Stepwise immunologic selection of antigenic variants during tumor growth

Using tumor-specific effector cells as probes, we have studied the immunologic changes that occur in tumor cells during continuous growth in a host. As a model, we used a highly immunogenic ultraviolet light (UV)-induced tumor that is rejected regularly by normal mice but grows progressively when transplanted into UV-irradiated mice. The immunogenic tumor growing continuously in these partially immunocompromised mice gave rise to genetically stable progressor variants that were poorly immunogenic. A sequence of changes in susceptibility to activated macrophages and tumor-specific cytolytic T cells was observed when serial reisolates from the continuously growing tumors were analyzed. First, the tumor cells developed resistance to the cytocidal effects of activated macrophages. This was followed by the loss of one and then a second tumor-specific antigen defined by syngeneic cytolytic T cells. The phenotypes of the developing antigen loss variants and their sequence of appearance were the same in several independent experiments, and the process was apparently determined by a hierarchy of the host's immune response to multiple independent tumor-specific antigens expressed by a single malignant cell. Our ability to generate the predicted variants in vitro before they actually appear in vivo suggests a possible approach to preventing the outgrowth of such immunoselected variants from a tumor.     

Secretion of stress protein grp170 promotes immune-mediated inhibition of murine prostate tumor

It is well established that certain stress proteins or molecular chaperones are highly efficient in cross-presenting tumor-derived antigens, resulting in a potent antitumor immune response. In this study we demonstrate that genetic modification of weakly immunogenic murine prostate tumor cells (TRAMP-C2) by stable transfection with a secretable form of endoplasmic reticulum resident chaperone grp170 significantly enhances its immunogenicity in vivo. Generation of systemic antitumor immunity is indicated by the growth suppression of distant parental tumors, which is associated with increased tumor infiltration, elevated effector functions of CD8⁺ T-cells. Immunization with inactivated grp170-secreting C2 cells augments a CD8⁺ T-cell dependent, tumor-protective effect. Furthermore, infection of C2 tumor cells with a nonreplicating adenoviral vectors encoding secretable grp170 promotes tumor immunogenicity more effectively than plasmid transduction, as shown by the increased production of pro-inflammatory cytokine TNF-α by dendritice cells and enhanced therapeutic efficacy in treating pre-established tumors. Given a repertoire of undefined antigens in prostate tumor, manipulation of cellular compartmentalization of immuno-stimulatory chaperone grp170 to elicit systemic tumor immunity may be used to improve treatment outcomes for prostate cancer when combined with other treatment modalities.     

The influence of major histocompatibility complex class I antigens on tumor growth and metastasis

The work described here demonstrates the importance of major histocompatibility complex class I antigens for the control of tumor growth and metastasis by the host's immune system. In certain murine tumor cells which have lost expression of H-2 class I antigens, a de novo expression of H-2 can be achieved by transfection with syngeneic class I genes. In contrast to the parental cells the transfected tumors do not grow any more in syngeneic mice, or in other cases they do not form metastases. The studies suggest that the de novo expression of the H-2 antigens renders the tumors highly immunogenic and leads to effective recognition of a tumor-associated antigen in conjunction with the transfected H-2 antigen. These conclusions were confirmed in other tumor systems. For example, separation of a heterogeneous tumor into clones expressing high or low amounts of H-2 showed that only the tumor cell with low H-2 grew well in syngeneic mice, whereas the H-2 high tumor clones were rejected. In other studies in vitro induction by IFN-gamma of H-2 antigen on H-2 negative tumors led to reduced tumor growth in vivo which was due to the increased immunogenicity. About 10% of human tumors are also low or defective for HLA class I expression and often these tumors appear to be more malignant. The class I negative tumors could either have arisen from class I low or negative tissues or are HLA loss variants which escaped the attack of the immune system. Altogether, our studies and the data of other laboratories demonstrate the important role of class I antigens for anti-tumor immunity and they suggest that modulation of class I expression by gene transfection or by induction with soluble mediators could be a useful tool for the manipulation of tumor immunity.     

Studies of the mechanisms for the induction of in vivo tumor immunity. VI. Induction of specific and nonspecific cell-mediated immunity in tumor-bearing hosts and its correlation with transplantation tumor immunity

Studies were performed to determine the development of cell-mediated cytotoxic response at tumor site in C57BL/6 mice bearing progressively growing FBL-3 ascites leukemia. The effectors isolated from tumor ascites are found to be highly cytotoxic for leukemic target cells. The levels of cytotoxicity obtained with effectors isolated from tumor site are generally higher than those obtained with immune mice. This cytotoxicity is both specific and nonspecific. The specific cytotoxicity against tumor-associated antigen is mainly mediated by T cells and the nonspecific cytotoxicity against unrelated tumor cells is mediated largely by macrophages. The T-cell-enriched preparation did not give significant natural killer activity. When testing the ability of these effectors to produce in vivo immunity against the challenge of FBL-3, it was found that only T cells could confer the transplantation-type immunity, but the immunity was transient. The macrophage-enriched preparation isolated from tumor ascites failed to give in vivo protection. These findings indicate that in FBL-3 system, mice with progressively growing tumors are able to develop immune response against tumor cells. However, this immunity is probably interfered with by a suppressor factor(s) or suppressor cells which restrict their activity to eliminate the tumor cells effectively.     

Independent immunodominant and immunorecessive tumor-specific antigens on a malignant tumor: antigenic dissection with cytolytic T cell clones

We have analyzed the complexity of a unique tumor-specific transplantation antigen expressed by the murine ultraviolet light-induced fibrosarcoma 1591-RE. This tumor is highly immunogenic and is regularly rejected by normal mice. We have derived a cloned cytolytic T cell line showing a reactivity pattern representative of the cytolytic response of the host rejecting this regressor tumor. Using this T cell line (anti-A), variants of 1591-RE (1591-A-) were selected in vitro that had lost the same antigen as progressor variants of 1591-RE selected by the host in vivo. The in vitro derived variant was then used to generate a second T cell clone (anti-B) that recognized an antigen on the parental tumor that had been retained by the variants derived in vitro. Host-selected progressor variants were also found to have retained this antigen. By selecting for variants in vitro from the parental tumor with the anti-B T cell line, it was shown that the two different antigens (A and B) present on the parental tumor were lost independently of each other. Despite the independence of these two antigens, the host T cell response to the parental regressor tumor was invariably restricted to only the "immunodominant" A antigen.     

The immunogenic properties of drug-resistant murine tumor cells do not correlate with expression of the MDR phenotype

Alterations in the immunogenic properties of tumor cells frequently accompany selection for multipledrug-resistant (MDR) variants. Therefore, studies were performed to examine the hypothesis that overexpression of membrane P-glycoprotein, commonly observed in MDR tumor cells, is associated with enhanced immunogenic properties. Immunogenicity was determined by (a) the ability of drug-sensitive parental UV2237M fibrosarcoma cells and drug-resistant UV2237M variant cells to immunize normal mice against rechallenge with parental tumor cells and (b) the ability of normal syngeneic mice to reject cell inocula that caused progressive tumor growth in immunocompromised mice. Variant UV2237M cell lines included subpopulations selected for a six- to ten-fold increase in mRNA for P-glycoprotein and expression of the MDR phenotype (resistance to doxorubicin) and cells sensitive to doxorubicin (and no expression of MDR properties) but resistant to ouabain. All UV2237M drug-resistant cells were highly immunogenic in immunocompetent mice, regardless of their MDR phenotype. Additional studies showed that CT-26 murine adenocarcinoma cells, sensitive or resistant to doxorubicin (expressing high levels of P-glycoprotein), injected into normal syngeneic Balb/c mice produced rapidly growing tumors. The data do not demonstrate a correlation between the immunogenic properties of drug-resistant tumor cells and the expression of P-glycoprotein.Supported in part by core grant CA-16672 R35-CA42 107 from the National Cancer Institute, and postdoctoral fellowship grant PF-3446 from the American Cancer Society (R. R.)     

Protective immunity to progressive tumors can be induced by antigen presented on regressor tumors

Immunization of animals with 1591-RE tumor cells, a highly immunogenic UV-induced epithelia cell tumor from C3H/HeN mice, that were haptenated with trinitrophenol (TNP) leads to protective immunity against a challenge of TNP-haptenated 3152-PRO tumor cells, a progressive highly malignant. MCA-induced fibrosarcoma from syngeneic mice. Animals that rejected TNP-1591-RE and subsequently TNP-3152-PRO tumor cells showed increased tumor-specific resistance to another challenge of 3152-PRO tumor cells, even when these fibrosarcoma cells had not been haptenated with TNP. Induction of protection required the presence of TNP-hapten groups on both 1591-RE and 3152-PRO during the initial immunization, and could be induced by immunization with other haptenated syngeneic highly immunogenic regressor tumor lines. In addition, TNP-haptenated progressor variants of the 1591-RE were ineffective in generating protection, suggesting that the immunogenicity of the haptenated tumor used for the initial immunization was a determining factor in whether or not protective immunity against the highly malignant tumor was later generated. Protection required at least two T cell types: a Lyt-1-2+ T cells, and a Lyt-1+2- T cell that also expressed I-J determinants and was Vicia villosa lectin adherent, suggesting it was not a classical helper T cell. These results suggest that presentation of a hapten by highly immunogenic tumor cells can lead to enhanced protective immunity to poorly immunogenic noncross-reactive tumors that co-express the same hapten, and rejection of these haptenated poorly immunogenic tumors leads to enhanced protection against a subsequent challenge of the same, but not noncross-reactive progressor tumors.     

TCR hypervariable regions expressed by T cells that respond to effective tumor vaccines

A major goal of immunotherapy for cancer is the activation of T cell responses against tumor-associated antigens (TAAs). One important strategy for improving antitumor immunity is vaccination with peptide variants of TAAs. Understanding the mechanisms underlying the expansion of T cells that respond to the native tumor antigen is an important step in developing effective peptide-variant vaccines. Using an immunogenic mouse colon cancer model, we compare the binding properties and the TCR genes expressed by T cells elicited by peptide variants that elicit variable antitumor immunity directly ex vivo. The steady-state affinity of the natural tumor antigen for the T cells responding to effective peptide vaccines was higher relative to ineffective peptides, consistent with their improved function. Ex vivo analysis showed that T cells responding to the effective peptides expressed a CDR3β motif, which was also shared by T cells responding to the natural antigen and not those responding to the less effective peptide vaccines. Importantly, these data demonstrate that peptide vaccines can expand T cells that naturally respond to tumor antigens, resulting in more effective antitumor immunity. Future immunotherapies may require similar stringent analysis of the responding T cells to select optimal peptides as vaccine candidates.     

Distinct T-cell proliferative responses to 13762A rat mammary adenocarcinoma and derived clones

We examined the in vitro responses of immune lymphocytes to the tumor antigens of the syngeneic rat mammary adenocarcinoma 13762A. This tumor readily metastasizes to lymph node and lungs and is poorly immunogenic. Rats were immunized with a highly immunogenic clone (18A) which was isolated as a spontaneous variant from the parental 13762A tumor. Clone 18A grew progressively in irradiated rats but regressed completely in normal rats. Animals immune to 18A tumor were also immune to parental 13762A. Lymphocytes obtained from the spleen and peritoneum of immune rats were tested for specific proliferation to parental 13762A tumor and clone 18A to determine whether similar cross-reactivity to these tumors occurred in vitro. We found an anatomical difference in localization of immune lymphocytes which reacted to the two tumor cell lines. Immune peritoneal exudate cells (PEC) responded strongly to clone 18A but poorly to 13762A, while immune spleen cells from the same animals responded predominantly to 13762A tumor. After 7 days culture, PEC proliferating in response to clone 18A contained 84-95% W3/25+ T-helper cells, and only 5-8% OX8+ cytotoxic/suppressor cells, while analogous cultures of spleen cells responding to parental 13762A tumor consisted of 60-80% W3/25+ cells and 20-23% OX8+ cells. Immune spleen cell cultures stimulated with 13762A tumor generated cytotoxic lymphocytes which specifically lysed both parental 13762A and clone 18A cells. We conclude that despite cross-reactivity in vivo and in vitro, antigens present on 13762A and 18A tumor cells stimulated different subsets of immune T cells.     

Immune suppression in the tumor microenvironment: a role for dendritic cell-mediated tolerization of T cells

Immune suppression remains a consistent obstacle to successful anti-tumor immune responses. As tumors develop, they create a microenvironment that not only supports tumor growth and metastasis but also reduces potential adaptive immunity to tumor antigens. Among the many components of this tumor microenvironment is a population of dendritic cells which exert profound immune suppressive effects on T cells. In this review, we discuss our recent findings related to these tumor-associated dendritic cells and how targeting them may serve to generate more durable anti-tumor immune responses.     

Murine tumor cells transduced with the gene for tumor necrosis factor-alpha. Evidence for paracrine immune effects of tumor necrosis factor against tumors

Studies of the anti-tumor activity of TNF-alpha in vivo have been hampered by the need to administer systemically toxic doses of the cytokine to obtain a curative response. To facilitate studies of the effect of high local concentrations of TNF-alpha on tumor growth and host immunity, a newly induced murine sarcoma was transduced with the gene for human TNF-alpha and the biologic characteristics of these cells were examined. We identified high and low TNF-producing tumor clones which exhibited stable TNF secretion over time. Significant amounts of membrane associated TNF were found in a high-TNF producing clone as well. No difference in the in vitro growth rates between TNF-producing and nonproducing cell lines was observed. In contrast, in vivo studies demonstrate that although unmodified parental tumor cells grew progressively when implanted s.c. in animals, tumor cells transduced with the TNF gene were found to regress in a significant number of animals after an initial phase of growth. This effect correlated with the amount of TNF produced and could be blocked with a specific anti-TNF antibody. Regressions of TNF-producing cells occurred in the absence of any demonstrable toxicity in the animals bearing these tumors. TNF-producing tumor cells could function in a paracrine fashion by inhibiting the growth of unmodified, parental tumor cells implanted at the same site. The ability of tumor cells to regress was abrogated by in vivo depletion of CD4+ or CD8+ T cell subsets and animals that had experienced regression of TNF-producing tumors rejected subsequent challenges of parental tumor. Our studies thus show that tumor cells elaborating high local concentrations of TNF regress in the absence of toxicity in the host and that this process requires the existence of intact host immunity. Studies of the lymphocytes infiltrating the gene modified tumors and attempts to use TNF gene modified tumor infiltrating lymphocytes to deliver high local concentrations of TNF to the tumor site without inducing systemic toxicity are underway.     

Induction of tumor-specific cytotoxic T lymphocytes and natural killer cells by tumor cells transfected with the interleukin-2 gene

To study the antitumor effect of local production of interleukin-2 (IL-2) from tumor cells, the poorly immunogenic murine colon cancer cells, colon26, was transfected with murine IL-2 cDNA in a bovine papilloma virus vector. IL-2 gene transfectants (mIL2+colon26) did not alter their growth rate compared with parental colon26 cells in vitro, but reduced their tumorigenicity in vivo. Immunization with mIL2+colon26 cells could induce protective immunity against parental colon26 cells. Following intravenous challenges, the colonies of lung metastasis were also inhibited. Moreover, inoculation of mIL2+ colon26 cells slowed the growth of challenged renal cell carcinoma cells, RenCa. Intraperitoneal inoculation of IL-2 gene transfectants generated a large number of peritoneal exudate cells and these cells had a highly cytolytic activity against colon26 and YAC-1. These results suggest that inoculation with IL-2 transfected tumor cells can stimulate not only cytotoxic T lymphocytes but also natural killer cells, and that these cells will act as antitumor effector cells in host animals.     

Sublethal,whole-body ionizing irradiation can be tumor promotive or tumor destructive depending on the stage of development of underlying antitumor immunity

Summary It was shown that sublethal (500 rads), whole-body -irradiation of mice bearing an established i.d. immunogenic tumor can result, after several days delay, in complete tumor regression and long-term survival, but only if radiation is given after the tumor is established and growing progressively. Exposing mice to the same dose of radiation several hours after tumor cells were implanted resulted, in contrast, in enhanced growth of the primary tumor and in earlier death from systemic disease. Irradiation-induced tumor regression failed to occur in mice that were incapable of generating antitumor immunity, because of having been made T cell deficient by thymectomy and irradiation. Again, irradiation-induced tumor regression could be blocked by infusion of spleen cells from donor mice bearing a well-established tumor. These and previously published results support the view that sublethal, whole-body ionizing irradiation causes tumor regression by preferentially destroying radiosensitive suppressor T cells, thereby enabling the host to generate a therapeutic level of concomitant immunity. It is suggested that the preferential destruction of suppressor cells by irradiation depends on the acquisition, during immunologic induction, of radioresistance by antigen-activated effector T cells, and that this is the reason irradiation causes regression only of established tumors. Not all tumors tested were immunogenic enough to undergo regression in response to -irradiation.This study was supported by Grants CA-16642 and CA-27794 from the National Cancer Institute, Grant RR-05705 from the Division of Research Resources, NIH; and a Grant-in-aid from RJR Nabisco     

Is an immune reaction required for malignant transformation and cancer growth?

Increasing evidence has shown that probably all malignant mouse cells, even those of spontaneous sporadic cancers, are endowed with tumor-specific antigens. Stimulation of cancer growth, rather than inhibition by the immune reaction, is seemingly the prevalent effect in the animal of origin (the autochthonous animal). Small initial dosages of even strong tumor antigens tend to produce stimulatory immune reactions rather than tumor inhibition in any animal. Thus, an immune response at a low level may be an essential growth-driving feature of nascent cancers, and this may be why all cancers apparently have tumor-specific antigens. Inasmuch as a low level of immunity is stimulatory to tumor growth while larger dosages are inhibitory, immuno-selection via this low response may tend to keep the antitumor immune reaction weak and at a nearly maximal stimulatory level throughout most of a tumor's existence. These facts suggest that both suppression of tumor immunity and a heightened immune reaction might each be therapeutic although very contrasting modalities.     

Oncomodulatory signals by regulatory proteins encoded by human cytomegalovirus: a novel role for viral infection in tumor progression

A high frequency of human cytomegalovirus (HCMV) genome and antigens in tumor samples of patients with different malignancies is now well documented, although the causative role for HCMV in the development of the neoplasias remains to be established. HCMV infection can modulate multiple cellular regulatory and signalling pathways in a manner similar to that of oncoproteins of small DNA tumor viruses such as human papilloma virus or adenoviruses. However, in contrast to these DNA tumor viruses, HCMV infection fails to transform susceptible normal human cells. There is now growing evidence that tumor cells with disrupted regulatory and signalling pathways enable HCMV to modulate their properties including stimulation of cell proliferation, survival, invasion, production of angiogenic factors, and immunogenic properties. In contrast to previously suggested "hit and run" transformation we suggest that persistence in tumor cells is essential for HCMV to fully express its oncomodulatory effects. These effects are observed particularly in persistent HCMV infection and are mediated mainly by activity of HCMV regulatory proteins. In persistently HCMV-infected tumor cell lines - a selection of novel, slowly growing virus variants with changes in coding sequences for virus regulatory proteins takes place. As a result, oncomodulatory effects of HCMV infection may lead to a shift to more malignant phenotype of tumor cells contributing to tumor progression.     

CD40 ligation in vivo can induce T cell independent antitumor effects even against immunogenic tumors

Antitumor effects of CD40 ligation appear to involve distinct antitumor effector cells in different experimental models. In this study, we tested whether T cells were required for antitumor effects of agonistic anti-CD40 mAb (alphaCD40) against immunogenic versus poorly immunogenic tumors. Treatment of mice bearing poorly immunogenic B16 melanoma and its more immunogenic variant, B16-hsp72.1, with alphaCD40 resulted in a similar level of tumor growth suppression. Depletion of T cells did not reduce the antitumor effects in these 2 tumor models. To generate antitumor T cell responses, C57BL/6 mice were immunized with irradiated B16-hsp72.1. Treatment of these vaccinated mice challenged with a high dose of B16-hsp72.1 tumor cells with alphaCD40 induced tumor growth suppression, which was reduced by T-cell depletion, demonstrating that T cells were involved in the antitumor effect of alphaCD40. However, immunized mice depleted of T cells and treated with alphaCD40 were still able to suppress tumor growth as compared to tumor growth in immunized, T cell-depleted mice not treated with alphaCD40, suggesting that T cells were not required for the antitumor effect of alphaCD40. To confirm a lack of correlation between tumor immunogenicity and T-cell requirement in antitumor effects of CD40 ligation, we found that alphaCD40 induced tumor growth suppression in nude and SCID/beige mice bearing highly immunogenic tumors such as Meth A sarcoma, suggesting that macrophages may play a role. Indeed, both poorly immunogenic and highly immunogenic tumors were sensitive to in vitro growth inhibition by macrophages from alphaCD40-treated mice. Taken together, our results indicate that antitumor effects induced by alphaCD40, even against immunogenic tumors, can be observed in the absence of T cells and may involve macrophages.     

Anti-ganglioside anti-idiotypic monoclonal antibody-based cancer vaccine induces apoptosis and antiangiogenic effect in a metastatic lung carcinoma

Anti-idiotype monoclonal antibody (mAb) 1E10 was generated by immunizing BALB/c mice with an Ab1 mAb which recognizes NeuGc-containing gangliosides, sulfatides and some tumor antigens. 1E10 mAb induces therapeutic effects in a primary breast carcinoma and a melanoma model. However, the tumor immunity mechanisms have not been elucidated. Here we show that aluminum hydroxide-precipitated 1E10 mAb immunization induced anti-metastatic effect in the 3LL-D122 Lewis Lung carcinoma, a poorly immunogenic and highly metastatic model in C57BL/6 mice. The therapeutic effect was associated to the increment of T cells infiltrating metastases, the reduction of new blood vessels formation and the increase of apoptotic tumor cells in lung nodules. Interestingly, active immunization does not induce measurable antibodies to the 1E10 mAb, the NeuGc-GM3 or tumor cells, which may suggest a different mechanism which has to be elucidated. These findings may support the relevance of this target for cancer biotherapy.     

Suppression of growth of guinea pig line-10 hepatocarcinoma. I. Effect of simultaneous administration of tumor cells and antibodies from normal rabbits.

Suppression of growth of the line-10 hepatocarcinoma in strain-2 guinea pigs occurred when line-10 cells were injected intradermally together with sera or immunoglobulins derived from normal rabbits. A significant number of animals were resistant to subsequent rechallenge with tumor cells. This immunity was specific, depended on contact of immunoglobulins with tumor cells and on the concentration of immunoglobulins. Repeated injections acted as potent vaccines and resulted in the development of immunity in 84.6% of recipients. Fc receptors were not detected on line-10 cells. Antibodies reacting with line-10 cell unique antigens as well as with antigens common to line-10, line-1 and normal guinea pig spleen cells were found in NRS. Injection of line-10 cells together with rabbit immunoglobulins from which antibodies reacting with antigens derived from line-10 cells had been removed did not result in tumor suppression. The specific antigen(s) recognized by antibodies that suppressed growth of the line-10 tumor in vivo was not determined.     

T-suppressors from tumor-bearing mice inhibit antitumor cytotoxic T-lymphocytes in monoculture]

The generation of specific antitumor cytotoxic T-lymphocytes (CTL) via 2-fold immunization in vivo and subsequent cultivation without tumor cells (in monoculture) was previously described. The spleen cells from B10 mice bearing progressively growing MX-11 sarcoma suppressed the maturation of CTL specific to MX-11 but not to EL-4 lymphoma in monoculture. It was observed, that the suppression was not the result of the inhibitory effect of suppressor cells upon the IL-2 production, because suppression took place in the presence of the exogenous IL-2 in monoculture. Since the treatment of the spleen cells with MoAb against both L3T4 and Lyt2.2 antigens plus C' considerably decreased the suppressive activity, it was suggested, that two distinct subsets of T-lymphocytes were required for suppression. It might be possible, that the presence of anti-idiotype on the effector suppressors was the cause of the suppressive specificity in the absence of tumor antigens in vitro.     

Bystander elimination of antigen loss variants in established tumors

Cancers express antigens that are targets for specific cytotoxic T lymphocytes (CTLs). However, cancer cells are genetically unstable. Consequently, sub-populations of cancer cells that no longer express the target antigen may escape destruction by CTLs and grow progressively. We show that cytotoxic T cells indirectly eliminate these antigen loss variants (ALVs) in a model system when the parental cancer cells express sufficient antigen to be effectively cross-presented by the tumor stroma. When the parental tumor expressed lower levels of antigen, cytotoxic T cells eradicated the antigen-positive parental cancer cells, but the ALVs escaped, grew and killed the host. By contrast, when the parental tumor expressed higher levels of antigen, cytotoxic T cells eradicated not only the parental cancer cells but also the ALVs. This 'bystander' elimination of ALVs required stromal cells expressing major histocompatibility complex (MHC) molecules capable of presenting the antigen, and occurred in tumors showing evidence of stromal destruction. ALVs were apparently eliminated indirectly when tumor-specific CTLs killed stromal cells that were cross-presenting antigen produced by and released from antigen-positive cancer cells. These results highlight the general importance of targeting the tumor stroma to prevent the escape of variant cancer cells.