Bacteriophage Resistance Alters Antibiotic-Mediated Intestinal Expansion of Enterococci

Enterococcus faecalis is a human intestinal pathobiont with intrinsic and acquired resistance to many antibiotics, including vancomycin. Nature provides a diverse and virtually untapped repertoire of bacterial viruses, or bacteriophages (phages), that could be harnessed to combat multidrug-resistant enterococcal infections.

ABSTRACT

Enterococcus faecalis is a human intestinal pathobiont with intrinsic and acquired resistance to many antibiotics, including vancomycin. Nature provides a diverse and virtually untapped repertoire of bacterial viruses, or bacteriophages (phages), that could be harnessed to combat multidrug-resistant enterococcal infections. Bacterial phage resistance represents a potential barrier to the implementation of phage therapy, emphasizing the importance of investigating the molecular mechanisms underlying the emergence of phage resistance. Using a cohort of 19 environmental lytic phages with tropism against E. faecalis, we found that these phages require the enterococcal polysaccharide antigen (Epa) for productive infection. Epa is a surface-exposed heteroglycan synthesized by enzymes encoded by both conserved and strain-specific genes. We discovered that exposure to phage selective pressure favors mutation in nonconserved epa genes both in culture and in a mouse model of intestinal colonization. Despite gaining phage resistance, epa mutant strains exhibited a loss of resistance to cell wall-targeting antibiotics. Finally, we show that an E. faecalis epa mutant strain is deficient in intestinal colonization, cannot expand its population upon antibiotic-driven intestinal dysbiosis, and fails to be efficiently transmitted to juvenile mice following birth. This study demonstrates that phage therapy could be used in combination with antibiotics to target enterococci within a dysbiotic microbiota. Enterococci that evade phage therapy by developing resistance may be less fit at colonizing the intestine and sensitized to vancomycin, preventing their overgrowth during antibiotic treatment.

INTRODUCTION

Enterococci are Gram-positive commensal bacteria native to the intestinal tracts of animals, including humans (1). Under healthy conditions, enterococci exist as minority members of the microbiota in asymptomatic association with their host. However, upon antibiotic disruption of the intestinal bacterial community, enterococcal populations can flourish, resulting in elevated intestinal colonization (2, 3). As dominant members of the intestinal microbiota, enterococci can breach the intestinal barrier, leading to bloodstream infections (3). The pathogenic success of the enterococci is largely attributed to the development of multidrug resistance (MDR) traits, including the emergence of vancomycin-resistant enterococci (VRE). Enterococcus faecalis and Enterococcus faecium represent the species most commonly associated with vancomycin resistance. In the hospital, MDR enterococcal strains can be transmitted rapidly, leading to dangerous outbreaks that put immunocompromised patients at risk (4, 5). This is especially troubling as clinical VRE isolates that are resistant to recently introduced “last-line-of-defense” antibiotics have been discovered (6–10). With limited treatment options to combat the continuing rise of MDR enterococci, it is imperative to develop alternative therapeutic approaches in addition to conventional antibiotic therapy.

Bacteriophages (phages), viruses that infect bacteria, could be used for the eradication of difficult-to-treat E. faecalis and E. faecium infections. Many of these are obligate lytic phages belonging to the Siphoviridae and Myoviridae families of tailed double-stranded DNA phages (11). Current efforts in the development of phages as antienterococcal agents have focused on the treatment of systemic infections or surface-associated biofilms (12–14), although they may also be effective in decolonizing the intestines of individuals in a hospital setting.

The utility of phages as effective antienterococcal therapeutics relies on having a detailed understanding of phage infection mechanisms to understand how enterococci subvert phage infection through the development of resistance. To date, only a single membrane protein, PIP_EF (phage infection protein of E. faecalis), has been definitively identified as a phage receptor for E. faecalis (15). The enterococcal polysaccharide antigen (Epa) is involved in phage adsorption to E. faecalis cells and may act as a phage receptor (16–18). Considering that at least a dozen well-characterized lytic enterococcal phages have the potential to be used for phage therapy (11), identification of receptors used by phages could allow for the generation of more-efficient phage cocktails to be used for the treatment of enterococcal infections. This is particularly important since phage therapies can employ the use of multivalent phage cocktails to limit the emergence of bacterial resistance (19, 20), and knowledge of phage receptors can lead to rational design of such cocktails. In addition, phages often target conserved components of the bacterial cell surface, which bacteria can mutate to subvert phage infection. If a phage receptor is essential to bacterial physiology, mutation often imposes a fitness cost (21–23). Therefore, therapeutic phages could be selected that force bacterial targets to trade a fitness benefit in return for phage resistance, making them less pathogenic and possibly more susceptible to current antimicrobials (21, 24).

The E. faecalis genome contains both broadly conserved and strain-variable epa genes (25). Using E. faecalis and a collection of uncharacterized virulent phages, we identify genes located in the variable region of the epa locus to be critical for phage infection. Epa is directly involved in phage attachment to the bacterial surface. Exposure of E. faecalis to certain phages, both in vitro and in the mouse intestine, selects for mutations in the epa locus, primarily in epa variable genes. Loss-of-function mutations in two epa variable genes, epaS and epaAC, resulted in cell surface alterations that increase the sensitivity of E. faecalis to cell wall-targeting antibiotics. During colonization of the mouse intestine, an E. faecalis epaS mutant had a colonization defect in both adult mice and juvenile mice shortly following birth. The epaS mutant also failed to outgrow in the intestine efficiently upon antibiotic-mediated perturbation of the native commensal bacteria. Together, these data suggest that during enterococcal intestinal dysbiosis, phages could be harnessed to selectively modify the enterococcal population in favor of epa mutants that could be targeted more efficiently with concurrent antibiotic therapies.

RESULTS

Host range and morphology of enterococcal bacteriophages.

We obtained a library of 19 enterococcus-specific bacteriophages through the Biological Defense Research Directorate of the Naval Medical Research Center (NMRC). These phages were isolated from environmental sources as described previously (26). Phage spot agar assays were performed (27, 28) to assess phage infectivity against 19 E. faecalis strains whose susceptibility profiles for these phages were unknown. Phage lysates formed clear, opaque, or no spots against specific E. faecalis strains, indicating strong infection, weak infection, or no infection, respectively (Fig. 1A). With the exception of phi44 and phi49, each phage infected at least one E. faecalis strain, and there was host range variability among the phages. Phages phi4, phi17, and phi19 had the broadest host range, infecting more than 75% of the E. faecalis strains tested. In contrast, phages phi16, phi35, phi47, phi48, and phi51 had restricted host ranges, infecting four or fewer E. faecalis strains (Fig. 1A). Hence, this phage collection includes both broad- and narrow-host-range phages.

FIG 1.

NMRC phages have broad and narrow E. faecalis host ranges. (A) Host ranges of 19 NMRC phages against 19 different E. faecalis strains. The strains are clustered by maximum likelihood alignment of the epa variable-region nucleotide sequences, indicated by the likelihood tree to the left of the strain designations. E. faecalis ATCC 4200 was not included in the maximum likelihood estimation due to a contig gap that omits a large portion of the epa variable region. A closer distance between nodes indicates greater nucleotide sequence similarity. (B to D) Phage sensitivity profiles of E. faecalis phage-resistant isolates indicate a high degree of cross-infectivity among NMRC phages. (B) phi4-, phi17-, and phi19-resistant strains have lost susceptibility to the majority of phages that can infect the V583 parental strain. (C) Compared to the wild-type strain X98, phi51-resistant mutants gained immunity against all phages capable of infecting X98. (D) Spot assays demonstrating that phi47-resistant E. faecalis strains are now resistant to phages phi4, phi17, and phi19.

We performed transmission electron microscopy to determine the structural features of three broad-host-range and two narrow-host-range NMRC phages. Phages phi4, phi47, and phi51 belong to the Siphoviridae family of long-noncontractile-tailed phages. phi4 has a cubic icosahedral capsid symmetry (Fig. 2A), whereas phi47 and phi51 have elongated prolate capsids (Fig. 2B and E) (29). Phages phi17 and phi19 belong to the Myoviridae family with icosahedral capsids and sheathed contractile tails (Fig. 2C and D) (29).

FIG 2.

Transmission electron microscopy of NMRC phages reveals diverse morphologies. All phages imaged are double-stranded DNA phages of the order Caudovirales. Phages phi4 (A), phi47 (B), and phi51 (E) are Siphoviridae, and phages phi17 (C) and phi19 (D) are Myoviridae.

NMRC phages infect E. faecalis independent of PIP_EF.

To determine how NMRC phages infect E. faecalis, we tested their ability to infect E. faecalis BDU50, a pip_EF mutant strain of E. faecalis V583 that is resistant to phage infection (15). Phages from the NMRC collection had identical tropisms for both wild-type E. faecalis V583 and BDU50, indicating that NMRC phages infect E. faecalis in a PIP_EF-independent manner (see Fig. S1A and S1B in the supplemental material). To determine the molecular mechanism underlying NMRC phage infection, we selected E. faecalis phage-resistant isolates using both broad-host-range (phi4, phi17, and phi19) and narrow-host-range (phi47 and phi51) phages, using E. faecalis strains V583 (phi4, phi17, and phi19), SF28073 (phi47), and X98 (phi51) (Table S1). Phages were mixed with E. faecalis in top agar, poured over the surface of an agar plate, and assayed for confluent lysis. Potential phage-resistant colonies emerged within zones of lysis after overnight (O/N) incubation. The spot agar assay was used to confirm phage resistance of the isolates (Fig. 1B to D).

We next determined the extent of phage cross-resistance by testing the E. faecalis phage-resistant isolates against all other phages in the NMRC collection (Fig. 1B to D). Our data show that regardless of the phage used to select for resistance, there is broad cross-resistance to other phages from the collection (Fig. 1B to D). These data suggest that even though NMRC phages have distinct host tropisms, they likely infect through a related mechanism.

Mutations in enterococcal polysaccharide antigen genes promote phage resistance.

To determine the genetic basis of PIP_EF-independent lytic phage infection in E. faecalis, we performed whole-genome sequencing of select spontaneous phage-resistant mutants and their corresponding wild-type parental strains. In all cases, no matter which phage was used to select for resistance, phage-resistant isolates harbored mutations in the enterococcal polysaccharide antigen (epa) gene cluster (Fig. 3 and Table S1), implicating epa mutation as a key contributor to phage resistance.

FIG 3.

epa variable regions show unique gene organizations. The final epa core gene, epaR (in gray), and downstream variable genes are shown for E. faecalis strains V583, X98, and SF28073, which were used in this study. Genes in the variable region are colored according to the genome annotations and BLASTP results. A contig gap in the E. faecalis X98 scaffold is depicted by a black bar. Lollipops indicate genes where mutations were found in sequenced phage-resistant E. faecalis isolates. All the genes are drawn to scale.

The epa locus harbors genes involved in the biosynthesis of a rhamnose-containing cell surface-associated polysaccharide (30), yet the biochemical functions of most Epa proteins are uncharacterized. The E. faecalis epa gene cluster consists of a conserved core set of 18 genes (epaA through epaR [epaA–epaR]) upstream of a group of variable genes beginning at epaS (EF2176 in V583) and ending at epaAC (EF2165 in V583) (Fig. 3) (17, 25). epaR, encoding a transmembrane glycosyltransferase, was the only core gene found to be mutated (Fig. 3 and Table S1). This is consistent with a recent study demonstrating that mutation of epaR in E. faecalis OG1RF results in resistance to infection by the phage NPV1 (16). The remaining mutations in the phage-resistant isolates mapped to variable-region epa genes, including epaS, epaW, epaX, and epaAC (Fig. 3 and Table S1). epaX was recently found to aid in the adsorption of the Podoviridae phage Idefix (18). Notably, there were no predictable phage infection patterns that correlated with epa variable-region nucleotide similarity among the various E. faecalis strains (Fig. 1A).

Since E. faecalis is a native inhabitant of the intestine, we determined whether phage-driven epa mutations arose in germfree C57BL/6J mice colonized with the E. faecalis strain V583 or SF28073. Mice were treated orally with 10¹⁰ PFU of phi4 or phi47 for 7 days, after which bacteria were isolated from the feces and screened for phage resistance by plating on agar plates containing either phi4 or phi47. We sequenced 6 E. faecalis V583 isolates resistant to phi4 and 10 E. faecalis SF28073 isolates resistant to phi47. Similar to the phage-resistant isolates acquired in vitro, all in vivo phage-resistant isolates had mutations that mapped to the epa locus (Fig. 3 and Table S2). Mutations were restricted to epaR and epaS in the E. faecalis SF28073 resistant isolates. Two of the six E. faecalis phi4-resistant isolates in the V583 background had mutations that mapped to epaX and epaAC. The remaining four isolates had mutations that mapped to epaY (Fig. 3 and Table S2). Interestingly, no epaY mutations were found in any of 16 in vitro-derived epa-specific phage-resistant isolates (Table S1), suggesting that in vivo phage selective pressure may be directed toward alternative epa variable genes.

Previous studies have demonstrated that core epa genes are important for phage infection of E. faecalis (16–18). To confirm the role of epa variable genes in facilitating phage infection of the NMRC phages, we generated in-frame deletion mutants of epaS and epaAC in E. faecalis V583 using allelic replacement. We attempted to make unmarked deletions in epaX and epaW; however, we were unable to generate these mutant strains. Because epa mutations resulted in widespread resistance of E. faecalis to many of the phages in the NMRC collection (Fig. 1B), we chose to confirm these isogenic mutants using phi4. As judged by bacterial growth on agar plates containing phi4, E. faecalis strains BDU61 (ΔepaS) and BDU62 (ΔepaAC) were phenotypically indistinguishable from the spontaneous phage-resistant isolates 4RS4 and 4RS9, respectively (Fig. 4A and B). phi4 susceptibility could be restored by complementation (Fig. 4A and B). The mutants achieved a slightly lower overall culture density in the stationary phase (Fig. 4C) but had similar doubling times during logarithmic growth compared to wild-type E. faecalis V583 (31 min for the wild type, 35 min for the ΔepaS strain, and 33 min for the ΔepaAC strain). Together, these data indicate that the loss of function of epaS and epaAC alone is sufficient to confer phage resistance.

FIG 4.

Mutations in the epa locus confer phage resistance. (A and B) phi4 susceptibility assays were performed on serially diluted cultures of specific bacterial strains grown overnight (E, empty vector; C, complemented; M1FS, frameshift mutation at start methionine; V84L, missense mutation at amino acid 84 substituting valine for leucine). Dilutions were spotted onto THB agar plates with or without 5 × 10⁸ PFU/ml of phi4. Representative spot plates (A) and the corresponding quantitative viable colony counts (B) are shown. Open bars below the limit of detection in panel B occur when one or more colonies arise for a single experimental replicate at a 10⁻³ dilution. ND, none detected. (C) Growth curves comparing wild-type E. faecalis V583 and isogenic epa mutant strains in BHI broth. The dashed horizontal line indicates the limit of detection based on the bacterial plating procedure. *, P < 0.0001 (by two-way analysis of variance [ANOVA]).

We performed complementation experiments on three additional phage-resistant isolates aside from the isogenic epaS and epaAC deletion mutants. Two of the isolates had mutations in either epaR or epaW in the form of premature stop codons, and the third isolate carries an IS256 insertion sequence in epaX. The introduction of the wild-type version of each of these epa genes into the respective mutated isolate resulted in the restoration of phi4 or phi47 susceptibility (Fig. S2).

Epa dictates phage adsorption but not phage infectivity of E. faecalis.

To investigate how Epa contributes to phage infection, we tested the ability of phages to adsorb to phage-resistant or wild-type E. faecalis cells. phi4 adsorption was higher for wild-type E. faecalis V583 than for the epa mutant isolates 4RS4, 4RS9, 17RS5, 19RS21, and 19RS28 (Fig. 5A). Consistent with this observation, in-frame deletion of epaS or epaAC abrogated the adsorption of phi4 to E. faecalis (Fig. 5A). These data indicate that Epa cell wall modification is essential for phage adsorption.

FIG 5.

NMRC phages adsorb to a broad array of enterococci through Epa. (A) Wild-type E. faecalis V583 but not spontaneous epa mutants or isogenic epa deletion strains of E. faecalis efficiently adsorb phi4. (B) phi4 adsorption profile of cognate (V583, X98, and SF28073) and noncognate (CH188 and ATCC 4200) E. faecalis strains. (C) phi47 adsorption profile of cognate (SF28073) and noncognate (V583, X98, CH188, and ATCC 4200) E. faecalis strains. (D and E) phi4 and phi47 adsorption to various strains of the related bacterium E. faecium.

We next assessed whether susceptibility to infection by specific phages is dictated by the ability of phages to adsorb to the surface of E. faecalis cells. For these studies, we chose the phages phi4 and phi47, which represent both broad- and narrow-host-range phages, respectively. We selected E. faecalis strains that were sensitive and resistant to phi4 and phi47 to determine whether phage infectivity is dependent on successful adsorption or if the two events may be mutually exclusive. phi4 adsorbed to both sensitive (V583, X98, and SF28073) and resistant (CH188 and ATCC 4200) E. faecalis cells (Fig. 5B). Similarly, phi47 adsorbed to E. faecalis SF28073 and adsorbed with ∼60 to 80% efficiency to strains that it does not infect (V583, CH188, and ATCC 4200) (Fig. 1A and Fig. 5C). Interestingly, phi47 did not adsorb efficiently to E. faecalis X98. The E. faecalis X98 epa variable region harbors two open reading frames that are absent from E. faecalis strains V583 and SF28073, EFOG_00553, encoding a putative teichoic acid transport family protein, and EFOG_00558, which encodes a putative polysaccharide polymerase protein. The inability of phi47 to adsorb to the E. faecalis X98 cell surface may reflect additional cell wall modifications directed by these two coding sequences.

To determine if the promiscuous adsorption observed for phi4 and phi47 was restricted to E. faecalis, we tested the ability of these phages to adsorb to the related enterococcal species E. faecium. More than 50% of phi4 and phi47 phage particles adsorbed to E. faecium strains Com12, Com15, and 1141733 (Fig. 5D and E). E. faecium harbors an epa locus that resembles the epa locus of E. faecalis (25). From these data, we conclude that Epa is important for primary phage adsorption prior to infection. Considering that phages phi4 and phi47 adsorb to strains that are naturally resistant to infection or killing, this suggests that either abortive infection drives this resistance or an unidentified receptor required for DNA entry dictates phage infectivity.

epa variable gene mutations increase susceptibility to cell wall-targeting antibiotics and alter cell surface properties.

Previous work showed that mutation of the genes epaI, epaR, and epaOX (epaX in strain V583) increases the susceptibility of E. faecalis OG1RF to the cell membrane-specific antibiotic daptomycin (16, 31). We sought to determine if the loss of epa variable genes, epaS and epaAC, conferred similar enhanced sensitivity to cell wall-targeting antibiotics. To test this, we compared the sensitivities of wild-type E. faecalis V583 and isogenic epaS and epaAC mutants to daptomycin and vancomycin. We chose to assess vancomycin sensitivity because its mechanism of action targets cell wall biosynthesis, and E. faecalis V583 is vancomycin resistant (32). The epaS mutant showed increased susceptibility to both vancomycin and daptomycin compared to wild-type E. faecalis V583 (Fig. 6A and B). To a lesser extent, the epaAC mutant also showed increased susceptibility to both antibiotics albeit at concentrations higher than those observed for the epaS mutant (Fig. 6C and D). These data indicate that similar to other epa genes, epaS and epaAC play a role in the structural integrity of the E. faecalis cell wall during antibiotic pressure.

FIG 6.

E. faecalis epa mutant strains are more susceptible to cell wall-targeting antibiotics. Shown are antibiotic susceptibility profiles of wild-type E. faecalis V583 and the epa mutant strains BDU61 (ΔepaS) and BDU62 (ΔepaAC). Vancomycin susceptibility (A and C) and daptomycin susceptibility (B and D) of the mutants were compared to those of wild-type E. faecalis V583 and complementation strains. E, empty vector; C, complemented. *, P < 0.01; **, P < 0.001; ***, P < 0.0008; ****, P < 0.0001 (by Student’s t test).

epa mutations likely result in a modified cell wall-anchored sugar composition (16, 17). To determine if these modifications influence the overall charge of the E. faecalis cell wall, we performed a protein-binding assay using the cationic protein cytochrome c (33). Cytochrome c bound less to the epaS and epaAC mutants than to the wild-type strain (Fig. S3). Complementation restored cytochrome c binding to wild-type levels (Fig. S3). These data suggest that the cell wall of the epaS and epaAC mutants has a higher net positive charge than wild-type E. faecalis V583, confirming that loss-of-function mutations in the epa variable genes influence cell surface charge.

EpaS supports colonization and antibiotic-mediated expansion of intestinal E. faecalis.

Recently, Rigottier-Gois et al. (34) reported that mutation of the E. faecalis epa variable gene epaX, encoding a group 2 glycosyltransferase domain protein, results in an intestinal colonization defect in mice. EpaS also contains a group 2 glycosyltransferase domain. Therefore, we tested whether an epaS deletion strain has an intestinal colonization defect. Groups of conventional C57BL/6J mice were colonized with either wild-type E. faecalis V583 or an isogenic epaS mutant strain by oral gavage, followed by the addition of the bacteria to the drinking water for 18 days (Fig. 7A). On day 18, the mice were given bacterium-free water, and the E. faecalis colonization levels were monitored for 10 days. Both the wild type and the epaS mutant established persistent colonization over the 10-day period (Fig. 7B). However, beginning 3 days after the removal of bacteria from the drinking water (day 21), the epaS mutant colonized at levels ∼33-fold lower than those of wild-type E. faecalis V583 (Fig. 7B). Thus, similar to epaX, epaS is a colonization factor. This also suggests that Epa glycosyltransferases are critical for intestinal colonization.

FIG 7.

Mutation of epaS ameliorates antibiotic-mediated expansion of intestinal E. faecalis. (A) Cartoon depicting the regimen of bacterial and antibiotic exposure to mice. (B) Colonization of conventional mice with either wild-type E. faecalis V583 or the isogenic epaS mutant strain. (C) Colonization of conventional mice with either wild-type E. faecalis V583 or the isogenic epaS mutant strain. At day 21 (indicated with an arrowhead), following the introduction of bacterium-free water, the mice were orally treated with 100 μg of vancomycin daily for 4 days. The dashed horizontal lines in panels B and C indicate the limit of detection based on the bacterial plating procedure. *, P < 0.04; **, P < 0.008 (by Student’s t test with Mann-Whitney U correction).

Enterococcal intestinal dysbiosis has been linked to antibiotic use in humans, and these individuals are at an increased risk of developing enterococcal bloodstream infections (3, 35). Having observed that the epaS mutant is more susceptible to vancomycin treatment in vitro (Fig. 6A), we asked whether functional EpaS would be beneficial during antibiotic-mediated E. faecalis expansion in the intestine. To test this, we performed an experiment identical to that described in the legend of Fig. 7B, except that starting on day 21, the mice were gavaged with 100 μg of vancomycin daily for 4 days (Fig. 7C). Immediately following the first dose of vancomycin, mice colonized with wild-type E. faecalis V583 experienced a 4-log increase in E. faecalis colonization compared to mice colonized with the epaS mutant strain (Fig. 7C). During the course of vancomycin treatment, the epaS mutant remained at a colonization level significantly lower than that of wild-type E. faecalis V583. Three days after vancomycin treatment, we observed a slight bloom of the epaS mutant; however, the epaS mutant did not achieve a level of colonization similar to that of wild-type E. faecalis V583. We hypothesize that this bloom may be due to vancomycin-mediated killing of commensal bacteria, freeing up previously occupied niches that allow the epaS mutant to expand its population. Considering that the epaS mutant strain does not rebound to the same levels as wild-type E. faecalis V583 following vancomycin treatment, these data show that antibiotic-induced intestinal expansion of the enterococci requires functional EpaS.

EpaS is required for successful transmission of E. faecalis to newborn mice.

Studies suggest that offspring acquire commensal E. faecalis from the mother’s breastmilk and the vaginal tract during birth (36). However, nothing is known about the mechanisms that contribute to the ability of antibiotic-resistant E. faecalis to transmit to and colonize the intestine following birth. Therefore, we tested the ability of wild-type E. faecalis V583 and the isogenic epaS mutant strain to be transmitted to naive mouse pups born to mothers colonized with the bacteria. Female C57BL/6J mice were impregnated while continuously being exposed to wild-type E. faecalis V583 or the epaS mutant in their drinking water. After 21 days of bacterial exposure, the pregnant mothers were switched to clean water. The mothers littered their pups within 5 to 8 days after transitioning to bacterium-free drinking water. Mothers were chronically colonized for the duration of the experiment; however, the epaS mutant was maintained at a lower level than wild-type E. faecalis V583 (Fig. S4). After weaning (3 weeks after birth), we determined the levels of wild-type E. faecalis V583 and the epaS mutant in the feces of the pups. The level of recovery of wild-type E. faecalis V583 from the pups was significantly higher than that of the epaS mutant (Fig. 8). Therefore, our data suggest that EpaS is an important factor for the colonization and transmission of E. faecalis to newborns.

FIG 8.

Mutation of epaS prevents transmission to and colonization of offspring born to chronically colonized mothers. Shown are fecal abundances of wild-type E. faecalis V583 and the epaS mutant strain BDU61 from juvenile mice born to chronically colonized mothers. Data show the CFU per gram of feces on day 1 (3 weeks after birth), day 4, and day 7 postweaning. The dashed horizontal line indicates the limit of detection based on the bacterial plating procedure. *, P = 0.001; **, P < 0.0001 (by Student’s t test with Mann-Whitney U correction).

DISCUSSION

Enterococci have developed and acquired resistance to antibiotics and continue to do so. Thus, there is renewed interest in the use of phages for the treatment of MDR infections. Understanding the molecular mechanisms underlying phage-host interactions could aid in the development of phage therapies by influencing the design of effective phage cocktails. In the present study, we assessed the infectivity of lytic phages that kill E. faecalis and demonstrated that genes in the variable region of the epa locus are involved in phage infection. Exposure to phages, both in vitro and in vivo, promoted the acquisition of phage resistance. Interestingly, phage-resistant E. faecalis strains harboring loss-of-function mutations in the variable genes epaS and epaAC resulted in the sensitization of the bacteria to cell wall-targeting antibiotics. Additionally, an epaS mutant was unable to colonize the mouse intestine of adult and juvenile mice efficiently in the presence of a conventional microbiota and failed to overgrow during vancomycin treatment. This suggests that overgrowth of vancomycin-resistant E. faecalis in patients could be prevented using phage therapy. More broadly, our data suggest that phages could be used to exploit the evolution of bacterial phage resistance as an adjuvant to antibiotic therapy, in cases where acquisition of phage resistance leads to new antibiotic sensitivities.

Bacterial cell wall-anchored polysaccharides directly interact with mammalian host surfaces and are key virulence factors (17, 34, 37–41). Previous studies identified glucose, rhamnose, N-acetylglucosamine, N-acetylgalactosamine, and galactose as major components of Epa (17, 34); however, there are gaps in our understanding of the Epa cell surface architecture. Epa is produced through the action of biosynthetic enzymes encoded by select core genes residing in the epaA–epaR gene cluster (17, 34, 42). It is less clear how the genes in the variable region contribute to the overall Epa composition. We discovered that the ability of phages to infect E. faecalis is mediated through Epa. Specifically, mutations in the core gene epaR and/or variable-region genes, including epaS, epaW, epaX, epaY, and epaAC, were sufficient to abrogate phage infection. Considering that the exact contribution of many of the epa variable genes to the structure of the mature Epa polymer is unknown, we hypothesize that epa variable genes encode proteins that contribute to the decoration of the core Epa structure. Single variable gene mutations likely result in modified Epa cell surface decorations that directly influence phage adsorption, promoting resistance. Recently, it was discovered that mutation of epaR and epaX in E. faecalis prevented phage infection by excluding phage adsorption (16, 18). Here, we found that modifications made by epa variable genes are important for initial phage adsorption, although adsorption does not always lead to successful phage infection.

Phage infection occurs in three distinct stages: phage adsorption, host receptor engagement, and phage DNA replication. Considering that the phages from our study adsorb to nonsusceptible bacterial strains, it is likely that noncognate hosts either have an incompatible cognate receptor, lack the required phage receptor, or abort phage DNA replication. It is curious that all 32 phage-resistant isolates reported in this study harbored mutations only in epa genes. This observation, combined with the knowledge that the phages adsorb to noncognate host strains, indicates that E. faecalis preferentially subverts phage infection by mutating epa. We hypothesize that another factor is required for productive phage infection. This second factor is likely a bona fide phage receptor that facilitates DNA entry and may be an essential protein, as only epa mutations arose in phage-resistant E. faecalis isolates. We propose that Epa is the attachment factor that positions phages in proximity of an unidentified receptor required for DNA ejection.

Previously, we isolated phages that infect E. faecalis through the integral membrane protein PIP_EF. It is intriguing that none of the NMRC phages from this study requires PIP_EF to infect E. faecalis. Although the exact reason why PIP_EF-dependent phages do not belong to the NMRC phage cohort is unknown, we speculate that regional differences in environmental enterococci and/or enterococcal phage diversity, abundance, or infection mechanisms may be involved. NMRC phages were isolated from wastewater retrieved from the mid-Atlantic region of the Northeastern United States (Washington, DC, area), whereas the PIP_EF-dependent phages were isolated from wastewater from the Southwest United States (Dallas-Fort Worth, TX).

Previous studies have demonstrated that inactivation of core epa genes in E. faecalis affects bacterial fitness (16, 17). However, we have a limited understanding of the contribution of epa variable genes in this context. Similar to a recent report that an epaX mutant strain of E. faecalis has an intestinal colonization defect (34), we demonstrated that an epaS mutant strain is also impaired in intestinal colonization. Importantly, we show that epaS is a colonization determinant within the context of an unperturbed microbiota. This shows that Epa cell surface decorations help E. faecalis compete in a complex microbial community. In addition, an epaS mutant was impaired in its ability to transfer from mother to infant and establish productive colonization. Considering that enterococci are lifelong colonizers of humans and animals, this observation raises interesting questions about whether enterococci are transferred directly from mother to infant or if they are acquired from the environment following birth.

E. faecalis intestinal adaptation is facilitated by its inherent resistance to environmental stressors encountered in the mammalian gut, such as low pH, high osmolarity, and bile salts (16, 31, 43–45). Therefore, Epa likely plays a critical role in the survival of E. faecalis when encountering intestinal environmental stresses. A recent study showed that the variable gene epaX facilitates E. faecalis penetration of both biotic and abiotic surfaces (46), suggesting that Epa is also involved in the ability of E. faecalis to translocate from the intestine and enter the bloodstream. In our study, we observed that epa genes aid in the ability of E. faecalis to tolerate environmental stress. This is demonstrated by phage resistance and susceptibility to cell wall-targeting antibiotics upon a loss of functional epa genes. When challenged with vancomycin, an epaS mutant of E. faecalis lacks the ability to expand its population efficiently in the mouse intestine when bacterial diversity is diminished. These data suggest that the epaS mutant cannot tolerate vancomycin selection and remains a minority member of the microbiota.

In conclusion, we believe that phages like those described in this study are candidates for the development of straightforward therapeutics that could be used in conjunction with current antibiotic therapies to curtail the overgrowth of multidrug-resistant enterococci in vulnerable patients. We also believe that these data emphasize the importance of understanding phage infection mechanisms for the future development of phage cocktails. This information could help reduce the risk of developing phage resistance during therapy.

MATERIALS AND METHODS

Bacteria and bacteriophages.

A list of the bacterial and bacteriophage strains used in this study can be found in Table S3 in the supplemental material. E. faecalis and E. faecium were grown with aeration on brain heart infusion (BHI) broth or on BHI agar at 37°C. Escherichia coli was grown on Lennox L broth (LB) with aeration or on LB agar at 37°C. When necessary, for the selection of E. coli or E. faecalis, 15 μg/ml chloramphenicol (Research Products International) was added to the medium. Growth conditions for the generation of mutant strains of E. faecalis by allelic exchange were described previously by Thurlow et al. (47). Phage sensitivity assays were performed on Todd-Hewitt broth (THB) agar. The library of 19 enterococcus-specific bacteriophages was obtained through the Biological Defense Research Directorate of the Naval Medical Research Center (NMRC).

Determination of phage host range.

The lytic activities of the 19 phages from the NMRC collection were screened against 19 different E. faecalis strains using a standard spot assay (27, 28). A total of 250 μl of a 1:5 dilution of a culture of E. faecalis grown overnight (O/N) was mixed with 5 ml of THB top agar (0.35% agar) and poured onto the surface of a THB agar plate (1.5% agar). Both top agar and base agar were supplemented with 10 mM MgSO₄. Five microliters of each phage lysate was spotted onto the bacterial overlay plate. The plates were incubated at 37°C O/N, and E. faecalis sensitivity to individual phages was indicated by either clear, opaque, or no clearing spots, which indicated infection, weak infection, and no infection, respectively.

Isolation of phage-resistant E. faecalis strains.

A total of 250 μl of a 1:5 dilution of a culture of host bacteria grown O/N was mixed with 10 μl of serially diluted phage and 5 ml of prewarmed THB top agar. Phage-bacterium mixtures were poured onto the surface of THB agar plates. The plates were incubated at 37°C until phage-resistant colonies appeared in the zones of clearing. The presumptive resistant colonies were passaged four times by streaking single colonies onto BHI agar. The phage-resistant phenotypes were confirmed by spot assays (27, 28).

Phage adsorption assay.

A bacterial culture grown overnight was pelleted at 3,220 × g for 10 min and resuspended to 10⁸ CFU/ml in SM-plus buffer (100 mM NaCl, 50 mM Tris-HCl, 8 mM MgSO₄, 5 mM CaCl₂ [pH 7.4]). The cell suspensions were mixed with phages at a multiplicity of infection of 0.1 and incubated at room temperature without agitation for 10 min. The bacterium-phage suspensions were centrifuged at 24,000 × g for 1 min, and the supernatant was collected to determine the phage concentration by a plaque assay. SM-plus buffer with phage only (no bacteria) served as a control. Percent adsorption was determined as follows: [(PFU_control − PFU_{test supernatant})/PFU_control] × 100.

Antibiotic susceptibility assay.

E. faecalis was added to 7 ml of BHI broth (final density of 5 × 10⁵ CFU/ml). Daptomycin (Tokyo Chemical Industry) or vancomycin (Alvogen) was added to obtain the desired concentrations indicated in Fig. 6. Cultures containing daptomycin were supplemented with 50 mg/ml CaCl₂. Cultures were incubated at 37°C with aeration O/N. To determine viable CFU after O/N growth, cultures were serially diluted in phosphate-buffered saline (PBS), and 10 μl was spotted onto BHI agar and incubated at 37°C O/N. Viable CFU per milliliter were determined by colony counting.

Animals.

C57BL/6J (conventional and germfree) male and female mice were used for these studies. For detailed information on specific animal experiments, see the supplemental materials and methods. All animal protocols were approved by the Institutional Animal Care and Use Committee of the University of Colorado School of Medicine (protocol number 00253).

Data availability.

The DNA sequencing reads associated with this study have been deposited at the European Nucleotide Archive under accession number PRJEB30526.