Differential Induction of Antimicrobial REGIII by the Intestinal Microbiota and Bifidobacterium breve NCC2950

Abstract

The intestinal microbiota is a key determinant of gut homeostasis, which is achieved, in part, through regulation of antimicrobial peptide secretion. The aim of this study was to determine the efficiency by which members of the intestinal microbiota induce the antimicrobial peptide REGIII and to elucidate the underlying pathways. We showed that germfree mice have low levels of REGIII-γ in their ileum and colon compared to mice with different intestinal microbiota backgrounds. Colonization with a microbiota of low diversity (altered Schaedler flora) did not induce the expression of REGIII-γ as effectively as a complex community (specific pathogen free). Monocolonization with the probiotic Bifidobacterium breve, but not with the nonprobiotic commensal Escherichia coli JM83, upregulated REGIII-γ expression. Induction of REGIII-γ by B. breve was abrogated in mice lacking MyD88 and Ticam1 signaling. Both live and heat-inactivated B. breve but not spent culture medium from B. breve induced the expression of REGIII-α, the human ortholog and homolog of REGIII-γ, in human colonic epithelial cells (Caco-2). Taken together, the results suggest that REGIII-γ expression in the intestine correlates with the richness of microbiota composition. Also, specific bacteria such as Bifidobacterium breve NCC2950 effectively induce REGIII production in the intestine via the MyD88-Ticam1 pathway. Treatment with this probiotic may enhance the mucosal barrier and protect the host from infection and inflammation.

INTRODUCTION

Antimicrobial peptides secreted by intestinal immune and epithelial cells are important effectors of innate immunity. These endogenous peptides are induced during exposure to enteric pathogens in an attempt to protect the host from infection (1). It is increasingly apparent that antimicrobial peptides also play an essential role in the maintenance of intestinal homeostasis by limiting microbial-epithelium interactions and preventing unnecessary microbe-driven inflammation (2). This is particularly important in the distal gut where microbiota load and density are high.

The intestinal microbiota consists of a complex community of bacteria with various physiological and immune-modulating capacities (3). A balanced composition of symbionts and pathobionts is thought to stimulate homeostatic responses in the host (3), while shifts in this balance (dysbiosis) have been associated with inflammatory disorders of the gut, such as inflammatory bowel disease (IBD) (4). Recently, it has been shown that the intestinal microbiota provides pivotal stimuli and cues necessary for the induction of antimicrobial peptides (5).

Regenerating islet-derived III (REGIII) proteins, which belong to the family of C-type lectins, are one class of antimicrobials that are expressed in the intestine. In mice, three distinct classes of RegIII, RegIII-α, -β, and -γ, have been identified. In contrast, only REGIII-α and -γ have been identified in humans. Human REGIII-α, also known as hepatocarcinoma-intestine-pancreas/pancreatic-associated protein (HIP/PAP), shares 67% homology with murine RegIII-γ, while human REGIII-γ shares 68% homology with murine RegIII-β. REGIII proteins bind to the peptidoglycan moieties of bacteria and induce damage to the bacterial cell wall (6–8). Different intestinal cell types express REGIII proteins (6, 9). We have previously shown that modulating intestinal microbiota composition, either by colonizing with microbiota devoid of pathobionts or supplementation with a probiotic bacteria, affects the expression of RegIII-γ in mice lacking intracellular microbial recognition receptors (10). However, it is unclear whether a specific component(s) of the intestinal microbiota differentially and directly regulates the expression of REGIII proteins by various intestinal cell types.

Here, we sought to determine whether different components of the intestinal microbiota and specific probiotics differ in their capacities to stimulate the expression of antimicrobial peptide REGIII by intestinal epithelial cells (IECs). For this, we investigated the effects of colonization with both diverse communities and specific components of the microbiota on REGIII-γ expression by ileal and colonic epithelial cells. REGIII-γ levels were quantified in whole ileal and colonic tissue of germfree (GF) mice and in mice colonized with specific-pathogen-free (SPF) microbiota, altered Schaedler flora (ASF), commensal Escherichia coli JM83, or the probiotic Bifidobacterium breve NCC2950. The importance of Toll-like receptor (TLR) signaling in this process was also investigated using GF MyD88^−/−; Ticam1^−/− mice. Furthermore, REGIII-α expression in human colonic epithelial cells was also quantified after stimulation with E. coli or B. breve.

MATERIALS AND METHODS

Mice.

MyD88^−/−; Ticam1^−/− mice on a C57BL/6 background were kindly provided by B. A. Beutler (La Jolla, CA, USA). SPF C57BL/6 mice were purchased from Taconic. Germfree C57BL/6 and MyD88^−/−; Ticam1^−/− mice were rederived at the McMaster University Axenic Gnotobiotic Unit (AGU) by an axenic two-cell embryo transfer technique previously described (11). Germfree mouse colonies were maintained in flexible film isolators at the AGU, and germfree status was routinely confirmed by a combination of culture- and non-culture-based techniques in fecal and cecal contents (11). Periodic serological testing was also performed for viruses, parasites, and known pathogens (Charles Rivers Laboratories). All mice had unlimited access to autoclaved food and water. Both male and female mice were used at the age of 8 to 12 weeks. All experiments were carried out in accordance with the McMaster University animal utilization protocols.

Bacterial strains.

Bifidobacterium breve (B. breve) NCC2950 was obtained from the Nestlé Culture Collection (Lausanne, Switzerland) and grown under anaerobic conditions in deMan-Rogosa-Sharpe (MRS) medium supplemented with 0.05% (vol/vol) l-cysteine hydrochloride. Escherichia coli JM83 (American Type Culture Collection [ATCC]) was grown under aerobic conditions in Luria broth (LB; Oxoid) medium. After 18 h at 37°C, bacterial cells were pelleted, washed in phosphate-buffered saline (PBS), resuspended at a concentration of 10¹⁰ CFU/ml in PBS with 20% (vol/vol) glycerol, and kept in frozen aliquots until used. Growth medium was spun down and sterile filtered to generate the spent culture medium (SCM). The heat-inactivated (HI) formulation was prepared by incubating 50-ml aliquots of B. breve cells at 90°C for 1 h.

Colonization and monocolonization experiments.

Monocolonization of GF mice was performed in dedicated flexible film gnotobiotic isolators as previously described (11). Briefly, Bifidobacterium breve or Escherichia coli JM83 cultures were prepared and imported into the isolator following strict aseptic procedures. A total of 10⁹ CFU of bacteria of each preparation was introduced by gavage into the stomach of each mouse. Three weeks after monocolonization, mice were used for experiments. Monocolonization was confirmed at sacrifice by plating both the fecal and cecal contents in deMan-Rogosa-Sharpe (MRS; Oxoid) agar supplemented with 0.05% (vol/vol) l-cysteine hydrochloride (Sigma-Aldrich) and mupirocin (Sigma-Aldrich) under anaerobic conditions for B. breve and in LB agar under aerobic conditions for E. coli. ASF consisted of Lactobacillus acidophilus, Lactobacillus salivarius, Bacteroides distasonis, a spiral-shaped bacterium, and four fusiform, extremely oxygen sensitive bacteria (12). ASF colonization was achieved by cohousing germfree mice with an ASF-colonizer from McMaster's standard colony and was confirmed by a combination of culture and molecular techniques (10).

Colonic epithelial cell line assays.

The Caco-2 cell line was obtained from the ATCC and used from passages 19 to 25. Cell monolayers were maintained in cell medium consisting of Dulbecco's modified Eagle medium, 20% heat-inactivated fetal bovine serum (FBS; PAA), and 1% minimal essential medium (MEM) and nonessential amino acids (Gibco) supplemented with 2 mM glutamine, and cells were cultured in a humidified atmosphere with 5% CO₂. For stimulation experiments, cells were seeded in 24-well tissue culture plates and used at 70 to 80% confluence. Cells were stimulated with live or HI B. breve or live E. coli bacteria at cell-to-bacterium ratios of 1:10 and 1:100 or with SCM (10%, vol/vol). PBS-glycerol (10%, vol/vol) was used as a negative control while interleukin-22 (IL-22; 10 ng/ml) (R&D Systems) was a positive control. All stimulations were performed in FBS-free medium with 1% penicillin-streptomycin (Pen-Strep) for 4 to 24 h. After stimulation, cells were washed twice with PBS containing Pen-Strep and 50 μg/ml gentamicin and stored at −80°C until used.

RNA isolation and RT-qPCR.

Total RNA from the ileum and colon of mice or cell lines was isolated using an RNeasy Mini Kit (Qiagen). Potential DNA contamination was removed by column DNase treatment (Qiagen). RNA quantity and integrity were checked with a NanoDrop instrument (Thermo Scientific) and agarose gel electrophoresis. Only samples with intact RNA were used for subsequent cDNA synthesis with iScript reverse transcriptase (Bio-Rad). A total of 500 μg of input RNA was used for each sample. Quantitative real-time PCR (RT-qPCR) was performed on an iQ5 Real-Time Detection System (Bio-Rad) with SSofast Evagreen Supermix (Bio-Rad). Primers used were as follows: murine RegIII-γ fwd, 5′-CGTGCCTATGGCTCCTATTGCT-3′; murine RegIII-γ rev, 5′-TTCAGCGCCACTGAGCACAGAC-3′; human REGIII-α fwd, 5′-TATGGCTCCCACTGCTATGCCT-3′; human REGIII-α rev, 5′-TCTTCACCAGGGAGGACACGAA-3′; GAPDH fwd 5′-CCATGGAGAAGGCTGGGG-3′; GAPDH rev 5′-CAAAGTTGTCATGGATGACC-3′. The iQ5 manager software (Bio-Rad) was used to calculate the relative fold change in expression normalized to GAPDH expression by the 2^−ΔΔCT (where C_T is threshold cycle) method. All procedures were performed according to the manufacturer's instructions.

IF for REGIII-γ proteins.

REGIII-γ protein expression in the mouse intestine was evaluated using immunofluorescence (IF) in formalin-fixed, paraffin-embedded tissue sections. Sections were cut (5 μm), deparaffinized, and then blocked with phosphate-buffered saline (PBS)-bovine serum albumin in 2% Tween 20 at 0.05% for 1 h. Samples were stained overnight (4°C) with rabbit anti-mouse REGIII-γ antibodies (1:100 dilution) (13) and for 1 h with secondary antibody at room temperature [goat anti-rabbit IgG(H+L); 1:250 dilution] (Molecular Probes) and mounted in Prolong Gold with 4′,6′-diamidino-2-phenylindole (DAPI). For quantification, specific fluorescence intensity from three different microscopic fields per animal was acquired using a Nikon Eclipse 90i instrument. All individual fields have been normalized by using the same fluorescence acquisition settings and by tissue autofluorescence. Specific REGIII-γ fluorescence staining was quantified using ImageJ software (NIH) and was reported per unit surface of tissue. Data were represented as fold increase of signal intensity compared to the control group (germfree C57BL/6 mice), which was arbitrarily reported as 1.

Statistics.

Data are presented as either dot plots or bar graphs (means ± standard deviations [SD]). Statistical analysis was performed by using one-way analysis of variance (ANOVA), followed by the Bonferroni test, or by using a two-tailed Student t test, when applicable. All statistical testing was performed using GraphPad Prism, version 6 (GraphPad software Inc.). A P value of <0.05 was considered significant.

RESULTS

Germfree mice had lower REGIII-γ expression levels than SPF-colonized mice.

Different regions of the gastrointestinal tract have unique microenvironments characterized by specialized cell types, microbial diversity, and load. Thus, we first sought to examine ileal and colonic expression of REGIII-γ in the absence of microbiota and to determine whether microbiota with different levels of diversity had an impact on its expression. We measured RegIII-γ in the ileum and colon of GF- and ASF-colonized and SPF C57BL/6 mice by quantitative real-time PCR. GF mice showed significantly lower RegIII-γ expression than SPF mice (Fig. 1). REGIII-γ protein expression was also examined by immunofluorescence (IF) staining. Consistent with RNA results, REGIII-γ protein expression was significantly lower in GF mice than in SPF mice, and the pattern of expression was different in the ileum from that in the colon (Fig. 2). In the ileum of SPF mice, REGIII-γ was expressed by IECs located both at the crypt and the villi, with the highest fluorescence often localized within the crypt. On the other hand, colonic expression of REGIII-γ in SPF mice was specifically expressed by colonocytes. Overall, our results show that REGIII-γ expression varies between the ileum and colon and is influenced by the type of microbiota it is colonized with.

Fig 1.

RegIII-γ RNA expression in ileum and colon of GF, ASF-colonized, and SPF mice. SPF mice had higher RegIII-γ RNA expression levels than ASF-colonized mice. Total RNA from ileum and colon of 8- to 12-week-old mice was extracted, and RegIII-γ expression was measured by RT-qPCR. RNA expression was normalized to Gapdh, and results are shown relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values are calculated using ANOVA, followed by a Bonferroni post hoc test.

Fig 2.

REGIII-γ protein expression in ileum and colon of GF, ASF-colonized, and SPF mice. SPF mice had higher REGIII-γ protein expression in ileum and colon than GF mice. (A) Representative immunofluorescence staining (anti-REGIII-γ; red) in each group. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; blue). (B) Quantification of REGIII-γ staining intensity was expressed as fold increase relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values were calculated using ANOVA, followed by a Bonferroni post hoc test.

Monocolonization with B. breve but not E. coli increased REGIII-γ expression.

There are important interindividual differences in gut microbial communities (14, 15), and this may have functional implications. We investigated this using a model of gnotobiotic monocolonization. The effects of Bifidobacterium breve NCC2950, previously found to restore REGIII in mice lacking Nod signaling (10), and of Escherichia coli JM83 on REGIII expression in the intestine were explored. Three weeks after monocolonization, mice were aseptically exported from flexible-film isolators, and tissue was collected to measure REGIII-γ expression. Successful monocolonization was confirmed by selective plating of both fecal and cecal contents collected at 21 days postcolonization and by fluorescence in situ hybridization (FISH) staining (see Fig. S1 in the supplemental material). B. breve- and E. coli-monocolonized mice displayed similar bacterial loads in stool and cecal content (Fig. 3). However, only B. breve-monocolonized mice exhibited higher expression of ileal and colonic REGIII-γ than GF controls, suggesting that induction of REGIII-γ was strain specific (Fig. 4 and 5).

Fig 3.

Bacterial load in the feces and cecum of B. breve- and E. coli-monocolonized mice. GF mice were monocolonized with either B. breve or E. coli. At 21 days postmonocolonization, the density of bacteria in the stool and cecal content was determined by selective plating. Data are pooled from two to three independent experiments.

Fig 4.

RegIII-γ RNA expression in ileum and colon of GF and B. breve- and E. coli-monocolonized mice. Monocolonization with B. breve induced higher RegIII-γ RNA expression than monocolonization with E. coli. At 21 days postmonocolonization, total RNA from ileum and colon was extracted, and RegIII-γ expression was measured by RT-qPCR. RNA expression was normalized to Gapdh, and results are shown relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values are calculated using ANOVA, followed by a Bonferroni post hoc test.

Fig 5.

REGIII-γ protein expression in ileum and colon of GF and B. breve- and E. coli-monocolonized mice. B. breve-monocolonized mice showed higher REGIII-γ protein expression than E. coli-monocolonized mice. (A) Representative immunofluorescence staining (anti-REGIII-γ; red) in each group. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; blue). (B) Quantification of REGIII-γ staining intensity was expressed as fold increase relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values are indicated on the figure and were calculated using ANOVA, followed by a Bonferroni post hoc test.

Monocolonization with B. breve did not increase RegIII-γ expression in MyD88^−/−; Ticam1^−/− mice.

To determine the host factors that govern the B. breve-induced REGIII-γ expression, we used mice that lack myeloid differentiation primary response 88 (Myd88) and TIR-containing adaptor molecule (Ticam1) genes, which are relevant downstream signaling adaptor proteins for all TLRs. Similar to wild-type GF C57BL/6 mice, GF MyD88^−/−; Ticam1^−/− mice displayed low levels of ileal and colonic REGIII-γ expression. Levels of REGIII-γ expression in MyD88^−/−; Ticam1^−/− mice, however, remained unchanged after B. breve monocolonization (Fig. 6 and 7). Together, these results suggest that microbial induction of REGIII-γ is mediated by MyD88-Ticam-dependent pathways, such as TLR signaling.

Fig 6.

RegIII-γ RNA expression in ileum and colon of Myd88^−/−; Ticam1^−/− mice monocolonized with B. breve. B. breve-monocolonized Myd88^−/−; Ticam1^−/− mice displayed RegIII-γ RNA expression levels similar to those of GF Myd88^−/−; Ticam1^−/− mice. At 21 days postmonocolonization, total RNA from ileum and colon was extracted, and RegIII-γ expression was measured by RT-qPCR. RNA expression was normalized to Gapdh, and results are shown relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values are calculated using ANOVA, followed by a Bonferroni post hoc test.

Fig 7.

REGIII-γ protein expression in ileum and colon of Myd88^−/−; Ticam1^−/− mice monocolonized with B. breve. B. breve-monocolonized Myd88^−/−; Ticam1^−/− mice displayed levels of RegIII-γ protein expression similar to those of GF Myd88^−/−; Ticam1^−/− mice. (A) Representative immunofluorescence staining (anti-REGIII-γ; red) in each group. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; blue). (B) Quantification of REGIII-γ staining intensity was expressed as fold increase relative to the mean of GF mice, which is set to 1. Each data point represents an individual mouse. Data were pooled from two to three independent experiments. P values are calculated using ANOVA, followed by a Bonferoni post hoc test.

Live B. breve induced the expression of RegIII-α in human colonic epithelial cells.

To quantify expression of REGIII-α, the human ortholog and homolog of murine RegIII-γ, we performed in vitro experiments using Caco-2 and HT-29 human colon epithelial cell lines. To optimize the expression of REGIII-α induced by bacteria, cells were incubated for 2, 4, 6, and 24 h at cell-to-bacterium ratios of 1:10 and 1:100. Results show that 4 h of incubation at a cell-to-bacterium ratio of 1:100 led to the highest REGIII-α expression level (see Fig. S2 in the supplemental material). Using these parameters, the amount of REGIII-α expression was measured, and our results showed that colonic epithelial cells significantly upregulated expression in response to B. breve stimulation but not to stimulation with E. coli (Fig. 8).

Fig 8.

REGIII-α RNA expression in Caco-2 cells stimulated with IL-22 (positive control), B. breve, or E. coli. Both IL-22 and B. breve induced RegIII-α RNA expression in Caco-2 cells. Total RNA from stimulated cells was extracted, and REGIII-α expression was measured by RT-qPCR. RNA expression was normalized to GAPDH, and results are shown relative to the mean of PBS-glycerol-stimulated cells, which is set to 1. Data were pooled from three to four independent experiments with two to four wells per group (means ± SD). P values are calculated using ANOVA, followed by a Bonferroni post hoc test.

We then asked whether B. breve actively secretes metabolites able to upregulate REGIII-α expression from IECs. For this, Caco-2 cells were stimulated with live, HI, and SCM formulations of B. breve. Our results showed that only live and HI preparations induced higher REGIII-α expression levels than the PBS-glycerol control, suggesting that the whole B. breve cell is required for the induction, independently of its metabolic activity (Fig. 9).

Fig 9.

REGIII-α RNA expression in Caco-2 cells stimulated with live and heat-inactivated (HI) B. breve and B. breve spent culture medium (SCM). Both live and HI B. breve (live or HI) induced REGIII-α RNA expression in Caco-2 cells. Total RNA from stimulated cells was extracted, and REGIII-α expression was measured by RT-qPCR. RNA expression was normalized to GAPDH expression, and results are shown relative to the mean of PBS-glycerol-stimulated cells, which is set to 1. Data were pooled from three to four independent experiments, with two to four wells per group (means ± SD). P values are calculated using ANOVA, followed by Bonferroni post hoc test.

DISCUSSION

Antimicrobial REGIII proteins play an important role in maintaining gut homeostasis through spatially segregating bacteria, preventing potentially harmful immune responses, and protecting the host from infection (16–18). In parallel with previous findings (19–22), this study shows that the intestinal microbiota affects the level of REGIII expression in the intestine. However, the level of expression differs depending on the region of the gut examined and the nature of microbiota to which it is exposed. We demonstrated that the probiotic Bifidobacterium breve NCC2950 but not the commensal Escherichia coli JM83 significantly induced REGIII expression in vivo in mice and in vitro in a human intestinal cell line, and this upregulation was independent from the metabolic activity of the strain and mediated through MyD88-Ticam1 signaling. Collectively, these results indicate that regulation of REGIII depends on the richness and specific components of the intestinal microbiota.

Colonization with the community of eight strains of bacteria that compose the ASF did not induce the same level of REGIII-γ expression as observed in SPF mice. Colonization with ASF has previously been shown to effectively reverse GF-related phenotypes (23); however, ASF-induced phenotypes are not always identical to those found in SPF mice (11, 24). This may be related to variability in SPF composition; the SPF mice used in this study came from Taconic and, unlike SPF mice from other suppliers, contain segmented filamentous bacteria (SFB), a potent inducer of T-helper 17 (Th17) cells (25). IL-22 produced by Th17 cells (26) induces the expression of REGIII-γ in both murine and human colonic epithelial cells (17, 27). In fact, monocolonization with SFB led to increased REGIII-γ production, comparable to that of ASF-colonized BALB/c mice (28). It has been shown that reducing microbiota diversity with broad-spectrum antibiotics decreases REGIII-γ expression (16, 22). Thus, in conjunction with previous work, our results support the notion that microbiota composition, especially with regard to the presence of pathobionts, is an important factor in REGIII-γ regulation.

We have previously shown that supplementation of B. breve upregulated REGIII-γ in Nod1^−/−; Nod2^−/− mice with lower baseline expression of this antimicrobial peptide (10), but the mechanisms underlying this specific probiotic stimulation were unclear. In this study, we monocolonized wild-type C57BL/6 mice with B. breve or E. coli and found that B. breve but not E. coli significantly increased REGIII-γ levels in the ileum and colon. Although ileal bacterial counts were not performed, fecal and cecal content showed similar bacterial loads of B. breve and E. coli, suggesting that the effects on REGIII-γ expression were not due to differential capacity of these two strains to colonize the gut. It should be noted that although B. breve stimulated REGIII-γ, the level of induction was lower than in SPF mice. In accordance with our findings, monocolonization with Bacteroides thetaiotaomicron, but not with noninvasive Listeria innocua, resulted in increased REGIII-γ expression in the small intestine that did not reach the level seen in SPF mice (6). Other investigators have also determined the effect of specific mono- or dicolonizations on REGIII-γ production in the colon and cecum, with variable results (28, 29). Overall, the data support the conclusion that the net effect of intestinal bacteria on REGIII-γ expression will be modulated by the presence of specific strains in the microbiota that include both commensals and potential pathobionts. One important aspect in colonization studies relates to variability in experimental design and the time point of tissue sampling. Dynamic changes in the microbiota load and diversity, as well as in immune responses, occur immediately after colonization (30). We chose to evaluate REGIII-γ at a steady state (21 days) postcolonization (31) since we were interested in defining REGIII responses under stable conditions. Indeed, REGIII levels have been shown to peak at 4 days postcolonization in the small intestine and to stabilize by day 16. Therefore, in addition to strain specificity, the time point chosen to determine colonization effects in antimicrobial peptides should be carefully defined.

Many cell types in the gastrointestinal tract are capable of producing REGIII proteins, including intestinal epithelial cells (IECs) and γδ intraepithelial lymphocytes (6, 32). Studies have proposed that IECs, particularly enterocytes, are producers of REGIII-γ in the colon (28, 33). In this study, we used immunofluorescence to investigate the main source of REGIII-γ after microbial exposure in vivo. We found that IECs are the cell types that predominantly express REGIII-γ in both the ileum and colon. Furthermore, our in vitro studies confirmed that incubation of human colonic epithelial cells with B. breve, but not E. coli, induced expression of REGIII-α, the human ortholog and homolog of REGIII-γ.

There is evidence to suggest REGIII-γ expression in the intestine is regulated by MyD88-mediated TLR signaling (18, 22, 34–36). B. breve has previously been shown to stimulate TLR2/MyD88 responses in CD103⁺ dendritic cells (DC). Thus, it is probable that B. breve potentially induces responses in epithelial cells through TLR2/MyD88 signaling as well (37). We found that REGIII-γ expression in MyD88-Ticam1 double knockout mice monocolonized with B. breve was low and comparable to that of GF mice, indicating that B. breve-induced REGIII-γ production requires TLR signaling. Direct signaling of B. breve through epithelial TLRs is consistent with the epithelial cell-autonomous model of REGIII-γ expression (38). The results may also explain our previous results in Nod1^−/−; Nod2^−/− mice in which B. breve led to normalization of REGIII-γ expression, likely through preserved TLR signaling in these mice. Recently, a new model of REGIII-γ production that involves IL-22 has been proposed in which luminal bacteria interact with TLR expressed by DC, leading to the release of cytokines which then prime innate lymphoid cells (ILC) to release the cytokine IL-22 (38). In our previous study, we did not detect an increase in IL-22 after B. breve-induced REGIII-γ expression (10). We propose that B. breve-induced REGIII-γ expression may occur in the absence of IL-22 supplementation through an epithelial cell-autonomous manner that involves the MyD88-Ticam1 pathway although, in this case, induction may be more moderate than in the presence of IL-22.

A number of studies have demonstrated that anti-inflammatory effects of probiotics can be elicited without live bacteria (4, 39, 40). Likewise, induction of antimicrobial human β-defensin 2 can be mediated by either live Escherichia coli Nissle 1917 or its structural flagellum protein (41). Here, we examined different probiotic preparations on REGIII-α expression by colonocytes. We found that live and heat-inactivated B. breve increased REGIII-α expression, whereas the spent culture medium did not induce any changes. These results suggest that a specific component of the structure of B. breve, and not its secreted metabolites, is responsible for REGIII-α epithelial expression.

In conclusion, we demonstrated that the effects of the microbiota on REGIII expression in the intestine correlate with microbial composition and that the effect is strain and formulation specific. We determined that the probiotic B. breve NCC2950 upregulates REGIII-γ expression through MyD88-Ticam1 signaling. We have previously shown that preventive administration of B. breve NCC2950 to genetically susceptible mice not only increased REGIII-γ expression but also ameliorated the severity of subsequent colitis (10). Based on these findings, we hypothesize that treatment with B. breve may regulate REGIII-γ production in a controlled manner that enhances barrier integrity and protects from inflammation. Our results support the use of microbiota-modulating strategies to target homeostatic regulation of antimicrobial peptides. This could be of benefit for IBD patients, their first-degree relatives, and patients undergoing chemotherapy or radiation therapy to prevent intestinal injury.