INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Transgenic models have been developed to study the mechanisms of tolerance and their implications for autoimmune or other immune-mediated diseases. In this regard, transgene-encoded neo-self antigen coupled with the corresponding T-cell receptor transgene has been particularly valuable (4, 20, 23, 35). In addition to central thymic selection, peripheral tolerance mechanisms, including peripheral deletion, anergy, and ignorance have been defined (5, 10, 27, 28, 36). In the latter case, it is often possible to break tolerance and induce autoimmunity, leading to immune-mediated tissue injury. Expression of neo-self antigens in the liver presents a particular interesting scenario because of the putative toleragenic role of the liver in immune response and the unique anatomy of the liver in which the fenestrated vasculature allows direct access of hepatocytes to circulating T cells (25). This intriguing question has been addressed in several transgenic models in which central and peripheral deletion of reactive T cells appears to confer a robust tolerance to the neo-self antigen expressed in the transgenic liver. In situations whereby peripheral anergy or ignorance induction is operative, tolerance at the T-cell level can be broken by either viral infection or dendritic cell or DNA immunization (23, 31, 35, 37). However, induction of hepatitis still requires adoptive transfer of a large quantity of antigen-specific T cells in most cases (35).

We have previously reported the generation of several transgenic lines expressing hepatitis C virus (HCV) structural proteins (core, E1, and E2) (16). The liver-specific expression of HCV mRNA and proteins could be demonstrated by reverse transcriptase PCR and by Western immunoblotting and immunohistochemistry, respectively. However the expression level is relatively low. The mice did not exhibit any long-term pathological effects from the expression of the HCV proteins. In this study, we analyze the effects of HCV DNA immunization on both wild-type and transgenic mice and demonstrate an interesting hierarchy of immune response and tolerance induction against the HCV structural proteins.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mice. Female FVB/n (H-2^q) and BALB/c (H-2^d) mice 6 to 8 weeks old were purchased from Charles River Laboratories (Wilmington, Mass.). AAD mice, namely, those expressing the transgene with the 1 and 2 domains from human HLA-A2.1 and the 3 domain of murine H-2D^d in the C57BL/6 background, have been described previously (32) and were obtained from Victor Engelhard of the University of Virginia. A transgenic mouse lineage (AC1-0) expressing HCV 1b structural proteins, core, E1, and E2, under a liver-specific albumin promoter was generated in the FVB/n background as described previously (16).

Plasmids for DNA immunization. For DNA immunization, two different plasmids, pHCVcore and pHCVSt expressing the HCV core (amino acids [aa] 1 to 191) and core/E1/E2 polyprotein (aa 1 to 830) under the control of the cytomegalovirus (CMV) promoter (Fig. 1), were constructed by PCR of HCV cDNA from a type 1b strain that was also used for the production of the transgenic mouse (16). The PCR product was cloned into a CMV expression plasmid, WRG7020, which is an efficient vector for DNA immunization in mice (30). These plasmids have been shown to direct the expression and proper processing of the HCV structural proteins in transfected HuH7 hepatoma cells. All plasmid DNAs were purified with an endotoxin-free plasmid extraction kit (Qiagen, Chatsworth, Calif.) and were dissolved in phosphate-buffered saline at a concentration of 2 µg/µl. For DNA immunization, mice were first injected with 0.25% bupivicaine into both quadriceps muscles and 24 h later were inoculated with 50 µl of the plasmid DNA into each of the quadriceps. Booster injections were given following the same protocol every 21 days for three times. All mice were bled by retro-orbital puncture prior to each immunization and 2 weeks after the last immunization. The sera were assayed for anticore and anti-E1/E2 antibodies.

View larger version (28K): [in this window] [in a new window]   FIG. 1.   Schematic diagrams of various HCV plasmids. The construction of these plasmids is described in Materials and Methods.

Recombinant HCV proteins. Full-length core protein of HCV 1b was purified from Sf9 insect cells infected with the recombinant baculovirus bvHCV.S at a multiplicity of infection of 5. The infected Sf9 cells were lysed with buffer containing 50 mM Tris-HCl, 50 mM NaCl, 0.5% NP-40, 1 mM EDTA, and 1× protease inhibitor cocktail (Boehringer Mannheim Biochemicals, Indianapolis, Ind.). The lysate was centrifuged at 20,000 × g to pellet cellular debris, and the supernatant was applied to preparative sodium dodecyl sulfate (SDS)-14% polyacrylamide gel electrophoresis (PAGE) using Model 491 Prep Cell (Bio-Rad, Richmond, Calif.). Collected fractions were assayed by immunoblotting with C1 anticore monoclonal antibody; positive fractions were pooled, concentrated with Centricon 3 (Millipore, Bedford, Mass.), and subjected to a repeated preparative SDS-PAGE for further purification. The resulting core protein is more than 95% pure on an SDS-PAGE gel. Recombinant truncated HCV core protein was purified using the QIAexpress System (Qiagen). Purification of HCV E1 and E2 proteins from BHK-21 cells infected with a recombinant vaccinia virus expressing the HCV structural proteins (vvHCV.S) has been described elsewhere (3).

Establishment of syngeneic cell lines expressing HCV structural proteins. To establish stably transfected syngeneic cell lines expressing HCV proteins, core (aa 1 to 192), E1(aa 137 to 383), E2 (aa 367 to 830), and core/E1/E2 (aa 1 to 830) expression plasmids (pEF-core, pEF-E1, pEF-E2, and pEF-St) were separately constructed by PCR. An AUG start codon and a stop codon were introduced by the PCR primers at each end. The PCR product was inserted into an EF-1 promoter-driven expression plasmid by replacing the enhanced green fluorescent protein gene of the pEF-EGFP plasmid; this plasmid was derived from the EGFP-N1 plasmid (Clontech, Palo Alto, Calif.) from which the CMV promoter was replaced with the EF-1 promoter from the pEF321-Neo plasmid (provided by Tatsuo Miyamura, Tokyo, Japan) (17). The P815 (H-2^d) and SQSV cell lines (H-2^q) (from Margaret Koziel, Beth Israel-Deaconess Hospital, Boston, Mass.) (18) were transfected with various HCV plasmids or the pEF-EGFP plasmid and were selected by Geneticin (Sigma, St. Louis, Mo.). Selected cells were screened by immunofluorescence and then confirmed by Western immunoblotting with antibodies against core, E1, or E2. Positive clones were expanded as stimulator and/or target cells for cytotoxic T-lymphocyte (CTL) assay.

Anti-HCV ELISA, T-cell proliferation, and cytokine assays. The enzyme-linked immunosorbent assay (ELISA) for detection of anticore or anti-E1/E2 antibodies has been described earlier (3). Spleens were removed from immunized mice 2 weeks after the last boost. After preparation of single-suspension cells, spleen cells were cultured in triplicate for 5 days by using a U-bottomed 96-well plate at 2 × 10⁵ cells/well in 200 µl of RPMI 1640 containing 10% fetal bovine serum, 4 mM L-glutamine, 100 U of penicillin/ml, and 100 µg of streptomycin/ml. The cells were stimulated with recombinant core or envelope protein at 1 µg/ml. Nonspecific activation was obtained using 1 µg of phytohemagglutinin (Sigma)/ml as positive stimulation. On day 6, [³H]thymidine was added (1 µCi/well), and the cells were incubated for an additional 18 h. The [³H]thymidine incorporation into DNA was measured after harvesting the cells. Incorporation of radioactivity was corrected for background activity. To determine cytokine production, the effector cells stimulated with 1 µg of recombinant proteins/ml were cultured as described above at 5 × 10⁵/well and the supernatants were harvested on day 3. Gamma interferon (IFN-) and interleukin-4 levels were measured by Quantikine M (R&D Systems, Minneapolis, Minn.).

CTL assay. Splenocytes derived from immunized mice (10⁷/well) were stimulated in six-well culture plates for 7 days with 5 × 10⁵ irradiated (10,000 rads) stably transfected cells expressing pEF-ST as stimulator cells in RPMI 1640 that contained 10% fetal bovine serum, 4 mM L-glutamine, 100 U of penicillin/ml, and 100 µg of streptomycin/ml. Culture medium was supplemented with 10% T-Stim (Collaborative Biomedical Products, Bedford, Mass.) on days 2 and 5. After restimulation of these effector cells for another week with the same stimulator cells and irradiated (3,000 rads) spleen cells from the FVB/n mouse as feeder cells, the CTL activity was determined in a standard ⁵¹Cr release assay using U-bottomed, 96-well plates containing 3,000 ⁵¹Cr-labeled target cells per well. For FVB/n and BALB/c mice, stably transfected syngeneic cells expressing various HCV structural proteins were used as target cells. Target cells (10⁶) were pulsed with 100 µCi of ⁵¹Cr for 1 h, washed three times, and added to the plates containing different number of effector cells in a final volume of 200 µl. Effector and target cells were cocultured in duplicate, and the supernatants were harvested for analysis after 5 h of incubation. The percentage of specific ⁵¹Cr release was calculated as 100 × (experimental release  spontaneous release)/(maximum release  spontaneous release). Spontaneous release was determined from target cells incubated without effector cells, and maximum release was determined in the presence of 1% SDS.

ELISpot assay. The ELISpot assay for IFN- has been described previously (21). Either recombinant proteins (core or E1/E2 protein) or peptides are incubated with splenocytes to determine the HCV-specific CD4 or CD8 responses, respectively. A peptide (aa 131 to 140; sequence ADLMGYIPLV) representing an A2-restricted epitope in the core region was used (32). Since the previously identified A2-restricted E2 epitopes were not conserved in our HCV strain, we screened a panel of overlapping peptides spanning the E2 region of our strain and identified several peptides that were positive in the A2-restricted CTL assay (K. Murata and T. J. Liang, unpublished results). One of the strongly reactive peptides (aa 614 to 622; RLWHYPCTI) was used for assaying E2-specific CTL activities in the AAD mice. The spots, as counted by KS Elispot-Axioplan 2I (Zeiss, Thornwood, N.Y.), are expressed as per 10⁶ cells.

Histological evaluation and adoptive transfer. Tissues were fixed in formalin buffered with phosphate-buffered saline. Sections of paraffin-embedded tissue were cut at 5-µm thickness and stained with hematoxylin and eosin. Wild-type FVB/n mice were immunized with pHCVSt plasmid four times. Splenocytes were harvested after immunization and were stimulated with syngeneic cells expressing core/E1/E2 for 1 week and were assayed for CTL activities. Stimulated splenocytes (n = 10⁷) were then injected into transgenic mice, and serum and liver were harvested 1 week later for alanine aminotransferase and histological analysis, respectively.

Statistical analysis. Comparisons of the mean antibody levels and proliferative and CTL activities between groups of mice were analyzed using the Student t test. All tests were two-tailed, and differences were considered significant when P was < 0.05. Percentages of positive responses between groups were compared using Fisher's exact test.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Antibody response. Wild-type FVB/n mice and HCV transgenic mice were immunized four times with the plasmid vector, pHCVcore, or pHCVSt. Two weeks after the last DNA immunization, the wild-type FVB/n mice showed a weak humoral immune response with low levels of anticore and/or antienvelope antibodies in only a fraction of the immunized animals (Fig. 2, top panel). Immunized transgenic mice demonstrated higher titers of antibody response, the reason for which is not clear at present. However, similar to the wild-type mice, only a fraction of animals had detectable humoral response (bottom panel).

View larger version (40K): [in this window] [in a new window]   FIG. 2.   Antibody responses in wild-type and HCV transgenic mice immunized with HCV plasmids. Wild-type FVB/n mice were immunized four times with the plasmid vector (n = 4), pHCVcore (n = 5), or pHCVSt (n = 6). HCV transgenic mice were similarly immunized with the plasmid vector (n = 4), pHCVcore (n = 5), or pHCVSt (n = 5). Two weeks after the last DNA immunization, serum was obtained and analyzed for anticore and anti-E1/E2 at a dilution of 100. The results are shown as optical density (O.D.) of the ELISA, and the cutoff of 0.2 represents a signal-to-noise ratio of 2.5 (dotted line).

T-cell proliferative response. To analyze the presence of cellular immune response, splenocytes were isolated from immunized mice 14 days after the last immunization and were assayed for proliferative T-cell response using either core or E1/E2 as stimulating antigens (Fig. 3). pHCVcore-immunized wild-type mice demonstrated little or no response against the core antigen, whereas several wild-type mice immunized with the pHCVSt construct exhibited positive activities against either the core (three of six) or envelope proteins (three of six). Two of five transgenic mice immunized with pHCVcore had a weakly detectable proliferative response against the core protein, but the overall difference from the wild-type response was not statistically significant. However, the pHCVSt-immunized transgenic mice were completely nonresponsive to either core or envelope proteins. Analyses of cytokine secretion of the proliferative assays showed IFN- production to be predominant (Fig. 3 legend) in positive samples, consistent with a TH1 response by intramuscular DNA immunization (9).

View larger version (39K): [in this window] [in a new window]   FIG. 3.   T-cell proliferative responses in wild-type and HCV transgenic mice immunized with various HCV plasmids. Immunized mice were sacrificed 2 weeks after immunization, and splenocytes were harvested for T-cell proliferative response against HCV antigens (1 µg/ml) as described in Materials and Methods. The results are shown as stimulation index (S.I.), and a stimulation index of greater than 3 is considered positive (dotted line). The wild-type animals which showed positive proliferative response produced IFN- and not interleukin-4 in the medium (animal 2, which showed positive response to core and E1/E2, produced 37 and 47 pg of IFN-/ml, respectively; animals 3 and 5, which were positive to E1/E2 alone, produced 96 and 224 pg of IFN-/ml, respectively)

CTL response. Splenocytes from immunized mice were stimulated with syngeneic cells expressing core, E1, and E2 for 14 days and were assayed for CTL activities. Typical CTL assays of each group against core-, E1-, E2-, or core/E1/E2-expressing cell lines were shown in Fig. 4A. The pHCVcore-immunized mouse showed positive cytolysis against core-, E1-, and core/E1/E2-expressing target cells. The E1-expressing construct contains the C-terminal region of core, therefore probably accounting for the cross-reactivity. pHCVSt-immunized wild-type mice exhibited strong CTL activities against the E2 and core/E1/E2-expressing target cells but none or little against the core- and E1-expressing cells. In transgenic mice, pHCVcore immunization led to positive cytolytic activities against the core-, E1-, and core/E1/E2-expressing cells. However, pHCVSt-immunized transgenic mice were completely negative for CTL activities against all HCV structural proteins, expressed either separately or together. The results shown in Fig. 4A were typical of the larger group of the animals summarized in Fig. 4B. Both wild-type and transgenic mice responded to immunization with core-expressing construct against the core-expressing target cells. On the other hand, wild-type mice immunized with pHCVSt exhibited CTL activities only against E2 and not against E1 or core. Transgenic mice were completely nonresponsive against any of the structural proteins after immunization with the pHCVSt construct. Control animals immunized with vector DNA showed no evidence of CTL activities against any of the HCV target cells.

View larger version (23K): [in this window] [in a new window]   FIG. 4.   CTL responses in wild-type and HCV transgenic mice immunized (Imm.) with HCV plasmids. Immunized mice were sacrificed 2 weeks after immunization, and splenocytes were harvested for CTL assay as described in Materials and Methods. (A) Representative results from each immunization are shown at various effector-to-target (E:T) ratios. (B) Results from all the immunized mice at an effector-to-target ratio of 30 are shown. The results are expressed as % specific lysis, and greater than 10% specific lysis is considered positive (dotted line). For core- and St-immunized wild-type mice: *, P = 0.001 for comparison of mean values and P = 0.005 for comparison of percentage of positive responses of the core CTL activities. §, P = 0.005 when the mean values of core CTL activities are compared for core- and St-immunized transgenic mice. For results of pHCVSt immunization of wild-type and transgenic mice: , P < 0.0001 when the mean values are compared and P = 0.002 when the percentages of positive response of the E2 CTL activities are compared. C/E1/E2, core/E1/E2.

Histological evaluation and adoptive transfer experiment. All immunized mice were evaluated for liver histology and serum ALT. None of the mice, including those with positive CTL responses, demonstrated any evidence of hepatocellular injury or inflammatory infiltrate. This finding suggests that, despite the induction of HCV-specific peripheral cellular immunity, the liver, which is the site of HCV protein expression, is not being targeted by the induced immune response. To further study this immune "ignorance" phenomenon, an adoptive immune transfer experiment was performed. Wild-type FVB/n mice were immunized with the pHCVSt plasmid, and splenocytes were harvested after immunization and were repeatedly stimulated with syngeneic cells expressing core/E1/E2. The stimulated splenocytes demonstrated strong cytolytic activities with 50% specific lysis of core/E1/E2-expressing target cells at an E/T ratio of 30. Stimulated splenocytes (n = 10⁷) were then injected intravenously into transgenic mice, and serum and liver were harvested 1 week later for ALT and histological analysis, respectively. Despite the presence of HCV-specific CTL activities in the adoptively transferred splenocytes, no evidence of hepatitis was seen. We have also labeled the splenocytes with bromodeoxyuridine (BUdR) and stained the liver after adoptive transfer with anti-BUdR antibodies (not shown). We did not observe any accumulation of BUdR-positive cells in the liver, suggesting the absence of targeting of the virus-specific T cells to the liver.

Hierarchy of cellular immune responses. Wild-type FVB/n mice were immunized with recombinant vaccinia virus expressing core/E1/E2 (vvHCV.S). Similar to the results of the DNA immunization experiment, the only significant CTL activities were detected against the E2- and core/E1/E2-expressing target cells, while little activity was detected against core-expressing cells and none against E1-expressing cells (not shown). This and the above findings suggest the induction of a hierarchy of CTL responses against the HCV structural proteins (E2 > core) when they are expressed as a polyprotein in either DNA or vaccinia immunization. To demonstrate that this hierarchical response is not restricted to the FVB/n strain, we immunized the BALB/c mice with the same constructs and assayed their CTL activities against the HCV structural proteins (Fig. 5). The results showed a similar pattern of CTL responses: the pHCVcore plasmid induced strong core-specific activities and the pHCVSt plasmid elicited a predominantly E2-specific CTL response with a weak or no core response. Finally, to test whether this hierarchy of cellular immune response occurs in the humanized HLA-A2.1 background, AAD transgenic mice were immunized with the pHCVSt plasmid, and the CD4 and CD8 responses were measured with ELISpot, a more quantitative assay than the proliferative T-cell and CTL assays. The responses to E1/E2 protein (102, 60, and 81 spots/million cells in triplicate) were much higher than those to core protein (0, 0, and 3 spots/million cells in triplicate; P = 0.0026). When HLA-A2.1-restricted CTL epitopes were used, the CD8 responses were also higher using the E2 peptide (7, 20, and 8 spots/million cells) than the core peptide (0, 3, and 0 spots/million cells; P = 0.07). Similar to results for the HCV transgenic mice in the FVB/n background, pHCVSt could not induce any core- or E2-specific CD4 and CD8 responses in the AAD/HCV double transgenic mice (0 to 1 spot for either the protein or peptide ELISpot assay).

View larger version (40K): [in this window] [in a new window]   FIG. 5.   CTL activities in various strains of mice immunized (Imm.) with HCV plasmids. BALB/c mice were immunized with pHCVcore or pHCVSt three times at 3-week intervals (n = 3 for each BALB/c group). Two weeks after the last immunization, splenocytes were harvested and assayed for CTL activities against HCV structural proteins at an effector-to-target ratio of 30 for BALB/c mice. The results are expressed as % specific lysis, and >10% specific lysis is considered positive (dotted line). *, P = 0.02 when the mean values of the core CTL activities for core-and St-immunized mice are compared.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Our study showed that genetic immunization of FVB/n mice could elicit both humoral and cellular immune responses, albeit the magnitude of the response was not strong (results summarized in Table 1). Our results are consistent with other publications on HCV DNA immunization in various strains of mice (11, 12, 14, 22, 30). However, we observed a hierarchy of cellular immune response, particularly that of CTL, against the HCV structural proteins (E2 > core >E1) when the proteins are expressed as a polyprotein in the immunizing plasmid. This hierarchy is probably not a result of difference in the expression level of the core protein between the constructs expressing either the core alone or core/E1/E2 polyprotein, because comparable core expression is observed between the two constructs in tissue culture transfection study (not shown). However we cannot completely eliminate a possible expression difference in vivo. Furthermore, E1 expressed either alone (not shown) or together with the other structural proteins appears to be a poor target for CTL induction.

Previous genetic immunization studies using HCV structural genes have not specifically analyzed the difference in cellular immune response among the individual HCV structural proteins. In one study, Saito et al. immunized BALB/c mice with six different constructs expressing different structural genes and studied HCV-specific antibody and cellular responses (30). They noted that only constructs containing the core (either alone or together with E1/E2) elicited CTL response against target cells expressing HCV structural proteins. Since the construct expressing E1 and E2 did not induce any CTL response, the authors concluded that the dominant response was likely against the core. However, CTL responses against the individual core and envelope proteins were not tested specifically in the core/E1/E2-immunized mice. In addition, the E1/E2 construct used in their study did not contain the C-terminal hydrophobic region of E2 (aa 706 to 809), which was included in our construct and might contain important CTL epitopes. Using a full-length E1/E2 construct for DNA immunization in FVB/n mice, we observed humoral and cellular immune responses against the envelope proteins similar to those for the pHCVSt construct (not shown), suggesting that the coexpression of core does not alter the immunogenicity of the envelope proteins. This hierarchy of cellular immune response against the HCV structural proteins is not mouse-strain-specific, because both BALB/C and FVB/n strains as well as AAD (HLA-A2.1 in the background of C57BL/6) mice responded similarly. Furthermore, analysis of CD8 response using individual CTL epitopes in the AAD background suggests that this difference is not a result of possibly more CD8 epitopes in the larger E2 protein. Hierarchy or immunodominance of class I-restricted T-cell responses to virus has been described (39), and the phenomenon is probably multifactorial (6). The explanation for this hierarchy in HCV structural proteins is not apparent at present. It has been shown that immunodominant determinants can suppress subdominant determinants in vivo (15). It is also possible that the core protein, when expressed together with the envelope proteins, may have a different fate in the antigen presentation pathway. These possibilities await further experimentation.

Does this observation have any relevance to cell-mediated immunity during HCV infection in humans? While it is clear that CTL response can be detected against the core in infected persons with either acute or chronic infection ( (2, 19, 26, 29), very few studies focused on the induction of cellular response against the envelope proteins, probably because of the highly variable sequences in this region. In one study, a strong CD8⁺ CTL response against the hypervariable region can be detected early during acute HCV infection and sequence variations in this region can lead to antagonism (33). Similar antagonism has also been reported for CD4⁺ T-helper epitopes (13). These authors postulated that this might be a mechanism for viral escape and persistence. Recent publications have shown that active cellular immune responses against the E2 are frequently present in persons infected with HCV (8, 38). A similar observation can be inferred from a study of cell-mediated immunity during acute infection of chimpanzees, in which several CTL epitopes were identified in the E2 but not in the core region and the strength of CTL responses against these epitopes was associated with viral clearance (7). The relevance to natural infection notwithstanding, our observation on the hierarchy of immune response could have implications in designing constructs for DNA vaccination trials in primates or humans. For example, if induction of core-specific, cell-mediated immunity is important, a core-only construct should be designed. This may also be relevant to other HCV polyprotein constructs, which may induce a set of immune responses qualitatively different from those induced by constructs expressing individual proteins.

The transgenic mouse model has been used to study immunological tolerance and immune response against viral antigens. In one hepatitis B virus transgenic model, DNA immunization with the hepatitis B virus expression construct broke immunological tolerance and induced hepatitis (24). In another model, immunological tolerance at the CTL level could only be broken by dendritic cell immunization, but no hepatitis was observed (31, 37). Hepatitis could be induced only by adoptive transfer of HBV-specific CD8⁺ CTL clones into these mice (1). In our transgenic mice expressing HCV structural proteins, antibodies against either the core or envelope proteins could be induced by DNA immunization. The transgenic mice did not appear to be tolerant to the core protein at the T-cell level. On the other hand, either CD4 or CD8 cellular immunity against the envelope proteins appeared to be tolerant and could not be broken by DNA immunization in the transgenic mice. It is particularly intriguing that the core protein in the context of polyprotein expression is both a poor immunogen as well as a weak toleragen. This observation is consistent with the present concepts of antigen presentation and tolerance induction (34).

Despite the induction of core-specific antibodies and cellular immune response, the transgenic mice exhibited no inflammatory cell infiltrate or pathology in the liver. Adoptive transfer of splenocytes with strong CTL activities also did not cause any hepatocellular injury. It is possible that a low-level expression of HCV antigens in our transgenic mice was insufficient to induce homing of these HCV-specific T cells into the liver. A similar observation was reported in a recent report in which liver-specific expression of a CTL epitope-containing viral antigen (lymphocytic choriomeningitis virus) was not sufficient to result in hepatitis despite the presence of peripheral T cells that were specific for the epitope (35). Only by lymphocytic choriomeningitis virus superinfection and adoptive transfer of a large number of virus-specific T lymphocytes did hepatocellular injury occur. Therefore, a much more vigorous immune response is necessary for targeting to the liver. This "immune avoidance" behavior of viral antigens expressed in the liver may partially explain the persistence of HCV infection. To completely eliminate HCV infection, it may be necessary to activate and target endogenous HCV-specific immune response specifically to the liver. Such a concept may underpin the success of future immunotherapy for hepatitis C.