Probiotic nutrition is frequently claimed to improve human health. In particular, live probiotic bacteria obtained with food are thought to reduce intestinal colonization by pathogens, and thus to reduce susceptibility to infection. However, the mechanisms that underlie these effects remain poorly understood. Here we report that the consumption of probiotic Bacillus bacteria comprehensively abolished colonization by the dangerous pathogen Staphylococcus aureus in a rural Thai population. We show that a widespread class of Bacillus lipopeptides, the fengycins, eliminates S. aureus by inhibiting S. aureus quorum sensing—a process through which bacteria respond to their population density by altering gene regulation. Our study presents a detailed molecular mechanism that underlines the importance of probiotic nutrition in reducing infectious disease. We also provide evidence that supports the biological significance of probiotic bacterial interference in humans, and show that such interference can be achieved by blocking a pathogen’s signalling system. Furthermore, our findings suggest a probiotic-based method for S. aureus decolonization and new ways to fight S. aureus infections.


There is increasing appreciation of the key role that the intestinal microbiota play in preventing the colonization and overgrowth of pathogens1,2. The mechanisms that have been implicated in this beneficial function of probiotic bacteria are mostly indirect, and include modulation of the immune system, enhancement of the intestinal epithelial barrier, or competition with pathogens for nutrients2,3,4,5. Whether there is direct interference between probiotic and pathogenic bacteria is less clear. Some probiotic strains produce bacteriocin proteins, which can kill phylogenetically related pathogenic bacteria2, and it has been shown that a bacteriocin-producing Escherichia coli strain inhibits colonization by related pathogenic bacteria in the inflamed gut of mice6. However, no evidence has been obtained to indicate that such mechanisms matter or are widespread in humans. Furthermore, it is not known whether there are mechanisms for direct probiotic bacterial interference that are not mediated by bacteriocins.

The genus Bacillus comprises different species of soil bacteria that form endospores with the ability to survive harsh environmental conditions, such as the high temperatures encountered during cooking procedures. Bacillus spores are commonly ingested with vegetables7. They can subsequently germinate to form metabolically active, vegetative cells8, which can temporarily colonize the intestinal tract9. Given the variability in dietary customs, the concentration of Bacillus spores in human faeces is also highly variable. It has been reported to be around 105 colony-forming units (CFU) per gram on average, occasionally reaching up to 108 CFU per gram7. Several probiotic formulae contain Bacillus species10, which are thought to reduce pathogen colonization by mechanisms that—except for a described immune-stimulatory effect on epithelial cells11—remain poorly defined.

Staphylococcus aureus is a widespread and dangerous human pathogen that can cause a variety of diseases, ranging from moderately severe skin infections to fatal pneumonia and sepsis12. Treatment of S. aureus infections is severely complicated by antibiotic resistance13, such as in methicillin-resistant S. aureus (MRSA), and there is no working S. aureus vaccine14. Therefore, alternative strategies to combat S. aureus infections are eagerly sought15. Because S. aureus infections commonly originate from previous asymptomatic colonization16,17, decolonization has recently gained considerable attention as a possible means to fight S. aureus infections in a preventive manner18. While the nares (nostrils) have traditionally been considered the primary S. aureus colonization site19, there is increasing evidence that the intestinal tract is also commonly colonized by S. aureus20,21,22 and forms an important reservoir for outbreaks of infectious S. aureus disease23,24. Several studies have reported levels of S. aureus in the faeces of human adults of around 103–104 CFU per gram25,26,27. Possibly, intestinal S. aureus colonization explains the failure of previous topical decolonization efforts aimed solely at the nose16,22,28.

Here we hypothesized that the composition of the human gut microbiota affects intestinal colonization with S. aureus. To evaluate that hypothesis, we collected faecal samples from 200 healthy individuals from rural populations in Thailand (Fig. 1a). This exemplary population was selected in order to rule out, as much as possible, the food sterilization and antibiotic usage that are common in highly developed urban areas, which potentially could diminish the abundance of probiotic bacteria in the food and intestinal tracts of the participating subjects. Our analysis revealed a comprehensive Bacillus-mediated S. aureus exclusion effect in the human population. By demonstrating that quorum sensing is indispensable for S. aureus to colonize the intestine, and discovering that secreted Bacillus fengycin lipopeptides function as quorum-sensing blockers to achieve complete eradication of intestinal S. aureus, we provide evidence that strongly suggests that this pathogen-exclusion effect in humans is due to a widespread and efficient probiotic-mediated mechanism that inhibits pathogen quorum-sensing signalling.

Fig. 1: Exclusion of S. aureus colonization by dietary Bacillus in a human population.
Fig. 1

a, Areas (in red) from which faecal samples were collected in rural populations and analysed for the presence of Bacillus and S. aureus. b, c, Intestinal (b) and nasal (c) colonization with S. aureus (yellow) in individuals that showed (green) or did not show (grey) intestinal colonization with Bacillus.

S. aureus exclusion by Bacillus

We found that 25/200 (12.5%) of human subjects carried S. aureus in their intestines, as determined by growth from faecal samples. Nasal carriage was similar in frequency (26/200; 13%), a result that is in accordance with previous findings showing a correlation between nasal and intestinal colonization22. These rates are considerably lower than those commonly found in adult populations during cross-sectional culture-based surveys that were performed mainly in hospital-admitted individuals in urbanized areas (on average, 20% for intestinal and 40% for nasal carriage)16,21,22.

To examine the hypothesis that bacterial interactions in the gut determine intestinal S. aureus colonization, we first analysed the composition of the gut microbiome by 16S ribosomal RNA sequencing. However, we did not detect substantial differences in the composition of the microbiome between S. aureus carriers and non-carriers (Extended Data Fig. 1).

By contrast, we found a striking correlation between the presence of Bacillus bacteria and the absence of S. aureus. Bacillus species (mostly B. subtilis; Extended Data Table 1) were found in 101/200 (50.5%) of subject samples. S. aureus was never detected in faecal samples when Bacillus species were present (P < 0.0001, Fisher’s exact test; Fig. 1b). Furthermore, this pathogen-exclusion effect was not limited to the site of interaction—the gut—but extended to S. aureus colonization in a general fashion. While Bacillus was generally absent from nasal samples, S. aureus nasal colonization was never detected when intestinal Bacillus was present (P < 0.0001, Fisher’s exact test; Fig. 1c). Notably, the levels of S. aureus colonization that we found in non-Bacillus-colonized individuals from rural Thailand approximately match those reported—using similar culture-based assays—in urbanized Western areas. These findings indicate a widespread mechanism exerted by Bacillus species that comprehensively inhibits colonization with S. aureus. Moreover, they suggest that S. aureus colonization is increased in urban populations because of the lack of a probiotic, Bacillus-containing diet. Of particular note, the results also indicate that the intestinal site has a previously underappreciated role in determining general S. aureus colonization, a notion in accordance with findings attributing a key role to faecal transmission in MRSA recolonization28.

When we analysed data from previous 16S rRNA-sequencing-based microbiome studies, we found strongly variant results and no correlation between the absence of S. aureus and the presence of B. subtilis: studies that reported considerable B. subtilis or S. aureus numbers (samples with more than 10% colonization by either species) did not reveal exclusion phenomena (average 14.89 ± 15.69% colonization by both species) (Extended Data Table 2). However, although we did not find a correlation, this might be due to the fact that such sequencing-based analyses are set up to detect high-order taxonomic shifts rather than specific differences on the species or genus level.

Quorum sensing and colonization

Our results, which show no substantial high-order taxonomic differences in the microbiome composition between S. aureus carriers and non-carriers, exclude an indirect effect of Bacillus on the microbiome composition. Rather, we hypothesized that the Bacillus isolates produce a substance that directly and specifically inhibits intestinal colonization by S. aureus. We first analysed whether there is a growth-inhibitory effect of the Bacillus isolates on S. aureus. However, only a minor growth inhibition occurred in just 6 out of 105 isolates (we saw a maximal 1-mm inhibition zone when using an agar diffusion test with a five-times-concentrated culture filtrate). Therefore, a growth-inhibitory effect fails to explain the observed complete correlation between the presence of Bacillus and the absence of S. aureus, and rules out a bacteriocin-mediated phenomenon.

The factors that are important for S. aureus intestinal colonization are poorly understood. One study in mice has implicated teichoic acids found in the bacterial cell wall, as well as the cell-surface protein clumping factor A (ClfA)29. Prompted by our previous finding that ClfA is positively regulated by the accessory gene regulator (Agr) quorum-sensing system30, we hypothesized that the Bacillus isolates secrete a substance that interferes with quorum-sensing signalling. Quorum sensing is responsible for sensing the density of the bacterial population (the ‘quorum’) and controlling a concomitant alteration in cell physiology31. Because quorum-sensing signals and sensors differ between different types of bacteria31, an underlying quorum-quenching mechanism could explain the specificity of the inhibitory effect that we detected.

Because the role of quorum sensing in S. aureus intestinal colonization is unknown, we first used a mouse model of S. aureus intestinal colonization to test whether Agr-based quorum sensing is involved (Fig. 2a). In all mouse models in our study, we included: first, a human faecal isolate belonging to a sequence type (ST) that was frequently detected in the faecal isolates that we obtained (ST2196), according to multi-locus sequence typing (MLST) that we performed (Supplementary Table 1); second, a mouse infection isolate (ST88)32; and third, a human infection isolate of the highly virulent MRSA type USA30033. In competition experiments with equal amounts of wild-type and isogenic agr mutant strains, only wild-type S. aureus was detected in the faeces and colonized the large and small intestines at the end of the experiment (competition index ≥ 100) (Fig. 2b and Extended Data Fig. 2a, b). Furthermore, in a non-competitive experimental set-up, only those bacteria expressing the intracellular Agr effector RNAIII34 achieved colonization; agr-negative control strains never did (Fig. 2c and Extended Data Fig. 2c). These data show that, in addition to its well-known role in infection30,35, the Agr quorum-sensing system is absolutely indispensable for intestinal colonization.

Fig. 2: Quorum-sensing dependence of S. aureus intestinal colonization.
Fig. 2

a, Experimental set-up of the mouse intestinal colonization model. Mice received, by oral gavage, either 100 μl containing 108 CFU ml−1 of wild-type (WT) S. aureus strain ST2196 F12 and another 100 μl of 108 CFU ml−1 of the corresponding isogenic agr mutant (n = 5 per group; competitive experiment, shown in b); or 200 μl containing 108  CFU ml−1 wild-type, isogenic agr mutant or Agr (RNAIII)-complemented agr mutant (n = 5 per group; non-competitive experiment, shown in c). CFU in the faeces were determined two, four and six days after infection. At the end of the experiment (day seven), CFU in the small and large intestines were determined. b, Competitive experiment. Total obtained CFU are shown as dot plots; also shown are mean ± s.d. Bars show the percentage of wild-type among total determined CFU, of which 100 were analysed for tetracycline resistance (which is present only in the agr mutant). No agr mutants were detected in any experiment; therefore, all bars show 100% wild type. Given that 100 isolates were tested, the competitive index of wild type/agr mutant in all cases is ≥100. c, Non-competitive experiment with genetically complemented strains. Wild-type and isogenic agr mutant strains all harboured the pKXΔ16 control plasmid; Agr-complemented strains harboured pKXΔRNAIII and constitutively expressed RNAIII, which is the intracellular effector of Agr. During the experiment, mice received 200 μg ml−1 kanamycin in their drinking water to maintain plasmids. Statistical analysis was performed using Poisson regression versus values obtained with the agr mutant strains. *P < 0.0001. Data are mean ± s.d. Note that no bacteria were found in the faeces or intestines of any mouse receiving S. aureus Δagr with vector control. The corresponding zero values are plotted on the x axis of the logarithmic scale. See Extended Data Fig. 2 for corresponding data obtained using strains USA300 LAC and ST88 JSNZ.

Source data.

Fengycin quorum quenchers

Having established that the Agr quorum-sensing regulatory system is essential for S. aureus intestinal colonization, we next analysed whether culture filtrates of the Bacillus isolates collected from human faeces can inhibit Agr. To that end, we used an S. aureus reporter strain, into the genome of which we had transferred the luminescence-conferring luxABCDE operon under the control of the Agr P3 promoter34, which controls production of RNAIII. Remarkably, culture filtrates from all 105 isolates reduced Agr activity in the S. aureus reporter strain by at least 80% (Fig. 3a and Extended Data Table 1). No growth effects were observed, substantiating that growth inhibition does not underlie the inhibitory phenotype. Furthermore, a culture filtrate from a reference B. subtilis strain suppressed the production of key Agr-regulated virulence factors (phenol-soluble modulins, α-toxin and Panton–Valentine leucocidin; Fig. 3b, c and Supplementary Fig. 1). These results indicate that the inhibitory effect of the Bacillus isolates on S. aureus colonization is due to a secreted substance that inhibits Agr signalling.

Fig. 3: Inhibition of S. aureus quorum sensing by Bacillus fengycin lipopeptides.
Fig. 3

a, Example of an Agr-inhibition experiment. The Bacillus isolate was considered inhibitory if luminescence after 4-h growth of S. aureus was less than or equal to half that of the control value (red arrow). RLU, relative light units; TSB, tryptic soy broth (control conditions). The experiment was performed with n = 2 biologically independent samples. The lines connect the means. b, Inhibition of expression of Panton–Valentine leucocidin (PVL) and α-toxin, using culture filtrate from the B. subtilis reference strain. Western blot analysis of n = 3 biologically independent samples was performed with filtrates from S. aureus cultures that had been grown for 4 h. See Supplementary Fig. 1 for the entire blots. c, Inhibition of expression of phenol-soluble modulins (PSMs) using culture filtrate from the B. subtilis standard strain. PSM expression was determined by RP-HPLC/ESI-MS after 4 h of S. aureus growth. d, Test for Agr-inhibitory capacity of Bacillus culture filtrate applied at a final concentration that represents the median concentration of total fengycin in the tested 106 Bacillus isolates. *P < 0.0001 (two-way analysis of variance (ANOVA) with Tukey’s post-test versus control). e, Total fengycin concentrations in stationary-phase culture filtrates of the 106 Bacillus isolates (see Extended Data Table 1 for details). f, Agr-inhibiting activities of B. subtilis wild-type (WT) in comparison to ΔfenA (fengycin-deficient) and ΔsrfA (surfactin-deficient) strains. *P < 0.0001 (two-way ANOVA with Tukey’s post-test versus wild type). The experiments shown in c, d, f were performed with n = 3 biologically independent samples. Data are mean ± s.d.

Source data.

To characterize the Agr-inhibitory substance(s), we performed experiments with culture filtrate of the reference B. subtilis strain. We found that the substance in question was thermostable and resistant to protease digestion (Extended Data Fig. 3a). In reversed-phase high-performance chromatography (RP-HPLC) (Extended Data Fig. 3b), substantial Agr-inhibiting activity was associated with two peaks, which we analysed by RP-HPLC/electrospray ionization mass spectrometry (ESI-MS) (Extended Data Fig. 3c). This analysis, together with the elution behaviour and published literature36, allowed us to identify the Agr-inhibiting substances as members of the fengycin cyclic lipopeptide family. Because fengycins can differ in specific amino acids and in the length of the attached fatty acid, which usually is β-hydroxylated (β-OH), and because different Bacillus strains produce different fengycin species37, we used further tandem mass spectrometric fragmentation analysis (MS/MS) to identify the specific fengycins present in the two active peaks (Extended Data Fig. 3d). Fengycins in the first peak were identified as β-OH-C17-fengycin A and β-OH-C16-fengycin B. The second peak consisted of one fengycin species, β-OH-C17-fengycin B. According to RP-HPLC/ESI-MS analysis, smaller, adjacent peaks also contained fengycin species, which we tentatively identified as β-OH-C17-fengycin A and the dehydroxylated versions of the identified three major fengycins (Extended Data Fig. 3e). For further analyses, we purified higher amounts of β-OH-C17-fengycin B to homogeneity from culture filtrate and verified the dose-dependent Agr-inhibiting activity of this pure substance (Extended Data Fig. 4).

Using RP-HPLC/ESI-MS analysis, we found fengycin production in all isolates, substantiating the general character of the inhibitory interaction (Extended Data Table 1). Although the production pattern of different fengycins varied between the analysed isolates, in many of them β-OH-C17-fengycin B was the most strongly produced type. Notably, almost complete inhibition of Agr was detected at a concentration of about 1.4 μM total fengycin (Fig. 3d). This corresponds to the median concentration of total fengycin (1.5 μM) produced by stationary-phase cultures of the Bacillus isolates (Fig. 3e).

To provide definitive evidence that fengycin production underlies the Agr-inhibiting capacity of Bacillus, we produced an isogenic mutant in the reference B. subtilis strain of the fenA gene, which is essential for fengycin production38. RP-HPLC/ESI-MS showed a specific absence of fengycins in that mutant strain, whereas surfactins—the predominant Bacillus lipopeptides—were still present (Extended Data Fig. 3f). Culture filtrate of the fenA mutant strain was devoid of Agr-inhibiting activity, in contrast to that of the isogenic wild-type strain (Fig. 3f). We also measured an isogenic surfactin-negative mutant strain, which showed Agr-inhibiting activity similar to that of the wild-type strain (Fig. 3f). These results confirmed that fengycin production is the source of the observed Agr inhibition.

Mechanism of fengycin-mediated inhibition

In the S. aureus Agr quorum-sensing regulatory circuit, the secreted Agr autoinducing peptide (AIP) interacts with an extracellular domain of AgrC, the histidine kinase part of a two-component signal-transduction system, to signal the cell-density status39 (Fig. 4a). Different Agr subgroups of S. aureus, as well as different staphylococcal species, produce distinct cyclic heptapeptide to nonapeptide AIPs35. AIPs from other subgroups or species frequently inhibit Agr signal transduction by competitive inhibition at the AgrC-binding site39,40,41. Given that fengycins, being cyclic lipopeptides, show structural similarity to AIPs (Fig. 4b), it appears likely that fengycins compete with the natural AIP for AgrC binding. The only other theoretically possible site of interference from the extracellular space would be the membrane-located AIP production/secretion enzyme AgrB. Using an S. aureus agrBD deletion strain and stimulation of AgrC by synthetic AIP, which led to complete Agr activation, we ruled out that the target of Agr inhibition by Bacillus is AgrB (Fig. 4c). In further support of a mechanism that works through competition with AIP for binding to the AgrC receptor, we found that fengycin inhibition could be reversed in a dose-dependent fashion by adding AIP (Fig. 4d). Finally, we determined the AIP concentration in early stationary growth phase (at 6–8 hours) to be about 1 μM (Extended Data Fig. 5a), which is approximately equal to the concentration of fengycin for which we found complete Agr inhibition (Fig. 3d). These findings indicate that fengycins inhibit Agr signal transduction by efficient competitive inhibition as structural analogues of AIPs.

Fig. 4: Competitive inhibition of S. aureus AIP activity by fengycins.
Fig. 4

a, Model of competitive Agr inhibition by fengycins. The agrBDCA operon (bottom right), whose expression is driven by the P2 promoter, encodes the AgrD precursor of the autoinducing-peptide (AIP), which is modified and secreted by AgrB. AIP binds to membrane-located AgrC, which, upon autophosphorylation, triggers phosphorylation and activation of the DNA-binding protein AgrA. In addition to stimulating transcription from the P2 promoter (autoinduction), AgrA drives expression of RNAIII, which in turn regulates the expression of target genes such as those encoding ClfA, α-toxin and leukotoxins. RNAIII also encodes the δ-toxin. Furthermore, AgrA drives the expression of phenol-soluble modulins (PSMs) in an RNAIII-independent fashion. b, Structural similarity of fengycins with AIPs. The structures of β-OH-C17-fengycin B and AIP-I are shown as examples. In red are structures and/or amino acids that may differ in different subtypes. c, Fengycins work by inhibiting AgrC. Shown is the inhibition of Agr by fengycin-containing Bacillus culture filtrate, using an agrBD-deleted S. aureus strain in which AgrC was stimulated by exogenously added AIP. *P < 0.0001 (two-way ANOVA with Tukey’s post-test; values obtained in ΔagrBD/AIP versus ΔagrBD/control (no AIP), and ΔagrBD/AIP/culture filtrate versus ΔagrBD/AIP). d, Competitive titration of fengycin-mediated Agr inhibition by increasing amounts of AIP, as assayed by the Agr luminescence assay. RLU, relative light units. Statistical analysis is by two-way ANOVA with Tukey’s post-test versus control. e, Inhibition of Agr in different Agr-subtype S. aureus and S. epidermidis (strain 1457) by β-OH-C17-fengycin B, as measured by relative expression of δ-toxin using RP-HPLC/ESI-MS. Statistical analysis is by two-way ANOVA with Tukey’s post-test versus intensity values obtained without addition of fengycin. Values were calculated as percentages relative to intensity values obtained without addition of fengycin, owing to different δ-toxin expression levels in the different strains. c–e, Experiments were performed with n = 3 biologically independent samples. Data are mean  ± s.d.

Source data.

The fact that AgrC–AIP interaction differs according to Agr subtype raises the question of whether fengycins have a general ability to inhibit Agr. We found that purified β-OH-C-17 fengycin B inhibited Agr in members of all S. aureus Agr subtypes, as well as in S. epidermidis (Fig. 4e). Furthermore, the S. aureus strains used in our mouse experiments belong to different Agr subtypes (strain USA300, type I; strain ST88, type III; strain ST2196, type I). These results indicate that fengycins have broad-spectrum Agr-inhibiting activity.

Bacillus spores eradicate S. aureus

To validate our findings in vivo and demonstrate the specific role of fengycins in the inhibition of S. aureus intestinal colonization, we compared the impact of the B. subtilis wild-type reference strain and its isogenic fenA mutant on S. aureus colonization in a mouse intestinal colonization model. We first performed a control experiment to analyse the colonization kinetics of B. subtilis when given as spores, which corresponds to the form in which Bacillus would be taken up with food or probiotic formulae (Extended Data Fig. 5b). We observed transient colonization that strongly declined within two days. Importantly, colonization by the B. subtilis fenA mutant was not different to that by the wild-type strain, ruling out the possibility that fengycin production as such affects B. subtilis colonization.

Feeding mice B. subtilis spores completely abrogated colonization of all tested S. aureus strains in the faeces and intestines, in experimental set-ups with or without antibiotic pretreatment to eliminate the pre-existing microbiota. (Fig. 5b, c and Extended Data Fig. 5c–f). By contrast, spores of the fenA mutant had no notable effect on colonization of any S. aureus test strain. As Bacillus intestinal colonization in humans has been shown to reach much higher levels than that by S. aureus7—a situation likely to be even more pronounced in the tested rural population—our mouse data obtained with S. aureus numbers approximately equal to or exceeding those of applied Bacillus spores suggest that fengycin-mediated interference in quorum sensing contributes to the exclusion of S. aureus colonization that we observed in humans.

Fig. 5: Inhibition of S. aureus colonization by dietary fengycin-producing Bacillus spores in a mouse model.
Fig. 5

a, Experimental set-up. n = 5 mice per group received 200 μl of 108 CFU ml−1 S. aureus strain ST2196 F12 by oral gavage. On the next day and every following second day, they received 200 μl of 108 CFU ml−1 spores of the B. subtilis wild-type (WT) or its isogenic fenA mutant, also by oral gavage. CFU in the faeces were determined two, four and six days after infection. At the end of the experiment (day seven), CFU in the small and large intestines were determined. The experiment was performed with (b) or without (c) antibiotic pretreatment. b, c, Experimental results. Statistical analysis was performed using Poisson regression versus values obtained with the B. subtilis WT spore samples. *P < 0.0001. Data are mean ± s.d. Note that no S. aureus were found in the faeces or intestines of any mouse challenged with S. aureus and receiving Bacillus wild-type spores. The corresponding zero values are plotted on the x axis of the logarithmic scale. See Extended Data Fig. 5 for corresponding data obtained using strains USA300 LAC and ST88 JSNZ.

Source data.


Scientific evidence to support the frequent claims that probiotic nutrients improve human health is scarce. However, this study provides evidence for a molecular mechanism by which probiotic bacteria found in food could directly interfere with pathogen colonization. In particular, our data underscore the often-debated10,42 probiotic value of B. subtilis. Notably, we found the responsible agents to work by quorum quenching, demonstrating that pathogen exclusion in the gut may work by inhibition of a pathogen signalling system. Furthermore, our findings emphasize the importance of quorum sensing for pathogen colonization.

Our study suggests several valuable translational applications regarding alternative strategies to combat antibiotic-resistant S. aureus. First, the quorum-quenching fengycins—which previously had been known only for their antifungal activity43—could potentially be used as quorum-sensing blockers in eagerly sought antivirulence-based efforts to treat staphylococcal infections15,44. Second, Bacillus-containing probiotics could be used for simple and safe S. aureus decolonization strategies. In that regard, it is particularly noteworthy that our human data indicate that probiotic Bacillus can comprehensively eradicate intestinal as well as nasal S. aureus colonization. Such a probiotic approach would have numerous advantages over the present standard topical strategy involving antibiotics, which is aimed exclusively at decolonizing the nose45.


No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment, except for when noted.

Sample collection and bacterial screening

Nasal swabs and faecal samples were obtained from 200 Thai healthy volunteers from four different locations in southern, central, northeastern and northern Thailand. One sterile nasal swab, a sample collection tube, a sterile container and tissue paper were given to each participant. All participants provided informed written consent. The study was performed in compliance with all relevant ethical regulations and approved by the Siriraj Institutional Review Board (approval no. Si 733/2015). All participants were at least 20 years old (age range 20–87 years; median age 57 ± 14.5 years; 131 women and 69 men) and without history of intestinal disease. None had received any antibiotic treatment or stayed at a hospital within at least three months before the study.

Nasal swabs and faecal samples were streaked on mannitol salt agar (MSA) and then incubated at 37 °C for 24 h. Positive or negative S. aureus or Bacillus colonization could easily be distinguished by either strong growth on the entire plate, or the absence of any colonies, respectively. At the time of this analysis, the purpose was to obtain and archive colonizing S. aureus strains. As the hypothesis regarding Bacillus/S. aureus exclusion was developed only after we obtained the results of this analysis, the staff performing the analysis were blinded as to the exclusion hypothesis. Isolates were easily recognized as S. aureus or Bacillus by colony morphology and colour; however, every isolate was confirmed for species identity using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI–TOF MS; see below), and Bacillus species were further distinguished by 16S rRNA sequencing (Extended Data Table 1). To that end, 16S rRNA genes were amplified by the polymerase chain reaction (PCR) using primers 27FB and 1492RAB46 and similarity analysis with the basic local alignment search tool (BLAST) was used to identify the species. Subjects were considered as permanently colonized by S. aureus if two positive samples were obtained, tested after a four-week interval. All individuals tested either negative or positive for S. aureus at both times. In total, 105 Bacillus isolates from 101 individuals were analysed. In the samples from four individuals, two isolates each were taken owing to their apparent phenotypic differences.

Bacterial identification using MALDI–TOF MS

Isolates were inoculated onto sheep blood agar and incubated for 24 h at 37 °C. Bacterial colonies were applied onto a 96-spot target plate and allowed to dry at room temperature. Subsequently, 2 µl of MALDI matrix (a saturated solution of α-cyano-4-hydroxycinnamic acid (HCCA) in 50% acetonitrile and 2.5% trifluoroacetic acid) was applied onto the colonies and allowed to dry before testing. Then the target plate was loaded into the MALDI–TOF MS instrument (MicroFlex LT mass spectrometer, Bruker Daltonics). Spectra were analysed using MALDI Biotyper automation control and the Bruker Biotyper 2.0 software and library (version 2.0, 3,740 entries; Bruker Daltonics). Identification score criteria used were those recommended by the manufacturer: a score of ≥2.000 indicated species-level identification; a score of 1.700–1.999 indicated identification to the genus level; and a score of <1.700 was interpreted as no identification. Isolates that failed to produce a score of <1.700 with direct colony or extraction methods were retested. S. aureus ATCC25923, E. coli ATCC25922 and Pseudomonas aeruginosa ATCC27853 were used as controls.

Bacterial strains and growth conditions

The reference B. subtilis strain and parent of the fenA and srfA mutants used in this study was strain ZK3814 (genotype NCIB3610). The S. aureus strains used in all experiments (except the experiment in which we analysed different Agr-subtype S. aureus) were: first, the human faecal isolate F12 of ST2196 (Supplementary Table 1); second, strain JSNZ of ST88, a mouse isolate previously described as mouse adapted32; and third, strain LAC of pulsed-field type USA300, an MRSA lineage predominantly involved in community-associated infections, but now generally representing the major lineage responsible for S. aureus infections in the United States47.

Isogenic mutants in agr were previously described (for strain LAC)48 or produced in this study (for strains JSNZ and F12) by phage transduction of the agr deletion from strain RN6911. The agr system is entirely deleted in these strains, except for a 3′ part of RNAIII, which is not transcribed owing to the absence of the corresponding promoter. All mutants were verified by analytical PCR.

Owing to the tetracycline resistance introduced in the agr deletion strains, kanamycin derivatives (pKXΔ) of the pTXΔ expression plasmid series were constructed and used for complementation of Agr. (This was not possible in strain LAC, which harbours resistance to multiple antibiotics, including kanamycin.) To that end, we treated plasmid pKX1549—provided by B. Krismer, University of Tübingen—as described48 to delete the xylR repressor gene, in order to make expression of any fragment cloned under control of the xyl promoter constitutive. To obtain plasmid pKXΔRNAIII, the RNAIII BamH1–MluI fragment was transferred from pTXΔRNAIII50. Plasmid pKXΔ16 is the corresponding empty control plasmid, derived from pKX16 by analogous deletion of the xylR repressor gene.

To construct the agrBD deletion mutant of strain LAC P3-lux, we used a 4.8-kilobase PCR product from USA300 genomic DNA that included the agrBDCA operon as well as 1 kb upstream and 1 kb downstream; we cloned this product into the SmaI site of plasmid pIMAY51 and used inverse PCR to delete agrBD. Allelic exchange was then performed, and the chromosomal deletion was confirmed by PCR using one primer outside of the 1-kb homology arm, followed by sequencing of the PCR product. See Supplementary Table 2 for the oligonucleotides used.

To construct the tetracycline-resistant derivatives of S. aureus ST88 and ST2196, we carried out φ11-phage-mediated transduction as described in order to transfer the tetracycline cassette in the donor strain (S. aureus RN4220 with integrated pLL29) to S. aureus strains ST88 and ST219652.

To construct the B. subtilis fengycin mutant strain, SPP1-phage-mediated transduction53 was performed to transfer the fenA deletion present in the donor strain (BKE18340, a fenA(ppsA)::erm mutant in B. subtilis strain 168 obtained from the Bacillus Genetic Stock Center) to B. subtilis strain ZK3814. This was necessary as B. subtilis strain 168 bears a mutation in the sfp gene, abolishing lipopeptide production.

Bacteria were generally grown in tryptic soy broth (TSB) with shaking unless otherwise indicated.

Typing of S. aureus isolates

S. aureus isolates were typed by MLST as described54. PCR amplicons of seven S. aureus housekeeping genes (arcC, aroE, glpF, gmk, pta, tpi and yqiL) were obtained from chromosomal DNA and their sequences compared with those available from the PubMLST database (https://pubmlst.org/saureus/). Previously undescribed alleles (arcC 520–521 and gmk 337) and sequence types (ST4630–ST4638) were deposited to the website. The Agr subtype of S. aureus isolates was determined using a modified multiplex quantitative reverse transcription PCR (qRT–PCR) protocol55. Two duplex qRT–PCR protocols, using the respective described primer sets and two coloured probes each, were set up for Agr types I and II, and III and IV, respectively. Isolates for which the Agr type could not be determined by that method were analysed for the type of AIP production using RP-HPLC/ESI-MS with the chromatography method also used for PSM detection (see below), integrating the three major m/z peaks for each AIP type.

Microbiome analysis

Genomic DNA from each faecal sample was extracted using a QIAamp DNA stool Minikit (Qiagen) according to the manufacturer’s instructions. The DNA was quantified using a Nanodrop spectrophotometer, and 16S rRNA paired-end sequencing of the V4 region of 16S rRNA was performed by Illumina using an Illumina MiSeq system as described56.

For all obtained paired-end sequences, the abundance of operational taxonomic units (OTUs) and alpha and beta diversity were identified using quantitative insights into microbial ecology (QIIME 1.9.1)57. This study used the Nephele (release 1.6) platform from the National Institute of Allergy and Infectious Diseases (NIAID) Office of Cyber Infrastructure and Computational Biology (OCICB) in Bethesda, Maryland, USA. The sequences were assigned to OTUs with the QIIME’s uclust-based58 open-reference OUT-picking protocol59 and the Greengenes 13_8 reference sequence set60 at 99% similarity. Alpha diversity was calculated using Chao1 and Shannon analyses61 and compared across groups using a non-parametric t-test with 999 permutations.

Growth-inhibition analysis

Growth inhibition of S. aureus by Bacillus culture filtrates was tested with an agar diffusion assay. To that end, 10 μl of Bacillus culture filtrate from each isolate was spotted on sterile filter disks. The filters were left to dry and the procedure was repeated four times, after which filters were laid on agar plates containing S. aureus, resulting in the analysis of five-times concentrated culture filtrate.

Fengycin purification

To identify the Agr-inhibiting active substance, 10 ml of culture filtrate from the B. subtilis reference strain grown for 48 h in TSB were applied to a Zorbax SB-C18 9.4 mm × 25 cm reversed-phase column (Agilent) using an AKTA Purifier 100 system (GE Healthcare). After washing with three column volumes of 100% buffer A (0.1% trifluoroacetic acid (TFA) in water) and five column volumes of 30% buffer B (0.1% TFA in acetonitrile), a 20-column volume gradient from 30% to 100% buffer B was applied. The column was run at a flow rate of 3 ml min−1. Peak fractionation was performed using the absorbance at 214 nm, and fractions were subjected to further analysis by RP-HPLC/ESI-MS and MS/MS and tested for Agr inhibition (see below).

To purify larger amounts of the main active peak containing β-OH-C17-fengycin B, we added acetonitrile to 200 ml filtrate from cultures grown under the same conditions to a final concentration of 10%; precipitated material was removed by centrifugation for 10 min at 3,700g using a Sorvall Legend RT centrifuge, and the obtained cleared supernatant was applied to a self-packed HR 16/10 column filled with Resource PHE (GE Healthcare) material (column volume 17 ml). After sample application, the column was washed with 10% buffer B for three column volumes and 25% buffer B for five column volumes, after which a gradient of 15 column volumes from 25% to 60% buffer B was applied. We collected 10-ml fractions and lyophilized positive fractions (as determined by RP-HPLC/ESI-MS). The lyophilisate was redissolved in 2 ml acetonitrile. We added 6 ml of water and removed the precipitated material through a 5-min centrifugation in a table-top centrifuge at maximum speed. The cleared supernatant was then further purified on a Zorbax SB-C18 9.4 mm × 25 cm reversed-phase column as described above.

PSM and lipopeptide detection by RP-HPLC/ESI-MS

PSMs were analysed by RP-HPLC/ESI-MS using an Agilent 1260 Infinity chromatography system coupled to a 6120 Quadrupole LC/MS in principle as described62, but with a shorter column and a method that was adjusted accordingly. A 2.1 mm × 5 mm Perkin-Elmer SPP C8 (2.7 μm) guard column was used at a flow rate of 0.5 ml min−1. After sample injection, the column was washed for 0.5 min with 90% buffer A and 10% buffer B, then for 3 min with 25% buffer B. Next, an elution gradient was applied from 25% to 100% buffer B in 2.5 min, after which the column was subjected to 2.5 min of 100% buffer B to finalize elution.

Bacillus culture filtrates or (partially) purified fractions containing lipopeptides (fengycins and surfactins) were analysed using the same column, system and elution conditions. To quantify the production of different fengycins, we used the two most abundant peaks, corresponding to double- and triple-charged ions, for the integration. Agilent mass hunter quantitative analysis version B.07.00 was used for quantification.

Measurement of Agr activity

To determine the Agr-inhibiting activity of Bacillus culture filtrates or purified fractions, we measured luminescence emitted by an Agr P3 promoter–luxABCDE reporter fusion construct that was inserted into the genome of S. aureus strain LAC34. Strain LAC P3–luxABCDE was diluted 100-fold from a preculture grown overnight in TSB before distribution into a 96-well microtitre plate. To 100 μl of that dilution, we added 100 μl of sterilized culture filtrate sample, unless otherwise indicated. Plates were incubated at 37 °C with shaking. Luminescence was measured with a GloMax Explorer luminometer (Promega) every 2 h for a total of 6 h. Inhibition was considered significant if the 4-h sample and control values differed by at least a factor of two. Of note, the quorum-quenching effect exerted by the one-time initial dose of fengycin or fengycin-containing culture filtrates was transient and was overcome at later times by the increasing intrinsic AIP production. The Agr-inhibiting activity of purified fengycin was also measured using quantitative real-time PCR of RNAIII as described30.

To determine the Agr-inhibiting activity with target strains other than LAC (Agr subtype I), we measured the production of δ-toxin, for which the gene is embedded in the Agr intracellular effector RNA, RNAIII, in most staphylococci. Production of δ-toxin was measured by RP-HPLC/ESI-MS as described above. Strains LAC (Agr subtype I), A950085 (Agr subtype II), MW2 (Agr subtype III) and A970377 (Agr subtype IV) were used for testing the effect of β-OH-C17-fengycin B on S. aureus of different Agr subgroups. Strain 1457 was used for S. epidermidis. All strains were diluted 100-fold from a preculture grown in TSB. β-OH-C17-fengycin B dissolved in dimethylsulfoxide (DMSO) was added to each sample to a final concentration of 20 μM and 100 μM. All samples were incubated at 37 °C with shaking for 4 h. Samples were centrifuged and supernatant was collected for RP-HPLC/ESI-MS detection.

Analysis of PVL and α-toxin expression

S. aureus strain LAC was diluted 100-fold from a preculture grown in TSB and inoculated into 500 µl TSB. Then, 250 μl of B. subtilis culture filtrate was added into the sample. Samples were incubated at 37 °C with shaking for 4 h. Samples were centrifuged in a table-top centrifuge at maximum speed for 5 min; the supernatants were collected and loaded onto 12% SDS–polyacrylamide gel electrophoresis (PAGE) gels, which were run at 160 V for 1 h. Proteins were transferred to nitrocellulose membranes using an iBlot western blotting system. Membranes were incubated with Odyssey blocking buffer for 1 h at room temperature. Anti-α-toxin antibodies (polyclonal rabbit serum; Sigma S7531; dilution 1:5,000) or anti-LukF-PV antibodies (affinity-purified rabbit IgG specific for a peptide region of LukF-PV, produced by GenScript USA and provided by F. DeLeo, NIAID; dilution 1:500) were added to the blocking buffer and membranes were incubated overnight at 4 °C. Then, membranes were washed five times with Tris-buffered saline containing 0.1% Tween-20, pH 7.4, and incubated with Cy5-labelled goat anti-rabbit IgG (diluted 1:10,000 in blocking buffer) in the dark for 1 h at room temperature. Membranes were washed five times with the washing buffer and scanned with a Typhoon TRIO+ variable mode imager.

Preparation of Bacillus spores

B. subtilis wild-type or isogenic fengycin mutant strains were inoculated from a preculture (1:100) into 1 litre of 2× SG medium63 and allowed to sporulate for 96 h. Cells were pelleted, washed with water, and resuspended in 20% metrizoic acid (Sigma). Five different concentrations (w/v) of metrizoic acid (60% to 20%) were added stepwise to a 50-ml centrifuge tube to obtain a density gradient. A cell suspension was added to the top of the gradient, and was followed by centrifugation at 40,000g for 60 min at 4 °C (as described previously64). Spores were found in the middle layers and were collected. They were washed three times with 10 ml water. The total obtained number of viable spores per ml was determined by serial dilution, plating on TSA, and counting of CFU. The total number of heat-resistant spores per ml was determined by submerging the spores in a water bath at 80 °C for 20 min, followed by serial dilution and quantification of CFU per ml as described above.

Mouse intestinal colonization model

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 of the 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.

All C57BL/6J mice were female and six to eight weeks of age at the time of use. In one set-up, before S. aureus was given by oral gavage, mice were pretreated to eradicate the pre-existing intestinal microbiota using an antibiotic mix consisting of ampicillin (1 g l−1), metronidazole (1 g l−1), neomycin trisulfate (1 g l−1) and vancomycin (1 g l−1) in the drinking water. The last day before gavage, antibiotics were omitted from the drinking water. No bacteria could be found in the faeces or intestines of mice for seven days after this treatment in a control experiment. In another set-up, antibiotic pretreatment was omitted. In all set-ups, S. aureus strains were grown to mid-exponential growth phase, washed, and resuspended in sterile phosphate-buffered saline (PBS) at 108 CFU ml−1. Mice were inoculated by oral gavage with 200 μl of a 108 CFU ml−1 suspension of the indicated S. aureus strains, or 1:1 mixtures of wild-type and isogenic agr mutants to reach the same final concentration and volume. For the experiments with strains containing plasmids of the pKXΔ type, mice received kanamycin (0.2 g l−1) in the drinking water during the experiment to maintain plasmids. For the B. subtilis spore competition experiment, oral gavage with 200 μl of spores of wild-type Bacillus or its isogenic ΔfenA fengycin mutant (108 CFU ml−1 in sterile PBS) was performed on the day following the S. aureus gavage, and repeated every second day thereafter for a total of three times (days 2, 4 and 6). Intestinal colonization was evaluated by quantitative cultures of mouse stool samples and samples from the small and large intestines of mice. In detail, stool was collected and suspended to a final volume of 1 ml of PBS, diluted and plated on TSB agar. Plates were incubated for 24 h at 37 °C, and colonies were enumerated. Moreover, after mice were euthanized seven days after infection, the small and large intestines were collected, resuspended each in 1 ml PBS, and homogenized. Serial dilutions of the homogenates were plated on TSB agar and incubated at 37 °C. Bacterial colonies were enumerated on the following day. In the experiments without antibiotic pretreatment, extracts were plated on MSA plates containing 4 μg ml−1 oxacillin (for strain USA300 LAC) or 3 μg ml−1 tetracycline (for tetracycline-resistant derivatives of strains ST88 and ST2196), incubated for 48 h at 37 °C, and enumerated.


Statistical analysis was performed using GraphPad Prism version 6.05 with one-way or two-way ANOVA, or Fisher’s exact test, as appropriate, except for the experiments shown in Figs. 2c, 5b, c, and Extended Data Figs. 2b, c, 5c–f, for which Stata Release 15 and Poisson regression were used, owing to the exclusive presence of 0 values in one group (no variance). For ANOVAs, Tukey post-tests were used, which correct for multiple comparisons using statistical hypothesis testing. All data show the mean and standard deviation (s.d.). All replicates are biological.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability

Microbiome sequencing data are available from Bioproject with accession number 483343. All other data generated or analysed during this study are included in the published Article or in the Supplementary Information.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Guarner, F. & Malagelada, J. R. Gut flora in health and disease. Lancet 361, 512–519 (2003).

  2. 2.

    Kamada, N., Chen, G. Y., Inohara, N. & Núñez, G. Control of pathogens and pathobionts by the gut microbiota. Nat. Immunol. 14, 685–690 (2013).

  3. 3.

    Gourbeyre, P., Denery, S. & Bodinier, M. Probiotics, prebiotics, and synbiotics: impact on the gut immune system and allergic reactions. J. Leukoc. Biol. 89, 685–695 (2011).

  4. 4.

    Macpherson, A. J. & Harris, N. L. Interactions between commensal intestinal bacteria and the immune system. Nat. Rev. Immunol. 4, 478–485 (2004).

  5. 5.

    Bermudez-Brito, M., Plaza-Díaz, J., Muñoz-Quezada, S., Gómez-Llorente, C. & Gil, A. Probiotic mechanisms of action. Ann. Nutr. Metab. 61, 160–174 (2012).

  6. 6.

    Sassone-Corsi, M. et al. Microcins mediate competition among Enterobacteriaceae in the inflamed gut. Nature 540, 280–283 (2016).

  7. 7.

    Tam, N. K. et al. The intestinal life cycle of Bacillus subtilis and close relatives. J. Bacteriol. 188, 2692–2700 (2006).

  8. 8.

    Casula, G. & Cutting, S. M. Bacillus probiotics: spore germination in the gastrointestinal tract. Appl. Environ. Microbiol. 68, 2344–2352 (2002).

  9. 9.

    Duc, L. H., Hong, H. A., Barbosa, T. M., Henriques, A. O. & Cutting, S. M. Characterization of Bacillus probiotics available for human use. Appl. Environ. Microbiol. 70, 2161–2171 (2004).

  10. 10.

    Hong, H. A., Duc, L. H. & Cutting, S. M. The use of bacterial spore formers as probiotics. FEMS Microbiol. Rev. 29, 813–835 (2005).

  11. 11.

    Fujiya, M. et al. The Bacillus subtilis quorum-sensing molecule CSF contributes to intestinal homeostasis via OCTN2, a host cell membrane transporter. Cell Host Microbe 1, 299–308 (2007).

  12. 12.

    Lowy, F. D. Staphylococcus aureus infections. N. Engl. J. Med. 339, 520–532 (1998).

  13. 13.

    Lowy, F. D. Antimicrobial resistance: the example of Staphylococcus aureus. J. Clin. Invest. 111, 1265–1273 (2003).

  14. 14.

    Septimus, E. J. & Schweizer, M. L. Decolonization in prevention of health care-associated infections. Clin. Microbiol. Rev. 29, 201–222 (2016).

  15. 15.

    Dickey, S. W., Cheung, G. Y. C. & Otto, M. Different drugs for bad bugs: antivirulence strategies in the age of antibiotic resistance. Nat. Rev. Drug Discov. 16, 457–471 (2017).

  16. 16.

    Wertheim, H. F. et al. The role of nasal carriage in Staphylococcus aureus infections. Lancet Infect. Dis. 5, 751–762 (2005).

  17. 17.

    von Eiff, C., Becker, K., Machka, K., Stammer, H. & Peters, G. Nasal carriage as a source of Staphylococcus aureus bacteremia. N. Engl. J. Med. 344, 11–16 (2001).

  18. 18.

    Simor, A. E. & Daneman, N. Staphylococcus aureus decolonization as a prevention strategy. Infect. Dis. Clin. North Am. 23, 133–151 (2009).

  19. 19.

    Williams, R. E. Healthy carriage of Staphylococcus aureus: its prevalence and importance. Bacteriol. Rev. 27, 56–71 (1963).

  20. 20.

    Mody, L., Kauffman, C. A., Donabedian, S., Zervos, M. & Bradley, S. F. Epidemiology of Staphylococcus aureus colonization in nursing home residents. Clin. Infect. Dis. 46, 1368–1373 (2008).

  21. 21.

    Eveillard, M. et al. Evaluation of a strategy of screening multiple anatomical sites for methicillin-resistant Staphylococcus aureus at admission to a teaching hospital. Infect. Control Hosp. Epidemiol. 27, 181–184 (2006).

  22. 22.

    Acton, D. S., Plat-Sinnige, M. J., van Wamel, W., de Groot, N. & van Belkum, A. Intestinal carriage of Staphylococcus aureus: how does its frequency compare with that of nasal carriage and what is its clinical impact? Eur. J. Clin. Microbiol. Infect. Dis. 28, 115–127 (2009).

  23. 23.

    Senn, L. et al. The stealthy superbug: the role of asymptomatic enteric carriage in maintaining a long-term hospital outbreak of ST228 methicillin-resistant Staphylococcus aureus. MBio 7, e02039-e15 (2016).

  24. 24.

    Squier, C. et al. Staphylococcus aureus rectal carriage and its association with infections in patients in a surgical intensive care unit and a liver transplant unit. Infect. Control Hosp. Epidemiol. 23, 495–501 (2002).

  25. 25.

    Lindberg, E. et al. High rate of transfer of Staphylococcus aureus from parental skin to infant gut flora. J. Clin. Microbiol. 42, 530–534 (2004).

  26. 26.

    Bhalla, A., Aron, D. C. & Donskey, C. J. Staphylococcus aureus intestinal colonization is associated with increased frequency of S. aureus on skin of hospitalized patients. BMC Infect. Dis. 7, 105 (2007).

  27. 27.

    Ray, A. J., Pultz, N. J., Bhalla, A., Aron, D. C. & Donskey, C. J. Coexistence of vancomycin-resistant enterococci and Staphylococcus aureus in the intestinal tracts of hospitalized patients. Clin. Infect. Dis. 37, 875–881 (2003).

  28. 28.

    Klotz, M., Zimmermann, S., Opper, S., Heeg, K. & Mutters, R. Possible risk for re-colonization with methicillin-resistant Staphylococcus aureus (MRSA) by faecal transmission. Int. J. Hyg. Environ. Health 208, 401–405 (2005).

  29. 29.

    Misawa, Y. et al. Staphylococcus aureus colonization of the mouse gastrointestinal tract is modulated by wall teichoic acid, capsule, and surface proteins. PLoS Pathog. 11, e1005061 (2015).

  30. 30.

    Cheung, G. Y., Wang, R., Khan, B. A., Sturdevant, D. E. & Otto, M. Role of the accessory gene regulator agr in community-associated methicillin-resistant Staphylococcus aureus pathogenesis. Infect. Immun. 79, 1927–1935 (2011).

  31. 31.

    Miller, M. B. & Bassler, B. L. Quorum sensing in bacteria. Annu. Rev. Microbiol. 55, 165–199 (2001).

  32. 32.

    Holtfreter, S. et al. Characterization of a mouse-adapted Staphylococcus aureus strain. PLoS One 8, e71142 (2013).

  33. 33.

    Diep, B. A. et al. Complete genome sequence of USA300, an epidemic clone of community-acquired meticillin-resistant Staphylococcus aureus. Lancet 367, 731–739 (2006).

  34. 34.

    Dastgheyb, S. S. et al. Role of phenol-soluble modulins in formation of Staphylococcus aureus biofilms in synovial fluid. Infect. Immun. 83, 2966–2975 (2015).

  35. 35.

    Novick, R. P. & Geisinger, E. Quorum sensing in staphylococci. Annu. Rev. Genet. 42, 541–564 (2008).

  36. 36.

    Pathak, K. V., Keharia, H., Gupta, K., Thakur, S. S. & Balaram, P. Lipopeptides from the banyan endophyte, Bacillus subtilis K1: mass spectrometric characterization of a library of fengycins. J. Am. Soc. Mass Spectrom. 23, 1716–1728 (2012).

  37. 37.

    Cochrane, S. A. & Vederas, J. C. Lipopeptides from Bacillus and Paenibacillus spp.: a gold mine of antibiotic candidates. Med. Res. Rev. 36, 4–31 (2016).

  38. 38.

    Chang, L. K. et al. Construction of Tn917ac1, a transposon useful for mutagenesis and cloning of Bacillus subtilis genes. Gene 150, 129–134 (1994).

  39. 39.

    Lyon, G. J., Wright, J. S., Muir, T. W. & Novick, R. P. Key determinants of receptor activation in the agr autoinducing peptides of Staphylococcus aureus. Biochemistry 41, 10095–10104 (2002).

  40. 40.

    Ji, G., Beavis, R. & Novick, R. P. Bacterial interference caused by autoinducing peptide variants. Science 276, 2027–2030 (1997).

  41. 41.

    Otto, M., Echner, H., Voelter, W. & Götz, F. Pheromone cross-inhibition between Staphylococcus aureus and Staphylococcus epidermidis. Infect. Immun. 69, 1957–1960 (2001).

  42. 42.

    Brisson, J. HSO’s part 2—is Bacillus subtilis dangerous? Fix Your Gut http://fixyourgut.com/hso-probiotics-part-2-danger-supplementing-bacillus-subtilis/ (2014).

  43. 43.

    Vanittanakom, N., Loeffler, W., Koch, U. & Jung, G. Fengycin—a novel antifungal lipopeptide antibiotic produced by Bacillus subtilis F-29-3. J. Antibiot. 39, 888–901 (1986).

  44. 44.

    Khan, B. A., Yeh, A. J., Cheung, G. Y. & Otto, M. Investigational therapies targeting quorum-sensing for the treatment of Staphylococcus aureus infections. Expert Opin. Investig. Drugs 24, 689–704 (2015).

  45. 45.

    Poovelikunnel, T., Gethin, G. & Humphreys, H. Mupirocin resistance: clinical implications and potential alternatives for the eradication of MRSA. J. Antimicrob. Chemother. 70, 2681–2692 (2015).

  46. 46.

    Miranda, C. A., Martins, O. B. & Clementino, M. M. Species-level identification of Bacillus strains isolates from marine sediments by conventional biochemical, 16S rRNA gene sequencing and inter-tRNA gene sequence lengths analysis. Antonie van Leeuwenhoek 93, 297–304 (2008).

  47. 47.

    Carrel, M., Perencevich, E. N. & David, M. Z. USA300 methicillin-resistant Staphylococcus aureus, United States, 2000–2013. Emerg. Infect. Dis. 21, 1973–1980 (2015).

  48. 48.

    Wang, R. et al. Identification of novel cytolytic peptides as key virulence determinants for community-associated MRSA. Nat. Med. 13, 1510–1514 (2007).

  49. 49.

    Gauger, T. et al. Intracellular monitoring of target protein production in Staphylococcus aureus by peptide tag-induced reporter fluorescence. Microb. Biotechnol. 5, 129–134 (2012).

  50. 50.

    Queck, S. Y. et al. RNAIII-independent target gene control by the agr quorum-sensing system: insight into the evolution of virulence regulation in Staphylococcus aureus. Mol. Cell 32, 150–158 (2008).

  51. 51.

    Monk, I. R., Shah, I. M., Xu, M., Tan, M. W. & Foster, T. J. Transforming the untransformable: application of direct transformation to manipulate genetically Staphylococcus aureus and Staphylococcus epidermidis. MBio 3, e00277-11 (2012).

  52. 52.

    Luong, T. T. & Lee, C. Y. Improved single-copy integration vectors for Staphylococcus aureus. J. Microbiol. Methods 70, 186–190 (2007).

  53. 53.

    Yasbin, R. E. & Young, F. E. Transduction in Bacillus subtilis by bacteriophage SPP1. J. Virol. 14, 1343–1348 (1974).

  54. 54.

    Enright, M. C., Day, N. P., Davies, C. E., Peacock, S. J. & Spratt, B. G. Multilocus sequence typing for characterization of methicillin-resistant and methicillin-susceptible clones of Staphylococcus aureus. J. Clin. Microbiol. 38, 1008–1015 (2000).

  55. 55.

    Francois, P. et al. Rapid Staphylococcus aureus agr type determination by a novel multiplex real-time quantitative PCR assay. J. Clin. Microbiol. 44, 1892–1895 (2006).

  56. 56.

    Caporaso, J. G. et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 6, 1621–1624 (2012).

  57. 57.

    Caporaso, J. G. et al. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7, 335–336 (2010).

  58. 58.

    Edgar, R. C. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26, 2460–2461 (2010).

  59. 59.

    Rideout, J. R. et al. Subsampled open-reference clustering creates consistent, comprehensive OTU definitions and scales to billions of sequences. PeerJ 2, e545 (2014).

  60. 60.

    McDonald, D. et al. An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea. ISME J. 6, 610–618 (2012).

  61. 61.

    Lozupone, C. A., Hamady, M., Kelley, S. T. & Knight, R. Quantitative and qualitative beta diversity measures lead to different insights into factors that structure microbial communities. Appl. Environ. Microbiol. 73, 1576–1585 (2007).

  62. 62.

    Joo, H. S. & Otto, M. The isolation and analysis of phenol-soluble modulins of Staphylococcus epidermidis. Methods Mol. Biol. 1106, 93–100 (2014).

  63. 63.

    Nicholson, W. L. & Setlow, P. in Molecular Biological Methods for Bacillus (eds Harwood, C. R. & Cutting, S. M.) 391–450 (John Wiley, Chichester, 1990).

  64. 64.

    Fukushima, T. et al. Characterization of a polysaccharide deacetylase gene homologue (pdaB) on sporulation of Bacillus subtilis. J. Biochem. 136, 283–291 (2004).

Download references


We thank R. Kolter, Harvard Medical School, for providing the B. subtilis srfA mutant; D. Dubnau, Rutgers University, for the SPP1 phage; S. Holtfreter, University of Greifswald, and W. P. Zeng, Texas Tech University Health Sciences Center, for providing strain JSNZ/ST88; B. Krismer, University of Tübingen, for plasmid pKX15; F. DeLeo, National Institute of Allergy and Infectious Diseases (NIAID), for anti-Panton–Valentine-leucocidin; and N. A. Amissah for technical assistance. This work was supported by the Intramural Research Program of the NIAID, US National Institutes of Health (NIH) (project ZIA AI000904-16, to M.O.); and the Thailand Research Fund through the Royal Golden Jubilee PhD Program (grant number PHD/0072/2557, to P.P. and P.K.). P.K. was also supported by the Faculty of Medicine, Siriraj Hospital, Mahidol University, grant number (IO) R015833012; P.P. by the Graduate Partnership Program of the NIH; and S.W.D. by the Postdoctoral Research Associate Program of the National Institute of General Medical Sciences (1FI2GM11999101).

Reviewer information

Nature thanks A. Baumler, M. Parsek and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Author information

Author notes

    • Hwang-Soo Joo

    Present address: Department of Pre-PharmMed, College of Natural Sciences, Duksung Women’s University, Seoul, South Korea

  1. These authors contributed equally: Yue Zheng, Thuan H. Nguyen


  1. Pathogen Molecular Genetics Section, Laboratory of Bacteriology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD, USA

    • Pipat Piewngam
    • , Yue Zheng
    • , Thuan H. Nguyen
    • , Seth W. Dickey
    • , Hwang-Soo Joo
    • , Amer E. Villaruz
    • , Kyle A. Glose
    • , Emilie L. Fisher
    • , Rachelle L. Hunt
    • , Barry Li
    • , Janice Chiou
    • , Gordon Y. C. Cheung
    •  & Michael Otto
  2. Department of Microbiology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand

    • Pipat Piewngam
    • , Sujiraphong Pharkjaksu
    •  & Pattarachai Kiratisin
  3. Faculty of Veterinary Science, Rajamangala University of Technology Srivijaya, Nakhon Si Thammarat, Thailand

    • Sunisa Khongthong


  1. Search for Pipat Piewngam in:

  2. Search for Yue Zheng in:

  3. Search for Thuan H. Nguyen in:

  4. Search for Seth W. Dickey in:

  5. Search for Hwang-Soo Joo in:

  6. Search for Amer E. Villaruz in:

  7. Search for Kyle A. Glose in:

  8. Search for Emilie L. Fisher in:

  9. Search for Rachelle L. Hunt in:

  10. Search for Barry Li in:

  11. Search for Janice Chiou in:

  12. Search for Sujiraphong Pharkjaksu in:

  13. Search for Sunisa Khongthong in:

  14. Search for Gordon Y. C. Cheung in:

  15. Search for Pattarachai Kiratisin in:

  16. Search for Michael Otto in:


P.P., S.P. and S.K. collected human samples and analysed bacterial isolates by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI–TOF MS). Y.Z., H.-S.J. and M.O. performed analytical and preparative chromatography. P.P., T.H.N., E.L.F., R.L.H., J.C. and G.Y.C.C. performed animal studies. S.W.D. constructed the S. aureus agrBD mutant and A.E.V. constructed all other agr mutants and complemented strains. K.A.G., A.E.V. and B.L. performed MLST. P.P. performed reporter assays, the microbiome study, and all further analyses not specifically mentioned. P.K. supervised the human analyses and M.O. all other parts of the study.

Competing interests

The authors declare no competing interests.

Corresponding author

Correspondence to Michael Otto.

Extended data figures and tables

  1. Extended Data Fig. 1 Microbiome analysis of S. aureus carriers versus non-carriers.

    The microbiota of n = 20 randomly selected S. aureus carriers (red) and n = 20 non-carriers (blue) were analysed in faecal samples. a–c, Rarefaction (species-richness) curves based on 16S rRNA gene sequences. Data are mean ± s.d. a, Shannon index. b, Observed species. c, Chao1 index. d, Comparison of relative taxa abundance between S. aureus carriers (red) and non-carriers (blue). e, f, Beta diversity, represented by a principal coordinate analysis plot based on unweighted UniFrac (e) and weighted UniFrac (f) metrics for samples from S. aureus carriers (red) and non-carriers (blue).

  2. Extended Data Fig. 2 Quorum-sensing dependence of S. aureus intestinal colonization.

    Data from strains USA300 LAC and ST88 JSNZ. The experimental set-up is the same as in Fig. 2: mice received by oral gavage either 100 μl containing 108 CFU ml−1 of wild-type S. aureus strain USA300 LAC or ST88 JSNZ plus another 100 μl of 108 CFU ml−1 of the corresponding isogenic agr mutant (n = 5 per group; competitive experiment shown in a, b); or 200 μl containing 108 CFU ml−1 wild-type, isogenic agr mutant or Agr (RNAIII)-complemented agr mutant (n = 5 per group, non-competitive experiment shown in c). CFU in the faeces were determined two, four and six days after infection. At the end of the experiment (day seven), CFU in the small and large intestines were determined. a, b, Competitive experiment. Total obtained CFU are shown as dot plots; also shown are mean ± s.d. Bars show the percentage of wild-type among total determined CFU, of which 100 were analysed for tetracycline resistance that is present only in the agr mutant. No agr mutants were detected in any experiment; thus, all bars show 100%. Given that 100 isolates were tested, the competitive index wild-type/agr mutant in all cases is ≥100. c, Non-competitive experiment with genetically complemented strains. Wild-type and isogenic agr mutant strains all harboured the pKXΔ16 control plasmid; Agr-complemented strains harboured pKXΔRNAIII and thus constitutively expressed RNAIII, which is the intracellular effector of Agr. During the experiment, mice received 200 μg ml−1 kanamycin in their drinking water to maintain plasmids. Statistical analysis was performed using Poisson regression versus values obtained with the agr mutant strains. *P < 0.0001. Data are mean ± s.d. Note that no bacteria were found in the faeces or intestines of any mouse receiving S. aureus Δagr with vector control. The corresponding zero values are plotted on the x axis of the logarithmic scale. Source data

  3. Extended Data Fig. 3 Analysis of Agr-inhibitory substances.

    a, Influence of heat and proteases on Agr inhibition. B. subtilis culture filtrate was subjected to heat (95 °C for 20 min) or digestion with proteinase K (50 μg ml−1, 37 °C, 1 h) and the effect on inhibition of Agr activity was measured using the luminescence assay with the USA300 P3–luxABCDE reporter strain (see Fig. 3a). RLU, relative light units. The experiment was performed with n = 2 independent biological samples. Lines connect the means. (The observed additional suppression of Agr activity in the proteinase-K-treated sample at 6 h, compared with the B. subtilis culture filtrate sample, is expected owing to proteolytic inactivation of intrinsic AIP.) b, Preparative RP chromatography of B. subtilis culture filtrate to determine the Agr-inhibiting substance. The peaks labelled 2 and 3 showed substantial Agr-inhibiting activities in the Agr-activity assay and were identified as fengycins using subsequent RP-HPLC/ESI-MS and MS/MS analysis (see c, d). The peaks labelled 1 and 4–6 also contained fengycin species (see e). AU, arbitrary units. The applied gradient (% buffer B) is shown in green. c, Fractions corresponding to Agr-inhibitory peaks 2 and 3 from the preparative RP run (b) were subjected to RP-HPLC/ESI-MS. Top, total ion chromatograms (TICs) of the RP-HPLC/ESI-MS runs; bottom, ESI mass spectrogram of the major peaks. d, MS/MS analysis of the peak 2 and 3 fractions. Peaks that are characteristic of a given fengycin subtype (A or B in this case) are marked in colour. ‘Parent’ refers to the relevant numbered peak in the spectrograms above. e, Analysis of further fengycin-containing fractions. Peaks 1, 4, 5 and 6 from the preparative RP run (b) were also found to contain fengycin species as determined by subsequent RP-HPLC/ESI-MS analysis. Shown are the mass spectrograms of the major peaks of those runs and the tentative characterization for fengycin type. The preparative and analytical chromatography and RP-HPLC/ESI-MS analyses (as shown in b, d) were repeated multiple (more than ten) times for fengycin purification, with similar results. MS/MS analyses were not repeated. f, Analysis of fengycin and surfactin lipopeptide expression by the B. subtilis wild-type strain and its isogenic ΔfenA mutant. Source data

  4. Extended Data Fig. 4 Assessment of purity and functionality of purified β-OH-C17-fengycin B.

    a, RP-HPLC run. b, Agr inhibition at different concentrations in the luminescence assay. RLU, relative light units. Statistical analysis was by two-way ANOVA with Tukey’s post-test. Comparisons shown are those versus DMSO control. c, Agr inhibition as measured by inhibition of expression of RNAIII by qRT–PCR. *P < 0.0001 (one-way ANOVA with Tukey’s post-test; comparisons shown are those versus 0 μM value). The experiments in b, c were performed with n = 3 independent biological samples. Data are mean ± s.d. Source data

  5. Extended Data Fig. 5 Inhibition of S. aureus colonization by dietary fengycin-producing Bacillus spores in a mouse model.

    a, Concentration of AIP-I during S. aureus growth. Strain LAC (USA300) was grown in TSB, and AIP-I concentrations were measured by RP-HPLC/ESI-MS. Calibration was performed using synthetic AIP-I. The detection limit of this assay is around 0.3 μM. The experiment was performed with n = 3 independent biological samples. Data are mean ± s.d. b, B. subtilis colonization kinetics in the mouse intestinal colonization experiment. Mice (n = 5) received 200 μl of a 108 CFU ml−1 suspension of wild-type B. subtilis or ΔfenA mutant spores by oral gavage; CFU in the faeces were analysed up to five days afterwards. Data are mean ± s.d. c–f, Inhibition mouse model with strains USA300 LAC and ST88 JSNZ. The experimental set-up was as shown in Fig. 5a. In brief, n = 4 or 5 mice per group received 200 μl of 108 CFU ml−1 S. aureus strains USA300 LAC or ST88 JSNZ by oral gavage. On the next day and every following second day, the mice received 200 μl of 108 CFU ml−1 spores of wild-type B. subtilis or its isogenic fenA mutant, also by oral gavage. CFU in the faeces were determined two, four and six days after infection. At the end of the experiment (day seven), CFU in the small and large intestines were determined. The experiment was performed with (c, d) or without (e, f) antibiotic pretreatment. Statistical analysis was performed using Poisson regression versus values obtained with wild-type B. subtilis spore samples. *P < 0.0001. Data are mean ± s.d. Note that no S. aureus were found in the faeces or intestines of any mouse challenged with any S. aureus strain that also received Bacillus wild-type spores. The corresponding zero values are plotted on the x axis of the logarithmic scale. Source data

  6. Extended Data Table 1 Fengycin production and Agr-inhibition potency of Bacillus faecal isolates
  7. Extended Data Table 2 Analysis of previous microbiome studies for correlation between the presence of S. aureus and B. subtilis in the human intestinal tract

Supplementary information

  1. Supplementary Information

    This file contains Supplementary Tables 1-2 and Supplementary Figure 1, the uncropped gels.

  2. Reporting Summary

Source data

About this article

Publication history




Issue Date




By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.