Intestinal epithelial cell
Delayed type hypersensitivity
The development of immunological tolerance to orally fed antigens depends on the sampling, processing and transportation events followed in the intestinal epithelium. We present here a description of a ''tolerosome'': a supra-molecular, exosome-like structure assembled in and released from the small intestinal epithelial cell. The tolerosome is a ≈ 40 nm large vesicular structure that carries MHC class II (MHC II) with bound antigenic peptides sampled from the gut lumen. Tolerosomes isolated from serum shortly after antigen feeding or from an in vitro pulsed intestinal epithelial cell line are fully capable of inducing antigen specific tolerance in naive recipient animals. Purified tolerosomes represent a structure by which fed antigens can be efficiently presented to the immune system. Removal of the tolerosomes from serum by ultracentrifugation or absorption of MHC II results in abrogated tolerance development.
Oral administration of soluble protein antigens gives rise to a profound suppression of the subsequent immune responsiveness to the fed antigen. The crucial role for CD4+ T cells in the induction and maintenance of peripheral tolerance to non-self antigens emanating from the intestinal content of microbial and dietary antigens is now well established 1–5. The suppression is mediated by antigen specific CD4+ T cells and their production of the suppressive cytokine TGF-ß 1, 6–8.The true nature of the uptake and processing of luminal antigens giving rise to these regulatory T cells is largely unknown. It has previously been reported that gut processing of the antigen 9, 10 and portal drainage through the liver 11, 12 is a prerequisite for establishing this form of tolerance. Previous studies have described the involvement of a serum factor, obtained shortly after feeding an antigen, that can induce tolerance in recipient animals 10. However, no satisfactory definition of its biochemical characteristics was put forward 9. We have recently confirmed these data in our rat model and additionally shown that the serum factor induce an active antigen specific tolerance with suppression of the response to a bystander antigen and the involvement of CD25+ lymphocytes 13.
It has been shown that SCID mice lack the ability to produce the tolerogenic factor 14. Interestingly, the SCID mouse lack MHC class II (MHC II) expression in intestinal epithelial cells 15 but expression can be induced by treatment with IFN-γ which enables the SCID mouse to produce the tolerogenic serum factor (our unpublished data). Furthermore, there is a need for enzymatic processing for the tolerogen to be formed 16–18. It is therefore feasible that MHC II expression of epithelial cells and luminal processing of the antigen is required for the production of the tolerogenic factor. This is also supported by previous findings, that the ability to induce active suppression after antigen feeding correlates with the expression of MHC II in intestinal epithelial cells which takes place at about 5–6 weeks of age 19, 20.
Since CD 4+ T cells require the antigens to be presented on MHC II and since indirect evidence indicate that expression of MHC II in the intestinal epithelial cells (IEC) is a prerequisite for the development of active suppression to orally administered antigens 19–22 we examined the MHC II compartment constitutively present in the IEC in normal rodents and in man. The abundance and morphological appearance of this putative antigen sampling and processing site prompted us to investigate the possibility of an MHC II bearing vesicular structure as a vehicle for soluble intestinal protein antigens.
As shown in the autoradiographic trace of 125I-labeled ovalbumin obtained 1 h after feeding the antigen was rapidly pinocytosed from the intestinal lumen and co-localized with MHC II in an apical and baso-lateral compartment of the IEC (Fig. 1b). A careful histological examination of this compartment reveals that the vesicularity of the MHC II compartment is obligate and apparently released small MHC II-stained vesicles are present basolaterally of and in the space between the epithelial base and the dendritic cells (DC) in the lamina propria (Fig. 1a). In humans the epithelial MHC II compartment is also rich in HLA-DM (Fig. 1d), a molecule necessary for effective peptide loading onto the MHC II molecule 23. The presence invariant chain, CD68 and LAMP1 with an apical staining pattern identical to that of HLA-DM (not shown) within these vesicle structures supports the notion of a complete antigen processing compartment within the IEC.
Serum from rats taken 2 h after an ad libitum feed of an OVA-containing diet, readily transfers OVA-specific tolerance to naive recipients with a significant reduction of the DTH response to OVA (p<0.01) as compared with rats fed a control diet (Fig. 2). When an equivalent volume of this serum was subjected to preparative ultracentrifugation a fraction sedimenting at 70,000×g was obtained that was equally effective in inducing tolerance in recipient animals (p<0.01) (Fig. 2). Furthermore, the supernatant from the serum ultracentrifuged twice had now lost its capacity to induce tolerance (Fig. 3a), which implied that all the tolerogenic activity in the serum from the fed donor rats could be retrieved in a supra-molecular fraction that sediments at 70,000×g. Indeed, this fraction mediated significant suppression of both DTH reactivity (p<0.01) and antibody response (p<0.01) (Fig. 4) to the antigen fed to the donor rat. Dot blot analysis of this fraction revealed that it was rich in MHC class II (Fig. 5), which led us to examine whether the MHC II content was crucial for its tolerogenic activity. Removing MHC II from the 70,000×g fraction by the use of paramagnetic beads (MACS) and a monoclonal antibody to rat MHC II totally eliminated its capacity to induce tolerance in the recipients as compared to an isotype control treated 70,000×g fraction (Fig. 3b).
The staining of MHC II in cultured IEC-17 cells revealed that these cells do not express MHC II molecules, unless they are stimulated with IFN-γ (Fig. 6). The MHC II molecules were, as in the normal epithelial cells, located exclusively in intracellular vesicles (Fig. 1a and 6). As Fig. 7 shows, the group of rats that received the 70,000×g preparation from the cells cultured in the presence of IFN-γ and peptides from pre-digested OVA actually induced a substantial down-regulation of the DTH reaction to OVA (40% reduction, p =0.04), while the response to HSA was unaffected. This demonstrates that cultured IEC-17 cells can indeed produce a tolerogenic supra-molecular structure. Transmission electron microscopy of this preparation revealed that it consisted of small vesicles with a size of ≈40 nm (Fig. 8), that stained positive for MHC II (Fig. 5d).
Previous reports have revealed that soluble protein antigens are rapidly pinocytosed from the intestinal lumen and co-localize with MHC II in a vesicular compartment of the IEC 24, 25. It was originally proposed that the epithelial cell itself presented the sampled antigens on its surface to a local T cell population 26. This view is hampered by the lack of surface MHC II staining of the IEC as shown in the present study and the almost complete absence of CD4+ T cells in its vicinity (our unpublished observations). It is, however, compatible with a released MHC II-bearing vesicular structure that mediate peptide presentation and disseminate from the epithelial cell. From the histological data obtained by light microscopy presented in this report it can only by concluded that the MHC II compartment of the IEC is vesicular and present apically and basolaterally and can not distinguish released from unreleased material. However, earlier electron-microscopy studies of rat small intestine elegantly performed by Mayerhofer and Spargo clearly show MHC II-stained vesicular structures released basolaterally between IEC 27, 28. In the present report we provide evidence for the generation of such structure by the IEC both in vitro and in vivo. The sedimentation characteristics, size and MHC II content resemble these of the previously described exosome produced by B cells and DC 29–31. Due to these characteristics and its tolerogenic properties we will refer to the structure presently described as a tolerosome.
We have clearly shown that all the tolerogenic activity found in serum shortly after an antigen is contained in a fraction sedimenting at 70,000×g. This fraction contains cell organelles and large aggregates >1,000 kDa of protein molecules. Theoretically this fraction could contain large aggregates of OVA but we have not been able to detect any OVA in this fraction using a sensitive ELISA (not shown). In addition we and others have previously described that aggregated OVA is not tolerogenic and rather induce a protective immune response when injected intravenously 9, 13. It has also been shown that tolerance is inversely related to the serum levels of OVA in fed animals 16 and that the tolerogenic activity of the serum transferred to naive recipients is unrelated to its content of native OVA 10.
Interestingly, lack of MHC II in the IEC due to immaturity 19, 20, 32 or in the MHC II knockout mouse 21, results in failure to produce regulatory T cells after oral administration of the antigen. In addition, the serum factor described here is not produced in SCID mice 14 lacking MHC II in their IEC 15 and this can be reversed by treatment of the donor SCID with IFN-γ that induces MHC II in the IEC and allows the SCID mouse to produce the transferable tolerogenic serum factor i.e. tolerosomes (to be published, Evertsson, S., Taube, M., E.Telemo.).
Given these characteristics of the serum factor and its capacity to profoundly influence the immune responsiveness we propose a direct connection to the IEC and its MHC II content.
The intracellular MHC II compartment of the IEC in situ shows great resemblance to the antigen loading compartment (MIIC) of DC comprising a high constitutive MHC II expression (Fig. 1a–c ), multivesicular compartments 27 and antigen loading machinery which includes HLA-DM (Fig. 1c), and co-localization of invariant chain, CD68 and LAMP1 along with lysosomal proteases 33. It should be stressed that all these components of the antigen processing compartment are all present in the IEC under normal non-inflammatory conditions which suggest a continuos handling of antigenic peptides taken up from the lumen. In vivo this machinery is fulled by IFN-γ mainly produced by the intraepithelial lymphocytes (IEL) 34 in response to bacterial flora-derived stimulatory molecules 35, 36. Thus MHC II is absent in the IEC of IFN-γ and IFN-γ receptor knockout mice, in SCID mice lacking IEL, and is poorly developed in germ-free rodents (our own unpublished observations). Accordingly, we here show that the rat duodenal epithelial cell line IEC17 develops a prominent vesicular MHC II expression when cultured with IFN-γ and is capable, when grown in the presence of OVA peptides, of producing tolerosomes that sediments at 70,000×g, contains MHC II and induce tolerance to OVA in vivo. The preparation obtained from the cells cultured with native OVA failed to affect the immune response to OVA, which suggests that to be able to produce the tolerogen, the epithelial cells need extracellular pre-processing of the antigen. Here we used digestion by pepsin and trypsin, to try to generate peptides similar to the ones found in the intestinal lumen. The limited processing capacity of native protein antigens by IEC has previously also been shown by others 37. In addition, lack of luminal processing of protein antigens due to the presence of protease inhibitors has been shown to abrogate tolerance development in vivo16.
These results show that intestinal epithelial cells are, at least in vitro, capable of producing tolerosomes with similar characteristics and biological activity as those isolated in vivo from serum shortly after an antigen feed. The apparent differences in functional activity i.e. protection vs. tolerance between the previously described exosomes derived from DC and the tolerosomes isolated from IEC could be due to content of second messages and/or different homing patterns. The exosomes used in vaccination studies 30 were derived from in vitro propagated and activated DC and coexpressed the co-stimulatory molecules B7.1 and 2 and also contained heat shock protein along with several potential homing structures like MFG-E8, ICAM-1 and CD63 31. The existence of these structures on tolerosomes needs to be determined, however, histological examination of human small intestinal IEC has revealed basolateral staining of LAMP 1, CD63 and CD68 but not B7 or ICAM 1 (to be published). This indicates differences in the composition between DC-derived exosomes and tolerosomes which may be crucial in the resulting immune response.
The fate of the tolerosomes produced by the IEC in vivo hence remains to be investigated. It is, however, likely that the liver is important since it is the major draining site of the intestinal circulation and contains APC that effectively clears particulate matter of similar size as the tolerosomes (≈40 nm) from the blood 38. In addition, our preliminary results shows that blood drawn from the portal vein contained significantly more MHC II stained material than blood that had passed through the liver (to be published). This notion is also supported by the role of the liver in tolerance development to orally administered antigens 11, 12, 39 and its ability to prolong graft survival in transplantation 40, 41.
We hypothesize the presence of a tolerance inducing system where under normal non-inflamed conditions antigens are sampled from the small intestinal content, assembled with MHC II and loaded onto tolerosomes in the epithelial cell. The basolaterally released tolerosomes enter the circulation from where they could be endocytosed by liver sinusoidal DC which then migrate to the celiac lymph nodes 38. Alternatively they could be taken up by the sinusoidal endothelial cells that have been shown to be effective APC. Under non-inflammatory conditions i.e. low endotoxin levels these cells are conditioned by the hepatic microenvironment to produce IL-10 and express low levels of accessory molecules like B7 42 providing circumstances that would facilitate the development of regulatory T cells in response to the presented antigens 4, 43. It is interesting to note that the commensal bacterial flora and the presence of bacterial components in the small intestine is intimately involved in this tolerance system in at least two ways. One is by facilitating sampling and processing of luminal antigens and MHC II expression of the IEC 44 via induction of IFN-γ production 34–36, the other is by providing a normal endotoxin load to the liver aiding the proper conditioning of the liver APC 42, 45. Accordingly, it has been shown that bacterial lipopolysaccharide (LPS) facilitates the development of oral tolerance in various experimental systems 46–48. The present report provides a basis for the understanding of the regulatory events observed in response to "harmless" orally encounteredenvironmental antigens aiming to avoid pathological inflammatory reactions in the intestine and highlights the implications suggesting a role for a well developed commensal flora for these events to occur.
4 Materials and methods
4.1 Treatment of donor animals
Male 12–14-week-old Wistar rats (BK universal, Sweden) were starved overnight. One group was fed OVA-containing diet (R380 AnalyCen, Sweden) for 2 h (estimated OVA intake was 350 mg/rat) and the control group was fed standard OVA-free diet. After 2 h ad libitum feeding the blood was collected by heart puncture. The obtained sera were pooled in one OVA-fed pool and one control-fed pool. The average yield from each donor rat was 3.5 ml of serum that was the amount injected into the recipients.
All animal experiments were performed with the approval from Swedish ethics committee for animal experiments; Permission Nr: Dnr 321/98.
4.2 Cell lines and preparation of OVA peptides
CHO-211A cells were kindly provided by Dr. A. Tarkowski and the rat small intestinal epithelial cell line IEC-17 was kindly provided by Dr. N. Lycke. The cell lines were grown in ISCOVES medium supplemented with 10% fetal calf serum, 1% penicillin-streptomycin, 1% Na-pyruvate and 1% L-glutamine.
Ovalbumin (OVA, Grade V; Sigma Chemical Co., St. Louis, MO), diluted to 20 mg/ml in 0.1 M citrate buffer (pH 2.2) was enzymatically digested with 1,300 u/ml pepsin A (No P-6887; Sigma) for 17 h at 37°C, followed by neutralization with 1 M Tris(hydroxymethyl)-aminomethane and further digestion by 550 u/ml trypsin (crude type II, No T-8128; Sigma) for 24 h at 37°C. Finally, trypsin-chymotrypsin inhibitor (T-9777; Sigma) was added to reach 45 μg/ml. The peptide soup was then purified on a Sephacryl S-100 (Pharmacia, Uppsala, Sweden) column, and the fractions containing fragments smaller than 3 kDa (peak absorbance at 1.3 kDa) were collected, pooled and lyophilized.
4.3 Preparative centrifugation
The blood was allowed to clot for 2 h at RT and the serum was collected after an initial centrifugation at 3,000×g for 15 min followed by a 10,000×g centrifugation for 30 min to remove cell debris and clotted material. The different serum pools were then ultracentrifuged either once or in a separate experiment (see below) twice at 70,000×g for 1 h (ultracentrifuge tubes; Quick-Seal, Beckman, 70 Ti rotor; Beckman). The pellets were resuspended in a volume of PBS corresponding to the initial volume of serum. All serum fractions were injected i.p. into groups of 8-week-old male Wistar furth recipients. Each rat received 3.5 ml that corresponds to the average yield from one donor rat. For the in vitro preparation IEC-17 cells were grown in 150 ml tissue culture flasks, in the presence of 100 u/ml IFN-γ and 0.5 mg/ml native OVA, or IFN-γ and 0.01 mg/ml OVA-peptides, or IFN-γ only. After 5 days of culture when the cell number had reached 5×107, the cell suspension was collected and sequentially centrifuged for 10 min at 1,000×g, followed by 30 min at 20,000×g (70 Ti rotor; Beckman Instruments, Palo Alto, CA),and 60 min at 70,000×g (50 Ti rotor; Beckman). The 70,000×g pellet was resuspended in 10 ml 20 mM Hepes buffer, once again centrifuged for 60 min at 70,000×g, and finally resuspendedin 1 ml 20 mM Hepes buffer. Before transfer into 8-week-old inbred Wistar furth recipient rats, the exosome suspensions were diluted to 8 ml with PBS.
4.4 Immunization of recipients
In the serum transfer experiment the immunization was performed as previously described 5, 8. Briefly, 100 μg of OVA (grade V, Sigma) mixed in complete Freund's adjuvant (DIFCO Labs, Detroit, MI) was injected subcutaneously in the hind legs 1 week after serum transfer. Blood samples for antibody detection were collected from the tip of the tail at the time of immunization and at 1 and 2 weeks thereafter. Recipients of the in vitro preparation were randomly divided into four treatment groups (n=6 rats/group), that received either PBS only, or exosome suspensions from IEC-17 cells grown in the presence of either native OVA and IFN-γ, or OVA-peptides and IFN-γ, or IFN-γ only. All injections were made intraperitoneally, in 1 ml. One week after the injection all rats immunized subcutaneously in the hind legs with 50 μg OVA mixed with 50 μg human serum albumin (HSA, Fraction V; Sigma) as bystander antigen in Freund's complete adjuvant.
4.5 Delayed type hypersensitivity reaction (DTH)
In the serum transfer experiment the DTH reaction was performed as described previously 5. Briefly, rats were intracutaneously challenged 2 weeks after the immunization with50 μg OVA (grade V, Sigma) in 20 μl PBS in each ear or in the in vitro preparation experiments by injecting 50 μg of OVA in 20 μl of PBS subcutaneosly in one ear and the same amount of HSA in the other ear. The ear thickness was measured in a blinded fashion before and 24 h after injection, using a micrometer caliper (Oditest, Kröplin, Germany). The DTH-reaction was expressed as the increase in ear thickness.
4.6 Specific IgG anti-OVA antibody determination in serum
IgG anti-OVA antibodies in serum taken week 1 and 2 after the immunization were measured by an enzyme linked immumosorbent assay (ELISA) as previously described 5. Briefly, PVC microtiter plates were coated with OVA and the serum samples or a standard hyperimmune serum were serially diluted in PBS-Tween in the plates. The plates were incubated with rabbit anti-rat IgG(Fc) (Zymed) antibodies for 2 h followed by an alkaline phosphatase conjugated goat anti-rabbit IgG (Sigma). The enzyme activity was visualized by the substrate p-nitrophenyl phosphate and the absorbance was measured in a spectrophotometer at 405 nm. All incubations were done at room temperature. The antibody levels were expressed in arbitrary units calculated from the standard curve obtained from the standard hyperimmune serum added to each plate.
Proximal small intestinal samples from normal adult Wistar rats or jejunal biopsies from healthy adult human volunteers were rapidly embedded in OCT-medium (Sakura Finetechnical, Japan) and frozen in isopentane (2-metyl-propane), pre-cooled in liquid nitrogen, and finally transferred into liquid nitrogen. The tissue blocks were stored at –70°C prior to cutting. Sections (5-μm) were fixed in cooled acetone (+4°C; 30 s in 50% acetone followed by 5 min in 100% acetone) and then air-dried for 5 min and washed three times in PBS-Tween (0.05 % Tween; 3×5 min). The endogenous peroxidase activity was blocked by incubation in glucose oxidase 8. The IEC 17 cells were grown in chamber slides with or without IFN-γ, air-dried and then fixed as above and to permeabilize the cells 0.5 % Saponin was added to the buffers.
The slides of human tissue were incubated overnight at 4°C with the following mouse monoclonal antibodies to human markers: HLA-DR 8 (TAL 15B, Dako), HLA-DM (MapDM, Pharmingen), CD68 (KP1, Dako), LAMP1 (CD107a, KP1, Southern Biotech) and for rat tissue the anti-MHC II ( MRC OX6, OX3 and OX17, the clones were obtained from European collection of Animal cell Cultures, Salisbury, GB) diluted in PBS-Tween. Control slides were incubated with isotype matched control antibodies (Pharmingen). After washing, the sections were incubated for 1 h with a biotinylated affinity-purified rabbit-anti-mouse-IgG antibody diluted (rat tissue: STAR44, Serotec; human tissue: E0413, DAKO), followed by avidin-conjugated peroxidase (ABC-complex, DAKO). The peroxide staining was revealed with amino-ethyl-carbazole (Sigma), followed by a light counter staining with Mayer's hematoxylin, and mounting in Mount-Quick "Aqueous" (Daido Sangyolo, Japan). The microscopical examination was performed using a Leica DMB microscope.
OVA (20 μg; grade V, Sigma) was labeled with 1.0 mCi 125I (Amersham) using the Chloramine-T method and purified by gel filtration (PD10, Sephadex G25, Pharmacia). After an overnight fast 10 μg of 125I-labeled OVA in 1 ml PBS with 5% BSA was fed to a 7-week-old Wistar rat. One hour later the rat was sacrificed and samples of the small intestine were obtained andcut and stained as described above. The slides were then dipped in photographic emulsion (NTB-II, Kodak) incubated in the dark for 20 days and then developed.
4.9 Absorption of MHC class II
In a separate experiment two batches of 14 ml serum from OVA fed donor rats were centrifuged twice 70,000×g for 1 h and the pellet was resuspended in 1 ml PBS pellet and either 1 μg mouse anti-rat MHC class II (OX-6) or 1 μg of isotype control antibody (mouse IgG1 anti TNP, Pharmingen) was added. A second incubation was done with FITC-conjugated goat anti-mouse IgG antibodies (Star 70, Serotec) and 1% rat serum followed by 100 μl anti-FITC coated paramagnetic beads (Mini Macs, Miltenyi Biotec). All incubations were done on ice for 20 min and the pellets were washedonce and ultracentrifuged at 70,000×g between each step. The preparation was then added to Mini Macs separation columns (Miltenyi, Biotech). The fraction not retained in the columns was adjusted to the original volume with PBS and injected into four recipient rats giving the same donor recipient ratio of 1:1 as in the other experiments.
4.10 Transmission electron microscopy (TEM)
The resuspended 70,000×g pellets from the in vitro preparation was loaded onto EM-grids. After loading, the grids were fixed for 30 min in 2% paraformaldehyde and washed 3×1 min in distilled water. Thereafter, they were contrast stained with 2% uranylacetate, air-dried, and subjected to TEM using a JEOL 100CX microscope.
4.11 Dot blot
A 20-μl sample was bound to a nitro-cellulose membrane (Bio-Rad), using a Bio-blot apparatus (Bio-Rad). The membrane was washed once with TBS buffer, and blocked with 2% BSA, over night at4°C. In all subsequent steps, 0.5% BSA was added to the buffers. The presence of MHC II in the sample was detected with a biotinylated mouse anti-MHC class II antibody [OX6, (Serotec), 1 μg/ml,biotinylated isotype-matched anti-TNP 1 μg/ml, (Pharmingen) was used as control] added for 2 h at 20°C, followed by a streptavidin peroxidase complex (ABC, DACO), for 1 h. Between every step, the membranes were washed for 3×10 min with TBS buffer.
Finally, the membrane was transferred to the dark, dipped in chemoluminescent ECL reagent (Amersham) and overlayed with photographic film (Kodak Biomax) that was exposed for 5 min and subsequently developed.
4.12 Data analysis
Experiments were done at least twice with reproducible results. The experimental groups comprised seven to nine animals. All data are expressed as mean ±s.d. and the results were comparedstatistically using the Mann-Whitney U test; ** p<0.02.
We thank Sibylle Widen for carrying out the electron microscopy, Inger Petersson, Andreas Järemo, Christina Eklund and Maria Larsson for skillful technicalassistance, and Dr. Vincent Collins for constructive criticism on the manuscript. The work was supported by the Swedish Medical Research Council grant No. K2000–16x 13062–02B and the VÅRDAL foundation grant No. A2000 088.