Anti-virulence probiotic strategy to prevent enterococcal bacteremia

Abstract

Whether and how probiotics benefit human health is an often and controversially discussed issue. Their claimed benefit for gastrointestinal infection is linked predominantly to reducing pathogen abundance within the intestinal microbiota. Much less understood mechanistically is the reported value probiotics have in reducing systemic infections such as sepsis. Enterococcus is the epitome of an opportunistic pathogen that causes severe systemic infection as a result from translocation through the intestinal epithelium, which happens particularly in hospitalized, immune-depleted patients receiving antibiotic therapy. Here, we show that quorum-sensing controls enterococcal translocation into the blood. Furthermore, we demonstrate that orally administered probiotic Bacillus subtilis can block that process by a quorum-quenching mechanism, thereby abolishing systemic dissemination and infection. These findings demonstrate that a key aspect of Enterococcus pathogenesis is controlled by quorum-sensing and can be targeted by a combination of a probiotic with an anti-virulence approach. Notably, our study thus reveals a specific mechanism by which a probiotic may be used to prevent systemic bacterial infection and which in this or similar form may underlie the reported benefit probiotics have in reducing sepsis.

One Sentence Summary:

Pathogenesis of systemic Enterococcus infection is controlled by quorum-sensing and can be inhibited by oral administration of a Bacillus probiotic.

Introduction

Probiotics, which are dietary formulas that contain living microorganisms, are often claimed to benefit human health (1, 2). More recently, the field has started to decipher the underlying mechanisms and several studies have been published that revealed, for example, how specific probiotic bacteria reduce the abundance of enteric pathogens such as Salmonella or enterohemorrhagic Escherichia coli in the intestine of laboratory animals by stimulation of the intestinal immune system (3–6) or direct bacterial interference via bacteriocins (7, 8). We recently also showed how Bacillus subtilis probiotic spores reduce intestinal colonization by Staphylococcus aureus in mice via signaling interference (9). In contrast to probiotic interference with enteric pathogens or asymptomatic colonizers in the gut, there is much less evidence to support an impact of probiotics on systemic infections such as sepsis. Clinical trials have reported that systemic infections in infants generally are reduced after administration of probiotic bacteria (10, 11), but this phenomenon remains poorly understood on a mechanistic level.

Enterococci are among the most important nosocomial pathogens. They show widespread antibiotic resistance, particularly to vancomycin (vancomycin-resistant enterococci, VRE) (12, 13); and as a result, enterococcal infections show high morbidity and mortality (13, 14). Notably, Enterococcus is one of the most infamous examples of a gut colonizer that can cause severe systemic infection after translocation from the intestinal tract into the bloodstream (13, 15–17). Such infections occur primarily in patients who received multiple courses of antibiotics during prolonged hospitalization (13, 18). This is because such treatment can rapidly disrupt the homeostasis of the intestinal microbiota, creating a window for Enterococcus dominance in the gut (19). The resulting high intestinal loads of Enterococcus are considered an important factor facilitating enterococcal dissemination into the bloodstream and subsequent infection (15, 20, 21). Furthermore, enterococcal bacteremia usually occurs in immune-depleted patients (22), such as those with hematologic malignancy undergoing hematopoietic stem-cell transplantation (HSCT) (23). In these patients, VRE are the most common bacteria to cause bacteremia (24).

Enterococcus faecalis is, together with E. faecium, the most frequent intestinal colonizer and most important pathogen among the enterococci (18, 19). E. faecalis has a quorum-sensing system for cell density-dependent regulation of virulence called Fsr (fecal streptococci regulator) (25) that is a homolog of the staphylococcal Agr (accessory gene regulator), one of the best-studied Gram-positive quorum-sensing systems (26). The E. faecalis Fsr system is encoded by four genes, fsrA, fsrB, fsrD, and fsrC, which are homologs of the corresponding genes in the staphylococcal agr locus. The fsr locus positively regulates the transcription of two important virulence factors, the metalloproteases gelatinase (encoded by gelE) and serine protease (encoded by sprE) (25, 27–30).

We have recently shown that the Agr system controls asymptomatic intestinal colonization of S. aureus, which can be targeted by orally administered Bacillus spores that secrete Agr-inhibiting lipopeptides (9). The similarity of the E. faecalis Fsr to the staphylococcal Agr system prompted us to investigate the role of Fsr in E. faecalis intestinal colonization and infection, assuming that it may provide a means to reduce intestinal colonization or even subsequent infection by that opportunistic pathogen via a probiotic quorum-quenching mechanism using Bacillus spores. We found that – in remarkable contrast to S. aureus – the E. faecalis quorum-sensing system does not impact colonization. Rather, we discovered that it plays a key role in facilitating translocation into the bloodstream. Furthermore, we demonstrate that Bacillus spores efficiently inhibit Fsr quorum-sensing and their oral application in mice abolishes bloodstream translocation and subsequent systemic E. faecalis infection. Our data thus demonstrate how probiotics can be used to directly target a key step in the pathogenesis of systemic bacterial infection.

Results

Mouse model of E. faecalis intestinal colonization and systemic infection of intestinal origin

To monitor colonization of the intestine and associated systemic disease by E. faecalis, we set up a mouse model in part based on a previous study (31), mimicking the clinical situation of antibiotic therapy and immune depletion that commonly leads to enterococcal overgrowth and subsequent bacteremia in humans. To that end, we pre-treated mice with antibiotics, to deplete the intestinal microbiota, and cyclophosphamide (CY), which induces leukopenia and is frequently used in cancer chemotherapy (32). After pre-treatment, the mice received E. faecalis by oral gavage, and colonization and dissemination into the blood and organs, as well as additional parameters, were analyzed on the indicated days post administration. One-time gavage of 2 × 10⁹ CFU was used when analyzing translocation into the bloodstream. This was increased to two-time gavage (24 h apart) with the same amount each when analyzing subsequent infiltration into organs (liver, spleens). The latter was necessary because systemic dissemination into the organs required a higher dose not achievable with one-time gavage and a higher bacterial concentration in the gavage was not technically possible. Very high CFU in the gavage are necessary to achieve intestinal CFU that reflect the E. faecalis overgrowth that is a clinically observed prerequisite for E. faecalis bacteremia (19), because only small numbers of bacteria pass through the stomach. The specific conditions are illustrated in the figures for every mouse experiment. We generally used the most commonly studied E. faecalis wild-type strain OG1RF in our experiments, unless specifically noted when we confirmed important in-vitro and in-vivo phenotypes also with clinical VRE strains.

Lack of Fsr impact on intestinal colonization

We first analyzed whether the E. faecalis quorum-sensing system Fsr contributes to colonization of the intestine using an isogenic mutant in fsrB (ΔfsrB), an essential component of the Fsr system (29). This experiment was performed using a setup without antibiotic and CY pretreatment (Fig. 1A) and with one-time and two-time gavage models with antibiotic/CY treatment (Fig. 1B,C), representing the conditions used for the mouse experiments in this study. In all experiments, similar numbers of wild-type E. faecalis and isogenic ΔfsrB mutant were observed in the intestines, indicating that the Fsr quorum-sensing system does not influence E. faecalis intestinal colonization. These results contrast the role of S. aureus Agr, which may be explained by the much larger set of target genes that Agr controls as compared to Fsr (33, 34), which likely include yet unidentified factors controlling intestinal colonization in S. aureus. Of note, they are in accordance with the clinical observation that approximately 50% of E. faecalis isolates from patient stools are fsr-negative (35).

Fig. 1.

The E. faecalis Fsr quorum-sensing system does not impact intestinal colonization. A, Colonization without pretreatment. E. faecalis [wild-type (WT) or isogenic ΔfsrB mutant] was administered (2 × 10⁹ CFU in 200 μl each) by oral gavage, and CFU in the entire small or large intestines were determined after sacrifice 3 days afterwards. n=5/group. B, C, Colonization with pretreatment. Mice received CY for 3 days (−6 to −4), then antibiotics for 3 days (−3 to −1), after which E. faecalis [wild-type (WT) or isogenic ΔfsrB mutant] was administered on one (B) or 2 (C) consecutive days (2 × 10⁹ CFU in 200 μl each) by oral gavage, and CFU in the entire small or large intestines were determined after sacrifice 3 days afterwards (B, n=5/group, C, n=9/group,). A-C, All error bars show the geometric mean and geometric SD. No statistically significant differences were found in any WT-ΔfsrB comparison by unpaired two-tailed t (A,B) or Mann-Whitney tests (C). Parametric (t) or non-parametric (Mann-Whitney) tests were used dependent on the results of normality and log normality analyses of all data (see methods).

Fsr control of translocation through the gut epithelium

While it has been shown that antibiotic therapy can induce depletion of a specific host protein, facilitating enterococcal translocation into the bloodstream (20), the bacterial factors mediating translocation are less well understood. However, previous reports using ex-vivo experimentation indicated that E. faecalis can destroy the intestinal epithelium by gelatinase-mediated proteolytic activity (36, 37). Based on these latter reports, we hypothesized that Fsr promotes bacterial translocation from the intestine into the bloodstream. In the mouse model (Fig. 2A), viable wild-type E. faecalis and ΔsprE (isogenic mutant in the sprE serine protease gene) bacteria were detected in the blood one day after oral administration, whereas ΔfsrB, ΔgelE (isogenic mutant in the gelE gelatinase gene) or ΔgelE/sprE (isogenic double mutant in the gelE and sprE genes) bacteria were never detected (Fig. 2B). Furthermore, intestinal permeability was higher in mice that received WT or ΔsprE than ΔfsrB, ΔgelE, or ΔgelE/sprE bacteria (Fig. 2C). Finally, we did not observe differences in leukocyte numbers in the blood between wild-type- and ΔfsrB-infected mice, a finding that, given the central importance of leukocytes for bacterial elimination, is in accordance with the notion that the differences in CFU in the blood were due to differences in translocation rather than subsequent elimination (Fig. 2D). These results indicate that the Fsr quorum-sensing system is critical for E. faecalis dissemination from the GI tract into the bloodstream and its primary mediator is the Fsr-controlled gelatinase.

Fig. 2.

In-vivo translocation of E. faecalis through the gut epithelium is due to quorum-sensing control of gelatinase. A, Experimental setup. Mice were pre-treated with cyclophosphamide (CY) (days −6 to −4) and an antibiotic mix (days −3 to −1) and received E. faecalis wild-type (WT) or the indicated isogenic mutants by oral gavage (2 × 10⁹ CFU in 200 μl) the next day. 24 h later, mice were sacrificed, the blood was collected by cardiac puncture and CFU were determined by plating. B, CFU in the blood. n=5. Data under the detection limit (LOD, dashed line) were set to 0 and are depicted below the x axis. Statistical analysis is by Poisson regression versus WT. C, Permeability of the gut epithelium. Permeability was determined by administration of FITC-dextran 4 h before sacrifice. n=5. Statistical analysis is by 1-way ANOVA with Dunnett’s post-test versus values obtained with the WT. D, Leukocyte numbers in the blood (WBC, total white blood cells; Neu, neutrophils; Lymph, lymphocytes; Mono, monocytes). n=5. Statistical analysis is by 2-way ANOVA with post-tests versus values obtained with the WT within each leukocyte group. B, Error bars show the geometric mean and geometric SD. C,D, Error bars show the mean ± SD.

Fsr control of systemic infection of intestinal origin

Focusing on the Fsr system and GelE as the main Fsr-controlled effector mediating epithelial translocation, we then monitored the development of systemic infection by determining bacterial CFU in the blood and organs at days 1 and 3 after administration of wild-type, ΔfsrB, or ΔgelE bacteria by oral gavage (Fig. 3A). While only wild-type E. faecalis but not ΔfsrB or ΔgelE mutant bacteria were detected at day 1 in the blood (similar to the results achieved with one-time gavage shown in Fig. 2B), no bacterial CFU were detected in the organs at that time point (Fig. 3B). In contrast, we could not detect any CFU in the blood after 3 days post gavage, indicating that bacteria were efficiently cleared from the bloodstream at that time by the immune system despite immune depletion. However, at day 3 we detected considerable amounts of wild-type E. faecalis CFU in the livers and spleens of animals, indicating systemic infection (Fig. 3B). Notably, we could not detect any CFU of the ΔfsrB or ΔgelE mutants in the blood or organs at either of the two time points. Measurement of bacterial epithelial translocation in this experiment reproduced the results of the one-time gavage experiment shown in Fig. 2C (Fig. 3C). Furthermore, we performed histological analyses of the epithelium of the small intestine, which showed severe structural compromise with a lack of crypts, villi, and Goblet cells and infiltration of immune cells in mice infected with the wild-type E. faecalis strain, while epithelia of mice infected with the ΔfsrB or ΔgelE mutant strains appeared normal (Fig. 3D). These data demonstrate that intestinal infection by E. faecalis bacteria has a profound, Fsr quorum-sensing-dependent, GelE-mediated deleterious impact on the structural integrity of the intestinal epithelium and that Fsr and the Fsr-regulated effector GelE are crucial for systemic dissemination of E. faecalis bacteremia of intestinal origin.

Fig. 3.

The E. faecalis Fsr quorum-sensing is essential for translocation through the intestinal epithelium and subsequent systemic infection. A, Mice received CY for 3 days (−6 to −4), then antibiotics for 3 days (−3 to −1), after which E. faecalis (WT, ΔfsrB, or ΔgelE isogenic deletion strains) were administered on two consecutive days (2 × 10⁹ CFU in 200 μl each) by oral gavage. On day 1 or 3 the indicated parameters were determined in two different experiments (n=9–10/group). B, CFU in blood and organs. Data under the detection limit (LOD, dashed line) were set to 0 and are depicted below the x axis. Statistical analysis is by Poisson regression versus WT. C, Permeability of the gut epithelium. Permeability was determined by administration of FITC-dextran 4 h before sacrifice. Statistical analysis is by Kruskal-Wallis and Dunn’s multiple comparisons tests. D, Histological examination of intestinal epithelia (jejunum). Areas encircled with white dashed lines mark examples of areas with pronounced immune cell infiltration. Green arrows, crypts; yellow arrows, villi. B, Error bars show the geometric mean and geometric SD. C, Error bars show the mean ± SD.

The Fsr system and its regulatory targets gelatinase and serine protease have been shown to impact virulence in endophthalmitis and peritonitis mouse as well as systemic Caenorhabditis elegans infection models (25, 29, 38, 39). It is therefore possible that the organ invasion and bacteremia effects we detected may be impacted by a gut epithelium-translocation-independent, direct effect on systemic virulence. However, it is not known whether Fsr also impacts bacteremia and associated organ invasion in vertebrates. We therefore first tested whether Fsr has a direct effect on persistence and dissemination into organs in a murine bacteremia model, in which we injected the bacteria directly into the bloodstream (intravenously, without any antibiotic or CY pretreatment). Three days post infection, there was a significantly higher bacterial load, at ~ 2 logs difference, in the liver and kidneys when mice received wild-type as compared to the fsrB mutant strain (Fig. 4A). Interestingly, this effect appeared to be due to a combination of GelE and SprE activity, as both the ΔgelE and ΔsprE mutants showed significantly diminished presence in the organs. The fact that the ΔgelE/sprE double mutant was similarly impaired as the ΔfsrB mutant indicates that these two proteases are the major factors underlying Fsr-dependent bacteremia and associated organ invasion. Thus, the Fsr quorum-sensing system has a significant, GelE- and SprE-mediated direct impact on avoiding clearance by immune cells during systemic E. faecalis blood infection in mice.

Fig. 4.

The impact of the E. faecalis Fsr quorum-sensing system on experimental systemic infection of intestinal origin is mainly due to translocation effects. A, Direct impact on systemic infection. Mice received 2 × 10⁸ CFU E. faecalis wild-type (WT) or the indicated isogenic mutant bacteria by intravenous injection, without any CY or antibiotic pre-treatment. Three days post infection cardiac puncture was performed, and the blood, liver and kidneys were collected. There were no bacteria found in the blood. n=4–5. Statistical analysis is by two-tailed unpaired Mann-Whitney test. B, Co-colonization model. Mice received CY for 3 days (−6 to −4), then antibiotics for 3 days (−3 to −1), after which E. faecalis WT and ΔfsrB bacteria were administered in equal amounts (1 × 10⁹ CFU in 200 μl each) twice (on two consecutive days) by oral gavage and compared to separate infection with 2 × 10⁹ WT or ΔfsrB bacteria (also given twice on two consecutive days). On day 3 the indicated parameters were determined (n=5/group). C, Histological examination of the intestinal epithelium (jejunum). Areas encircled with white dashed lines mark examples of areas with pronounced immune cell infiltration. Green arrows, crypts; yellow arrows, villi. D, CFU in the blood and intestines. Statistical analysis is by Mann-Whitney tests (intestines) or Poisson regression (organs, blood). Data under the detection limit (LOD, dashed line) were set to 0 and are depicted below the x axis. A,D, Error bars show the geometric mean and geometric SD.

Then, to analyze whether the systemic infection effects that we detected in our intestinal-origin infection model were due to differential epithelial translocation in the gut or a direct impact on systemic infection, we set up a co-colonization model in which mice received equal amounts of wild-type and ΔfsrB mutant E. faecalis simultaneously and compared directly to mice that received the same amounts of only wild-type or ΔfsrB mutant E. faecalis (Fig. 4B). When co-colonizing, the deficiency that the ΔfsrB mutant bacteria have in transgressing through the gut epithelium can be assumed to be overcome, at least to some extent, by Fsr-controlled gelatinase that is secreted from co-colonizing wild-type bacteria, a notion supported by the finding by Roberts et al. that neither fsr nor gelatinase is associated with disease origin of E. faecalis isolates (40). That this assumption is correct was first confirmed by histological analysis, which showed similarly strong compromise of the intestinal epithelium in co-colonization and separate colonization by wild-type E. faecalis, while the intestinal epithelium in mice colonized by ΔfsrB bacteria appeared normal (Fig. 4C). If the Fsr impact was mostly due to a post-translocation systemic effect, one would expect similar effects on bacterial CFU in blood and organs as we showed in Fig. 3 in the separate intestinal-origin systemic infection model also in the co-colonization model, while greatly diminished or absent differences would suggest that the main reason for the observed effects were differences in translocation. In stark contrast to the highly significant difference in the separate colonization model, we detected no significant differences in organ bacterial load between wild-type and ΔfsrB bacteria in the co-colonization model (Fig. 4D). These data strongly suggest that the Fsr-dependent differences in systemic infection of intestinal origin that we observed were due to differential capacity for translocation through the intestinal epithelium rather than systemic effects after translocation.

Efficient inhibition of Fsr by B. subtilis lipopeptides

Interference with quorum-sensing control is considered a promising way for anti-virulence therapy (41–43). Fsr inhibitors have been identified (44, 45), but their therapeutic usefulness in particular for systemic infection is unclear. Treatment using probiotic bacteria with quorum-quenching activity constitutes an interesting alternative to the application of synthetic quorum-quenchers and has many advantages, such as most importantly the constantly maintained production of the quorum-quenching molecules. However, direct systemic administration of a live microorganism obviously is not a therapeutic option. On the other hand, orally administered probiotics are often claimed to reduce systemic infection, especially in neonates (10), but how this is achieved other than by preventing general dysbiosis (46) is largely unclear. We recently showed that orally administered probiotic B. subtilis spores, which germinate in the gut to produce Agr-inhibiting lipopeptides called fengycins, can eliminate S. aureus intestinal colonization via quorum-quenching (9). Together with these previous results (9), the findings presented herein suggested that blocking the Fsr system by probiotic intervention in the intestine via orally administered B. subtilis spores may directly inhibit the development of enterococcal bacteremia by decreasing the capacity of intestinal E. faecalis to disseminate into the bloodstream.

To test this hypothesis, we first analyzed whether B. subtilis culture filtrate can inhibit the E. faecalis Fsr system. As readouts for Fsr activity, we measured proteolytic activity fsr and gelE/sprE gene expression. In addition to the standard E. faecalis OG1RF strain used in the previous experiments, we included the clinical isolates E. faecalis V583, the first vancomycin-resistant Enterococcus (VRE) clinical isolate reported in the United States (47), and E. faecalis TX0104, a VRE isolate from a case of bacteremia. All strains showed similar in-vitro growth patterns, while proteolytic activity fsr and gelE/sprE gene expression was lower in V583 than in the other two strains (Fig. 5A). B. subtilis culture filtrate suppressed proteolytic activity in all E. faecalis strains, indicating an efficient quorum-quenching effect (Fig. 5B). Further proteolysis assays and qRT-PCR measurements of fsrBDC and gelE/sprE expression performed with isogenic deletion mutants in fenA (fengycin-deficient, ΔfenA), srfA (deficient in the structurally similar surfactin lipopeptides, ΔsrfA) (9), and sfp (deficient in both fengycins and surfactins, Δsfp) (48) as well as with purified fengycin and surfactin, pinpointed the inhibitory activity to those lipopeptides. They also showed that – in contrast to the fengycin-specific quorum-quenching activity toward S. aureus Agr (9) – E. faecalis Fsr is inhibited by both B. subtilis lipopeptide classes (Fig. 5B–F). Attenuation of quorum-quenching by the extracellular signal peptide of the E. faecalis Fsr system, GBAP (gelatinase biosynthesis-activating pheromone), in a dose-dependent fashion demonstrated Fsr specificity of the effect (Fig. 5G–I). Of note, in all experiments with B. subtilis culture filtrates or pure lipopeptides, there were no lipopeptide-dependent differences in E faecalis growth at the used volumes or concentrations (Supplementary Fig. S1).

Fig. 5.

B. subtilis fengycin and surfactin lipopeptides inhibit E. faecalis Fsr quorum-sensing. A, Growth, proteolytic activity, fsr and gelE/sprE gene expression during growth of the indicated E. faecalis strains. B, Suppression of proteolytic activity by culture filtrates of the indicated strains. C,D Suppression of proteolytic activity by purified fengycin (C) or surfactin (D). E,F Suppression of fsrBDC (E) or gelE/sprE (F) expression by purified fengycin or surfactin (10 μM). G, Growth and proteolytic activity during growth of E. faecalis OG1RF under influence of the Fsr-autoinducing peptide GBAP. Statistical analysis is by unpaired two-tailed t-tests. H,I Reversion of the fengycin- (H) or surfactin- (I) caused suppression effect on E. faecalis OG1RF proteolytic activity by GBAP. B-F,H,I, Statistical analysis is by 1-way ANOVA with Dunnett’s post-tests versus the respective control groups. A-I, n=3. All error bars show the mean ± SD, except for growth curves, where error bars show the geometric mean and geometric SD. B-I, *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001. Proteolytic activity or gene expression was determined in cultures grown for 3 h.

Inhibition of E. faecalis epithelial translocation and systemic infection by Bacillus probiotic

B. subtilis is used as a probiotic regimen in the form of orally administered spores. In notable contrast to many other probiotic medications, the spore character of B. subtilis probiotics offers the benefit of increased resistance to stomach passage (49, 50). To directly test whether application of B. subtilis spores can limit translocation of E. faecalis from the intestine into the bloodstream, we fed mice spores of wild-type B. subtilis or its isogenic Δsfp mutant and a day later challenged with E. faecalis. Because of their more pronounced gelatinase activity (see Fig. 5A), we used strains OG1RF and TX0104 for this experiment (Fig. 6A). Remarkably, in mice that received B. subtilis wild-type spores, E. faecalis translocation into the bloodstream was completely inhibited (Fig. 6B) and epithelial permeability strongly decreased (Fig. 6C). This was not the case when administering Δsfp mutant spores, showing that the effects were due to B. subtilis lipopeptide production. Furthermore, there was no difference in leukocyte numbers between mice receiving B. subtilis wild-type and Δsfp mutant spores, or PBS, in mice that received wild-type or ΔfsrB E. faecalis, indicating that the effect was not due to divergent immune responses (Fig. 6D).

Fig. 6.

Oral administration of B. subtilis spores decreases E. faecalis epithelial translocation and eliminates E. faecalis bacteremia by a quorum-quenching mechanism. A, Experimental setup. Mice were pre-treated with CY (days −6 to −4) and an antibiotic mix (days −3 to −1) and received B. subtilis spores (of wild-type or fengycin/surfactin lipopeptide-deficient Δsfp mutant) by oral gavage on the two subsequent days (2 × 10⁷ CFU in 200 μl each), and E. faecalis OG1RF, OG1RF ΔfsrB, or TX0104 (2 × 10⁹ CFU in 200 μl) the next day. 24 h later, mice were sacrificed, the blood was collected by cardiac puncture and CFU were determined by plating. Epithelial permeability was determined by administration of FITC-dextran 4 h before sacrifice. n=5/group. B, CFU in the small and large intestine and the blood. Data under the detection limit were set to 0 and are depicted on/below the x axis. Statistical analysis is by 1-way ANOVA or Kruskal-Wallis test dependent on the result of normality and log-normality tests of the data to be compared (intestines) or Poisson regression (blood). No comparison in the intestine data was statistically significant. n=4–5. C, Epithelial permeability. Permeability was determined by administration of FITC-dextran 4 h before sacrifice. Statistical analysis is by 1-way ANOVA and Tukey’s post-tests (OG1RFΔfsrB, TX0104) or Kruskal-Wallis test with Dunn’s multiple comparisons test (OG1RF). n=5/group. D, Leukocyte numbers. WBC, total white blood cells; Neu, neutrophils; Lymph, lymphocytes; Mono, monocytes. n=4–5. B, Error bars show the geometric mean and geometric SD. C,D, Error bars show the mean ± SD.

Then, we set up an additional experiment with strain OG1RF in which we measured intestinal colonization, epithelial translocation, white blood cell count, bacteremia, and organ infiltration, at days 1 and 3 post oral gavage with E. faecalis. To achieve systemic infection and organ infiltration, we had to increase the E. faecalis dose in comparison to the previous experiment that only monitored translocation into the blood, by administering the same CFU twice on subsequent days (Fig. 7A). Our data showed that oral intake of Bacillus spores, but not of those deficient in lipopeptide production, completely prevented epithelial translocation, compromise of the intestinal epithelium, bacteremia, and subsequent organ infiltration by E. faecalis, while having no impact on E. faecalis intestinal colonization (Fig. 7B–E).

Fig. 7.

Oral administration of B. subtilis spores decreases systemic E. faecalis infection of intestinal origin via quorum-quenching. A, Experimental setup. Mice were treated with CY for 3 days (−6 to −4), then with antibiotics for 3 days (−3 to −1), and received B. subtilis spores (of wild-type or fengycin/surfactin lipopeptide-deficient Δsfp mutant) by oral gavage on the two subsequent days (2 × 10⁷ CFU in 200 μl each), and E. faecalis OG1RF on the next two days (2 × 10⁹ CFU in 200 μl each). CFU in the small and large intestine, organs, and the blood were determined after sacrifice on day 1 or 3 in two different experiments (n=8/group, n=10/group, respectively). Epithelial permeability was determined by administration of FITC-dextran 4 h before sacrifice. Histological analysis of the intestinal epithelium (jejunum) was performed. B, CFU in the intestines, organs (liver, spleen) and the blood. Non-plotted data were under the detection limit (20, dashed line) and set to 0. Statistical analysis is by Poisson regression. C, Epithelial permeability. Statistical analysis is by Friedman and Dunn’s multiple comparisons tests. D, Leukocyte numbers. WBC, total white blood cells; Neu, neutrophils; Lymph, lymphocytes; Mono, monocytes. E, Histology. Areas encircled with white dashed lines mark examples of areas with pronounced immune cell infiltration. Green arrows, crypts; yellow arrows, villi. B, Error bars show the geometric mean and geometric SD. C,D, Error bars show the mean ± SD.

We performed a series of further control experiments to confirm the validity of these findings. First, we verified that B. subtilis intestinal colonization remained substantial over the 3 days post inoculation when E. faecalis was applied in the experimental schemes employed in Fig. 6 and 7 (Supplementary Fig. S2). Furthermore, we ruled out that B. subtilis lipopeptide production has an effect on B. subtilis intestinal colonization by confirming that intestinal colonization by wild-type and Δsfp B. subtilis was not different under the used experimental conditions (Supplementary Fig. S2). Also, B. subtilis has been reported to exert immune-stimulatory effects in human trials (51, 52), which in mouse experiments have been linked to lipopeptide-unrelated mechanisms (53, 54). Our results all demonstrate lipopeptide specificity, but to rule out the hypothetical possibility that lipopeptides have yet unrecognized immune-stimulatory properties that affect the phenotypes assessed in this study, we determined leukocyte numbers in the blood and performed histological analysis of the intestinal epithelium in mice receiving B. subtilis wild-type or Δsfp mutant spores, or PBS, the absence of E. faecalis. (Analyses in the presence of E. faecalis are shown in Fig. 6D and 7D). We did not observe changes in leukocyte numbers or appearance of the intestinal epithelium (Supplementary Fig. S2), in support of the notion that the strong B. subtilis lipopeptide-mediated effects we observed in the presence of E. faecalis are not due to a hypothetical lipopeptide-mediated immune-stimulatory mechanism independently of quorum-sensing interaction or the interaction with E. faecalis. Finally, it is important to note that we never found B. subtilis in the blood, which is noteworthy given reports that orally administered probiotics can themselves cause bacteremia (55).

Together, these results show that orally administered probiotic B. subtilis spores can prevent the development of E. faecalis bacteremia and systemic infection in a lipopeptide-dependent manner by inhibiting E. faecalis-caused harm to the intestinal epithelium and subsequent translocation into the bloodstream.

Discussion

In this study we show that in the important nosocomial pathogen E. faecalis, translocation through the intestinal epithelium, an established major route of enterococcal sepsis development (13, 15–17), is regulated by the Fsr quorum-sensing system. While we found that the Fsr system also has a direct impact on blood infection, our results indicate that the main effect that Fsr has on systemic E. faecalis infection of intestinal origin occurs by its control of epithelial translocation (Fig. 8).

Fig. 8.

Model of Fsr quorum-sensing and B. subtilis probiotic-mediated reduction of E. faecalis bacteremia. The Fsr system comprises the fsrA, fsrB, fsrD, and fsrC gene operon. FsrD is secreted and post-translationally modified by FsrB, to the mature GBAP peptide, which stimulates the FsrC-FsrA two-component system by binding to the FsrC transmembrane histidine kinase receptor. The activated FsrA response regulator then increases transcription of the fsrABDC locus (quorum-sensing feedback loop) in addition to that of the two target genes, gelE (gelatinase) and sprE (serine protease), located immediately downstream of the fsrABDC operon. As shown in this study, orally administered B. subtilis spores germinate in the gut and produce fengycin/surfactin lipopeptides that inhibit GBAP-mediated induction of the Fsr quorum-sensing system of E. faecalis. As a result, gelatinase production-mediated destruction of the gut epithelium is avoided and E. faecalis cannot transgress into the bloodstream, abolishing E. faecalis systemic infection.

We then investigated whether this novel insight into a key step of E. faecalis pathogenesis can be used as a basis to prevent systemic E. faecalis infection by a combined probiotic and anti-virulence approach. Quorum-sensing is a frequently proposed target for anti-virulence therapy, because in many pathogens, quorum-sensing systems regulate several key virulence factors (41, 56). The biological reason for this regulatory concept, as developed for systemic bacterial infection, is that expression of aggressive substances such as toxins, degradative exoenzymes and other virulence factors is delayed until there is sufficient bacterial cell density to cope with the attacks of host defenses. This notion is in good accordance with the idea that enterococcal overgrowth in the intestine is a prerequisite of epithelial translocation, which we show is due to quorum-sensing control of the secreted protease, gelatinase.

To determine whether Fsr-mediated pathogenesis can be targeted by a quorum-quenching approach, we investigated whether Fsr as a homolog of the staphylococcal Agr quorum-sensing system can be inhibited by lipopeptides purified from B. subtilis, which we previously have shown efficiently inhibit Agr (9). We found that both the surfactin and fengycin classes of Bacillus lipopeptides inhibit Fsr, which is notable as only fengycins inhibit Agr and surfactins are generally produced at higher amounts than fengycins (Fig. 8). Traditional quorum-quenching strategies aim to apply quorum-sensing blocking drugs in purified form, which particularly for systemic application is often problematic due to low drug stability or high hydrophobicity and thus low bioavailability. In contrast, the probiotic approach proposed in this study, which uses orally administered live B. subtilis spores that germinate in the intestine (57), has the advantage that the inhibiting drugs are constantly being produced by an administered probiotic live organism.

Importantly, the target of the probiotic anti-virulence strategy described in this study is not asymptomatic colonization as a potential reservoir for infection of distant body sites, as in the case of probiotic interference with staphylococcal Agr (9, 50), but direct interference with a key step in the pathogenesis of systemic infection. This combination of a probiotic with an anti-virulence approach represents a categorically new strategy for the prevention of systemic infection of a pathogen of gastro-intestinal origin. Such a strategy may add to conventional therapy as well as recently reported probiotic approaches aimed at reducing the abundance of VRE in the gut (58). As probiotic therapy represents a generally healthy way of medical intervention, it may be particularly valuable for the high-risk group of immune-compromised patients that frequently develop enterococcal bacteremia. These include patients receiving HSCT or other treatment for underlying hematologic malignancies, such as CY, which can lead to damage of the intestinal epithelium (59–61).

Finally, our findings on probiotic prevention of systemic blood infection by a major nosocomial pathogen help to mechanistically explain previous reports indicating that probiotics reduce the incidence of sepsis (10, 11) and generally confirm the value of probiotic therapy for the prevention of infections.

Limitations of our study include the use of a mouse model to mimic human bacterial colonization and infection. However, this is commonly accepted, and specifically for the study of intestinal colonization and systemic disease caused by Enterococcus.

Materials and Methods

Study design

The present study comprises controlled laboratory in-vitro and mouse experiments to verify the hypotheses stated in the main text. Numbers of repeats and numbers of animals were determined without power analyses based on common numbers of similar experiments. Investigators were not blinded as for the distribution of animals to specific groups. All data were included, except when samples could not be analyzed due to technical reasons.

Bacterial strains and growth conditions

The reference B. subtilis strain and parent of the ΔfenA, ΔsrfA and Δsfp mutants used in this study was strain ZK3814 (genotype NCIB3610). The E. faecalis strains used in this study were strain OG1RF, which is a human caries-associated strain isolated in the early 1970s (62) and the most commonly studied non-VRE strain, strain V583, the first VRE isolated in the United States, which is resistant to high amounts of aminoglycosides, macrolides, lincosamides, and streptogramin B (47), and strain TX0104, which was isolated from a case of endocarditis-associated bacteremia in 2002 (B. E. Murray, unpublished). Isogenic mutants in fsrB (TX5266, ΔfsrB), gelE (TX5264, ΔgelE), sprE (TX5243, ΔsprE) and the gelE/sprE double mutant (TX5471, ΔgelE/sprE) were kindly provided by Dr. Barbara E. Murray, The University of Texas Health Science Center at Houston (UTHealth). When necessary, isogenic mutant E. faecalis were distinguished from wild-type bacteria using their introduced resistance to kanamycin by plating on kanamycin (2 mg/ml)-containing agar plates.

To construct the B. subtilis lipopeptide mutant strain (Δsfp), SPP1-phage-mediated transduction (63) was performed to transfer the sfp deletion present in the donor strain (BKE03570, a sfp::erm mutant in B. subtilis strain 168 obtained from the Bacillus Genetic Stock Center) to B. subtilis strain ZK3814. The resulting strain was verified by HPLC/MS to be avoid of fengycin and surfactin production (Supplementary Fig. S3). Bacteria were generally grown in tryptic soy broth (TSB) with shaking unless otherwise indicated.

Measurement of proteolytic activity

To determine the Fsr-inhibiting activity of Bacillus culture filtrates or purified fractions, we measured gelatinase activity using an azocasein assay as described (64). To avoid assay interference from B. subtilis proteases (65), we boiled all B. subtilis culture filtrates prior to co-incubation with E. faecalis cultures. Briefly, E. faecalis strains were diluted 100-fold from a preculture grown overnight in TSB before distribution into 15-ml culture tubes. To 3 ml of that dilution, we added 1 ml of sterilized Bacillus culture filtrate sample. Culture tubes were incubated at 37 °C with shaking for 3 h. 400 μl of E. faecalis supernatant were pipetted in an Eppendorf tube; then 1% azocasein in 0.05 M Tris/HCl (pH 8.5) was mixed with the supernatant. The mixture was incubated at 37 °C for 1 h. After incubation, 135 μl of 35% trichloroacetic acid (TCA) were added to stop the reaction, and the mixture was kept on ice for 15 min. The mixture was centrifuged at 13,000 rpm in a tabletop centrifuge for 10 min. After centrifugation, 750 μl of supernatant were collected, 750 μl of 1 M NaOH was added and the mixture was mixed, after which absorbance was measured with a GloMax Explorer luminometer (Promega) at 490 nm.

RNA extraction

Bacterial cultures were harvested by centrifugation at 4,122 × g for 10 minutes. Pellets were resuspended in 1 ml RNase-free water, centrifuged at 4,122 × g for 10 minutes, and pellets were resuspended in 700 μl buffer RLT (Qiagen) supplemented with β-mercaptoethanol. The bacterial suspensions were transferred to lysing matrix, homogenized in a Fast Prep bead beater (Thermo Savant) at 6.0 m s⁻¹ for 20 s, incubated on ice for 5 minutes, and centrifuged at 18,213 × g for 15 minutes. Ethanol was added to the lysate. Finally, RNA was extracted using an RNeasy mini kit (Qiagen) according to the manufacturer’s instructions.

qRT-PCR Analysis

Fsr activity was measured using quantitative real-time PCR (qRT-PCR) of the fsrBCD and gelE/sprE transcripts in cultures grown for 3 h. To that end, RNA obtained from three independent samples was subjected to One-Step SYBR^® Green real-time PCR using the SuperScript^® III Platinum^® SYBR^® Green One-Step qRT-PCR Kit (Invitrogen Life Technologies) and an ABI 7500 thermocycler (Applied Biosystems). The expression level of fsrBDC and gelE/sprE transcripts was normalized against that of 16S rRNA. Previously published primer sequences (44) were used and primers were synthesized by Sigma.

Intestinal permeability assay

In-vivo intestinal permeability was assessed through oral administration of the non-digestible marker, fluorescein-isothiocyanate (FITC)-conjugated dextran-4000, which crosses the intestinal mucosal barrier by non-mediated diffusion into the blood (66). 4-kDa FITC-dextran (Sigma) was dissolved in phosphate buffered saline (PBS) to a concentration of 60 mg ml⁻¹. Mice received 150 μl FITC-dextran four hours before sacrifice by oral gavage. Whole blood was collected by cardiac puncture and then centrifuged at 1,300 × g for 10 min at 4 °C to obtain serum. The fluorescence of the obtained serum samples was quantified at an excitation wavelength of 488-nm and an emission wavelength of 519-nm using a Tecan Spark multimode microtiter plate reader.

Peptides

The Bacillus lipopeptides surfactin and fengycin were prepared from culture filtrate of B. subtilis ZK3814 as previously described (9) with a two-step reversed-phase chromatography protocol using SOURCE Phenyl and Zorbax SB-C18 columns. Fractions containing different fengycins or surfactins were combined for a total fengycin and a total surfactin preparation, respectively. GBAP was synthesized and quality-controlled (> 95% purity) in-house by the NIAID Research Technologies Branch.

Mouse experiments

In-vivo studies were approved by the Institutional Animal Care and Use Committee of the NIAID. Animal work was conducted by certified staff in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). All animal work adhered to the institution’s guidelines for animal use and followed the guidelines and basic principles in the US Public Health Service Policy on Humane Care and Use of Laboratory Animals, and the Guide for the Care and Use of Laboratory Animals.

To study E. faecalis intestinal colonization, bacteremia, and systemic infection, mice received the immunosuppressant drug cyclophosphamide (CY) and antibiotics, or antibiotics only (as indicated) as described (31) with some modifications. 6 to 8-week-old C57BL-6J mice were first intraperitoneally pre-treated with 150 mg kg⁻¹ CY once a day for 3 days to induce neutropenia, followed by oral gavage with metronidazole (100 mg kg⁻¹), kanamycin (100 mg kg⁻¹) and vancomycin (100 mg kg⁻¹) once a day for 3 days to eradicate the pre-existing microbiota. In all set-ups, E. faecalis strains were grown to mid-exponential growth phase, washed, and resuspended in sterile PBS at 10¹⁰ CFU ml⁻¹. Mice were inoculated by oral gavage with 200 μl of a 10¹⁰ CFU ml⁻¹ suspension (2 × 10⁹ CFU) of the indicated E. faecalis strains once or on two consecutive days, as indicated. Intestinal colonization was evaluated by quantitative cultures of the small and large intestine samples on subsequent days, as indicated. Bacteremia and systemic infection was evaluated by quantitative cultures of mouse blood obtained via cardiac puncture as well as samples of collected livers and spleens. Blood samples were also analyzed on an IDEXX ProCyte DX^™ instrument to determine leukocyte numbers. Intestinal samples were diluted to a final volume of 1 ml of PBS for counting. The organs of each mouse were placed into a 2-ml tube containing 1 ml of sterile PBS with 500 mg of 2-mm borosilicate glass beads (Sigma). The organs were homogenized in a Fast Prep bead beater (Thermo Savant) at 6.5 m s⁻¹ for 60 s. Organ homogenates, blood samples, and PBS-diluted intestinal samples were plated on Enterococcus selective media (BBL^™ Enterococcosel^™ Agar) (Becton Dickinson), and incubated 24 – 48 h at 37 °C for CFU counting. In control experiments with or without the antibiotic treatment specified above, we did not detect any bacteria in the feces of mice on the Enterococcus selective media, indicating there is no source of enterococci in our experimental setup other than from oral gavage.

For the B. subtilis competition experiment, peri-oral gavage with 200 μl of a 10⁸ CFU ml⁻¹ solution (2 × 10⁷ CFU) in sterile PBS of spores of Bacillus wild-type or its isogenic Δsfp lipopeptide mutant was performed once a day on the two days following the antibiotic treatment. The preparation of Bacillus spores was performed as described (9). Peri-oral gavage with 200 μl of a 10¹⁰ CFU ml⁻¹ suspension (2 × 10⁹ CFU) of E. faecalis was performed on the following day and in some experiments repeated on the next day, as indicated.

For the direct systemic infection study, mice were intravenously injected with 200 μl of a 10⁹ CFU ml⁻¹ suspension (2 × 10⁸ CFU) of the indicated E. faecalis strains. Three days post infection cardiac puncture was performed, and the livers, kidneys, and hearts were collected. There were no bacteria found in the blood and hearts.

Histopathology

Intestinal samples were cut into approximately 1-cm pieces and fixed in 10% formalin. Fixed tissue samples were trimmed, dehydrated, cleared, and infiltrated with paraffin wax. Following infiltration, the tissues were embedded into paraffin blocks and cut at 5 μm on a microtome. The sections were then placed on APEX SAS slides. These slides were then deparaffinized and stained with hematoxylin and eosin, and blued to provide the resultant H&E staining. Finally, the H&E slides were scanned at 40 × magnification (0.25μm/pixel).

Statistics

Statistical analysis was performed using Graph Pad Prism version 8.3.0 for Mac. For experiments with n=3, unpaired two-sided t-tests were used for comparing two groups and 1-way ANOVAs with the indicated post-tests for multiple comparison of three or more groups. For experiments with n > 3, the four normality tests included in Prism (Anderson-Darling, D’Agostino-Pearson, Shapiro-Wilk, Kolmogorov-Smirnov) were performed and unpaired two-sided t-tests or 1-way ANOVAs, as appropriate, were only performed if all calculated tests computed normal distribution for all groups. Otherwise, non-parametric tests were used. Zero-inflated Poisson regression was used for comparisons where many values in one group were under the detection limit and set to zero. Used tests are always indicated in the figure legends. All error bars show the mean ± standard deviation (SD) for arithmetic y axes and the geometric mean and geometric SD for logarithmic y axes. Note error bars and means cannot be depicted for geometric means/SDs if the group contains 0 values. All replicates are biological.

Data and materials availability:

All data is available in the main text or the supplementary materials. Unique biological material is available subject to completion of simple transfer agreements (MTAs) with the NIAID.