Large number of T cells traffic through the liver. In order to examine the effects of such traffic on the phenotype of hepatocytes, we vaccinated mice using DNA vaccines encoding antigens with MHC class I-binding epitopes. Small numbers of activated CD8+ T blasts (105–106/liver) changed the surface phenotype and cytokine expression profile of hepatocytes (HCs). HCs upregulate surface expression of major histocompatibility complex (MHC) class I molecules and CD1d but not MHC class II molecules Qa-1, CD80, CD86, CD54, or CD95; in addition, they expressed/secreted interleukin (IL)-10 and IL-4 but not IL-1, IL-6, IL-13, interferon (IFN)-γ, tumor necrosis factor (TNF), IL-4, or IL-27 (i.e., they acquire the HC* phenotype). HCs* (but not HCs) induced specific activation, proliferation, and IFN-γ, TNF, and IL-13 release of cocultured naïve CD8+ T cells. In contrast to the specific activation of naïve CD8+ T cells by dendritic cells (DCs), specific CD8+ T cell activation by HC* was not down-modulated by IFN-αβ. Only recently activated CD8+ T blasts (but not recently activated CD4+ T blasts or activated cells of the innate immune system, including natural killer T [NKT] cells) induced the HC* phenotype that is prominent from day 10 to day 20 postvaccination (i.e., time points at which peak numbers of recently primed CD8+ T blasts are found in the liver). In conclusion, recently activated CD8+ T blasts that enter the liver postimmunization in small numbers can transiently modulate the phenotype of HC, allowing them to activate naïve CD8+ T cells with unrelated specificities. (HEPATOLOGY 2004;39:1256–1266.)
Large numbers of naïve recently activated or memory T cells traffic through the liver. CD8+ T cells expressing the αβ T cell receptor for antigen (TCR) represent more then 60% of the hepatic T cell population but only about 5% of the liver nonparenchymal cell (NPC) population. Sinusoidal endothelial cells of the liver selectively trap CD8+ αβ T cells in a postactivated (rather than a resting) state.1–3 We asked if recently activated CD8+ T cells entering the liver change the phenotype of hepatocytes (HCs).
Extrahepatic priming of CD8+ αβ T cells by a vaccine (which specifically activates 0.5%–5% of the CD8+ T cell population) results in the influx of specific CD8+ T cells into the liver in the second week postimmunization.4 These specific, recently activated, intrahepatic CD8+ T cells are functional and can be found in the liver for up to 3–5 months postpriming (depending on the vaccination protocol used). This experimental system differs from many preclinical, immunological hepatitis models that rely on extensive polyclonal activation of T cells induced by injecting anti-CD3 antibody, superantigen, or mitogen into mice (or the relevant peptide into TCR-transgenic mice). In these models, (1) CD8+ T cell influx into the liver is rapid and massive; (2) presentation of the activating stimulus or antigen by HCs seems likely in most of these experimental models; and (3) “bystander” hepatitis develops during the intrahepatic accumulation of large numbers of recently activated CD8+ T cells.5, 6 The studies model extreme clinical situations (e.g., toxic shock syndrome). Our interest is focused on changes in the phenotype of HCs induced by the appearance in the liver of low numbers of CD8+ T blasts that have been recently activated by extrahepatic immunization. These studies may help us to understand early events in the development of bystander or autoimmune hepatitis.
We vaccinated mice intramuscularly using DNA vaccines encoding antigens with defined MHC class I–binding epitopes.7–11 We tested if the postvaccination influx of specific, recently activated CD8+ T cells into the liver modulates the surface and functional phenotype of HCs. We further tested if specific priming of naïve CD8+ T cells by HCs is enabled or modulated in a liver in which a small number of recently activated CD8+ T cells is found. We describe the unexpected observation that the phenotype of HCs is changed by the influx of low numbers of recently activated CD8+ T cells.
TCR, T cell receptor for antigen; NPCs, nonparenchymal cells; HCs, hepatocytes; MHC, major histocompatibility complex; IL, interleukin; IFN, interferon; TNF, tumour necrosis factor; DCs, dendritic cells; NKT cells, natural killer T cells; RAG, recombination-activating gene; HBsAg, hepatitis B virus surface antigen; HBs-tg mice, transgenic mice expressing HBsAg in the liver; HBcAg, hepatitis B virus core antigen; HBc-tg mice, transgenic mice expressing HBcAg in the liver; αGalCer, α-galactosyl ceramide; FCM, flow cytometry; OVA, ovalbumin; ELISA, enzyme-linked immune assay; NK cells, natural killer cells.
Materials and Methods
C57BL/6J (H-2b) (B6) mice and their mutant MHC class II–deficient (Aα−/− or Aβ−/−), MHC class I–deficient (β2m−/−), severely immunodeficient (recombination-activating gene [RAG]1−/− or RAG2−/−), and TCR-transgenic OT-I and OT-II (and their RAG2−/− OT-I) sublines were bred and kept under standard pathogen-free conditions in the animal colony of Ulm University (Ulm, Germany). TgN(Alb1HBV)44Bri-transgenic B6 mice that express hepatitis B surface antigen (HBsAg) in the liver (HBs-tg mice) were obtained from The Jackson Laboratory (Bar Harbor, ME). CexL B6 mice that express hepatitis B virus core antigen (HBcAg) in the liver (HBc-tg mice) have been described.12, 13 Female mice were used at 12–16 weeks of age. All animal experiments were conducted according to the guidelines of the local Animal Use and Care Committees and were carried out according to the National Animal Welfare Law.
DNA immunization and the DNA vaccines used have been described previously.7–9, 14 We used three B6 mice per group in all experiments because the variability of CD8+ T cell responses to the tested antigens between individual B6 mice induced by our vaccination protocol is low (see SEM shown in the figures). Representative data from one of at least three independent experiments are shown.
In Vivo Activation of Natural Killer T Cells by α-Galactosyl Ceramide.
To activate natural killer T (NKT) cells in vivo, α-galactosyl ceramide (αGalCer) (kindly provided by Y. Koezuka, Kirin Brewery, Pharmaceutical Research Laboratory, Gunma, Japan) dissolved in phosphate-buffered saline was injected intraperitoneally (100 ng/mouse). HCs were isolated 18 hours postinjection.15, 16
Isolation of HCs, Dendritic Cells, and CD8+ and CD4+ T Cells.
HCs and liver NPCs were isolated as described previously.15 The purity of the isolated HCs was greater than 98% as verified morphologically and by flow cytometry (FCM). Magnetic bead-activated cell sorting of splenic CD11c+ dendritic cells (DCs) (Miltenyi Biotec, Bergisch-Gladbach, Germany) has been described.17, 18 CD8+ and CD4+ T cells were purified using magnetic bead-activated cell sorting kits (Miltenyi Biotec; CD8α+ T cell isolation kit cat.no.130-090-859; CD4+ T cell isolation kit cat.no.130-090-860). The purity of the isolated CD8+ and CD4+ T cells was greater than 96% as verified by FCM.
FCM Analysis and Measuring Messenger RNA Expression by Real-Time Polymerase Chain Reaction (TaqMan).
FCM analyses using FITC-, PE- or biotin-conjugated antibodies from BD Pharmingen have been described.15 SA-Red 670 was obtained from Gibco-BRL (Berlin, Germany; cat.no.19543-024). We and others have described the intracellular staining of CD8+ T blasts for interferon (IFN)-γ and tumor necrosis factor (TNF)-α,4 RNA isolation, and real-time polymerase chain reaction (TaqMan, PE Applied Biosystems, Norwalk, CT) analysis.19
Cells were cultured in 200-μL flat-bottom microwells in RPMI-1640 medium supplemented with 5% fetal calf serum, 2 mM L-glutamine, and antibiotics. Usually, 1 × 105 CD8+ or CD4+ T cells/well were cocultured with 1 × 104 HCs or DCs/well. In some experiments, titrated numbers of CD8+ T cells (ranging from 2 × 104 to 2 × 105/well) were cocultured with a constant number of stimulating HCs (1 × 104/well). HCs or HCs* were pulsed with increasing concentrations of: (1) the Kb-binding OVA257-264 peptide from ovalbumin (recognized by the transgene-encoded TCR expressed by OT-I mice)20; the Ab-binding OVA323-339 peptide (recognized by the transgene-encoded TCR expressed by OT-II mice)21, 22; or αGalCer. HCs pulsed for 4 hours were washed twice (with medium) and cocultured with purified, naïve CD8+ T cells (from OT-I mice), CD4+ T cells (from OT-II mice), or NKT cells (from Aα−/− KO B6 mice). To some cocultures, either 100 U/mL, or titrated amounts of recombinant, murine interferon IFN-α or IFN-β were added. Supernatants were collected after 24-, 48-, 72-, and 96-hour incubations.
Cytokine Determination by Enzyme-Linked Immunosorbent Assay.
Cytokines were detected in supernatants by double-sandwich enzyme-linked immunosorbent assay (ELISA). Antibodies and recombinant mouse cytokine standards were from BD Biosciences. Extinction was analyzed at 405/490 nm on a TECAN micro plate-ELISA reader (TECAN, Crailsheim, Germany) using the EasyWin software (TECAN).
The Surface Phenotype of HCs Changes When Recently Primed CD8+ T Cells Enter the Liver.
CD8+ T cells primed by an intramuscular injection of a peptide- or DNA-based vaccine migrate to the liver, with peak numbers appearing in the second week postvaccination.4 This was observed when the appearance in the liver of specific Kb- and/or Db-restricted CD8+ T blasts recognizing epitopes from OVA, the large T antigen of simian virus 40, HBcAg, or HBsAg were followed after the injection of antigen-encoding expression plasmid DNA. Between day 10 and day 20 postvaccination, 0.5%–5% of the splenic or hepatic CD8+ T cell populations displayed specificity for epitopes of the immunizing antigen as shown by tetramer staining or specific ex vivo restimulation for 5 hours followed by intracellular IFN-γ staining.4 The specific intrahepatic CD8+ T blasts are viable, produce cytokines, are cytolytic, and can be found for more than 90 days in the liver.4 We tested if HCs respond to the influx of a low number (i.e., 105 to 106) of specific CD8+ T blasts into the liver.
At day 12 postvaccination, HCs freshly isolated from immune (HC*) but not naïve (HC) mice showed upregulated surface expression of MHC class I (Kb, Db) and CD1d (Fig. 1A). Surface expression of MHC class II, Qa-1, CD95, CD54, and costimulator (CD80, CD86) molecules was unchanged or undetectable in HCs* (Fig. 1A; data not shown). Most (>96%) freshly isolated HCs and HCs* were viable (annexin V−), but some cells showed evidence of apoptosis after a 24-hour in vitro culture. In FCM analysis, HCs were gated according their typical size and shape. The isolated HC populations contained no contaminating cells (CD11c+ or B220+) DCs (CD11b+), macrophages (CD3+), T cells, or natural killer (NK) (NK1.1+) or NKT cells (Fig. 1B). The described changes in the HC* surface phenotype were found after vaccinating mice with OVA-, HBsAg-, HBcAg-, or large T antigen of simian virus 40–encoding plasmid DNA (data not shown). The changes in the HC* surface phenotype appeared early in the second week postimmunization, were prominent between day 10 and day 20 postvaccination, and gradually declined thereafter. Up-regulation of MHC class I and CD1d molecules by most HCs indicated that a fairly small number of 105–106 recently primed CD8+ T cells entering this large organ elicits a signal to which a greater than 104-fold number of HCs can respond. We used quantitative real-time polymerase chain reaction analysis and ELISA to test if HCs* show changes in the expression profile of cytokines (or cytokine receptors) that distinguish them from HCs.
Upregulated Expression of Interleukin-10 and Interleukin-4 in HCs*.
We tested if freshly isolated HCs* show upregulated expression of pro- or anti-inflammatory cytokines by quantitatively estimating transcript levels by real-time polymerase chain reaction. The transcript levels of the proinflammatory cytokines TNF-α, interleukin (IL)-1β, IFN-γ, IL-13, or IL-6 in HCs* and HCs were comparable (Fig. 2A, data not shown). Similarly, expression of the IL-19–, IL-20–, and IL-24–binding IL-20 receptor β-chain, a subunit of the class II receptor family for IL-10-homologous cytokines,23 was unchanged, although its expression is strikingly upregulated in HCs in the acute phase response (Fig. 2A, data not shown). In contrast, transcript levels of the anti-inflammatory cytokines IL-10 and IL-4 were upregulated in HCs* compared with HCs (Fig. 2A). These data were confirmed by determining IFN-γ, TNF-α, IL-6, IL-13, IL-10, and IL-4 by ELISA in supernatants conditioned by HCs or HCs* (Fig. 2B). IL-10 and IL-4, but not IFN-γ, TNF-α, IL-6, or IL-13 were detected in supernatants of cultured HCs* but not HCs (Fig. 2B, data not shown). In addition to changes in surface phenotype, HCs* thus show changes in their cytokine expression profile.
Induction of the HC* Phenotype Requires Successful CD8+ T Cell Priming.
HBc-tg mice are tolerant to HBcAg but respond to HBsAg; HBs-tg mice are tolerant to HBsAg but respond to HBcAg. We vaccinated HBs-tg or HBc-tg mice with DNA vaccines encoding either HBsAg or HBcAg to test if the induction of the HC* phenotype requires successful priming of a detectable CD8+ T cell response. The pCI/C DNA vaccine (encoding HBcAg) but not the pCI/S DNA vaccine (encoding HBsAg) induced the HC* phenotype in HBs-tg mice; similarly, the pCI/S DNA vaccine but not the pCI/C DNA vaccine induced the HC* phenotype in HBc-tg mice (Fig. 3A). Expression of the HC vs. HC* phenotype correlated well with successful priming of a CD8+ T cell response, as was evident by the changes in the surface phenotype (data not shown) and the IL-10 (Fig. 3B) and IL-4 (data not shown) expression profile of HCs.
HCs* (But Not HCs) Support Specific Priming of Naïve CD8+ T Cells.
HCs and HCs* were cocultured with purified, naïve CD8+ T cells from OT-I B6 mice. The purity of the isolated CD8+ OT-I T cells was greater than 99% with no contaminating CD4+ T cells, DCs, NKT cells, or NK cells detectable, and the viability of this responder cell population was greater than 98% (Fig. 1B). When naïve CD8+ T cells were cocultured with nonpulsed HCs*, IL-10 and IL-4 but no IFN-γ, TNF-α, IL-6, or IL-13 release was detected (Fig. 4A). When naïve CD8+ T cells were cocultured with HCs* (or HCs) pulsed with increasing amounts of the Kb-binding OVA peptide, they were activated and proliferated only in response to pulsed HCs* (but not HCs). IFN-γ, TNF-α, and IL-13 (but not IL-6, IL-4, or IL-10) were detected in supernatants of cultures in which CD8+ T cells were stimulated with pulsed HC* (Fig. 4A). Cytoplasmic staining showed that IFN-γ and TNF-α were produced by CD8+ T cells cocultured with HCs* but not HCs (Fig. 4B). Specifically activated CD8+ T cells seem to suppress IL-10 and IL-4 release by HCs* in a dose-dependent way that correlates inversely with their release of IFN-γ and TNF-α (Fig. 4A). Specific cytokine induction in cocultures of peptide-pulsed HCs* with naïve CD8+ T cells showed a dose response (Fig. 4A), kinetics (Fig. 5A), and an optimal responder/stimulator ratio (Fig. 5B): a pulse of HCs* with 1 μg/mL antigenic peptide primed optimal, specific IFN-γ and TNF-α release by T cells; the response peaked at 48 hours of coculture; and IFN-γ/TNF-α release were optimal at a responder/stimulator ratio of 10. Culture of OT-I CD8+ T cells without HCs* did not induce IFN-γ release (Fig. 5A). Naïve OT-I CD8+ T cells cocultured with pulsed HCs* (but not pulsed HCs) showed proliferation during 72 hours of culture (Fig. 5C). Release of IFN-γ by HC*-stimulated CD8+ T cells was not downregulated by IFN-α or IFN-β, although these type I IFNs suppressed in vitro priming of OT-I CD8+ T cells by peptide-pulsed DCs (Fig. 5D, E), confirming published reports.24–26 HCs* pulsed with the Ab-binding OVA peptide did not activate purified, naïve CD4+ T cells from OT-II B6 mice (data not shown). Mice vaccinated intramuscularly with the OVA-encoding DNA plasmid pCI/OVA did not contain spleen cells, hepatocytes, or liver NPCs that present the Kb-restricted OVA epitope to naïve or primed OT-I CD8+ T cells without being pulsed with the respective antigenic peptide (data not shown). HCs* induced in response to the influx of a small number of recently primed CD8+ T cells are thus characterized by the upregulation of MHC class I and CD1d and the induction of anti-inflammatory cytokines, but also by the ability to support CD8+ T cell priming when pulsed with an antigenic peptide. We tested in vivo if the entry of recently primed CD8+ T cells only, recently primed CD4+ T cells, or recently activated cells of the innate immune system also induces the HC* phenotype.
Only Recently Primed CD8+ T Cells Induce the HC* Phenotype.
B6 mice immunized with a single intramuscular injection of the pCI/OVA DNA vaccine develop specific CD4+ and CD8+ T cell responses, and a CD4+ T cell–dependent, OVA-specific IgG2a serum antibody response. Immunized MHC class II–deficient (Aα−/− KO) B6 mice develop potent CD8+ T cell responses but no CD4+ T cell or antibody responses to OVA. Immunized MHC class I–deficient (β2m−/− KO) mice develop CD4+ T cell and antibody responses but no CD8+ T cell responses to OVA. Severely immunodeficient (RAG1−/− KO) B6 mice develop no immune response to OVA but activate lymphoid or myeloid cells of the innate immune system in response to plasmid DNA injection. We isolated HCs* 12 days postinjection from these four groups of vaccinated, normal, or mutant B6 strain mice (as well as from noninjected control B6 mice). HCs* expressed the immunostimulating phenotype for naïve OT-I CD8+ T cells only in the groups in which CD8+ T cells were primed (Fig. 6, groups 1 and 2). Hence recently primed CD8+ T cells but not recently primed CD4+ T cells can induce the HC* phenotype.
A single pCI/OVA DNA vaccination of RAG2−/− OT-I B6 mice specifically activates approximately 10% of all CD8+ T cells (data not shown). To confirm that (1) HCs* prime naï CD8+ T cells in vitro and (2) CD8+ T cells are providing the critical signal that induces the HC* phenotype, we performed experiments using naïve CD8+ T cells from nonimmunized RAG2−/− OT-I B6 mice (that contain only naïve CD69− CD44− CD8+ T cells specific for OVA) as responder cells (Fig. 7A) or HCs* from immunized RAG2−/− OT-I B6 mice as stimulator cells (Fig. 7B).
In the first series of experiments, we cocultured naïve RAG2−/− OT-I responder CD8+ T cells (not contaminated by CD8+ T cells with either other specificities, or preactivated or memory subsets) with different HC* (or HC) stimulator cell populations. Naïve CD8+ T cells were efficiently activated by pulsed HC* (but not pulsed HC) from immunized (RAG2−/− or normal) OT-I B6 mice (Fig. 7A).
In the second series of experiments, we used pulsed HCs* from immunized RAG2−/− OT-I B6 mice, and tested their immunostimulating capacity for naïve CD8+ T cells from either (RAG2−/− or normal) OT-I B6 mice or normal B6 mice (Fig. 7B). Naïve responder CD8+ T cells from RAG2−/− OT-I mice were most efficiently activated by pulsed HCs*, although naïve responder CD8+ T cells from normal OT-I B6 mice (but not CD8+ T cells from nontransgenic B6 mice) were also activated. Thus: (1) specifically induced CD8+ T cell activation (without CD4+ T cell coactivation) efficiently induces the immunostimulating HC* phenotype and (2) naïve responder CD8+ T cells (but not contaminating preactivated or memory T cells) are primed without interference of other regulatory T cell subsets.
We tested in vitro if either soluble factors produced by activated CD8+ T blasts, or a direct contact between primed CD8+ T blasts and HCs can induce the HC* phenotype. The cytokines IFN-α, IFN-β, IFN-γ, TNF-α, IL-1β, IL-6, IL-10, and IL-4 did not induce the HC* phenotype in a 48-hour culture (when used in different concentrations). Similarly, neither supernatants from activated CD8+ T blasts, nor the coculture of (naïve or activated) CD8+ T cells with HCs induced the HC* phenotype (data not shown). Thus we have not defined a factor or cell that induces the HC* phenotype, but we have excluded some obvious possibilities.
NKT Cell Activation Does Not Induce the HC* Phenotype.
NKT cells represent a major population of intrahepatic αβ T cells. We considered whether or not activation of NKT cells induces the HC* phenotype. HCs from mice injected intravenously or intraperitoneally with 100 ng or 1 μg αGalCer did not upregulate MHC class I or CD1d surface expression (data not shown), did not show enhanced expression of IL-10, IL-4, or other cytokines (Fig. 8A), and did not specifically activate naïve OT-I CD8+ T cells when pulsed with the relevant peptide (data not shown). NKT cells thus cannot induce the HC* phenotype.
We also considered whether or not HCs* activate NKT cells more efficiently than HCs. When liver CD4+ NKT cells (from MHC class II–deficient Aα−/− B6 mice) were cocultured with αGalCer-pulsed HCs*, they released IL-4 and IFN-γ more efficiently than NKT cells cocultured with pulsed HCs (Fig. 8B). Nonpulsed HCs or HCs* did not stimulate IFN-γ or IL-4 release of cocultured NKT cells. The HC* phenotype specifically induced by recently activated CD8+ T (but not NKT) cells can thus facilitate NKT cell activation.
The novel finding of this study is the change in phenotype induced in the large majority of HCs by the influx of a comparatively small number of recently activated CD8+ T cells. Only activated CD8+ T cells, but not activated CD4+ T cells or NKT cells, induce the HC* phenotype that is characterized by an altered surface phenotype and cytokine expression profile (IL-10, IL-4), as well as the capacity to specifically prime naïve CD8+ T cells.
The changes of the HC* phenotype we describe were induced in vivo. Because immunized and nonimmunized (sex- and age-matched) mice were always analyzed in the same experiments, and their HCs were preparatively isolated by identical protocols using the same reagents, the changes in phenotype did not result from the preparative isolation and purification of HCs. HCs cultured in vitro in RPMI-1640 medium/5% fetal calf serum (with or without cytokines) did not gain the HC* phenotype but started to lose viability after 24 hours of culture. These data exclude that the HC* phenotype results from isolation and/or culture conditions used to transfer HCs in vitro. Furthermore, clustering of naïve OT-I CD8+ T cells with HCs* but not HCs was observed in cocultures.
Changes detected in the surface phenotype of HCs* included upregulation of surface expression of conventional MHC class I (Kb, Db) and CD1d molecules but no changes in the surface expression of MHC class II, costimulator, adhesion or apoptosis-inducing molecules. This suggested that the interaction of conventional CD8+ αβ T cells and NKT cells with HCs is facilitated if they express the HC* phenotype that was confirmed. MHC class II expression was not (or only barely) detectable on the surface of HCs and HCs*, and peptide-pulsed HCs or HCs* did not activate cocultured TCR-transgenic OT-II CD4+ T cells (data not shown).
Only CD8+ T cells induced the HC* phenotype. In the absence of antigen, the upregulation of IL-10 and IL-4 expression by HCs* may counteract specific T cell activation or induce anergy in activated T cells. This has been described in systems in which HCs presented antigen to naïve or activated CD8+ T cells.27–29 Other data suggest that HCs can prime CD8+ T cell responses in vivo,30, 31 but the difference between the HC and the HC* phenotype was not accounted for in these studies. Distinguishing HCs* and HCs may resolve some conflicting data in the field. In epitope-presenting HCs* IL-10 may support CD8+ T cell priming initially, although its production is downregulated once CD8+ T cell activation is underway. IL-10 is not only immunosuppressive—it supports priming of CD8+ T cell and NK cell responses; enhances lipopolysaccharide-stimulated TNF production by myeloid cells32; facilitates T helper cell type 1 development under some conditions33; is required for generating tumor-rejecting CD4+ T cells34; is chemotactic for CD8+ T cells and enhances proliferation and cytotoxicity of primed CD8+ T cells35; and enhances (in synergy with IL-18) IL-2–dependent proliferation, cytotoxicity, and IFN-γ production by NK cells.36, 37 The context in which IL-10 is expressed thus determines its effect on T cells and NK cells.
Interestingly, IL-20R2 and IL-6 expression were not upregulated in HCs*, indicating that the response to recently activated CD8+ T cells differs from the acute phase response of HCs. This is supported by the observation that expression of neither TNF nor IL-1β is markedly upregulated in HCs* (although TNF-α and IL-1β messenger RNA expression was up to twice as high in HCs* as in HCs). The data support the notion that HCs respond with a unique pattern to the influx of recently activated CD8+ T cells into the liver.
Induction of the HC* phenotype by activated CD8+ T cells may be direct or indirect. Factors derived from CD8+ T blasts or by hepatic NPCs stimulated by them may rapidly diffuse through the organ to induce the HC* phenotype in most parenchymal cells. As few as 105 CD8+ T blasts in the liver can induce this phenotype in 1010–1011 HCs. Because a large fraction of CD8+ T cells may pass through the liver in the weeks following immunization—and many of these CD8+ T blasts may undergo apoptosis in the liver—the CD8+ T cell/hepatocyte ratio may not be as extreme as analysis at a single time suggests. The molecular nature of the factor that induces the HC* phenotype is not known but is under active investigation. Type I IFNs, IL-1α, IL-6, IL-10, IL-4, or TNF-α alone do not seem to be involved (data not shown). Chemokines or defensins may be interesting candidates. Alternatively, a focal event in situ triggered by an incoming CD8+ T blast that interacts with a HC may induce a signal that diffuses through the liver by gap junctions connecting HCs.38, 39 This could lead to the rapid intracellular diffusion of the signal exclusively through the parenchymal compartment of the organ. Further experiments are required to define the molecular nature of the HC*-inducing signal and its mode of diffusion throughout the liver.
Selective, epitope-independent retention of activated CD8+ T blasts by the normal liver is a well-established phenomenon,6 but their fate in the liver is controversial. Activated CD8+ T blasts have been reported to proliferate and express specific function while in the liver and to migrate through its different compartments.4, 40, 41 Alternatively, the liver has been proposed to represent a “graveyard” for activated CD8+ T blasts,3 a suggestion supported by the observation that HCs and other liver NPCs can induce apoptosis in activated CD8+ T blasts.28 We found “physiologically” low numbers of specific CD8+ memory/effector T cells primed extrahepatically by a vaccine for more than 3 months within the liver that were not apoptotic and expressed readily inducible specific function.4 HCs are targets for specific CD8+ effector/memory T cells restricted by MHC class Ia42 or class Ib (Qa-1),43 but not for specific CD4+ T effector/memory cells5 that may nevertheless injure the hepatic parenchyma indirectly.44, 45 HCs can prime naïve CD8+ T cell precursors in a CD4+ T cell– and DC-independent way in vitro27 and in vivo.29–31 Our data indicate that HCs must have converted to the HC* phenotype to be able to support CD8+ T cell priming. Two events are initiated when HCs* specifically interact with naïve CD8+ (but not CD4+) T cells: (1) the responding T cells are specifically activated to clonally expand and to produce proinflammatory cytokines, and (2) the immunosuppressive phenotype of the presenting cell is down-modulated. The “danger” signal (i.e., the influx of a small number of recently activated CD8+ memory/effector T cells or their products) can thus change strikingly the intrahepatic, immunomodulatory microenvironment.
Could this observation contribute to our understanding of bystander or autoimmune hepatitis? If the peripheral activation of a “minor” CD8+ T cell response translates into the facilitated presentation of unrelated epitopes by HCs* to naïve CD8+ T cells that overrides the “suppressive” intrahepatic conditions, it would be expected that this would facilitate priming of self-reactive or other potentially injurious CD8+ T cell responses. Once primed, these T blasts could be rescued and further expanded by accessory cells within the hepatic NPC population that cross-present antigen from damaged HCs. It is thus conceivable that early, subtle changes in the HC* phenotype in response to fairly trivial peripheral immune responses pave the way for ensuing T cell immunopathologies that, fortunately, would be a rare event.
We greatly appreciate the excellent technical support of Daniela Schey and Ellen Allmendinger.