Paneth cells directly sense gut commensals and maintain homeostasis at the intestinal host-microbial interface

We previously performed a genome-wide expression analysis to assess bacterial regulation of Paneth cell antimicrobial responses (5). Our analysis revealed that bacteria orchestrate expression of a complex antimicrobial transcriptional program that includes RegIIIγ, an antimicrobial factor that binds peptidoglycan and is expressed in both Paneth cells and enterocytes (5) (Fig. 1D); RegIIIβ, which is closely related to RegIIIγ and binds peptidoglycan (H. L. Cash and L.V.H., unpublished data); CRP-ductin, which agglutinates Gram-positive and Gram-negative bacteria (7); the inflammatory modulator resistin-like molecule β (RELMβ) (8, 9); and several members of the α-defensin family of antimicrobial peptides.

To delineate the host factors that govern expression of this antimicrobial transcriptional program, we analyzed gene-targeted mice that are deficient in the sensing of conserved microbe-associated molecular patterns. At least two protein families recognize microbial products in mice and humans: the cytoplasmic NOD-like proteins and the membrane bound toll-like receptors (TLRs). NOD2 regulates Paneth cell expression of key α-defensins, including Defcr-rs10 (10). Examination of Nod2^−/− mice revealed that, unlike Defcr-rs10, small intestinal expression of RegIIIγ, RegIIIβ, CRP-ductin, and RELMβ was NOD2 independent (Fig. 1A). In contrast, mice lacking the TLR signaling adaptor MyD88 revealed that expression of these genes was MyD88 dependent (Fig. 1A), consistent with previous findings of MyD88-dependent RegIII expression in the small intestine (11, 12). We further established that MyD88-dependent expression of these transcripts occurred in Paneth cells recovered by laser capture microdissection (Fig. 1B). Treatment of wild-type conventional mice with a mixture of broad-spectrum antibiotics (13) showed that expression of the MyD88-dependent antimicrobial program was reversible upon partial microflora depletion (Fig. 1 C and D). Thus, intestinal microbes elicit a dynamic and reversible MyD88-dependent antimicrobial expression program in Paneth cells that is distinct from the NOD2-dependent antimicrobial response.

MyD88 is involved in signaling through TLRs, the IL-1 receptor (IL-1R), and the IL-18 receptor (IL-18R) (14). Paneth cells express several TLRs, including TLR2, 4, 5, and 9 (15–17). Although we were unable to identify a TLR-deficient mouse that phenocopied the profoundly reduced antimicrobial factor expression observed in MyD88^−/− mice, two lines of evidence suggest that TLRs, rather than IL-1R and IL-18R, direct expression of MyD88-dependent antimicrobial factors. First, we directly addressed the possibility that IL-1R or IL-18R might direct antimicrobial gene expression by analyzing Caspase-1^−/− mice, which are unable to produce active forms of IL-1β and IL-18 (18). We detected no statistically significant differences in RegIIIγ, RegIIIβ, CRP-ductin, and RELMβ transcript levels between conventional wild-type and Caspase-1^−/− mice [supporting information (SI) Fig. S1A], indicating that IL-1R and IL-18R do not direct MyD88-dependent antimicrobial responses. Second, oral administration of the TLR4 ligand lipopolysaccharide (LPS) to germ-free mice elicited dose-dependent expression of RegIIIγ, RegIIIβ, and RELMβ (Fig. S1B). Together, these findings suggest that TLRs, rather than IL-1R or IL-18R, direct expression of the MyD88-dependent antimicrobial response.

We next used this MyD88-dependent antimicrobial expression program to investigate whether Paneth cells directly detect bacteria in vivo or whether other mucosal cells detect and relay bacterial signals to Paneth cells. Prior studies have established a role for hematopoietic cell MyD88 in relaying bacterial signals to intestinal epithelial cells during intestinal injury (13, 19). However, adoptive transfer of wild-type bone marrow into lethally irradiated MyD88^−/− mice was insufficient to restore expression of the MyD88-dependent antimicrobial program (12) (Fig. S2), indicating that MyD88 signaling in bone marrow-derived cells does not govern epithelial antimicrobial factor expression. However, these findings were consistent with three possibilities. First, it was possible that nonhematopoietic, nonepithelial cells such as myofibroblasts (20) could detect and relay bacterial signals to Paneth cells. Second, enterocytes could capture and relay bacterial signals to Paneth cells. Third, Paneth cell-autonomous MyD88 activation could govern antimicrobial program expression.

To directly test whether Paneth cells sense bacteria through cell-autonomous mechanisms, we performed a gain-of-function genetic experiment. Here we used the Paneth cell-specific cryptdin-2 (CR2) promoter (21) to direct expression of a FLAG-tagged MyD88 transgene, and then crossed these transgenic mice to MyD88^−/− mice to yield a mouse model (CR2-MyD88 Tg) in which Paneth cells were the sole cell lineage expressing MyD88 (Fig. 2Aand Fig. S3). MyD88 expression in Paneth cells was sufficient to restore RegIIIγ expression in the MyD88^−/− background (Fig. 2 B and C). Although RegIIIγ is expressed in both Paneth cells and enterocytes in wild-type conventionally reared mice, its expression was restored specifically in Paneth cells (Fig. 2B). Quantitation of RegIIIγ, RegIIIβ, CRP-ductin, and RELMβ transcripts revealed that the broader MyD88-dependent antimicrobial transcriptional program was also restored through Paneth cell-specific MyD88 expression (Fig. 2C). Antimicrobial factor expression in the transgenic model was reversible upon antibiotic treatment (Fig. 2 B and C). This was not due to reversal of transgene expression, as the FLAG-tagged MyD88 was detected in Paneth cells from antibiotic-treated mice (Fig. 2B). Thus, activation of transgenic MyD88 requires bacteria and is not due to nonspecific autoactivation through overexpression of MyD88. Together, these findings show that Paneth cells directly detect intestinal bacteria via cell-autonomous MyD88 activation, revealing a direct dialogue between enteric bacteria and the Paneth cell lineage of the gut epithelium.

Our finding that Paneth cells directly sense bacteria through MyD88-dependent pathways suggested that intestinal epithelial MyD88 might play a key functional role in maintaining homeostasis with the commensal microbiota. To investigate this role, we further examined wild-type and MyD88^−/− mice. Conventionally reared wild-type and MyD88^−/− mice revealed no significant differences in luminal densities of commensal bacteria (Fig. 3A). However, bacteria that localize near the apical surface of epithelial cells are sampled by dendritic cells (DCs) and translocated alive to mesenteric lymph nodes (MLN) (22–24) and we reasoned that perhaps MyD88-dependent antimicrobial factors specifically modulate the numbers of bacteria that closely associate with the mucosal surface and are therefore available for uptake to MLN. Consistent with this idea, we recovered ≈10-fold more bacteria from the MLN of MyD88^−/− mice as compared with wild-type mice (Fig. 3A). The difference in MLN numbers was not due to MyD88 deficiency in bone marrow-derived cells, as transfer of MyD88^−/− bone marrow into wild-type recipients did not result in increased numbers of MLN bacteria relative to wild-type mice (Fig. 3A). In contrast, transfer of wild-type bone marrow into MyD88^−/− recipients resulted in numbers of MLN bacteria that were similar to MyD88^−/− mice, suggesting that epithelial MyD88 was necessary to limit bacterial translocation to MLN (Fig. 3A). To determine whether expression of MyD88 in the epithelium was sufficient to limit commensal penetration, we examined the CR2-MyD88 Tg mice (on the MyD88^−/− background). Transgenic expression of MyD88 in Paneth cells restored wild-type MLN numbers (Fig. 3A), indicating that expression of MyD88 in a single epithelial lineage is sufficient to limit commensal penetration to MLN. It is unlikely that these differences are due to differences in microflora composition among different strains, as we took care to cohouse mice (wt, MyD88^−/−, CR2-MyD88 Tg mice, and bone marrow chimeric mice) for 1 week before sacrifice.

To assess whether Paneth cell MyD88 could also limit pathogen penetration and dissemination, we orally inoculated conventionally reared wild-type, MyD88^−/−, and CR2-MyD88 Tg mice with Salmonella typhimurium. After 48 h luminal Salmonella colonization densities were ≈3 logs lower than those of the conventional flora (Fig. 3B), reflecting the fact that the commensal microbiota provides colonization resistance against Salmonella (25). Numbers of luminal Salmonella did not differ among wild-type, MyD88^−/−, and CR2-MyD88 Tg mice but increased numbers of Salmonella penetrated to MLN and spleen in MyD88^−/− as compared with wild-type mice. Paneth cell-specific expression of MyD88 restored MLN numbers to wild-type levels and significantly reduced the numbers of splenic bacteria (Fig. 3B). The decrease in spleen bacteria was not due to altered bacterial killing, as similar numbers of bacteria were recovered from the spleens of MyD88^−/− (8 × 10⁶ ± 3 × 10⁶ CFU) and CR2-MyD88 Tg mice (9 × 10⁶ ± 5 × 10⁶ CFU) following intraperitoneal injection of Salmonella. Thus, expression of MyD88 in Paneth cells is sufficient to limit Salmonella penetration across the mucosal barrier and systemic dissemination.

These findings indicated that expression of MyD88 in a single small-intestinal epithelial lineage was sufficient to limit penetration of commensals and pathogens across the epithelial barrier. Though Salmonella elicited Paneth cell antimicrobial responses when introduced into germ-free mice (Fig. S4A), the relatively low Salmonella numbers in conventional mice were insufficient to stimulate MyD88-dependent responses in excess of the basal levels elicited by the commensal microbiota (Fig. S4B). Thus, the protection afforded against Salmonella penetration and dissemination is likely the result, in part, of commensal stimulation of MyD88. This suggests that commensals protect their host against pathogen invasion not only through colonization resistance but also by activating MyD88-dependent signaling in the host epithelium.

These findings suggested that a key function of intestinal epithelial cells is to actively limit bacterial penetration of the mucosal barrier. To test this idea, we examined a transgenic mouse model in which Paneth cells are specifically ablated by expressing a diphtheria toxin fragment (tox176) under the control of the cryptdin-2 promoter (21). Although isolated CR2-tox176 crypts show reduced antimicrobial peptide secretion (3), the mice are overtly healthy, harbor a normal distribution of gut-associated lymphoid tissue, and exhibit no differences in the numbers of luminal bacteria (21) (Fig. 4A). Given our finding that Paneth cell-intrinsic MyD88 limits bacterial penetration of the mucosal barrier, we hypothesized that Paneth cell-deficient (CR2-tox176) mice might exhibit increased bacterial penetration of the mucosa. Though conventional CR2-tox176 mice and their wild-type littermates harbored similar luminal bacterial loads, we recovered higher numbers of bacteria from CR2-tox176 MLNs (Fig. 4A). Similarly, oral challenge of germ-free CR2-tox176 mice and their wild-type littermates with B. thetaiotaomicron resulted in ≈200-fold more MLN bacteria in CR2-tox176 mice (Fig. 4B). This was not due to higher luminal bacterial loads, as luminal numbers were similar between the two groups (Fig. 4B). No bacteria were recovered from the spleens of either CR2-tox176 or wild-type mice. Thus, Paneth cells are essential for limiting translocation of commensals across the intestinal barrier, but have little impact on overall small intestinal bacterial load.

To determine whether Paneth cells also play an essential role in limiting pathogen penetration, we challenged germ-free CR2-tox176 mice orally with S. typhimurium. Though luminal colonization levels were similar 48 h later, we recovered ≈7-fold more Salmonella from CR2-tox176 as compared with wild-type MLN. Moreover, 90-fold more Salmonella were recovered from the spleens of CR2-tox176 as compared with wild-type mice (Fig. 4C), revealing an essential role for Paneth cells in limiting pathogen translocation and dissemination.

Conventionally reared wild-type and MyD88^−/− C57BL/6 mice were maintained in the barrier facility at the University of Texas Southwestern Medical Center at Dallas. NOD2^−/− mice on the C57BL/6 background were obtained from Jackson Laboratories and were bred in the barrier facility at University of Texas Southwestern (UT Southwestern). Caspase-1^−/− mice were obtained from Dr. Richard Flavell (18). All experiments were performed using protocols approved by the Institutional Animal Care and Use Committees of the UT Southwestern Medical Center. All conventional mice within an experiment were cohoused for 1 week to normalize the microflora among the different mouse strains. This was done either by placing mice into a single cage, or in the case of experiments involving larger numbers of mice, by daily mixing of bedding among multiple cages. Germ-free C57BL/6 mice were maintained in plastic gnotobiotic isolators as previously described (5).

FLAG-MyD88 was amplified using primers containing BglII restriction sites (Table S1) from pCMV-FLAG-MyD88 (a kind gift of R. Medzhitov). The TAg fragment of pCR-TAg-hGH (21) was removed by digestion with BamHI and replaced with the BglII-digested FLAG-MyD88 fragment. Correct orientation of the cloned FLAG-MyD88 was determined by restriction mapping and sequencing. This yielded pCR2-FLAG-MyD88-hGH, which contained FLAG-MyD88 under the control of nucleotides −6,500 to 134 of the mouse Cryptdin-2 (CR2) gene (32) and immediately upstream of hGH^{+3 to +2150}. The hGH sequences enhance the efficiency of transgene expression; however, hGH protein is not produced from the RNA transcript (21). A 10.2 kb fragment containing CR2-FLAG-MyD88-hGH was released from pCR2-FLAG-MyD88-hGH by digestion with EcoRI, and was used to generate transgenic mice (UT Southwestern Transgenic Core Facility). The CR2-FLAG-MyD88 mice were backcrossed to MyD88^−/− mice. Expression of the MyD88 transgene was assessed by RT-PCR using specific primers (Table S1). Endogenous MyD88 genotype was verified using the primers directed against the MyD88 gene (Table S1).

Cryosections 7 μm thick were cut from OCT-embedded small intestine using previously published protocols (33). Paneth cells were harvested by laser capture microdissection using an Arcturus PixCell II system and CapSure HS LCM caps (Arcturus) as previously described (5). Total RNA was prepared from captured cells using the Arcturus PicoPure RNA Isolation Kit, and was used as template for cDNA synthesis as previously described (5).

Total RNA was isolated from small intestinal tissues using the Qiagen RNeasy RNA Isolation Kit and was used to synthesize cDNA. SYBR green-based Q-PCR used Stratagene Brilliant Q-PCR Master Mix and specific primers (Table S1). Signals were normalized to 18S rRNA levels within each sample using published primers (5), and normalized data were used to quantitate relative levels of gene expression using ΔΔC_t analysis.

Bouins-fixed, paraffin-embedded tissue sections were stained with anti-RegIIIγ antiserum (6) or preimmune serum and detected using a goat anti-rabbit IgG Cy3 conjugate (Biomeda). Tissues were counterstained with Hoechst dye. Anti-FLAG immunohistochemistry was carried out using a polyclonal anti-FLAG antibody raised in rabbits (Sigma) and goat anti-rabbit IgG Cy3 conjugate. All images were captured using a 250 ms exposure.

Age-matched conventional C57B6 wild-type and CR2-MyD88 Tg mice were given ampicillin (1g/L; Sigma), vancomycin (500 mg/L; Sigma), neomycin sulfate (1g/L; Sigma), and metronidazole (1g/L; Sigma) in drinking water for 4 weeks as described previously (13). Depletion of the commensal microbiota was verified by aerobic and anaerobic culture of feces. At the end of the treatment, mice were killed and intestinal tissues were snap frozen for RNA extraction or prepared for immunohistochemistry. Depletion of the small intestinal microbiota was verified by aerobic and anaerobic culture of luminal contents following sacrifice.

Age-matched conventional C57B6 wild-type, Myd88^−/−, and CR2-MyD88 Tg mice were cohoused together in the same cage for 1 week to minimize microflora composition differences among the different mouse strains and among mice that had been housed in different cages. The mice were killed, and small intestinal colonization levels were measured by dilution plating of luminal contents. Bacterial levels in spleen and mesenteric lymph nodes were determined by dilution plating of homogenized tissue. For additional information, see SI Materials and Methods.