Cross-presentation of oral antigens by liver sinusoidal endothelial cells leads to CD8 T cell tolerance



After ingestion, oral antigens distribute systemically and provoke T cell stimulation outside the gastrointestinal tract. Within the liver, scavenger liver sinusoidal endothelial cells (LSEC) eliminate blood-borne antigens and induce T cell tolerance. Here we investigated whether LSEC contribute to oral tolerance. Oral antigens were efficiently cross-presented on H-2Kb by LSEC to naive CD8 T cells. Cross-presentation efficiency in LSEC but not dendritic cells was increased by antigen-exposure to heat or low pH. Mechanistically, cross-presentation in LSEC requires endosomal maturation, involves hsc73 and proteasomal degradation. H-2Kb-restricted cross-presentation of oral antigens by LSEC in vivo induced CD8 T cell priming and led to development of CD8 T cell tolerance in two independent experimental systems. Adoptive transfer of LSEC from mice fed with antigen (ovalbumin) into RAG2–/– knockout mice, previously reconstituted with naive ovalbumin-specific CD8 T cells, prevented development of specific cytotoxicity and expression of IFN-γ in CD8 T cells. Using a new transgenic mouse line expressing H-2Kb only on endothelial cells, we have demonstrated that oral antigen administration leads to tolerance in H-2Kb-restricted CD8 T cells. Collectively, our data demonstrate a participation of the liver, in particular scavenger LSEC, in development of CD8 T cell tolerance towards oral antigens.


gut-associated lymphatic tissue


liver sinusoidal endothelial cells


It has been clearly shown that the gut-associated lymphatic tissue (GALT) 1, and especially mesenteric lymph nodes, are essential for induction of oral tolerance 2. However, several reports demonstrated that oral antigens are not restricted to the gut 35. After oral ingestion a rapid systemic distribution of antigen is observed, which results in MHC class II-restricted antigen presentation in the spleen and antigen-specific T cell proliferation 6, 7. These observations not only argue against strict gut-specific compartmentalization of immune responses against oral antigens, but also raise the question how oral tolerance is induced in view of the widespread antigen distribution. Blood draining from the intestine via the portal vein reaches the liver as the next anatomical location, where metabolization of oral antigens occurs. Intestinal venous drainage through the liver is important for the development of oral tolerance. Diverting blood draining from the intestine away from the liver by creation of porto-caval or mesenterico-caval shunts profoundly affects development of oral tolerance 8, 9. Early transplantation experiments revealed that the liver is a unique tolerance-inducing organ 10, 11. Moreover, the liver has been shown to contribute to generation of antigen-specific immune tolerance in situations where soluble exogenous antigens or cell-associated antigens reach the liver with the blood stream (12, 13 and reviewed in 14). Other hepatic cell populations, such as hepatocytes, may also be involved in tolerance induction towards endogenous antigens 15. Induction of T cell tolerance towards exogenous antigens such as oral antigens, however, requires a competent APC that takes up antigen and presents processed peptides in an MHC-restricted fashion to CD4 or CD8 T cells.

It has been demonstrated that oral antigens are presented within the liver by dendritic cells (DC), leading to generation of IL-4-expressing CD4 T cells 16. However, antigen uptake is performed most efficiently by liver sinusoidal endothelial cells (LSEC), which line the hepatic sinusoidal wall and are constantly in contact with portal venous blood. LSEC are known for their professional scavenger function, thereby clearing small molecules from portal venous blood 17. Previously, we have shown that LSEC combine their scavenger function with potent antigen-presenting capacity not only for CD4 18, but also for CD8 T cells 12. Presentation of exogenous antigens on MHC class I molecules to CD8 T cells, termed cross-presentation, is a function initially thought to be restricted to professional APC 19, but it is also proficiently carried out by LSEC. LSEC bear the molecules necessary for cognate interaction with T cells, and are capable of cross-priming of naive CD8 T cells 12. Importantly, cross-priming by antigen-presenting LSEC leads to antigen-specific induction of CD8 T cell tolerance 12. Thus, LSEC contribute by virtue of their extraordinary scavenger function and their immune-regulatory capacity to the unique tolerance-inducing hepatic phenotype.

Here, we address the question whether scavenger liver endothelial cells participate in shaping the immune response towards oral antigens. We provide evidence that oral antigens are cross-presented by LSEC in the liver to CD8 T cells. Denaturing conditions typically encountered by oral antigens, such as exposure to low pH, increase efficiency of LSEC-mediated cross-presentation. As a result preferential presentation of antigens by LSEC rather than DC is observed. Importantly, cross-presentation of oral antigens by LSEC leads to induction of CD8 T cell tolerance, assigning a role for the liver in oral CD8 T cell tolerance.


Oral antigens are cross-presented by LSEC

The first organ encountered by oral antigens, which have entered the blood circulation, is the liver, which receives blood draining from almost the entire gastrointestinal tract via the portal vein. Elimination of molecules from portal venous blood in the liver is achieved by the extraordinary scavenger activity of the LSEC and Kupffer cells 17, both of which have antigen-presenting capacity. It was recently demonstrated that oral antigens are readily detected in the liver a few hours after ingestion 16. To determine whether hepatic cell populations were also able to cross-present oral antigens via MHC class I molecules, we applied OVA to C57BL/6 mice and isolated LSEC and Kupffer cells 2 h later. Isolation of LSEC via portal vein perfusion, enzymatic digestion, density gradient centrifugation and centrifugal elutriation yielded highly pure cell populations (⩾98% as judged by uptake of AcLDL and binding of the endothelial cell-specific antibody ME9F1) (Fig. 1a). Absence of CD11c+ cells in LSEC cell preparations confirmed the absence of contaminating DC (Fig. 1a). LSEC cross-primed naive OVA-specific H-2Kb-restricted CD8 T cells (OT-I) as measured by cytokine expression (Fig. 1b). Kupffer cells isolated from these mice failed to stimulate OVA-specific T cells (data not shown), which is consistent with our previously reported observation that Kupffer cells are less efficient APC than LSEC with regard to cross-presentation of soluble antigens 12. These results demonstrate that oral antigens quickly reach the liver and are cross-presented by organ-resident LSEC to CD8 T cells.

Figure 1.

(A) LSEC were isolated as described in methods. Purity of LSEC was determined by uptake of AcLDL-Alexa-488 (5 µg/mL for 30 min) and staining for the endothelial cell specific antibody ME9F1 (kindly provided by A. Hamann, DRFZ, Berlin). Biotinylated ME9F1 was detected with streptavidin-allophycocyanin (1:200). LSEC cell preparations were stained with mAb specific for CD11c+ to exclude contamination with DC. LSEC cell preparations yielded more than 98% pure LSEC. (B) C57BL/6 mice were fed OVA for 1 day, and LSEC were isolated and incubated with OVA-specific H-2Kb–SIINFEKL-restricted B3Z cells. After 24 h, IL-2 concentrations in cell culture supernatants were determined by ELISA.

Mechanisms of cross-presentation in LSEC

The molecular mechanisms underlying cross-presentation have recently been elucidated in DC revealing a role for phagosome-ER fusion 20. Certain antigen modifications, such as immune complexation of antigen, improve cross-presentation in B cells as well as DC 21, 22, presumably as a result of receptor-mediated antigen uptake that may modify cellular localization of antigen compared to uptake by (macro)pinocytosis. However, immune complexation of antigen did not increase cross-presentation by LSEC any more (Fig. 2a), although LSEC express Fc receptors (data not shown and 23). Uptake of OVA into LSEC occurred exclusively by receptor-mediated endocytosis and involved mainly mannose and scavenger receptors (data not shown). Consequently, cross-presentation in LSEC critically depends on receptor-mediated endocytosis because blockade of mannose and scavenger receptors but not inhibition of macropinocytosis down-regulated cross-presentation (data not shown). Exposure of OVA to heat (90°C for 10 min) improved cross-presentation in LSEC to H-2Kb–SIINFEKL-specific CD8 T cells (Fig. 2a). The difference between heat-treated and native OVA persisted over a large range of antigen concentrations tested (Fig. 2b). Treatment of OVA with low pH (pH 3 for 10 min), likely to be encountered during gastric passage of food, improved cross-presentation by LSEC in a similar fashion (Fig. 2b). To exclude that improved cross-presentation was the result of increased antigen uptake, we heat-treated fluorochrome-labeled OVA and quantified uptake of native or heat-treated antigen into LSEC by flow cytometry. Clearly, heat treatment of antigen did not result in increased antigen uptake by LSEC (Fig. 2c). To investigate the relevance of heat treatment for antigen presentation in vivo, we injected mice i.v. with different concentrations of native or heat-treated OVA, subsequently isolated LSEC as well as splenic DC, and determined cross-presentation of these APC populations ex vivo. As expected, injection of decreasing concentrations of OVA led to a decline in cross-presentation by both LSEC and DC (Fig. 2d). In comparison to in vitro experiments, lower levels of IL-2 were observed in these experiments, which is presumably related to different kinetics of antigen uptake and degradation in LSEC in vivo. The injection of heat-treated OVA, however, dramatically improved cross-presentation by LSEC but not by DC (Fig. 2d). These results support the notion that modified antigens are preferentially cross-presented in vivo by LSEC but not DC.

Figure 2.

(A) LSEC isolated from C57BL/6 mice as described were incubated for 3 h with OVA (10 nM) treated in different ways: native, immune-complexed with anti-OVA-Ig or heat treated (90°C for 10 min). H-2Kb-restricted cross-presentation to B3Z cells was determined as described in methods. (pOVA – OVA heat-treated ⩽0.001) (B) Native OVA, heat-treated OVA (90°C for 10 min) or low pH-treated OVA (pH 3 for 15 min) was titrated on LSEC and cross-presentation was determined. (C) LSEC were incubated for 30 min with native or heat-denatured Tx-Red-labeled OVA (1 µM), detached from the culture disc with EDTA and Tx-red fluorescence in LSEC was determined by flow cytometry. (D) Different concentrations of native or heat-treated OVA were i.v. injected into C57BL/6 mice. APC populations were isolated (LSEC by differential centrifugation and centrifugal elutriation; splenic CD11c+ DC by immuno-magnetic separation) and equal numbers of cells were used for determination of H-2Kb-restricted cross-presentation ex vivo to B3Z cells (pLSEC – DC for heat treated OVA =0.011).

Mechanistically, improved cross-presentation of heat-treated antigen by LSEC required proteasomal processing, as demonstrated by complete inhibition of cross-presentation of native and heat-treated OVA through the proteasome inhibitor lactacystin (Fig. 3a), excluding the possibility of final endosomal processing of heat-treated OVA and class I peptide loading in this compartment. Blockade of the mannose receptor through mannan or of the scavenger receptor through polyinosinic acid yielded a reduction in cross-presentation, supporting the notion that receptor-mediated endocytosis is the most important pathway used by LSEC for antigen acquisition. Macropinocytosis is not used by LSEC for antigen ingestion, because blockade of macropinocytosis through dimethylamiloride (DMA) did not affect antigen uptake (data not shown) nor cross-presentation (Fig. 3b). Complete inhibition of cross-presentation in LSEC by Brefeldin A provides further evidence for involvement of the classical TAP-dependent pathway for antigen processing (data not shown). Moreover, endosomal maturation is required for cross-presentation in LSEC. Inhibition of the vacuolar proton pump (v-ATPase) through bafilomycin or concanamycin, or prevention of endosomal acidification by NH4Cl affected cross-presentation in LSEC (Fig. 3c). At early time points after antigen uptake (15–30 min), we detected OVA in vesicular compartments where it co-localized with heat-shock cognate (hsc)73 (Fig. 3d). More direct evidence for involvement of heat-shock proteins in cross-presentation came from experiments where a heat-shock protein inhibitor (desoxyspergualin, DSG) as well as antibodies to hsc73 blocked cross-presentation in LSEC (Fig. 3e). Antibody-mediated blockade of heat-shock proteins did not affect antigen-uptake nor peptide-binding to MHC I (Fig. 3e), which attributes a role for heat-shock proteins in transport and processing of endocytosed antigen in LSEC. Heat-shock proteins bind and release substrates in an ATP-dependent fashion 24. We reasoned, therefore, that energy was required for binding of OVA to heat-shock proteins in an endosomal compartment. Indeed, depletion of ATP from endosomal compartments by pre-incubation of LSEC with the ATP-hydrolyzing enzyme apyrase resulted in profound reduction of cross-presentation (Fig. 3e). It is unlikely that apyrase was transferred into the cytoplasm in an intact form. Rather, we believe that apyrase hydrolyzed endo-lysosomal ATP required for chaperon function of hsc73. Likewise, a decrease in cross-presentation by the metabolic inhibitor sodium azide was observed in LSEC that exceeded the influence of peptide binding to MHC class I molecules (Fig. 3f). Taken together, these results demonstrate that receptor-mediated endocytosis, endosomal maturation and binding to heat-shock proteins as well as proteasomal processing are essential steps in efficient cross-presentation by LSEC.

Figure 3.

(A) LSEC were incubated with lactacystin (4 µM) or mock-treated for 30 min before exposure to OVA (10 µM) or heat-denatured OVA (100 nM) for 3 h. LSEC were fixed with glutaraldehyde and H-2Kb-restricted cross-presentation to B3Z cells was determined. Inset shows peptide controls. (B) LSEC were separately incubated with an inhibitor of macropinocytosis (DMA, 100 µM), a competitive inhibitor of the mannose receptor (mannan 10 mg/mL) and a non-competitive inhibitor of the scavenger receptor (PIA, 100 µg/mL) for 30 min before incubation with OVA (10 nM) for 3 h or SIINFEKL (40 nM). H-2Kb-restricted presentation to B3Z in comparison to untreated cross-presenting LSEC was determined by quantitation of IL-2 release (pmock - Mannan =0.02; pmock -PIA =0.034). (C) During 3 h of exposure to OVA (10 nM) or SIINFEKL (40 nM) LSEC were incubated with concanamycin (1 nM), bafilomycin (1 µM) or NH4Cl (1 mM) to demonstrate relevance of endosomal maturation. H-2Kb-restricted presentation to B3Z in comparison to untreated cross-presenting LSEC was determined by quantitation of IL-2 release (pmock – concanamycin =0.04; pmock – bafilomycin =0.019; pmock – NH4Cl =0.02). (D) LSEC were incubated with Tx-Red OVA (10 µg/mL) for 15 min and were then incubated with anti-hsc73 mAb (13D3) or an isotype control antibody. Bound antibody was detected using Alexa 488-conjugated chicken anti-mouse Ab (Molecular Probes). Cells were analyzed by confocal laser scanning microscopy (Leica TCS). Scale bar denotes 2 µm. (E) LSEC were incubated with either OVA (10 nM) or SIINFEKL (40 nM) together with mAb against hsc73 (25 µg/mL) or the hsc73 inhibitor DSG (20 µg/mL). H-2Kb-restricted presentation to B3Z in comparison to untreated cross-presenting LSEC was determined by quantitation of IL-2 release (pmock – DSGcross-presentation =0.035; pmock – 13D3 cross-presentation =0.021). Control mouse IgM (25 µg/mL) did not modify cross-presentation. Uptake of Tx-Red OVA was determined by flow cytometry. (F) LSEC were incubated with OVA (10 nM) and treated simultaneously with sodium azide (100 ng/mL) or apyrase (1 U/mL). H-2Kb-restricted cross-presentation of glutaraldehyde-fixed LSEC to B3Z cells was determined after 3 h (p⩽0.05). Incubation of azide-, apyrase- or mock-treated LSEC with SIINFEKL peptide (20 nM) was performed to exclude a nonspecific decrease in H-2Kb-restricted antigen-presentation (see inset).

LSEC cross-tolerize CD8 T cells towards oral antigens

Given the extraordinary scavenger activity and preferential presentation of modified antigens by LSEC, we addressed the question whether cross-presentation of oral antigens by LSEC leads to tolerization of CD8 T cells. We used adoptive transfer of LSEC, which is known to result in orthotopic implantation of LSEC into the hepatic sinusoid 12. LSEC from mice fed OVA were isolated and adoptively transferred into T cell-deficient RAG2–/– knockout mice that were previously reconstituted with OVA-specific TCR transgenic CD8 T cells from OT-I mice (RAG2–/–/OT-I). Functionally, transfer of OVA-loaded LSEC resulted in reduced IFN-γ expression in CD8 T cells after antigen-specific restimulation ex vivo (Fig. 4a). More importantly, OVA-specific CD8 T cells in animals receiving LSEC from OVA-fed mice showed reduced specific cytotoxicity, as determined by in vivo lysis of target cells injected into recipient mice (Fig. 4b). In contrast, transfer of OVA-loaded splenic CD11c+ cells did not decrease IFN-γ expression ex vivo nor specific cytotoxicity of CD8 T cells in vivo (Fig. 4a, b).

Figure 4.

LSEC from C57BL/6 mice fed with OVA or unfed control mice were adoptively transferred into RAG2–/– mice reconstituted with CFSE-labeled OT-I CD8 T cells (RAG2–/–/OTI). At 10 days after adoptive transfer of equal numbers of OVA-loaded LSEC or DC into RAG2–/– mice reconstituted previously with naive OT-I CD8 T cells, OVA-specific immune responses were determined. (A) Similar numbers of CD8 T cells were isolated from spleens and 105 CD8 T cells were re-stimulated with RMA-OVA cells ex vivo. IFN-γ concentration in cell culture supernatant was determined after 18 h by ELISA (pDC – LSEC =0.018). (B) Alternatively, OVA-specific in vivo cytotoxicity was determined as described in methods (pDC – LSEC =0.03).

To further determine the role of LSEC in the CD8 T cell response to oral antigens, we generated a new transgenic mouse line expressing the MHC class I molecule H-2Kb under control of the tie2 promoter. Sinusoidal cells expressed H-2Kb in a similar fashion to wild-type H-2b mice (Fig. 5a). Livers of DBA mice did not show any reactivity with the antibody to H-2Kb. LSEC isolated from tie2-H-2Kb mice express H-2Kb surface levels comparable to C57BL/6 mice and bear similar surface levels of CD54 and CD106 (Fig. 5b). LSEC bear the co-stimulatory molecules necessary to initiate priming of naive T cells as shown by substantial proliferation of naive CD8 T cells after contact with antigen-presenting LSEC (Fig. 5c). Moreover, LSEC from tie2-H-2Kb mice cross-present OVA ex vivo, albeit at lower levels than LSEC from C57BL/6 mice (Fig. 5d). Importantly, LSEC from tie2-H-2Kb mice cross-present OVA after isolation from OVA-fed mice to OVA-specific CD8 T cells (Fig. 5e), clearly demonstrating their ability to participate in shaping the immune response towards oral antigens. However, the tie2 regulatory sequence directs expression to endothelial as well as hematopoietic cells. For exclusive H-2Kb expression on endothelial cells, including LSEC, we generated bone marrow chimeras by reconstituting tie2-H-2Kb mice with bone marrow from B6.CH-2bm1 mice. APC from B6.CH-2bm1 mice are not capable of presenting SIINFEKL on H-2Kb molecules to CD8 T cells 25. Thus, in (B6.C-H2bm1 → tie2-H-2Kb) chimeric animals only endothelial cells, including LSEC, bear the ability to interact with H-2Kb-restricted CD8 T cells. Importantly, LSEC in chimeric animals were not replaced by endothelial cells from bone marrow (data not shown), which demonstrates liver autonomous repopulation of this cell population as reported previously 26. Feeding of OVA to (B6.C-H2bm1 → tie2-H-2Kb) mice, where only endothelial cells could present the OVA-derived peptide SIINFEKL to H-2Kb-restricted CD8+ T cells, resulted in induction of tolerance in adoptively transferred H-2Kb–SIINFEKL-specific CD8 T cells, as characterized by the lack of OVA-specific cytotoxicity (Fig. 5f). Collectively, these data provide evidence that LSEC induced CD8 T cell tolerance to oral antigens.

Figure 5.

(A) Paraffin-embedded sections of livers from DBA2, H-2b and tie2-H-2Kb mice were stained with R1.21.2 for H-2Kb expression using APAAP stain. Scale bars 100 µm. (B) LSEC were isolated from tie2-H-2Kb or H-2b mice and stained for surface expression levels of H-2Kb, CD54 and CD106. (C) Antigen presentation to naive Des-TCR CD8 T cells was determined in vitro for LSEC isolated from tie2-H-2Kb, DBA-2 and H-2b mice. Proliferation of naive CD8 T cells was determined by [3H]thymidine incorporation. (D) At 1 h after i.v. injection of OVA, LSEC from C57BL/6, tie2-H-2Kb or CBA mice were isolated and H-2Kb-restricted cross-presentation of OVA to B3Z cells was examined. (E) LSEC were isolated from tie2-H-2Kb 1 day after feeding OVA and cross-presentation to B3Z cells was determined. (F) Chimeric (B6.C-H2bm1→tie2-H-2Kb) mice were generated, where H-2Kb-restricted presentation of SIINFEKL was limited to endothelial cells. Animals received 5 × 106 naive OT-I CD8 T cells by adoptive transfer at day 0 and then received OVA by i.v. injection or were fed OVA. Specific cytotoxicity in vivo was determined after 10 days as described.


The ingestion of antigens via the oral route leads to antigen-specific immune tolerance. The development of immune tolerance has been reported to depend on the GALT, especially the mesenteric lymph nodes 2, and is characterized predominantly by generation of regulatory CD4+ T cells through tolerogenic DC 1, 27, 28. However, oral antigens within minutes gain access to mesenteric blood vessels 3, 5, 7, which drain into the liver via the portal vein. Work from independent laboratories suggested a role for the liver in oral tolerance 4, 8, 9. Tolerance induction in this situation was linked to hepatic non-parenchymal cells 29. Further studies attributed tolerance-promoting properties to Kupffer cells 30 and to DC, which induced IL-4-expressing CD4 T cells 16. Here we assign an important new role for the liver in oral tolerance, i.e., uptake of oral antigens and induction of immune tolerance in CD8 T cells by cross-presenting liver sinusoidal endothelial cells.

The liver is known for its capacity to remove blood-borne macromolecules from the circulation. The clearance function of the liver can be attributed mainly to two cell populations, Kupffer cells and scavenger LSEC. LSEC are most efficient in receptor-mediated uptake of macromolecules from portal venous blood 17, and represent a unique organ-resident APC population 12, 18. Here, we demonstrate that LSEC efficiently scavenge antigens that were originally ingested via the oral route and cross-present these antigens to CD8 T cells. Given the low reported efficiency for transport of oral antigens into the blood 3, it is astonishing to find presentation of antigen by so-called non-professional APC in the liver, such as LSEC. How is this efficient cross-presentation achieved by LSEC?

Cross-presentation in DC or B cells is clearly improved if antigen is taken up by receptor-mediated endocytosis and not by fluid-phase endocytosis 21, 22. Antigen-uptake in LSEC occurs almost exclusively by receptor-mediated endocytosis, and in case of the model antigen OVA involves both mannose and scavenger receptors. As endocytosis rates for the mannose receptor are among the fastest measured for any surface receptor 17, and as preventing mannose-receptor-mediated uptake diminishes cross-presentation in LSEC (data not shown), we assume that efficient cross-presentation is closely linked to scavenger function in these cells. Importantly, antigen modification by treatment with heat or low pH increased efficiency of cross-presentation in LSEC, but not in splenic DC (Fig. 2). This may lead to preferential presentation of oral antigens by LSEC rather than by splenic DC, especially at low antigen concentrations likely to be present in portal venous blood. Although the exact mechanisms underlying increased cross-presentation efficiency in LSEC for denatured antigen still remain unclear, it is evident that endosomal maturation and proteasomal processing of denatured antigen are involved. Time-lapse confocal microscopy of LSEC revealed rapid degradation of OVA-DQ in endosomal vesicles, as evidenced by generation of fluorescent proteolytic cleavage products within 15 min after exposure to antigen (data not shown). Proteolytic cleavage products of OVA generated in the endosomal compartment may reach the cytoplasm through the Sec61 20, 31 or the p97 channel 32. Since hsc73 enhances MHC class II-restricted antigen presentation 33, and blockade of endo-lysosomal hsc73 in LSEC results in inhibition of cross-presentation (Fig. 3d), it seems reasonable to assume that improved cross-presentation of denatured antigens in LSEC involved hsc73. Interestingly, heat-shock proteins have been demonstrated to convert the immune response from T cell tolerance to autoimmunity by improving cross-presentation by DC 34, and activating DC in an adjuvant-like fashion 35. In contrast, LSEC are resistant to changes in their tolerogenic phenotype even after incubation with TLR ligands (A. L. and P. K., unpublished). The presence of hsc73 in endo-lysosomal compartments has already been described 36, although the mechanisms of how this compartment is loaded with hsc73 remain unsolved. LSEC appear to contain constitutively sufficient amounts of heat-shock proteins in endo-lysosomal compartments to support their APC function. As LSEC are most efficient scavenger cells and are known to take up small liposomes (⩽100 nm) from the blood 37, it is possible that LSEC take up blood-borne exosomes by similar mechanisms. Exosomes are rich in heat shock proteins (hsc73), originate from APC 38 or intestinal epithelial cells 39, and are transported to the liver with portal venous blood. Collectively, our data imply that the anatomical location of the liver in combination with extraordinary scavenger function of LSEC and efficient antigen-processing in LSEC assure preferential cross-presentation of denatured antigen by LSEC rather than by splenic DC, especially when antigen concentration is limited. However, we do not exclude a contribution of hepatic immature DC in mediation of oral tolerance in liver-draining lymph nodes.

Presentation of antigen by LSEC to naive CD4 or CD8 T cells leads to subsequent induction of tolerance, as evidenced by absence of Th1 differentiation for CD4 T cells 18, absence of effector cytokine expression, and absence of specific cytotoxicity in CD8 T cells 12. We wondered whether minute concentrations of blood-borne antigen, such as those concentrations likely to be encountered during blood-borne distribution of oral antigens, would be sufficient to lead to LSEC-mediated T cell tolerance. We addressed this question using two different experimental systems. LSEC from OVA-fed mice were adoptively transferred into T cell-deficient mice (RAG2–/–) that were previously reconstituted with OVA-specific naive CD8 T cells from TCR-transgenic OT-I mice. In these mice, CD8 T cell tolerance developed after transfer of LSEC from OVA-fed animals. We defined CD8 T cell tolerance as absence of expression of IFN-γ following antigen-specific restimulation ex vivo (Fig. 4a), and, more importantly, as absence of antigen-specific cytotoxicity in vivo (Fig. 4b). In these models, we can not formally exclude the possibility that recipient APC populations contributed to the induction of T cell tolerance and that poor T cell priming leads to a weak cytotoxic response. However, the higher expression of IFN-γ in T cells obtained from mock-treated animals, and the presence of significant cytotoxicity in mock-treated animals as compared to mice adoptively transferred with LSEC, rather suggest that transfer of LSEC cross-presenting oral antigens changed the functional repertoire of CD8 T cells.

To confirm the contribution of cross-presenting LSEC in CD8 T cell tolerance towards oral antigens independent of adoptive transfer experiments, we generated a new transgenic mouse line. The tie2-H-2Kb transgenic mice express the MHC class I molecule H-2Kb under control of the vascular promoter tie2. LSEC from these transgenic mice are functional as APC, similar to LSEC isolated from H2b mice. To exclude any participation in MHC class I-restricted antigen presentation of bone marrow-derived APC in tie2-H-2Kb transgenic mice, we generated bone marrow chimeras employing B6.C-H2bm1 mice as donors. Cells from B6.C-H2bm1 mice are unable to engage in cognate H-2Kb-restricted interaction with CD8 T cells 25. Accordingly, in (B6.C-H2bm1 →tie2-H-2Kb) chimeric animals only endothelial cells were capable of H-2Kb-restricted interaction with CD8 T cells. Feeding of OVA to these chimeric animals resulted in absence of cytotoxicity in OVA-specific CD8 T cells, demonstrating that cross-presentation of oral antigens by endothelial cells contributed to CD8 T cell tolerance. Although these data suggest that LSEC are sufficient to induce CD8 T cell tolerance towards oral antigens, this does not prove that LSEC are essential for induction of tolerance. Furthermore, we cannot formally exclude participation of endothelial cells from organs other than the liver in induction of CD8 T cell tolerance. However, this seems unlikely because endothelial cells from other organs require an initial pro-inflammatory stimulus, such as contact with IFN-γ or TNF-α, to function as APC for naive T cells 4043, and because the capacity for cross-presentation has only been demonstrated for “professional” APC types, such as DC, macrophages, B cells and LSEC.

The extraordinary scavenger function in combination with preferential cross-presentation of antigens previously exposed to heat or low pH renders LSEC functional in generating CD8 T cell tolerance towards oral antigens. Further observations support the importance of preferential cross-presentation of oral antigens by LSEC for oral tolerance. Orally ingested antigens necessarily come in contact with low pH for considerable periods of time during gastric passage. It is therefore of interest to note that (i) micro-encapsulation of antigen, which prevents antigen contact with the gastric environment 44, and (ii) presence of large amounts of antibodies to oral antigens, which leads to formation of immune complexes and in turn improves antigen presentation by DC, lead to loss of oral tolerance 45. Moreover, diversion of portal venous blood from the liver directly into the systemic circulation prevents development of oral tolerance 8, 9. Along the same line, induction of oral tolerance is impeded in liver cirrhosis, where porto-systemic shunts develop as a consequence of portal hypertension. Liver cirrhosis is further accompanied by reduced scavenger activity of LSEC, increasing concentrations of antibodies against oral antigens and increased circulating levels of gut-derived bacterial degradation products 46. Contact of DC with bacterial degradation products such as LPS results in strong activation of immunity 47. Thus, failure of the liver to clear portal venous blood from oral antigens and bacterial degradation products results in systemic immunity against oral antigens. Interestingly, LPS challenge of LSEC does not modify induction of cross-tolerance in CD8 T cells, because LSEC do not show a change in their immune-regulatory phenotype (A.L. and P.K., unpublished) despite their constitutive expression of TLR4 48. During physiological situations, where LSEC are exposed to low levels of LPS in portal venous blood 49, 50 and contribute to LPS clearance 51, LSEC maintain their immune-regulatory potential. Contact with LPS may even license LSEC to engage in cognate interaction with T cells and induce immune tolerance, because animals kept under sterile, LPS-free conditions do not show endothelial cell expression of adhesion molecules relevant for T cell interaction in the liver 52 and fail to develop oral tolerance 53.

Most important to induction of oral tolerance is the generation of regulatory CD4 T cells through tolerogenic DC in the GALT 1. However, the rapid systemic distribution of orally ingested antigens via the blood stream requires a secure fail-safe mechanism. Indeed, in some cases oral ingestion of antigen was followed by induction of immunity and antigen-specific autoimmunity 53. Owing to its anatomic localization, enormous scavenger activity and physiological metabolic function the liver is involved in the induction of oral tolerance once oral antigens reach the portal-venous circulation. Thus, immune functions of the gastro-intestinal tract and the liver appear to be closely linked to anatomy and physiological metabolic function 1, 14. Together with other APC populations of the liver 16, 30, cross-presenting LSEC induce tolerance towards oral antigens in CD8 T cells, and thus ensure absence of inadvertent immune reactivity towards nutrient antigens that are metabolized in the liver.

Materials and methods

Reagents and antibodies

The peptide SIINFEKL (OVA aa 257–264) was synthesized by the bioengineering department at ZMBH (Heidelberg, Germany). OVA (grade II and VII), azide, apyrase, glutaraldehyde, glycine and lactacystin were obtained from Sigma (Deisenhofen, Germany). To eliminate LPS contaminations, OVA was passed over polymyxin-B columns (Pierce, Bonn, Germany). Lactacystin was purchased from ICN (Eschwege, Germany), saccharose from Merck (Darmstadt, Germany) and hsc73-specific inhibitor 15-deoxyspergualin (DSG; Nippon Kayaku, Tokyo). CFDA-SE was obtained from Molecular Probes (Leiden, The Netherlands). mAb for IL-2 or IFN-γ ELISA were obtained from BD Biosciences (Heidelberg, Germany) and the assay was performed according to the manufacturer's instruction. Anti-CD8α (53–6.7), anti-CD11c (HL3), anti-IFN-γ (XMG1.2), anti-TCRβ (H57–597), anti-Vα2 (B20.1), anti-Vβ5 (MR9–4) mAb, conjugated to different fluorochromes or biotin, and streptavidin conjugates were purchased from BD Biosciences. Anti-hsc73 mAb (3D3) and control IgM were purchased from Dianova (Hamburg, Germany), anti-ME-9F1 mAb was kindly provided by Alf Hamann (DRFZ, Berlin). AcLDL conjugated with Alexa 488 was obtained from Molecular Probes.

Mice and cell lines

Mice received humane care and were maintained under SPF conditions at the animal facilities of the ZMBH (Heidelberg, Germany), the DKFZ (Heidelberg, Germany) or the IMMEI (Bonn, Germany) according to German guidelines for animal care, and experiments were performed according to animal experimental ethics committee guidelines. C57BL/6 (H-2b), CBA (H-2k) and B6.C-H2bm1 mice were purchased from Jackson Laboratories (Bar Harbor, ME). OVA257–264-specific MHC class I-restricted T cell receptor transgenic animals (OT-I) bearing Vα2/β5 CD8 T cells 54, and RAG2–/–55 mice have been described previously. Transgenic tie2-H-2Kb mice, expressing H-2Kb under tie2-promoter control, were generated according to Constien et al. 56. Briefly, the H-2Kb gene was cloned as a 5.5-kb EcoRI fragment blunt end into the HindIII and EcoRV sites of the pTie-2–2.5 plasmid. The resulting plasmid was cut with XhoI and a 7.4-kb fragment containing the Tie-2 promoter and the H-2Kb gene was cloned blunt end into the SmaI site of the pTie-2–10.5 plasmid harboring the first intron sequences of the Tie-2 gene. This construct was used for pronuclear injections according to standard protocols. The Tie-2 regulatory sequences were kindly provided by T. N. Sato (University of Texas, Dallas, TX). Mice were kept on a DBA background. For generation of chimeric animals, bone marrow cells were depleted of T lymphocytes and 5 × 106 cells were intravenously transferred into irradiated (900 rad) recipient mice. Chimeric animals were analyzed 8–12 weeks later for bone marrow engraftment. The OVA257–264-specific CD8 T cell hybridoma B3Z was kindly provided by N. Shastri. The RMA tumor cell line transfected with OVA (RMA-OVA) was described previously 12.


In vitro cross-presentation assays were performed as described previously 12. APC (either LSEC 105 cells/well in collagen-I-coated 24-well plates or mature bone marrow-derived DC 105 cells/well in 96-well round-bottom plates) were pulsed with OVA at the indicated concentrations for 2 h, washed thoroughly to remove free OVA, and B3Z cells were added (105 cells/well in 96-well plates and 2 × 105 cells/well in 24-well plates) for 18 h. As a control for integrity of surface MHC class I molecules and cell viability, MHC class I molecules were loaded with peptide SIINFEKL (OVA257–264). Activation of B3Z was measured as release of IL-2 into the cell culture supernatant by specific Sandwich-ELISA. Where indicated, LSEC were fixed after OVA pulse with ice-cold glutaraldehyde (0.05%) for 2 min followed by quenching with glycine (0.2 M in PBS).

Oral delivery of OVA

For in vivo stimulation, OVA free of LPS was either injected i.v. (450 μg per mouse) or fed using OVA pellets. OVA pellets were generated with 500 g OVA grade II (crude egg white) mixed with 100 g saccharose and water and crisped up until dried. Estimated OVA uptake from this OVA-enriched diet exceeds 5 mg per animal per day. For antigen-specific restimulation in vitro, RMA-OVA cells (5 × 104 cells) were incubated with 105 CD8 T cells.

In vivo cytotoxicity assay

In vivo cytotoxicity assays were performed as described 57. In brief, spleen cells were either pulsed with SIINFEKL peptide (1 µg/mL, 45 min at 37°C) and labeled with a high concentration of CFSE (1 μM, 15 min at 37°C; CFSEhigh cells) or were mock treated and labeled with a low concentration of CFSE (0.1 μM; CFSElow cells). Cells were washed twice with PBS and equal numbers of cells were injected i.v. (1 × 107 target cells). Animals were killed 4 h later and target cells in spleen were analyzed with FACSCalibur or LSR II (BD Bioscience, Heidelberg, Germany). Data were processed with CellQuest Pro (BD Bioscience) or Flow Jo (Tri Star, San Carlos, USA) software. To calculate specific lysis of the in vivo cytotoxicity assay, the following formula was used: % specific cytotoxicity = 100 – [ratio (CFSEhigh/CFSElow)primed × 100/ratio (CFSEhigh/CFSElow)control].

Cell isolation procedures and adoptive transfer of cells

LSEC, Kupffer cells 18 and lymphocytes 58 from murine liver were isolated and cultured as described. Orthotopic implantation of adoptively transferred LSEC (5 × 106 cells/mouse) has been reported previously 12. Splenic CD8 OT-I T cells were purified by MACS (Miltenyi Biotech, Bergisch-Gladbach, Germany) according to the manufacturer's instruction and 5 × 106 CD8 T cells were injected i.v. into recipient mice. DC were obtained after organs were digested with collagenase D (Roche Diagnostics, Mannheim, Germany) and DNase I (Boehringer, Mannheim, Germany). DC/B> were isolated by positive selection with anti-CD11c-conjugated MACS beads (Miltenyi Biotech) and 5 × 106 cells were i.v. injected into recipient mice.

Statistical analysis

Results are expressed as mean ± SEM. Analysis was performed using Student's t-test.


The authors thank A. Hamann for providing ME9F1 mAb. This work was supported by grants from the DFG and the Volkswagenstiftung to P.A.K., B.A. and G.J.H. We would like to acknowledge the assistance of the Flow Cytometry Core Facility at the Institute of Molecular Medicine and Experimental Immunology, University of Bonn, supported in part by grant HBFG-109–517.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5