Main

The human gut contains a diverse population of nonpathogenic, commensal bacteria that contribute to gastrointestinal health and disease. Considerable clinical and experimental evidence links immune responses directed against the normal bacterial flora to the pathogenesis of human inflammatory bowel disease1,2,3,4. Paradoxically, nonpathogenic gut bacteria are also thought to contribute to immune homeostasis by altering microbial balance or by specifically interacting with the gut immune system, mechanisms purported to underlie the therapeutic basis for bacterial probiosis in inflammatory bowel disease2,5,6. Although the implication of these claims for gut health is potentially very important, the cellular and molecular mechanisms by which individual members of the commensal flora contribute to immune homeostasis have not been elucidated. Study of these mechanisms could underpin new therapies for inflammatory gut disorders.

In terms of target sites, intestinal epithelial cells provide the first point of contact for bacteria within the gut lumen; they also interface and segregate the gut immune system and therefore have a pivotal function in bacteria-host communication. By virtue of recognition receptors expressed on apical and/or basolateral surfaces of epithelial cells7, bacterial moieties common to commensal and pathogenic bacteria bind and activate signaling cascades that can trigger proinflammatory gene transcription. Although the presence of virulence factors on pathogenic bacteria amplifies inflammatory immune responses, the molecular basis of hyporeponsiveness to commensal bacteria is not fully appreciated.

Regulatory mechanisms driven by CD4+ regulatory T cells in the lamina propria are important in maintaining tolerance to enteric bacteria8. However, it is likely that other equally important mechanisms controlling innate immunity and inflammation operate at the level of epithelial cells. That nonpathogenic gut bacteria contribute to this process is also a relatively new idea9. Given the complex mechanisms defining host-pathogen adaptation, the possibility also exists that homologous systems evolved amongst the commensal bacterial species that reside permanently within the gastrointestinal tract and that such adaptations actively contribute to immunological tolerance and homeostasis within the healthy gut. Here we describe an anti-inflammatory mechanism activated by Bacteroides thetaiotaomicron, a prevalent anaerobe of the human intestine; this bacterium attenuates proinflammatory cytokine expression by promoting nuclear export of NF-κB subunit RelA, through a PPAR-γ-dependent pathway. This mechanism highlights new cellular targets and approaches for therapeutic intervention in the treatment of inflammatory disease.

Results

B. thetaiotaomicron attenuates gut inflammation

The commensal microflora, in addition to driving the expansion of lymphoid tissues in the gut, contribute to many other functions, including colonization resistance. That they also actively regulate local immune responses within the gut is also a distinct possibility. We therefore hypothesized that commensal bacteria influence immunological outcome by modulating the innate immune response mounted by intestinal epithelial cells. To investigate this, we examined the effect of B. thetaiotaomicron, a prevalent commensal anaerobic bacterium of the human gut microflora, on the acute inflammatory response triggered in intestinal cells by exposure to pathogenic Salmonella enterica serovar Enteritidis. Using cDNA microarray technology, we identified several inflammatory genes whose induction after exposure of Caco-2 cells to S. enteritidis was modulated by the presence of B. thetaiotaomicron. Genes relevant to inflammation and showing large responses included those encoding tumor necrosis factor (TNF), interleukin 8 (IL-8) and cyclooxygenase 2. We confirmed these gene expression changes by RNA hybridization, semiquantitative PCR (Fig. 1a,b) and real-time PCR10. The anti-inflammatory activity of B. thetaiotaomicron could also be demonstrated with other intestinal cells, including IEC-6 cells (Fig. 1c) and T84 cells (data not shown). The attenuating effect seemed to be specific to B. thetaiotaomicron, as a related aerotolerant strain, B. vulgatus, did not produce similar effects on these proinflammatory cytokines (Fig. 1d). Induction of IL-8 in Caco-2 cells by other inflammatory mediators, including IL-1α, IL-1β, TNF, phorbol 12-myristate 13-acetate (PMA), lipopolysaccharide (LPS), purified flagellin and enterohemorrhagic Escherichia coli 0157:H7 demonstrated that B. thetaiotaomicron had a specific antagonistic effect on the IL-8 expression induced by PMA, flagellin and E. coli 0157:H7, whereas the activation induced by IL-1α and IL-1β was unaffected (Fig. 1b and data not shown). LPS and TNF did not induce IL-8 expression in Caco-2 cells. The lack of an effect of B. thetaiotaomicron on the expression of IL-8 induced by IL-1α and IL-1β indicated that although the bacterium attenuated inflammatory cytokine expression induced by several stimuli, it was not universally effective, indicating either conditional regulation or elements of specificity in its mode of action.

Figure 1: B. thetaiotaomicron specifically attenuates inflammatory responses in Caco-2 cells and in vivo.
figure 1

(a) Caco-2 cells were incubated for 2 h with medium (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron (4) and cytokine expression was determined by RNA blot: Tnf, TNF; Il8, IL-8; Cxcl2, MIP-2α, Tgfb1, TGF-β; Ptgs2, cyclooxygenase 2; Gapd, GAPDH. (b) Caco-2 cells were incubated for 2 h with medium (1); bacteria or ligands as indicated above each gel (2); bacteria or ligands as indicated above each gel, but including B. thetaiotaomicron (3); or B. thetaiotaomicron (4). IL-8 expression was determined by PCR. (c) IEC-6 cells were incubated with bacteria as described in a and the expression of TNF was determined by RNA blot. (d) Caco-2 cells were incubated for 2 h with medium (1); S. enteritidis (2); S. enteritidis and B. vulgatus (3a); or B. vulgatus (4a) and cytokine expression (left margin) was determined by RNA blot. (e) Transepithelial migration of PMN cells through a Caco-2 monolayer, as determined by MPO assay. Treatments 1–4 were as described in a. ***, P < 0.001. (f) In vivo confirmation of the anti-inflammatory effect of B. thetaiotaomicron, as determined by MPO assay of rat ileal mucosa. Treatments 1–4 were as described in a. All experiments were repeated a minimum of three times. **, P < 0.01. (g) Histology of ileal tissue from a rat infected with S. enteritidis. (h) Histology of ileal tissue from a rat infected with S. enteritidis and B. thetaiotaomicron. Scale bars, 50 μm.

Full size image

We established the physiological relevance of the anti-inflammatory effect of B. thetaiotaomicron with an in vitro model of inflammation based on polymorphonuclear leukocyte (PMN) recruitment, monitored through myeloperoxidase (MPO) detection (Fig. 1e). The variable effect of S. enteritidis and B. thetaiotaomicron bacteria on PMN recruitment could be correlated with the secretion of IL-8 protein, a chemotactic factor produced by epithelial cells that is known to be essential for both PMN recruitment and transepithelial migration11,12. Consistent with the gene expression data (Fig. 1a), S. enteritidis induced 227 ± 33.1 pg/ml of IL-8 protein in culture supernatants over 4 h, whereas after coculture in the presence of B. thetaiotaomicron, the resultant concentration was significantly lower, at 104 ± 9.2 pg/ml (P < 0.001). We also demonstrated the efficacy of this anti-inflammatory activity with 'minimal flora' rats, in which the normal microbial flora were maintained at low levels by housing rat pups in isolator conditions. Pups were dosed orally with B. thetaiotaomicron, and successful colonization was confirmed by specific amplification of bacterial DNA from rat jejunal and ileal tissues and by conventional culture-based microbiological analysis (data not shown). Rats orally infected with S. enteritidis and B. thetaiotaomicron had significantly attenuated MPO concentrations compared with that of rats challenged with S. enteritidis without therapeutic administration of B. thetaiotaomicron (Fig. 1f). Consistent with these data, we found substantial histological disruption of the villus and crypt structure concomitant with extensive cellular infiltrate in the lamina propria of rats infected with S. enteritidis (Fig. 1g). In contrast, although some cellular infiltrate was present in the rats treated with S. enteritidis and B. thetaiotaomicron, their intestinal structure was largely preserved (Fig. 1h). The substantial anti-inflammatory effect derived from in vivo colonization with this anaerobic bacterium provided an important physiological confirmation of the responses seen in our Caco-2 cell model system.

B. thetaiotaomicron attenuates nuclear RelA

NF-κB is a nuclear factor involved in the transcriptional regulation of inflammatory genes. Several signal transduction cascades, including those generated in response to bacterial infection, trigger nuclear translocation of NF-κB as a sequel to the phosphorylation, ubiquitination and proteosome-mediated degradation of the inhibitory IκB proteins13,14. Consistent with this, we found that exposure of Caco-2 cells to S. enteritidis for 2 h enhanced the nuclear translocation of the NF-κB subunit RelA (Fig. 2a, top row). Gel-shift and supershift analyses showed that RelA was the main nuclear subcomponent of NF-κB after incubation of Caco-2 cells with salmonella. When Caco-2 cells were coincubated with salmonella and B. thetaiotaomicron, a similar amount of nuclear import of RelA was apparent at 30 and 60 min (Fig. 2b). However, unlike the cells treated with Salmonella alone, B. thetaiotaomicron induced nuclear export of RelA, with considerably reduced amounts in the nucleus and a mainly cytoplasmic redistribution at 2 h (Fig. 2a,b and data not shown). Given our earlier finding that B. thetaiotaomicron did not attenuate the IL-1α- or IL-1β-mediated induction of IL-8, we investigated if this difference was because of differential effects on the specifically activated NF-κB subunits. B. thetaiotaomicron had profound effects on all stimuli that propagated their inflammatory response through a sustained activation of RelA (S. enteritidis, E. coli, PMA and flagellin) by reducing nuclear RelA accumulation. IL-1α and IL-1β, however, produced only a transient activation of RelA evident at 1 h but negligible at 2 h after stimulation (data not shown). As was evident from the time course studies, the B. thetaiotaomicron effect on nuclear RelA was only apparent at times after 1 h of exposure to this bacterium (Fig. 2b), when the induction of RelA by IL-1α and IL-1β had subsided.

Figure 2: B. thetaiotaomicron attenuates inflammation by enhancing nuclear export of NF-κB RelA protein in Caco-2 cells.
figure 2

(a) Immunofluorescence microscopy of RelA (top row) and IκBα (bottom row). Treatments were for 2 h: medium alone (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron (4). Scale bars, 25 μm. (b) Nonshifted and supershifted EMSAs of Caco-2 nuclear extracts with a consensus NF-κB oligonucleotide and RelA antibody. Treatments were as described in a (times, below photos). (c) Nonshifted and supershifted EMSAs of Caco-2 nuclear extracts with consensus AP-1 oligonucleotide and JunD antibody. Treatments 1–4 were as described in a, for 2 h. (d) Immunoblot of immunoprecipitated IκBα and pIκBα from Caco-2 cells. Treatments 1–4 were as described in a (times, above blots). (e) RNA hybridization of Nfkbia (IκBα) in Caco-2 cells. Treatments 1–4 were as described in a. (f) Immunoblot of immunoprecipitated IκBα protein. Treatments 1–4 were as described in a. Data are representative of at least three independent experiments.

Full size image

Subsequent supershift assays demonstrated that the effect of B. thetaiotoamicron seemed to be highly specific for NF-κB RelA, as other nuclear transcription factors including activator protein 1 (AP-1), shown here to be composed mainly of JunD, were nearly unaffected after 2 h of incubation, when nuclear translocation of RelA had occurred (Fig. 2b,c and data not shown). The transcription factor requirements of certain inflammatory genes may explain their differential susceptibility to the attenuating effects of B. thetaiotaomicron. Heat-inactivated B. thetaiotaomicron, culture supernatant and conditioned culture media derived from Caco-2 cells exposed to B. thetaiotaomicron lacked this attenuating activity (data not shown), indicating that bacterial-epithelial cell contact was required. Moreover, the viability and growth of S. enteritidis and B. thetaiotaomicron bacterial strains were unaffected after coculture with intestinal epithelial cells (data not shown). Also, both epithelial attachment and invasion of S. enteritidis were unaffected by the presence of B. thetaiotaomicron (data not shown) and hence the observed effect of B. thetaiotaomicron on inflammatory cascades cannot be ascribed to differences in S. ente-ritidis recognition or activation of epithelial cell surface receptors.

We noted both phosphorylation and degradation of the NF-κB inhibitor IκBα after exposure to S. enteritidis in the presence or absence of B. thetaiotaomicron (Fig. 2d). There were no differences in the kinetics of IκBα phosphorylation or degradation, indicating that the regulation of NF-κB is downstream of IκBα activation and therefore is not due to the inhibition of IκBα kinase (IKK)activity, as shown for pathogens15. After 2 h of exposure to the bacteria, the total amounts of IκBα mRNA (Fig. 2e) and protein (Fig. 2a, bottom row, and f) were enhanced, indicative of transcriptionally active NF-κB16,17. Inhibition of ubiquitination and degradation of the IκBα protein has been postulated as being an important mechanism of immune regulation whereby commensal bacteria prevent or limit inflammation in the gut9. However, as shown here, the anti-inflammatory activity of B. thetaiotaomicron, a prevalent human gut anaerobe, cannot be explained by this mechanism.

Export of nuclear RelA independent of Crm-1

We postulated that the S. enteritidis–induced RelA-mediated transcription, which was initiated in the presence and absence of B. thetaiotaomicron, was reduced after 2 h of coculture with B. thetaiotaomicron because of precocious nuclear clearance of activated NF-κB RelA. Direct acetylation and deacetylation of RelA influences both its duration of action and nuclear export18. Furthermore, transforming growth factor-β (TGF-β), an anti-inflammatory cytokine that also inhibits NF-κB-mediated transcription, specifically modulates histone acetylation19. To investigate the possible function of acetylation in our system, we used the specific histone deacetylase inhibitor trichostatin A (TSA) to enhance nuclear acetylation in Caco-2 cells (data not shown). Although nuclear acetylation and the S. enteritidis–induced IL-8 expression were substantially increased, the attenuating effects mediated by B. thetaiotaomicron were unaffected by TSA treatment (data not shown). To characterize the nuclear export mechanism further, we then investigated whether the B. thetaiotaomicron–induced export of RelA occurred through Crm-1 (exportin 1), the only pathway defined so far for the nuclear export of NF-κB–IκBα complexes20. We achieved inhibition of this receptor pathway with the nuclear export inhibitor leptomycin B (LMB)21. LMB did not influence the attenuating effect of B. thetaiotaomicron on IL-8 expression, as determined by real-time PCR (data not shown). The nuclear export of RelA induced by B. thetaiotaomicron was not blocked by LMB treatment (Fig. 3) and therefore was independent of the Crm-1 pathway, indicating a different molecular mechanism of export.

Figure 3: The nuclear export of RelA induced by B. thetaioatoamicron is independent of the Crm-1 pathway.
figure 3

Caco-2 cells were incubated for 2 h with medium (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron (4) with (+ LMB) or without (No LMB) LMB. The cellular distribution of immunofluorescently detected RelA was determined. Nuc., nuclear; Cyt., cytoplasmic. Data are representative of at least three independent experiments ± s.d. ***, P < 0.001.

Full size image

B. thetaiotaomicron induces nuclear export of PPAR-γ

PPARs and their ligands have been identified as modulators of inflammation22,23,24,25. The modes of action of the PPARs in this context have not been fully defined, but both receptor-dependent and receptor-independent mechanisms have been documented26,27,28. Because PPAR-γ is a potentially important protein in the regulation of inflammatory cascades, we investigated the effects of B. thetaiotaomicron on its expression and localization in Caco-2 cells. Immunoblot and immunocytochemical analyses showed that challenge with S. enteritidis induced nuclear accumulation of PPAR-γ in Caco-2 cells (Fig. 4a,b). In contrast, PPAR-γ redistributed to the cytosol after coculture with S. enteritidis and B. thetaiotaomicron (Fig. 4a,b). PPAR-γ induced by salmonella was mostly detergent insoluble, indicating probable cytoskeletal association. However, PPAR-γ was present in approximately equal amounts in the detergent-soluble and detergent-insoluble extracts of Caco-2 cells cocultured with salmonella and B. thetaiotaomicron, and was mostly in the detergent fraction in cells incubated with B. thetaiotaomicron alone (Fig. 4a). This indicates differential regulation of the biochemical or biological properties of PPAR-γ by these bacteria. Further investigation of the time course of nuclear translocation of PPAR-γ after exposure to bacteria showed notable accumulation within 30 min with all treatments (Fig. 4c). This was sustained in cells cultured with S. enteritidis during the 2-hour culture period (Fig. 4c, top row), whereas in cells cocultured with B. thetaiotaomicron, clearance from the nucleus was evident by 60 min and completed by 2 h (Fig. 4c, bottom row). Notably, the nucleocytoplasmic shuttling of PPAR-γ induced in response to coculture of Caco-2 cells with S. enteritidis and B. thetaiotaomicron mimicked that of RelA. Furthermore, as with RelA (Fig. 3), the nuclear export of PPAR-γ was not blocked by LMB treatment (data not shown). These observations therefore cannot be attributed to export through the nuclear receptor Crm-1. The data are compatible with previous findings for other nuclear receptors29 but clearly denote a different mechanism for the nuclear export of RelA. We postulated that the nuclear export and cytosolic localization of PPAR-γ and RelA induced by B. thetaiotaomicron was due to the formation of a stable protein complex, possibly involving additional factors.

Figure 4: B. thetaiotaomicron induces cellular shuttling of PPAR-γ in Caco-2 cells.
figure 4

(a) Immunoblot of PPAR-γ. Caco-2 cells were incubated for 2 h with medium (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron. After incubations, cells were separated into nuclear and cytoplasmic extracts, or detergent-soluble (D. soluble) or detergent-insoluble (D. insoluble) fractions, and were blotted for PPAR-γ. (b) Immunofluorescence microscopy of PPAR-γ in Caco-2 cells. Treatments were as described in a. (c) Time course of PPAR-γ nuclear translocation. Immunofluorescence microscopy of PPAR-γ in Caco-2 cells treated with S. enteritidis (top row) or S. enteritidis and B. thetaiotaomicron (bottom row) for the times indicated below the photos. Scale bars, 25 μm. Data are representative of at least three independent experiments.

Full size image

B. thetaiotaomicron induces association of PPAR-γ and RelA

PPAR-γ and NF-κB can form a complex in solution30,31. Thus, we investigated whether direct associations between PPAR-γ and RelA represented a possible mechanism to explain their common cytoplasmic localization in intestinal cells exposed to S. enteritidis and B. thetaiotaomicron. Immunoprecipitation of PPAR-γ from cells treated with S. enteritidis in the presence of B. thetaiotaomicron resulted in the copurification of RelA (Fig. 5a). Although the functional domains for both the interaction between RelA and PPAR-γ and potentially for their nuclear export were unknown, we assessed a dominant negative PPAR-γ with two amino acid replacements in its C-terminal ligand-binding domain32 for RelA-binding activity. We reasoned that because this dominant negative PPAR-γ shows impaired release or recruitment of corepressor and coactivator proteins32, its interactions with other proteins might also be impaired. In support of this, when we tested the dominant negative PPAR-γ receptor32 in immunoprecipitation experiments with in vitro–translated proteins, we found no coassociation with RelA (Fig. 5b). In contrast, wild-type PPAR-γ coassociated with RelA, consistent with the immunoprecipitation experiments in Caco-2 cells. Both the DNA-binding and the C-terminal domains of the PPAR-γ protein are RelA-interacting regions33. Hence, in addition to identifying biologically important molecular interactions between PPAR-γ and RelA, our data indicate the functional importance of the C-terminal domain within the PPAR-γ protein as a potential RelA-interacting domain.

Figure 5: PPAR-γ, but not the dominant negative PPAR-γ, interacts with active RelA and facilitates its nuclear export.
figure 5

(a) Immunoblot of RelA derived from immunoprecipitation with PPAR-γ agarose antibodies in Caco-2 cells treated for 2 h with medium (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron. Right, a control with secondary (2°) antibody alone. B, blank (no cellular protein). (b) Immunodetection of in vitro–translated RelA after incubation with reticulocyte lysate only (1); reticulocyte lysate and in vitro–translated wild-type (WT) PPAR-γ (2); reticulocyte lysate and in vitro–translated dominant negative PPAR-γ (3) and immunoprecipitation with antibody to PPAR-γ. (c) NF-κB–luciferase reporter assay. Caco-2 cells were cotransfected with an NF-κB–luciferase reporter construct and a construct for the expression of the dominant negative (DN) PPAR-γ or wild-type (WT) PPAR-γ. Treatments were as described in a for 2 h; bacteria were then removed and cells were cultured for a further 6 h; luciferase expression was determined and data were normalized to the salmonella values to give percentage stimulation indices. (d) Caco-2 cells were transfected with a YFP-RelA chimeric construct and a CFP–wild-type (WT) PPAR-γ construct (top row) or a CFP–dominant negative (DN) PPAR-γ construct (bottom row). Then, 2 d after transfection, cells were incubated with S. enteritidis and B. thetaiotaomicron and the dual localization of RelA and PPAR-γ was examined by fluorescence microscopy with specific filter sets.

Full size image

To investigate whether the interaction of PPAR-γ and RelA after exposure to B. thetaiotaomicron led to the inhibition of RelA-induced transcription, we cotransfected cells with dominant negative PPAR-γ and a NF-κB–luciferase reporter construct. The salmonella-mediated induction of luciferase was not notably different in cells transfected with dominant negative PPAR-γ or wild-type PPAR-γ. In cells transfected with wild-type PPAR-γ and dominant negative PPAR-γ, S. enteritidis induced substantial reporter gene activity (Fig. 5c). In cells transfected with wild-type PPAR-γ, the previously identified attenuation effect of B. thetaiotaomicron was maintained, but in cells transfected with dominant negative PPAR-γ this was no longer evident (Fig. 5c). As the dominant negative PPAR-γ did not bind RelA in our in vitro translation and immunoprecipitation experiments, we postulated that although the dominant negative PPAR-γ was unlikely to reduce or limit the interaction between RelA and the native PPAR-γ in the Caco-2 cells, the overexpressed dominant negative PPAR-γ may have interfered with the export of this complex, possibly by binding to or interfering with as-yet-unidentified proteins essential for the export of RelA and PPAR-γ.

PPAR-γ-dependent nuclear export of RelA

We investigated the importance of PPAR-γ in relation to nuclear export of RelA after exposure to B. thetaiotaomicron with chimeric fluorescent protein constructs of wild-type and dominant negative PPAR-γ linked to the C-terminal domain of a cyan fluorescent protein (CFP) or chimeric RelA linked to yellow fluorescent protein (YFP)34. Much of the expressed PPAR-γ protein was localized within the nuclei of transfected Caco-2 (and Hela) cells, whereas the RelA was mainly cytoplasmic. In Caco-2 cells transfected with wild-type CFP-PPAR-γ and YFP-RelA and treated with S. enteritidis, the main localization of these two proteins was nuclear, but there was evidence of punctate cytosolic labeling and colocalization of PPAR-γ and RelA after coincubation with B. thetaiotaomicron (Fig. 5d, top row). In cells transfected with dominant negative CFP–PPAR-γ, this labeling was absent from the cytosolic compartment (Fig. 5d, bottom row), indicating that nuclear export of the dominant negative PPAR-γ and RelA in response to B. thetaiotaomicron was impaired. Hence, PPAR-γ seems to be essential for both nuclear export and the cytosolic distribution of RelA induced in intestinal cells by B. thetaiotaomicron. Our data indicate that conditionally regulated proteins like PPAR-γ can govern the distribution and function of RelA, analogous to a previously reported mechanism involving IκBα35,36.

The distribution of CFP–PPAR-γ and YFP-RelA was very similar in the nuclei of transfected Caco-2 and Hela cells after culture in the presence of S. enteritidis and B. thetaiotaomicron (Fig. 5d and data not shown). This common distribution may indicate that interaction between PPAR-γ and RelA occurs within the nuclear compartment.

Decrease in PPAR-γ abolishes anti-inflammatory mechanism

To support our evidence of the PPAR-γ-mediated nuclear export of RelA, we used RNA interference (RNAi)37,38 to decrease the amount of PPAR-γ. To allow for rapid and robust analysis of the RNAi effect, we monitored the inhibition of an engineered CFP–PPAR-γ chimeric construct in transiently transfected Caco-2 cells cotransfected with YFP-RelA chimeric construct as a control. This design provided the most accurate method for the determination of RNAi-mediated effects on PPAR-γ amounts exclusively in the transfected population of cells. As shown in dually transfected cells, we noted a considerably decrease in CFP–PPAR-γ protein with no effect on the amounts of YFP-RelA (Fig. 6a). This was also demonstrated with immunodetection (Fig. 6b).

Figure 6: RNA interference of PPAR-γ.
figure 6

(a) Caco-2 cells were transfected with YFP-RelA and CFP–PPAR-γ constructs and a PPAR-γ-RNAi construct (top row) or negative control RNAi construct (bottom row). Transfected cells were subject to fluorescence microscopy with specific YFP (left) and CFP (right) filter sets. (b) Immunoblots of Caco-2 cells transfected with YFP-RelA and CFP–PPAR-γ constructs and a PPAR-γ-RNAi or a negative control RNAi construct; extracts were immunostained with antibodies specific for PPAR-γ and RelA.

Full size image

When we repeated these experiments, including coculture with S. enteritidis and B. thetaiotaomicron, cells transfected with the PPAR-γ-specific RNAi construct showed nuclear localization of RelA at 2 h, which was in contrast to all of the adjacent, non-RNAi-transfected cells, which showed cytosolic distribution of RelA (Fig. 7a, top row). The negative control RNAi construct had no effect on RelA, with transfected cells showing a mainly cytosolic distribution (Fig. 7a, bottom row). To establish that a decrease in PPAR-γ affected specific NF-κB-mediated gene expression, we measured IL-8 mRNA by real-time PCR after incubation with S. enteritidis (Fig. 7b) or purified S. enteritidis flagellin (Fig. 7c) in the presence and absence of B. thetaiotaomicron. As in the previous studies with dominant negative PPAR-γ and NF-κB-luciferase, transfection with PPAR-γ RNAi constructs had no effect on ligand-induced IL-8 concentrations (Fig. 7b,c). The decrease in PPAR-γ by RNAi resulted in a significant reduction in the inhibitory effect of B. thetaiotaomicron on IL-8 expression (salmonella, 9.1 ± 2.1%; flagellin, 30.5 ± 8.2%; Fig. 7d). We attributed this reduction exclusively to the transfected cell population that showed reduced PPAR-γ levels. For these experiments, routine transfection efficiencies were in the range of 18–26% and hence correction of the data to assume total cell transfection would result in a loss of the inhibitory effect of B. thetaiotoamicron equivalent to that seen in dominant negative PPAR-γ–NF-κB luciferase experiments. Overall, these studies show that B. thetaiotoamicron attenuates inflammatory responses in intestinal epithelial cells by enhancing the nuclear export of the NF-κB RelA subunit through a PPAR-γ-dependent mechanism.

Figure 7: PPAR-γ RNAi abolishes the nuclear export of RelA and the transcriptional inhibition of IL-8 in Caco-2 cells treated with S. enteritidis and B. thetaiotaomicron.
figure 7

(a) Caco-2 cells were dually transfected with pECFP-C1 (CFP only; control vector) and a PPAR-γ RNAi construct (top row) or a negative control RNAi construct (bottom row) and were incubated for 2 h with S. enteritidis and B. thetaiotaomicron. Transfected cells were detected by CFP fluorescence (left) and RelA distribution was identified by immunofluorescence (right). Arrows indicate transfected cells. Data are representative of at least three independent experiments. (b) Caco-2 cells were transfected with pECFP-C1 and a PPAR-γ RNAi construct or a negative control RNAi construct. Transfected cells were incubated for 2 h with medium alone (1); S. enteritidis (2); S. enteritidis and B. thetaiotaomicron (3); or B. thetaiotaomicron alone (4). Expression of IL-8 was determined by real-time PCR. Data are normalized to the salmonella response and are representative of one of at least three independent experiments. (c) Caco-2 cells were transfected as in (a) and incubated for 2 h with medium alone (1); S. enteritidis recombinant flagellin (2); S. enteritidis flagellin with B. thetaiotaomicron (3); or B. thetaiotaomicron alone (4). Expression of IL-8 was determined by real-time PCR. Data are normalized to the S. enteritidis flagellin response and are representative of one of at least three independent experiments. (d) Cells were treated with S. enteritidis and B. thetaiotaomicron (1) or S. enteritidis flagellin and B. thetaiotaomicron (2) and IL-8 expression was determined. Data are normalized values for the B. thetaiotaomicron inhibition of IL-8 mRNA in the presence of either RNAi PPAR-γ or the RNAi negative control. Data are representative of at least three experiments for S. enteritidis and two experiments for flagellin ± s.d. **, P < 0.01, ***, P < 0.001. In each experiment, transfection efficiencies were determined by monitoring the percentage of CFP-expressing cells relative to cells labeled with 4,6-diamidino-2-phenylindole.

Full size image

Discussion

The healthy gut maintains a physiological level of inflammation in response to the established bacterial flora. However, the presence of a threshold number of pathogenic bacteria is sufficient to activate transcriptional factors, including NF-κB, which trigger proinflammatory gene expression, inducing both innate and adaptive mechanisms of defense39. Although such mucosal responses are essential in combating gut infections, the intervention of regulatory systems that quench inflammation and restore immune homeostasis is an absolute requirement to prevent progression from the acute to the chronic state. In this context, the anti-inflammatory cytokines IL-10 (ref. 40) and TGF-β19,41 derived from epithelial and regulatory T cells are recognized to be important.

Gut bacteria, typically pathogenic species, prevent or limit the inflammatory response by targeting NF-κB or, more specifically, by inhibiting its activation9,15. We found that B. thetaiotaomicron also acted on NF-κB but that its mode of action was distinct. B. thetaiotaomicron targeted transcriptionally active RelA and induced precocious nuclear clearance, thereby limiting the duration of NF-κB action. The physiological effect of this anti-inflammatory bacterium was measurable in vivo and, as a prevalent constituent of the normal human gut microflora, its contribution to immune homeostasis and innate and adaptive mechanisms of defense is likely to be important.

In addition to the effect on RelA, both pathogenic salmonella and nonpathogenic bacteroides induced PPAR-γ protein expression, but its cellular localization was differentially modulated. PPAR-γ protein was nuclear in the presence of S. enteritidis but cytoplasmic in the presence of B. thetaiotaomicron. We noted that the kinetics of the nucleocytoplasmic export of PPAR-γ induced by B. thetaiotaomicron were reminiscent of those of RelA, indicating a possible association of these proteins during nuclear export. Consistent with this supposition, both RNAi reduction of constitutive PPAR-γ and transfection with dominant negative PPAR-γ constructs prevented nuclear export of RelA, thereby confirming the importance of the B. thetaiotaomicron–mediated effects on PPAR-γ in determining the nucleocytoplasmic distribution of RelA. The RelA that remained in the nucleus during these studies retained transcriptional activity, demonstrated by both NF-κB reporter assays and the induction of IL-8 mRNA, indicating that PPAR-γ-dependent export of RelA was essential in attenuating the salmonella-mediated proinflammatory responses.

Immunoprecipitation of in vitro–translated proteins and epithelial cell extracts demonstrated that RelA and PPAR-γ were capable of physical association and that B. thetaiotaomicron could trigger this interaction. In vitro images also indicated that such an association might form in the nucleus. Although this is the first description to our knowledge of a nuclear PPAR-γ–NF-κB complex in epithelial cells, the formation of a similar complex that limits the transcriptional activity of the PPAR-γ protein was described in ST2 mesenchymal cells33. In vitro studies demonstrated that dominant negative PPAR-γ did not associate with RelA, and although additional experiments are required to further clarify the mechanism of association of these proteins in our system, both the DNA-binding and the C- terminal ligand-binding domain regions of PPAR-γ protein have been linked to NF-κB binding33. It is also possible that RelA and PPAR-γ are components of a much larger complex, as other proteins normally associated with PPAR-γ, such as PPAR-γ coactivator-1, can physically interact with RelA33.

Our data provide evidence for the IκBα- and Crm-1-independent export of RelA14, as LMB, a specific inhibitor of Crm-1, did not affect B. thetaiotaomicron–induced export of the PPAR-γ–RelA complex from the nucleus. This indicated an alternative nuclear export pathway, perhaps similar to that used by the thyroid receptor29. Consistent with the nuclear export function of PPAR-γ, the dominant negative PPAR-γ blocked RelA export from the nuclei of Caco-2 cells exposed to B. thetaiotaomicron, indicating that the dominant negative PPAR-γ interfered with the nuclear export of the endogenous receptor. The precise mechanism for how PPAR-γ facilitated the nuclear export of RelA remains to be elucidated, and will require further exploration. Collectively, however, these data provide insight into biologically important protein interactions and previously unknown nuclear export mechanisms that divert activated transcription factors to other cellular localizations, thereby limiting gene transcription. The nuclear export of NF-κB–PPAR-γ complexes, in addition to limiting inflammatory cascades, may also underpin other physiological functions for these proteins within the cell.

Our findings identify a previously unknown anti-inflammatory mechanism involving PPAR-γ and NF-κB and provide new insights into molecular mechanisms in epithelial cells that contribute to immune homeostasis. They also extend our understanding of the evolutionary adaptation of commensal bacteria within gut ecosystems and emphasize the potential for identifying bacterially derived mechanisms and/or bioactive molecules with immune-modulating functions. The identification of such factors has recently become more possible with the publication of the genomes of commensal bacteria, including B. thetaiotaomicron42. The exploitation of such findings may lead to the development of new therapeutics for human inflammatory bowel disease and other inflammatory conditions.

Methods

S. enteritidis and B. thetaiotaomicron coculture models.

Caco-2 and Hela cells were routinely cultured in 35-mm culture dishes. Based on optimized dose-response and time-course studies, typical experiments used cells incubated for 2 or 4 h with the following four treatments: medium alone; 1 × 108 S. enteritidis; 1 × 108 S. enteritidis plus 1 × 109 B. thetaiotaomicron; or 1 × 109 B. thetaiotaomicron. Bacteria were then removed by extensive washing. Alternative incubations included the following bacteria and/or ligands, where appropriate: E. coli 0157:H7 (1 × 108), B. vulgatus (1 × 109), PMA (300 ng/ml), IL-1α and IL-1β (20 ng/ml), TNF (20 ng/ml) and LMB (20 nM). Transepithelial migration of PMN cells through a Caco-2 cell monolayer grown on transwells was determined by MPO assay43. Caco-2 cells were incubated with bacteria for 2 h and thoroughly washed, and fresh medium was applied. Thereafter, the Caco-2 cells were incubated for a further 2 h before cells and media derived from the apical compartment were solubilized in PBS containing Triton X-100 (1%), and MPO concentrations were determined.

In vivo rat study.

Newly weaned (21-day-old) 'minimal flora' rats (fed on normal laboratory diets) were split into two groups and additionally fed anaerobically prepared jelly (0.5 g/d) or anaerobically prepared jelly containing 1 × 108 B. thetaiotaomicron for 19 d. Half of the rats in each group were then orally challenged with 1 × 108 S. enteritidis. MPO was measured in isolated ileal mucosa at 6 d after S. enteritidis infection. Experiments were undertaken at least three times with similar results.

Histology.

Rat ileal samples were fixed for 18–24 h at 20 °C in paraformaldehyde (4%) in 0.1 M phosphate buffer, pH 7.3. Tissues were subsequently washed in phosphate buffer, dehydrated in an ethanol series and embedded in Historesin (Leica). Sections 1 μm in thickness were stained with toluidine blue, and representative images were obtained with a Zeiss Axioskop microscope equipped with an AxioCam digital camera.

Cytokine analyses.

Cytokines in Caco-2 cells were analyzed with microarray (Atlas cytokine-receptor array; Clontech), RNA hybridizations and real-time and semiquantitative PCR. Total RNA or mRNA was isolated, cDNA was produced and experiments were done with standard conditions. IL-8 protein concentrations were determined by enzyme-linked immunosorbent assay.

PCR analysis.

IL-8 mRNA was estimated by semiquantitative and/or real-time PCR. For semiquantitative studies, reverse transcription of total RNA was done followed by trial PCR to determine the cycle range for the linear amplification of IL-8 and glyceraldehyde phosphodehydrogenase (GAPDH) sequences in separate reactions. Experiments were then done for the appropriate number of cycles to ensure that all samples were within the linear range for each product. Products were resolved on a 1% TAE agarose gel (0.5 μg/ml of ethidium bromide) and band intensities were determined by Quantity One software (Bio-Rad). IL-8 values were normalized relative to GAPDH values. Real-time PCR was done with predeveloped assay reagents for human IL-8 and GAPDH as described by the manufacturer (Applied Biosystems). Samples were run on an ABI Prism 7700 and analyzed with standard methodologies.

Analyses of NF-κB and AP-1.

Nuclear extracts of Caco-2 cells were analyzed by electrophoretic mobility-shift assay (EMSA). The probes used were double-stranded 32P-labeled oligonucleotides containing the consensus binding sequences for NF-κB (5′-AGTTGAGGGGACTTTCCCAGGC-3′) or AP-1 (5′-CGCTTGATGAGTCAGCCGGAA-3′). Products were separated by electrophoresis and visualized by autoradiography. EMSA supershifts were done with specific NF-κB and AP-1 subunit antibodies (RelA, sc-109X; JunD, sc-74X; Santa Cruz Biotechnology). The effects of B. thetaiotaomicron on NF-κB signaling were determined by EMSA, immunoblotting (RelA sc-109; IκBα, sc-371 and sc-371AC; pIκBα, sc-8404; Santa Cruz Biotechnology), NF-κB–luciferase reporter assays and target gene expression. Luciferase expression was determined with the Steady-Glo Luciferase assay system (Promega).

Immunoprecipitation experiments.

After experimental treatments, Caco-2 cells were washed and solubilized in PBS contianing Igepal CA-630 (1%), sodium deoxycholate (0.5%), SDS (0.1%) and protease inhibitor 'cocktail' (Sigma) and were incubated with agarose conjugates as described by the Santa Cruz Biotechnology immunoprecipitation and immunoblot protocols. The following antibody-agarose conjugates were used: RelA, sc-372AC; PPAR-γ sc-7273AC or sc-1984; and IκBα, sc-371AC.

Immunofluorescence analysis.

After experimental treatments, Caco-2 and Hela cells were fixed in paraformaldehyde (4%) and permeabilized in 0.2% Triton X-100 in PBS. Cells were incubated for 1 h at room temperature with primary antibodies (1 μg/ml) in PBS (RelA, sc-109; IκBα, sc-371; and PPAR-γ, sc-7196 and sc-1984) containing 1% serum from the species in which the secondary antibody was raised. Secondary antibodies (1 μg/ml) were either Alexa Fluor donkey antibody to goat or Alexa Fluor goat antibody to rabbit IgG (Molecular Probes), where appropriate. Labeled cells were mounted with Vectorshield (Vector) and examined on a Zeiss Axioskop 50 widefield fluorescence microscope or on a Bio-Rad Radiance 2100 laser-scanning microscope. Representative digital images were imported into Adobe Photoshop 6.0 for construction of dual-color overlays. In LMB experiments, digital images were obtained with standard conditions of illumination and exposure on a Zeiss Axiocam camera attached to a Zeiss Axioskop fluorescence microscope. Digital captures were converted to 8-bit grayscale images and mean pixel values were recorded with National Institutes of Health Image J software (http://rsb.info.nih.gov/ij/) for representative areas of nuclei and cytoplasm from 100 cells.

Immunofluorescence microscopy and immunoblotting.

Caco-2 cells were cotransfected with constructs encoding YFP-RelA and CFP-PPAR-γ, and either the PPAR-γ RNAi or the negative control pSilencer 3.0-H1 plasmid, at a ratio of 1:1:2, with lipofectamine 2000 (Invitrogen). Cells were cultured for 48 h before being examined on a Zeiss AxioVert fluorescence microscope with specific filters for the identification of YFP and CFP. All cell images were obtained with equal exposure times (4,000 ms). After being photographed, cells were washed with PBS and collected in SDS-PAGE gel loading buffer, resolved on a 4–12% gradient polyacrylamide gel and blotted onto a polyvinyldifluoride membrane. PPAR-γ and RelA were identified with specific primary antibodies (PPAR-γ, sc-7273; RelA, sc-109; Santa Cruz Biotechnology).

Construction and expression of fluorescence-tagged PPAR-γ.

The coding sequence of PPAR-γ and the dominant negative mutant of PPAR-γ32 were modified by PCR amplification with Pfu DNA polymerase to add an XhoI recognition sequence at the 5′ end and a SacII sequence at the 3′ end of the products. Products were digested with these restriction enzymes and cloned into pECFP-C1 and pEYFP-C1 (Promega). Successful construction was verified by DNA sequencing. Human RelA cloned into pEYFP-C1 was as reported34. Caco-2 or Hela cells were grown to 90% confluence and transfected with standard lipofectamine-mediated methods (Invitrogen). After 48 h, bacterial incubations were undertaken as described above; cells were fixed for 30 min at room temperature in paraformaldehyde (4%) in 0.1 M sodium phosphate buffer, pH 7.4, washed in PBS and examined on a Zeiss Axioskop 50 widefield fluorescence microscope equipped with custom filters for CFP and YFP. Representative digital images were imported into Adobe Photoshop 6.0 for construction of dual-color overlays. In some experiments, transfected cells were also subject to immunofluorescence as described above.

Preparation of PPAR-γ RNAi constructs.

Constructs to provide RNAi of human PPAR-γ were engineered with the pSilencer system (Ambion). After initial testing of five constructs, one was selected that showed the most robust RNAi effect on PPAR-γ. This construct contained the human PPAR-γ sequence from nucleotides 105–123 (5′-GCCCTTCACTACTGTTGAC-3′) in the plasmid pSilencer 3.0-H1. A negative control pSilencer 3.0-H1 plasmid containing 19 nucleotides with no known identity to any human sequence was provided with the system.