance; arna and rosb related to cationic antimicrobial peptide (CAMP) resistance; and vang related to vancomycin resistance.  Fig. S1. Composition of gut bacterial community at phylum level in three colon regions at different sampling points, in which letters A, T and D mean ascending, transverse, and descending colon, reprehensively. C-A1 and C-A2 represent samples collected from ascending colon at 14 and 21 days before AMX administration, AMX-A1 and AMX-A2 represent the samples collected from ascending colon at 24 and 28 days during 600 mg/day of AMX treatment, R1-A1 and R-A2 represent the samples collected from ascending colon at 35 and 42 days after termination of AMX. In this way, C-T1, C-T2, AMX-T1, AMX-T2, R-T1, and R-T2 refer the samples from transverse colon at the corresponding days and treatments. C-D1, C-D2, AMX-D1, AMX-D2, R-D1, and R-D2 refer the samples from descending colon at the corresponding days and treatments. Fig. S2.
Heatmap of the average nucleotide identity (ANI) for strains NKU_KlebA1 and NKU_Kleb8.
Colors reflect ANI of two strains from low (red) to high (blue). The area of the ellipse reflect ANI of two strains from low (large) to high (small).
Additional file 2: Excel S1. The mobile elements and metabolic systems of the K.
Additional file 3: Excel S2. Human disease-related systems of the K. pneumoniae