Tolerosome-induced oral tolerance is MHC dependent

Abstract

Oral administration of a protein antigen generates a serum factor that induces tolerance when transferred into naïve recipients. This serum factor has been described in rats as consisting of exosome-like structures or tolerosomes, which express major histocompatibility complex class II molecules (MHCII) and mediate antigen-specific tolerance. In this study, we investigated the functions of serum-derived tolerosomes both in vivo and in vitro. Tolerosomes were purified from the 100 000 g pellet fraction of serum from ovalbumin (OVA)-fed mice. When transferred into naïve recipient mice, the tolerosomes mediated OVA-specific tolerance. We also found that tolerosomes from OVA-fed mice induced the activation of OVA-specific T cells both in vivo and in vitro. The inoculation of severe combined immunodeficiency (SCID) mice with an interferon-γ-producing cell line normalized the expression of MHCII in the intestinal epithelial cells and restored their ability to generate tolerosomes. Syngeneic but not allogeneic transfer of tolerosomes from OVA-fed donors induced tolerance in the recipients. Our results show that tolerosomes can be isolated from mouse serum, that tolerosome-induced oral tolerance requires MHCII expression in intestinal epithelial cells, and that tolerosomes are functional only in syngeneic recipients.

Introduction

Feeding of protein antigens leads to the generation of CD4⁺ T cells that down-regulate the immune responses to the fed antigen^1–4 in a process that depends partly on the secretion of suppressive cytokines, such as interleukin (IL)-10 and transforming growth factor (TGF)-β.^5–9 The sites of induction and maintenance of these regulatory CD4⁺ T cells, which normally prevent harmful reactions to dietary antigens and antigens that originate from the normal flora, remain largely unknown.

In 1983, Strobel et al. discovered that serum collected from mice 1 hr after a single intragastric dose of ovalbumin (OVA) induced the suppression of delayed-type hypersensitivity (DTH) reactions when injected intraperitoneally (i.p.) into recipient mice.¹⁰ In addition, it has been shown that parenteral administration of native OVA does not produce the tolerogenic serum factor, which suggests that intestinal processing of the antigen is a prerequisite for the development of this factor.¹¹ Although various attempts have been made to mimic gut processing of the native protein, the modification does not appear to be a physicochemical alteration of the native protein, as parenterally administered denaturated or deaggregated OVA did not act as a tolerogenic serum factor.¹¹ Further, the tolerogenic activity in serum from OVA-fed animals does not seem to be directly related to the serum levels of immunoreactive antigen.¹²

Previously, we described a supra-molecular structure that was present in the serum from OVA-fed rats, and which could be isolated from the supernatant fluid of cultured antigen-pulsed intestinal epithelial cells (IECs) and was fully capable of inducing tolerance in naïve recipient animals.¹³ The tolerogenic factor was isolated in vitro by ultracentrifugation at 70 000 g, and shown to consist of exosome-like vesicles with an approximate size of 40 nm that stained positive for major histocompatibility complex class II molecules (MHCII). Similar structures have been isolated from a human and a mouse IEC line.¹⁴ Immunohistochemistry of the small intestines of rodents and humans reveals small MHCII-positive vesicles that are present in abundance in the apical and basolateral part of the IEC.^15–17 We have previously shown that these small vesicles, named tolerosomes, are produced by the IEC and constitute the tolerogenic fraction that is found in serum shortly after antigen feeding.¹³ Therefore, the induction of oral tolerance by the serum transfer model is referred to as ‘tolerosome-induced tolerance’.

Because regulatory CD4⁺ T cells are MHCII restricted, the fed antigens must at some stage enter an MHCII processing pathway in order to be recognized. The IECs that line the gut epithelium normally possess the machinery needed for antigen processing, and they express all the key markers of professional antigen-presenting cells.^13,18 Fed OVA is rapidly endocytosed at the apical membrane of the IEC, and is subsequently transported to the basolateral side of the cell in lysosome-associated membrane protein (LAMP)-1- and MHCII-positive vacuoles.¹⁶ Furthermore, it has been shown in vivo that orally administered OVA is only targeted to late endosomes that contain LAMP-1 and MHCII in the IECs in SCID mice that were treated with interferon (IFN)-γ.¹⁷ Interestingly, SCID mice are unable to produce the tolerogenic serum factor,¹⁹ which may be a result of incomplete antigen processing and the lack of MHCII in the IECs of the SCID mouse.

In this study, we isolated exosome structures from mouse serum after an antigen feed, and studied their functions both in vivo and in vitro. In addition, we investigated whether the inability of SCID mice to produce tolerosomes is related to the MHCII deficiency of their IECs, and whether serum-derived tolerosomes are functionally MHC restricted.

Methods

Mice

Female and male inbred CB.17 scid/scid H2^d (SCID) mice were originally obtained from Bomholtgärd Ltd (Ry, Denmark). Female and male CB.17 non-scid H-2^d (CB.17) mice, male BALB/c H-2^d (BALB/c) mice and male C57BL/6 H-2^b (C57B/6) mice were purchased from Bomholtgård Ltd and maintained under standard conditions. Female and male DO11·10 H-2^d[OVA T-cell receptor (TCR) transgenic] mice were bred and maintained under standard conditions. All of the animals were housed in the animal facilities of the Department of Rheumatology and Inflammation Research, University of Göteborg under standard conditions of temperature and light.

A low percentage of SCID mice are ‘leaky’, in that they produce low levels of immunoglobulins.²⁰ These mice were identified by assaying the serum immunoglobulin G (IgG) and IgM levels by enzyme-linked immunosorbent assay (ELISA), as previously described,²¹ prior to the initiation of the experiment. Those mice that had >5 µg/ml serum Ig were excluded from the study.

Antigens

Ovalbumin (OVA, grade V; Sigma Chemical Co., St Louis, MO) was used throughout this experiment. For DTH testing, the OVA solution (2 mg/ml) was denatured by heating at 80° in a water bath for 1 hr.

OVA feeding and isolation of exosome fractions from mouse serum

The mice were starved overnight for 8 hr, and then fed by gavage with 50 mg OVA dissolved in phosphate-buffered saline (PBS) (at a concentration of 100 mg/ml). One hour after feeding, the mice were bled by cardiac puncture. The blood samples were pooled for each group, allowed to clot, and then centrifuged twice to remove red blood cells, at 5000 g for 10 min at room temperature. The sera were collected and centrifuged for 1 hr at 100 000 g. The pellet fraction was suspended in PBS and the supernatant was collected. To exclude more exosomes from the supernatant and to wash the pellet, both fractions were centrifuged once more at 100 000 g for 1 hr. The pellet fraction was suspended in PBS and the supernatant was collected.

Adoptive transfer of tolerosomes

Naïve BALB/c mice were injected i.p. with a serum pellet that corresponded to 0·9 ml of serum dissolved in 0·9 ml of PBS or with 0·9 ml supernatant from OVA- or PBS-fed mice. All of the mice were immunized and tested for DTH reactions, as described below.

Tolerosome-induced activation of OVA-specific T cells in vivo

DO11·10 mice were injected i.p. with tolerosomes or supernatants as described above. Ten hours after transfer, the peripheral lymph nodes (PLNs) (popliteal, axillary, subiliac and cervical lymph nodes) and mesenteric lymph nodes (MLNs) were collected and prepared as single cell suspensions. The cells were stained and analysed by fluorescence-activated cell sorter (FACS) with the following monoclonal antibodies (mAbs): phycoerythrin (PE)-conjugated anti-CD69 and rat IgG2a (PharMingen, San Diego, CA), biotin-conjugated anti-glucocortico-induced TNF receptor family gene (GITR) and goat IgG (R & D Systems, Minneapolis, MN).

Tolerosome-induced activation of OVA-specific T cells in vitro

CD4⁺ T cells were isolated from PLNs, and CD11c⁺ dendritic cells (DCs) were isolated from the spleens of DO11·10 mice using magnetic antibody cell sorting (MACS) microbeads (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) according to the manufacturer's instructions. The T cells (2 × 10⁵) and DCs (2 × 10³) were cultured in 96-well U-bottom plates in Iscoves medium that was supplemented with 10% fetal calf serum (FCS), 1%l-glutamine, 1% sodium pyruvate and 1% penicillin-streptomycin. The cells were stimulated with OVA (100 µg/well) and the serum pellet isolated from OVA- or PBS-fed mice, as described above. This set-up was used in order to monitor both the stimulatory and the suppressive abilities of the isolated exosome fraction. Pellets from four pools of OVA serum and four pools of PBS serum (two or three animals per pool), which corresponded to 0·8, 0·08 and 0·008 ml of serum, were diluted in 50 µl of medium and added to the wells. The cells were incubated in a 5% CO₂ atmosphere for 96 hr, and then pulsed with 1 µCi/well ³H-thymidine (Amersham International, Buckinghamshire, UK) for 8 hr. The cells were then harvested (Inotech; Ninolab AB, Upplands Visby, Sweden) and dried and the levels of incorporated ³H-thymidine were measured in a Matrix 96 β-counter (Canberra Packard, Uppsala, Sweden).

In vivo treatment of donor SCID mice with IFN-γ

Chinese hamster ovary (CHO-211A) cells were used as a prolonged source of IFN-γ in the SCID mice. The CHO-211A cell line incorporates a gene that produces 10⁴−10^4.5 units IFN-γ/ml in the supernatant fluid when grown in vitro, while the wild-type (wt) CHO-K1 clone has no detectable IFN-γ production. The cells were cultured in 50-cm³ culture flasks in Iscoves medium that was supplemented with 10% FCS, 1% sodium pyruvate, 1%l-glutamine and 1% penicillin-streptomycin and incubated at 37° in a 5% CO₂ atmosphere. The cells were split in a 1 : 8 ratio every third day until harvest.

SCID mice were injected i.p. with either IFN-γ-producing cells (CHO-211A) or wt cells (CHO-K1) (2 × 10⁶ cells in a 0·5-ml volume). CB.17 mice were injected i.p. with 0·5 ml of PBS.

Immunoperoxidase staining of specimens from the small intestine

The small intestines of SCID mice, which were inoculated with IFN-γ-producing cells or wt cells, were stained for MHCII. SCID mouse biopsies from the small intestine were removed and placed in specimen moulds (Tissue-Tek Cryomould Biopsy; Miles Inc. Diagnostic Division, Elkhart, IN) with Tissue-Tek O.C.T. compound (Sakura Finetek Europe BV, Zoeterwoude, the Netherlands). The biopsies were frozen instantly in isopentane cooled by liquid nitrogen, and then stored at −70°. Cryostat sections (6-µm thickness) were fixed in cold acetone (30 seconds in 50% acetone followed by 5 min in 100% acetone). Endogenous peroxidase activity was blocked by incubating the slides for 20 min in a solution of 1 U/l glucose oxidase (Type V-S; Sigma), 10 mm glucose and 1 mm NaN₃. The slides were incubated overnight at 4° with biotinylated mouse anti I-A^d mAb (Pharmingen, San Diego, CA), which was diluted 1/1000 in PBS-Tween. The slides were then incubated with avidin-conjugated peroxidase (ABC-complex; Dako, Glostrup, Denmark) for 30 min. Finally, the peroxidase-stained tissues were visualized with amino-ethyl-carbazole followed by light counterstaining with Mayer's haematoxylin. The assessment of MHCII-positive cells was carried out under a Leica Q500MC microscope (Leica, Cambridge, UK). The degree of MHCII staining in the lamina propria (LP) was recorded as the relative stained area (%) of five individuals (three sections per individual).

Adoptive transfer of serum from SCID to CB.17 mice

SCID mice were starved overnight 1 week after inoculation with IFN-γ-producing or wt CHO cells and then fed by gavage with 50 mg OVA dissolved in PBS (at a concentration of 100 mg/ml). One hour after feeding, the mice were bled by cardiac puncture. The blood samples were pooled for each group, allowed to clot, and then centrifuged at 2500 g for 10 min at room temperature. The serum collected for each experimental group was injected i.p. into the recipient CB.17 mice at a volume of 0·7 ml.

Syngeneic and allogeneic transfer of serum between BALB/c and C57B/6 mice

Sera from OVA-fed BALB/c and C57B/6 donor mice were collected as described. The serum was transferred to syngeneic recipients (BALB/c to BALB/c; C57B/6 to C57B/6) and allogeneic recipients (BALB/c to C57B/6; C57B/6 to BALB/c) by i.p. injection at a volume of 0·7 ml.

Immunization of recipients

One week after serum transfer, the recipients were immunized intramuscularly (i.m.) in the hind legs with 50 µg of OVA emulsified in 0·05 ml of Freund's complete adjuvant (FCA) (Difco Laboratories, Detroit, MI). Three weeks later, the mice were booster-immunized subcutaneously (s.c.) at the base of the tail with 100 µg of OVA emulsified in 0·1 ml of incomplete Freud's adjuvant (IFA) (Difco Laboratories).

Delayed-type hypersensitivity (DTH) testing

One week after booster immunization, all the recipient mice were challenged s.c. with 100 µg of denatured OVA in the left hind footpad. The footpad thickness was measured before and 24 hr after challenge with a micrometer caliper (Oditest; Kroplin, Hessen, Germany). The difference between the two measurements gave an index of footpad swelling in millimetres, which was used for group comparisons.

Passive cutaneous anaphylaxis (PCA)

The levels of circulating anti-OVA IgE antibody were assayed by PCA in Sprague-Dawley rats. In brief, the rats were anaesthetized by i.p. injections of Hypnorm® (0·032 mg of fetanyl citrate and 1 mg of fluanisone per rat) and Stesolid Novum® (0·5 mg/rat). Test mouse sera were serially diluted twofold in PBS, and 50 µl of each dilution was injected intradermally into the shaved dorsal skin of the rats. Seventy-two hours later, the rats were challenged intravenously with 5 mg of OVA diluted in 1 ml of PBS that contained 1% Evans blue (Sigma). One hour later, the rats were killed and the skin areas were examined for the appearance of blue colour. The PCA titres were defined as the reciprocal of the highest dilution of serum giving a blue spot of <5 mm in diameter.

FACS analysis

The cells to be analysed were dissolved in FACS buffer [PBS containing 1% FCS, 0·5 mm ethylenediaminetetraacetic acid (EDTA) and 0·1% sodium azide], aliquoted in 96-well V-bottom plates and pelleted (3 min at 300 g and 4°). The cells were incubated with FcRII/III blocking antibody for 5 min, before incubation with predetermined optimal concentrations of either flourochrome- or biotin-labelled mAbs for 20 min at 4° in the dark. Streptavidin-allophycocyanin (APC) (PharMingen) were added to the cells stained with biotin-conjugated Ab, for 20 min at 4°. After washing, the cells were suspended in FACS buffer and 20 000 cells were analysed in the FACScalibur (Becton-Dickinson, New Jersey, NJ), which was equipped with the CellQuest software. The percentages of positive cells were recorded.

Analysis of splenocytes from IFN-γ and sham-treated SCID mice

Spleens were removed from SCID mice inoculated with IFN-γ-producing or wt CHO cells, 1 week after cell transfer, and single cell suspensions were prepared and analysed by FACS with the following mAbs: PE-conjugated anti-CD80, anti-CD86, anti-CD40, hamster IgG1, rat IgG2a, fluorescein isothiocyanate (FITC)-conjugated anti-IAd, biotin-conjugated IAb, anti-CD11c and F4/80 (PharMingen).

Phenotypic characterization of tolerosomes

Streptavidin-conjugated magnetic beads 2·8 µm in diameter (Dynabeads M-280, Dynal beads; Dynal Biotech, Oslo, Norway) were incubated with biotin-conjugated anti-IA^d Ab or control Ab (anti-IA^b) (PharMingen) for 30 min at room temperature with agitation. The beads were washed three times in PBS + bovine serum albumin (BSA) (1%) using a magnetic particle separator (Dynal Biotech). Tolerosomes isolated from the OVA-fed BALB/c mice were added to the beads and incubated for 2–3 hr at room temperature with agitation. Unbound streptavidin and biotin were blocked with the Avidin-Biotin Blocking kit (Vector Laboratories, Inc., Burlingame, CA) according to the manufacturer's instructions. The beads were washed three times, and stained for surface markers as described previously, except that a magnetic separator was used in the washing steps. After washing, the beads were suspended in FACS buffer and 20 000 beads were analysed in the FACScalibur (Becton-Dickinson) equipped with the CellQuest software. The percentage of positive beads was recorded using the following mAbs (all from Pharmingen): PE-conjugated α-CD86, α-CD40, hamster IgG1, rat IgG2a, FITC-conjugated α-IAd, α-Mac-3, α-LAMP-1, α-CD68, α-IAb, rat IgG2a and biotin-conjugated α-intercellular adhesion molecule type 1 (ICAM-1), followed by streptavidin-APC as a second step. The IEC marker A33 was detected using a polyclonal rabbit anti-A33 (α-A33; a kind gift from Dr Joan Heath, Ludwig Institute for Cancer Research, Sidney, Australia) or normal rabbit serum as control, followed by a secondary biotinylated goat anti-rabbit IgG (Sigma).

Electron microscopy analysis

Exosomes were attached to magnetic beads and stained with biotinylated Abs as described above. ExtrAvidin® 10 nm Gold Conjugate (Sigma), diluted 1 : 5, was added to the beads and after washing the beads were fixed in Na-cacodylate buffer (0.1%, pH 7.35) with glutaraldehyde (3%). Exosome–bead complexes were washed in PBS and embedded in agar. The agar was cut into small pieces, and fixed overnight in 1% osmium tetroxide. After washing in PBS, the sample was dehydrated in a graded series of ethanol: 70% for 15 min, 95% for 15 min and 100% ethanol for 4 × 15 min followed by 100% propylenoxid for 2 × 5 min. The sample was polymerized with epoxy resin and 100% propylenoxid, 1 : 1, for 2 hr, and finally pure epoxy resin was added and left at room temperature overnight. The sample was transferred to a mould and polymerized at 60° for 24 hr, and was then passed to Sibylle Widehn of the Department of Pathology, University of Gothenburg, Sweden for analysis in an electron microscope.

Statistical analysis

Experiments were performed at least twice, with reproducible results. All data are expressed as mean ± standard deviation (SD) and the results were compared statistically using the Mann–Whitney U-test with levels of significance of P < 0·05 and P < 0·01.

Results

Tolerosome isolation from mouse serum

To show that the tolerogenic factor in the serum was present in the pellet fraction that contained the tolerosomes, mouse serum was collected 1 hr after feeding of OVA or PBS, and the exosome fraction was isolated by centrifugation of the serum at 100 000 g. The pellet fraction and supernatant were transferred into naïve recipient mice at a dose corresponding to 0·9 ml of serum. The pellet fraction isolated from OVA-fed mice induced tolerance in the recipients as compared with the controls that received the pellet from PBS-fed mice (P < 0·01). The supernatant from the OVA-fed mice failed to induce tolerance, which indicates that tolerosomes were abundant in the exosome fraction of the serum (Fig. 1). The pellet fraction did not contain any immunoreactive OVA, as detected by ELISA, while the supernatant contained 0·5–1·0 µg/ml OVA.

Figure 1.

Delayed-type hypersensitivity reaction (DTH) to ovalbumin (OVA) in pellet or supernatant recipients. Serum from OVA- or phosphate-buffered saline (PBS)-fed donors was centrifuged at 100 000 g and the pellet or supernatant was injected into recipients at a dose corresponding to 0·9 ml of serum. All mice were immunized with 100 µg of OVA in complete Freund's adjuvant (CFA) 1 week after the injection of the donor pellet or supernatant and then booster-immunized with 100 µg of OVA in incomplete Freud's adjuvant (IFA) 3 weeks later. DTH was assayed 5 weeks after serum or pellet transfer as the mean increment in footpad thickness at 24 hr following challenge with 100 µg of OVA. The bars represent the mean increment in footpad thickness ± the standard error of the mean for eight individuals. Statistical analyses were performed using the Mann–Whitney U-test. **P < 0·01.

Tolerosomes can activate OVA-specific T cells both in vivo and in vitro

To test the functionality of tolerosomes, we isolated the exosome fractions from OVA-fed or PBS-fed mice and injected them i.p. into OVA-transgenic DO11·10 mice. Ten hours post-transfer, we removed the peripheral and mesenteric lymph nodes and stained the cells for surface markers. We found that the tolerosomes isolated from OVA-fed mice activated OVA-specific T cells, both in the PLNs and in the MLNs, as shown by increased expression of the activation markers CD69 (P < 0·05) and GITR (P < 0·05) compared with tolerosomes isolated from PBS-fed mice. The level of activation was comparable to that induced by the supernatant from OVA-fed mice, which contained 0·5–1·0 µg OVA (PLN: Fig. 2; MLN: data not shown). We also investigated the in vitro capacity of tolerosomes to stimulate OVA-specific CD4⁺ T cells. Tolerosomes were isolated from mouse serum 1 hr after feeding of OVA or PBS, and added to cultures together with OVA at doses corresponding to 0·8, 0·08 and 0·008 ml of serum. Tolerosomes isolated from OVA-fed mice significantly increased by twofold the proliferation of OVA-specific CD4⁺ T cells as compared with cultures with tolerosomes that were isolated from PBS-fed mice (P < 0·01) or with the positive control (P < 0·05), which was stimulated with OVA alone (Fig. 3).

Figure 2.

In vivo activation of ovalbumin (OVA)-specific T cells. The serum pellet and supernatant from OVA- or phosphate-buffered saline (PBS)-fed mice were injected intraperitoneally (i.p.) into naïve DO11·10 mice. Ten hours after transfer, peripheral lymph nodes were removed and the OVA-specific KJ.126⁺ cells were analysed in the fluorescence-activated cell sorter (FACS) for expression of the activation markers CD69 and glucocortico-induced TNF receptor family gene (GITR). The bars represent the mean value ± the standard error of the mean for six individuals. Statistical analyses were performed with the Mann–Whitney U-test. *P < 0·05. The supernatant contained 0·1–0·5 µg/ml OVA and in the pellet the amount of OVA was undetectable.

Figure 3.

In vitro activation of ovalbumin (OVA)-specific T-cells. CD4⁺ T cells were separated by magnetic antibody cell sorting (MACS) from DO11·10 mice and stimulated with 100 µg of OVA together with CD11c⁺ antigen-presenting cells (APCs) and the serum pellet from OVA- or PBS-fed donors. ³H-thymidine incorporation was estimated after 4 days of cell culture. The pellet added to the cultures corresponds to the indicated amount of serum (0·8, 0·08 and 0·008 ml). The bars represent the mean value ± the standard error of the mean for five individuals. Statistical analyses were performed with the Mann–Whitney U-test. *P < 0·05. The supernatant contained 0·1–0·5 µg/ml OVA and in the pellet the amount of OVA was undetectable.

IFN-γ induces MHCII expression in SCID mouse IECs

SCID mice were inoculated with CHO cells that produce mouse IFN-γ (CHO-211A cells) or with wt CHO cells (2 × 10⁶ cells per mouse) by i.p. injection. One week after inoculation, the mice were killed and specimens were removed from the small intestine, sectioned, and stained for MHCII (I-A^d). Mice that were inoculated with the IFN-γ-producing CHO-211A showed positive staining for MHCII in the IECs (Fig. 4a). The inoculation with CHO-211A did not completely restore the MHCII expression in the SCID mice IECs to the wt appearance, but the staining patterns were similar to those found in wtCB.17 mice, showing both apical and baso-lateral staining of the IECs (Fig. 4c). SCID mice that were inoculated with wt CHO cells were negative for MHCII expression in the IECs (Fig. 4b). The two groups of treated SCID mice showed comparable staining for MHCII-positive cells, with dendritic morphology in the LP (Figs 4a and b) resembling that of non-treated SCID mice (Table 1). FACS analysis of splenocytes showed that the mice inoculated with the IFN-γ-producing cells had a higher frequency of macrophages (F4/80⁺), DCs (CD11c⁺), and MHC II-positive cells than the wt CHO-inoculated SCID, although the macrophages and DCs expressed similar levels of activation markers CD80, CD86, CD40 on a per cell basis (Table 2).

Figure 4.

The expression of major histocompatibility complex class II molecules (MHCII) in small intestinal epithelial cells (IECs) of mice. Frozen sections (6 µm) of small bowel biopsy specimens were taken from severe combined immunodeficiency (SCID) mice injected intraperitoneally (i.p.) with interferon (IFN)-γ-producing Chinese hamster ovary (CHO) cells or wild-type (wt) CHO cells (2 × 10⁶ cells per mouse). Biopsy specimens were also taken from wt CB.17 mice. The biopsies were collected 1 week after i.p. injection of CHO cells. The sections were labelled with an antibody against MHCII (I-A^d). Bars, 30 µm. (a)SCID mice administered IFN-γ-producing CHO cells demonstrated abundant expression of intracellular MHCII in the IECs, comparable to that of wt CB.17 mice (c).(b)SCID mice administered wt CHO cells were negative for MHCII expression in the IECs. The MHCII staining of the dendritic cells (DCs) in the lamina propria (LP) of both IFN-γ- and sham-treated SCID mice was comparable to that observed in normal CB.17 wt mice (c). (c)wt CB.17 mice that constitutively expressed intracellular MHCII in the IECs.

Table 1.

Expression of major histocompatibility complex class II molecules (MHCII) in the lamina propria

	Relative stained area (%) in lamina propria
	wt CHO1	CHO-211A2
I-Ad	3·9 ± 2·2	5·9 ± 3·1
CD11c	7·4 ± 2·2	6·3 ± 3·2
CD86	0·4 ± 0·4	0·3 ± 0·4

The small intestines of severe combined immunodeficiency (SCID) mice, which were inoculated with interferon (IFN)-γ-producing cells or wild-type (wt) cells, were stained for MHCII and examined under the microscope. The degree of MHCII staining in the lamina propria was recorded as the relative stained area (%) for five individuals (three sections per individual).

¹SCID mice inoculated with IFN-γ-producing Chinese hamster ovary (CHO)-211A cells.

²SCID mice inoculated with wt CHO cells.

Table 2.

Phenotype and frequency of antigen-presenting splenocytes

	Surface marker/cell type
	F4/80		CD11c		MHCII
Positive cells (%)	wt CHO1	CHO-211A2	wt CHO	CHO-211A	wt CHO	CHO-211A
Total	7·6 ± 1·2	9·5 ± 0·5*	8·4 ± 1·1	10·9 ± 0·9*	10·8 ± 1·8	27 ± 13·5*
CD80	44 ± 7·9	31·7 ± 4·4	34·2 ± 3·9	25·1 ± 5·6	27·4 ± 2	41·1 ± 6·6
CD86	32 ± 10·3	23 ± 6·3	17·9 ± 3·3	18·2 ± 3·7	23·4 ± 2·4	33·5 ± 6·1
CD40	19·4 ± 4·2	13·5 ± 2·3	14 ± 3·2	10 ± 3·1	6·3 ± 2·4	13·8 ± 6·7
I-Ad	30·9 ± 10·4	37·3 ± 4·4	27·4 ± 4·9	24·4 ± 10·3

Spleens were removed from severe combined immunodeficiency (SCID) mice inoculated with interferon (IFN)-γ-producing or wild-type (wt) Chinese hamster ovary (CHO) cells, 1 week after cell transfer, and single cell suspensions were prepared and analysed by fluorescence-activated cell sorter (FACS) for surface markers.

¹SCID mice inoculated with wt CHO cells.

²SCID mice inoculated with IFN-γ producing CHO-211 A cells.

^*P < 0·05.

MHCII, major histocompatibility complex class II molecules.

SCID mice inoculated with IFN-γ-producing cells are able to produce a tolerogenic serum factor

Serum taken from wt CB.17 mice 1 hr after a single feed of OVA induced tolerance in naïve recipient mice with a significant reduction of the DTH response (P < 0·05) as compared with control mice, which received serum from PBS-fed mice (Fig. 5a). In addition, the SCID mice inoculated with the IFN-γ-producing CHO-211A cells were able to generate tolerosomes, as shown by the transfer of serum from OVA-fed CHO-211A-inoculated SCID mice to naïve recipients. These recipient mice showed significant reductions in DTH responses as compared with the recipients of serum from OVA-fed wt CHO-inoculated SCID mice (P < 0·05) and the recipients of serum from PBS-fed CHO-211A-inoculated SCID mice (P < 0·01). (Fig. 5a). The IgE responses of the recipients suggested that antibody responses were also suppressed by tolerosomes from CHO-211A-inoculated SCID mice. To determine the IgE anti-OVA response of the recipient mice, sera were collected 5 weeks after serum transfer and analysed by PCA testing in rats. The PCA titres were significantly lower in the group of mice that received serum from OVA-fed, CHO-211A-inoculated SCID donors (P < 0·01), as compared with the control group, in which the mice received serum from OVA-fed wt CHO-inoculated SCID donors (Fig. 5b).

Figure 5.

Delayed-type hypersensitivity reaction (DTH) to ovalbumin (OVA) in recipient mice. Serum from OVA- or phosphate-buffered saline (PBS)-fed donors [interferon (IFN)-γ-producing Chinese hamster ovary (CHO)-211A-inoculated severe combined immunodeficiency (SCID) mice, wild-type (wt) CHO-inoculated SCID mice and wt CB.17 mice] was injected into wt CB.17 recipients at a dose of 0·7 ml. All mice were immunized with 100 µg of OVA in complete Freund's adjuvant (CFA) 1 week after the injection of donor serum and then booster-immunized with 100 µg of OVA in incomplete Freund's adjuvant (IFA) 3 weeks later. (a)DTH was assayed 5 weeks after serum transfer as the mean increment in footpad thickness at 24 hr following challenge with 100 µg of OVA. Treatment of the donor mice is indicated below the bars (K1 is sham-treated SCID, 211A is IFN-γ-treated SCID and wt is untreated CB.17). The bars represent the mean increment in footpad thickness ± the standard error of the mean (SEM). (b)OVA-specific immunoglobulin E (IgE) antibody levels in serum of recipient mice as determined by passive cutaneous anaphylaxis (PCA) at 5 weeks after the serum transfer. The bars represent the mean PCA titres ± SEM. Statistical analyses were performed with the Mann–Whitney U-test. **P < 0·01.

Tolerosomes are tolerogenic only when transferred to syngeneic recipients

Serum from OVA-fed or PBS-fed BALB/c and C57B/6 mice was transferred to syngeneic mice (BALB/c to BALB/c; C57B/6 to C57B/6) or allogeneic mice (BALB/c to C57B/6; C57B/6 to BALB/c). The syngeneic recipients of serum from OVA-fed donors showed significant suppression of the DTH response (BALB/c, P < 0·01; C57B/6, P < 0·01), as compared with the control recipients, which received serum from PBS-fed syngeneic donors (Figs 6a and b). The allogeneic transfer of serum did not result in tolerized recipients, as there was no difference in the DTH responses of the allogeneic serum recipients, regardless of whether the donor mice were fed OVA or not (Figs 6a and b).

Figure 6.

Delayed-type hypersensitivity reaction (DTH) to ovalbumin (OVA) and OVA-specific immunoglobulin E (IgE) antibody levels in serum in recipient mice. Serum from OVA- or phosphate-buffered saline (PBS)-fed donors was transferred to syngeneic recipients (BALB/c to BALB; C57B/6 to C57B/6) and allogeneic recipients (BALB/c to C57B/6; C57B/6 to BALB/c) by intraperitoneal (i.p.) injection at a dose of 0·7 ml. All mice were immunized with 100 µg of OVA in complete Freund's adjuvant (CFA) 1 week after the injection of donor serum and then booster-immunized with 100 µg of OVA in incomplete Freund's adjuvant (IFA) 3 weeks later. (a, c) DTH was assayed 5 weeks after serum transfer as the mean increment in footpad thickness at 24 hr following challenge with 100 µg of OVA. (b, d) OVA-specific IgE antibody levels in serum of recipient mice as determined by passive cutaneous anaphylaxis (PCA) at 5 weeks after the serum transfer. The bars represent the mean increment in footpad thickness ± the standard error of the mean (SEM) and mean PCA titres ± SEM. Statistical analyses were performed with the Mann–Whitney U-test. *P < 0·05, **P < 0·01.

Syngeneic serum transfer resulted in a significant suppression of PCA titres in the group of mice that received serum from OVA-fed BALB/c donors (P < 0·05), as compared with the syngeneic control recipients (Fig. 6c). Although the PCA titres were also lower in the C57B/6 syngeneic transfer mice, the difference did not reach statistical significance (P = 0·1) (Fig. 6d). Allogeneic serum transfer did not result in any difference in PCA titre between any of the recipient groups (Figs 6c and d).

FACS and electron microscopy (EM) analyses of the serum-derived exosome fraction

The exosome fraction isolated from serum from OVA-fed normal BALB/c mice was incubated with magnetic beads that were precoated with anti-MHCII or control Ab, before being stained for different surface markers, and analysed by FACS and EM. The exosome fraction was strongly positive for MHCII and MAC-3 (Fig. 7d), and also showed positive staining for CD68 (Figs 6a–c), CD86, CD40, LAMP-1, ICAM-1 and the epithelial specific marker A33 (Fig. 7d). This analysis is mainly qualitative and was performed in bead excess in order to obtain a rough estimate of the abundance of the markers on the MHCII captured exosome fraction. EM also visualized the exosome fraction and Gold-labelled ICAM-1 is shown on exosomes attached to the magnetic beads (Fig. 8).

Figure 7.

Characterization of serum pellet by fluorescence-activated cell sorter (FACS) analysis. Major histocompatibility complex class II molecule (MHCII)-positive material from the serum pellet was attached to magnetic beads using beads precoated with MHCII-specific antibody (Ab) or control Ab. The beads were then stained and analysed by FACS to detect different surface markers on the attached material. (a) Typical FACS appearance of beads precoated with MHCII Ab and stained for CD68. (b) Beads precoated with MHCII Ab and stained for control Ab. (c) Beads precoated with control Ab and stained for CD68. (d)Per cent positively stained beads for the marker indicated from a typical 100 000 g serum fraction from ovalbumin (OVA)-fed mice. The bars represent one value and are representative of three independent experiments.

Figure 8.

Intercellular adhesion molecule type 1 (ICAM-1)-expressing exosomes visualized by electron microscopy. Tolerosomes were prepared from mouse serum and incubated with magnetic beads precoated with major histocompatibility complex class II molecule (MHCII)-specific antibody (Ab) or contol Ab. The MHCII-positive material from the serum pellet was attached to the magnetic beads and stained for ICAM-1 labelled with Gold Conjugate and visualized by electron microscopy. The arrows show 50 nm large exosomes (mostly in aggregates) attached to the magnetic bead and the black dots are ICAM-1 expressed on the exosomes. The inset picture is the negative control, showing no exosomes attached to the magnetic bead that was precoated with control Ab and stained for ICAM-1. Bar = 100 nm.

Discussion

As previously shown in the rat, a transferable tolerogenic factor can be isolated by preparative ultracentrifugation from the serum 1 hr after feeding of an antigen. The tolerogenic factor showed exosome characteristics, stained positive for MHCII, and mediated all the tolerizing activity found in the serum. In the present study, we confirmed these results in a mouse system, and further analysed the role of MHCII in this process. We demonstrated that the inoculation of SCID mice with IFN-γ-producing CHO cells induced the expression of MHCII in IECs, which, while not completely restored to the wt appearance, produced similar staining patterns to those found in normal wt mice. This treatment rendered the SCID mice capable of producing the exosome-like tolerogenic serum factor named tolerosomes. The sham-treated SCID mice that were inoculated with wt CHO cells were still negative for MHCII in their IECs and could not produce tolerosomes. There was no other obvious difference between the mice that were inoculated with IFN-γ-producing CHO cells and those that received the wt CHO cells, in terms of morphology, MHC II expression or the number of DCs in the LP. Nevertheless, we observed an increase in the number of macrophages and DCs in the spleens of CHO-211A-inoculated SCID mice and we cannot exclude the possibility that these cells also generate exosomes. However, as the macrophages and DCs were equally activated in the wt CHO- and CHO-211A-inoculated SCID groups, it seems unlikely that their potential exosome release would be different. Considering these results together, we suggest that the major and functional difference between SCID mice inoculated with IFN-γ-producing cells and sham-treated SCID mice was the IEC MHCII expression.

IFN-γ has a well-established effect in inducing the expression of MHCII in IECs both in vivo^22,23 and in vitro.^13,24 The IECs of normal mice constitutively express intracellular MHCII,¹³ while the IECs of SCID mice are devoid of MHCII in the intestinal epithelium,^16,17 an observation that was confirmed in the present study.

The difference in MHCII expression between normal mice and SCID mice is probably linked to the fact that SCID mice are deficient in functional T lymphocytes. In normal mice, intraepithelial lymphocytes (IELs) produce IFN-γ,²⁵ which induces the expression of MHCII in the IECs²⁶ in response to microbial presence in the gut lumen. Thus, germ-free mice lack MHCII expression in the small intestinal epithelium^27,28 (and our own unpublished observations) and also have very few IELs.^29,30 This finding implies that constitutive MHCII expression in the IELs of normal mice is caused by constant exposure to microbial products from the normal gut flora.^31,32

In orally tolerized mice, CD4⁺ T cells are responsible for the observed down-regulation of immunological effector functions in response to the fed antigen.^1,33 This requires the assembly of antigen and MHCII for activation of the CD4⁺ regulatory T cells that are responsible for the oral tolerance. Previous reports have revealed that soluble protein antigens are rapidly pinocytosed from the intestinal lumen and directed to an MHCII-containing vesicular compartment of the IEC.^13,15 Furthermore, Zimmer et al. have shown that the passage of OVA through the intestinal epithelium is different in SCID mice than in wt mice.¹⁶ OVA reaches the paracellular side of the IEC in LAMP-1-positive vesicles that contain MHCII antigens in BALB/c mice, whereas in SCID mice OVA is transported in LAMP-1-negative vacuoles that are devoid of MHCII antigens. In addition, the same group has recently shown that IFN-γ treatment of SCID mice normalizes the transport and loading of OVA in the MHCII compartment¹⁷ and induces the budding of antigen-MHCII-positive exosome-like vesicles in the baso-lateral IEC compartment. These findings strongly suggest that intestinal antigens pass through the IECs in SCID mice without being displayed on MHCII molecules and that IFN-γ can profoundly alter this route and allow the formation of the functional exosomes described in the present study.

The findings of the present study accord with previous observations showing that SCID mice are unable to generate a tolerogenic serum factor after oral administration of OVA.¹⁹ We propose that the incapability of SCID mice to process OVA in a tolerogenic fashion is a consequence of their MHCII-deficient pathway through the IECs. This is further supported by the observations that MHC class II-deficient mice cannot be orally tolerized³⁴ and that MHCII depletion abolishes the tolerogenic effect of serum from Ag-fed animals.¹³

The present study suggests that the tolerogenic conversion of OVA takes place in the IECs via an MHCII-dependent pathway, as discussed above. This is consistent with the discovery that MHCII-bearing tolerosomes are responsible for oral tolerance induction¹³ and, as shown in the present study, that tolerosomes are tolerogenic only when they are transferred to syngeneic recipients, which indicates MHC restriction. Tolerosomes isolated from mice are also able to activate OVA-specific T cells in vivo and in vitro. This was shown in vivo by injecting tolerosomes i.p. into DO11·10 mice, in which T cells increased the expression of the activation markers CD69 and GITR. Tolerosomes from OVA-fed mice also increased by twofold the OVA-specific proliferation of DO11·10 T cells. These experiments do not explain the tolerogenic effect of tolerosomes, but rather demonstrate the ability of tolerosomes to activate the immune system. The activation of OVA-specific T cells by tolerosomes from OVA-fed mice may reflect an initial step in tolerance induction, which is clearly not caused by native OVA contamination of the serum, as no OVA could be detected in the pellet fraction, whereas the serum supernatant contained detectable levels of OVA. The flow cytometric analyses of the serum-derived exosome fraction reveal similarities to the human and mouse IEC line-derived exosomes, which have been extensively investigated. The exosome fraction stained positive for MHCII, MAC-3, CD68, CD40, LAMP-1, ICAM-1 and A33. The A33 antigen is exclusively expressed on epithelial cells³⁵ and has been found on exosomes released by mouse IECs. This shows that at least part of the exosome fraction found in serum is epithelial cell derived.³⁶ It should be noted that the percentages shown in Fig. 6 are not quantitative and were generated with an excess of beads, and will therefore not reveal the actual proportion of exosomes carrying the different markers.

Unlike the tolerosomes described in the present study, mouse IEC line-derived exosomes have been shown to induce antigen-specific immunity in naïve recipient mice. This apparent contradiction probably reflects the flexibility of an immune reaction that is to a great extent affected by surrounding factors such as the cytokine milieu created by other cells. It should also be noted that we have in a previous study showed the in vivo tolerogenic capacity of in vitro-derived exosomes from rat IECs.¹³ Nevertheless, the ability of intestinal epithelial cell-derived exosomes to influence the immune system is apparent, and points to an important pathway for antigens that enter the gut mucosa. Although the mechanism by which the MHCII-positive antigen-loaded exosomes act in inducing tolerance is still not understood, the primary events are likely to take place in lymph nodes that have access to the tolerosomes, either directly by stimulation of naïve T cells or indirectly after being taken up by DCs that subsequently migrate to the draining lymph node.³⁷ The tolerosomes present in the circulation probably first enter the portal vein and end up to a large extent in the liver, an organ that is known to filter particles of similar size to tolerosomes that are approximately 40–50 nm. The liver represents an environment with suppressive cytokines, i.e. IL-10 and prostaglandin E₂, which would favour ‘tolerogenic’ presentation of antigens by local APCs. Tolerosomes entering the liver might also be taken up by the liver sinusoidal endothelial cells (LSECs).³⁸ These cells are specialized endothelial cells with antigen-presenting capacity.³⁹ The tolerosomes could also be taken up by liver DCs, and it has been shown recently that, shortly after feeding of FITC-labelled OVA, CD11c⁺ MHCII DCs in the liver acquire the FITC stain, which indicates that native or processed OVA gains access to these cells in vivo⁴⁰ and can induce OVA-specific regulatory T cells locally in the livers of DO11·10 mice.⁴¹ In addition, we have shown that T cells in the liver draining celiac lymph nodes have also been shown to be engaged very early in response to fed antigens and gain a regulatory phenotype (Hultkrantz et al., Department of Rheumatology and Inflammation Research, Góteborg University, unpublished results). Indeed, liver DCs under homeostatic conditions have been shown to possess a unique tolerogenic capacity both for orally administered antigens and for intraportal administration of both particulate and soluble transplantation antigens.^42,43

In conclusion, we show here that the exosome fraction obtained from serum shortly after an antigen feed contains the tolerosomes that effectively can induce tolerance when transferred to naïve MHC-matched recipients. We also provide indirect evidence that the production of functional tolerosomes requires the presence of MHCII in the small intestinal epithelial cells. Because MHCII expression in IECs is ultimately dependent on the presence of luminal microbes, we propose that the degree and quality of this stimulation can influence the capacity to become tolerant to environmental antigens.

Abbreviations

CHO

Chinese hamster ovary

DTH

delayed-type hypersensitivity

IEC

intestinal epithelial cell

IEL

intraepithelial lymphocyte

wt

wild type