Collateral sensitivity to pleuromutilins in vancomycin-resistant Enterococcus faecium

The acquisition of resistance to one antibiotic sometimes leads to collateral sensitivity to a second antibiotic. Here, we show that vancomycin resistance in Enterococcus faecium is associated with a remarkable increase in susceptibility to pleuromutilin antibiotics (such as lefamulin), which target the bacterial ribosome. The trade-off between vancomycin and pleuromutilins is mediated by epistasis between the van gene cluster and msrC, encoding an ABC-F protein that protects bacterial ribosomes from antibiotic targeting. In mouse models of vancomycin-resistant E. faecium colonization and septicemia, pleuromutilin treatment reduces colonization and improves survival more effectively than standard therapy (linezolid). Our findings suggest that pleuromutilins may be useful for the treatment of vancomycin-resistant E. faecium infections.

T he patterns of evolutionary cross-resistance to clinical antibiotics are a major driving force for accelerating the emergence and dissemination of multidrug resistant (MDR) bacteria 1 . Potential avenues for antibiotic resistance are co-selection for resistance driven by antimicrobial-producing organisms and harmful substances in natural environments where active compounds compete for the same targets 2 . The possibility of such crossresistance to clinical antibiotics has received intensive attention previously. Antibiotic resistance often carries various fitness costs in the absence of selective pressures, and such trade-offs in turn occasionally result in rugged fitness landscapes to channel the evolutionary trajectory 3 . However, the understanding that whether and how negative responses to antibiotic selective pressure modulate the trajectory of evolution in bacteria remains unclear 4,5 . Indeed, pioneering works have recently implicated the patterns of collateral sensitivity in Escherichia coli 6 , Pseudomonas aeruginosa 7 , Staphylococcus aureus 8,9 and other pathogens, wherein resistance to one antibiotic simultaneously induces susceptibility to another by forming either homogeneous or heterogeneous populations 10 (Fig. 1a). The clinical implications of collateral sensitivity therefore supply prioritized rational therapies such as the sequential or concurrent deployment of reciprocal collateral sensitivity antibiotic pairs, to combat the antibiotic resistance crisis.
The increasing vancomycin-resistant enterococci (VRE) particularly vancomycin-resistant Enterococcus faecium (VRE fm ), with high genomic plasticity and metabolic flexibility, seriously compromise the effectiveness of existing antibiotics 11 . The persistence and spread of VRE fm within health care settings has become one of the most challenging nosocomial pathogens, leading to at least 5400 estimated deaths and more than $500 million in excess health care costs annually in the United States (U.S.) 12 and accounting for 37% nosocomial infections in Germany 13 . Although vancomycin resistance is widely mediated by mega-plasmids containing diverse van gene clusters, these resistance alleles have been persistently maintained in VRE with low or without fitness cost 14 . Therefore, a better mechanistic understanding of antibiotic resistance is urgently required to combat VRE fm associated infections. However, the evolutionary and pharmacological consequences of dedicated van genes remain largely unknown. In this work, we exploit the evolutionary trade-offs to identify the collateral sensitivity patterns of 102 clinical VRE fm isolates nationwide for antibiotic repurposing. Our observations indicate that pleuromutilin antibiotics are promising candidates targeting VRE fm in vitro and in animal models, shedding light on the evolution-directed rational design of antibiotic therapies against MDR bacterial infections.
Results and discussion VRE fm show collateral sensitivity to pleuromutilins. To identify the trade-offs in VRE fm , we first determined the susceptibility and constructed the network of collateral sensitivity of 10 main classes of antibiotics routinely used in the clinic against both 20 VRE fm and 20 vancomycin sensitive E. faecium (VSE fm ) isolates. Markedly, compared to VSE fm , we observed that VRE fm exhibited specific susceptibility to lefamulin (Fig. 1b, Supplementary Tables 1, 2), an approved pleuromutilin antibiotic by the U.S. Food and Drug Administration (FDA) for community-acquired bacterial pneumonia (CABP) in 2019 15 , although they displayed general cross-resistance to multiple antibiotics as expected (Fig. 1c). The susceptibility to lefamulin increased more than 500 folds, from ≥16 μg/mL to 0.03 μg/mL. To extend whether other ribosome-targeting antibiotics display similar behaviors, we tested eight antibiotics with different binding sites 16 . It confirmed a strong collateral sensitivity to lefamulin with the minimum R e s is t a n t S e n s it iv e R e s is t a n t S e n s it iv e  inhibitory concentration required to inhibit the growth of 50% bacteria (MIC 50 ) of 0.03 μg/mL in VRE fm , whereas most VSE fm were resistant to lefamulin with MIC 50 of 16 μg/mL (Fig. 2a). Therefore, we hypothesized that such collateral response could be due to the distinct modes of action of pleuromutilins (PLEs).
To test the generality of VRE fm channeled evolution toward elevated susceptibility to pleuromutilins, we first performed the structure-activity relationship (SAR) analysis of five pleuromutilins routinely used in human and veterinary medicine (Fig. 2b, Supplementary Fig. 2a). We expanded the number of E. faecium isolates to 210, including 102 VRE fm isolates of human origins and 109 VSE fm isolates of human, animals and probiotics in China. These VRE fm isolates validated the robust collateral sensitivity to pleuromutilins, particularly the decreased MICs of lefamulin, retapamulin and valnemulin with resistance ratios of <0.002 (Fig. 2b, Fig. 1a). Notably, the vanA gene clusters are dominant (95%, 20/21) in 21 clinic VRE fm isolates ( Supplementary Fig. 1b), consisting with the vanA genotype most frequently recorded 11,13 . To further assess the generality of collateral sensitivity in other van gene clusters, we found that a clinical vanB-type E. faecium isolate showed similar collateral patterns to five pleuromutilins (Supplementary Table 2). Then, we tested the activity of lefamulin against E. faecalis including vancomycin resistant/ sensitive isolates. In contrast to VRE fm , vancomycin-resistant E. faecalis (VRE fs ) were resistant to lefamulin (Fig. 2c), implying the species specificity of collateral sensitivity in enterococci. The collateral pattern indicates the species-specific effect 19 , where the collateral sensitivity to pleuromutilins in VRE fm is contingent upon the intrinsic genomics. It is consistent with previous studies 20,21 that the van gene clusters are located on pMG1-like and pheromone sensing plasmids in VRE fm and VRE fs , respectively. Additionally, the ubiquity of plasmid-encoded toxin-antitoxin gene systems may also account for the species-specific effect 22 . Collectively, it suggests that the antibiotic pairs of vancomycin and pleuromutilins exhibit ubiquitous collateral sensitivity in VRE fm .
To further characterize the collateral response, we noticed the susceptibility of 210 E. faeciums isolates in four patterns that approx. 90% isolates (89/102) of VRE fm display collateral sensitivity to pleuromutilins using lefamulin as a model (P s V r ) (Fig. 2d). Intriguingly, we observed the contrary pattern of pleuromutilin susceptibility in VSE fm , in which approx. 90% isolates (96/109) of VSE fm were resistant to pleuromutilins (P r V s ). Therefore, E. faecium shape the divergent evolution upon pleuromutilins to four patterns, including P s V r , P s V s , P r V s and P r V r , and in turn such contingency can be used to design rational approaches to treating the prevalent VRE fm associated infections.
Heterogeneous collateral responses are channeled by ribosomes. The unique collateral sensitive antibiotics suggest that the interaction between pleuromutilins and the target may shed light on elucidating the evolutionary conservation in VSE fm . First, lefamulin showed moderate bacteriostatic activity against VRE fm at a high level (1.2 μg/mL, 40 × MIC) ( Supplementary Fig. 3). Furthermore, we quantified the accumulation of lefamulin in eight E. faecium isolates including all four phenotypes of heterogeneous collateral responses and found no difference (Supplementary Fig. 5a, b). Last, we confirmed there was almost no   further modifications for the accumulated lefamulin in E. faecium ( Supplementary Fig. 5c-j). These data denote that the intrinsic properties of pleuromutilins should be excluded and the main target of pleuromutilins should eventually dominate the trajectory of collateral response in VRE fm . Pleuromutilins exhibit distinctive recognition to the peptidyl transferase center (PTC) by blocking bacterial protein biosynthesis 16 . Pleuromutilins mainly interact with eight nucleotides in the PTC domain. We first confirmed that there were no mutations at such sites in 20 VRE fm and 20 VSE fm isolates based on whole-genome sequence analysis (Fig. 4a, Supplementary  Figs. 2c, 6b). Moreover, we found no genes such as cfr 23 encoding methyltransferases to confer universal resistance ( Supplementary  Fig. 7a-c). Interestingly, we found that pleuromutilins had a higher affinity to ribosomes in P s V r than P r V s type strains through fluorescence polarization analysis 24 ( Supplementary  Fig. 8a, b), implying that VRE fm may modulate the susceptibility to pleuromutilins through the mechanism of ribosomal protection. The ATP-binding cassette F (ABC-F) protein family protects bacterial ribosomes from multiple classes of ribosome-targeting antibiotics 25 . We found the presence of four prevailing genes (msrC, eatA/eatAv and lsaE) of the 26 members of the ABC-F family 26 in 40 E. faecium isolates ( Supplementary Fig. 6b). Nevertheless, both lsaE and eatA/eatAv were partially carried by clinical isolates and expressed at similar levels in the presence and absence of pleuromutilins ( Supplementary Fig. 9a, b). Extremely, we noticed that the species-specific gene msrC 27 expressed more than 70-fold higher in model strain VSE fm CAU310 than VRE fm CAU369 ( Supplementary Fig. 9c), suggesting that the low expression of msrC may potentiate the efficacy of pleuromutilins. Meanwhile, we observed that the decreased transcription of msrC in two lefamulin sensitive VRE fm isolates (Fig. 3a, Supplementary  Fig. 10a) is in a dose-dependent manner of lefamulin. Furthermore, we found the increased transcription of msrC in all 12 lefamulin resistant isolates as well (Fig. 3b, Supplementary Figs. 10b, 11). Constantly, we obtained the increased expression of MsrC in a lefamulin resistant VRE fm CAU378 treated with lefamulin for 1 h (Fig. 3c), based on proteomics analysis. Additionally, the MsrC overexpression strain was constructed in a pleuromutilins-sensitive strain using conjugative transformation ( Supplementary Fig. 12a). We found that the conjugant (pAM401 + msrC) shows high expression of msrC and are resistant to all pleuromutilins (Supplementary Fig. 12b-d). Taken together, these results indicate that msrC is linked with the collateral sensitivity in VRE fm .
To explore how MsrC reduces the susceptibility to pleuromutilins, we performed a simulation analysis on the interaction between MsrC and the PTC domain 28 . Compared to the affinity between pleuromutilins and PTC with Z-docker interaction energy of −146.324 kcal, the residues of R241 and L242, and K233 and K246 in MsrC competitively bound to the shared binding sites U2504 and C2063 of MsrC and lefamulin in PTC domain, respectively, with Z-docker interaction energy of −170.611 kcal (Fig. 3d). These findings indicate that high expression of species-specific MsrC tends to block pleuromutilins targeting PTC in VSE fm , whereas the low expression of MsrC facilitate the binding of pleuromutilins in VRE fm accordingly.
Epistasis between van gene clusters and msrC. To understand how msrC mediates collateral response in VRE fm , we first observed the delayed growth curves of a constructed conjugant containing the vanA-plasmid from E. faecium CAU369 (pCAU369) in E. faecium BM4105-RF 20 , in the presence of sub-inhibitory levels of lefamulin ( Supplementary Fig. 14b, c). It suggested that the plasmid carrying the van gene cluster plays a crucial role in the increased susceptibility to pleuromutilins in VRE fm . Subsequently, we showed that most van gene clusters are dominant vanA-type (95%, 19/20) ( Supplementary  Fig. 6b), consisting with previous reports that the increasing dissemination of vanA-VRE fm is prevalent world widely, particularly in the U.S. and Europe 29,30 . Therefore, we compared 14 plasmids containing vanA gene clusters in global E. faecium isolates (Supplementary Table 5) with that in E. faecium CAU369, and noticed only the vanA gene clusters were present in all isolates ( Supplementary  Fig. 13). Thus, we deduced that the vanA gene clusters (~6-7 kbp), instead of the other motifs in the vanA-type megaplasmids 31 (3 0-150 kbp), contribute to the collateral sensitivity to pleuromutilins in VRE fm .
Given that the decreased expression of msrC in VRE fm (Fig. 3a, Supplementary Fig. 9c), we hypothesized a negative feedback between msrC and the vanA gene clusters. To verify such epistasis, we found the sequence in the promoter of msrC shared high similarity (83.3%, 10/12) to the promoters of vanR/H/Y 32 (Fig. 4a) and the promoters are conserved in both vanA-and vanB-type isolates globally ( Supplementary Fig. 15), indicating that phosphorylated VanR may simultaneously induce van transcription but inhibit msrC transcription. Toward this goal, we exploited transcriptome analysis on VRE fm treated with lefamulin. Remarkably, we observed vanRS activation and decreased msrC transcription in pleuromutilin-sensitive VRE fm CAU369, whereas there was no vanS activation and 33.1% vanR transcription, and increased msrC transcription in pleuromutilin-resistant VRE fm CAU378 ( Supplementary Fig. 16). Consistently, we confirmed the opposite patterns of vanS and msrC transcription in pleuromutilin-resistant isolates under lefamulin treatments based on qRT-PCR and proteomics analysis (Fig. 4b-d, Supplementary Fig. 17a, b). To validate that vanR modulates msrC expression, we constructed a conjugant by receiving a recombinant vanRS plasmid in a pleuromutilin resistant E. faecium ( Supplementary Fig. 18). The transcription of vanR and vanS were activated in a dose-dependent manner under lefamulin treatments ( Supplementary Fig. 19a, b), in turn, the transcription of msrC in the conjugant (pAM401 + vanRS) was dramatically inhibited (Supplementary Fig. 19c). Correspondingly, the conjugant with vanRS expression is sensitive to pleuromutilins, with more than 16-fold decreased MICs (Supplementary Fig. 19d). These results support our hypothesis that the negative epistasis between the van gene cluster and msrC is responsible for the collateral sensitivity to pleuromutilins in VRE fm .
To further verify that P-VanR regulates msrC transcription, we first demonstrated that P-VanR binds to the msrC promoter using electrophoretic mobility shift assay (Fig. 4e, f). Furthermore, we calculated the EC 50 (effective concentration for 50% response) as 1.01 μmol/L for P-VanR binding to the fragment of msrC promoter, which is much lower than the binding affinity between P-VanR and the vanH promoter (Fig. 4g). When in the presence of inducers such as classic vancomycin and lefamulin, VanS switches its activity from phosphatase to kinase, phosphorylating the cognate response regulator VanR 33 . Phospho-VanR then binds to similar promoter sequences triggering the transcription of the van gene cluster, and inhibits msrC transcription to induce VRE fm sensitive to pleuromutilins accordingly (Fig. 4h). Altogether, these results indicate that P-VanR co-regulates vanR/ vanH and msrC by binding to similar promoter fragments, to facilitate pleuromutilins against VRE fm .
In addition, antibiotic resistance/tolerance can be regulated by metabolic reprogramming 34,35 , which, in principle, could boost the effectiveness of antibiotics, consistent with the observation that anaerobic glycolysis further exacerbated the growth of VRE fm with vanA-plasmid in the presence of subinhibitory levels of lefamulin ( Supplementary Fig. 14c). To test this possibility, we dissect whether metabolites modulate collateral response in VRE fm . Since there is no Krebs cycle in enterococci 36 , the dominant pyruvate metabolism kept steadily ( Supplementary  Fig. 21a-c). Considering the abundance of metabolic versatility and the challenge to genetically manipulate clinical isolates, its contribution to collateral sensitivity in most VRE fm with megaplasmids remains unclear.
Collateral sensitivity in vivo. Given that VRE usually resists multiple classes of antibiotics, we hypothesized that pleuromutilins might be used to treat VRE associated infections based on our observation. Lefamulin was recently approved in the U.S. and European Union for the treatment of CABP in adults. To further assess the efficacy of lefamulin, we empirically evaluated its efficacy in two mouse models including an intestinal colonization model (VRE fm CAU369) and in a peritonitis-septicemia model in mice (VRE fm CAU427) (Fig. 5a). First, we found that VRE fm CAU369 mainly colonized in the cecum and colon with remarkably decreased bacterial burden after a single dose of lefamulin (Fig. 5b). Remarkably, lefamulin showed better antibacterial efficacy than linezolid, the only antibiotic with the U.S. Food and Drug Administration (FDA) approval for treating VRE infections, especially in the cecum and colon on the first day after administrations. Meanwhile, we collected daily the feces for bacterial counting. Compared to clinically recommended linezolid, lefamulin similarly produced time-dependent reduction in bacterial burden of feces over the 7-day treatment period (Fig. 5c). In addition, VRE fm promptly dominated mouse gut microbiota after intragastric inoculum based on bacterial community analysis, whereas lefamulin administration notably reduced its abundance through the upregulated the genera of Akkermansia and Klebsiella (Fig. 5d). After 7 days, all mice nearly eliminated the pathogens and appeared healthy. Intriguingly, lefamulin could restore the homeostasis of fecal microbiota faster than the untreated group according to the Chao1 and Shannon indexes ( Supplementary Fig. 22a, b), particularly the patterns of Bacteroides, Prevotella, Clostridium, and Escherichia/Shigella on the 7th day ( Supplementary Fig. 22c-f). In the peritonitissepticemia model, the survival curve showed that all mice survived under the treatment of a dose of 10 mg/kg lefamulin after 96 h (Fig. 5e), with reduced bacterial counts in organs (Fig. 5f). Altogether, our results suggest that lefamulin efficiently remedies the severity of VRE fm infections, which may be extended to treat and diminish bacterial colonization since that there are limited effective agents available to treat such infections in clinic.
For the foreseeable future, VRE especially VRE fm will remain an important nosocomial problem 37,38 , particularly the steady increase in the U.S. since the early 1990s in hospitals 11 . Because VRE fm has developed resistance to essentially every antibiotic used in clinic, novel therapeutic strategies to explore collateral sensitivity may be attempted. A better understanding of the evolutionary biology particularly epistasis by which the van genes manipulate ribosome protection could shed light on intervention strategies against VRE fm pathogens. Notably, plasmid mediated epistasis is not restricted to E. faecium, and we postulate that such collateral response may be widespread phenomenon in other bacteria. Mobile plasmids drive the spread of many critical antibiotic resistance genes in clinical pathogens 39 . Most recently, it has been demonstrated the collateral sensitivity associated with transferable plasmids in clinical E. coli isolates 40 , compared to previous empirical observations and mutations in chromosome and plasmids [5][6][7][8][9] . Our results suggest that the evolutionally collateral response has already dominated the prevalent VRE fm carrying conjugative plasmids (Supplementary Fig. 14a). To argue  for preclinical development of pleuromutilins as leads against VRE fm , continued studies should further focus on the evaluation whether pleuromutilins reduce VRE fm colonization in patients and maintain microbiota recovery after its expansion.
In conclusion, our results demonstrate the epistasis between van genes and msrC mediating collateral sensitivity in most clinically relevant VRE fm strains, to potentiate the efficacy of pleuromutilins. These observations in vitro and in vivo provide proof-of-concept for an efficient therapeutic option against the increasingly aggravating VRE fm associated infections by revitalizing existing antibiotics. Pleuromutilins not only exhibit robust bacteriostatic activity against clinical isolates nationwide but also attenuate the colonization possibly improving clinical outcomes. Overall, our work expands the knowledge of evolutionary   ATP determination. Extracellular and intracellular levels of ATP were determined using an Enhanced ATP Assay Kit (Beyotime). VRE fm CAU369 grown overnight at 37°C with shaking at 200 r.p.m. were washed and resuspended to obtain an OD600 of 0.5 with 0.01 M of PBS (pH 7.4). After treatment with 1 × MIC, 5 × MIC and 10 × MIC lefamulin for 1 h, bacterial cultures were centrifuged at 12,000 r.p.m. and 4°C for 5 min, and the supernatants were collected for the determination of extracellular ATP levels. Meanwhile, bacterial precipitates were lysed by lysozyme, and centrifuged, then the supernatants were prepared for the measurement of intracellular ATP levels. The detecting solution was added to a 96-well plate and incubated at room temperature for another 5 min. Last, the supernatants were added to the well and mixed quickly, before recording in the model of luminescence using the Infinite M200 Microplate reader (Tecan).
Antibiotic accumulation test. The accumulation of intracellular antibiotics in bacteria was determined based on the established liquid chromatography with tandem mass spectrometry (LC-MS/MS) method 41 . Overnight cultures of bacterial cells (VRE fm CAU369, VRE fm CAU372, VRE fm CAU378, VRE fm CAU419, VSE fm CAU259, VSE fm CAU277, VRE fm CAU309, VRE fm CAU310, S. aureus 29213, MASA T144, E. faecalis 29212, VRE CAU 475) were diluted to 100 mL fresh BHI broth at 1:100 and resuspended to OD600 of 0.5 at 37°C. Bacteria were then centrifuged at 3,000 g for 10 mins at 4°C and the supernatants were collected for three times. Subsequently, bacterial cells were diluted to 10 10 CFU per mL with fresh PBS and aliquoted in 1.5 mL tubes. The aliquots were treated with subinhibitory levels of lefamulin at 37°C for 1 h and the samples were collected to destroy the cell envelopes for LC-MS/MS analysis (Waters 2695). Additionally, structure elucidation of accumulated lefamulin in four phenotypes of E. faeciums isolates were based on UHPLC-Q-Orbitrap analysis (UHPLC-Q-Exactive Plus, Thermo Fisher Scientific). Extracted ion chromatogram of lefamulin obtained in the positive ESI mode at retention time of 7.08 min. Elucidation of lefamulin and its product ions based on the molecular weights of m/z 508.31 (lefamulin), and product ions of m/z 188.07 and 206.08.
Whole-genome sequencing. Genomic DNA were extracted from the overnight culture of E. faeciums (including 20 VRE fm and 20 VSE fm ) isolates in BHI, according to the manufacturer instructions (Genome Extraction Kit, Magen). Obtained DNA was sequenced by Illumina Seq and the whole-genomes were aligned with the resistance gene database from the Center for Genomic Epidemiology (CGE).
Molecular simulation. Models of the complexes of lefamulin-ribosome and MsrCribosome were built on Discovery Studio 2018 Client. The model of MsrC was generated using the sequence from VRE fm CAU369. The receptor-ligand interaction and the Z-Docker reaction energy in the 50S and 30S subunits, and tRNA were performed according to a previous study 28 .
Methylation modification analysis. Methylation modification analysis is based on the SELECT approach according to a previous study 42 . Briefly, total RNA (1500 ng) was mixed with 40 nM up primer, 40 nM down primer and 5 M dNTP in 17 μL CutSmart buffer. SELECT qPCR was performed with the following program: 90°C for 1 min, 80°C for 1 min, 70°C for 1 min, 60°C for 1 min, 50°C for 1 min and 40°C for 6 min. Afterward, the qRT-PCR was performed using SYBR premix Ex Taq qPCR Kit (TaKaRa). qRT-PCR was performed with the following program: 95°C, 5 min; 95°C, 10 s then 60°C, 35 s for 40 cycles; 95°C, 15 s; 60°C, 1 min; 95°C, 15 s. Primers for SELECT qPCR or qRT-PCR are listed in Table S6, respectively. Ct values of samples were normalized to their corresponding Ct values of control. All assays were performed with three independent experiments. Fig. 4 Epistasis between the van gene cluster and msrC. a Promoters in genes vanR, vanH, vanY and msrC. b Fold changes in log 2 [FPKM] values of relative expression of vanS and msrC. Transcriptome analysis of E. faeciums under the treatment of lefamulin at levels of 1× and 10 × MIC for 1 h. c Volcano plot represents the protein expression ratios of lefamulin treated bacterial cells (VRE fm CAU378). For each protein, the -log 10 (P-value) is plotted against its log 2 (fold change). Proteins upregulated (P < 0.05, fold change > 2) in 1× and 10 × MIC lefamulin treated samples are colored in red, proteins downregulated (P < 0.05, fold change < −2) are colored in blue, while unchanged in black. d Proteomics analysis of VRE fm CAU378 treated with 1× and 10 × MIC lefamulin for 1 h. Proteins were identified as significantly different with fold changes of log 2 [fold changes] values of at fold-increase or fold-decrease of expression levels. e Binding reactions between P-VanR (12-1.5 μM) and msrC promoter fragment (212-bp, 0.3 ng) based on the gel electrophoretic mobility shift assay. Free probe: biotin-labeled promoter; Competitor: Unlabeled promoter; BSA: Bovine serum albumin. Experiments were performed as three biologically independent experiments. f Calculated protein-binding rates of P-VanR and vanH/vanR/msrC promoters, based on the gray values in Fig. 4e. g Binding constants of vanH/vanR/msrC promoters and P-VanR. h Scheme of collateral sensitivity in VRE fm . The decreased transcription of msrC by van genes enhances ribosome-targeting pleuromutilins binding to PTC to block protein synthesis. Data were presented as means ± S.D. n.s., not significant, determined by non-parametric one-way ANOVA (n = 3).
Fluorescence polarization assay. The fluorescence polarization value (FP value) was detected as described previously 43   Electrophoretic mobility shift assay. VanR purified and P-VanR prepared according to a previously described protocol 46 . Binding reactions between P-VanR (12-1.5 μM) and msrC promoter fragment (212-bp, 0.3 ng) based on the gel electrophoretic mobility shift assay. Briefly, P-VanR was incubated with different concentrations. After 15 min incubation, the reaction mixture was subjected to 5% nondenaturing polyacrylamide gel and electrophoresis was performed at 80 V in ice-cold bath. The images were visualized and acquired by the PharosFX imaging system (Bio-Rad, CA, USA).
Intestinal colonization model. To evaluate the in vivo efficacy of lefamulin, a mouse intestinal colonization model was involved according to a previous publication 44 . Briefly, 6-8 weeks old female ICR mice (20 g, n = 6 per group) were treated with PBS, lefamulin and linezolid. First, mice were administered 0.5 g per L ampicillin in drinking water for 5 days. On day 0, mice were infected with VRE fm CAU369 (1 × 10 9 CFUs) by gavage. After the treatments with drugs, CFUs of VRE fm CAU369 were enumerated in the feces from 1 to 7 days. Meanwhile, intestinal contents (including ileum, cecum and colon) were collected on the first and seventh day for bacterial counting and the evaluation of the abundance and diversity of intestinal flora, using a previously described method 44 . The alpha analysis and the beta diversity analysis were performed based on PCoA/NMDS analysis. GraphPad Prism 8 was used to plot graphs Mouse peritonitis-septicemia model. To construct an immunosuppressive mice model, the 6-8 weeks old ICR female mice (20 g, n = 6 per group, 18 mice were used in total) were treated with 200 mg/kg cyclophosphamide intraperitoneally at the 1 st day and 3 rd day. After the second cyclophosphamide was treated for 24 h, mice were infected with 0.5 mL of VRE fm CAU427 suspension (1 × 10 10 CFU per mouse) via intraperitoneal injection. At 1 h post-infection, all mice were treated with lefamulin and linezolid with 10 mg/kg every 24 h. PBS was used as the negative control. The survival rates of treated mice were recorded during a 96 h period and the bacterial number in the mice organs was counted on the plates.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
DNA and RNA sequencing data are available in NCBI SRA with accession number PRJNA628015. The MS raw files and proteome sequences data used in this study are available in the Proteome Xchange Consortium under accession code PXD030906. The metabolomic data have been deposited in the CNCB-NGDC repository under accession code PRJCA008528 and in MetaboLights under submission code MTBLS4498. Other data generated in this study are provided in the Supplementary Information and the Source Data file. Source data are provided with this paper. Fig. 5 Efficacy of lefamulin against VRE fm in vivo. a Scheme of mouse intestinal colonization model and VRE fm peritonitis-septicemia model. b Bacterial loads in the ileum, cecum and colon. Mice (n = 6 per group) were given 1 × 10 9 CFUs of VRE fm CAU369, under the treatment of a single dose of 5 mg/kg lefamulin or linezolid. c Bacterial loads in feces. Mice (n = 6 per group) were given 1 × 10 9 CFUs of VRE fm CAU369 by oral gavage, with administration of 5 mg/kg lefamulin or linezolid. d Fecal microbiota was profiled of mice infected with VRE fm CAU369 in the presence of 5 mg/kg lefamulin at different points using 16S rRNA gene sequencing. e Survival rates of peritonitis-septicemia mice infected with VRE fm CAU427 (1.0 × 10 10 CFUs) in the presence of lefamulin (10 mg/kg). f Lefamulin decreased bacterial loads in different organs in the mouse peritonitis-septicemia model (n = 6 per group). Data in b, c and d were presented as mean values ± S.D, f was presented as mean values ± SEM. P-values in b, c and d were determined by non-parametric one-way ANOVA and in f was determined by two-tailed t-test.