Enterococcus faecalis was isolated from clinical specimens obtained from patients at the First Hospital of Jilin University (Changchun, Jilin province, China). All of the isolated strains were confirmed to be E. faecalis using 16S rRNA and SodA sequence analysis after PCR amplification with the universal primers E1 (5′-TCAACCGGGGAGGGT-3′) and E2 (5′-ATTACTAGCGATTCCGG-3′), and FL1 (5′-ACTTATGTGACTAACTTAACC-3′) and FL2 (5′-TAATGGTGAATCTTGGTTTGG-3′) (Layton et al., 2010). E. faecalis, Enterococcus faecium, Staphylococcus aureus, and Streptococcus species were cultured in Brain Heart Infusion (BHI) broth (tryptone 10 g/l; bovine brain leaching powder 12.5 g/l; bovine heart leaching powder 5 g/l; NaCl 5 g/l; glucose 2 g/l; Na₂HPO₃⋅12⋅H₂O 2.5 g/l. Becton, Dickinson and Company, USA), whereas other species (including Bacillus subtilis, Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae) were cultured in Luria-Bertani (LB) medium (tryptone 10 g/l; NaCl 10g/l; yeast 5 g/l. Becton, Dickinson and Company, USA). All bacterial strains were stored in 30% glycerol at -80°C and routinely cultivated at 37°C.

Bacterial genomic DNA samples obtained from E. faecalis isolates were amplified using randomly amplified polymorphic DNA-analysis (RAPD-PCR), as previously described (Yang et al., 2015). The oligonucleotide that was used in the RAPD-PCR was 5′-AACGCGCAAC-3′ (Martin et al., 2005). The RAPD products were then separated using electrophoresis and analyzed using BioNumerics (Applied Maths, St. Martens-Latem, Belgium).

The antibiotic susceptibility of E. faecalis isolates was evaluated using the Kirby-Bauer disk diffusion method according to the criteria of the Clinical and Laboratory Standards Institute (Performance standards for antimicrobial susceptibility testing. M100–S20 CLSI, Wayne, PA, USA, 2010). Antibiotics that are commonly used in the hospital were selected and bought from the Hangzhou Binhe Microorganism Reagent Development Company, Hangzhou, China, including vancomycin (30 μg), erythromycin (15 μg), streptomycin (300 μg), chloromycetin (30 μg), ampicillin (30 μg), gentamicin (120 μg), tetracycline (30 μg), nitrofurantoin (300 μg), ciprofloxacin (5 μg), rifampicin (5 μg), fosfomycin (200 μg), daptomycin (15 μg), linezolid (30 μg).

The E. faecalis strain GF29 was used to isolate phages from sewage samples that were collected from the Changchun sewerage system (Gong et al., 2016). Phages were detected and purified using the double-layer agar plate method (Gu et al., 2011). The phages were amplified and stored at either 4°C or at -80°C in glycerol (3:1 [vol/vol]).

The concentration and purification of the phages was conducted as previously described with some modifications (Matsuzaki et al., 2003). After we performed large-scale culturing to grow the phage EF-P29, the lysates were centrifuged at 8,000 × g for 10 min at 4°C to remove cell debris. The phage particles were precipitated from the culture supernatants using 10% polyethylene glycol 8,000 and 1 M NaCl and then dissolved in 5 ml phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 50 mM Na₂HPO₄ and 10 mM KH₂PO₄, pH 7.4). The phage suspension was placed on top of a discontinuous CsCl gradient (1.45, 1.50, and 1.70 g/ml) and centrifuged at 35,000 × g for 3.5 h at 4°C. The phage band was then collected and dialyzed. The concentrated EF-P29 samples were placed on carbon-coated copper grids and allowed to absorb for 15 min. They were then negatively stained with phosphotungstic acid (PTA, 2% w/v). The morphology of EF-P29 was examined using a transmission electron microscope (TEM) (JEOL JEM-1200EXII, Japan Electronics and Optics Laboratory, Tokyo, Japan) at an acceleration voltage of 80 kV.

The host spectrum of the phage EF-P29 was characterized using spot testing (Chang et al., 2005) and the double-layer agar plate method (Gu et al., 2011), as previously described, with some modifications.

The multiplicity of infection (MOI) was defined as the ratio of virus particles to host cells (Gu et al., 2011). The E. faecalis strain GF29 was cultivated to log phase in BHI broth (OD₆₀₀ = 0.6) and infected with the phage EF-P29 at different MOIs (0.001, 0.01, 0.1, 1, 10, and 100). After the cultures were incubated while shaking for 8 h at 37°C, the titers of the lysates were quantified (Gu et al., 2011).

The one-step growth curve of the phage EF-P29 was determined as previously described, with some modifications (Pajunen et al., 2000). A culture of GF29 was grown to mid-exponential phase and then harvested and resuspended in fresh BHI broth. Phages were added to mid-exponential phase cultures of GF29 at an MOI of 0.01 and allowed to adsorb for 10 min at 37°C. The mixture was centrifuged at 4°C for 10 min at a speed of 10,000 × g, and the pellets were then resuspended in 10 ml of BHI. This suspension was incubated at 37°C while shaking. Samples of the mixture were obtained at 10-min intervals, immediately diluted, and then plated for phage titration.

Phage genomic DNA was extracted from concentrated and purified EF-P29 preparations using a viral genome extraction kit (Omega Bio-Tek Inc., Norcross, GA, USA). Whole genome sequencing was performed by Suzhou GENEWIZ Biotechnology Co. Ltd using IlluminaHiSeq 2500 sequencing. The bioinformatics analysis of the complete genome sequence of the phage EF-P29 was performed as previously described (Gong et al., 2016). The circle map of the EF-P29 genome was constructed using CGView¹ (Grant and Stothard, 2008). Coding sequences (CDSs) and potential open reading frames (ORFs) were predicted and analyzed using BLAST and GeneMarkS (Besemer et al., 2001). Genes encoding tRNA were identified by tRNAscan-SE (Lowe and Eddy, 1997). The gene and deduced amino acid sequences were used to perform database searches (using standard parameters) in BLASTN and BLASTP, respectively, using the National Center for Biotechnology Information (NCBI) network service (Altschul et al., 1997).

The ability of EF-P29 to inhibit the growth of the vancomycin-resistant E. faecalis (VREF) strain 002 was analyzed using colony counting. Briefly, a culture of the VREF 002 strain was grown to mid-exponential phase, harvested and resuspended in fresh BHI broth. The VREF 002 strain and previously prepared EF-P29 dilutions in PBS buffer were inoculated in 2× BHI. The mixture was diluted to make 1× BHI and then normalized to reach a final bacterial count of 10⁸ CFU/ml. The phages were added at an MOI of 0.0001, 0.001, 0.01, 0.1, or 1 in separate tubes. Consequently, the in vitro lytic efficiency of the phages could be examined. Bacterial growth in the absence of phages was analyzed as the control. The tubes were incubated at 37°C for 8 h while shaking, and the bacterial density of the cultures was measured. Each independent trial was repeated three times.

Female BALB/c mice aged 6–8 weeks old (weight, 18 to 20 g) were purchased from the Experimental Animal Center of Jilin University, Changchun, China. Water and normal mouse chow were provided ad libitum, and the mice were monitored daily. Groups of five mice per experiment were intraperitoneally (i.p.) injected with different inocula of the VREF strain 002 (1 × 10⁶, 1 × 10⁷, 1 × 10⁸, 1 × 10⁹, and 1 × 10¹⁰ CFU/mouse) to determine the minimal dose that would produce 100% mortality over a 7-day follow-up period (the minimal lethal dose [MLD]) (Zhang et al., 2013). Once the MLD was determined, 2× MLD was used as the infective inoculum (the challenge dose).

To determine the effect of the phage treatment in mice, a single dose of EF-P29 (4 × 10³, 4 × 10⁴, 4 × 10⁵, 4 × 10⁶, or 4 × 10⁷ PFU/mouse) was administered i.p. 1 h after bacterial challenge. Each dose group contained 10 mice. In the control group, the infected mice were treated with PBS buffer. The ability of EF-P29 to protect the bacteremic mice was investigated using colony counting. At predetermined intervals, bacterial counts were determined in 10 μl of peripheral blood obtained from the caudal veins of mice that were treated with either EF-P29 or buffer.

To verify the effect of phage therapy on the health of the mice, body weight was measured initially, weekly and at the end of the experiment in all of the mice (Chang et al., 2015; Wang et al., 2016). Health scores were determined using a method that was modified from a previous report (Xia et al., 2015). The mice were scored for their state of health on a scale of 5 to 0 based on disease progression at 24 h after treatment with a specific therapy. A score of 5 indicates normal health and an unremarkable condition. Slight illness was defined as decreased physical activity and ruffled fur and was scored as 4. Moderate illness was defined as lethargy and a hunched back and was scored as 3. Severe illness was defined as the aforementioned signs plus exudative accumulation around partially closed eyes and was scored as 2. A moribund state was scored as 1; death was scored as 0. The investigator was blinded to the group allocation during the experiment and when assessing the outcome.

Female Balb/c mice aged 6–8 weeks old were used in the experiments. Twenty-four mice were randomly divided into four groups (EF, EF+P, P, and WT) that contained six animals each. The mice in the EF group were challenged i.p. with 2 × 10⁹ CFU of E. faecalis strain 002 per mouse. The mice in the EF+P group were challenged with 002 as in the EF group and then injected i.p. with 4 × 10⁵ PFU of EF-P29 after 1 h. The mice in the P group were injected i.p. with 2 × 10⁹ PFU EF-P29, and the mice in the WT group were not treated. The feces of each mouse was collected at 24 h post-infection. The feces samples (24 in all) were snap-frozen in liquid nitrogen and then stored at -80°C until DNA extraction.

For each fecal sample, total genomic DNA was extracted using a QIAamp DNA stool mini kit (Qiagen, West Sussex, United Kingdom) according to the manufacturer’s instructions. DNA was frozen at -80°C prior to PCR amplification. Broadly conserved primers with the barcodes 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Evans et al., 2014) were used to amplify the V4 region (300 bp) of 16S rRNA. All PCRs were performed using Phusion@ High-Fidelity PCR Master Mix (New England Biolabs). The PCR products were purified using a Qiagen Gel Extraction Kit (Qiagen, Germany). Sequencing libraries were generated using a TruSeq^® DNA PCR-Free Sample Preparation Kit (Illumina, USA) according to the manufacturer’s recommendations, and index codes were then added. Library quality was assessed using a Qubit@ 2.0 Fluorometer (Thermo Scientific) and an Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on an Illumina HiSeq2500 platform to generate 250 bp paired-end reads.

Paired-end reads were assigned to the samples based on their unique barcode and then truncated by cutting off the barcode and primer sequence. The paired-end reads were then merged using FLASH (V1.2.7²) (Magoc and Salzberg, 2011). Quality filtering of the raw tags was performed under specific filtering conditions to obtain high-quality clean tags (Bokulich et al., 2013) using the QIIME (V1.7.0³) (Caporaso et al., 2010). The tags were compared to the reference database (Gold database⁴) using the UCHIME algorithm (UCHIME Algorithm⁵) (Edgar et al., 2011) to detect chimeric sequences, and the chimeric sequences were then removed (Haas et al., 2011).

Sequences analyses were performed using Uparse software (Uparse v7.0.1001⁶) (Edgar, 2013). Sequences with ≥97% similarity were assigned to the same OTUs. The representative sequence for each OTU was screened for further annotation. Multiple sequence alignments were constructed using MUSCLE software (Version 3.8.31⁷) (Edgar, 2004). Alpha diversity was applied to analyze the complexity of species diversity in each sample using six indices, including Observed-species, Chao1, Shannon, Simpson, ACE, and Good-coverage. All of the indices for our samples were calculated using QIIME (Version 1.7.0) and displayed using R software (Version 2.15.3). A beta diversity analysis was used to evaluate differences in species complexity between the samples. This consisted of principal component analysis (PCA), principal coordinate analysis (PCoA), and non-metric multidimensional scaling (NMDS), which were calculated using QIIME software (Version 1.7.0).

SPSS version 13.0 (SPSS, Inc., Chicago, IL, USA) was used to statistically analyze the experimental data using one-way analysis of variance. A P-value < 0.05 was considered to indicate statistical significance. The error bars represent the standard deviation of the mean.

The 16S rRNA and SodA gene fragments of 40 putative E. faecalis strains were amplified using PCR. Their sizes were approximately 733 and 360 bp according to gel electrophoresis and sequencing, respectively. The BLAST analysis showed that the SodAs of these 40 strains were more than 99% identical to those of E. faecalis strains that were identified in the GenBank database (such as CP008816.1 and AE016830.1).

Most of the E. faecalis isolates were sensitive to ampicillin and fosfomycin and resistant to erythromycin, streptomycin, chloramphenicol, tetracycline, ciprofloxacin, rifampicin, daptomycin, and linezolid (Supplementary Table S1). Interestingly, 22 of the strains were vancomycin-resistant. The results of RAPD indicated that there was genetic diversity among the E. faecalis isolates (Supplementary Figure S1). However, several of the isolates also had highly similar fingerprints (e.g., 2GF-2 and 2GF-25, GF-2 and GF-27, and GF26 and E028), and there was limited correspondence between genetic type and the observed pattern of antibiotic susceptibility.

In BHI agar plates that were covered with a bacterial lawn of E. faecalis GF29, clear plaques were observed wherever the phage lysate was spotted. This verified that the lysate did contain phages. After purification, the host ranges of five phages were determined, and the phage with the broadest host spectrum was chosen and named EF-P29. As shown in Supplementary Table S2, the spot tests showed that EF-P29 lysed 17 (42.5%) of the 40 E. faecalis isolates that were tested, and efficiency of plating tests showed that it produced large, clear plaques on 11 of the E. faecalis strains (27.5%). Based on the limited data, the phage EF-P29 was unable to lyse E. faecium or any other tested species (Supplementary Table S2). The results of electron microscopy showed that phage EF-P29 had an icosahedral head and a non-contractile tail, suggesting that it is a member of the Siphoviridae family, as shown in Figure 1A. The width of the head was approximately 60 ± 10 nm, its length was approximately 90 ± 10 nm, and its tail was approximately 105 ± 5 nm long. When the MOI was 0.01, the phage titer was the highest, reaching >10⁹ PFU/ml (Figure 1B). A one-step growth curve experiment (MOI = 0.01) revealed that the latent periods was approximately 30 min long and the burst size was estimated at approximately 90 PFU per infected cell, as shown in Figure 1C.

The complete genome sequence of EF-P29 is available at GenBank under the accession number GKY303907. The EF-P29 phage genome is a 58,984 bp contiguous sequence of linear, double-stranded DNA with an overall G+C content of 39.97%. The complete genome of EF-P29 encodes 101 putative ORFs, but no tRNA was identified. The arrangement of putative ORFs at the whole-genome level was mapped (Supplementary Figure S2A). The entire genome can be approximately divided into two distinct regions that are divergently transcribed, as shown in Supplementary Figure S2A and Table S3. As shown in Supplementary Table S3, of the identified ORFs, 39 gene products that presented an obvious similarity to proteins of known function were tentatively assigned a corresponding function. In addition, two gene products were unique and produced no database match to any publicly available viral or prokaryotic sequence. According to the functions identified for all of the predicted ORFs, the genome of EF-P29 comprises three discrete modules, including nucleotide metabolism and replication, lysis, and morphogenesis, but does not possess a lysogeny module. Of these modules, the lysis module is the smallest one, and it includes only two discrete ORFs (ORFs 62 and 81).

Phage EF-P29 shared identity at the genome level with several other E. faecalis phages, such as IME198, VD13, IME-EF1, BC-611, SAP6 and the Streptococcus phage SP-QS1, as shown in Supplementary Table S4 and Figure 2. Additionally, several ORFs showed homology at the amino acid level with putative functional proteins in other phages that infect other species, such as Bacillus and Listeria.

Open reading frame 79 and ORF 80 of EF-P29 shared high identity at both the gene sequence and amino acid level with two components of the terminase large subunit (originating from the E. faecalis phages IME-EF1, IME198, and SAP6 and the Streptococcus phage SP-QS1). However, there is a 401 bp insertion sequence between ORF 79 and ORF 80 of EF-P29 that is not present in these homologous and whole terminase large subunit genes (Supplementary Figure S2B).

The effect of the phage EF-P29 on the growth of VREF strain 002 was tested in vitro. As shown in Figure 3, the bacteria numbers increased continuously in the control group. In contrast, the VREF strain 002 showed a rapid decrease after incubation with EF-P29 at a MOI of 1, 0.1, 0.01, and 0.001. EF-P29 at the higher MOI showed more efficiently bactericidal effect. The bacteria numbers of the phage-treated cultures reached the lowest peak at different time (MOI = 1, reached 4.6 log units at 1 h; MOI = 0.1, reached 4.3 log units at 1.5 h; MOI = 0.01, reached 3.9 log units at 2 h; MOI = 0.001, reached 3.5 log units at 4 h). From this peak point, the bacteria numbers began to increase, suggesting the phage-resistant bacteria proliferated massively. In contrast, when EF-P29 was added at the MOI of 0.0001, the bacteria numbers decreased slowly (reached the lowest peak of 5.7 log units at 6 h).

Injecting ≥1 × 10⁹ CFU/mouse of VREF 002 (i.p.) was sufficient to produce bacteremia and caused a 100% mortality rate within 2 days (as shown in Figure 4A). The average MLD of 002 was approximately 1 × 10⁹ cells. When 2×MLD was used as the challenge dose, high bacteria counts (>10⁶ CFU/ml) were observed in blood samples (as shown in Figure 4B), and the infection became systemic within 1 h after challenge. Three of the mice treated with this dose had died and the others were dying within 24 h after challenge, as shown in Figure 5A.

Consequently, different doses of EF-P29 were administered (i.p.) at 1 h after the challenge inoculation. A dose-dependent effect was observed in the state of the animals after 24 h. All of the mice recovered and bacteremia was substantially reduced when 4 × 10⁵ and 4 × 10⁶ PFU of EF-P29 were administered i.p. at 1 h after VREF challenge (Figures 4A,B). When the bacteremic mice were treated with 4 × 10⁵ PFU/mouse, their health scores were higher than those of the control group, and their body weights recovered after phage treatment (Figures 5A,B). When the dose of phage was increased to 4 × 10⁶ PFU/mouse, the mice appeared healthier. A lower dose of phage (4 × 10⁴ PFU/mouse) was sufficient to protect seven out of 10 tested mice, while a dose of 4 × 10³ PFU protected only three out of 10 tested mice (Figure 4A). The mice that were treated with EF-P29 (4 × 10⁵ PFU/mouse) at 1 h after inoculation exhibited a drop in blood bacterial counts that reached approximately 3.8 log units at 24 h after treatment with EF-P29 (Figure 4B). In contrast, the bacterial loads in the control bacteremic mice, which were treated with PBS instead of phages, had reached approximately 8.5 log units at 24 h after treatment.

The intestinal microflora in the VREF challenge and phage-treated groups were analyzed. Because the experimental mice in the EF group (challenge but no treatment) would have all died at later timepoints, fecal samples were obtained from the mice at 24 h after treatment. A total of 1,613,040 previously selected effective tags were generated, and each fecal sample (n = 6 in each group) produced an average of 53,768 effective tags. The average OTUs of each sample was approximately 550. The rarefaction curve and rank abundance indicated that the sequencing depth covered rare new phylotypes and most of the observed diversity (Supplementary Figures S3A,B). UniFrac-based PCoA revealed that there was a distinctly clustered microbiota composition in each treatment group (Figure 6A). When the mice were challenged with VREF 002, the microbial composition of 4/6 of the mice shifted significantly away from the composition observed in the WT group. However, no significant shift in microbial composition was observed across the clusters observed in the P, EF+P, and WT groups, which could not be completely separated.

Taxonomic profiling demonstrated that when VREF 002 was injected i.p. into the mice, there were initially more Firmicutes and Bacteroidetes and fewer Proteobacteria increased than were observed in the normal group (Figure 6B). Notably, VREF 002 challenge altered the normal balance of the gut microbiota. Specifically, bacteria in the genus Enterococcus were present in higher numbers in the challenge group than in the normal mice (Figures 7A,B). Moreover, in comparison to normal mice, the VREF challenge mice also exhibited an enhanced variety of bacterial genera, including Streptococcus, Aeromonas, Parasutterella, and Lactococcus (Figures 6C,D and Supplementary Figure S4). At the same time, Coprococcus_1, Desulfovibrio, Moheibacter, Rhodococcus, Rhodanobacter, Lactobacillus, Ruminococcae, Anaerotruncus, Lachnospiraceae, Roseburia, Ruminiococcaceae_UCG-014, Rikenella, Lachnospiraceae_UCG-001, Ruminiclostridium_9, Ruminiclostridium_5, and Odoribacter genera were significantly reduced by VREF 002 challenge (Figure 6C). However, in the phage-treated group (4 × 10⁵ PFU/mouse), balance was largely restored in the gut microbiota. The changes that had been observed in Bacteroidetes and Proteobacteria were reversed (Figure 6B), and phage treatment reduced the level of the genus Enterococcus and family Enterococcaceae that were observed in the challenge mice to a level similar to that in the normal mice (Figures 6C, 7A,B and Supplementary Figure S5). These results suggest that the observed increases in Enterococcus in the gut microbiota were reduced by treatment with 4 × 10⁵ PFU/mouse EF-P29. Additionally, all of the genera that were different in the challenge group compared to the normal group, including Streptococcus, Aeromonas, Parasutterella, Lactococcus, Coprococcus_1, Desulfovibrio, Moheibacter, Rhodococcus, Rhodanobacter, Lactobacillus, Ruminococcae, Anaerotruncus, Lachnospiraceae, Roseburia, Ruminiococcaceae_UCG-014, Rikenella, Lachnospiraceae_UCG-001, Ruminiclostridium_9, Ruminiclostridium_5, and Odoribacter, returned to normal levels after treatment with 4 × 10⁵ PFU/mouse EF-P29 (Figure 6C). Collectively, these results show that the phage EF-P29 modulates the gut microbiota in VREF-challenged mice by suppressing the genus Enterococcus, which results in a microbiota composition that is similar to that observed in normal mice.

When a large dose of phage EF-P29 (4 × 10⁹ PFU/mouse) was administered to normal mice, bacteria in the genus Enterococcus were significantly reduced in number, which substantially altered the composition of the gut microbiota from that observed in the normal group (Figures 6C, 7A,B). Nevertheless, there was no significant difference in health scores or body weight between the normal mice that were treated with the large dose of EF-P29 (4 × 10⁹ PFU/mouse) and normal untreated mice, as shown in Figures 5A,B.