Systemic antigen cross-presented by liver sinusoidal endothelial cells induces liver-specific CD8 T-cell retention and tolerization


  • Potential conflict of interest: Nothing to report.


Peripheral CD8 T-cell tolerance can be generated outside lymphatic tissue in the liver, but the course of events leading to tolerogenic interaction of hepatic cell populations with circulating T-cells remain largely undefined. Here we demonstrate that preferential uptake of systemically circulating antigen by murine liver sinusoidal endothelial cells (LSECs), and not by other antigen-presenting cells in the liver or spleen, leads to cross-presentation on major histocompatibility complex (MHC) I molecules, which causes rapid antigen-specific naïve CD8 T-cell retention in the liver but not in other organs. Using bone-marrow chimeras and a novel transgenic mouse model (Tie2-H-2Kb mice) with endothelial cell-specific MHC I expression, we provide evidence that cross-presentation by organ-resident and radiation-resistant LSECs in vivo was both essential and sufficient to cause antigen-specific retention of naïve CD8 T-cells under noninflammatory conditions. This was followed by sustained CD8 T-cell proliferation and expansion in vivo, but ultimately led to the development of T-cell tolerance. Conclusion: Our results show that cross-presentation of circulating antigens by LSECs caused antigen-specific retention of naïve CD8 T-cells and identify antigen-specific T-cell adhesion as the first step in the induction of T-cell tolerance. (HEPATOLOGY 2009.)

In the absence of inflammation, circulating antigens, such as autoantigens or innocuous food antigens, induce peripheral immune tolerance rather than immunity.1 Likewise, circulating viral antigens skew virus-specific T-cell immune responses promoting viral immune escape.2 Immune regulation is achieved by different means, including regulatory CD4 T-cells, or direct tolerization of CD8 T-cells after contact with immature antigen-presenting cells (APCs).3, 4 Antigen clearance from the blood, however, is mainly achieved by scavenger cell populations in spleen and liver, by macrophages, microvascular endothelial cells, and only to some degree by dendritic cells (DCs).5 The liver is particularly prominent in scavenger function because of its physiological function in carbohydrate, protein and lipid metabolism, and removal of toxic and degradation products. Moreover, the liver has immune-regulatory functions.6, 7

It is not clear, however, how the scavenger cell populations in the liver act as APCs and present exogenous systemic antigens to circulating T-cells, potentially contributing directly to peripheral immune regulation in direct competition with other antigen-presenting cell populations. Presentation of exogenous antigens on major histocompatibility complex (MHC) class I molecules to CD8 T-cells,8 i.e., cross-presentation, is restricted to specific APC.9 Thus, antigen scavenging by hepatic cell populations needs to be accompanied by the capacity to cross-present circulating antigen to CD8 T-cells in order to establish immune-regulatory function. Several studies have investigated CD8 T-cell tolerance toward circulating antigens. However, these studies used minimal MHC-binding peptides2 or endogenous antigens,10, 11 and thus failed to consider that exogenous circulating protein antigens require cross-presentation for recognition by CD8 T-cells. We have previously reported that LSECs upon adoptive transfer can cross-present antigen to naïve CD8 T-cells initiating T-cell tolerance rather than immunity.12 However, the initial steps in CD8 T-cell tolerance in vivo, i.e., how tolerogenic LSECs establish contact with circulating naïve CD8 T-cells, remained unclear. Here, we provide evidence that cross-presentation of circulating antigens by LSECs is sufficient to retain antigen-specific circulating naïve CD8 T-cells. Thus, antigen-specific adhesion to LSECs in the absence of inflammation represents the first step in T-cell tolerance toward systemic antigens.


Ab, antibody; AcLDL, acetylated low-density lipoprotein; APC, antigen-presenting cell; BM, bone marrow; DC, dendritic cell; LN, lymph node; LSEC, liver sinusoidal endothelial cell; MHC, major histocompatibility complex; OVA, ovalbumin.

Materials and Methods

Mice and Bone-Marrow Chimeras.

C57Bl/6J, B6.C-H2bm1 (H-2Kbm1 does not bind the SIINFEKL-peptide), H-2KbSIINFEKL-restricted T-cell receptor (TCR)-transgenic animals (OT-1), and CD54−/− mice were maintained under specific pathogen-free (SPF) conditions. Tie2-H-2Kb mice, expressing H-2Kb under tie2-promoter control, were generated as described13 on a DBA background to assure endothelial cell (EC)-specific H-2Kb-expression.14 Bone marrow (BM) chimeras were generated as follows: 107 CD90+-depleted BM cells were injected into irradiated (9 Gy) mice. After 6 weeks, BM reconstitution was determined with anti-H-2Kb specific 5-F-I that fails to recognize Kbm-1.

Cell Isolation.

LSECs were isolated and cultured as described.12, 15 Hepatic stellate cells were isolated as described16 and visualized at 355 nm excitation yielding an autofluorescent signal at 425-475 nm. DCs were isolated from collagenase-digested spleens with CD11c MACS-beads. Lymphocytes were isolated from spleen and inguinal lymph nodes (LNs) by mechanical separation; liver and lung lymphocytes were isolated by collagenase digestion and Percoll (GE Healthcare, Sweden) gradient centrifugation.

Flow Cytometry.

Cells were stained with fluorochrome-conjugated antibody (Ab) for 15 minutes at 4°C in the presence of FcγR blocking Ab (2.4G2); biotin-conjugated antibodies were detected with streptavidin conjugates. Dead cells were excluded by 7-AAD (Invitrogen, Germany) or Hoechst 33258 staining. Intracellular cytokine staining was performed according to the manufacturer's recommendation (BD Bioscience, Germany). Cells were analyzed with a FACS Calibur or Canto II (BD Bioscience) and data were processed with Flow Jo (Tree Star, CA) software. Antibodies and secondary reagents were purchased from BD Bioscience or eBioscience (San Jose, CA).

In Vitro Adhesion Assays.

OT-1 cells stained with 1 μM carboxy-fluorescein-succinimidyl-ester (CFSE, Invitrogen, Germany) for 7 minutes at 37°C were cultured for 4 hours with LSECs. Nonadherent cells were washed away and adherent cells harvested with 2 mM EDTA.

Statistical Analysis.

All experiments were performed at least three times with groups of three mice. Results are expressed as mean ± standard error of the mean (SEM). Statistical significance was calculated using a Student t test (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001).

More experimental procedures are found in the supporting materials online.


Preferential Rapid Uptake of Soluble Antigen by Hepatic Scavenger LSECs In Vivo.

To investigate which hepatic cell populations in vivo take up soluble antigens from the circulation, we intravenously injected fluorescently labeled ovalbumin (OVA). Within minutes circulating OVA was mainly localized in the liver in cells lining the hepatic sinusoids (Fig. 1A), as we have previously shown.12 Although the spleen clears circulating antigens from blood, we failed to detect comparable OVA-uptake by splenic cells (Fig. 1A). Acetylated low-density lipoprotein (AcLDL) was taken up equally well by the same hepatic cell population that internalized OVA, indicating that OVA-scavenging cells were LSECs (Fig. 1B). OVA-positive cells were negative for CD11b+ or CD11c+, excluding Kupffer cells and DC, but were ME9F1+ (Fig. 1C), which further identified endothelial cells. Furthermore, no significant OVA uptake was observed by stellate cells (Fig. 1D), which can function as organ-resident APC.17 Because LSECs are also APCs,12 we further investigated the functional consequences of their interaction with CD8 T-cells.

Figure 1.

LSECs, but not other hepatic cell populations, scavenge circulating soluble antigens. Fifteen minutes after injection of Alexa647-labeled OVA (5 μg), Alexa-488-labeled AcLDL (12 μg) or the combination of both into mice uptake into liver or spleen cells was analyzed by (A) confocal microscopy or (B-D) flow cytometry. (B) PBS-injected animals served as control. (C) Liver cells were stained for ME9F1, CD11b, CD11c, and analyzed by flow cytometry. (D) Antigen uptake by stellate cells was determined by flow cytometry.

Migratory Arrest and Stimulation of Naïve CD8 T-cells on Cross-Presenting LSECs In Vitro.

Within lymphatic tissue, antigen recognition on DC leads to migratory arrest of naïve CD8 T-cells.18 Employing in vitro video microscopy we observed that the interaction of naïve OVA-specific H-2Kb-restricted OT-1 CD8 T-cells with cross-presenting OVA-loaded LSECs induced a rapid migratory arrest of naïve OT-1 cells as determined by their diminished trajectory length and distance (Fig. 2A). This led to increased OT-1 adhesion that correlated directly with the antigen dose (Fig. 2B), suggesting that the numbers of peptide-loaded MHC I molecules determined the extent of T-cell adhesion. The absence of CD54 on LSECs reduced T-cell adhesion (Fig. 2B). Interestingly, LSECs further increased CD54 expression on antigen-specific interaction with T-cells (Fig. 2C), which likely stabilizes subsequent LSEC/T-cell interactions. Furthermore, antigen cross-presentation by LSECs resulted in stimulation and proliferation of naive CD8 T-cells (Fig. 2D), but ultimately T-cells activated by LSECs became tolerant, lacking interleuken-2 (IL-2) / interferon-gamma (IFN-γ) production after TCR-mediated restimulation (Fig. 2E and Ref.12). Although IL-10 is associated with T-cell tolerance, LSEC-tolerized OT-1 T-cells did not produce IL-10 (data not shown). Furthermore, IL-10 did not affect OT-1 T-cell activation or proliferation or their tolerization by LSECs (Supporting Fig. 1). Together, these data indicate that cross-presentation by LSECs initiates antigen-specific immobilization of naïve CD8 T-cells, eventually leading to tolerance.

Figure 2.

Antigen-specific adhesion of naïve CD8 T-cells to cross-presenting LSECs in vitro. (A) Migratory activity of naïve OT-1 cells on cross-presenting LSEC was observed over 1 hour by video microscopy. (B) Adhesion of naïve OT-1 cells to wildtype or CD54−/− LSECs in the presence or absence of OVA after 4 hours. (C) CD54 expression on cross-presenting LSECs after 24 hours of coculture with naïve OT-1 cells (black), or in the absence of T-cells (gray); isotype control (shaded). (D) CD69 expression on OT-1 cells cultured on LSECs or DCs in the presence (black line) or absence (dashed line) of OVA after 24 hours. Isotype control: shaded area. After 48 hours proliferation was measured by CFSE dilution (OVA-black line; PBS-shaded area). (E) After 5 days of coculture with antigen-presenting LSECs, OT-1 cells were restimulated with anti-CD3ϵ-antibody for 18 hours and IL-2/IFN-γ was determined in the supernatant.

Cross-Presentation of OVA In Vivo Leads to Specific Hepatic Retention of Naïve CD8 T-cells.

Naïve CD8 T-cells recognizing a ubiquitously expressed self-antigen10 accumulate in the liver. Therefore, we examined whether there is also antigen-specific recruitment of naïve CD8 T-cells to the liver in response to soluble circulating protein antigen. We found that in OVA-injected mice adaptively transferred naïve OT-1 cells accumulated in the liver within 4 hours (Fig. 3A). Naïve OT-1 cells did not accumulate in phosphate-buffered saline (PBS)-injected mice nor did antigen-nonspecific CD8 T-cells in OVA-injected mice (Fig. 3A), indicating that hepatic cross-presentation is required for antigen-specific CD8 T-cell retention. This antigen-specific T-cell retention occurred rapidly within 2 to 4 hours after adoptive transfer (Fig. 3B), but no antigen-specific accumulation of naïve OT-1 cells in lung, spleen, or LN was observed (Fig. 3B). Intravital microscopy confirmed adhesion of naïve OT-1 cells to sinusoidal liver cells, which occurred rapidly within minutes after adoptive transfer preferentially in the periportal area (data not shown). Counting of adherent OT-1 cells per liver lobule confirmed that naïve CD8 cells were retained antigen-specifically (Fig. 3C). Although CD54 enhanced antigen-specific adhesion to LSECs under static conditions in vitro, we observed only a small but significant decrease in antigen-specific retention of naïve CD8 T-cells in the liver in vivo (Fig. 3D), indicating that CD54/LFA-1 interactions do not play a major role in this process.

Figure 3.

Antigen-specific hepatic accumulation of naïve CD8 T-cells in vivo. (A,B) Equal numbers of CFSEhigh naïve OT-1 cells and CFSElow C57BL/6 CD8 T-cells were injected into (A,B) wild-type mice that received PBS or OVA 2 hours previously. Lymphocytes from liver, lung, spleen, and lymph nodes were analyzed for CD8α and analyzed by flow cytometry. (C) At 4 hours after transfer, total numbers of OT-1 cells were determined by intravital microscopy per hepatic lobule analyzing 20-30 lobules per treatment group. (D) Naïve OT-1 cells were injected into CD54−/− mice and analyzed as described in (A).

Organ-Resident LSECs Mediate Antigen-Dependent Naïve CD8 T-cell Retention in the Liver.

To identify the cell population responsible for antigen-specific T-cell retention in vivo, we generated BM chimeras, in which either organ-resident cells but not BM-derived immune cells were capable of cross-presenting OVA on H-2Kb ([B6.CH-2bm1->C57Bl/6]) or vice versa ([C57Bl/6->B6.CH-2bm1]). Hepatic retention of naïve OT-1 cells was only observed in OVA-injected [B6.CH-2bm1->C57Bl/6] and not in [C57Bl/6->B6.CH-2bm1] chimeras, indicating that organ-resident radiation-resistant APC were involved (Fig. 4A). Again, no antigen-specific T-cell retention was observed in the spleen (Fig. 4B). Using tie2-H-2Kb mice, where H-2Kb is expressed under control of an endothelial cell-specific tie2 promoter,14 we found that retention of OT-1 cells in the liver occurred with the same efficiency as in [B6.CH-2bm1->C57Bl/6] chimeras (Fig. 4C). As expected, after quantifying total OT-1 numbers, higher numbers of naïve T-cells were found in the spleen, but after OVA challenge only in the liver the OT-1 numbers almost doubled (Fig. 4D).

Figure 4.

Antigen-specific naïve CD8 T-cell retention by LSECs. (A,B) Naïve CFSE+ OT-1 cells were transferred into OVA/PBS-challenged BM chimeras. After 4 hours lymphocytes from (A) liver and (B) spleen were isolated, stained for CD8α, and analyzed by flow cytometry. Antigen-specific accumulation of naïve CD8 T-cells was calculated in fold-increase as follows: (%CFSE+ of total CD8α+)OVA / (%CFSE+ of total CD8α+)PBS. (C,D) CFSEhighOT-1 / CFSElow C57BL/6 CD8 T-cells were transferred into (C,D) tie2-H-2Kb animals, previously depleted of macrophages by clodronate-liposomes or (E) splenectomized [B6.CH-2bm1->C57Bl/6] chimeras. (C) Antigen-specific retention and (D) total numbers of OT-1 and C57BL/6 CD8 T-cells were determined in liver and spleen. (E) Antigen-specific OT-1 cell retention was determined in livers of splenectomized [B6.CH-2bm1->C57Bl/6] chimeras. The percent OT-1 or C57BL/6 T-cells within total CD8 T-cells is depicted.

The liver can trap and eliminate activated CD8 T-cells.19, 20 To exclude that T-cells retained in the liver were initially activated in the spleen and then subsequently trapped in the liver, we used splenectomized [B6.CH-2bm1->C57Bl/6] chimeras. Although naïve T-cells could not have been activated in the spleen, hepatic antigen-specific T-cell retention was still present (Fig. 4E). The total numbers of CD8 T-cells in the livers of splenectomized mice were greatly increased (data not shown), explaining the low percentages of adaptively transferred CD8 T-cells in total CD8 T-cells. Together these data demonstrate the unique functional capacity of LSECs to recruit and retain circulating naïve T-cells that are specific for soluble circulating protein antigens.

Functional Consequences of Antigen-Specific Naïve CD8 T-cell Retention in the Liver.

Antigen-specific T-cell retention in the liver was followed by rapid stimulation in situ characterized by rapid induction of CD69 expression on OT-1 cells retained in the liver (Fig. 5A), which at the early timepoints is not observed in spleen, LN, or lung (Fig. 5B). The induction of CD69 expression was due to the recognition of cross-presented antigen on LSECs as CD69 induction was also observed in OVA-injected tie2-H-2Kb mice (Fig. 5C). Induction of CD69 expression on hepatic OT-1 cells also occurred in OVA-challenged splenectomized [B6.CH-2bm1->C57Bl/6] chimeras (Fig. 5D) and, therefore, did not result from prior stimulation in the spleen.

Figure 5.

In situ stimulation of naïve CD8 T-cells in the liver. (A,B) CD69 expression levels were determined on CFSEhighOT-1 / CFSElow C57BL/6 CD8 T-cells under experimental conditions described in Fig. 4 after (A) 4 hours or (B) the indicated timepoints. (C,D) CD69 expression levels on CFSEhigh OT-1 and CFSElo C57BL/6 cells transferred into (C) OVA-challenged tie2-H-2Kb mice or (D) splenectomized [B6.CH-2bm1->C57Bl/6] chimeras.

Next we addressed the influence of hepatic T-cell retention on the subsequent quality of the CD8 T-cell response. We adaptively transferred naïve OT-1 cells into C57Bl/6 mice without OVA (I), C57Bl/6 mice immunized with OVA/IFA subcutaneously (II), or [B6.CH-2bm1->C57Bl/6] chimeras that received OVA systemically (III). This allowed us to discriminate immunogenic priming of CD8 T-cells in secondary lymphatic tissue (II) from LSEC-mediated CD8 T-cell activation (III). As expected, in the absence of antigen (I) OT-1 cells did not proliferate or expand (Fig. 6A,C), whereas after immunogenic priming (OVA/IFA) (II) they proliferated and expanded extensively over a period of 4 days (Fig. 6A), yielding a more than seven-fold expansion (Fig. 6B). Importantly, priming of naïve OT-1 cells in OVA-challenged [B6.CH-2bm1->C57Bl/6] chimeras also led to significant proliferation (Fig. 6A), yielding more than three-fold expansion compared to naïve T-cells (Fig. 6B). This lower expansion was not due to increased apoptosis, as active caspase-3 levels in the CD8 T-cells did not differ ex vivo (Fig. 6C). In summary, this demonstrates that antigen-presenting organ-resident LSECs can sustain significant antigen-specific CD8 T-cell expansion in vivo.

Figure 6.

Clonal expansion of CD8 T-cells by cross-presenting LSEC. (A-C) CFSE-labeled OT-1 cells were transferred into congenic mice that received PBS (I), OVA/IFA (II), or into [B6.CH-2bm1->C57Bl/6] chimeras that received OVA intraperitoneally (III). For details, see supporting materials. (A) After 4 days lymphocytes were isolated from (A,C) spleen or (B,C) liver and analyzed by flow cytometry. (A) Depicted are percentages Vα2β5 double-positive OT-1 cells (upper panel) or percentage proliferated OT-1 cells (lower panel) within the CD8pos fraction. (B) Total numbers of Thy1.1+ OT-1 cells recovered from liver at day 4 after adoptive transfer. (C) Caspase 3 expression in OT-1 T-cells recovered from spleen and liver.

To examine the function of LSEC-stimulated OT-1 cells in vivo, T-cells were antigen-specifically restimulated ex vivo. OT-1 cells primed in an inflammatory environment (II) produced large amounts of IFN-γ consistent with the induction of immunity (Fig. 7A). In contrast, T-cells activated by cross-presenting LSECs in vivo failed to produce IFN-γ (Fig. 7A), indicating that T-cells were eventually rendered tolerant. Quantification of OT-1 cells by way of Vα2β5 showed that after immunogenic stimulation (II) over 85% of OT-1 cells in the liver and 50% in the spleen produced IFN-γ upon restimulation (Fig. 7B). In contrast, IFN-γ or IL-2 production was low or absent in restimulated OT-1 cells primed by LSECs cross-presenting OVA in vivo (Fig. 7B). Interestingly, OT-1 cells tolerized by cross-presenting LSECs expressed significant levels of the inhibitory molecule PD-1, whereas fully activated or naïve OT-1 cells did not express comparable PD-1 levels (Fig. 7C). Thus, recognition of cross-presented soluble antigens on LSECs in vivo leads to proliferation and expansion of antigen-specific naïve CD8 T-cells but these cells lack effector function.

Figure 7.

LSECs tolerize CD8 T-cells in vivo. (A-C) Lymphocytes were obtained as described in Fig. 6 and were (A,B) restimulated with SIINFEKL-loaded, LPS-matured BMDC for 5 hours and analyzed for IFN-γ/IL-2 expression. (C) PD-1 expression was determined on OT-1 cells isolated 4 days after antigen challenge in vivo. Dot plot in (A) is gated on CD8-positive lymphocytes; in (C) gated on the congenic marker Thy1.1 and CD8.


The liver is known for its ability to induce tolerance rather than immunity. Various cell populations are involved in induction of CD8 T-cell tolerance: hepatocytes,21 immature DC,22 and nonparenchymal cells such as Kupffer cells23 and LSECs.12 Induction of tolerance by hepatocytes leads to clonal elimination of CD8 T-cells specific for antigens expressed in hepatocytes, i.e., autoantigens or in case of infection viral antigens.21, 24 Induction of CD8 T-cell tolerance toward circulating antigens, however, requires cross-presentation of antigen on MHC I molecules, a feature restricted to professional APC such as DC or macrophages8, 9 and to liver-resident LSECs.12 Although the mechanisms mediating contact between antigen-presenting DC and naïve CD8 T-cells in secondary lymphatic tissue are well known,18 the mechanisms promoting adhesion of circulating naïve CD8 T-cells to antigen-presenting endothelial cells are not. Therefore, we addressed the question of how LSECs, which are the predominant cell population scavenging antigen from the blood stream (Fig. 1 and Ref.5) and are capable of cross-priming,12 establishing physical contact with circulating naïve CD8 T-cells.

Under static conditions in vitro cross-presentation of soluble antigen by LSECs rapidly promoted migratory arrest of naïve CD8 T-cells (Fig. 2). Thus, antigen-specific adhesion to cross-presenting LSEC initiates physical contact with naïve CD8 T-cells, which is then followed by stimulation of naïve CD8 T-cells and tolerogenic maturation of LSECs, eventually leading to CD8 T-cell tolerance through B7-H1/PD-1-dependent signals.15 Systemic application of antigen induced antigen-dependent retention of naïve CD8 T-cells within several hours in the liver (Fig. 3). T-cells were not retained in the lung or secondary lymphoid organs, indicating that the liver has a unique role in initiating immune responses toward circulating antigens. Compared to the antigen-specific hepatic retention of autoreactive naïve CD8 T-cells10 in response to ubiquitously expressed peptide-MHC I molecules, the naïve CD8 T-cell retention in response to cross-presented circulating antigens was less pronounced (data not shown). This suggested that only a highly specialized hepatic cell population bearing the ability to cross-present caused antigen-specific arrest of naïve CD8 T-cells. Hepatic retention of naïve CD8 T-cells was still observed in BM chimeras in which BM-derived immune cells fail to present antigen. Additionally, also in tie2-H-2Kb mice in which endothelial cells, but not hepatocytes or stellate cells, can present the antigen, antigen-specific retention was still observed. These data together indicate that LSECs, which are the major antigen scavenging cells, are also responsible for the observed T-cell retention.

Naïve CD8 T-cell retention in the liver did not require additional proinflammatory signals. Those signals are normally required to increase expression of adhesion molecules (CD54, CD106, or VAP-1) or chemokines (CXCL16, CCL19, and CCL21), and trigger T-cell adhesion to EC in other organs.25, 26 CD54 also contributes to hepatic antigen-specific accumulation of activated CD8 T-cells after peptide injection27 and promotes stable antigen-specific interactions.28 Neither CD106 nor VAP-1 contributed to antigen-specific T-cell adhesion to cross-presenting LSECs in vivo (data not shown). Although CD54 on LSECs contributed to T-cell interaction in vitro (Fig. 2), there was only a marginal influence on T-cell adhesion in vivo (Fig. 3), which is in contrast to intrahepatic T-cell retention in response to ubiquitously presented antigen.10 However, the contribution of chemokines to antigen-specific CD8 T-cell retention in the liver still remains to be determined.

The unique features of hepatic sinusoidal blood circulation such as low perfusion pressure, irregular perfusion, and selectin-independent lymphocyte adhesion29 are ideal for antigen-specific adhesion of naïve CD8 T-cells to cross-presenting LSECs in vivo. T-cells in lymphatic tissue need chemotactic signals and adhesion molecules to physically interact with both antigen-presenting and nonpresenting DC.30 Such an interaction of T-cells with LSECs would obstruct hepatic sinusoidal blood flow. Allowing T-cell retention only after recognition of antigenic peptide-MHC class I complexes thus may prevent unnecessary vessel obstruction.

Trapping of previously activated T-cells occurs preferentially in the liver, leading to apoptosis.20 However, in splenectomized mice antigen-specific CD8 T-cell retention was not lost, indicating that initial stimulation of naïve CD8 T-cells takes place in the liver and not in the spleen. Antigen-specific hepatic recruitment of naïve CD8 T-cells within 4 hours and upregulation of CD69 within 2 hours after antigen application suggests that antigen-uptake, antigen-processing, and antigen-presentation by the same hepatic APC directly mediated antigen-specific T-cell retention. Importantly, in BM chimeras CD8 T-cells primed by LSECs in vivo were rendered tolerant, confirming our previous observation that adoptive transfer of in vitro antigen-loaded LSECs induces CD8 T-cell tolerance.12

Although the total number of naïve CD8 T-cells retained in the liver within 4 hours was low compared with the number redistributing to the spleen (≈10%), T-cells had massively expanded after 4 days. At this time, tolerant CD8 T-cells reached more than 40% of the numbers of CD8 T-cells that were activated under inflammatory conditions. Thus, hepatic retention of T-cells in the liver by cross-presenting leads to their expansion and tolerization, which is in contrast to the deletional tolerance induced by hepatocytes and immature DC.11, 21 We speculate that competition for priming of naïve T-cells between liver and lymphatic tissue plays a role in shaping the immune response toward circulating antigens. Competition for priming of naïve T-cells can occur between hepatocytes presenting auto-antigens and lymphatic DC, the latter of which failed to induce immunity if T-cells had first recognized antigen on tolerizing hepatocytes.24

T-cell tolerance induction by LSECs in vitro is mediated by co-inhibitory signals through B7-H1/PD-1 receptor-ligand interaction.15 Interestingly, CD8 T-cells tolerized by cross-presenting LSECs in vivo expressed PD-1, whereas naïve and activated CD8 T-cells did not. Therefore, PD-1-expression on CD8 T-cells may function as a marker for LSEC-tolerized T-cells. Although LSEC-induced tolerance does not depend on high antigen dose (Supporting Fig. 2), as is the case in the PD-1 positive exhausted CD8 T-cell population found during chronic viral infection,31 the fact that PD-1 is induced on LSEC-tolerized T-cells raises the question of whether within the PD-1 positive T-cell population some cells could originate from tolerogenic priming by LSECs.

Taken together, antigen-specific retention of naïve CD8 T-cells in the liver represents the first step in the induction of CD8 T-cell tolerance toward circulating antigens that is executed by liver-resident scavenger LSECs. Understanding the principles of peripheral CD8 T-cell tolerance induction will not only help us to understand the pathophysiological principles underlying tolerance induction toward circulating viral antigens but also to manipulate tolerance induction for immunotherapy.


We thank A. Dolf, P. Wurst, and E. Endl for technical assistance, C. Weber (Aachen), D. Adams, and T. Lalor (Birmingham, UK) for help with flow-adhesion assays, T. Schwand for help with splenectomy, and B. Arnold for supplying transgenic animals.