Induction of Protective IgA by Intestinal Dendritic Cells Carrying Commensal Bacteria

See allHide authors and affiliations

Science  12 Mar 2004:
Vol. 303, Issue 5664, pp. 1662-1665
DOI: 10.1126/science.1091334

This article has a correction. Please see:


The enormous number of commensal bacteria in the lower intestine of vertebrates share abundant molecular patterns used for innate immune recognition of pathogenic bacteria. We show that, even though commensals are rapidly killed by macrophages, intestinal dendritic cells (DCs) can retain small numbers of live commensals for several days. This allows DCs to selectively induce IgA, which helps protect against mucosal penetration by commensals. The commensal-loaded DCs are restricted to the mucosal immune compartment by the mesenteric lymph nodes, which ensures that immune responses to commensal bacteria are induced locally, without potentially damaging systemic immune responses.

Mammals and other vertebrates coexist with an extremely dense and diverse microflora of nonpathogenic bacteria in the lower intestine, which salvage the energy of otherwise indigestible dietary carbohydrates and compete with the growth of pathogenic organisms. Although this mutualism can break down in individuals with inflammatory bowel disease, coevolution of commensals and their hosts has ensured that inflammatory intestinal immunopathology is relatively rare.

Comparisons between experimental germ-free animals and those containing commensal bacteria have revealed that the mucosa has become highly adapted to the presence of the commensals, although the full functional significance of these adaptations is not completely understood (13). In the presence of commensal bacteria, IgA secreted across the intestinal mucosa accounts for >70% of total-body immunoglobulin production. However, it is unclear whether secretory IgA simply limits the growth of commensal bacteria in the intestinal lumen (4), or whether IgA-coated commensals are also inhibited from penetrating the surface epithelium.

Like their pathogenic counterparts, commensal bacteria express molecular patterns that can be detected by Toll-like receptors of the innate immune system. Yet although commensal bacteria or their products can induce intestinal B and T cells, they do this without triggering the neutrophil infiltrate or classical inflammation that is characteristic of a pathogenic infection.

To assess how the mucosal immune system deals with commensal intestinal bacteria that have penetrated the surface epithelial layer, we intragastrically challenged mice with doses of Enterobacter cloacae and other mouse commensals (5). After administration of 109 E. cloacae into the stomach of C57BL/6 wild-type mice, 200 to 800 bacteria could be recovered from washed mesenteric lymph node (MLN) cells for about 60 hours following the oral dose (Fig. 1A). Bacteria were not recovered from splenocytes (Fig. 1A) or other systemic tissues (6) after the challenge; the live bacteria were confined to the mucosal immune system, Deliberate intraperitoneal injection resulted in clearance to the spleen rather than the MLNs (6), so penetration in orally dosed animals was direct rather than via an intraperitoneal leak.

Fig. 1.

Penetration of E. cloacae after intestinal challenge. (A) Culturable bacteria from lymph nodes (filled symbols) or the spleen (open) of C57BL/6 wild-type mice after intragastric gavage (circles) of 109 CFU or intravenous injection (squares) of 107 CFU E. cloacae. C57BL/6 animals without MLNs are shown after gavage (red circles). Results are means ± SD for three or four mice per group, representative of 14 experiments. (B) Bacterial recovery following FACS sorting of mesenteric leukocytes 22 hours after gavage with 109 (white bars) or 108 (shaded bars) CFU. E. cloacae or S. typhimurium. (C) Antibiotic-resistant bacteria in mesenteric leukocytes after challenge with 109 naladixic acid (Nal) and 109 rifampicin (Rif)–resistant E. cloacae into separate disconnected intestinal loops of the same animal. Lysis of the washed cells before plating increased the bacterial colonies by 20 ± 7% (means ± SD), which showed that some cells contained multiple bacteria. Colonies from three experiments were analyzed for dual antibiotic resistance by replica plating [experiment (Exp) 1, n = 261; Exp 2, n = 235; Exp 3, n = 113)]. Control splenocytes (n = 200) are shown after intravenous injection of 106 of the two resistant strains or control mesenteric leukocytes (n = 300) after administration of both strains into a single loop. (D to F) FACS of C57BL/6 Peyer's patch leukocytes (D and E) or intestinal leukocytes (F) gated on CD11c+ 18 hours after gavage with 109 E. cloacae constitutively expressing GFP or (inset) with GFP E. cloacae. GFP+ frequencies (means ± SD, n ≥ 3) are representative of four experiments. Fluorescence microscopy (G) of the E. cloacae GFP (inset, control GFP strain) and (H) of a cryosection of Peyer's patch after gavage (inset, GFP gavaged Peyer's patch).

The live commensal bacteria isolated from MLNs were identified as residing in dendritic cells (DCs), as determined by bacterial culture after fluorescence-activated cell sorting (FACS) (Fig. 1B). Unexpectedly, we were unable to culture bacteria from sorted (CD11b+CD11c) macrophages (7) either from MLNs (Fig. 1B) or the Peyer's patches (6). However, ex vivo macrophage killing assays confirmed that macrophages are highly efficient at killing penetrant commensals within about 4 hours [(8), fig. S1], whereas DCs are relatively ineffective for killing bacteria (8). In contrast to challenge with E. cloacae, mice challenged with 108 Salmonella typhimurium showed penetration and growth of organisms that could be cultured from the spleen [2.76 ± 0.52 × 105 colony-forming units (CFU) per spleen at 18 hours; means ± SD n = 3]. Culturable Salmonella were also detected in both DCs and the sorted macrophage fraction from the MLNs (Fig. 1B). Therefore, unlike pathogens (9), commensal bacteria are efficiently killed by macrophages, so they can only survive in low numbers within local DCs, and they are prevented from entering the systemic immune compartment.

We next determined whether commensal bacteria were carried to the MLNs by the DCs, rather than penetrating freely and being taken up by DCs only after arrival in the MLNs. To do this, we challenged mice in one surgically isolated intestinal loop (fig. S2) with rifampicin-resistant E. cloacae and in another loop in the same animal with naladixic acid–resistant E. cloacae. Lysis of the washed DCs before plating increased the numbers of bacterial colonies by an average of 20%, so a proportion of DCs contained more than one bacteria. Because both loops were connected via lymphatics to the MLNs, we reasoned that dual antibiotic resistance of bacteria isolated from individual DCs would only occur if free bacteria had been able to penetrate and become associated with DCs in the MLNs themselves. In fact, colonies with dual resistance were found only if both antibiotic resistant strains were given into the same loop (Fig. 1C), which indicated that bacteria were taken up by DCs at mucosal sites and transported within DCs to the MLNs. In a different approach, we measured penetration of a challenge dose of E. cloacae in mice deficient for the chemokine receptor CCR7 [hence, harboring a defect in dendritic cell homing (10)]; in this case, bacteria that could be cultured were almost absent from the MLNs (Table 1). These results indicated that bacterial penetrance occurred as a result of migration within DCs.

Table 1.

Bacterial recovery in mesenteric leukocytes after intestinal challenge with E. cloacae. Animals of the strains shown maintained under SPF conditions were challenged intragastrically with different doses of live E. cloacae. C57BL/6 conditioned mice had been previously treated with E. cloacae as described in the legend to Fig. 2 and rested for 10 days after conditioning before carrying out the challenge. Results show the numbers of bacterial colonies in washed mesenteric leukocytes (means ± SD, n = 4) and are representative of two experiments. The results were significant at P < 0.005 (paired Wilcoxon test) between each of the different groups tested.

Challenge dose of E. cloacae C57BL/6 JH-/- CCR7-/- C57BL/6 conditioned
1 × 109 640 ± 60 980 ± 110 6 290 ± 65
3.3 ± 108 42 ± 25 182 ± 61 0 18 ± 3
1 × 108 19 ± 7 45 ± 7 1 3 ± 1
PBS 3 ± 1 5 ± 2 0 1 ± 1

To visualize bacteria within Peyer's patch DCs, E. cloacae were transformed with a plasmid expressing green fluorescent protein (GFP) under the constitutive control of the rpsM ribosomal protein promoter (11). Following intragastric challenge with these bacteria, GFP-positive CD8+ and CD8 cells in the Peyer's patch CD11c+ cell fraction could be detected by FACS (Fig. 1D). However, these cells were CD86+ (Fig. 1E), which indicated that the DCs had become activated after E. cloacae phagocytosis.

Epithelial antigens have been shown to be sampled by DCs (12), and pathogens can be taken up by DCs protruding through the villous epithelial layer (13). However, in our experiments with specific pathogen–free (SPF) mice gavaged with GFP–E. cloacae, GFP-positive DCs could not be isolated from the intestinal lamina propria by FACS (Fig. 1F). In addition, we could not detect live bacteria by culture of the lamina propria DC fraction at 5 or 22 hours after challenge, whereas Peyer's patch DCs did contain live bacteria at these times (table S1). Green fluorescent staining in the cytoplasm of surface-positive CD11c+ cells in the Peyer's patches (Fig. 1, G and H), but not the lamina propria (6), could also be visualized by immunohistochemistry. These results suggested that sampling of commensal bacteria is likely to occur mainly through the specialized M cell layer of Peyer's patches and lymphoid follicles (14, 15).

When mice were repeatedly conditioned with 108 live E. cloacae by intestinal gavage, IgA was selectively induced in the mucosa and serum (Fig. 2, table S2). No mucosal or serum IgA induction was seen in mice treated in parallel at each time point with the identical preparation of E. cloacae, which had been split and killed by heat treatment before administration (Fig. 2, table S2). Similar induction was also seen when the conditioning protocol was carried out with other mouse commensals Enterobacter faecalis, Bacillus puminatus, and Staphylococcus saprophyticus (table S3). This response was partly T cell–independent, as only 19 to 24% of wild-type IgA levels could be induced in mice lacking the T cell receptor [TCRβ–/–δ–/– (fig. S3)].

Fig. 2.

Intestinal and splenic histology of C57BL/6 mice conditioned by gavage with live 108 E. cloacae every third day for 27 days and sampled at 34 days. The same conditioning protocol was followed for heat-killed bacteria after the washed bacterial preparations were split before each treatment. Sham-conditioned mice were gavaged with 350 μl phosphate-buffered saline (PBS). Where indicated, the MLNs had been surgically excised 4 weeks before (MLN). Completeness of mesenteric lymphadenectomy was confirmed in all MLN animals at autopsy, and continuity of lymphatics (seen as white lines after administration of 200 μl olive oil 12 hours before) was confirmed at the beginning of the conditioning period in postoperative controls. Numbers of IgA-secreting intestinal cells and serum IgA concentrations (means ± SD) are shown on the corresponding panels (n = three or four mice per group). Splenic size was considerably increased in the MLN animals (6.9 ± 1.1 × 108 splenocytes in conditioned C57BL/6 MLN and 1.1 ± 0.46 × 108 splenocytes in conditioned C57BL/6 MLN+ controls).

We next addressed whether MLNs are obligatory either for restriction of the effects of commensals to the mucosal immune system or as sites for IgA induction by commensal bacteria. To do this, mesenteric nodes from wild-type (C57BL/6) mice were surgically excised by microdissection along the superior mesenteric artery; reanastomosis of the mesenteric lymphatics occurred during healing (5). Conditioning of these MLN-deficient animals by repeated administration of commensal bacteria induced IgA with the same efficiency seen in unmanipulated controls (Fig. 2, table S2). However, splenic size was considerably increased in MLN-deficient mice treated with commensal bacteria, and histology also revealed considerable expansion of their splenic marginal zones (Fig. 2). We also showed directly that live bacteria could be cultured from the spleens of MLN-deficient animals after gavage with E. cloacae, whereas bacteria did not penetrate beyond the MLNs of intact wild-type mice (Fig. 1A). The MLNs therefore confine the live commensals within DCs to the mucosal immune system and avoid systemic priming by these organisms (6, 16).

We next examined whether purified DCs, loaded with bacteria, could directly stimulate IgA production by B cells. To do this, B cells with or without CD4+ and T cells purified from the MLNs of unmanipulated mice were cultured with CD11c+ DCs, purified from the Peyer's patches of mice that had been pulsed with E. cloacae 18 hours earlier. FACS analysis confirmed that IgA+ B cells had been induced only in cultures containing E. cloacae–pulsed DCs (Fig. 3, A and E). IgA secretion by B cells was confirmed by using enzyme-linked immmunosorbent assay (ELISA) (Fig. 3F).

Fig. 3.

Induction and in vivo function of IgA by commensal-loaded DCs. (A to E) FACS after 3 days of coculture of 104 CD11c+ DCs purified by magnetic cell sorting (MACS) from the Peyer's patches of mice that had been pulsed with 109 E. cloacae in vivo 18 hours before with (A) 2 × 105 B cells or (B) 2 × 105 B cells and 2 × 105 CD4+ T cells MACS-purified from the MLNs of unmanipulated B6 mice [coculture of CD4+ T cells and DCs only shown as inset to (B)]. (C) and (D) as for (A) and (B), respectively, but DCs purified from Peyer's patches of mice gavaged with heat-killed E. cloacae (inset, cultures without added DCs). (E) FACS after 3-day coculture of 104 CD11c+ DCs from MLNs of mice lacking T cell receptor (TCRβ–/–δ–/–) pulsed in vivo with live E. cloacae and 2 × 105 B cells from MLNs of unmanipulated TCRβ–/–δ–/–mice. Values represent IgA+/B220+ frequencies. (F) IgA concentrations at day 6 in the culture supernatants of (A to E and B, insert). (G) Cecal bacterial densities during recolonization of antibody-deficient (JH–/–, open symbols) or control C57BL/6 (filled symbols) germ-free mice. Aerobic (squares) and anaerobic (circles) bacteria are shown. (H) Penetration of aerobic bacteria into the MLNs in the same two groups of germ-free mice during recolonization with an SPF flora as in (G). All results are means ± SD for three or four mice and are representative of at least three separate experiments.

The in vivo function of intestinal IgA antibody secretion on the commensal flora was next addressed in two ways. First, bacterial penetration and luminal densities were compared between germ-free C57BL/6 and JH–/– (antibody-deficient) mice undergoing recolonization with an SPF flora in the same cage. Despite no significant differences in the final densities of intestinal bacteria or their subtypes in this very restricted flora [Fig. 3G and (6)], antibody-deficient animals took >20 days longer to adapt to the presence of intestinal bacteria and to prevent aerobic commensals from reaching the MLNs from the intestinal lumen. Second, we measured the penetration of challenge doses of E. cloacae to the MLNs in SPF C57BL/6 and JH–/– mice that already had an intestinal flora (Table 1). Wild-type mice showed less commensal penetration than their antibody-deficient JH–/– counterparts, which was reduced even further after conditioning to induce secretory IgA levels (Table 1).

In summary, mutualism between commensal bacteria and their host is established by the competence of the host mucosal immune system to restrict the penetration of commensals to low levels. Whereas pathogens can avoid being killed by subverting the biocidal mechanisms of macrophages (9), it is likely that the commensals have coevolved with their hosts not to do this. Efficient killing of commensals prevents an inflammatory response in the intestinal mucosa, which would be detrimental to the bacteria as it would increase intestinal secretion and motility, as well as disrupt the commensals' luminal habitat. At the same time, commensals do persist in DCs, allowing them to induce local protective mucosal immune responses. The commensal-loaded DCs are restricted to the mucosal inductive sites through the barrier of the MLNs. This allows a productive mucosal immune response without inducing unnecessary systemic immunity to these organisms. Provided the MLNs are intact, the result of repeated intestinal commensal priming is to induce IgA selectively and locally, forming one of the layers of the mucosal barrier that limits bacterial penetration.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 to S3


References and Notes

Stay Connected to Science

Navigate This Article