Potential conflict of interest: Nothing to report.
Supported by National Institutes of Health (NIH) grants R01-DK075274 and NIH R01-AI064463 (to INC)
Antigen cross-presentation is a principal function of specialized antigen-presenting cells of bone marrow origin such as dendritic cells. Although these cells are sometimes known as “professional” antigen-presenting cells, nonbone marrow-derived cells may also act as antigen-presenting cells. Here, using four-way liver cell isolation and parallel comparison of candidate antigen-presenting cells, we show that, depending on the abundance of antigen-donor cells, different subsets of liver cells could cross-present a hepatocyte-associated antigen. This function was observed in both liver sinusoidal endothelial cells and Kupffer cells even at very low antigen concentration, as well as when using soluble protein. Antigen cross-presentation by liver cells induced efficient CD8+ T-cell proliferation in a similar manner to classical dendritic cells from spleen. However, proliferated cells expressed a lower level of T-cell activation markers and intracellular interferon-gamma levels. In contrast to classical spleen dendritic cells, cross-presentation by liver antigen-presenting cells was predominantly dependent on intercellular adhesion molecule-1. Conclusion: Hepatic sinusoids are an environment rich in antigen cross-presenting activity. However, the liver's resident antigen-presenting cells cause partial T-cell activation. These results clarify how the liver can act as a primary site of CD8+ T-cell activation, and why immunity against hepatocyte pathogens is sometimes ineffective. (HEPATOLOGY 2011;54:1379–1387)
Specialized antigen-presenting cells (APCs) such as dendritic cells (DCs) are capable of capturing exogenous antigens from other cells and indirectly presenting them on class I major histocompatibility complex (MHC) molecules, a capacity that has been referred to as “antigen cross-presentation.” Such cross-presentation plays an essential role in the induction of effective cytotoxic T lymphocyte (CTL) immunity to various antigens, including minor histocompatibility antigens, tumor-associated antigens, and viral antigens, a process referred to as “cross-priming.”1, 2
Inflammation and poor immunity are characteristic of many liver infections. The persistence of liver pathogens is often accompanied by a weak CD8+ T-cell response to hepatocellular antigens,3 in part because liver pathogens subvert antigen presentation mechanisms.4 However, cross-presentation can allow the antigen to be presented by a noninfected, and thus nonsubverted cell. In addition, weak CD8+ T-cell responses can be augmented by CD4+ T-cell help.5 Hepatocytes are the main target of liver pathogens and lack MHC class II antigen presentation. In this situation, cross-presentation on an MHC class II-positive APC could also allow the priming of CD4+ T cells. Therefore, cross-presentation of antigen, whether on bone marrow-derived cells or on nonhematopoietic APCs, is likely to be critical in determining the outcome of hepatocellular infection.
Nonparenchymal liver cells include liver sinusoidal endothelial cells (LSECs), Kupffer cells (KCs), and hepatic stellate cells (HSCs). These subsets of liver cells have been shown to present antigens to T cells,6, 7 and may be responsible for delivery of either tolerogenic or immunogenic signals to T cells. Soluble ovalbumin (OVA) protein has been used to investigate liver antigen presentation8-10; however, the more relevant forms of antigens in vivo may be cell-associated. Cross-presentation of soluble antigens can occur in stable early or in recycling endosomes, whereas cell-associated proteins engage phagosomes and late endosomes where they are temporally separated from MHC class II loading.11, 12 Illustrating this distinction, mannose receptor-deficient DCs show impaired endocytosis of soluble OVA and abrogated antigen presentation, whereas their uptake of cell-associated OVA and activation of T cells was unaffected.13
In this study we investigate the relative capacity of different liver APCs to engage in direct presentation, cross-presentation of soluble antigens, and cross-presentation of cellular antigens. Our results show that many liver cells can cross-present antigens and induce T-cell proliferation. However, cross-presentation by liver APCs induces partial T-cell activation, which is dependent on intercellular adhesion molecule-1 (ICAM-1) expression. These results support a model of liver immunity that achieves primary T-cell activation but fails to induce an effective immune response.
Eight to 10-week-old C57BL/6 wildtype, OVA transgenic, ICAM-1 deficient, OVA-specific H-2Kb-restricted T-cell receptor (TCR) transgenic (OT-I), and B6.C-H-2bm8 (bm8) mice were used in accordance with Institutional Animal Care and Use Committee guidelines.
Candidate liver APCs were isolated to high purity using a novel multistep isolation technique (Supporting Fig. S1). Spleen mDCs were isolated using magnetic antibody cell sorting against CD11c (MACS, Miltenyi Biotec). CD8+ T cells were isolated from spleen and peripheral lymph nodes of OT-1 mice as described.14 Following isolation, OT-1 CD8+ T cells were labeled with 0.7 mM carboxy-fluorescein-succinimidyl-ester (CFSE) and used in antigen presentation experiments. Antibodies used are described in the Supporting Information.
Antigen Cross-Presentation Experiments.
For studying antigen cross-presentation, 2.5 × 104 isolated liver APCs or spleen DCs were seeded in 96-well round-bottom plates, and soluble OVA protein (final concentration of 100 μg/mL, grade VII) was added. On the following day, OT-I CD8+ T-cells were isolated, CFSE-labeled, and added to the cultures at 105 cells per well. In antibody blocking experiments, antibodies (10 μg/mL) were added 1 hour before the OT-I CD8+ T cells. For cross-presentation from bm8-OVA hepatocytes, 1 day before isolation of APCs, hepatocytes from bm8-OVA mice were isolated and seeded in the plates at 102, 103, or 104 cells per well. All cells were cultured in RPMI supplemented with 10% fetal bovine serum (FBS), 50 μM beta-mercaptoethanol, glutamine, sodium pyruvate, and antibiotics. We characterized the uptake, intracellular processing, and presentation of OVA using fluorescein-OVA, DQ OVA, and H-2Kb-SIINFEKL staining, respectively. DQ is a self-quenched conjugate of OVA that exhibits bright green fluorescence upon proteolytic degradation. We also quantified the relative basal expression of H-2Kb in each cell type.
Cross-Presentation of Hepatocyte-Derived Antigen by Liver APCs.
Isolated hepatocytes from bm8-OVA transgenic mice were used as the source of cell-derived antigen. Cells derived from these mice were unable to present the OVA peptide to OT-I CD8+ T cells due to a mutation in the H-2Kb molecule.15 OT-I CD8+ T-cells express transgenic TCRs against the peptide derived from OVA257-264, peptide sequence SIINFEKL, in the context of H-2Kb. In parallel experiments we tested the capacity of the different liver APCs to cross-present hepatocyte-derived OVA to CD8+ OT-I T cells. We isolated candidate APCs from B6 mice and cocultured them with bm8-OVA hepatocytes and OT-I CD8+ T cells. The proliferation of OT-I cells was assessed by the dilution of CFSE staining. When the antigenic cells were hepatocytes at 104 and 103 per well, all liver APCs were efficient in the cross-presentation of OVA associated with bm8-OVA hepatocytes, and all induced more than five rounds of T-cell proliferation (Fig. 1A,B). However, HSCs were not as efficient as the other liver cells in the induction of T-cell proliferation (Fig. 1D, P < 0.01 at all concentrations of antigen). Even normal hepatocytes were better than HSCs in cross-presentation of OVA antigen (Fig. 1D, P = 0.014 at both 103 or 104 bm8-OVA hepatocytes/well). KCs and LSECs strongly cross-presented cell-associated antigens and induced high levels of T-cell proliferation, similar to the level that we observed with spleen mDCs (Fig. 1D).
It has been postulated that antigen dose can regulate cross-presentation function,16 therefore limiting the availability of antigen may create conditions that are not favorable for cross-presentation. By decreasing the source of antigen in cultures, cross-presentation activity was attenuated in both HSCs and hepatocytes (Fig. 1C,D). Plating 102 bm8-OVA hepatocytes per well, LSECs and KCs were the only liver cells that could efficiently cross-present OVA to OT-I CD8+ T cells. Both LSECs and KCs were as efficient as spleen DCs in the induction of CD8+ T-cell proliferation (Fig. 1C,D). The capacity of LSECs and KCs to cross-present hepatocyte-associated antigens could explain the potential of the liver as a primary site of T-cell activation when antigen is in hepatocytes.
Cross-Presentation of Soluble Antigen by Liver APCs.
In separate studies, HSCs, KCs, and LSECs have been proposed to cross-present soluble OVA protein and activate CD8+ T cells.8-10 However, from such studies it is difficult to draw conclusions about the relative ability of each cell type to induce CD8+ T-cell activation. Using direct back-to-back comparison of the different liver APCs, we can conclude that among the liver APCs, LSECs induced the most robust T-cell proliferation, and in fact appeared to be slightly more potent than mDC (Fig. 2A,C, P = 0.058). KCs were also capable of inducing significant levels of CD8+ T-cell proliferation; however, they tended to be less potent than LSECs (Fig. 2A,C, P = 0.055). Both hepatocytes and HSCs were inefficient in the presentation of soluble OVA proteins to CD8+ T cells (Fig. 2A,C, P < 0.01 in comparison to mDCs).
Presentation of Endogenous Antigen by Liver Cells.
Direct presentation of endogenous antigen by nonparenchymal liver cells is likely an important component of liver immunity.17-19 To evaluate direct presentation of antigen by liver APCs, we took advantage of cells from OVA transgenic mice. Back-to-back comparison of these cells showed that all liver APCs can induce proliferation of CD8+ T cells very efficiently. The level of this induction was comparable to OVA transgenic mDCs (Fig. 2B,C).
LSECs Show Superior Uptake, Processing, and Presentation of Soluble Antigen.
The SIINFEKL antigen needs to be in complex with MHC class I H-2Kb in order to engage an OT-I CD8 TCR. The basal expression of H-2Kb was lower in hepatocytes compared to the other liver cells or mDC (Fig. 3C, P < 0.05). This may suggest a lower capacity of these cells to induce a T-cell response (Fig. 3). Following 4 and 24 hours of incubation with fluorescein OVA conjugate, LSECs took up more OVA compared to other APCs (Fig. 3A; Supporting Fig. S3). Concordantly, LSECs processed more DQ OVA and displayed higher levels of H-2Kb-SIINFEKL on their cell surface (Fig. 3B,D, respectively). These results explain how LSECs show such strong cross-presentation of soluble proteins and induce T-cell proliferation. Hepatocytes and HSCs could uptake OVA protein, but less efficiently than the other liver APCs (Figs. 3; Supporting S3). Unexpectedly, we also noticed high levels of H-2Kb-SIINFEKL on the surface of HSCs after 24-hour incubation with OVA protein.
Cross-Presentation Is Dependent on ICAM-1.
The events that lead to T-cell activation are critically regulated by costimulatory molecules, such as CD28 and ICAM-1 located at the immunological synapse.20, 21 With a focus on LSECs, KCs, and spleen mDCs, we tested whether liver APCs exhibit distinctive costimulatory requirements during antigen cross-presentation and activation of CD8+ T cells. During initial experiments using blocking antibodies, we observed an important role for ICAM-1 in antigen presentation by liver cells (Fig. S2). Thus, to further address the role of ICAM-1 in cross-presentation of OVA by liver APCs, we used APCs isolated from ICAM-1-deficient mice. ICAM-1-deficient LSECs and KCs could not cross-present soluble OVA to CD8+ T cells and failed to induce T-cell proliferation (Fig. 4, wildtype [WT] versus ICAM-1-deficient APCs, P = 0.029 for LSECs and P = 0.018 for KCs). However, ICAM-1-deficient spleen mDCs induced robust proliferation of CD8+ T cells similar to WT mDCs (Fig. 4). This suggests that ICAM-1 is particularly important in T-cell activation by liver APCs, in contrast to its smaller contribution to T-cell activation by spleen mDCs.
Cross-Presentation by Liver APCs Induced Partial T-Cell Activation.
The CD8+ T-cell activation response relies on the differentiation of a small number of specific naive CD8+ T lymphocytes into potent effector CTLs. One of the many facets of this activation is up-regulation of cell adhesion molecules including the hyaluronic acid receptor (CD44), and the expression of the CD25, receptor for IL-2, an important cytokine for T-cell proliferation.22, 23 We evaluated the expression of these two markers of CD8+ T-cell activation following antigen cross-presentation by liver APCs or spleen mDCs. Compared to mDCs, we observed that liver APCs induced less CD25 and CD44 expression on proliferated CD8+ cells (Fig. 5A-C). Increasing bm8-OVA hepatocyte density failed to elevate CD44 or CD25 induced by liver APCs to levels comparable to spleen mDCs (Fig. 5A,B). Cross-presentation of soluble OVA by liver APCs also failed to induce CD44 or CD25 to the same degree as mDCs, although LSECs and KCs caused greater CD44 and CD25 induction than other liver APCs (Fig. 5C).
Because PD-L1 on APCs can interact with PD-1 to reduce activation of CD8+ T cells, we explored whether a PD-L1 blocking antibody (MIH5 clone)24 could increase the T-cell activation phenotype following antigen presentation by liver APCs. However, in our setting, blocking PD-L1 did not restore CD44 or CD25 in proliferated CD8+ T cells (Fig. S4).
It is also plausible that antigen presentation by liver APCs promote activation-induced cell death (AICD) of highly activated T cells. In this scenario, strongly activated T cells would be predicted to die more rapidly. To address this, we measured CD8+ T-cell death following cross-presentation of Bm8-OVA hepatocytes by liver APCs or spleen DCs. The results showed no significant difference in the frequency of dead cells among cultures of KCs, LSECs, or spleen DCs (Fig. S5). Additionally we measured expression of CD44 and CD25 in CD8+ T cells that are 7-AAD permeable. These dead CD8+ T cells from cultures of liver APCs were still lower in expression of CD44 and CD25 when compared with T cells from cultures of spleen DCs (P < 0.05, Fig. S5). We found the highest frequency of 7AAD+ CD8+ cells in the cultures of OT-1 T cells with hepatocytes and HSCs, in which the T cells were not so strongly activated. This does not fit with an AICD model (see Discussion).
Reduced Intracellular Levels of Interferon-γ (IFN-γ) Following Antigen Cross-Presentation by Liver APCs.
Following activation, CD8+ T cells produce IFN-γ, an important signature cytokine of antiviral responses. Although measurements of IFN-γ in a culture supernatant can give us a crude estimate of IFN-γ secretion from both CD8+ T cells and their APCs, intracellular staining of CD8+ TCR+ cells for IFN-γ can accurately reveal the levels of IFN-γ. Consistent with lower CD44 and CD25 expression, we observed that CD8+ T cells cross-primed by liver APCs produced less IFN-γ (Fig. 6A). To understand whether a low level of IFN-γ production is exclusive to the conditions in which antigen is cross-presented, we assessed IFN-γ levels following direct presentation of OVA protein. Direct presentation of OVA by liver APCs, but not spleen mDCs, was also accompanied by inefficient IFN-γ production (Fig. 6B). This was particularly interesting because the levels of proliferation were similar. These experiments showed that the attenuated level of IFN-γ in T cells exposed to liver APCs is independent of the source of antigen. These results, taken together with the CD44 and CD25 expression data, show that antigen cross-presentation by liver APCs can induce T-cell proliferation but not full T-cell activation.
The goal of this study was to conduct an unbiased back-to-back comparison of the major resident liver cell populations as APCs. This approach led us to precisely characterize cross-presentation of hepatocyte-associated and soluble protein antigens by liver APCs. Our results show that the liver is very rich in cell types that can perform antigen cross-presentation. Antigen cross-presentation by liver cells was dependent on the amount of available antigen, and was specialized to LSECs and KCs in the case of soluble protein antigen, or with a low dose of a cell-associated antigen. HSCs were not efficient in activation of T cells through cross-presentation of antigens. However, HSCs were quite good in direct presentation of endogenous antigens. Cross-presentation by liver APCs resulted in proliferation of CD8+ T cells to a level comparable to spleen mDCs under certain conditions. Interestingly, classical features of fully activated CD8+ T cells, such as high CD44, high CD25, and high IFN-γ, did not accompany liver APC-induced T-cell proliferation. We believe these features of intrahepatic antigen presentation strongly influence liver immune responses.
Hepatocytes are the predominant target cells in liver infection. In this study we observed that at high antigen levels these cells could cross-present cell-associated antigens from neighboring hepatocytes to CD8+ T cells and induce a substantial level of proliferation. However, it has been shown that activation of CD8+ T cells on hepatocytes promotes helpless CTLs and suboptimal T-cell activation.25, 26 Thus, upon infection of a small number of hepatocytes, it is the cross-presentation of hepatocyte antigens by liver nonparenchymal cells that has the potential to engage both CD4+ and CD8+ T cells simultaneously and deliver an appropriate effective immune response. Presumably, CD4+ help is one of the microenviromental cues that can influence the ultimate fate of adaptive immune response in the context of partial CD8+ T-cell activation. In this study, we were able to show that the function of cross-presentation is not limited to classic professional APCs, such as mDCs. Assessment of both LSECs and KCs showed that they efficiently cross-presented antigens to CD8+ T cells under all of the conditions tested. However, the cross-presentation by these liver APCs was accompanied by suboptimal CD8+ T-cell cross-priming.
Hepatic stellate cells have recently been proposed as professional liver-resident APCs that efficiently present antigens and drive proliferation of NKT cells,10 but there have been few follow-up studies to characterize the antigen presentation activity of these cells. One concern is that the standard protocols for isolation of HSCs could result in some level of cross-contamination from liver macrophages. We isolated and compared several APC candidates in parallel, making special efforts to deplete CD11b+ macrophages from HSC cultures. This allows us to offer a more detailed assessment of antigen presentation by HSCs. Our data show that purified HSCs were indeed very competent in presentation of endogenously expressed antigens to CD8+ T cells. However, in our parallel comparison, HSCs poorly cross-presented antigens at various concentrations. It was therefore unexpected to observe such high levels of H-2Kb-SIINFEKL on surface of HSCs that were incubated with OVA protein. Because HSCs did not induce T-cell proliferation, this suggests that HSCs either actively inhibit T-cell proliferation or fail to express costimulatory factors that allow HSCs to function as efficient cross-presenting cells.27, 28
Following TCR engagement, T-cell activation can be outwardly measured by proliferation, expression of surface molecules such as CD44 and CD25, or cytokine secretion. Whether these aspects of the T-cell response are triggered concomitantly or independently following engagement of the TCR complexes on naive CD8+ T cells has been the target of several investigations; however, it is difficult to propose a cohesive model.29, 30 Tight linkage between proliferation and function is evident in some experimental models of differentiation, but not others.29, 30 In our study, we observed that even with a similar level of proliferation between liver APCs and spleen mDCs, there are distinct differences in CD44, CD25, and IFN-γ levels. This may suggest that T cells primed by liver APCs could require additional cell divisions to achieve the same level of activation, or alternatively, that cell division and cytokine synthesis are not strictly coupled in this context. The exact underlying mechanisms that shape this partial activation state warrant further investigation. However, one possible mechanism is that liver APCs promote cell death in highly activated T cells.
Studying the dead cells in relation to the level of proliferation and CD44 and CD25 expression, the data do not support the activation-induced cell death hypothesis. Instead, less efficient APCs (hepatocytes and HSCs) induced lower levels of proliferation, and less induction of CD44 and CD25, but higher levels of cell death. This suggests that suboptimally activated CD8+ T cells are less viable. This phenotype is particularly evident in hepatocyte-mediated cross-presentation where CD8+ T cells divide, but fail to up-regulate CD25 or IFN-γ production and die. It is important to consider that proliferation and IFN-γ production by the CD8+ T cells are influenced by the spectrum of cytokines secreted by the APCs. For instance, in the fibrotic liver KCs are an important source of transforming growth factor beta (TGF-β), which can skew naïve CD8+ T-cell activation and IFN-γ production.31-33 Thus, this T-cell phenotype is additionally regulated by further microenvironmental cues during the course of infection.
The liver microenvironment is constitutively exposed to intestinal bacterial products that engage pathogen recognition receptors and result in continuous low-level stimulation of the MyD88-NF-κB axis.6 In such an environment, it makes sense to have a higher threshold for effector T-cell activation. Thus, in the liver the CD8+ T-cell response may be tuned down in order to meet the physiologic requirement of this organ and prevent uncontrolled inflammation or immune response. However, this tolerance may also allow pathogens to achieve hepatic chronicity.
This study demonstrates that the chronicity of many liver infections is not likely to be due to the lack of efficient antigen cross-presentation; in fact, the liver is an organ with diverse cells that can cross-present antigens. Instead, antigen cross-presentation in the liver may expand a CD8+ T-cell population with an atypical phenotype, the purpose of which has yet to be understood. Taken together, these results help to explain how the liver can be a primary site for T-cell proliferation, and yet liver-activated T cells are prone to deliver ineffective immunity.
We thank Ms. Rebekah Brown and Mr. Dat Mai as well as flow core facilities at Seattle BioMed and the University of Washington for their technical support during this study. We also thank Ms. JoAnne Dyer for help with preparation of the article.