IFN-γ, a pleiotropic immune regulator, is implicated in both tumor immune surveillance and selection of tumor variants resistant to immune control, i.e., immunoediting. In uveal melanoma patients, elevated serum levels of IFN-γ correlate with the spread of metastasis and represent a negative prognostic marker. Treatment with IFN-γ boosted the MHC class I presentation machinery in uveal melanoma cells but suppressed their MHC class I-restricted CTL lysis. Tumor cells exposed to IFN-γ efficiently activated specific CTL but were less susceptible to permeabilization by perforin and exhibited a decreased capacity to bind and incorporate granzyme B. These results define a novel mechanism of resistance to granule-mediated CTL lysis in human tumors. Furthermore, the data suggest that immunoediting is not limited to genetic or epigenetic changes resulting in stable cellular phenotypes but also involves an inducible modulation of tumor cells in response to a microenvironment associated with immune activation.
Interferon-γ is a pleiotropic cytokine implicated in the immune control of infections and tumors (1, 2). The importance of IFN-γ for host resistance to bacterial pathogens is unequivocally demonstrated by an increased susceptibility to mycobacterial infections in patients with genetic alterations of the IFN-γ or IL-12 receptor signaling pathways (3). In contrast, the importance of IFN-γ, in particular, and the immune system, in general, in the control of cancer is still contested (4). Nevertheless, a growing body of evidence supports the existence of cancer immunosurveillance and a major contribution of IFN-γ in this process (5, 6, 7). Thus, several studies have shown that mice with different genetic immunodeficiencies develop tumors with increased frequencies as compared with immunocompetent control animals. Likewise, significantly more tumors of apparently nonviral etiology are observed in immunodeficient humans than in the appropriate control groups (reviewed in Ref. 8). Mice lacking T and B lymphocytes or deficient for IFN-γ develop similar histological types of malignancies with comparable frequencies (9). This indicates that IFN-γ plays a critical role in tumor immunosurveillance probably due to the enhancing effects of the lymphokine on the MHC class I-processing and presentation pathway. However, direct or indirect effects of IFN-γ exerted on tumors are not limited to the inhibition of tumor development or growth, because tumors arising in animals lacking lymphocytes or deficient for their critical functions are more sensitive to immune control than tumors arising in immunocompetent hosts (9). These data suggest that, in addition to immunosurveillance, IFN-γ also promotes the process of “cancer immunoediting,” which results in the resistance of tumor cells to immunological control (reviewed in Ref. 8). The molecular basis of cancer immunoediting is poorly defined. Defects in MHC class I expression, as well as down-regulation or loss of expression of tumor-specific Ags, frequently observed in many types of tumors (10) suggest that immunoediting also takes place in the course of tumor progression in humans. It remains unclear, however, to what extent epigenetic or functional changes in response to the microenvironment may contribute to the development of less immunosusceptible tumor phenotypes. To study the mechanisms of IFN-γ-mediated immunoediting in human tumors, we turned to primary uveal melanoma (UM),3 the most frequent malignancy of the eye in adults (reviewed in Ref. 11). Leukocytes of late-stage UM patients secrete increased amounts of IFN-γ. Elevated IFN-γ levels in the serum of UM patients correlate with the spread of metastases and were suggested as a negative prognostic marker (12). In agreement with this observation, expression of IFN-γ was frequently detected in UM tumor samples (13) and high levels of MHC class I and II expression in the primary UM lesions correlated with significantly decreased patient survival (14, 15). These data indicate that UM cells may be resistant to IFN-γ and/or subjected to changes promoting progression of the disease in response to this lymphokine.
In this study, we demonstrate that the MHC class I- processing pathway undergoes coordinated changes in UM cells in response to IFN-γ and TNF-α, another proinflammatory cytokine known as a potent systemic immune modulator. However, IFN-γ-treated tumors were less efficiently lysed by CD8+ MHC class I-restricted CTL as compared with untreated parental cells or TNF-α-treated counterparts. This decreased sensitivity of target cells to CTL was due to IFN-γ-induced resistance to granule-mediated cytolysis.
To our knowledge, these results represent the first demonstration of inhibitory effects of an immunostimulatory cytokine on the MHC class I-restricted elimination of tumor cells due to an impaired sensitivity to granule-mediated cell killing. The data also suggest that immunoediting of the tumor phenotype may involve inducible changes in tumor cells in response to an inflammatory microenvironment.
Materials and Methods
Tumor cell lines
The UM cell lines OCM-1, -3, and -8 were provided by J. Kan-Mitchell (University of California, San Diego, CA). The UM cell line Mel290 was a generous gift from B. R. Ksander (Schepens Eye Institute, Harvard Medical School, Boston, MA). The melanoma cell line DFW was derived from a metastatic lesion from a patient at Radiumhemmet, Karolinska Hospital (Stockholm, Sweden). C1R/A11 is an HLA-A11 transfectant of the C1R lymphoblastoid cell line (LCL) carrying B-type EBV (16). The LCL JAC-B2 was obtained by transformation of B cells with the B95.8 EBV strain. T2 is an LCL/T cell hybrid defective in TAP and proteasomal subunits (17). The HLA-A, -B, and -C-deficient lymphoblastoid cell line 721.221 and a peptide derived from the signal sequence of HLA-G (VMAPRTLFL), capable of binding to and stabilizing HLA-E, were provided by C. Teixeira de Matos (Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Sweden). Cell lines were maintained in IMDM supplemented with 10% heat-inactivated FCS (Invitrogen Life Technologies), 100 IU/ml penicillin, and 100 μg/ml streptomycin (complete medium).
The generation and characterization of the CD8+ HLA-A11-restricted CTL clones BK289 and CAR13, specific for the EBV nuclear Ag-4-derived peptide IVTDFSVIK (IVT), were previously described (18). The CTL clone 435 recognizing HLA-A2 was produced by stimulating PBMC from a HLA-A2-negative donor with irradiated HLA-A2-positive PBMC.
Abs and chemicals
The Abs specific to LMP2, LMP7, MECL-1, and the α2/HC3 subunit were purchased from Affinity Research Products. The TAP1- and TAP2-specific Abs were provided by Dr. J. Trowsdale (University of Cambridge, Cambridge, U.K.). Human rTNF-α was obtained from Cetus and IFN-γ from Boehringer Ingelheim International. The FITC-labeled Ab specific for perforin, IgG1, and IgG2a isotype controls were purchased from BD Pharmingen. The HLA-ABC-specific Ab (clone W6/32) conjugated with R-PE and the HRP-conjugated sheep anti-goat Ab were obtained from Dakopatts. The hybridomas producing the HLA-A11-specific Ab (clone HB-164) and HLA-A2-specific Ab (clone HB-54) were obtained from the American Type Culture Collection. The R-PE-conjugated mouse Abs of IgG2a isotype and R-PE-conjugated rabbit anti-mouse F(ab′)2 were purchased from Dakopatts. The anti-CD94 Ab (clone HP-3D9) was provided by Dr. M. Lopez-Botet (University Pompeu Fabra, Barcelona, Spain). The Fas ligand (FasL)-specific mouse mAb NOK-2 was purchased from BD Pharmingen. The recombinant soluble TRAIL receptor 2 (R2) was obtained from Alexis. Human purified granzyme B (grB) was either prepared as described previously (19) or obtained from Alexis. The R-PE-conjugated mouse anti-grB mAb was purchased from Serotec. Human perforin was either purified as described earlier (20) or purchased from Kamiya Biomedical. Propidium iodide (PI) and actin-specific mouse mAb were purchased from Sigma-Aldrich. Goat Abs specific to Bid were purchased from R&D Systems Europe. Tetramethylrhodamine ethyl ester perchlorate (TMRE) was purchased from Molecular Probes and the annexin V binding kit from BD Pharmingen.
Western blot analysis
Protein electrophoresis was performed using a Multiphor II Electrophoresis System and ExelGel SDS homogeneous precast gels (Amersham Pharmacia Biotech).
Expression of molecules involved in MHC class I-restricted Ag presentation
Lysates of cells cultured in complete medium alone (herein referred to as control) or in the presence of TNF-α (30 ng/ml) or IFN-γ (500 IU/ml) for 48 h at 37°C corresponding to 105 cells were separated by SDS-PAGE. Immunoblots were detected by ECL (Amersham Pharmacia Biotech) according to the manufacturer’s protocol.
Analysis of Bid cleavage
Tumor cells were pulsed with IVT peptide for 1 h at 37°C, mixed with CTL at a 1:1 E:T ratio, and pelleted down at 100 × g for 5 min. At the indicated time points, cells were collected and lysed in sample buffer. Western blots were performed as described above. Blots were visualized using a Fuji Film Image reader LAS-1000 (Fuji Film Sverige). The intensity of Bid bands was monitored using Fuji Film ImageGauge 4.0 software.
MHC class I expression
Total HLA class I expression on tumor cells was measured after a 48-h treatment with TNF-α or IFN-γ, using a R-PE-labeled HLA-ABC-specific Ab (clone W6/32) or an IgG2a isotype control, and analyzed on a FACSCalibur flow cytometer (BD Biosciences). Expression of the HLA-A11 or HLA-A2 alleles was assessed using HLA-A11 (clone HB-164)- or HLA-A2-specific Abs (clone HB-54). Cells incubated with mouse serum were used as control. FITC-conjugated rabbit anti-mouse F(ab′)2 were used as secondary Abs.
Peptide binding and cytotoxicity assays
The HLA-A11-restricted CTL peptide epitope, IVTDFSVIK, was synthesized by Alta Biosciences and biotinylated as previously described (21). Binding of the peptide to HLA-A11-positive cells was visualized by R-PE-conjugated streptavidin (BD Biosciences) and analyzed by FACS. Standard 51Cr release assays were performed as previously described (22). For blocking of FasL or TRAIL, the effector cells were preincubated with 10 μg/ml Nok-2 (anti-FasL), soluble TRAIL R2 (10 μg/ml) or isotype control for 45 min before addition to the assay. In the CD94-blocking experiments, effectors were preincubated with 10 μg/ml HP-3D9 Ab or IgG1 isotype control for 30 min on ice before cytotoxicity assays. To assess the blocking efficacy of the HP-3D9 Ab, the 721.221 cell line was prepulsed with the leader sequence peptide from HLA-G at 26°C for 16 h and subsequently incubated with CD94/NKG2a-expressing polyclonal NK cultures in a standard 4-h 51Cr release assay.
The IVT-specific CTL clone BK289 was coincubated with IVT-pulsed target cells as described above. After 1 or 5 h of coincubation, cells were permeabilized using Ortho Permeafix (Ortho Diagnostic Systems), stained with a FITC-conjugated anti-perforin Ab, and analyzed by FACS. Perforin release in activated T cells was related to the percentage of perforin-positive T cells in a control sample (targets not pulsed with IVT peptide).
Tumor cell permeabilization by perforin
OCM1 and OCM8 cells, either treated with 500 IU/ml IFN-γ for 48 h or left untreated, were detached by versene treatment, washed twice in RPMI 1640 containing 1% fatty acid-free BSA, and resuspended at a concentration of 2 × 106 cells/ml in the same medium. To an Eppendorf tube were added 100 μl of cell suspension and 50 μl of “high-calcium” HEPES solution (20 mM HEPES, 150 mM NaCl, 5 mM CaCl2, and 10 μg/ml BSA). Perforin was diluted in Ca2+-free HEPES (20 mM HEPES, 150 mM NaCl, and 10 μg/ml BSA) immediately before the experiment at a concentration that lyses ∼50% of untreated UM cells (batch dependent); 50 μl of this solution was added to the cell suspension. The suspension was mixed by careful pipetting and incubated at 37°C for 15 min. The samples were transferred to FACS tubes and 10 μg/ml PI (final concentration) was added before analysis by flow cytometry.
grB binding and internalization assays
All steps of the grB surface-binding procedure were performed at +4°C to prevent grB uptake. UM cells, either untreated or treated with the indicated cytokine, were washed and incubated for 10 min in PBS containing 0.1% BSA in a V-bottom polypropylene 96-well plate (2 × 105 tumor cells/well). Human purified grB was added at a final concentration of 50 ng/ml. Equal volumes of PBS/0.1% BSA were added as control. Following 45 min of incubation on ice, binding of grB was assessed by flow cytometry. grB internalization was assessed by incubation of UM cells with grB at 37°C for 60 or 90 min. Surface-associated grB was removed by treatment with low pH buffer as described previously (23). Cells were then permeabilized and stained with grB-specific Ab as described above.
Assessment of apoptotic cell death
UM cells, treated with the indicated cytokine as previously described, were pulsed with 10−7 M IVT peptide for 1 h and subsequently incubated with CTL at a 2:1 E:T ratio for 2 h at 37°C. Assessment of apoptosis was performed using annexin V and TMRE as described previously (24).
Statistical analysis was performed using the GraphPad PRISM 4 software. A two-tailed paired t test was used to analyze differences between groups of IFN-treated and untreated cells. Differences between the cytokine treated and untreated control groups were assessed using a one-way ANOVA-repeated measures test followed by Dunnett’s multiple comparison posttest (see Fig. 9B).
Proinflammatory cytokines up-regulate components of the MHC class I-processing and presentation pathway in UM cells
Previous studies have shown that IFN-γ and TNF-α up-regulate the expression of MHC class I molecules, the components of the TAP heterodimer and induce the immunoproteasome, leading to more efficient processing and presentation of MHC class I-associated peptides at the cell surface in a variety of cells including different types of tumors (23, 25). Both lymphokines enhanced the expression of TAP1 and TAP2 as well as the immunoproteasomal subunits LMP2, LMP7, and MECL-1 in all UM cell lines analyzed in this study (Fig. 1 and data not shown). This was paralleled by an increased expression of assembled MHC class I molecules at the cell surface of lymphokine-treated cells (Fig. 2 A and data not shown). The effects of IFN-γ on the MHC class I-processing machinery were always more prominent than those induced by TNF-α. Increasing the concentration of TNF-α or prolonging the treatment did not compensate for this difference (data not shown).
Treatment with IFN-γ decreases the lysis of UM cells by MHC class I-restricted CD8+ CTL
We next investigated whether the modulation of the Ag presentation machinery caused by the proinflammatory cytokines is translated into an enhanced recognition of UM cells by MHC class I-restricted CD8+ CTL. Two HLA-A2-positive tumor cell lines, UM Mel290 and cutaneous melanoma DFW, were used as targets for an allogeneic HLA-A2-specific CD8+ CTL clone in 4-h 51Cr release assays. Recognition of MHC-peptide complexes by allogeneic T cells is usually peptide independent or can be induced by a set of distinct peptides (reviewed in Ref. 26). Therefore, changes in the efficiency of allorecognition are more likely to reflect general changes in the expression of a given MHC allele at the cell surface rather than changes in the presentation of any particular endogenously processed peptide. Despite the 5-fold up-regulation of HLA-A2 at the cell surface (Fig. 2,A, left panel), IFN-γ-treated Mel290 cells were less efficiently killed by HLA-A2-specific allogeneic CTL as compared with untreated or TNF-α-treated targets (Fig. 2,A, right panel), although the latter up-regulated HLA-A2 by only 40% on average. Notably, the same effectors recognized both IFN-γ- and TNF-α-treated cutaneous melanoma cells more efficiently than untreated control cells (Fig. 2,B, right panel). Results similar to those presented in Fig. 2 A were obtained when the HLA-A11-positive OCM1, OCM3, and OCM8 cell lines were used as targets for a HLA-A11-specific allogeneic CTL clone (data not shown).
Much higher levels of MHC class I expression on IFN-γ-treated cells may mask the inhibitory effects of the lymphokine if the killing is assessed using allospecific effectors. Conventional peptide-pulsing experiments using peptide-specific CTL may suffer from the same drawback due to an increased expression of peptide-receptive MHC class I molecules on the surface of IFN-γ-treated cells. Moreover, endogenous presentation of certain peptide epitopes may be suppressed by IFN-γ due to degradation by the immunoproteasome (27, 28, 29). Therefore, we developed a system to analyze how IFN-γ affects CTL recognition of UM cells when the levels of Ag presentation are comparable on control and IFN-γ-treated cells. We synthesized a polypeptide in which the sequence of the EBV-derived HLA-A11-restricted CTL epitope, referred to as IVT, was attached to biotin through a flexible polypeptide linker as previously described (21). Binding of this peptide to cells was dependent on HLA-A11 expression and could be revealed using fluorescently labeled streptavidin. As shown in Fig. 2,C, IFN-γ treatment significantly increased peptide binding by OCM1 and OCM8 cells and to achieve similar levels of peptide presentation, IFN-γ-treated cells had to be pulsed with at least 27 times lower amounts of the biotinylated peptide, as compared with untreated control cells. Cytotoxicity assays performed with target cells pulsed with accordingly adjusted amounts of nonbiotinylated IVT nanomer revealed that IFN-γ treatment inhibits specific CTL lysis of OCM1 and OCM8 cells by 70–80% when lymphokine-treated and control cells present comparable levels of the specific Ag (Fig. 2 D).
IFNγ-treated UM cells do not exert negative effects on activation or viability of specific CTL
Next, we performed a series of experiments to examine whether the inhibitory effect of IFN-γ might be due to damage or inhibition of CTL by IFN-γ-treated targets. We first assessed the ability of UM cells to hamper T cell activation in a bystander manner using cold target inhibition assays. Untreated or cytokine-treated UM cells did not differ in their capacity to inhibit specific CTL lysis when added at increasing numbers into standard cytotoxicity assays as third-party cells (Fig. 3). NK cell-mediated killing of UM lines has been previously shown to be inhibited by factors secreted from tumor cells (30). However, conditioned culture supernatants of UM lines did not affect the viability or cytotoxic activity of the effectors used in our study (data not shown), consistent with the results of the cold target inhibition assays.
To investigate whether UM cells inhibit T cell activation while serving as targets for MHC class I-restricted recognition, we measured a number of parameters of T cell activation. A 4- to 5-fold increase in the level of CD69 expression was detected upon stimulation of T cells with several UM lines and this effect was not modulated by cytokine treatment of the tumor cells. Triggering with control or lymphokine-treated UM cells induced comparable levels of TCR down-regulation in peptide-specific CTL. Likewise, specific CTL produced comparable levels of TNF-α and IFN-γ upon activation with control or lymphokine-pretreated UM cells (data not shown).
Expression of mRNA encoding HLA-E, which serves as a ligand for the heterodimeric inhibitory receptor CD94/NKG2a (reviewed in Ref. 31), is enhanced by IFN-γ in UM cell lines (32). We tested the effect of the 3D9 Ab, capable of blocking CD94 heterodimers, on the cytotoxic activity of BK289 CTL, which expressed NKG2a, against the UM cell lines OCM1, OCM3, and OCM8 prepulsed with the IVT peptide. The Ab failed to increase CTL-mediated lysis of peptide-pulsed IFN-γ-treated OCM1, OCM3, or OCM8 cells (Fig. 4 and data not shown); however, killing of HLA-E-expressing control targets was increased after preincubation of NK cells with the CD94-blocking Ab (Fig. 4 B).
High expression of MHC class I complexes at the cell surface of targets induced by IFN-γ may trigger inhibitory receptors expressed by effector cells, thereby interfering with T cell activation and function (reviewed in Ref. 33). Thus, we determined concentrations of TNF-α and IFN-γ, which induced comparable levels of surface MHC class I complexes on UM cells. Under such conditions, only TNF-α but not IFN-γ treatment resulted in an enhanced recognition of peptide-pulsed targets by specific CTL (Fig. 5).
Collectively, we found no evidence for inhibitory effects of IFN-γ-treated OCM cells on T cell viability or activation.
UM cells are killed via the perforin/granzyme pathway and IFN-γ does not affect their capacity to stimulate the release of cytolytic granules by specific T cells
Effector mechanisms utilized by CTL to eliminate their targets include expression of FasL and TRAIL, secretion of lymphokines, and release of cytolytic granules (reviewed in Ref. 34). To identify the effector components of T cells causing death of UM cells in our experimental system, we performed FasL- and TRAIL-blocking experiments using the NOK-2 Ab (35) and soluble TRAIL R2 in a standard cytotoxicity assay. Incubation of FasL- and TRAIL-sensitive Jurkat cells with BK289 CTL induced detectable 51Cr release, which was substantially reduced by FasL-blocking Abs and slightly decreased by the addition of soluble TRAIL R2 (data not shown). The same reagents had no effect on CTL lysis of UM cells, untreated or treated with IFN-γ (Fig. 6,A). In contrast, specific CTL lysis of peptide-pulsed UM cells was strongly decreased by concanamycin A, an inhibitor of granule-mediated cytolysis, regardless of whether or not they were treated with IFN-γ (Fig. 6 A).
We next measured the content of perforin in the effector cells before and after triggering with lymphokine-treated or untreated UM cells. Activation of CTL by any of the tested targets resulted in a comparable decrease in the proportion of perforin-positive T cells 5 h after triggering (Fig. 6,B). Accordingly, no significant difference in the time kinetics of granule release was observed in CTL exposed to untreated or cytokine-treated tumor cells, since the perforin content decreased to comparable levels in all samples already after 1 h of stimulation (Fig. 6 C).
IFNγ strongly decreases the sensitivity of UM cells to perforin-mediated permeabilization
The data described above suggested that the inhibitory effect of IFN-γ on CTL lysis may result from an increased resistance of UM cells to granule-mediated cytolysis. Therefore, we set out to analyze how IFN-γ affects the interaction of UM cells with perforin and grB, the major granule components required for killing of target cells by CTL. Incubation of OCM cells with perforin at 37°C induced UM cell membrane permeabilization as was revealed by a 5- to 6-fold increase in the percentage of PI-positive cells. However, treatment with IFN-γ rendered OCM1 and OCM8 cells virtually resistant to this effect of perforin (Fig. 7).
IFN-γ hampers the association of grB with the cell membrane of UM cells
According to the current view on granule-mediated apoptosis, grB binds to the cell membrane (reviewed in Ref. 36) and, following internalization and activation facilitated by perforin, initiates the apoptotic cell death program in the target cell through activation of caspases (37). We hypothesized that the association of grB with the cell membrane and/or its internalization may also be affected by IFN-γ treatment in UM cells. Both OCM1 and OCM8 cells bound grB less efficiently after pretreatment with IFN-γ (Fig. 8 A) while TNF-α-treated cells bound grB as efficiently as untreated targets (data not shown).
To measure grB internalization, cells were incubated with the protein at 37°C to allow uptake of the molecule, treated with a low pH buffer to remove the surface-associated pool of the protein, permeabilized, and stained with grB-specific Ab. As shown in Fig. 8, B and C, treatment with IFN-γ significantly decreased the capacity of OCM1 and OCM8 cells to incorporate purified grB.
Bid is a substrate for direct cleavage by grB (19, 38). To test whether or not the processing of Bid in UM cells is affected by IFN-γ treatment, we measured the time kinetics of Bid cleavage induced by BK289 CTL in IVT-pulsed OCM8 cells (Fig. 9,A). Approximately 75% of Bid was processed in TNF-α-treated cells in contrast to only 25% of the protein in IFN-γ-treated OCM8 cells after 30 min of incubation with CTL (Fig. 9,B). Importantly, no detectable cleavage of Bid was observed in UM cells in the absence of the IVT peptide (data not shown). This argues against a contribution of Fas ligation by FasL-expressing BK289 cells in the processing of Bid (38). To test whether reduced Bid cleavage seen in IFN-γ-treated UM cells translated into decreased apoptosis, peptide-pulsed UM cells were analyzed for loss of mitochondrial membrane integrity following coincubation with CTL. As shown in Fig. 9 C, tumor cell apoptosis correlated well with the extent of Bid cleavage, as only ∼40% of IFN-γ-treated UM cells showed annexin V positivity and loss of mitochondrial membrane potential after 2 h of exposure to CTL, as compared with almost 70% of control cells and 80% of TNF-α-treated cells.
In this study, we show that IFN-γ strongly up-regulates MHC class I expression on the surface of UM cells but, paradoxically, decreases their sensitivity to CTL lysis. A moderate MHC class I up-regulation induced by TNF-α was associated with enhanced CTL recognition of UM cells, whereas IFN-γ-treated cells expressing a severalfold higher level of MHC class I were generally more resistant to CTL lysis, as compared with untreated control cells. The recognition of UM cells in our experimental settings was mediated by either HLA-specific allogeneic effectors or by peptide-specific CTL directed against targets with exogenously loaded peptide. The inhibition of killing was profound (70–80%) when IFN-γ-treated and untreated OCM cells presented similar amounts of the specific Ag.
Different scenarios could account for the phenomenon described above. IFN-γ could induce soluble and membrane-bound factors in malignant cells that cause damage, death, or defective activation of CTL or, alternatively, render tumors resistant to CTL-mediated lysis. Although, UM were reported to inhibit T cell proliferation in vitro via cell-to-cell contact (39), our results demonstrated that IFN-γ-treated UM cells do not inhibit specific cytolytic activity of CTL against a third-party target, either through cell-to-cell contact or release of soluble factors (Fig. 3 and data not shown). We also showed that enhanced triggering of NK inhibitory receptors by classical MHC class I molecules or by HLA-E, up-regulated in response to IFN-γ in UM cells, cannot account for the inhibitory effect of the lymphokine, although at least one inhibitory receptor, CD94/NKG2a, was expressed by the specific CTL used in this study. In agreement, IFN-γ treatment of UM cells did not affect their capacity to trigger T cell activation as measured by lymphokine release, up-regulation of CD69, TCR down-regulation, or degranulation of specific CTL (data not shown and Fig. 6). Moreover, the decreased sensitivity of IFN-γ-treated UM cells to CTL lysis was not accounted for by an overall reduction in their susceptibility to apoptosis, since tumor cell death induced by either FasL-expressing T cells or by the agonistic CH-11 Ab was not inhibited, but rather enhanced, upon IFN-γ-treatment (K. Hallermalm, unpublished results).
Cytotoxicity assays performed in the presence of conconamycin A or blocking of FasL and TRAIL clearly showed that killing of OCM cells in our experimental system was mediated primarily by cytotoxic granule release. IFN-γ strongly decreased the sensitivity of UM cells to permeabilization by human purified perforin. The mechanisms accounting for this effect of the lymphokine remain to be investigated. We also showed that IFN-γ decreases the capacity of grB to bind to the membrane of OCM cells. grB is secreted by CTL in a macromolecular complex bound to serglycin (40). The calcium-independent mannose 6-phosphate receptor (CI-MPR) was shown to act as a receptor for grB (19, 41, 42), although its importance for granzyme-mediated apoptosis has been questioned (43, 44). Other molecules suggested to serve as grB receptors include heat shock protein 70 (45) and CD44 (40, 46). We speculated that the efficiency of grB interaction with the membrane of IFN-γ-treated UM cells could be decreased. Indeed, IFN-γ-treated UM cells bound lower amounts of grB than their untreated counterparts (Fig. 8). This correlated with decreased cleavage of Bid and partial protection from CTL-mediated apoptosis in IFN-γ-treated cells (Fig. 9). Although the relative importance of different granzyme receptors on different cell types is still unknown, we assessed the expression of CI-MPR and CD44 on cytokine-treated UM cells. Interestingly, the expression of both CI-MPR and CD44 was reduced on UM cells following IFN-γ treatment (K. Hallermalm, unpublished data), suggesting that grB receptor down-regulation may contribute to the phenomenon described in our study.
IFN-γ, a high-affinity binder to the heparan sulfate portion of perlecan, causes its transcriptional repression in various cells with distinct histogenetic backgrounds (47). In agreement with these data, IFN-γ reduced the level of heparan sulfate glycosaminoglycans in some UM cell lines (data not shown), which led us to speculate that the observed decreased binding of grB may, in part, be attributable to the down-regulation of surface heparan sulfate glycosaminoglycans that in turn deprives the granzyme of its high-affinity binding sites (48). Perforin binding, on the other hand, appears to involve a calcium-dependent interaction with cell surface phospholipids. The latter may explain the reduced susceptibility of IFN-γ-treated UM cells to permeabilization by perforin, since this cytokine alters the composition of plasma membrane phospholipids in UM cells (49). The relative contribution of different modes of grB and perforin interaction with the target cell membrane in a process of CTL-mediated killing of UM cells needs to be explored further.
The inhibitory effect of IFN-γ is likely to operate in vivo due to high levels of this cytokine found both in the peripheral blood (12) and at the tumor site (50) in patients with UM (13). Consistent with the protective role of IFN-γ, elevated serum levels of IFN-γ and high expression of MHC class I and class II Ags in primary UM lesions have been established as markers of poor prognosis (14, 15). The presence of tumor-infiltrating lymphocytes within the intraocular tumor milieu and the expression of immunogeneic tumor Ags do not prevent UM growth and metastatic spread. A number of factors, such as anterior chamber-associated immune deviation, in situ suppression of delayed-type hypersensitivity effector cells, suppression of NK cell activity in oculi, and inactivation of the complement cascade by regulatory proteins expressed on UM cells, were suggested to underlie this phenomenon (reviewed in Ref. 51). In this study, we describe a new mechanism of resistance of UM cells to immune control. Paradoxically, this mechanism is triggered by a proinflammatory cytokine, IFN-γ, which is commonly viewed as a key positive regulator and executor of CTL effector functions. Given the capacity of IFN-γ to suppress angiogenesis, it is tempting to speculate that vasculogenic mimicry (52) may be another outcome of the IFN-γ-UM interaction that allows direct contact of tumor cells with blood components, including effector cells of the immune system. It is possible that, during hematogeneic spread, a population of UM cells pre-exposed to IFN-γ may not only escape NK-mediated recognition due to high MHC class I levels at the surface, but also evade T cell-mediated HLA class I-restricted killing by tumor-specific CTL due to the acquired resistance to perforin/granzyme-mediated lysis. Several studies have demonstrated that tumor cells rendered resistant to CTL lysis in vitro also escape CTL-mediated control in vivo (53, 54). Similar correlations were observed even for tumor cell lines manifesting relatively modest resistance to specific CTL, supporting the potential in vivo relevance of our data (55, 56).
The concept of tumor immunoediting is supported by experimental data obtained in animal models that demonstrated a critical role of IFN-γ in the selection of immunoresistant tumor phenotypes (9). The exact mechanisms of this process remain to be defined. We have recently shown that IFN-γ-treated ovarian carcinoma cells escape CTL killing through up-regulation of HLA-E, which interacts with NK inhibitory receptors expressed on activated CTL (57). Our current study describes a novel, principally different mechanism of inhibition of CTL lysis exerted by IFN-γ via active acquisition of a CTL-resistant phenotype by malignant cells. Collectively, these results suggest that escape from the immunostimulatory effects of IFN-γ is a common phenomenon for human tumors of different cellular origin. Conceptually, the results of this study indicate that selection of tumor cells with new stable phenotypes resulting from genetic or epigenetic changes may not be the only mechanism of immunoediting and that tumor cells may also acquire resistance to immune effector cells through inducible changes that require continuous presence of IFN-γ in the tumor microenvironment. This conclusion is important for designing new approaches to the analysis of immunoediting in different tumor models as well as for developing new methods for improving the efficiency of tumor immunotherapy.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work has been supported by grants from the Swedish Cancer Society, the Cancer Society of Stockholm, and the King Gustav the Vth Jubilee Fund.
Abbreviations used in this paper: UM, uveal melanoma; LCL, lymphoblastoid cell line; grB, granzyme B; FasL, Fas ligand; PI, propidium iodide; TMRE, tetramethylrhodamine ethyl ester perchlorate; R2, receptor 2; CI-MPR, calcium-dependent mannose 6-phosphate receptor.