Identification of potent immunostimulatory commensals from the human microbiota

To understand the impact of different commensal taxa on immunity, we cultured a broad panel of organisms from the major phyla from the human intestinal microbiota: the Bacteroidetes, Firmicutes, Actinobacteria and Proteobacteria[6]. We assembled a panel of 35 bacteria, the majority (27/35) from the dominant two phyla within the intestinal microbiota: Firmicutes and Bacteroidetes (Fig 1A). Our panel of commensals included members of three of the most common bacterial classes in the intestinal microbiota: Clostridia (Firmicutes), Bacilli (Firmicutes), and Bacteroidia (Bacteroidetes)[6]. We wanted to understand how these commensals influence innate immunity as this arm of the immune system is heavily influence by commensal bacteria[25]. To begin to understand the immunostimulatory capacity of our commensal panel, we stimulated the macrophage-like cell line (J774A.1) with each commensal individually and measured the production of the cytokines IL6 and TNF. We chose macrophages as a model because in vivo these cells are heavily influenced by signals from the microbiota[26], they are positioned within tissues at the interface with the microbiota[27], and express a broad complement of PRRs[28]. We measured IL6 and TNF as the production of these cytokines in macrophages relies on PRRs[29,30] and can therefore be used as a “readout” of commensal-induced innate immune activation (hereafter referred to as “immunostimulation”); IL6 and TNF are critical host signals in the intestine that mediate many of the effects of the microbiota on the immune system[27,31,32]; and they are products of primary (Tnf) and secondary (Il6) response genes[29,30,33], thus relying on different mechanisms to control their production. To begin our analysis of this, we used heat-killed bacteria to separate the impact of structural PRR ligands produced by each commensal, which are insensitive to heat treatment, from the metabolic products of commensals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Identification of commensal microbes that are robust activators of innate immunity.

(a) Cladogram of intestinal commensal species used in this study. Heat map displaying fold induction in macrophage TNF and IL6 production relative to unstimulated cells (mean values for each species are shown, n = 6–15 biological repeats, at MOI of 1:10). (b) Absolute levels of TNF produced by each commensal species ranked from low to high stimulators within each phyla. (c) Comparison of average induction of TNF and IL6 by members of the Bacteroidetes and Firmicute phyla. Each data point is mean level of cytokine production for a single commensal species, horizontal lines indicate median values. (d) Intestinal IL6 levels 24 hours post oral inoculation with 1×10⁸ CFU indicated commensal species. Each point represents a single mouse and horizontal lines indicate median values (HIGH STIMULATORS: Actinobacteria: Bifidobacterium adolescentis; Bacteroidetes: Bacteroides caccae; Firmicutes: Eubacterium infirmum; Proteobacteria: Acinetobacter radioresistens) (LOW STIMULATORS: Actinobacteria: Bifidobacterium breve; Bacteroidetes: Bacteroides cellulosilyticus; Firmicutes: Enterococcus raffinosus; Proteobacteria: Citrobacter koseri). All mice were antibiotic treated unless indicated otherwise prior to commensal administration. (e,f,g) Intestinal cytokine levels 4 (e) and 24 (f,g) hours post oral inoculation with 1×10⁸ CFU of indicated commensal consortium (BACTEROIDETES consortium: Bacteroides caccae, Bacteroides cellulosilyticus, Bacteroides dorei, Bacteroides eggerthii, Bacteroides finegoldii, Bacteroides ovatus, Bacteroides thetaiotaomicron, Bacteroides uniformis, Parabacteroides johnsonii; FIRMICUTES consortium: Clostridium histolyticum, Clostridium orbiscindens, Enterococcus faecalis, Enterococcus raffinosus, Eubacterium limosum, Faecalibacterium prausnitzii, Lactobacillus reuteri, Lactobacillus crispatus, Paenibacillus barengoltzii, Peptostreptococcus anaerobius). For (g) indicated mice were treated with clodronate liposomes or empty liposomes 2 days prior and concomitant with consortia inoculation. All mice were antibiotic treated unless indicated otherwise prior to commensal administration. Horizontal lines indicate mean values (d-g) and statistical comparisons were by Mann Whitney test *p < 0.05, **p < 0.01, and NS, not significant.

https://doi.org/10.1371/journal.ppat.1009191.g001

Across the panel as a whole, we found that there was significant variation in the immunostimulatory activity of commensals, as measured by the levels of TNF and IL6 they induce in macrophages (Fig 1A and 1B). This variation was evident between commensal species within the major phyla Bacteroidetes and Firmicutes. For example, Bacteroides caccae (Bacteroidetes) and Lactobacillus crispatus (Firmicutes) were both potent innate immune activators. Bacteroides caccae induced an 35-fold increase in TNF and 332-fold increase in IL6, above unstimulated controls. Similarly, Lactobacillus crispatus induced a 58-fold increase in TNF and 401-fold increase in IL6. By contrast, Bacteroides cellulosilyticus (Bacteroidetes) and Clostridium symbiosum (Firmicutes) both induced weak immune stimulators. Bacteroides cellulosilyticus induced an 9-fold increase in TNF and 0.4-fold increase in IL6, above unstimulated controls. Similarly, Clostridium symbiosum induced a 4-fold increase in TNF and 0.1-fold increase in IL6. Our screen of commensal-driven innate immune activation was performed with a macrophage-like cell line (J774A.1), next, therefore, we used bone marrow-derived macrophages (BMDMs) to confirm that the results from our cell line were broadly consistent between cells and thus a reliable reflection of the immunostimulatory capacities of our commensal panel. We stimulated BMDMs with a number of our commensal species and measured cytokine production. We found a strong correlation between the level of cytokine production after commensal stimulation of our cell line and BMDMs (S1 Fig), supporting the notion that cytokine production by our macrophage cell line is a fair reflection of the immunostimulatory capacity of our commensal panel. As both the J774A.1 cells and bone marrow-derived macrophages are murine cells, we wanted to ensure that our commensals elicited a similar pattern of immune activation in human cells too. To do this we stimulated differentiated THP-1 cells with members of our commensal panel. Using this human macrophage-like cell line, we found a strong correlation between the levels of cytokine elicited by commensals stimulation from these cells and our murine cell line (S2 Fig) indicating that our murine models were relevant to humans and that the commensal species in our panel are likely to have a similar effect on human and mouse cells. Finally, to ensure that any differences in cytokine production were not due to the differential induction of cell death in macrophages by commensals, we measured lactate dehydrogenase release by macrophages after commensal stimulation. We found that none of our commensals induced macrophage cell death confirming that any differences in cytokine induction was solely down to differences in the immunostimulatory capacity of each species (S3 Fig).

Bacteroidetes and Firmicutes have differential effects on primary and secondary responses genes in the intestine

Next, we wanted to understand whether there were any trends in innate immune activation that reflected the taxonomy of our commensal panel. To investigate whether there was any difference in immunostimulatory capacity between the dominant intestinal phyla, we compared mean cytokine production induced by each Bacteroidetes species with the mean cytokine production elicited by each Firmicute species. We found no significant difference in the levels of TNF induced by Bacteroidetes and Firmicutes (Fig 1C). By contrast, we found that commensal Bacteroidetes tended to induce significantly more IL6 than the Firmicutes (Fig 1C). This indicates that, as a whole, there is no distinction in the ability of Bacteroidetes and Firmicutes to induce cytokines that are the products of primary response genes, however, Bacteroidetes tend to be better at inducing the production cytokines from secondary response genes.

Next, we wanted to determine whether the immunostimulatory potential of commensals identified in our in vitro macrophage screen reflected their effect on host immunity in vivo. We examined the ability of commensals from our panel identified as either high or low cytokine stimulators to regulate intestinal IL6 production as this cytokine plays a central role in the intestinal immune system during both homeostasis and disease [11,31]. To do this, we depleted the microbiota in mice using broad spectrum antibiotics. We found that IL6 production in the intestine is reduced after microbiota depletion demonstrating the importance of the microbiota for driving the homeostatic production of this cytokine in the intestine (Fig 1D). We then examined the ability of immunologically potent and immunologically weak stimulators of macrophages from all of the major intestinal phyla to regulate intestinal cytokine production after microbiota depletion. All of the potent macrophage activators from each phyla (Actinobacteria: Bifidobacterium adolescentis; Bacteroidetes: Bacteroides caccae; Firmicutes: Enterococcus raffinosus; Proteobacteria: Acinetobacter radioresistens) were able to rescue IL6 production after microbiota depletion (Fig 1D). All of the weak macrophage activators (Actinobacteria: Bifidobacterium breve; Bacteroidetes: Bacteroides cellulosilyticus; Firmicutes: Eubacterium infirmum; Proteobacteria: Citrobacter koseri) were significantly poorer at rescuing IL6 production than the robust stimulator from the same phylum (Fig 1E). This demonstrates that our macrophage screen accurately reflects the immunostimulatory activity of commensals in the intestine. Next, we wanted to test whether we could use this information on the immunostimulatory properties on individual commensals to predict the effect of more complex commensal communities on immunity in vivo. At the phylum level, we found that members of the Bacteroidetes tended to induce higher IL6 production, a product of a secondary response gene, than Firmicutes, but they had similar effects on TNF, the product of a primary response gene. This indicates that commensals from these phyla are likely to have different effects on programming our immune system and therefore led to two questions. Firstly, do multi-species communities of Bacteroidetes and Firmicutes have differential effects on these cytokines in vivo? And secondly, are the effects of these phyla generalizable to other primary and secondary response genes? To answer this, we inoculated microbiota-depleted mice (using antibiotic treatment) with consortium of either Bacteroidetes or Firmicutes and then measured the intestinal levels of the products of primary response genes (including TNF, CXCL1, CXCL2 and GM-CSF) and the products of secondary response genes (including IL6, IL12b and Lipocalin2) [29,30,33]. We found that our Bacteroidetes and Firmicutes consortia both induced primary gene product production (after both 4 and 24 hours) and there was no difference in the levels of primary gene products induced by these consortia (Fig 1E and 1F). By contrast, we found little induction of secondary genes by either the Bacteroidetes or the Firmicutes consortium at our early timepoint (Fig 1E), but at the later timepoint we found that Bacteroidetes induced significantly higher levels of secondary response genes than a consortium of Firmicutes at (Fig 1F), consistent with the slower induction kinetics of secondary response gene products. This pattern was consistent whether the commensals were alive or heat-killed (S4 Fig). To further corroborate these findings, we inoculated germ-free mice with either the Bacteroidetes or Firmicutes consortium. As with our microbiota depletion model using antibiotics, we found no difference in the ability of Bacteroidetes and Firmicutes to induce primary gene products but Bacteroidetes were better able to induce secondary gene products compared to Firmicutes (S5 Fig). Next, we wanted to understand whether macrophages were involved in the production of these cytokines downstream of commensals in the intestine. To do this we used liposome clodronate which specifically depletes macrophages prior to commensal inoculation. We found that after clodronate treatment there was reduced production of cytokines which were the product of both primary and secondary response genes (Fig 2G). By contrast, depletion of neutrophils had no effect on cytokine levels in the intestinal mucosa (S6 Fig). Collectively, these data begin to reveal taxonomic patterns in the effect of commensals on innate immunity, identify commensals with the potential to be major regulators of the innate immune system, and allow us to predict the immunological impact of complex commensal communities on the host based on their individual immunostimulatory properties. These taxonomic patterns in immune activation suggest that commensals within these two dominant phyla may engage different innate immune receptors and/or rely on different signaling pathways to exert their effects on the innate immune system.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Patterns of commensal immunostimulation within the Bacteroidetes and Firmicutes phyla reflect and rely upon the activation of different TLRs.

(a) TLR-dependent SEAP production by HEK293 cells 24 hours post-stimulation with indicated commensal species (at MOI of 1:10). Corresponding frequency distribution of TLR-dependent SEAP production for all commensals in our panel. (b) Comparison of TLR activation by members of the Bacteroidetes and Firmicute phyla. Each data point is mean fold increase in TLR-dependent SEAP activity elicited by a single commensal species, horizontal lines indicate median values. (c) Correlation between overall immunostimulatory capacity of commensals in indicated phyla (measured by macrophage IL6 production (using data from Fig 1A)) and TLR-dependent SEAP activation. Correlations were determined using the Pearson’s test. (d,e) Intestinal cytokine levels 24 hours post oral inoculation with 1×10⁸ CFU of indicated commensal consortium ((d) Clostridia consortium: Eubacterium infirnum, Clostridium histolyticum, Clostridium orbiscindens, Clostridium symbiosum, Eubacterium limosum, Faecalibacterium prausnitzii, Peptostreptococcus anaerobius. (e) Bacteroidia consortium: Bacteroides caccae, Bacteroides cellulosilyticus, Bacteroides dorei, Bacteroides eggerthii, Bacteroides finegoldii, Bacteroides ovatus, Bacteroides thetaiotaomicron, Bacteroides uniformis). Indicated groups of mice were treated with TLR neutralizing antibodies, or isotype control, (100 μg/mouse) concomitant with consortia inoculation. All mice were antibiotic treated unless indicated otherwise prior to commensal administration. Each point represents a single mouse and horizontal lines indicate mean values. (b) statistical comparisons were by Mann Whitney test, (e) statistical comparisons were by Kruskal–Wallis test with Dunn’s multiple comparison test. *p < 0.05, **p < 0.01, and NS, not significant.

https://doi.org/10.1371/journal.ppat.1009191.g002

Bacteroidetes and Firmicutes exhibit different patterns in TLR activation and reliance for their effects on intestinal immunity

To examine potential differences in innate receptor stimulation, next we assessed the ability of our commensal panel to stimulate individual TLRs. TLRs are one of the principal classes of PRRs which mediate the effect of commensals on immunity[9]. We specifically focused on the activation of TLR2, TLR4 or TLR5 as these are the major TLRs which recognize bacteria and have established roles in mediating the effect of the microbiota on the immune system[11,12,14,16,22]. We assessed TLR activation by our commensal panel using in vitro HEK cell reporter cells expressing individual TLRs which couple receptor activation to the production of a secreted alkaline phosphatase which can them be measured spectrophotometrically[13]. Analogously to stimulation of macrophages, we found significant variation in commensal stimulation of TLRs (Fig 2A). Across the panel as a whole there was a broad distribution of TLR2 and TLR4 stimulatory potential (Fig 2A). By comparison, the distribution of TLR5 stimulation was more bimodal indicating that commensals were either potent TLR5 activators or provoked minimal activation of this receptor (Fig 2A). Comparison of mean SEAP production downstream of TLR stimulation elicited by individual species of the Bacteroidetes and Firmicutes revealed that Bacteroidetes tended to be more potent TLR4 activators than Firmicutes (Fig 2B). In comparison, there was no difference in TLR2 or TLR5 activation between these two major phyla (Fig 2B). Within the Firmicutes, commensals from the Clostridia class more potently activated both TLR4 and TLR2 compared to commensals from the Bacilli class (Fig 2B). As TLR4 is classically thought to be a receptor for LPS from Gram-negative organisms, rather than Gram-positives like Firmicutes, we did further controls to ensure that there was no contaminating LPS in our models. Firstly, to ensure that the overall immunostimulatory capacity of the Firmicutes that activate TLR4 was not due to LPS contamination, we stimulated macrophages with these organisms (E. limosum, E. infirmum, C. histolyticum, P. anaerobius, C. orbiscindens, F. prausnitzii, L.reuteri, P. barengoltzii, and S. aureus) with and without the LPS binding antibiotic polymyxin B present. We found that cytokine induction by these Firmicutes was unaffected by the presence of polymyxin B, whereas the stimulatory capacity of LPS was greatly diminished by polymyxin B (S7A Fig). Secondly, we stimulated our TLR4 reporter cells with these Firmicutes with and without polymyxin B present. Again, we found that TLR4 activation by these Firmicutes was unaffected by the presence of polymyxin B, whereas TLR4 activation by LPS was diminished by polymyxin B (S7B Fig). These data suggest that TLR4 activation by these Firmicutes was not due to contaminating LPS. Finally, to ensure that there was no LPS in the microbial growth media used to culture our Firmicutes which could have been carried over into our cell stimulation assays after washing the commensals, we stimulated macrophages with the media used to culture these Firmicutes with or without polymyxin B present. We found that the culture media elicited minimal cytokine production in macrophages and was unaffected by polymyxin B treatment indicating that there was no contaminating LPS in the growth media used to culture these Firmicutes (S7C Fig). Next, we wanted to determine whether we could identify the dominant TLR(s) responsible for mediating the immunostimulatory effects of commensals within our panel. To do this, we correlated macrophage cytokine production elicited by each commensal species with the degree to which they activate each TLR (Fig 2C). We found that for members of the Bacteroidia class their immunostimulatory capacity (measured by their ability to stimulate IL6 production in macrophages) correlated with how strongly they activated TLR4, but not TLR2 or TLR5 (Fig 2C). By contrast, the immunostimulatory capacity of Clostridia correlated with the robustness of both TLR2 and TLR4 activation (Fig 2C). The immunostimulatory effect of Bacilli did not correlate with the activation of any of the TLRs tested (Fig 2C). This led us to test whether members of these commensal classes within the Bacteroidetes and Firmicutes phyla rely on different TLRs to program host immunity in vivo. To test this, we blocked TLR signaling in microbiota-depleted mice using antibody neutralization prior to oral inoculation with a consortium of either Bacteroidia or Clostridia, and then determined the effects of these consortia on intestinal cytokine levels. We found that the stimulatory effect of Clostridia on intestinal IL6 production was only fully blocked after inhibition of both TLR2 and TLR4 indicating that the effect of this class of organisms is mediated by TLR2 or TLR4 activity (Fig 2D). By contrast we found that the regulation of intestinal IL6 production by Bacteroidia was blocked after neutralization of TLR4, but not TLR2 (Fig 2E). Together, these data reveal phylogenic patterns in TLR utilization required for different commensal taxa to exert their effects on the immune system.

Canonical NF-κB signaling calibrates the immunostimulatory power of commensal communities

The transduction of microbial TLR stimulation into changes in cytokine production and immune cell behavior occurs via a limited number of intracellular signaling components[34]. These conserved intracellular pathways must integrate information from the concomitant activation of multiple TLRs by commensals, each activated at differing intensities and with different TLR ligand combinations, to determine changes in immune cell behavior and function downstream of commensal stimulation. How a limited number of signaling components achieve this information transfer to grade changes in immune cell behavior and cytokine production elicited by different commensals is poorly understood[35,36]. Next, therefore, we wanted to examine the intracellular signaling components required to transduce activation of innate receptors effect of each of our commensal species and identify any critical signaling components which determine the degree of immunostimulation elicited by commensals in our panel (Fig 3A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Canonical NF-κB is the ultimate determinant of the immunostimulatory power of commensals.

(a) Schematic of important intracellular signaling components downstream of TLRs, components inhibited in this study underlined in red. (b) Heat map displaying percentage inhibition in IL6 production after pharmacological inhibition of IRAK4 (100 nM PF06650833), P38 (100 μM SB203580), JNK (10 μM SP600125), ERK (10 μM U0126), Syk (20 μM Piceatannol), PKC (10 nM GO6983), canonical NF-κB inhibitor IKKβ (10 μM BOT-64), non-canonical NF-κB inhibitor IKKε/TBK1 (100 μM Amlexanox) and AP-1 (50 μM SR11302). Data are relative to macrophages treated with vehicle control, with mean values for percentage inhibition for each species are shown (n = 4–10 biological repeats, at MOI of 1:10). Corresponding frequency distribution of cytokine inhibition for all commensals in our panel. (c) Correlation between overall immunostimulatory capacity of all commensals in our panel (measured by macrophage IL6 production) and their dependence on the indicated signaling component (measured by cognate reduction in cytokine production incurred after inhibition of each of the intracellular signaling components). (d) Correlation between overall immunostimulatory capacity of all commensals in our panel (measured by macrophage IL6 production) and their dependence on canonical NF-κB signaling (measured by cognate reduction in cytokine production incurred after inhibition of IKKβ). Correlations were determined by Spearman’s test with False Discovery Rate correction for multiple comparisons, adjusted p-values stated with p < 0.05 considered significant.

https://doi.org/10.1371/journal.ppat.1009191.g003

We began by examining whether the IRAK4 signaling is required for the immunostimulatory effect of commensals in our panel. IRAK4 is a threonine/serine kinase that is essential for signaling downstream of MYD88 and hence is required for all TLR signaling except for TLR3[37]. Inhibiting IRAK4 therefore gives an indication of the role of TLRs in recognizing commensals in our panel. After pharmacological inhibition of IRAK4, we found that this inhibitor caused a significant reduction commensal-induced cytokine production for 27/35 of the commensals within our panel (Figs 3B and S8) highlighting the importance of TLRs for the immunostimulatory effect of commensals. The requirement for TLR signaling did, however, vary across our panel suggesting commensal species relied on TLRs to different extents to exert their immunostimulatory effects (Fig 3B). Next, we examined whether there were specific components of the intracellular signal transduction machinery required downstream of receptor and IRAK4 activation which control the immunostimulatory effect of commensals. To do this, we pharmacologically inhibited the MAP kinases (P38, ERK and JNK), protein kinase C and Syk. All of these signaling components are known to be important in meditating the effects of individual TLR activation on cytokine production and immune cell behavior[38–41]. We found that inhibition of the MAPKs had a pronounced effect on the immunostimulatory capacity of some commensals, for example P. barengoltzii and E. limosum, but had minimal effect on cytokine production driven by other commensal species, such as B. angulatum (Fig 3B). Inhibition of Syk and protein kinase C (α, β, γ, δ, ζ and μ isoforms) had a similar range of effects, with the immunostimulatory effect of some commensals highly dependent on Syk signaling, for example B. thetaiotamicron whereas the immunostimulatory effect of other species was independent of these pathways, for example E. faecium (Fig 3B). We then used this information to understand whether there were any broad trends in signaling component dependence across our commensal panel which could reveal critical parts of the intracellular signaling machinery which play a conserved role in dictating the degree of immunostimulation commensal communities exert. To do this, we correlated the immunostimulatory effect of each commensals in our panel with their dependence on components of the intracellular signal transduction machinery: MAPKs, Syk and PKC signaling. Specifically, we correlated mean macrophage IL6 production elicited by each commensal species (as a marker of commensal species immunostimulation) with the cognate reduction in IL6 production incurred after inhibition of each of the intracellular signaling components (as a marker for the dependence of each commensal species on that pathway for its immunostimulatory effect) (Fig 3C). We found no correlation between the level of cytokine production elicited by commensals in our panel and dependence on P38, JNK, ERK, Syk or PKC signaling (Fig 3C). Collectively, these data suggest that each commensal species has a distinct profile of intracellular signal components required to translate its pattern of innate receptor stimulation into changes in cytokine production. Furthermore, these data indicate that none of these components likely act as a conserved regulator dictating the level of immunostimulation commensal communities will elicit on host immunity. We then interrogated whether the transcription factors NF-κB or AP-1 acts as this critical control. We found that there was a significant positive correlation between the immunostimulatory capacity of commensals and their dependence on IKKβ of the canonical NF-κB signaling complex, but not AP-1 or non-canonical NF-κB (Fig 3C and 3D). To further substantiate these findings, we used siRNA to inhibit IKKβ of the canonical NF-κB signaling pathway (S9A Fig). As with pharmacological inhibition of IKKβ, we found that there was a significant positive correlation between the immunostimulatory capacity of commensals and their dependence on IKKβ (S9B Fig). This pattern was consistent within the Firmicute and Bacteroidetes phyla (S9B Fig). In contrast, and again in agreement with our data using pharmacological inhibition, we found there was no correlation between the immunostimulatory capacity of commensals and their dependence on JNK signaling (rho = 0.31, p = 0.07). Thus, signaling through the canonical NF-κB complex may act as a control hub to calibrate the immunological impact of commensal communities on the host.

Rational design of commensal consortia to treat intestinal infections and inflammation

Having determined the immunological fingerprint of commensals in our panel, next we wanted to use this information to rationally design communities of commensals to treat intestinal diseases by modulating specific components of the innate immune system and examine whether the effects of these commensal groups are determined by canonical NF-κB signaling. One of the most significant clinical problems caused by microbiota dysbiosis is increased susceptibility to enteric infection, a problem exacerbated by increasing antibiotic resistance[4,42]. Based on our commensal analysis, we assembled a cross-phylum consortium of commensals that demonstrated potent stimulation of macrophages (each species demonstrating induction of >500 pg/mL of TNF) and were robust activators of TLRs (each species demonstrating >50-fold activation of at least two TLRs) (Fig 4A). We called this our “high immune-stimulating” consortium. We chose to construct a cross-phylum consortium comprising both Firmicutes (from the Clostridia class), which rely on TLR2 and TLR4 for their effects on intestinal immunity (Fig 2D), and Bacteroidetes (from the Bacteroidia class), which rely on TLR4 only for their effects on intestinal immunity (Fig 2E), to ensure that the immunostimulatory effect of our consortium was not dependent on a single TLR pathway. As a comparison, we also assembled a consortium of commensals that were poor macrophage stimulators (each species demonstrating induction of <500 pg/mL of TNF) and weak TLR activators (each species demonstrating <50-fold activation of all TLRs), our “low immune-stimulating” consortium (Fig 4A). We then tested the ability of these commensals to enhance innate immune defenses against intestinal colonization by a major antibiotic resistant pathogen: vancomycin resistant Enterococci (VRE)[43]. In mice with an intact microbiota VRE was unable to establish intestinal colonization, by contrast after microbiota depletion VRE was able to establish a high level of intestinal colonization (Fig 4B), demonstrating that the microbiota is important for host defense against colonization by these organisms. To test the efficacy of the commensal consortia we had designed in the regulation of intestinal defenses, we orally inoculated microbiota-depleted mice with either the “high immune-stimulating” consortium or the “low immune-stimulating” consortium, prior to oral inoculation with VRE. As predicted from our analysis, the “high immune-stimulating” consortium provided significantly more protection against intestinal colonization by VRE than the “low immune-stimulating” group (Fig 4C). Furthermore, we found that the “low immune-stimulating” consortium did not inhibit the protective effect of the “high immune-stimulating” consortium against VRE (Fig 4C). The same pattern of protection if we established VRE colonization and then administered our consortia (Fig 4D). This pattern was maintained after heat-inactivation of our consortia (S10 Fig) supporting a role for PRR stimulation rather than metabolic products from our commensals regulating intestinal defenses. To ensure that there was equivalent engraftment of our “high immune-stimulating” and “low immune-stimulating”, we used phyla- and species-specific 16s rRNA gene qPCR, combined with bacterial culturing, of fecal samples to determine the in vivo levels of the commensals we administered to mice. Via 16s qPCR, we found similar levels of total bacteria, Firmicutes and Bacteroidetes between mice administered the “high immune-stimulating” and “low immune-stimulating” consortia (S11A Fig). There was similar relative abundance of each member of each consortium suggesting similar levels of commensal engraftment (11B and 11C and S11B Fig). To ensure that our consortia were viable within the intestine, we used selective fecal culturing. In comparison to feces from microbiota-depleted mice, and in agreement with our 16s qPCR data, the feces of mice administered our “low immune-stimulating” consortium had high levels of Lactobacilli, Bacteroidetes, and Clostridia (S10B and S10C Fig). Likewise, in comparison to feces from microbiota-depleted mice, and again in agreement with our 16s qPCR data, the feces of mice administered our “high immune-stimulating” consortium had high levels of Bacteroidetes, Clostridia and P. anaerobius (S11B and S11C Fig). Collectively these data indicate the differences in the effects of these consortia was not due to different intestinal engraftment. As our immunological profiling had been based on stimulation of innate immune cells, next we wanted to determine whether the effect of our commensal consortia in vivo was solely mediated by the innate immune system or whether the adaptive immune system was also involved. To do this, we orally inoculated our “high immune-stimulating” and “low immune-stimulating” consortium into microbiota-depleted SCID mice prior to oral infection with VRE. We found that our “high immune-stimulating” consortium was still more effective than our “low immune-stimulating” consortium even in the absence of adaptive immunity supporting the notion that our consortium is enhances innate defenses in the intestine (Fig 4E). We then wanted to examine the role of macrophages in vivo. To do this, we use clodronate liposomes to deplete intestinal macrophages[44]. Clodronate treatment abrogated the difference between the “high immune-stimulating” and “low immune-stimulating” consortia (Fig 4F), supporting the central role of macrophages downstream of consortia stimulation in enhancing intestinal antibacterial defences. We then tested whether this protection occurred through the stimulation of TLRs. Concomitant neutralization of TLR2, TLR4 and TLR5 abrogated the protective effect of our commensal consortia (Fig 4G), demonstrating that this consortium programs a protective immune barrier against a major antibiotic resistant pathogen through TLRs. Finally, given the central role of IL6 in intestinal immunity we tested whether this cytokine was involved in the protective effect of our “high immune-stimulating” consortium. After antibody neutralization, our “high immune-stimulating” consortium was less effective at promoting the clearance of VRE from the intestine (Fig 4H).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Rational design of commensal consortia to treat intestinal infection and intestinal inflammation.

(a) Scatter plot of TLR activation (measured by TLR2-, TLR4- and TLR5-dependent SEAP activation) versus overall immunostimulatory capacity (measured by macrophage cytokine production) of indicated commensals from the Bacteroidetes and Firmicutes phyla. Commensals were grouped together into the “low immune-stimulators” (defined as <500 pg/mL TNF and <50-fold activation of all TLRs) and “high immune-stimulators” (defined as >500 pg/mL TNF and >50-fold activation of at least two TLRs). (b) VRE burden in the faeces of WT adult mice two days post-oral inoculation. Each point represents a single mouse and horizontal lines indicate median values. (c) VRE burden in the faeces of WT adult mice two days post-oral inoculation. Animals were orally inoculated with 1×10⁸ CFU of the “low immune-stimulating” commensal consortium, 1×10⁸ CFU of the “high immune-stimulating” commensal consortium, or 1×10⁸ CFU of the “low immune-stimulating” and “high immune-stimulating” commensal consortia combined, two days prior to VRE inoculation. (d) VRE burden in the faeces of WT adult mice four days post-oral inoculation. Animals were orally inoculated with 1×10⁸ CFU of the “low immune-stimulating” or “high immune-stimulating” commensal consortia two days post-VRE inoculation. (e) VRE burden in the faeces of adult SCID mice two days post-oral inoculation. Indicated animals were orally inoculated with 1×10⁸ CFU of the “low immune-stimulating” or “high immune-stimulating” commensal consortia two days prior to VRE inoculation. (f) VRE burden in the faeces of WT adult mice two days post-oral inoculation. Animals were orally inoculated with 1×10⁸ CFU of the “low immune-stimulating” or “high immune-stimulating” commensal consortia two days prior to VRE inoculation. Indicated mice were treated with clodronate liposomes or empty liposomes 2 days prior and concomitant with consortia inoculation. (g) VRE burden in the faeces of WT adult mice two days post-oral inoculation. Animals were orally inoculated with 1×10⁸ CFU of the “high immune-stimulating” commensal consortia two days prior to VRE inoculation. Concomitant with consortium inoculation mice were administered either isotype control or a cocktail of TLR neutralizing antibodies (αTLR2, αTLR4 and αTLR5, (100 μg/mouse)) via intraperitoneal injection. (h) VRE burden in the faeces of adult orally inoculated with the “high immune-stimulating” commensal consortia two days post-oral inoculation. Commensal consortia were administered two days prior to VRE inoculation. Concomitant with consortium inoculation mice were administered either isotype control or an IL6 neutralizing antibody via intraperitoneal injection (i,j) VRE burden in the faeces of WT adult mice orally inoculated with the “high immune-stimulating” commensal consortia two days post-oral inoculation with VRE. Commensal consortia were administered two days prior to VRE inoculation. For pharmacological inhibition of canonical NF-κB, mice were administered BOT-64 (30 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation. For pharmacological inhibition of P38 mice were administered SB203580 (2 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation. (k) Intestinal cytokine levels 24 hours post oral inoculation with 1×10⁸ CFU of indicated commensal consortium. Each point represents a single mouse and horizontal lines indicate median values. (l) Plasma FITC-dextran levels in mice with DSS-induced intestinal inflammation. For the induction of intestinal inflammation, antibiotic treated mice received 2% (w/v) DSS in drinking water for seven days. Indicated groups of mice were orally administered commensal consortia on day 2, day 4 and day 6 of DSS treatment. One day after cessation of DSS treatment, mice were orally administered 200 μL of FITC-dextran 4000 (80 mg/mL) and FITC-dextran 4000 levels in blood were measured after four hours. (m) Weight change in DSS-treated mice six days after start of DSS treatment. (n) Plasma FITC-dextran levels in mice with DSS-induced intestinal inflammation. For the induction of intestinal inflammation, antibiotic treated mice received 2% (w/v) DSS in drinking water for seven days. Indicated groups of mice were orally administered commensal consortia on day 2, day 4 and day 6 of DSS treatment. One day after cessation of DSS treatment, mice were orally administered 200 μL of FITC-dextran 4000 (80 mg/mL) and FITC-dextran 4000 levels in blood were measured after four hours. In indicated groups, concomitant with consortium inoculation mice were administered either isotype control or a cocktail of TLR neutralizing antibodies (αTLR2, αTLR4 and αTLR5, (100 μg/mouse)) or IL6 neutralizing antibody (100 μg/mouse), or isotype control via intraperitoneal injection. (o) Weight change in DSS-treated mice six days after start of DSS treatment. In indicated groups, concomitant with consortium inoculation mice were administered either isotype control or a cocktail if TLR neutralizing antibodies (αTLR2, αTLR4 and αTLR5, (100 μg/mouse)) or IL6 neutralizing antibody (100 μg/mouse), or isotype control via intraperitoneal injection. (p) Plasma FITC-dextran levels in mice with DSS-induced intestinal inflammation. For the induction of intestinal inflammation, antibiotic treated mice received 2% (w/v) DSS in drinking water for seven days. Indicated groups of mice were orally administered commensal consortia on day 2, day 4 and day 6 of DSS treatment. One day after cessation of DSS treatment, mice were orally administered 200 μL of FITC-dextran 4000 (80 mg/mL) and FITC-dextran 4000 levels in blood were measured after four hours. In indicated groups, concomitant with consortium inoculation, mice were treated with clodronate liposomes or empty liposomes concomitant with consortia inoculation. (q) Weight change in DSS-treated mice six days after start of DSS treatment. In indicated groups, concomitant with consortium inoculation, mice were treated with clodronate liposomes or empty liposomes concomitant with consortia inoculation. Each point represents a single mouse and horizontal lines indicate mean values. (r) Plasma FITC-dextran levels in mice with DSS-induced intestinal inflammation. For the induction of intestinal inflammation, antibiotic treated mice received 2% (w/v) DSS in drinking water for seven days. Indicated groups of mice were orally administered commensal consortia on day 2, day 4 and day 6 of DSS treatment. One day after cessation of DSS treatment, mice were orally administered 200 μL of FITC-dextran 4000 (80 mg/mL) and FITC-dextran 4000 levels in blood were measured after four hours. In indicated groups, for pharmacological inhibition of canonical NF-κB, mice were administered BOT-64 (30 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation concomitant with consortium inoculation. (s) Weight change in DSS-treated mice six days after start of DSS treatment. In indicated groups, for pharmacological inhibition of canonical NF-κB, mice were administered BOT-64 (30 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation concomitant with consortium inoculation. (b-h) Statistical comparisons were by Mann Whitney test, (i-s) statistical comparisons were by Kruskal–Wallis test with Dunn’s multiple comparison test. *p < 0.05, **p < 0.01, and NS, not significant.

https://doi.org/10.1371/journal.ppat.1009191.g004

We then wanted to examine whether canonical NF-κB signaling calibrates the level of antibacterial defenses our commensal consortia elicit in vivo. To do this, microbiota-depleted mice were treated with an inhibitor of canonical NF-κB signaling concomitant with oral inoculation of our commensal consortia. As a comparison, we treated separate groups of microbiota-depleted mice with a P38 inhibitor concomitant with commensal consortia inoculation, which from our earlier analysis plays a role in translating the effect of commensals into changes in host immunity but would not be the critical determinant of the degree of immunostimulation our commensal consortia elicit. If our model is correct, then the dependence of the “high immune-stimulating” consortium on NF-κB signaling to promote intestinal defenses will be larger than the NF-κB-dependence of the “low immune-stimulating” consortium. By contrast, for MAPK signaling the magnitude of dependence on P38 signaling should not reflect the immunostimulatory power of our consortia. Consistent with our hypothesis, in mice given the “high immune-stimulating” consortium, we found that inhibition of NF-κB caused an approximately 500-fold increase in intestinal VRE levels (Fig 4I), whereas in animals given the “low immune-stimulating” consortium NF-κB inhibition caused only a 10-fold increase in intestinal VRE levels (Fig 4J). By contrast, in mice given the P38 inhibitor, VRE levels increased approximately 10-fold in both “high immune-stimulating” and “low immune-stimulating” consortia treated mice (Fig 4I–4J). The regulation of intestinal cytokine levels by are rationally designed commensal consortia displayed a similar pattern of dependence (Fig 4K). Collectively, our data show that whilst a variety of intracellular signaling pathways are involved in translating commensal stimulation into changes in immune behavior it is ultimately NF-κB through IKKβ which acts as a critical control hub to calibrate the strength of the effect of commensal communities on immunity.

Stimulation of TLR signaling by the microbiota not only enhances intestinal antibacterial defenses but it is also required to restore intestinal homeostasis during pathological inflammation[11]. We therefore reasoned that the ability of the commensal consortia that we had designed to effectively engage TLR signaling would not only enhance antibacterial defenses in the intestine but could also be used to combat intestinal damage due to pathological inflammation. To test this, microbiota-depleted mice were treated with dextran sulfate sodium (DSS) to induce intestinal inflammation and then orally administered FITC-dextran 4000 to assess the integrity of the mucosal barrier. Consistent with previous work, we found FITC-dextran 4000 in the circulation of DSS-treated mice indicating that inflammation disrupted mucosal barrier function, and we found that circulating levels of FITC-dextran 4000 decreased after administration of the “high immune-stimulating” consortium but not the “low immune-stimulating” consortium (Fig 4L). Similarly, we found that weight loss induced by DSS treatment was reduced after administration of the “high immune-stimulating” but not the “low immune-stimulating” consortium (Fig 4M). Next, we wanted to examine the mechanistic basis for how our “high immune-stimulating” consortium restores intestinal barrier function. First, we wanted to examine whether protection occurred through the stimulation of TLRs. Using antibody neutralization, we found that blockade of TLR signaling eliminated the protective effect of our “high immune-stimulating” consortium on intestinal homeostasis (Fig 4N–$O). We then wanted to understand what signals are translating TLR stimulation into intestinal homeostasis. As IL6 plays a fundamental role in intestinal homeostasis we neutralized this pathway to test its role in mediating the effect of our “high immune-stimulating” consortium on intestinal homeostasis. We found that the beneficial effects of our consortium were lost after inhibition of IL6 (Fig 4N–4O) demonstrating that this pathway plays an integral role in mucosal homeostasis downstream of our “high immune-stimulating” consortium. To examine the role of macrophages, we depleted these cells using clodronate treatment prior to commensal consortia inoculation. After clodronate treatment, our “high immune-stimulating” consortium could no longer restore mucosal barrier function and prevent weight loss, indicating that macrophages are required for the restorative effect of the “high immune-stimulating” consortium (Fig 4P–4Q). Finally, we wanted to test whether analogously to the protective effect of our “high immune-stimulating” consortium on intestinal defenses, we wanted to know whether canonical NF-κB signaling played a role in restoring mucosal barrier function during intestinal inflammation. In mice given the “high immune-stimulating” consortium, we found that inhibition of canonical NF-κB prevented the “high immune-stimulating” consortium restoring mucosal barrier function and protecting against weight loss (Fig 4R–4S). Collectively, our study shows that we can rationally design commensal groups based on their immunological profiles to combat infectious and inflammatory diseases associated with microbiota dysbiosis through manipulation of intestinal innate immunity.

Ethics statement

The use of mice was performed under the authority of the UK Home Office outlined in the Animals [Scientific Procedures] Act 1986 after ethical review by Imperial College London Animal Welfare and Ethical Review Body (PPL 70/7969 and PF93C158E).

Bacterial strains

Bacteria used in this study and their growth conditions are outlined in S1 Table.

Macrophage culture, stimulation and inhibitor treatment

The murine macrophage cell line J774A.1 was cultured in DMEM supplemented with 10% v/v FBS, penicillin (100 units/mL) and streptomycin (100 μg/mL) and maintained at 37°C and 5% (v/v) CO₂. The human monocyte cell line was maintained in RPMI 1640 with 1×GlutaMAX, 10% FBS, penicillin (100 units/mL) and streptomycin (100 μg/mL), 1 mM sodium pyruvate and 10 mM HEPES. Cells were passaged via gentle scraping. THP-1 monocytes were differentiated to macrophages by treatment with PMA (100 ng/mL) for 48 hours. For bone marrow derived macrophages, C57BL/6 mice were humanly sacrificed using a Schedule 1 method. All fur and skin were removed from the legs, the muscle cut away until the bone was completely exposed. To remove the leg, a cut was made above the leg joint without removing the top of the femur. The bones were cleaned of any remaining tissue and placed in a petri dish containing PBS. Bones were transferred into a petri dish containing 70% ethanol for 1 minute then returned to the PBS. To remove the femur, the hip joint was cut off whilst holding the bone with sterile tweezers then similarly the knee joint was cut. To remove the tibia, the knee joint was removed, then the tibia was cut just below where the red bone marrow ends. The bone marrow was flushed into a 50 mL falcon tube by inserting a 26-gauge needle attached to a 20 mL syringe filled with PBS. The marrow was dispersed by aspirating it into an empty 20 mL syringe through a 19-gauge needle twice, then the cells were centrifuged at 1,500 rpm for 5 minutes at 4°C. Cells were then re-suspended in 15 mL of R10 medium containing 15% L-cell conditioned media (LCM). The cell suspension was used to seed flasks containing 20 mL R10 media + 15% LCM which were incubated at 37°C and 5% CO₂. For macrophage stimulation, J774A.1 macrophages or BMDMs were removed from culture flasks by gentle scrapping and seeded at a density of 2.5×10⁴ cells/well in a 96 well plate in 100 μL of growth media and incubated overnight at 37°C and 5% CO₂. Supernatant was removed and 100 μL of fresh growth media added. Macrophages were challenged with heat-killed bacteria at an MOI of 1:10 and incubated overnight in the same conditions. The supernatant was then removed, and cytokine production measured by ELISA. 100 μL of fresh growth media was added to the cells with inhibitors, as described in figure legends. The cells were incubated at 37°C and 5% CO₂ for the pre-incubation time stated then heat-killed (65°C for 30 mins) bacteria were added at MOIs of 1:10. Cells were incubated overnight at 37°C and 5% CO₂ then the supernatant was removed, and the cytokine production measured by ELISA. For treatment with inhibitors, macrophages were incubated with indicated inhibitors for 1 hour prior to commensal stimulation. Inhibitors used were: U0126 (ERK 1/2 inhibitor (Cell Signaling Technologies)); SP600125 (JNK inhibitor (Fluorochem)); SB203580 (P38 inhibitor (InvivoGen); PF 06650833 (IRAK4 inhibitor (TOCRIS); BOT-64 (IKK-β inhibitor (Sigma)); GO 6983 (PKC isoforms α, β, γ, δ, ζ and μ inhibitor (Abcam)); Piceatannol (Syk inhibitor (Sigma)); and SR11302 (AP-1 inhibitor (TOCRIS)) at indicated concentrations. All inhibitors were tested to ensure they were not toxic at concentrations used. Polymyxin B (Tokyo Chemical Industries) was used at indicated concentrations. siRNA oligonucleotides against IKK-β (Ambion), JNK1 and JNK2 (Ambion), and cognate control siRNAs were used to transfect J774A.1 cells using Lipofectamine RNAiMAX transfection reagent according to the manufacturer's instructions. Cells were used 24 hr post-transfection. Cell death was determined by measuring Lactate Dehydrogenase release using the CyQUANT LDH Cytotoxicity Assay kit according to the manufacturer’s instructions (Invitrogen).

HEK cell culture and stimulation

HEK Blue cells expressing human TLR2 (catalog number hkb-htlr2), TLR4 (catalog number hkb-htlr4) and TLR5 (catalog number hkb-htlr5) (InvivoGen), and control HEK Blue Null cells (catalog number hkb-null) (InvivoGen) were cultured at 37°C in 5% CO₂ in Dulbecco’s modified Eagle's media supplemented with 10% (v/v) fetal bovine serum (Invitrogen), penicillin (100 units/mL), streptomycin (100 mg/mL), blastocidin (30 mg/mL), and zeocin (100 mg/mL). Cells were then seeded at 2.5×10⁴ cells/well in a 96 well plate in growth media and allowed to adhere overnight. Supernatant was removed and 100 μL of fresh growth media was added. HEK-blue cells were challenged with heat-killed bacteria at a multiplicity of infection of 1:10. HEK-blue NULL2 cells being utilized as a negative control for receptor stimulation. Cells were then incubated at 37°C and 5% CO₂ for 24 hours. 50 μL of supernatant was removed and heat-inactivated at 65°C for 5 minutes then added to 150 μL of Quanti-Blue detection media (InvivoGen). Alkaline phosphatase activity was measured on a spectrophotometer at 620 nm (OD620).

Cytokine quantification

Cytokines levels were determined using ELISA kits for TNF, IL6, GM-CSF (all Biolegend), CXCL2, CXCL1, IL-12B, LCN2 (all Thermo Fisher Scientific) according to the manufacturer’s instructions.

Mice, microbiota and cell depletion

Wild-type C57BL/6 mice were purchased from Envigo (UK) and CB17 SCID mice were from Charles River (UK), germ-free Swiss Webster mice were purchased from Taconic. All adult mice were female and normally between 6 − 14 weeks old, except germ-free mice which were male and female. Adult mice were housed no more than five per cage with Aspen chip 2 bedding with a 12 hour light and 12 hour dark cycle at 20°C − 22°C. Mice were randomly assigned to experimental groups, water was provided ad libitum and mice were fed RM1 (Special Diet Services). To deplete the microbiota, mice were given broad-spectrum antibiotics (metronidazole 1 g/L, neomycin sulfate 1 g/L, ampicillin 1 g/L, and vancomycin 0.5 g/L) in drinking water for 7–10 days[13,80]. Macrophages were depleted using clodronate-containing liposomes (FormuMax Scientific) as described previously [13,44] confirmation of depletion shown in (S12 Fig). Neutrophils were depleted using the anti-mouse Ly-6G antibody (1A8) (Biolegend) and confirmed by measuring intestinal MPO levels (S13 Fig).

Models of bacterial colonization in the intestine

To establish intestinal colonization by VRE unanaesthetized mice were orally inoculated with approximately 5×10⁵ CFU of VRE in 200 μL of PBS. For consortia inoculation mice were orally inoculated with 5×10⁸ CFU of the indicated commensal consortia two days prior to VRE inoculation. Macrophages were depleted using clodronate liposomes as described in[44]. Engraftment of consortia was measured by qPCR of 16s rRNA gene as described in[13,44] using the following primers: 16s forward 5’-ACTCCTACGGGAGGCAGCAGT-3’; 16s reverse 5’-ATTACCGCGGCTGCTGGC-3’; Firmicutes forward 5’-GGAGYATGTGGTTTAATTCGAAGCA-3’; Firmicutes reverse 5’-AGCTGACGACAACCATGCAC-3’; Bacteroidetes forward 5’-GGARCATGTGGTTTAATTCGATGAT-3’; Bacteroidetes reverse 5’-AGCTGACGACAACCATGCAG-3’. B. dorei forward 5’-GGAAACGGTTCAGCTAGCAATA-3’; B. dorei reverse 5’-AGTCTTGTCAGAGTCCTCAGCATC-3’. B. ovatus forward 5’-CCGGATAGCATACGAATATC-3’; B. ovatus reverse 5’-CGTAGGAGTTTGGACCGTGT-3’. L. johnsonii forward 5’-GATTTTAGTGCTTGCACTAAA-3’; L. johnsonii reverse 5’-ATTACCTTACCAACTAG-3’. C. symbiosum forward 5’-GGGAACTCTTGGGAGACTGC-3’; C. symbiosum reverse 5’-TGACTCCCATGGTGTGACG-3’. L. rhamanosus forward 5’-GTGCTTGCATCTTGATTTAATTTT-3’; L. rhamanosus reverse 5’-TGCGGTTCTTGGATTTATGCG-3’. C. histolyticum forward 5’-AGAAGCAAGACCGTGAGGTG-3’; C. histolyticum reverse 5’-CGGACTTCGGGTGTTACCAA-3’. P. anaerobius forward 5’-CGATGAGTACTAGGTGTCGGG-3’; P. anaerobius reverse 5’-AGCCCCGAAGGGAAGGTGTG-3’. B. ovatus forward 5’-TCTTCCGCATGGTAGAACTATTA-3’; B. ovatus reverse 5’-ACCGTGTCTCAGTTCCAATGTG-3’. B. cellulosilyticus forward 5’-AGGATGACTGCCCTATGGGT-3’; B. cellulosilyticus reverse 5’-CTGCACTCAAGACGACCAGT-3’. In mice administered our “high immune-stimulating” and “low immune-stimulating” consortia fecal levels of viable Lactobacilli were determined by culture on Lactobacillus-selective MRS media, Bacteroides by culture on Bacteroides-selective Bacteroides Bile Esculin (BBE) media, and Clostridia by culture on Modified Reinforced Clostridial media (MRCA) (with Bacteroidetes level subtracted as MRCA can support growth of Bacteroidetes as well as Clostridia). P. anaerobius levels were determined by culture on TSA media +/- sodium polyanethol sulfonate at 0.05%. Sensitivity to sodium polyanethol sulfonate is used to identify P. anaerobius[81] and the other members of the consortium P. anaerobius is in are insensitive to it.

Signaling pathway neutralization in vivo

For antibody neutralization, mice were administered via the intraperitoneal injection αTLR2 (InvivoGen), αTLR4 (eBiosciences), αTLR5 (InvivoGen), or αIL6 (eBiosciences) neutralizing antibodies (100 μg/mouse) or pertinent isotype control, as indicated, three days prior to, and concomitant with, consortia inoculation. For pharmacological inhibition of canonical NF-κB, mice were administered BOT-64 (30 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation. For pharmacological inhibition of P38, mice were administered SB203580 (2 mg/kg/dose) via intraperitoneal injection concomitant with consortia inoculation.

Intestinal inflammation and permeability assay

For the induction of intestinal inflammation, antibiotic treated mice received 2% (w/v) DSS in drinking water for 7 days[11]. Indicated groups of mice were orally administered commensal consortia on day 2, day 4 and day 6 of DSS treatment. Intestinal barrier function was examined by measuring FITC-dextran 4000 levels in blood. Mice were orally administered 200 mL of FITC-dextran 4000 (80 mg/mL) one day after cessation of DSS treatment and fluorescence intensity in the blood was measured four hours after FITC-dextran inoculation (excitation: 495 nm and emission:519 nm).

Phylogenetic tree construction

The phylogenetic tree was constructed using the NCBI’s Taxonomy database to create a common tree which was then visualized using FigTree v1.4.4.

Statistical analysis

Statistical comparisons were performed using GraphPad Prism software. To compare differences between two groups, the Student’s t-test or the Mann–Whitney test were used. For all other comparisons a one-way analysis of variance (with post-hoc Sidak’s test or Dunnett’s test), a Kruskal–Wallis test with Dunn’s multiple comparison test, were used as appropriate. Correlations were determined using either the Pearson’s or the Spearman’s correlation test with FDR used to correct for multiple comparisons[82]. P-values <0.05 were considered significant.