Potential conflict of interest: Nothing to report.
The reason the adaptive immune system fails in advanced liver tumors is largely unclear. To address this question, we have developed a novel murine model that combines c-myc–induced autochthonous tumorigenesis with expression of a cognate antigen, ovalbumin (OVA). When c-myc/OVA transgenic mice were crossed with liver-specific inducer mice, multifocal hepatocellular carcinomas co-expressing OVA developed in a tetracycline-dependent manner with a short latency and 100% penetrance. Transferred OVA-specific T cells, although infiltrating the tumor at high numbers, were hyporesponsive, as evidenced by a lack of in vivo cytotoxicity and interferon gamma production. This allowed the tumor to progress even in the presence of large numbers of antigen-specific T cells and even after vaccination (OVA+CpG-DNA). Interestingly, T cell receptor down-modulation was observed, which may explain antigen-specific hyporesponsiveness. This model is helpful in understanding liver cancer–specific mechanisms of T cell tolerance and dissection of antigen-specific and nonspecific mechanisms of immunotherapies in the preclinical phase. (HEPATOLOGY 2009.)
Primary and secondary malignant tumors in the liver often have a dismal prognosis. Hepatocellular carcinoma (HCC) is one of the major causes of cancer-related deaths worldwide.1 Because systemic chemotherapies show limited success, immunotherapies represent an interesting alternative treatment for patients with nonresectable disease.2
Clinical observations suggest that the immune system is important in keeping tumor growth under control.3–5 However, the immune system is frequently unable to control tumor burden in patients suffering from advanced cancer arising in solid parenchymatous organs, such as the liver. One reason could be that the liver represents a tolerogenic environment for T cells.6, 7 Another reason could be that tolerance occurs because of a tolerizing microenvironment specific for advanced autochthonous tumors that arise spontaneously and slowly over time. Third, tumor antigen may be transported to lymphatic tissues for induction of cross-tolerance, which has been described as a mechanism of peripheral CD8+ T cell tolerance induction against self-antigens.8 Recent reports indicate that immune tolerance in spontaneous tumors occurs because of a failure of the adaptive immune system.9, 10 However, the mechanism of tolerance induction in autochthonous tumors remains poorly defined.
Tumor antigen–specific immune responses are difficult to track, both in autochthonous tumor models and in patients, because antigens in spontaneous tumors are largely unknown or not sufficiently immunogenic. To circumvent this limitation, we generated a novel “patient-like” model of HCC, in which the adaptive T cell response can be studied. In particular, we aimed to characterize the adaptive T cell response in advanced autochthonous tumors and to identify mechanisms of T cell recognition of tumor antigen in vivo.
By analyzing the distribution, activation, and effector function of tumor antigen–specific CD8+ T cells, we show that antigen-specific T cells gain access to the tumor microenvironment rapidly. T-cell activation kinetics and distribution of antigen-presenting dendritic cells (DCs) in this tumor model suggest primary antigen recognition in the liver as well as in lymphatic organs, resulting in profound T cell tolerance.
CFSE, succinimidyl ester of carboxyfluorscein diacetate [5(6)]; DC, dendritic cell; HCC, hepatocellular carcinoma; HPF, high-power field; IFN-γ, interferon gamma; LAPtTA, liver-enriched activator protein tetracycline-controlled transactivator; LLN, liver-draining lymph node; OT I, ovalbumin-specific CD8 T cells; OVA, ovalbumin; PCR, polymerase chain reaction; TCR, T cell receptor.
Materials and Methods
C-myc/ovalbumin (OVA) transgenic mice were generated using a bidirectional cytomegalovirus tetracycline-controlled transactivator containing plasmid, pBI-4.11 On one side, the human c-myc complementary DNA, exons 2 and 3,12 was introduced using NotI and SalI and OVA encoding complementary DNA on the other side (provided by F. Momburg, Heidelberg) using MluI and NheI restriction sites. C-myc/OVA transgenic mice were generated by nuclear microinjection of linearized plasmid DNA into the male pronucleus of fertilized mouse oocytes derived from C57BL/6J × F1 C57BL/6J/DBA/2 mice or C57BL/6J mice by standard techniques. C-myc/OVA mice were back-crossed to the C57BL/6J background for at least eight generations. Liver-enriched activator protein tetracycline-controlled transactivator (LAPtTA) transgenic mice13 (Charles River Laboratories, Sulzfeld, Germany) were back-crossed onto the C57BL/6J background for at least 10 generations for subsequent experiments. For tumor induction experiments, c-myc/OVA or c-myc14 transgenic mice were crossed with LAPtTA transgenic mice to obtain double transgenic (c-myc/OVA tg+ or c-myc tg+, respectively) mice and littermate controls (c-myc/OVA tg−).
OVA257-264-specific major histocompatibility complex class I (OT I) T cell receptor (TCR) transgenic mice with and without the Thy1.1 marker on the C57BL/6J background were used as T cell donors for cell adoptive transfer experiments. C-myc transgenic mice were provided by D. Felsher. C57BL/6J and DBA/2 mice were purchased from Harlan Winkelmann GmbH (Borchen, Germany) or Charles River Laboratories. Mice were kept in specific pathogen-free housing in accordance with international guidelines (Federation of European Laboratory Animal Science Associations) and national ethical guidelines and were analyzed at 4 to 10 weeks of age.
Genotypic analysis was performed for c-myc, OVA, and LAPtTA using genomic DNA from tail biopsies: A 319-bp fragment was detected for c-myc using forward (Myc-f: 5′-AGCTTGTACCTGCAGGATCTGAGC-3′) and reverse (Myc-r: 5′-ATCCAGACTCTGACCTTTTGCCAG-3′) primers, a 361-bp fragment using OVA-specific forward (OVA-f: 5′-GATGTTTATTCGTTCAGCCTTGCC-3′) and reverse primers (OVA-r: 5′-CAATCTGGTACATCATCTGCACAGG-3′) and a 450-bp fragment using LAPtTA-specific forward (CamK2tTA-fwd 5′-CGCTGTGGGGCATTTTACTTTAG-3′) and reverse (CamK2tTA-rev 5′-CATGTCCAGATCGAAATCGTC-3′) primers.
Cell Lines and Reagents.
The OVA257-264-specific CD8+ T cell hybridoma B3Z was provided by N. Shastri. The peptide SIINFEKL (OVA257-264) was obtained from Pinedar (Berlin, Germany), ovalbumin from Serva (Heidelberg, Germany), and saponin from Sigma (Munich, Germany).
Isolation of Primary Cells from Murine Liver and Antigen Presentation Assay.
Splenic DCs were isolated by using conventional techniques. After dissection and mechanical disruption, spleen cells were incubated for 30 minutes at 37°C in Grey′s balanced salt solution (Sigma) containing 0.05% collagenase D (Sigma). Afterward, splenic CD11c+ DCs were isolated by two rounds of immunomagnetic separation using CD11c+-labeled MACS microbeads (81%–97% pure) (Miltenyi Biotec, Bergisch Gladbach, Germany). DCs from murine liver (normal liver: 1 preparation pooled from 7–10 mice, tumor liver: 1 liver per experiment) were isolated as described.15 5e4 DCs (CD11c+ MACS purified) were co-cultured with 1e5 B3Z hybridoma cells in 96-well plates for 24 hours in Roswell Park Memorial Institute 1640 + 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid, 10% fetal bovine serum, 1% penicillin/streptomycin, 1% glutamine, 50 μM β-mercaptoethanole. Subsequently, interleukin-2 was quantified using cytometric bead arrays (Beckon Dickinson, San Jose, California).
Tumor Induction and Cell Adoptive Transfer Experiments.
Tumorigenesis was induced in c-myc × LAPtTA and c-myc/OVA × LAPtTA mice at birth by removing doxycycline from the drinking water (2.5% sucrose, 100 mg/L doxycycline). Subsequently, the animals were observed for obvious tumor burden by palpation or signs of distress. For cell adoptive transfers, 1e6 naive, CD8+ MACS-purified (Miltenyi Biotec) OT I T cells (>95% pure) were carboxyfluorescein succinimidyl ester (CFSE) labeled (1 μM, 10 minutes at 37°C) and transferred into 4-week-old mice by tail vein injection. For CD69 assessment, 1e7 T cells or 1e6 T-cells were used. OT I T-cell proliferation was quantified by the reduction of CFSE intensity with cell division.16
Reverse Transcription Polymerase Chain Analysis.
RNA isolation from tissue was carried out with RNAeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol including a deoxyribonuclease I digestion step. One hundred nanograms total RNA was used for complementary DNA synthesis with First Strand Synthesis Kit (Invitrogen, Karlsruhe, Germany). Polymerase chain reaction (PCR) reactions were carried out at 30 cycles using the following primer pairs: Act-F: 5′- GGTCAGAAGGACTCCTATGT-3′ and Act-R: 5′- ATGAGGTAGTCTGTCAGGTC-3′. PCR analysis using Myc-f, Myc-r, OVA-f, and OVA-r primer pairs was carried out as described previously.
Histological and Immunohistochemical Analysis.
Two-micron paraffin sections of formalin-fixed tissue were stained with hematoxylin-eosin. For OVA immunohistochemical staining, a rabbit polyclonal antiserum was generated by immunization and separation from albumin using saturated ammonium sulfate solution. Paraffin sections were stained with concentrated rabbit antiserum against OVA protein (1:1000) after epitope retrieval (citrate buffer, pH6.0) and developed using the DakoREAL Detection System and Peroxidase/AEC (Dako, Hamburg, Germany). Specificity of staining was assessed by lack of staining in c-myc × LAPtTA transgenic mice. Apoptotic tumor nuclei were semiquantified using the ApopTag in situ apoptosis kit (Chemicon, Nürnberg, Germany) by counting numbers of positive staining nuclei in 10 high-power fields (HPF) showing tumor areas. Subsequently, the mean and standard deviation of terminal deoxynucleotidyl transferase-mediated nick-end labeling–positive nuclei per HPF for each liver section was determined.
For immunostaining of cryosections, 6-μm frozen, acetone-fixed sections were blocked using the Biotin Blocking System (Dako) and FcγR-block (anti CD32/CD16, clone 2.4G2, 1:100, Immunotools, Friesoythe, Germany). Subsequently, sections were stained with the primary antibody overnight (anti CD3 Alexa 488, clone 500A2, 1:100, Invitrogen, and anti-CD8 bio, clone 53.-6.7, 1:400, BD Biosciences), subsequently with the secondary reagent (avidin Alexa 568, 1:1000, Invitrogen), counterstained with 5 μg/mL 4′-6-diamidino-2-phenyl-indole, dilactate (Invitrogen) and viewed using a DM LB fluorescence microscope (Leica) and appropriate software (Diskus, Hilgers, Königswinter, Germany).
Flow Cytometric Analysis.
Spleen and lymph node cell suspensions were prepared by mechanic dissociation. For surface staining, cells were stained with fluorochrome-conjugated antibodies (fluorescein isothiocyanate, R-phycoerythrin, PerCP-Cy5.5, phycoerythrin-cyanin-5.5, phycoerythrin-cyanin-7, allophycocyanin, or allophycocyanin-cyanin-7) for 20 minutes at 4°C. FcγR-blocking reagent (anti-CD32/16, 1:200, Immunotools) was added to prevent nonspecific binding. All monoclonal antibodies were purchased from BD Bioscience (Heidelberg, Germany), Immunotools, or Invitrogen. In the case of biotin-conjugated antibodies, cell-bound antibody was detected with streptavidin conjugates (SA-allophycocyanin-Cy7, 1:1000, Pharmingen, Hamburg, Germany). Seven-amino-actinomycin D (BD Biosciences, Heidelberg, Germany), or TO-PRO-3 iodide (Invitrogen) was added to exclude nonviable cells. Data analysis was performed using a FACSCanto flow cytometer (BD Biosciences) and BD FACSDiva or Flow Jo software (Tree Inc., San Carlos). Detection intracellular interferon gamma (IFN-γ) production was performed as described.17
In Vivo Cytotoxicity Assay.
For cell adoptive transfers, 1e6 CD8+ purified naive OT I T cells were injected in 200 μL phosphate-buffered saline via the tail vein. Five days later, in vivo cytotoxicity assays were performed as described.16 To prepare target cells, splenocytes from C57BL6/J mice were either labeled with a low concentration of CFSE (Invitrogen, 0.25 μM, 20 minutes at 37°C, CFSElow cells) or pulsed with SIINFEKL peptide (1 μg/mL, 15 minutes at 37°C) and labeled with a high concentration of CFSE (2.5 μM, CFSEhigh cells). The 1e7 targets at a 1:1 ratio of CFSElow and CFSEhigh cells were injected intravenously. Mice were sacrificed after 4 hours, and target cells in the liver and spleen were quantified by flow cytometry. To calculate specific lysis, the following formula was used: % specific cytotoxicity = 100 − [100 × (CFSEhigh/CFSElow) primed/(CFSEhigh/CFSElow) control].
Survival analysis was performed by the Kaplan-Meier method. Independent Student t tests for k = 2 or simple analysis of variance with post hoc tests (Bonferroni) for k > 2 groups were performed using SPSS 14.0. P-values of <0.05 were considered significant.
Generation and Characterization of an Autochthonous Liver Tumor Model Co-expressing a Cognate Tissue-Specific Antigen.
Mice carrying two tetracycline-dependent transgenes, the proto-oncogene c-myc and a well-characterized model antigen, cytosolic chicken ovalbumin (OVA), were generated. Subsequently, these c-myc/OVA mice were crossed to transgenic mice expressing the tetracycline-controlled transactivator element under the control of a hepatocyte-specific promotor (LAPtTA mice)13 to allow liver-specific expression of c-myc and OVA (Fig. 1A). Transgene expression and tumorigenesis were induced in double transgenic c-myc/OVA × LAPtTA mice (c-myc/OVA tg+) by withdrawing doxycycline from the drinking water at birth. Subsequently, the animals were observed for obvious tumor burden by palpation or signs of distress and analyzed at different time points as indicated.
Double transgenic c-myc/OVA tg+ mice became moribund with multifocal HCC within 5 to 8 weeks (Fig. 1B), similarly to c-myc × LAPtTA mice (c-myc tg+).18 Initial tumor foci were observed as early as 4 weeks after induction in some mice (data not shown). Macroscopic and microscopic findings (Fig. 1C) in 5-week-old tumor-bearing c-myc tg+ mice resembled HCC14 with invasive tumor cells in glandular and solid pattern infiltrating adjacent normal liver tissue. This indicates that OVA protein expression or an endogenous immune response against OVA-expressing tumor cells did not compromise tumor development.
OVA protein expression was specifically detected in tumor cells but not in adjacent normal hepatocytes by immunohistochemistry (Fig. 1C). OVA was not expressed in tumor cells of c-myc × LAPtTA mice (data not shown). By reverse transcription PCR analysis, OVA encoding messenger RNA transcripts were detected in tumor nodules as well as within surrounding non-neoplastic liver tissue of induced c-myc/OVA × LAPtTA mice (Fig. 1D). However, importantly, OVA transcripts were not detected within lymphatic organs. Littermate control mice (c-myc/OVA tg−) did not develop hepatic tumors (Fig. 1C) and did not express OVA by reverse transcription PCR (Fig. 1D) or western blot analysis (data not shown) or immunohistochemistry (Fig. 1C). These data confirm liver-specific transgene expression using the LAPtTA system13 and indicate increased OVA expression in malignant cells.
Antigen-Specific T Cells Migrate into the Tumor Microenvironment of Autochthonous Liver Tumors.
OVA protein expression in tumor cells allowed us to test where this tumor antigen was recognized by naïve CD8+ T cells in vivo and whether tumor antigen–specific T cells were able to mount an efficient immune response to autochthonous hepatic tumors. Endogenous CD8+ T cells recognizing the immunodominant OVA peptide (SIINFEKL) were similarly low in number in the spleen of tumor-bearing c-myc/OVA tg+ and littermate control mice, as evidenced by clonotypic antibody staining and kb/SIINFEKL pentamer staining reagents (data not shown).
Therefore, to study the tumor-specific cellular immune response in greater detail, naïve OVA-specific TCR transgenic CD8+ T cells (OT I) were transferred into c-myc/OVA tg+ mice. Antigen-specific T cell homing into the liver tumor did occur within 24 hours, because CFSE-labeled OT I T cells were detected primarily in tumor nodules of c-myc/OVA tg+ mice (Fig. 2A). In contrast, in c-myc/OVA tg− control mice, T cell homing occurred primarily into the T cell zone of the spleen and lymph nodes (Fig. 2A). Likewise, CFSE-labeled OT I T cells were found at a high frequency in the liver by flow cytometry 24 hours after transfer (Fig. 2B). Five days after adoptive transfer, CD8+ T cells were present and increased within the tumor microenvironment (Fig. 2C, D). This suggests that T cells had gained access to the tumor stroma, and a vascular barrier did not prevent T cell migration. However, T cell infiltration did not lead to necrosis or apoptosis of tumor cells as shown by standard morphology and terminal deoxynucleotidyl transferase-mediated nick-end labeling staining (Fig. 2C). Apoptotic tumor nuclei were not increased, when comparing c-myc/OVA tg+ mice injected with OT I T cells (100.9 nuclei/HPF ± 22.9, n = 4) with noninjected c-myc/OVA tg+ mice (105.2 nuclei/HPF ± 11.5, n = 4)(P = 0.75) (Fig. 2D). Furthermore, a change of apoptotic nuclei was not observed in the normal surrounding liver, indicating that OT I T cells did not cause liver cell injury.
Early activation and expansion of antigen-specific T cells in the liver and spleen of tumor-bearing mice results in T cell tolerance. Because no obvious tumor rejection had occurred after T cell transfer, we investigated whether transferred OT I T cells had recognized the tumor antigen in vivo. Proliferation and expression of the early activation marker CD69 in antigen-specific OT I T cells was used as a surrogate marker for antigen recognition.19 Interestingly, most transferred OT I T cells up-regulated CD69 both in the liver and in the spleen of c-myc/OVA tg+ mice, in contrast to T cells transferred into c-myc/OVA tg− control mice at 24 hours after adoptive transfer (Fig. 3A, P < 0.001). At 24 hours, OT I T cell expansion was greater in the liver (6.6% ± 1.2, P < 0.001) than in the spleen (0.3% ± 0.1; P < 0.001) (P < 0.001; n = 4–6 per group), suggesting that initial antigen recognition and expansion occurred mostly in the liver (Fig. 2B). Whereas at 66 hours after adoptive transfer 47.4% of all CD8+ T cells in the liver and only 4.4% in the spleen of c-myc/OVA tg+ mice were antigen specific, 5 days later, OT I T cells were found in relative high numbers in the liver, but also were significantly increased in the liver-draining lymph node (LLN) > spleen > inguinal lymph node (Fig. 3B, P < 0.05). This indicates that in secondary immune organs, expansion of antigen-specific T cells is slightly delayed.
A central question in developing effective immunotherapies is how and where tumors induce T cell tolerance.20, 21 In our model, rapid antigen-specific T cell activation (CD69) and synchronous proliferation in the local tumor microenvironment (liver) and the spleen had already taken place 24 hours after transfer (Fig. 3A). Thus, naïve tumor antigen–specific T cells had gained access to the tumor microenvironment very early on, whereas a small fraction also recognized tumor antigen in the spleen. Supporting the possibility that antigen was recognized in the spleen as well as in the liver, DCs presenting endogenous OVA could be isolated from both organs of c-myc/OVA tg+ tumor-bearing mice. Interestingly, DCs were greatly increased in number in livers of c-myc/OVA tg+ tumor-bearing mice compared with normal liver (Fig. 3C). DCs from the liver and the spleen were able to present endogenous OVA to antigen-specific T cells (B3Z) in vitro, as shown by interleukin-2 synthesis (Fig. 3D; P < 0.05, P < 0.001).
Antigen-Specific T Cell Tolerance in Autochthonous Liver Tumors.
Recent reports with sporadic immunogenic tumors arising in multiple organ systems suggest that progressed autochthonous tumors induce T cell tolerance. We therefore analyzed the fate of tumor antigen–specific T cells in our autochthonous tumor system. Transferred OT I T cells failed to synthesize IFN-γ in the liver and spleen 66 hours (Fig. 4A; P < 0.05) and 5 days (Fig. 4A; P < 0.05) after adoptive T cell transfer. In addition, OT I T cells did not show significant in vivo cytotoxicity against OVA-peptide–loaded target cells in c-myc/OVA tg+ tumor-bearing mice, when compared with c-myc/OVA tg− littermates (Fig. 4B, liver: P = 1.0, spleen: P = 0.174). In contrast, co-administration of cytosine-guanosine dinucleotide synthetic oligonucleotide 1668 and OVA (CpG OVA) in wild-type mice, which has been shown to promote OVA-specific cytotoxic T cell responses,22 did not result in cytotoxic T lymphocyte effector cells. In summary, the lack of in vivo cytotoxicity and IFN-γ synthesis in c-myc OVA tg+ tumor-bearing mice provides evidence that tumor-specific expression of OVA resulted in systemic T cell tolerance of OT I T cells.
T Cell Receptor Down-Regulation in Transferred Antigen-Specific T Cells.
Next, we investigated, why OT I T cells failed to be successfully primed and become cytotoxic effector cells. Interestingly, beginning at 24 hours and lasting up to 5 days after transfer (Fig. 5A and data not shown), antigen-specific T cells partially down-modulated the TCR, as evidenced by antibody staining with clonotypical antibodies (P < 0.001). TCR down-modulation occurred specifically in transferred clonotypical T cells in contrast to endogenous Vβ5.1, 5.2 TCR-expressing T cells (Fig. 5A, right histogram). Down-regulation did not occur in activated T cells (Fig. 5A, left histogram), only occurred in antigen-specific but not wild-type transferred T cells (data not shown), and was sustained up to 5 days after transfer (Fig. 5B), indicating that this was not a transient phenomenon during T cell activation.
Immunization Using Toll-Like Receptor-9 Ligands Does Not Overcome Tolerance Induced in Autochthonous Tumors.
Because tolerance was induced on antigen recognition of naïve OT I T cells in tumor-bearing c-myc OVA tg+ mice, we investigated whether tolerance could be broken when c-myc OVA tg+ mice were immunized using a combination of Toll-like receptor 9 stimulation and antigenic challenge (CpG and OVA protein). Even after vaccination of tumor-bearing mice with CpG/OVA, transferred OT I T cells failed to become cytotoxic effector cells in contrast to immunized control mice (Fig. 6; P < 0.01). These findings demonstrated that cytotoxic T lymphocyte tolerance induction was robust and could not be broken by pro-inflammatory stimuli.
Although much has been learned about recognition of tumor antigens by T cells in tumor transplant models, most experimental setups investigating antigen-specific responses involve transplantation of cultured tumor cell lines expressing a cognate antigen into syngeneic mice. This approach likely creates an artificial pro-inflammatory context, which may resemble the early but not the late stage of tumorigenesis.23, 24 The tumor stroma in these models contains cells and factors specific for tissue injury, not invasive tumor growth over a long period.
Here we present a liver tumor model in which tumors arise endogenously, and with a latency that depends on the time point of transgene expression. This allowed us to investigate the antigen-specific T cell response in a tumor microenvironment that more closely mimics the clinical situation.
In autochthonous liver tumors, we observed profound OVA-specific T cell tolerance. Interestingly, in transgenic mouse models with OVA expression in normal hepatocytes,25 OVA-specific CD8+ T cells cause autoimmune hepatitis, not tolerance. In our model, we could not use serum alanine aminotransferase as a measurement of liver cell injury, because this parameter was elevated because of tumor growth. However, in our model no hepatocellular damage was observed histologically in adjacent normal liver (data not shown). Although other parameters such as administration of CD4+ OVA-specific T cells or CD1d-restricted invariant natural killer T cells26 still need to be tested in our model, OVA-specific T cells in c-myc OVA tg+ tumor-bearing mice do not induce overt bystander autoimmune hepatitis. Interestingly, when OVA is expressed in normal hepatocytes,25 OT I T cells proliferate in the liver and the spleen in a time-delayed fashion, suggesting priming of CD8+ T cells in the liver and redistribution to the spleen. This is in contrast to our tumor model, where priming apparently occurs at the same time in the liver and in secondary immune organs based on activation and proliferation kinetics.
CD8+ T-cell tolerance in c-myc OVA tg+ tumor-bearing mice may be one explanation for why advanced liver tumors are not easily eliminated by the endogenous adaptive immune response. Tolerance in CD8+ T-cells has also been observed in autochthonous tumor models affecting other organs, such as prostate cancer27 and breast cancer.28 The c-myc/OVA model is applicable to other organ sites where c-myc plays a role in tumorigenesis, such as breast, ovarian, prostate, colon, small cell lung, and cervical carcinomas, osteosarcomas, glioblastomas, melanomas, and myeloid leukemias,29 and thus may facilitate dissection of antigen-specific and antigen-nonspecific immune mechanisms of tumor-induced tolerance in other organs.
Interestingly, when both transgenes were expressed in the liver, overexpression of OVA protein in tumor cells was observed predominantly in tumor nodules, but not in the normal nonneoplastic surrounding liver tissue. This could be attributable to an overall increase in protein expression in tumor cells or a c-myc–dependent regulation of OVA expression. Although OVA expression was restricted to the liver, OVA antigen presentation was detected in splenic as well as liver DCs. Because moribund c-myc/OVA tg+ mice did not show macroscopic or microscopic evidence of metastases (data not shown) and OVA messenger RNA and protein was only detected in the liver but not in the spleen or lymph nodes, most likely, OVA was derived from the primary tumor. Tumor antigen could be transported by DCs from the liver to the spleen or escape from the tumor liver into the circulation via apoptotic bodies or tumor-derived exosomes.30 C-myc–induced tumors have a very high turnover, mitotic activity, and apoptotic rate,14 which may cause tumor antigen to be produced in large amounts leading to systemic distribution in the body.
In this model of autochthonous liver cancer, we observed TCR down-regulation and subsequent profound tumor antigen-specific tolerance. TCR down-regulation in our system may explain why T cells do not recognize their target antigen in vivo efficiently and instead become hyporesponsive, as continuous TCR signaling is required for optimal cytotoxic T lymphocyte function in vivo.31 Interestingly, TCR down-regulation has been reported as a mechanism of T cell tolerance to hepatic auto-antigen.32 Here, a partial down-modulation of the TCR was reported with low level, compared with a complete down-modulation with high level of major histocompatibility complex I/peptide expression on hepatocytes. T cell tolerance to orally fed antigen is also associated with TCR down-regulation followed by a deletion of antigen-specific T cells in vivo and probably caused by a high dose of antigen.33 Antigen taken up by the gastrointestinal tract has been shown to be presented by tolerogenic antigen-presenting cells within the liver34 and spleen.19 It remains to be shown whether in autochthonous liver tumors liver resident or tumor-associated antigen-presenting cells are responsible for T-cell tolerance induction. Furthermore, it would be interesting to investigate other mechanisms of intrahepatic T cell tolerance induction, such as Bim-dependent T cell death.35
Our findings may be of considerable importance in understanding immune failure in cancer patients with hyporesponsive CD8+ T cells.36 Interestingly, in some clinical situations effector function can be restored on removal of persistent antigen,37 indicating that here, the amount of tumor antigen plays a role in determining immunological outcome. Interestingly, in patients with malignancies,38–40 TCR signaling is interrupted via down-regulation of the TCR ζ chain.
How stable is T cell tolerance in autochthonous tumors? Even after vaccination of tumor-bearing c-myc/OVA tg+ mice with CpG/OVA, transferred OT I T cells failed to become cytotoxic effector cells. These findings suggest that tolerance induction in autochthonous liver tumors is a dominant effect that cannot simply be overcome by immunization. Although other modes of immunization need to be tested, our observations may explain why experimental immunotherapies targeting tissue-specific antigens are effective against transplanted tumor cells but inefficient against spontaneous tumors.21
We conclude that induction of tumor antigen-specific CD8+ T cell tolerance and TCR down-modulation represents an obstacle for future immunotherapies in liver tumors. Possibly, vaccination or adoptive T cell transfer41 and therapies targeting the organ-specific tumor stroma42 need to be combined to be successful in patients. These results emphasize the need for preclinical testing in autochthonous tumor models.
The authors thank A. Egert for performing pronuclear injections in the C57BL/6J background, M. Berg, B. Schumak, S. Klein, and Z. Luo for assistance with experiments, M. Hessler and S. Steiner for technical assistance, T. Höller for assistance with the statistical analysis, L. Heukamp, R. Jäger, and S. Weber for technical advice, assistance, and discussions, and H. Kalthoff for helpful discussions.