Enterococcus faecium EF332 strain used in this study was isolated in the Laboratory Medicine Center of the Second Hospital of Shandong University. The laboratory maintains an opportunistic pathogen library isolated from patient samples and routinely screens them for possible new findings. The strain was identified by the analysis of 16S rDNA sequence.

Antibiotic susceptibility tests were conducted according to CLSI/EUCAST guidelines (EUCAST, 2017; CLSI, 2018). The tested antibiotics include: penicillin (PEN), ampicillin (AMP), ciprofloxacin (CIP), gatifloxacin (GAT), chloramphenicol (CHL), tetracycline (TET), tigecycline (TGC), fosfomycin (FOF), erythromycin (ERY), linezolid (LZD), rifampicin (RIF), vancomycin (VAN), and daptomycin (DAP). K-B Disk diffusion assays were performed as previously documented (Vading et al., 2011). Agar dilution method (for FOF) and broth microdilution method (for the rest antibiotics) were adopted to determine MICs of antibiotics. Staphylococcus aureus ATCC 25923 and Enterococcus faecalis ATCC 29212 were used as control strains for disk diffusion and dilution method, respectively, according to CLSI and EUCAST standards (EUCAST, 2017; CLSI, 2018).

The genomic DNA of E. faecium EF332 and constructed strains (29212-pDL278, EFDO-cls, EF332-cls) was extracted following previous reported protocol (Qian and Zhao, 2014), and the purity and integrity of the DNA were confirmed by agarose gel electrophoresis. DNA samples of E. faecium EF332 were constructed into a 10-kb SMRTbell DNA library, and PacBio single-molecule sequencing was used to obtain at least 50× of sequencing data. In addition, a 350-bp small fragment library was constructed, and Illumina NovaSeq PE150 platform was used for paired-end sequencing to obtain at least 100× clean data for auxiliary assembly. DNA samples of constructed strains were used as templates for quantitative polymerase chain reaction (qPCR) to detect the relative copy number of plasmids.

To verify the transferability of plasmid pEF332-2, we conducted conjugation and transformation experiments. A chloramphenicol-resistant and vancomycin-sensitive E. faecalis 3–147 strain was used as the recipient for conjugations, and vancomycin-sensitive E. faecalis ATCC 29212 was used as recipient for transformation. Transconjugants were identified on the plates containing 32 μg/ml of vancomycin and 50 μg/ml of chloramphenicol, and transformants were identified on the plates containing 32 μg/ml of vancomycin.

Genomic DNA of E. faecium EF332 strain was used as template to amplify the presumed DAP resistance genes cls and gdpD by PCR using Phanta super-fidelity DNA polymerase (Vazyme Biotech Co. Ltd. Nanjing, China). All the PCR products were analyzed by 1% agarose gel electrophoresis and purified by DNA clean-up kit (TIANGEN, Beijing, China).

All primer sequences are shown in Supplementary Table 1, in which the cleavage sites of BamHI and SalI were added to the 5′ ends for the next step of cloning. The shuttle plasmid pDL278 with spectinomycin resistance was used for molecular cloning (Genbank accession number AF216802.1). The cls, gdpD fragments and pDL278 vector, were digested by BamHI and SalI endonuclease, respectively, and then, the gene fragments were ligated to pDL278 vector by T4 DNA ligase to generate recombinant shuttle plasmids (pDL278-cls, pDL278-gdpD). These recombinant plasmids were transferred into Escherichia coli DH5α chemically competent cells, respectively, according to the manufacturer’s protocols, and Luria-Bertani (LB) agar plates with 500 mg/L spectinomycin were used to select for positive transformants. The plasmids of the positive clones were extracted and transformed into the electroporation competent cells of E. faecalis ATCC 29212, and tryptic soy broth (TSB) agar plates with 1,000 mg/L spectinomycin were used to select for positive transformants. The successfully transferred gene fragments were validated by Sanger sequencing using primer M13 (as shown in Supplementary Table 1). Finally, we obtained the EF332-cls and EF332-gdpD strains for downstream analysis. A summary of constructed strains is shown in Table 1 for easier comparison.

In order to further confirm the role of mutations in cls and gdpD genes, site-directed mutagenesis was performed using overlap extension PCR (SOE-PCR; Ho et al., 1989). The cls gene was divided into four fragments at three mutated sites, and gdpD gene was divided into two fragments at one mutated site. The primers used are shown in Supplementary Table 1.

The SOE-PCR process consisted of two consecutive reactions. First-round PCR: Genomic DNA of E. faecium EF332 was used as a template to obtain the corresponding gene fragments. PCR products obtained were purified using the DNA purification kit (TIANGEN, Beijing, China). Second-round PCR: The purified PCR products obtained from the previous step were used as templates, and corresponding primers were added to the reaction system for amplification.

These point mutations led to the amino acid substitutions T269I, V203I, T298S in Cls, and S201P in GdpD. Plasmids carrying these mutated gene fragments were transformed to obtain EFDO-cls and EFDO-gdpD strains by the method described above.

Poly-_L-lysine conjugated to fluorescein isothiocyanate (PLL:FITC) was used to assay cell surface charges according to previous studies (Prater et al., 2021). In brief, the strains grew until OD₆₀₀ reached 0.5 in TSB. Cells were then washed three times with sterilized HEPES buffer (20 mM, pH 7.0) and diluted to OD₆₀₀ of 0.1. Cells were incubated with 50 μg/ml PLL:FITC (final concentration) at room temperature with shaking for 10 min and were then washed once using HEPES buffer to remove unbound PLL:FITC. Treated cells were spread on poly-_L-lysine-treated glass slides, followed by the addition of 20 μl fluorescent anti-quenching agent. Fluorescence images were observed under NIKON Ti-E inverted fluorescence microscope (Nikon Instruments Co., LTD, Japan) and quantified by ImageJ, with duplicates per strain.

The fluorescent probe 10-N-nonyl acridine orange (NAO) specifically binds to cardiolipin, and the change of cardiolipin distribution caused by cls mutation can be determined by observing the NAO fluorescence distribution on the membrane (Mileykovskaya et al., 2001). NAO fluorescence assay was performed according to previous studies (Tran et al., 2013b). In brief, the bacteria were activated in TSB overnight and transferred to new TSB to grow to exponential phase (OD₆₀₀ of ~0.5), after which NAO was added to the medium at a final concentration of 2.5 μM. The sample was oscillated at 100 rpm and 37°C for 4 h under dark conditions. The cells were washed three times (5,000 rpm, 3 min) with sterilized 0.9% saline and immediately mixed with 2× volume of fluorescent anti-quenching agent and fixed on the poly-_L-lysine-treated glass slides. Fluorescence observation was performed using a laser scanning confocal microscope with 100× objective and airyscan detector (LSM880, ZEISIS, Germany).

Membrane lipids were extracted using previous methods (Komagata and Suzuki, 1988). The lipid components of tested strains were separated by two-dimensional thin-layer chromatography (2D-TLC, Silica 60 F254 TLC plates; Merck) and stained with phosphomolybdic acid (blue for all lipids) and ninhydrin (red for lipids with amino groups), respectively, with three replicates for each strain. The first dimension was developed with chloroform/methanol/water (65:25:4, by volume), and the second dimension was developed with chloroform/acetic acid/methanol/water (80:15:12:4, by volume). The dried TLC plates were heated for 15 min in a 110°C oven to find phospholipid composition.

Bacteria were inoculated in 30-ml TSB and grew to exponential phase by shaking at 160 rpm and 37°C. The cultures were centrifuged at 5,000 rpm for 3 min to pellet cells. Cells were fixed at 4°C for 1 h with 500 μl 2.5% glutaraldehyde, washed with PBS (pH 7.2, 0.01 M) for five times (20 min each), fixed with 1% osmium acid at 4°C for 1 h, and washed with PBS (pH 7.2, 0.01 M) for three times, 15 min each. The samples were sequentially dehydrated with 30, 50, 70, 90, and 100% ethanol (twice) and replaced twice by acetone, 15 min each time. Different proportions of embedding agent (EPON812: acetone =1:3,1:1,3:1) were used to penetrate for 1 h successively, and then, pure embedding agent (100% EPON812) was used for penetration overnight. The samples were embedded in EPON812 and then subjected to temperature-programmed curing for 48 h before ultra-thin section. The sections were stained with 2% uranium acetate solution for 20 min (in dark), washed with ddH₂O three times for 5 min each, stained with 1% lead citrate solution for 10 min, and then washed with ddH₂O three times for 5 min each. The prepared samples were observed by Focused Ion Beam-Scanning Transmission Electron Microscope (Crossbeam 550, ZEISIS, Germany).

qPCRs were used to assay the copy numbers of pDL278 vector and variants in E. faecium strains (29212-pDL278, EFDO-cls, EF332-cls) to verify the stability of this vector. Plasmid levels were calculated using the 2^−ΔΔCt method. 16S rDNA was used as the housekeeping gene. The primers used are listed in Supplementary Table 1. The qPCR system (20 μl) contains the following: 10 μl of 2× SYBR green premix pro taq HS premix (Accurate Biotechnology Co., Ltd., China), 7.8 μl dd H₂O, 0.4 μl of forward and reverse primers each (10 μM), 0.4 μl of ROX reference dye (20 μM, Accurate Biotechnology Co., Ltd., China), and 1 μl template DNA (100 ng/μl). The qPCRs were performed with the following programs: denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 5 s, annealing at 60°C for 30 s. Three biological replicates were performed for each sample, and each qPCR was conducted in triplicate using the ABI StepOnePlus system (Applied Biosystems Inc., Waltham, MA, United States).

Survival analysis was performed according to the previously published method to compare the survival of different strains under daptomycin stress (Grein et al., 2020). After the bacteria grew to exponential phase (OD₆₀₀ = 0.5) in 5 ml Müller Hinton broth containing 0.05 g/L Ca²⁺ (CAMHB), 100 μl of the cultures were diluted 10⁵-fold. One hundred microliters of the diluted bacterial solution was inoculated on TSB agar plates (three replications) for CFU calculation. The remaining cultures were incubated with 32 μg/ml (4 × MIC) daptomycin (final concentration) at 37°C with shaking at 180 rpm for 1 h. During incubation, 100 μl cultures were taken every 15 min, diluted, and counted as described above. In order to avoid the influence of the change of bacterial solution volume during incubation, an equal volume of CAMHB was added after 100 μl bacterial solution was taken each time. Colony counts at different time points were recorded after overnight incubation at 37°C, and survival rate was calculated as the percentage of colony counts to those without treatment of daptomycin.

Whole-genome sequencing reads were assembled into a preliminary genome using SMRT Link v5.1.0 software (Ardui et al., 2018), and arrow software was used to optimize the assembly results. The optimized assembly results were analyzed and compared with original data to discriminate chromosomes and plasmid sequences. Final circular genomes were obtained, followed by gene prediction using GeneMarkS v4.17 software (Besemer et al., 2001). Genomic Islands (GIs) were predicted using IslandPath-DIOMB v0.12 software (Hsiao et al., 2003). Prophage prediction was completed by online prediction server PHASTER (Zhou et al., 2011; Arndt et al., 2016). Functional annotation of coding genes was performed using GO (Ashburner et al., 2000), COG (Galperin et al., 2014), NR (Li et al., 2002), Pfam (Punta et al., 2011), TCDB (Saier et al., 2013), and SWISS-PROT (Bairoch and Apweiler, 2000) databases. ARG annotation used Resistance Gene Identifier (RGI) v5.1.0 software provided by CARD database (Jia et al., 2016). Multilocus sequence typing (MLST) and plasmid classification of E. faecium EF332 were performed using MLST¹ and PlasmidFinder databases,² respectively, provided by Center for Genomic Epidemiology. AlphaFold 2 software was used to construct Cls structure model (Jumper et al., 2021). Pfam database was used to predict Cls protein domain structure (Mistry et al., 2021), and DOG 2.0 software was used to draw the protein domain structure map (Ren et al., 2009). The relative intensity of fluorescence was quantified by ImageJ 1.52a, resulting in the mean of two biological duplicates, and the significance level was analyzed by Student’s t-test. The ggplot2 package in R Studio 1.3.929 software (based on R 3.6.3) was used to draw the box charts of fluorescence intensity, and the gggenes and ggplot2 packages were used to generate the gene structure diagrams.

Enterococcus faecium EF332 strain used in this study was isolated in the Laboratory Medicine Center of the Second Hospital of Shandong University during routine screening of pathogens. Antimicrobial susceptibility tests were performed for this strain using 13 antibiotics in 10 classes by both Kirby–Bauer disk diffusion method and determination of minimum inhibitory concentration (MIC) levels. It was shown that E. faecium EF332 was resistant to most antibiotics (Table 2), suggesting that it is multidrug-resistant. A striking observation is that E. faecium EF332 is resistant to both last-line antibiotics against enterococcal infections: vancomycin and daptomycin (Bender et al., 2018). This unusual phenomenon drove us to further investigate the mechanisms underlying multidrug resistance of this strain, particularly for the two last-resort antibiotics vancomycin and daptomycin.

To fully understand the genetic basis of multidrug resistance in E. faecium EF332, whole-genome sequences were obtained with third-generation PacBio sequencing accompanied by second-generation Illumina sequencing (The sequencing data were deposited in GenBank with accession numbers CP058891-CP058895). Full sequences for a circular chromosome and four plasmids were obtained. The size of the chromosome is 2,755,510 bp, similarly to previously sequenced E. faecium strains. The four plasmids, pEF332-1, pEF332-2, pEF332-3, and pEF332-4, are, respectively, 12,341, 87,675p, 28,308, and 203,652 bp long (Supplementary Figure 1; Supplementary Table 2). Multilocus sequence typing (MLST) result showed that E. faecium EF332 belongs to ST192, one type of the CC17 clone complex (clade A1), which is believed to have evolved in infections of hospital environment (Freitas et al., 2010).

With accurate gene prediction algorithms, a total of 3,121 genes were predicted to be encoded by E. faecium EF332, of which 2,735 are chromosome-borne and 386 are plasmid-borne (Supplementary Table 2). Annotation with Comprehensive Antibiotic Resistance Database (CARD) showed that 9, 0, 12, 0, and 2 ARGs were encoded on the chromosome, pEF332-1, pEF332-2, pEF332-3, and pEF332-4, respectively (Table 3; Supplementary Table 2). Determinants for all antibiotics were found, most of which are chromosome-borne. Only pEF332-2 and pEF332-4 are antibiotic-resistant plasmids. While pEF332-4 only carries ARGs for aminoglycosides and was previously found in E. faecium VRE1 (Sun et al., 2020), pEF332-2 is a new plasmid that carries both vanA and vanM gene clusters. To the best of our knowledge, plasmids carrying both vanA and vanM genes are very rare, and only one such plasmid (pELF1) has been found so far in Japan, which is associated with high vancomycin resistance and high risk of transmission (Hashimoto et al., 2019). pEF332-2 is therefore the second plasmid that bears both vanA and vanM clusters identified so far. This is also in consistency with the high level of vancomycin resistance (128 μg/ml) observed with E. faecium EF332. In addition, vanA gene clusters are encoded on a genomic island GIs011 that is also part of a multidrug-resistant prophage (Figure 1; Supplementary Table 3). This is a strong implication that these ARGs are acquired by horizontal gene transfer. Therefore, this new pEF332-2 plasmid is a strong vancomycin-resistant plasmid that also has a strong implication for ARG mobility. Further phylogenetic analysis of E. faecium EF332 plasmids and 55 other plasmids, along with subtyping of pEF332-2 with PlasmidFinder, suggests pEF332-2 belongs to type rep2 (Figure 2). Previous report suggested type rep2 plasmids originate from a variety of sources, including hospitals, poultry, and foods (Schwarz et al., 2001; Sletvold et al., 2007; Sun et al., 2020), suggesting that these plasmids may spread between multiple environments. Nevertheless, attempts to transfer pEF332-2 from the host E. faecium EF332 strain to another bacterium via conjugation or transformation were unsuccessful, showing the limited efficiency of horizontal gene transfer.

While comparing the genomic sequences of E. faecium EF332 and daptomycin-sensitive E. faecium DO strain, mutations were found on cls and gdpD that, respectively, encode cardiolipin synthase and glycerophosphoryl diester phosphodiesterase (Table 4). It was previously reported that mutations in these genes can lead to daptomycin resistance (Humphries et al., 2012; Davlieva et al., 2013; Kelesidis et al., 2013; Lellek et al., 2015; Prater et al., 2019). However, known mutations that confer DAP resistance on these genes were not present in E. faecium EF332, leading to the hypothesis that new mutations of the two genes could lead to daptomycin resistance.

In order to find out which mutations on cls and/or gdpD lead to daptomycin resistance in E. faecium EF332 strains, a series of clones were constructed using daptomycin-sensitive E. faecalis ATCC 29212 as the parent strain: the EF332-cls strain that harbors cls from E. faecium EF332; the EFDO-cls strain that harbors cls from E. faecium EF332 with identified mutations (V₂₀₃, T₂₆₉, T₂₉₈) reverted to I₂₀₃, I₂₆₉, and S₂₉₈; the EF332-gdpD strain that harbors gdpD from E. faecium EF332; and the EFDO-gdpD strain that harbors gdpD from E. faecium EF332 with identified mutation S₂₀₁ reverted to P₂₀₁; the 29212-pDL278 strain that harbors pDL278 empty vector as control for constructed strains (Table 1). It is shown that the DAP MIC of E. faecalis ATCC 29212 increased from 1 to 8 with the introduction of pDL278 vector (Supplementary Table 4). A twofold increase in MIC values for daptomycin was found for EF332-cls strain (16 μg/ml) over the EFDO-cls strain (8 μg/ml), whereas no change of MIC values for daptomycin was found for EF332-gdpD strain (8 μg/ml) over the EFDO-gdpD strain (8 μg/ml; Supplementary Table 4). These results suggest that mutations in cls can lead to daptomycin resistance, while mutations in gdpD cannot.

Further susceptibility tests including K-B disk diffusion assays and survival assays support this finding. The inhibition zones of daptomycin with the EF332-cls strain are significantly smaller than the EFDO-cls strain (p = 2.09 × 10⁻⁵, Figure 3), while no significant difference was found between EFDO-cls strain and 29212-pDL278. Similarly, in survival assays, a significantly higher daptomycin resistance strength was found for EF332-cls in comparison with EFDO-cls, while EFDO-cls and 29212-pDL278 respond to daptomycin stress similarly (Figure 4). Additional analysis of plasmid copy numbers confirmed that introducing variants of cls did not lead to plasmid copy number fluctuation (Figure 5). It is further interesting to find that in both liquid-media-based analyses, aka the MIC determination and survival analysis, pDL278 introduction increased daptomycin resistance (Figure 4), while in the plate-based K-B disk diffusion assay, pDL278 introduction led to little difference in daptomycin resistance (Figure 3). This prompted us to hypothesize that maybe pDL278 can result in growth status-specific response to daptomycin, but without further evidence we cannot suggest the specific mechanism behind this phenomenon. Despite this difference, in all scenarios and in all assays, introducing cls from E. faecium EF332 leads to significantly stronger daptomycin resistance than introducing cls from E. faecium DO, confirming the above suggestion that mutations in cls found in E. faecium EF332 lead to daptomycin resistance.

Both cls and gdpD genes are related to phospholipid metabolism in CM: The cls gene encodes cardiolipin synthase, and gdpD gene encodes glycerophosphoryl diester phosphodiesterase (Arias et al., 2011). We therefore hypothesized that mutations in these two genes alter membrane properties, which in turn leads to alteration in membrane–daptomycin interactions. To confirm this, two-dimensional thin-layer chromatography (2D-TLC) was performed to analyze the changes of membrane lipids components caused by mutations. As shown in Figure 6, mutation in gdpD altered the aminolipid composition of CM, while mutations in cls did not significantly alter the lipid composition of CM. This finding suggests a putative role of gdpD in the biosynthesis of relatively poorly characterized aminolipid. Furthermore, membrane surface charges were quantified using positively charged fluorescence dye poly-_L-lysine conjugated to fluorescein isothiocyanate (PLL: FITC). A significant reduction of binding to PLL: FITC was found between EF332-cls and E. faecalis ATCC 29212 (p = 6.27 × 10⁻⁶⁵), as well as between EF332-cls and EFDO-cls (p = 2.20 × 10⁻¹⁰¹; Figure 7). This significant change suggests that a reduction of membrane negative charges followed cls mutations, which can lead to weakened binding of positively charged DAP-Ca²⁺ to the membrane, the target for the bactericidal effect of daptomycin. This finding suggests that cls mutations lead to daptomycin resistance by reduction of membrane negative charges. Only minor changes of membrane negative potentials were found for gdpD mutation strain (Figure 7), in consistent with the finding that gdpD mutation did not lead to increase of daptomycin resistance.

We further sought to investigate whether cls mutations also alter the distribution of cardiolipin produced by cls-encoded cardiolipin synthase. A specific cardiolipin binding fluorescent probe, 10-N-nonyl acridine orange (NAO), was used to observe the distribution of cardiolipins in CM. A redistribution of cardiolipin in CM was found (Figure 8). In E. faecalis ATCC 29212 (Figures 8A,D) and EFDO-cls (Figures 8B,E), the cardiolipin-rich regions were mainly located at the cell septum or at both poles, while in the cls-mutant strain EF332-cls, the cardiolipin-rich regions were scattered all over the cell, leading to a more even distribution and reduced abundance at septa (Figures 8C,F). This is in agreement with previous suggestion that daptomycin resistance may be resulted from redistribution of phospholipids and diverting DAP from effective targets at the septum (Heidary et al., 2018).

Pfam was used to predict Cls domain structures, leading to the finding that the identified I269T mutation in this work is located in one of phospholipase D-like domains (Figure 9). As a three-dimensional structure of Cls is not available, we predicted its three-dimensional structure by AlphaFold 2 (Figure 10). Interestingly, T269 of the Cls mutant is located at the bottom of a pocket that could potentially bind substrate for catalysis (Figure 10B). This is in agreement with domain structure binding, which leads to the suggestion that mutations of Cls may affect the phospholipase activity of this cardiolipin synthase.

Using transmission electron microscopy, we were able to compare whether cls mutations caused changes in cell envelope and septa between cls mutant and parent strains. As shown in Figure 11, for parent strain E. faecalis ATCC 29212 and mutant strain EF332-cls, the cell surfaces and septum of both strains were smooth, and both single-septum (red arrows) and multiple-septum morphology (blue arrows) were observed and their morphology were similar. This result suggests that cls mutations does not change the structure of cell envelope.