Fitness Restoration of a Genetically Tractable Enterococcus faecalis V583 Derivative To Study Decoration-Related Phenotypes of the Enterococcal Polysaccharide Antigen

E. faecalis strain VE14089 was derived from V583 cured of its plasmids. Although VE14089 had no major DNA rearrangements, it presented significant growth and host adaptation differences from the reference strain V583 of our collection. To construct a strain with better fitness, we sequenced the genome of VE14089, identified single nucleotide polymorphisms (SNPs), and repaired the genes that could account for these changes. Using this reference-derivative strain, we provide a novel genetic system to understand the role of the variable region of epa in the enterococcal lifestyle.

such as reduced immune function (3,4). Data from clinical studies indicate that intestinal colonization precedes systemic infection and that antimicrobial therapies increase the risk of colonization with vancomycin-resistant enterococci (5). More than 70 virulence factors have been identified, with different models used to study enterococcal pathogenesis (reviewed in references 6 and 7). Cell wall glycopolymers, including the ubiquitous rhamnopolysaccharide enterococcal polysaccharide antigen (Epa), have been shown to play a major role in enterococcal pathogenesis. Epa is implicated in adhesion to intestinal mucus and in translocation into epithelial cells, resistance to phagocytosis, colonization and virulence in animal models, and antibiotic resistance (8)(9)(10)(11)(12)(13). Colonizing or invasive enterococcal isolates from hospitalized patients are commonly resistant to antibiotics in general and to vancomycin treatment in particular. Indeed, most hospital isolates belong to specific high-risk enterococcal clonal complexes (HIRECC) of E. faecalis (14,15). Isolates of HIRECC-2 (clonal complex 2 [CC2]) are among the most common causes of E. faecalis infections in the United States and in several European countries (16,17). Strain V583 belongs to CC2 and was the first vancomycin-resistant enterococcus (VRE) isolate reported in the United States (18). Since the report by Hancock and Perego was published (19), several research groups have attempted to genetically modify E. faecalis reference strain V583. Due to the presence of at least two plasmids (20), genetic manipulation of E. faecalis strain V583 remains challenging (21). To avoid problems related to the presence of replicating plasmids and to simplify genetic modification in this strain, we used a plasmid-cured derivative of strain V583 named VE14089 that was obtained after novobiocin and thermic shocks (22). Strain VE14089 was generated from our clone of the V583 strain that was renamed VE14002 and that lacks the pTEF3 plasmid (23). The VE14089 strain has previously shown a growth defect in comparison to the VE14002 parental strain; the defect was more pronounced under aerobic than anaerobic conditions (22). However, with the exception of a 20.5-kb region of pTEF1 (from efa0063 up to efa0006) integrated between chromosomal genes ef3209 and ef3210, the VE14089 strain had not undergone any other major DNA rearrangements (22). Although a large dissemination of VE14089 in specialized research laboratories has been reported (see, e.g., [24][25][26][27][28][29][30][31][32], it remains characterized only partially and its genome sequence and comparison to V583 have not yet been described. Here we report the construction of the genetically tractable E. faecalis strain VE18379 representative of the enterococcal clonal complex CC2 that shows restored fitness to permit functional studies. First, modifications potentially affecting fitness of strain VE14089 were identified by whole-genome sequencing. Then, after deletion of the 20.5-kb region of pTEF1, nonsilent mutations were sequentially restored to obtain the E. faecalis VE18379 strain. Genome resequencing and in-depth phenotypic characterization confirmed that strain VE18379 growth fitness and host adaptation were restored to a level equivalent to that of the reference E. faecalis strain VE14002. Finally, as a first proof of the genetic tractability of E. faecalis strain VE18379, deletion and complementation of the 16.8-kb epa variable (epa var ) region corroborated the role of this variable region in host adaptation and resistance to antibiotics.

RESULTS
Identification of mutations in E. faecalis strain VE14089 induced by plasmid curing. The genome of the plasmid-cured E. faecalis VE14089 strain was sequenced and compared to the genome of the reference V583 strain (33). In addition to the 20.5-kb insertion between genes ef3209 and ef3210, 46 differences, including deletions, insertions, and single nucleotide polymorphism (SNPs), were detected between the VE14089 and V583 genome sequences. Of these, 34 differences were reported previously by Palmer et al. (34), leading us to consider them to be real differences which may be due to sequencing errors of the initial V583 genome and/or may reflect the evolution of independent isolates of the original V583 strain (see Table S1 in the supplemental material). The other 12 differences were confirmed after PCR amplification and sequencing of the corresponding regions of VE14089 and VE14002 (strain V583 of our collection [devoid of pTEF3]) ( Table 1). Nine differences are located within open reading frames (ORFs), and four are found in intergenic regions. One deletion and three missense SNPs are found in both strains VE14002 and VE14089, being specific to our laboratory strains. The 5-bp deletion in ef1007 generates a predicted protein of 243 residues annotated as a regulator of sugar fermentation. Noticeably, ef1007 is annotated as mutated in the V583 genome due to a frameshift mutation that corresponds to the 5 bp missing in our strains. This insertion is also absent in resequenced V583 genomes found in GenBank (accession numbers AHYN01000020.1 and ASWP01000005.1) and in all available E. faecalis genomes, suggesting an assembly error in the reference genome. The remaining eight SNPs were specific to VE14089 and are likely due to the plasmid curing treatment (Table 1; see also Fig. 1). The single nonsense SNP was found in gene ef0295, predicted to encode subunit J of a V-type ATP synthase. This protein acts as a primary ion pump transporting Na ϩ or K ϩ ions. The three other substitutions predicted as nonneutral are in annotated transcription factors. The substitution in the DNA-binding domain of the predicted sugar-binding transcriptional regulator EF0172 may affect efficient regulation of its unknown targets. E. faecalis Spx regulator EF2678 is a global transcriptional regulator, which modulates RNA polymerase specificity in response to oxidative stress under aerobic conditions. It promotes colonization of the peritoneum and dissemination in the blood in mice (35). Transcrip- tion elongation factor GreA acts on the fidelity and processivity of RNA polymerase (36). However, the effect of the substitutions on these two regulators was difficult to predict. Finally, two SNPs are predicted as silent in ef0573 and ef2060. Overall, the plasmid-cured VE14089 strain lacks pTEF2 and pTEF1 (with the exception of the 20.5-kb region integrated between ef3209 and ef3210) and harbors 6 point mutations in ORFs relative to our VE14002 strain, conferring 1 truncated gene (ef0295) and 3 missense mutated ORFs (in ef0172, ef2678, and ef2914) (Fig. 1). E. faecalis strains VE18379 and VE14002 show comparable growth characteristics. Starting from E. faecalis strain VE14089, a set of isogenic strains was constructed by removing the pTEF1 insertion and by the sequential restoration of wild-type alleles of genes ef0295, ef2914, ef0172, and ef2678 for which missense and nonsilent SNPs were observed (see details in Fig. 1; see also Fig. S1 in the supplemental material). The growth on solid and in liquid (static and aerated conditions) media of the whole set of strains (VE14089 to VE18379) was compared with the growth characteristics of strain VE14002 ( Fig. 2; see also Fig. S1). In solid medium, strains VE18379 and VE14002 showed colonies that were similarly larger than those shown by strain VE14089 (Fig. S1). A subtle but progressive increase in colony size was observed at sequential steps in restoration from VE14089 to VE18379. In liquid medium, the significant growth defect was confirmed for strain VE14089. Generation time and final cell density were progressively restored from strain VE14089 to VE18379 (Table S2). The absence of significant differences in generation time and final cell density between strains VE18379 and VE14002 confirmed the growth fitness recovery of the VE18379 to the level of the reference strain under the tested conditions, with the exception of a slight difference of the generation time under aerated conditions. The removal of a pTEF1-integrated region portion and restoration of the regulatory genes spx and greA (ef2914) strongly affected both the generation time and the cell density. These results indicated that FIG 1 Genome comparison between E. faecalis strains V583, VE14002, VE14089, and VE18379. SNPs appears in white for VE14002 compared to V583, in green for VE14089 compared to VE14002, and in red for VE18379 compared to VE14089. Accession numbers are AE016830, CP039296, and CP039548 for V583, VE14089, and VE18379, respectively. Variants identified between VE14089 and V583 were confirmed by PCR using VE14089 DNA and VE14002 DNA as the templates. IR, inverted repeat. strain VE18379 has growth properties similar to those of reference strain VE14002. Thus, strain VE18379 was considered a reference-derivative strain of enterococcal clonal complex 2 suitable for further genetic manipulations. Gene repair of SNPs and the absence of plasmid insertion were confirmed by genome sequencing of the restored E. faecalis VE18379 strain. Genome sequence comparison of VE18379 and VE14089 uncovered 5 SNPs in VE18379 (Table 2) ( Fig. 1). Three SNPs were found to be located within ORFs with no assigned function in E. faecalis (ef0077, ef1054, and ef1538). One gene (ef1538) was found to have a synonymous substitution, and the other two (ef0077 and ef1054) showed missense mutations. EF1054 is a predicted ABC transporter permease, and EF1538 is ScpA (segregation and condensation protein A). Sequencing of the intermediary strains from VE14089 to VE18375 revealed the order of appearance of SNPs. SNPs in ef1538 and tRNAAsp2 appeared after excision of pTEF1. Mutations in ef0077 and ef1054 appeared after repair of ef0295 and ef2914 (greA), respectively. The last SNP in the intergenic region between tRNALys3 and ef1451 appeared in strain E. faecalis VE18379 after repair of ef2678 (spx). In sum, each sequential event of gene repair induced an average of one mutation per cycle but no major DNA rearrangement.
Host adaptation of E. faecalis strain VE18379. The reference and derivative strains were examined in in vivo assays to assess their ability to resist the host-relevant conditions encountered by the bacteria. The Galleria mellonella insect model ( Fig. 3A) was used as a surrogate to evaluate virulence. Strains VE14002 and VE18379 induced comparable levels of lethality for the two doses tested (ϳ1 ϫ 10 6 and ϳ2 ϫ 10 6 cells). A dose of 2 ϫ 10 6 of VE14089 cells was necessary to achieve the same lethality as that seen with strains VE14002 and VE18379 at 1 ϫ 10 6 cells, demonstrating recovery of the virulence potential of strain VE18379 to a level comparable to that of strain VE14002. To examine colonization abilities in the gut, we performed orogastric inoculation of E. faecalis strains in mice harboring their complex microbiota and imbalanced with  clindamycin, a lincosamide targeting the anaerobic bacteria and favoring the proliferation of enterococci (37). To compare the colonization potentials of E. faecalis strains, we performed independent mouse experiments (n ϭ 6 to 8 mice for each strain) whose grouped results are presented in Fig. 3B. Similar results representing the behavior of the E. faecalis strains were observed postgavage but with different levels between strains. The levels of carriage of strains VE14002, VE14089, and VE18379 were significantly (C), and VE18379 (R) in the gastrointestinal tract of mice after clindamycin treatment and of reemergence after a second clindamycin treatment from D8 to D10 noted ATB D11 for VE14089 (C) and VE18379 (R) (n ϭ 23 per group), and from D11 to D13 noted ATB D14 for VE14002 (WT, n ϭ 6) and VE14089 (C, n ϭ 13). Statistical analysis was based on the Mann-Whitney test. different at day 1 (D1) after inoculation, with median values of 5.1 ϫ 10 8 , 2.1 ϫ 10 7 , and 1.2 ϫ 10 8 CFU/g, respectively. For all strains, the carriage decreased after the antibiotic treatment was stopped, corresponding to the resilience of the gut microbiota. No significant differences between the VE14002 and VE18379 strains were observed at D4, D6, and D8. On the same days, significant differences were observed between both strains VE14002 and VE18379 and the cured strain VE14089. The levels of persistence of strains VE14002 and VE18379 after 8 days were also similar (detected in 83% of mice [5/6 and 19/23, respectively]) and were markedly different from that of strain VE14089 (detected in only 30% of mice [11/36]). To assess the ability of strains to reemerge, a second treatment of clindamycin was performed at days 11, 12, and 13 for all mice inoculated with strain VE14002 and for a selection of 13 mice inoculated with strain VE14089. To mitigate the possibility of VE14089 washout, the second treatment of clindamycin was advanced to days 8, 9, and 10 for 23 mice inoculated with VE14089 or with VE18379. The level of VE14002 recovered at day 14 from all mice was equal to the level recovered on day 1. In contrast, VE14089 reappeared at lower levels in 46% of the mice, evincing the lower fitness of the cured strain with respect to reemergence within the gastrointestinal ecosystem. In contrast to strain VE14089 and similarly to strain VE14002, strain VE18379 proliferated in all mice at day 11. However, as for the D1 results, a significant difference in viable counts between VE14002 and VE18379 was observed after the second antibiotic treatment, suggesting a role for plasmids pTEF1 and/or pTEF2 in the initial step of proliferation. These results demonstrate that strain VE18379 provides a V583-like platform that is markedly improved over that represented by VE14089. VE18379 is well-adapted, in terms of colonization persistence, for further functional and basic research studies on host adaptation and intestinal colonization of a representative member of the CC2 clonal complex. Deletion of epa var region in the reference-derivative E. faecalis strain recapitulates epa-associated phenotypes. The use of isogenic strains to analyze the function of specific Epa decoration is a prerequisite for elimination of other genetic variations. Enterococcal polysaccharide antigen (Epa) biosynthesis is encoded by a locus composed of two gene clusters, one conserved and one presenting genetic variability featuring major differences in Epa between E. faecalis isolates (8,10,38). Using the reference-derivative E. faecalis strain VE18379, we constructed E. faecalis VE18395, which is devoid of the variable region of the epa locus (Δepa var ). Genome analysis of VE18395 revealed no SNPs compared to VE18379 and no major rearrangement (accession number CP039549). Growth characterization revealed no significant differences between the wild-type and deletion strains, with the exception of a lower maximum optical density (OD) for the deleted mutant (Fig. S2). The isogenic strains were subjected to a series of assays that examined the in vitro ability of the strains to resist detergents on solid medium. The deleted strain had increased sensitivity to SDS, sodium cholate (0.5%), and sodium deoxycholate (0.03%) that could be complemented with a plasmid harboring the 16.8-kb epa var region to reach the sensitivity level of the wild-type (WT) strain (Fig. 4). Morphology and cell wall structure in the Δepa var , complemented, and wild-type strains were examined using electron microscopy (EM). Strain Δepa var displayed a rounder shape than the wild-type strain and lacked the wild-type strain's peripheral electron-dense zone, called the pellicle. These morphological characters were recovered in the complemented strain (Fig. 5). Analysis of surface polysaccharides from wild-type and isogenic strains showed that the blue band corresponding to the rhamnopolysaccharide Epa disappeared in the mutant strain and was recovered in the complemented strain (Fig. 5). Antibiotic susceptibility testing of wild-type and Δepa var strains was performed in the presence of 33 different antibiotics by the standard disk diffusion method, and the MIC was determined for 3 selected antibiotics and daptomycin (Table 3). An increase in the susceptibility of the Δepa var strain was observed for vancomycin and for daptomycin and was further confirmed for daptomycin by disk diffusion in the presence of CaCl 2 (data not shown). In addition, the ΔepaX strain, deleted for EpaX only, showed a slightly increased cefoperazone MIC compared to the Δepa var strain, but no increase in the vancomycin MIC was seen. The host adaptation potential of the mutant, complemented, and wild-type strains was analyzed in in vivo assays. We compared the levels of virulence of the isogenic strains at two doses of infection in the Galleria mellonella insect model (Fig. S3). At 2 ϫ 10 6 CFU, a significant difference was observed between the complemented and deleted strains, which showed virulence attenuation in the absence of the epa var region and a tendency toward attenuation in comparisons of the deleted and wild-type strains. No difference between the strains was observed at 8 ϫ 10 5 CFU. The wild-type, Δepa var , and complemented strains were also tested in mice to evaluate their ability to colonize the gastrointestinal tract after clindamycin treatment. As observed previously, the wild-type strain transiently colonized the gastrointestinal tract, reaching a median level of 3.2 ϫ 10 8 CFU/g 1 day after oral inoculation (Fig. 6). The same level of transient colonization was observed with the complemented strain. Conversely, the Δepa var strain was significantly impaired in intestinal colonization. The loss of the epa var region reduced the level of day 1 colonization by 1,000-fold (median value of 1.4 ϫ 10 5 CFU/g). The absence of the epa var region also affected persistence, as the deletion mutant strain was below the detection level 11 days after inoculation for 13 of 23 mice. In contrast, the mice inoculated with the wild-type and complemented strains were all persistently colonized, confirming that the epa variable region is important for intestinal persistence.

DISCUSSION
Here, we report on E. faecalis V583-derived strain VE18379, which showed restored fitness compared to the corresponding VE14089 parental strain, and provide a genetic system to investigate the influence of the specific epa var region on host adaptation independently of other genetic variation between strains.
Several loci were identified as linked with fitness during construction of VE18379. This work revealed a role for the transcriptional elongation factor GreA (EF2914) in growth fitness. Yuzenkova   tion of elongation pauses. In E. faecalis OG1RF, GreA is induced in response to copper exposure (36). However, this induction may be indirect and may occur in response to a secondary stress generated by exposure to copper (i.e., oxidative stress). Similarly, global proteome patterns of E. faecalis V583 examined after a pulsed light treatment revealed a level of overproduction of GreA (40) that was consistent with an indirect role of GreA in a general stress response to environmental perturbations. Notably, four of the five mutations of the restored VE18379 strain appeared before greA repair, and no SNPs were detected after subsequent deletion of the epa var region, suggesting that repair of greA lowers the overall mutation rate. Systematic sequencing of future mutants using VE18379 will help to quantify results indicating that repair of GreA stabilized the mutation rate. The recovery of growth rate and cell density after the repair of the ef2678 gene encoding Spx reveals a role of this protein in the growth fitness of E. faecalis. Spx is a conserved major global stress regulator in low-GC Gram-positive bacteria and was first identified as a suppressor of ClpP and ClpX phenotypes in Bacillus subtilis (41). In E. faecalis OG1RF, inactivation of the spx gene impairs growth under aerobic conditions whereas it does not do so under anaerobic  conditions (35). That finding mirrors our own observations, where impaired growth was detected only under aerobic conditions. In strain OG1RF, deletion of spx resulted in a higher sensitivity to oxidative stress and to in vitro macrophage killing using cell line J774. Likewise, a role of Spx in colonization of the peritoneum and spleen in a mouse model of foreign body-associated peritonitis was also reported previously (35). Thus, fitness recovery upon spx restoration is in line with the role of Spx in host adaptation of V583 reference-derivative strain VE18379. During construction of the final VE18379 strain, we noticed improved growth rates after excision of the integrated pTEF1 plasmid sequence that suggested a biological impact of pTEF1 insertion through the activity of pTEF1-carrying genes and/or by polar effects on the surrounding genes. A putative sucrose assimilation operon, efa0069-efa0070, is present within the integrated region; however, the role of this operon and its potential metabolic advantage remain to be determined. Contiguous to this operon, genes efA0072 and efA0071 encode products corresponding to PemI/PemK family proteins that are homologous to the type II toxin-antitoxin system MazE/MazF. Toxinantitoxin systems are known to be implicated in plasmid inheritance by mediating postsegregation killing (42). In this system, the toxin causes growth inhibition and eventual death of the bacteria but is usually inhibited with its cognate antitoxin (43). The presence of PemI/PemK on the pTEF1 integrated fragment could result in part in a growth disadvantage for E. faecalis VE14089. Insertion of the plasmid sequence was located between two divergent genes, ef3209 and ef3210, each of which is the first gene of its own operon. EF3209 is a predicted ABC transporter ATP-binding protein within an ABC transporter (EF3209-EF3208), and EF3210 is a putative subunit IIA system transporter in a phosphotransferase (PTS) system (EF3210-EF3213), with a sugar specificity for mannose based on KEGG (EC 2.7. 1.191). No study has yet revealed a role of these two operons in E. faecalis V583, but transcriptomic analyses revealed very low or undetectable levels of transcripts of genes between ef3210 and ef3218 in strain V583 (23,44). In contrast, analysis of a transcriptome on strain VE14089 (ArrayExpress accession number E-MEXP-3068) revealed that these genes are overexpressed, further suggesting an additional cost for the bacterium of the pTEF1 insertion by overexpression of genes within this region (22). As the artefactual insertion no longer exists in the V583-derived VE18379 strain, it is not influencing the expression of the surrounding genes. In our study, we observed that plasmid pTEF1 or plasmid pTEF2 or both conferred an advantage in intestinal colonization at the time of the peak of antibioticinduced dysbiosis (D1) but did not confer an advantage in intestinal persistence. This is somewhat in contradiction with the deleterious effect of pTEF2 on intestinal colonization of V583. Using an in vitro assay to estimate the competitive growth of E. faecalis V583 in the human gut microbiota, Gilmore et al. (20) reported that V583 was actively killed by endogenous E. faecalis and that the killing was mediated by an heptapeptide pheromone produced by endogenous E. faecalis and was dependent on the presence of plasmid pTEF2. This discrepancy may be explained by the conditions employed for the experiments, as we worked with a mouse model using clindamycin to provide the conditions for an overgrowth of enterococci, as was observed in human clinical cases.
Genes within the epa locus have been reported to play a role in several processes such as biofilm formation, adhesion to intestinal mucus and translocation of epithelial cells, resistance to phagocytosis, virulence in infection models, and antibiotic resistance (8,10,13,45). We have shown previously that Epa is a key player in colonization of the intestine by E. faecalis (10). Specifically, we proposed a role for the predicted glycosyl transferase EpaX and, more generally, for the variable region in the decoration of Epa rhamnan backbone putatively coded by the conserved epa region. EpaX is involved in cell wall integrity and resistance to bile salts, and a related glycosyltransferase, EpaOX of E. faecalis OG1RF, contributed to biofilm-associated antibiotic resistance (8,9). The important genetic diversity between E. faecalis strains, occurring independently of the variation of the epa locus, may be a confounding factor in assessing a specific role of Epa decorations in host adaptation. Construction of isogenic strains by switching of the epa var regions could provide a way to circumvent this confounding diversity, as exemplified previously with capsular serotype-switched strains of Streptococcus pneumoniae (46). This work shows that the complete epa var region is dispensable under laboratory growth conditions and confirms a key role of epa var in host adaptation, in resistance to biliary salts and detergents and to antibiotics, and in the cell morphology. The Δepa var strain recapitulates phenotypes observed with other single-gene knockouts ΔepaX or of genes ΔepaB, ΔepaI, and ΔepaR in the conserved region of epa. Our results obtained with daptomycin support the idea of a modification in the physicochemical properties of the cell surface (charge and/or hydrophobicity) due to the Epa structure that modifies resistance. In particular, the absence of the Epa variable region confers higher sensitivity to daptomycin, in line with previous reports (8,9). Daptomycin is a cyclic lipopeptide antibiotic used to treat infections caused by staphylococci and vancomycin-resistant enterococci (47). The mode of action of daptomycin in S. aureus is that of depolarization of the bacterial membrane and is linked to increased MprF activity of phosphatidylglycerol lysinylation, modifying physicochemical properties of the membrane (48). In contrast to the results seen in studies of S. aureus, the daptomycin resistance of E. faecalis 12030 is not linked to mutation in MprF (49) but involves modification of physicochemical properties by cardiolipin synthase mutation (34), also modifying charge interaction of the bacterial surface. Interestingly, we also established a role of Epa decoration in resistance to vancomycin and to cefoperazone, two antibiotics targeting peptidoglycan synthesis, suggesting a complex form of interplay between the peptidoglycan and Epa decoration chains. Very recently, loss of Epa was shown to be associated with increased resistance to ceftriaxone (another cephalopsporin) mediated by an unknown mechanism (50,51). Of note, Hoff et al. were the first researchers to observe increased resistance to ceftriaxone of a conditional ΔepaE mutant (52). The location and mechanisms of attachment of Epa to the cell wall are not known but have been predicted to be buried within the cell envelope and linked to the peptidoglycan (53). Recently, Smith et al. (54) showed that Epa decoration is required for the peptidoglycan cross-linking and that its defect may impair cell envelope integrity. Additional studies are required to determine how Epa decoration functions in resistance to peptidoglycan-targeting antibiotics. Although the consequences of deletion of epa genes have been studied, our understanding of Epa biochemical structure, including the specificity of the secondary chains and how it affects enterococcal pathogenesis and antibiotic resistance, remains incomplete. Our system combining strain VE18395 with plasmid complementation opens avenues to evaluate the links between genetic and biochemical diversity. Construction and investigation of isogenic strains expressing diverse epa var regions will help decipher the role of specific moieties of Epa in host adaptation and in resistance to antibiotics. More generally, VE18395 provides a tractable genetic platform for this important opportunistic human pathogen.

MATERIALS AND METHODS
Bacterial strains and culture conditions. Enterococcus faecalis, Escherichia coli, and Lactococcus lactis strains used and constructed in this study are listed in Table S3 in the supplemental material. E. faecalis strains were grown in brain heart infusion (BHI) medium or M17 medium supplemented with 0.5% glucose (GM17) at 37°C. L. lactis cells were grown on GM17 at 30°C. E. coli strains were grown at 37°C in LB medium with shaking. Growth characterization of E. faecalis strains was performed in GM17 broth in under static and agitation conditions. Briefly, 10 6 bacteria from an exponential-growth-phase culture were inoculated in a volume of 40 ml in plastic tubes (Sarstedt) (50 ml) and were incubated at 37°C. Growth was monitored as the optical density at 600 nm (OD 600 ) every 60 min over 10 h. Inocula of E. faecalis for Galleria mellonella and for mouse experiments were prepared as dried frozen pellets as previously described (10,22). Erythromycin was used at concentrations of 100 g/ml for E. faecalis, 150 g/ml for E. coli, and 3 g/ml for L. lactis for selection of pGh9 and derivatives and at a at concentration of 3 g/ml for E. faecalis for pIL252 and derivatives.
Genome resequencing and identification of polymorphisms. The VE14089 genome was sequenced on a Life Technologies 5500XL NGS system at the MetaQuant platform, INRA Jouy-en-Josas (www.mgps.eu). Mapping of the sequencing reads to the E. faecalis VE14089 reference sequence was performed along with SNP analysis using Lifescope v2.0 software (Life Technologies). To identify variants (insertion, deletions, duplications), mapping data were compared to reference or parental strains by analysis using aligned BAM files in Tablet software, a graphical viewer for next-generation sequence assemblies and alignments (55). Variants identified between VE14089 and V583 (NCBI accession number AE016830) were confirmed by targeted PCR of the affected open reading frame (ORF) on VE14089 and VE14002 DNA (Table S4) followed by sequencing at GATC Biotech (France). Genomic DNA of E. faecalis VE18379 and VE18395 was sequenced on Illumina instruments at GATC Biotech (Constance, Germany), and confirmation of SNPs was performed as described for strain VE14089.
Construction of E. faecalis strain VE18379 deleted of PTEF1 fragment and with sequential repair of VE14002 genes ef0295, ef2914, ef0172, and ef2678. Construction of strain VE18379 deleted of the pTEF1 fragment and sequential repair of VE14002 genes ef0295, ef2914, ef0172, and ef2678 were performed through double crossing-over as described previously (56). The regions of interest were subjected to PCR amplification from VE14002 chromosomal DNA. PCR amplifications were made with the primers listed in Table S4. Plasmids obtained in this study are listed in Table S3. Sequencing of the deletion sites was done to confirm the excision of the pTEF1 fragment and the repair of VE14002 wild-type genes using primers detailed in Table S4.
Galleria mellonella and mouse models to evaluate enterococcal host adaptation. Galleria mellonella virulence assay was performed as described previously (22). Mouse experiments were carried out in accordance with the European guidelines for the care and use of laboratory animals (Directive 2010/63/UE). The studies received ethical approval from the local ethics committee (COMETHEA) and from the French Ministry of Higher Education and Research (no. 00680.01) for the period 2015 to 2018. The animal facility was accredited by the Direction des Services Vétérinaires (reference A78-187). Mouse experiments were performed in this study as described previously (10). To assess persistence and the ability of E. faecalis to reemerge, subcutaneous clindamycin injections were readministered daily for 3 days at days 8, 9, and 10 or at days 11, 12, and 13. To monitor the inoculated strain of E. faecalis in feces, dilutions of stool samples were plated onto BEA (bile esculin azide agar) supplemented with vancomycin at 6 g/ml. Comparisons of E. faecalis growth kinetics at each time point were performed by Mann-Whitney test (GraphPad Prism).
Construction of E. faecalis strain VE18395 deleted of the variable region of epa and construction of a plasmid for complementation of the large deletion. To optimize the genetic tractability of E. faecalis VE18379, we compared the transformation efficiencies of plasmid pGh9 propagated in E. coli JM101 (VE14037) and E. coli GM1674 (VE18916), the latter deleted for the dam and dcm genes, and in L. lactis MG1363 (see Fig. S4 in the supplemental material). A markerless deletion of the epa var region (16,832 bp) from EF2176 to EF2164 was constructed by double homologous recombination using pVE14383, a pGh9-derivative plasmid (57). Two DNA fragments flanking the 5= and 3= ends of the epa var region gene were amplified by PCR from VE14002 chromosomal DNA with primer pairs OEF823-OEF824 and OEF825-OEF826 (Table S4), respectively. The two PCR products were fused by PCR using the external primers and cloned in pGh9, yielding plasmid pVE14383, using Lactococcus lactis MG1363 repA as host. The resulting plasmid was then introduced into E. faecalis VE18379. A markerless in-frame deletion mutant was selected, and the deletion of the variable region was confirmed by sequencing of the chromosomal locus. To complement the Δepa variable region in trans, the 16.8-kb epa var region was amplified with overlapping primers in 3 fragments as described for Fig. S5 and cloned into low-copynumber Gram-positive plasmid pIL252 (58). Assembly of the four DNA molecules was performed as described previously by Gibson et al. (59) by the single-reaction method using the combined enzymatic activities of a 5= exonuclease, a DNA polymerase, and a DNA ligase. The resulting pVE14388 plasmid was electroporated into Δepa var strain VE18395 to obtain complemented strain Δepa var :pVE14388 (Cplepa var ). Sequential PCRs and a total of 34 sequencing reactions using primers detailed in Table S4 were performed to verify the sequence of plasmid pVE14388. Control strain Δepa var :pIL252, containing an empty plasmid, was also constructed. The ΔepaX and complemented strains derived from VE18379 were similarly constructed as described previously in reference 10.
In vitro characterization of E. faecalis ␦epa var and its derivatives. To test SDS and biliary salt sensitivities, E. faecalis cells were grown in BHI medium and were collected 1 h after reaching the stationary phase. Cells were washed twice with a 0.9% saline solution. Serial dilutions of E. faecalis from 5 ϫ 10 6 to 5 CFU were spotted on plates with BHI medium alone and with BHI medium containing 0.5% and 1% concentrations of sodium cholate, 0.02% and 0.03% concentrations of sodium deoxycholate, and 0.005% and 0.01% concentrations of SDS. Growth was compared in three independent experiments. Transmission electron microscopy (TEM) and polysaccharide profile analysis were performed as described previously (10). Antimicrobial susceptibility testing was performed on VE18379 wild-type, epa mutant, and complemented strains and on parental strains VE14002 and VE14089 using the disk diffusion method. The antibiotics tested were penicillin G, ampicillin, oxacillin, piperacillin, imipenem, ceftriaxone, cefoperazone, vancomycin, teicoplanin, bacitracin, polymyxin B, colistin, gentamicin, netilmicin, kanamycin, chloramphenicol, tetracycline, tigecycline, erythromycin, spiramycin, lincomycin, nitrofurantoin, ciprofloxacin, clindamycin, quinupristin-dalfopristin, linezolid, trimethoprim plus sulfamethoxazole, norfloxacin, ofloxacin, enrofloxacin, rifampin, fusidic acid, and mupirocin. Results were interpreted according to the guidelines of the EUCAST (The European Committee on Antimicrobial Susceptibility Testing) committee breakpoint tables for interpretation of MICs and zone diameters (version 9.0, 2019; http:// www.eucast.org/). The disk diffusion method was also applied with various concentrations of daptomycin (0.5 to 10 g·ml Ϫ1 ) on BHI medium supplemented with CaCl 2 at 50 mg/ml. MICs for gentamicin, vancomycin, lincomycin, cefoperazone, and netilmicin were determined by a 2-fold dilution method as described previously (60).