The Role of Dendritic Cells, B Cells, and M Cells in Gut-Oriented Immune Responses

Immune responses generated to orally administered Ags have certain local and systemic features that distinguish them from other immune responses. For example, secretory IgA, a characteristic local feature of the mucosa, generates a barrier that prevents Ags from attaching to and penetrating the mucosal epithelium and forms immune complexes with Ags within epithelial cells of the gut lamina propria (1). Immune responses to fed Ags are also characterized by specific types of local and systemic T cell responses. Locally, in the Peyer’s patches that line the gut, oral administration of Ag leads to the induction of Th3 cells that produce TGF-β (2), a cytokine that promotes B cell switching to the production of IgA (3). Systemically, the typical response to fed Ag is characterized by a decrease in both delayed-type hypersensitivity (DTH)² and Th cell proliferation upon a subsequent immunization with the same Ag (4). Other characteristics of the systemic T cell response depend on the dose of fed Ag. High doses can lead to deletion of the Ag-specific T cells within Peyer’s patches (5, 6), whereas low doses can lead to a state in which the immune response has switched from the production of a DTH/Th1 response to the production of Th2/Th3-type responses (7, 8, 9, 10, 11) and, in some cases, the production of IgA (12, 13, 14, 15). In addition, T cells from animals fed low doses can be adoptively transferred to depress the DTH and proliferative responses of normal naive recipients (16). In the original host, fed T cells have suppressive bystander effects on DTH responses to new Ags when given together in an immunization with the original fed Ag (17). Both the transferable suppression and the bystander effect have been associated with Th3 cells producing TGF-β, as well as with Th2 cells that produce IL-4 and IL-10 (18), although subsequent high-dose feeding may change these characteristics (19). Thus, the response to fed Ag, whether called oral vaccination (because it produces IgA and protects against pathogens) or oral tolerance (because it reduces systemic DTH responses and ameliorates the symptoms of experimental autoimmune diseases), appears to differ in well-defined characteristics from responses induced at other sites.

There are several ways in which the intestinal environment might influence an immune response. First, the cells of the gut itself produce TGF-β (20), vasoactive intestinal peptide (which controls the secretory piece of IgA) (21), and other immunologically relevant molecules (22). Second, the luminal side of the intestinal epithelium contains a set of specialized Ag-handling cells called M (microfold) cells (23). These cells overlay intestinal lymphoid follicles and Peyer’s patches (Fig. 1 A), sending down long protrusions that form pockets in which T cells, B cells, and macrophages can be found (24). M cells can take up and transport Ags such as ferritin (25) and pathogens such as Salmonella (26) and HIV (27) by active transport. By their Ag handling, or perhaps by the secretion of specialized cytokines, the M cells might influence the effector class of local immune responses. Third, in contrast to other lymph nodes, B cells far outnumber all other lymphoid cells in the Peyer’s patches (28). They might therefore serve as a specialized set of APCs to induce tolerance or to drive immune responses toward particular effector classes.

To see whether B or M cells were required for the specialized response to oral Ag, we studied the responses of μMT mice, which contain no B cells due to a targeted mutation in the transmembrane region of the IgM heavy chain (29) and no M cells or Peyer’s patches because of the lack of B cells (30). Despite these deficiencies, the μMT mice made perfectly normal systemic T cell responses to oral Ag. Aside from a lack of IgA, their in vivo and in vitro responses were identical to those of wild-type (WT) mice over a large range of Ag doses. Thus, systemic Th cell responses to orally administered soluble Ags require neither the specialized Ag presentation properties of B cells, nor the microenvironment provided by M cells or Peyer’s patches.

When we tested the effects of Ag feeding on the dendritic cells (DCs) of μMT and WT mice, we found that naive T cells responded to these fed APCs by proliferating less than to normal APCs and producing high levels of IL-4 and IL-10. We suggest that the special features of systemic immune responses to orally fed Ag are therefore most likely due to characteristics of the professional APCs (DCs) that capture Ags entering through the gut.

We obtained C57BL/10 (B10 WT), C57BL/10 IgM^−/− μMT (B10 μMT), B10.A IgM^−/− μMT (B10.A μMT), and B10.A (transgenic (Tg)) TCR-Cyt-5CC7 RAG2^−/− (5CC7) mice from Taconic Farms (Germantown, NY). The mice were housed at the National Institute of Allergy and Infectious Diseases facility, which is fully accredited by the American Association of Laboratory Animal Care. In our colony, <0.01% of the lymph node T cells from the 5CC7 mice express high levels of CD44, suggesting that the T cells are naive.

We chose OVA for feeding because it is the Ag most commonly used in oral tolerance experiments and because OVA feeding protocols are well established. We fed mice, using a 20-gauge blunt-ended animal feeding needle, either PBS or low (0.1 mg) or high (10 mg) doses of OVA every other day for five feeds. In the experiment shown in Fig. 7, we added OVA to the drinking water at concentrations of 0.02, 0.2, and 2 mg/ml, resulting in ingestion of 0.1, 1, and 10 mg OVA/day/mouse, respectively (our mice drank ∼5.2 ml/day, calculated from a decrease of 26 ml of water over 24 h in a cage of five mice). Two days after the last OVA feeding (or drinking), we immunized the mice with 50 μg OVA emulsified in CFA in a total of 50 μl, divided equally between the base of the tail and one footpad. For DTH experiments, we immunized only at the base of the tail (Fig. 3,A). When assaying the number of cells in draining nodes, we immunized the mice only in the footpad (Fig. 3,B). In the experiments shown in Fig. 7, we added pigeon cytochrome c (PCC) to the drinking water of B10.A or B10.A μMT mice at a concentration of 0.02 mg/ml and harvested the draining mesenteric lymph nodes (MLNs) on day 4, or injected 0.1 mg PCC into the right footpad and harvested the popliteal lymph nodes (pLNs) on day 4.

We measured footpad thickness 7 days after OVA-CFA immunization using a dial thickness gauge recorder (Popper and Sons, New York, NY).

In mice immunized only at the base of the tail, we injected 50 μg OVA intradermally into the footpad 7 days after the immunization and measured the footpad thickness 48 h later.

To assay Th proliferation, we purified CD4 T cells from the draining lymph nodes of mice. For this purpose, we labeled lymph node cells with anti-B220, anti-CD8, and anti-MHC class II magnetic beads (Miltenyi Biotec, Auburn, CA) for 15 min at 4°C, washed once, passed the cells through Midi-MACS columns (Miltenyi Biotec), and collected the effluent, which consisted of >96% CD4 cells. To prepare 5CC7 responders, we labeled lymph node cells from 5CC7 mice with anti-MHC class II magnetic beads and collected the effluent. More than 97% of the resultant cells were T cells.

We depleted T cells from spleens of B10 (WT) mice using anti-Thy-1.2 (CD90) beads (Miltenyi Biotec) and used the remaining cells as stimulators in proliferation assays. To purify DCs for the experiments in Fig. 7, we labeled the MLN or pLN cells with CD11c magnetic beads, passed them through Midi-MACS columns, and collected the fraction within the column. The final population consisted of >94% CD11c⁺ DCs.

We incubated 2 × 10⁵ responder CD4 cells with 3 × 10⁵ irradiated (1500 rad) T-depleted spleen stimulators in triplicate in 96-well round-bottom microtiter plates, each well containing a total of 200 μl of culture medium (IMDM supplemented with10% FCS, 50 U/ml penicillin, 50 μg/ml streptomycin sulfate, 50 μg/ml gentamicin sulfate, 4 mM glutamine, and 50 μM 2-ME) with or without graded doses of OVA. The cells remained in culture for 4 days and were pulsed with [³H]thymidine for the last 8 h. For the experiments shown in Fig. 7, we incubated 2 × 10⁴ responder 5CC7 T cells with 1 × 10⁵ irradiated (1500 rad) enriched DC stimulators. The cells remained in culture for 3 days and were pulsed with [³H]thymidine for the last 8 h.

We assayed culture supernatants for IL-4 and IFN-γ at 48 h using specific sandwich ELISA (Endogen, Woburn, MA). In the experiments shown in Fig. 7, we measured IL-4 and IFN-γ (Endogen) at 72 h. To measure TGF-β (Promega, Madison, WI), we cultured cells in vitro for 4 days, washed and restimulated them in serum-free medium (QBSF 56; Sigma, St. Louis, MO) with irradiated APCs (T-depleted spleen cells) and Ag, and collected supernatants at 72 h, acid treated them to measure total TGF-β, and used the standard ELISA technique to quantify TGF-β levels. The results are shown in pg/ml. We calculated these amounts using standard curves with known amounts of recombinant cytokines. For intracellular IL-4 staining, we used the anti-IL-4 Ab 11B11 either unlabeled (to block) or conjugated to allophycocyanin, and the protocol provided by PharMingen (San Diego, CA).

We prepared single-cell suspensions of splenocytes from PBS- or PBS/OVA-fed (and otherwise unimmunized) WT and μMT mice and adoptively transferred them into the peritoneal cavities of unirradiated syngeneic mice (1 × 10⁸ WT cells into WT mice and 3 × 10⁷ μMT cells into μMT mice, which equals about one spleen equivalent each). One day after the adoptive transfer, we immunized the recipient mice at the base of the tail and one footpad with CFA-OVA, as described above.

We fixed the entire intestinal tissue from WT and μMT mice in 4% paraformaldehyde and then embedded it in paraffin. Thin sections (6 μm) from these paraffin blocks were than stained with hematoxylin and eosin.

We incubated unstained thin sections with biotinylated Ulex europaeus 1 (UEA 1) for 60 min at room temperature. We then washed the tissue sections in PBS and incubated them for 30 min with the avidin-biotin complex (ABC) linked to HRP (Vectastain Elite ABC kit; Vector Laboratories, Burlingame, CA), prepared according to the manufacturer’s instructions. After a 5-min wash, tissue sections were incubated with 0.05% (w/v) 3,3-diaminobenzidine (Sigma), 0.05% NiCl, and 0.03% H₂O₂, and examined under a light microscope.

Intestinal tissue was fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in PBS (pH 7.3). After washing, the tissue was postfixed in 2% OsO₄, dehydrated in graded series of ethanol, and embedded in Spurr’s resin blocks. From these blocks, 70-nm ultrathin sections were cut. The sections were then mounted on 200-mesh copper grids and stained with uranyl acetate and Reynolds lead citrate. The sections were examined with a JEOL 1200EX at 80 kV.

The two-tailed Student t test was used.

Because orally administered Ag comes into contact with several cell types that are not available to parenterally administered Ag, we asked whether such cells might be involved in setting the unique characteristics of the response to oral Ag. First, a large majority of the cells found in the Peyer’s patches that line the gut are B cells, known to be semiprofessional APCs that do not stimulate, but instead tolerize naive T cells (31, 32, 33). A second of these structures peculiar to the gut are M cells, a specialized set of epithelial cells that are found throughout the intestine that are especially concentrated in the epithelial layer overlying the follicles of Peyer’s patches (named follicle-associated epithelium or FAE), where they can make up to 50% of the epithelial lining, whereas in other parts of the gut, such as villus epithelium, their frequency can be as low as 1% (23). They are easily recognized by a set of features such as short microvilli, basally oriented nuclei, abundant mitochondria, and long cytoplasmic extensions into the lamina propria that can enfold several leukocytes (Fig. 1 A) (34). Ultrastructural studies have shown that M cells contain IgA-immune complexes, suggesting that they may be involved in Ag transport (35). Although the direction of transport cannot be known from these static pictures, M cells are unlikely to secrete IgA into the intestinal lumen because, unlike other intestinal epithelial cells, they do not have receptors for polymeric Ig (36). A hint that the transport of these Ag-Ab complexes might be from the lumen to the lymphoid follicle comes from studies showing that M cells can carry ferritin (24), liposomes (37), and latex beads (38) from the gut lumen and that several kinds of bacteria and viruses (reovirus, HIV) use M cells as entry points into the body from the gut (24).

Using transmission and scanning EM, Golovkina et al. (30) recently showed that M cells are absent in mice that lack B cells. Fig. 1 shows the typical patterns seen by EM in WT and μMT mice. Fig. 1,B shows that the guts of WT mice contain Peyer’s patches covered with smooth FAE, rather than the villus epithelium found in the rest of the gut. In contrast, the μMT mice do not have any Peyer’s patches, but only a few, rare, lymphoid aggregates that do not have significant FAE. In EM photographs of the FAE of WT mice, one can easily detect M cells both in the FAE and within villus epithelium. Fig. 1 C, for example, depicts M cells within the FAE of WT mice in contact with a lymphocyte. Among the aggregates in the μMT mice, however, we could detect no M cells by EM, in agreement with the results reported by Golovkina et al. (30).

Searching for M cells using EM, however, has two problems. First, because each field covers such a small area, it does not give information on the overall frequency of these cells. Second, EM only identifies M cells based on their structural properties and may miss M cells that look structurally very similar to intestinal epithelial cells. To determine whether M cells are involved in the response to oral Ag, we needed to be sure that the μMT mice had no cryptic M cells that might have a different shape due to the lack of interactions with B cells. We therefore used a method to search for M cells, using the lectin U. europaeus 1 (UEA 1) to stain for α-l-fucose, which is expressed on the surface of M cells, but not the epithelial cells of the gut (39). By this technique, we also found no evidence for M cells in the μMT gut (Fig. 1 D).

Thus, it appears that μMT mice lack both M cells and Peyer’s patches, enabling us to ask whether their absence would have any effect on systemic responses to Ags entering through the gut. For this purpose, we undertook an extensive series of in vitro and in vivo experiments to compare the effects of OVA feeding in WT and μMT mice, focusing on the well-documented characteristics of the systemic T cell response to oral Ag.

Because a decrease in systemic Th cell proliferation is a standard characteristic of the response to immunization with a previously fed Ag (40), we started our analysis by examining Th cell proliferation after feeding Ag. To take into account potential differences in Ag handling that might occur in mice with and without B cells, M cells, and Peyer’s patches, we tested a dose titration in vivo, feeding amounts of OVA that differed over a 10,000-fold range (from 0.001 to 10 mg), every other day for a total of five feeds, immunized the mice 2 days after the last feeding with 50 μg of OVA in CFA, and harvested the draining lymph nodes 7 days later to test for OVA responsiveness in vitro. Surprisingly, the WT and μMT mice responded in the same way, as assayed both by proliferation of CD4 cells (Fig. 2,A) and by cytokine production (Fig. 2,B). As we increased the dose of fed OVA, the in vitro Th cell proliferation to OVA decreased in a graded manner; the higher the dose, the less the proliferation. We found similar results for IFN-γ and opposite results for IL-4, in which feeding resulted in a dose-dependent increase in IL-4 levels in both WT and μMT mice (Fig. 2 B).

From the data above, we concluded that the systemic effects of oral immunization, as measured by in vitro tests, were induced with equal efficacy in the presence or absence of B cells, M cells, and Peyer’s patches. We next asked whether we could find a difference using in vivo functions.

We first looked at DTH responses in mice given the same feeding and immunization protocol as described in Fig. 2, except that they were immunized only at the base of the tail. Seven days after immunization, we tested DTH responses by injecting OVA in PBS intradermally into the footpad and measuring the footpad thickness 48 h later. Fig. 3 A, representing a summary of two different experiments, shows that both WT and μMT mice responded to feeding by producing smaller DTH responses than PBS-fed controls. In this assay, as in the in vitro tests, we found no significant difference between the two types of mice.

Because a decreased DTH response may reflect decreased cellular infiltration and/or proliferation in vivo, we asked whether we could see the same picture in the lymph nodes draining the immunization site. We fed WT and μMT mice either PBS or low (0.1 mg) or high (10 mg) dose OVA, immunized at the footpad, and counted the number of cells in the pLNs 7 days after immunization. Fig. 3 B shows that both WT and μMT mice that had been fed OVA responded to OVA-CFA immunization with a decreased cellular infiltration of the draining pLNs.

Having found that WT and μMT mice showed no differences in their direct responses to fed Ag, we next asked about their ability to mediate the bystander effects that have classically been associated with oral tolerance. There are two types of situations in which fed cells have been shown to have an influence on other T cells. First, in adoptive transfer systems, T cells from fed animals can have an Ag-specific suppressive effect on the DTH, proliferation (10), and cytokine (40) responses of naive host T cells. Second, in the original host, orally immunized T cells can affect the naive systemic T cell response against new Ags, as long as the immunizing mixture also contains the fed Ag (10). We compared both of these bystander effects in WT and μMT mice.

Bystander effect on naive T cells against the same Ag.Earlier studies show that the systemic effect of fed T cells on naive T cells is dose dependent. It is mediated only by T cells in animals that have been fed low doses of Ag. High doses induce deletion and abrogate both the direct response and the bystander suppression to the fed Ag (5, 6). To compare the Ag-specific bystander effect of WT and μMT mice, we therefore harvested spleen cells from low (0.1 mg) and high (10 mg) dose OVA-fed (five times every other day) animals and transferred them to the peritoneal cavities of unirradiated syngeneic hosts (WT spleen cells into WT hosts and μMT spleen cells into μMT hosts). One day after the transfer, we immunized the recipients with OVA in CFA and assayed for in vitro Th cell proliferation 7 days later. We found that the fed WT and μMT spleen cells behaved identically and characteristically. Spleen cells from low dose fed animals had a strong effect on their adoptive hosts, whereas spleen cells from high-dose fed mice were only minimally suppressive. Fig. 4 shows that the effects were evident in both the proliferation and cytokine assays after injection of OVA in CFA. Spleen cells from low-dose fed animals induced approximately a 100-fold decrease in the titrated proliferative response to Ag, a 10- to 20-fold decrease in IFN-γ production and an 8- to 10-fold increase in IL-4. In contrast, spleen cells from animals that were fed high doses of OVA (10 mg) had only a slight effect when compared with spleen cells from PBS or low-dose fed donors. Here again, transfer of low-dose fed cells was the most effective, and μMT mice behaved similarly to WT.

Bystander effect on naive T cells against new Ags that are given concomitantly with the fed Ag. To compare the bystander effect against new Ags, we looked at footpad swelling 7 days after immunizing with OVA in CFA, asking whether a shift in response to a fed Ag can affect the strong primary response normally seen to whole mycobacteria emulsified in oil. Fig. 5,A shows that the injected feet of naive mice fed only PBS swelled to about twice their size after injection of PBS-CFA and that prior OVA feeding had no effect. In contrast, prior OVA feeding markedly reduced the response if the OVA was also given along with the CFA injection. This bystander effect could also be adoptively transferred. Splenocytes from mice fed OVA, when transferred into naive mice, were able to suppress footpad swelling induced by immunization with OVA-CFA (Fig. 5 B). The effect waned, as would be expected because of the deletional effects, at higher doses of OVA. In both of these types of bystander assays in vivo, as we had previously seen in all other tests, the μMT mice behaved identically to the WT mice.

A possible complication resulting from Ag administration using a feeding needle is accidental abrasion to the esophagus, which may lead to spillage of Ag into the blood. There have been reports that i.v. administration of Ag can lead to responses with characteristics similar to those seen with low-dose oral administration, such as a decrease in DTH (41, 42), proliferation, and cytokine production (43) concomitant with an increase in bystander suppression (39) and, occasionally, production of secretory IgA (42). It has also been reported that a large bolus of fed OVA can lead to low levels of blood-borne OVA (44). To determine whether the T cell response and the bystander suppression we had seen after feeding an Ag were due to abrasion-induced systemic spill, we fed titrated amounts of OVA to both WT and μMT mice by adding it into their drinking water and then compared their footpad-swelling responses with those of animals fed using a feeding needle. Fig. 6,A shows that both groups responded the same way, and that there was little difference between WT and μMT mice. Furthermore, both WT and μMT Th cells from mice that drank OVA made high levels of TGF-β and IL-4 in vitro compared with unfed controls (Fig. 6 B).

The finding that there were no discernible differences between WT and μMT mice suggested that the characteristic properties of the systemic response to fed Ag might be established by an APC common to both strains, e.g., the DCs that drain from the gut to the mesenteric nodes. We therefore fed PCC to B10.A WT and μMT mice by adding it into their drinking water, harvested the MLNs on day 4, and purified the CD11c⁺ DCs by positive selection using magnetic columns. We then irradiated these DCs and used them as APCs in vitro for a monoclonal population of naive TCR Tg T cells specific for PCC (5CC7 TCR Tg bred onto a B10.A RAG2^−/− background) and compared them to APCs from the MLNs of unfed mice or from pLNs from mice immunized in the footpads to PCC or from control mice. Fig. 7,A shows that MLN DCs from the fed mice induced the naive T cells to make higher levels of IL-4 and IL-10 and less IFN-γ than did APCs from control mice or APCs draining a peripheral site injected with PCC. In contrast, APCs from the pLNs of mice injected at the footpad with PCC induced strong IFN-γ production, low level IL-10 production, and no IL-4. To confirm that the IL-4 measured by ELISA was being made by the naive Tg T cells rather than the DCs themselves, we stained the in vitro cultured T cells for intracellular IL-4. Fig. 7 B shows that, after 3 days in culture, when stimulated with fed DC and Ag, the TCR Vα11.1-positive Th against PCC are themselves staining for intracellular IL-4. This is a striking result, as it demonstrates for the first time that naive T cells can make large quantities of IL-4 in vitro, without a requirement for restimulation (45), if they receive appropriate stimuli (e.g., DCs from the MLNs of Ag-fed mice).

This study demonstrates two aspects of immunity to Ags that enter through the gut. The first relates to the systemic immune response, showing that the absence of GALT and the consequent inability to coordinate the local gut-oriented immune response have no effect on the characteristic systemic effects of orally administered Ag. The second relates to the APCs involved in gut-oriented immunity and shows that APCs from the MLNs of Ag-fed mice are able to elicit copious amounts of IL-4 from naive T cells, a surprising finding in light of the fact that naive T cells are thought to be able to produce only IL-2 and not IL-4. We will discuss both of these aspects separately.

The immune response to fed Ags has both local and systemic components. To determine whether the GALT is involved in the systemic response, we analyzed the response of μMT mice to orally administered OVA. We had previously reported that parenterally immunized systemic T cell responses seem to be normal in μMT mice, showing that neither B cells nor M cells are necessary for in vitro cytokine production or proliferation, the in vivo production of granulomas against parasite eggs, the rejection of skin grafts, priming for CTL function, or the maintenance of CTL memory (46, 47, 48). In effect, we found that μMT T cells respond normally to several different types of Ags, given by several different routes. We had not, however, tested responses to oral Ag, and it could be argued that the lack of B cells, M cells, and Peyer’s patches might have a serious effect on responses to Ags that enter through the gut. We therefore undertook an extensive series of tests to determine whether the lack of B cells, M cells, and Peyer’s patches would have any effect on the systemic T cell response to fed Ags. We found that all of the well-described features of oral tolerance were intact. Whether tested in DTH or 7-day footpad-swelling tests in vivo or by proliferation and cytokine production in vitro, over a 10,000-fold dose range, the T cells from fed WT and μMT mice behaved identically, both in their own responses to the fed Ag and in their ability to influence the responses of naive T cells to the fed Ag and to bystander Ags. From these data, we conclude that neither the lack of B cells nor the absence of M cells or Peyer’s patches had any measurable effect on the systemic T cell response to fed Ag.

Because feeding by gavage may cause damage to the esophagus, and because the systemic effects of oral tolerance are often similar to those generated by i.v. administration of Ag, we considered the possibility that systemic blood-borne Ag, rather than the GALT response, was the basis of the feeding effect. The lack of an organized GALT would then be expected to have little influence. We found, however, that the immune response to fed Ag is the same whether the Ag is fed by gavage or delivered in the water, suggesting that Ag delivery without trauma, and thus without major overt spillage into the blood, also results in the same effect. In fact, we found (Fig. 6) that Ag delivered daily through the drinking water seemed to be slightly more effective than Ag given every 2 days by gavage. As the final doses were the same, this difference may be due to the increased frequency with which Ag was delivered. One might even envisage the contrasting scenario that the effect of i.v. administration of Ag is actually due to seepage of Ag into the gut environment and activation of a gut-oriented immune response. In either case, however, whether the skewing of the response is due to leakage out of the gut or into it, it appears that neither B cells, M cells, or Peyer’s patches are necessary.

These data are in line with the earlier studies showing that B cells are semiprofessional (49), rather than professional APCs (31), that are able to stimulate memory, but not naive T cells (32, 33, 50, 51). They also eliminate one role for M cells, leaving open the possibility that, contrary to all of the structural evidence (52), M cells may have no role in initiating T cell immunity to soluble Ags. M cells and Peyers patches may only be involved in local gut responses to particulate Ags and microorganisms, leaving other cells to deal with the systemic response to fed Ag.

If B cells, M cells, and Peyer’s patches are not necessary for the systemic response to Ags entering through the gut, then what initiates the response?

Gut epithelial cells are unlikely candidates. Although these express MHC class II constitutively (53), they are not professional APCs (54) and should therefore induce deletion rather than activation of T cells. They are thus apt to be involved in the Ag-specific deletion of T cells that occurs after high-dose administration of oral Ag. This leaves us with macrophages and DCs. The observation that oral tolerance is enhanced by in vivo treatment with Flt3 ligand (which expands the number of DCs) (55) supports the view that, as with other responses, the immune response to oral Ags is most likely initiated by DCs.

But the initiation of the response is not the whole story. Gut-oriented immune responses are characterized by a particular class of effector function such as IgA (56, 57), TGF-β (4), IL-4, and IL-10 (58), which are known alternatively as oral tolerance (4) or immune deviation (8). What gives rise to these qualities?

We would like first to suggest that the gut-oriented immune response is neither tolerance nor deviation. It is simply an immune response tailored to the environment of the gut, which has specific immunological needs (the production of IgA) as well as certain sensitivities (it can be harmed by a DTH response (59, 60, 61, 62)). In these respects, it is similar to the eye (63). To protect itself and to ensure that a local immune response is nevertheless effective, the gut suppresses DTH responses (61) and enhances IgA production (20, 21). It appears that it may not do this through specialized Ag handling, nor through cytokines produced by B or M cells, but via the conditioning (or education) of DCs.

It has recently been shown that DCs from naive Peyer’s patches induce T cells to make IL-6, IL-10, and IL-4, but only very low levels of IFN-γ in vitro (64, 65), suggesting that the effector class of the local Peyer’s patch response may be controlled by the resident DCs. Our data show that the systemic T cell response to fed Ag might also be influenced by DCs, but regulated at a different lymph node rather than the Peyer’s patches. We found that DCs from the MLNs of fed animals had the startling capacity to elicit IL-4 production from naive T cells in primary cultures. They also induced high secondary levels of TGF-β, but little IFN-γ, whereas DCs from the MLNs of mice that were not fed Ag did not elicit these responses. The fed DCs were equally active at inducing this Th2/Th3 response regardless of whether they were derived from WT animals or from those lacking B cells, M cells, and Peyer’s patches. Thus, DCs draining the gut of Ag-fed animals seem to carry class-specific instructions with them regardless of the existence of an organized GALT microenvironment.

We do not know whether the MLN DCs are a separate population or whether they are influenced by local signals as they traffic through the gut. It has been shown in other systems that the effector function of DCs can be altered by interactions with Th cells (66, 67, 68) or by signals, such as PGE₂, which is produced by many different tissues (69). We think therefore that the most likely scenario is that gut-oriented immune responses, like those occurring elsewhere, are initiated by activated local professional APCs such as DCs, and that resident T cells in the gut, or the cells of the gut itself, set the effector class by educating the DCs, as well as by producing such immunomodulatory cytokines as TGF-β and vasoactive intestinal peptide. In this way, a tissue may control several aspects of a response. It initiates the response by sending danger signals to activate local APCs (70, 71), and it also influences the class of response to enhance efficacy and to ensure its own safety (63, 72).

We are indebted to T. Kamala for her ideas, suggestions, critiques, and help throughout the study and the preparation of the manuscript. We thank Ron Schwartz, Albert Bendelac, Bana Jabri, Warren Strober, and David Volkman for reading and improving the manuscript, and Ron Germain for suggesting the “drinking” experiment. We also thank Zeynep Melis Alpan for all of her inspirations.