Combination of T-Cell Therapy and Trigger of Inflammation Induces Remodeling of the Vasculature and Tumor Eradication¹

Tumors grow progressively in the host, despite their antigenicity and the presence of potentially tumor-reactive lymphocytes. In metastatic melanoma patients, T cells specific for tumor antigens have been detected in the circulation and in tumor-infiltrated lymph nodes (1, 2). Although these cells have encountered antigen and are capable of proliferating, they do not elicit protective immunity (1). Similarly, T cells recovered from tumors can display lytic activity in vitro but frequently fail to reject the tumors from which they were isolated (3, 4). It is widely accepted that insufficient effector functions account for the failure of tumor immunity. Thus, animal experiments and clinical trials aim at allowing effector cells to proliferate, to acquire lytic activity, and/or to release large amounts of cytokines. However, antitumor immune strategies that have been effective in controling tumor growth in mice have failed in patients with malignancies (5, 6).

Once successfully activated, effector T cells must be able to recirculate, enter tumor sites, and function within the tumor microenvironment to trigger the rejection process. Although tumor infiltration by effector T cells is a prerequisite for rejection (7), it is no guarantee for an immunologically specific interaction (8). There is accumulating evidence that local factors such as tumor-derived stroma (9, 10) and the cytokine micromilieu (11) play a crucial role in tumor immunity. Thus, for immune-mediated destruction of solid tumors, the effector phase is as important as the initial priming of T cells. There is also a clear correlation between tumor size and therapeutic success (6). Immunotherapy in classical transplantation models and experimentally induced metastasis is usually started simultaneously with tumor challenge or at a low tumor burden and is, therefore, not suited to destroy solid tumors. It is not surprising then, that the translation of tumor immune therapy to the clinic has been proven ineffective, and the numbers of responders remain small (5). Transgenic mice developing autochthonous tumors might better reflect the clinical situation in terms of tissue tropism and growth kinetics. As is apparent in human cancers, however, these mouse models of de novo tumorigenesis confirm that currently available immunotherapeutic strategies are most effective either in a preventive setting (12, 13) or in a setting in which the growth of solid tumors is merely delayed (14, 15).

The RIP1-Tag5 transgenic mouse is created by expressing Tag³ under the control of the RIP (16). Onset of oncogene expression in pancreatic islets of Langerhans in adult mice causes transformation of β-cells and progressive tumor development, although an immune response against Tag is evident (17). During multistage carcinogenesis, normal and hyperproliferative islets are infiltrated by lymphocytes. However, the lymphocytic infiltration is lost during the transition from angiogenic islets to highly vascularized, solid tumors. We have previously shown that increasing the frequency of Tag-specific lymphocytes and providing costimulation on tumor cells failed to provoke immune-mediated tumor destruction (18, 19). Thus, it became apparent that a potent tumor antigen, together with highly activated, self-reactive lymphocytes, is not sufficient for tumor rejection, although these factors are nonetheless considered to be essential for T-cell based immunotherapy (20). We, therefore, postulated that a permissive tumor microenvironment is required for effector function and subsequently demonstrated that irradiation of transgenic mice renders solid tumors accessible for lymphocytic infiltration (19, 21). Until now, however, the potential therapeutic efficacy of this treatment was not further investigated.

In normal tissue, radiation causes an inflammatory response that includes cytokine release, up-regulation of adhesion molecules, and a subsequent increase in lymphocyte adhesion (22, 23). Therefore, if irradiation is an inflammatory stimulus, is it able to promote inflammation in a highly angiogenic tumor? We evaluated the impact of irradiation on established tumors in transgenic mice and its efficacy for immunotherapy. We demonstrate that even persistently high levels of tumor-specific T cells, used alone, failed to eradicate solid tumors, but a combination of irradiation and multiple T cell transfers led to complete tumor regression. This antitumor effect was mediated by a proinflammatory environment and a dramatic lymphocytic influx. In addition, the combination of irradiation and adoptive transfers reestablished an almost normal vasculature within tumors early after the initiation of therapy, thus breaking the vicious cycle of tumor-induced angiogenesis.

RIP1-Tag5 mice, kindly provided by Douglas Hanahan (16), were generated in the C3HeB/Fe background. Tag expression in RIP1-Tag5 mice starts at 10 weeks of age and leads to the formation of insulinomas and premature death at ∼30 weeks. TagTCR1 mice express the MHC class II-restricted T-cell receptor specific for Tag; peripheral lymphocytes represent 10% transgenic CD4⁺ T cells with a normal CD4⁺:CD8⁺ T-cell ratio. TagTCR1 mice, kindly provided by Irmgard Förster (24), and Rag-1^−/− mice were backcrossed to C3HebFe (Jackson Laboratory, Bar Harbor, ME) and kept under pathogen-free conditions at the German Cancer Research Center.

RIP1-Tag5 mice were lethally irradiated with 10 Gy (X-ray source, Buchler, Braunschweig, Germany, 0.701 Gy/min). One day after irradiation, 5 × 10⁶ bone marrow cells from Rag-1^−/− or C3HebFe mice were injected i.v. Three weeks later, 5.0 × 10⁶ or 2.5 × 10⁶ in vitro activated TagTCR1 cells were injected i.v. and i.p. TagTCR1 cells were isolated from lymph nodes and 1.5 × 10⁷ cells/six-well plate were cultured in RPMI medium supplemented with 10% FCS, 2 mm glutamine, 100 units/ml penicillin/100 μg/ml streptomycin, 0.05 mm 2-mercaptoethanol, 10 units recombinant IL-2/ml and 25 nm Tag peptide 362–384. C3H-derived lymphocytes were cultured with 300 ng/ml ConA. Cells were harvested after 72 h.

Lymph node or tumor cell suspensions were stained with biotinylated anti-clonotype (TagTCR) antibody 9H5 [10 μg/ml; from I. Förster (24)], followed by streptavidin-Red670 (1.7 μg/ml, Life Technologies, Inc., Eggenstein, Germany), and PE-conjugated anti-CD4 (GK1.5, 2.0 μg/ml; BD PharMingen, Heidelberg, Germany) for two-color analysis. Cells were analyzed on a FACScan (Becton Dickinson).

Solid tumors were dissected from pancreatic tissue after perfusion of 1× HBSS/5 mm glucose through the common bile duct into the pancreas. Tumor size was measured with a caliper and the formula [volume = 0.52 × (width)² × length] was applied to determine the tumor burden (mm³) per mouse.

Pancreases were disrupted in 50 mm Tris (pH 8.0), 1% 2-mercaptoethanol, 0.3% SDS, and 1 tablet of protease inhibitors/10 ml (Roche, Mannheim, Germany). Extracts were heated for 10 min at 95°C and clarified by centrifugation. Fifteen μg of total protein extract were separated on 10% SDS/PAGE. The membrane was blocked in 1× PBS, 0.1% Tween 20, and 5% milk powder and probed with 1 μg/ml SV40T-Ag polyclonal mouse antibody (416; Dianova, Hamburg, Germany), antimouse IgG-horseradish peroxidase (10 ng/ml; Vector/Alexis Biochemicals, Grünberg, Germany) and SuperSignal chemiluminescent substrate (Pierce, Rockford, IL).

Staining was performed as described (19) with the following reagents: anti-Tag (rabbit polyclonal, 1:1000; from D. Hanahan); anti-CD4 (GK1.5, 10 μg/ml; BD PharMingen); anti-CD8 (Ly-2, 10 μg/ml; BD PharMingen), and anti-CD11b (Mac-1, 10 μg/ml; BD PharMingen). For in situ hybridization the following PCR fragments were subcloned into Bluescript vector: 639 bp Mig: forward primer (FP): 5′-CATCATCTTCCTGGAGCAGTGT-3′, reverse primer (RP): 5′-GACACTTGCCCAGATCTGATGT-3′; 341 bp IP10: FP: 5′-CAAGGGATCCCTCTCGCAAGGA-3′, RP: 5′-AGGGCAATTAGGACTAGCCATC-3′. Digoxigenin-labeled (dig-11-UTP; Roche) sense and antisense probes were synthesized from XbaI or BamHI linearized plasmids, using T3 and T7 polymerase (Roche), respectively. Ten-μm cryosections were treated with 4% paraformaldehyde and 0.2 m HCl, acetylated with acetic anhydride for 10 min, and hybridized with 250 ng of RNA probe/section in 100 μl of hybridization buffer [1% blocking reagent (Roche; 50% formamide, 5× SSC, 1 mg/ml yeast RNA, 0.5% Tween 20, 0.1 mg/ml heparin, 0.1% CHAPS, and 5 mm EDTA) at 55°C overnight. Sections were washed in 2× SSPE/0.3% CHAPS, treated with RNase A (20 μg/ml at 37°C for 30 min), followed by a wash in 50% formamide and 2× SSPE for 1.5 h at 50°C. RNA hybrids were detected with antidigoxigenin, alkaline phosphatase Fab-fragments (7.5 units/ml; Roche), followed by incubation with chromogenic substrate (0.45 mg/ml 4-nitroblue tetrazolium chloride and 0.18 mg/ml 5-bromo-4-chloro-3-indolyl-phosphate; Roche) in a buffer containing 0.1 m Tris (pH 9.5), 1 m NaCl, 0.05 m MgCl₂, 0.1% Tween 20, and 0.24 mg levamisole (Sigma Chemical Co., Taufkirchen, Germany). Sections were counterstained in 1% methyl green.

Staining for apoptotic cells was performed on 5-μm frozen sections using in situ cell-death detection kit-fluorescein (Roche). Subsequently, endothelial cells were stained with MECA32 (5 μg/ml; BD PharMingen) and Cy3-conjugated antirat IgG Fab-fragments (2 μg/ml; Dianova).

Intravital microscopy was performed as described previously for the rat (25). The pancreatic head was immobilized in a 37°C immersion chamber with Ringer’s solution. Animals received i.v. injections of 1 μg/g FITC-labeled, autologous mouse erythrocytes to measure erythrocyte velocity, followed by 50 μg/g FITC-labeled albumin (Sigma Chemical Co.) to obtain maximal vessel contrast for video recording of morphology and to evaluate vessel diameters.

RNA was prepared from snap-frozen tumors using RNeasy mini kit (Qiagen, Hilden, Germany). Three μg of RNA were treated with DNaseI (Promega, Mannheim, Germany), and cDNA synthesis was primed with random hexamers (Amersham Pharmacia, Freiburg, Germany), followed by reverse transcription using 1 unit of M-MLV (Life Technologies, Inc.)/15 μl. Reaction was stopped with 160 μl of H₂O. cDNA (2.5 μl) was analyzed using real-time PCR TaqMan technology (Applied Biosystems, Weiterstadt, Germany). The mouse hypoxanthine phosphoribosyltransferase (HPRT) gene served as the internal standard. The following primers and probes were used: HPRT (amplicon length: 93 bp): FP, 5′-ACACCTGCTAATTTTACTGGCAACA-3′; RP, 5′-TGGAAAAGCCAAATACAAAGCCTA-3′; probe, 5′-CTCGTATTTGCAGATTCAACTTGCGCTCATC-3′; IFN-γ (95-bp): FP, 5′-ATCTGGCTCTGCAGGATTTTCA-3′; RP, 5′-TCAAGTGGCATAGATGTGGAAGAA-3′; probe, 5′-CCTTTTGCCAGTTCCTCCAGATATCCAAGAAG-3′; TNF-α (94 bp): FP, 5′-ACAAGGCTGCCCCGACTA-3′; RP, 5′-CTTGACGGCAGAAGGAGGTT-3′; probe: 5′-CCCACACCGTCAGCCGATTTGC-3′; Mig (144 bp): FP, 5′-GAGGAACCCTAGTGATAAGGAATGC-3′; RP, 5′-TCTTCAGTGTAGCAATGATTTCAGTTT-3′; probe: 5′-ATCAGCACCAGCCGAGGCACG-3′; IP10 (108 bp): FP, 5′-CCTTCACCATGTGCCATGC-3′; RP, 5′-TCTTACATCTGAAATAAAAGAGCTCAGGT-3′; probe, 5′-CCCCACACCCTCCTTGTCCTCCC-3′.

β-cell tumors in RIP1-Tag5 transgenic mice are free of infiltrating lymphocytes and grow progressively, despite an initial immune reactivity toward the tumor antigen. We had postulated that effective immune-mediated tumor therapy requires both activated antitumor lymphocytes and their access into tissue (19, 21). In the present study, tumor-bearing RIP1-Tag5 mice were lethally irradiated with a dose of 10 Gy, which was insufficient to cause tumor cell death. All of the irradiated mice were reconstituted with bone marrow derived from syngeneic wild-type (C3H) or T-cell-deficient (Rag-1^−/−) mice as indicated. Solid tumors of mice receiving irradiation treatment alone showed a varying but generally low degree of T cell infiltration (Fig. 1, A and B), which was not significantly increased when compared with untreated tumors. For adoptive transfers, lymph node cells from transgenic mice expressing a T-cell receptor specific for Tag (TagTCR1) were activated in vitro. Transferred cells represent a pool of lymphocytes with 10% transgenic CD4⁺ T cells, but otherwise there is a normal CD4⁺:CD8⁺ T-cell ratio (24). Adoptive transfer in nonirradiated RIP1-Tag5 mice resulted in an accumulation of T cells at the periphery of solid tumors as early as 2 days after transfer. These cells, however, failed to efficiently penetrate into tumor tissue (Fig. 1, C and D). In contrast, adoptively transferred T cells in irradiated mice migrated into solid tumors and accumulated over time (Fig. 1, E and F). This experiment demonstrates that accessibility of tumor-specific lymphocytes to solid tumors requires “conditioning” of the organ by irradiation.

It has been reported that radiation therapy correlates with up-regulation of ICAM-1 and VCAM-1 on endothelial cells (23). Consistent with this finding, quantitative PCR analyses of solid tumors obtained from irradiated RIP1-Tag5 mice showed enhanced expression of ICAM-1 (4- to 6-fold) and VCAM-1 (3- to 4-fold) 3 weeks after irradiation, at the time of the first infusion of activated T cells (data not shown).

Tag expression in RIP1-Tag5 mice starts at 10 weeks of age and results in the stepwise development of insulinomas (19). The incidence of multifocal islet tumors increased with age and correlated with the amount of Tag present in a given pancreas (Fig. 2,A). To examine whether enhanced T-cell extravasation correlates with the rejection of already established tumors in RIP1-Tag5 mice, 22-week-old mice were irradiated and received a single adoptive transfer of 5 × 10⁶ TagTCR lymphocytes 3 weeks later. Treated animals showed no survival advantage over the untreated control group (data not shown). All of the transgenic mice developed insulinomas and died of hypoglycemia between week 27 and 34. Clearly, infiltrating activated antitumor lymphocytes are either poor effectors or they do not survive long enough to impact tumor growth. To monitor the fate of adoptively transferred T cells, we analyzed migration and frequency of TagTCR and CD4⁺ double-positive cells for up to 20 days after transfer by using flow cytometry. Tag-specific CD4⁺ T cells preferentially expanded in the draining lymph nodes of the pancreas and accumulated in tumor tissue reaching maximum numbers at day 7 (Fig. 2,B). Although no tumor cells were found within draining lymph nodes (data not shown), Tag is efficiently presented in pancreatic lymph nodes of RIP1-Tag5 mice and naïve 5,6-carboxy-succinimidyl-fluorescein-ester (CFSE)-labeled TagTCR lymphocytes are primed to proliferate.⁴ In addition, up-regulation of T-cell receptor expression was exclusively observed on T cells populating the pancreatic lymph nodes and solid tumors at day 7 after adoptive transfer (Fig. 2 C). In vivo expansion, however, was short lived and rapidly declined within 20 days after transfer. These data demonstrate that, after irradiation and adoptive transfer, highly activated antitumor lymphocytes are initially found at tumor and tumor-draining sites. It is likely that these cells disappear by activation-induced cell death (26, 27). The small remaining T-cell population infiltrating tumors at day 20 is unable to prevent progressive tumor growth.

Because the survival of activated, tumor-specific lymphocytes may be crucial for long-term therapeutic success, we repeatedly transferred ex vivo activated lymphocytes at intervals of 10 days and monitored the frequency of TagTCR/CD4⁺ T cells. This treatment regimen ensures consistently high numbers of Tag-specific T cells in the tumor-draining lymph nodes (Fig. 3,A). A group of 25-week-old tumor-bearing mice was again irradiated, followed by repeated injections of 2.5 × 10⁶ activated antitumor TagTCR1 cells. Control groups consisted of transgenic mice that received: (a) repetitive adoptive transfers without prior irradiation; (b) irradiation alone; or (c) irradiation and repeated injections of ConA-activated lymphocytes. Blood glucose levels and body weight were monitored every 10 days. None of the mice developed diabetes or dramatic weight loss during therapy (data not shown). All of the control animals became hypoglycemic because of the growth of insulin-secreting tumors, and mice in groups (b) and (c) died between weeks 26 and 34 (Fig. 3,B), similar to untreated RIP1-Tag5 mice. Mice receiving adoptive transfers without irradiation (group a) had a minor survival advantage compared with other control mice (Fig. 3,B; P = 0.0129); a similar survival advantage is observed in RIP1-Tag5 × TagTCR1 double transgenic mice.⁴ Notably, ConA-activated T cells infiltrated tumor tissue but were unable to prevent tumor growth (data not shown). In contrast, the life span of transgenic mice was dramatically increased when irradiation was combined with repeated transfers of Tag-specific lymphocytes. At the age of 46 weeks, 27 of 32 treated mice were still alive. Therapy with Tag-primed lymphocytes derived from C3H mice instead of TagTCR cells was equally efficient (data not shown). The same therapeutic outcome was observed in T-cell-deficient as well as immunocompetent mice, which indicated that potential intrinsic regulatory mechanisms are negligible during the effector phase in these studies (Fig. 3 B).

The dramatic survival advantage of treated animals is a first demonstration of the efficacy of our combined irradiation/transfer therapy. To determine whether therapeutic intervention merely causes stagnation of tumor growth rather than tumor rejection, long-term therapy was repeated, and the average tumor burden per mouse was evaluated after each adoptive transfer. The tumor load increased in all of the treatment groups up to the second adoptive transfer. Thereafter, a steady decline of tumor volume was observed in the irradiation/transfer group but not in control mice. After the fifth transfer, no macroscopic tumors were detected in the irradiation/transfer group, whereas most of the control mice had died (Table 1). Again, this result shows that the efficacy of adoptive transfers depends strictly on an initial radiation stimulus and, in addition, leads to complete macroscopic tumor rejection.

Strikingly, when tumors of different experimental groups were compared, the typical red appearance caused by intense vascularization and hemorrhaging was lost in mice that later showed complete tumor regression (Fig. 4,A). Only in rare cases, did irradiation treatment alone cause some discoloring that resulted in tumors with a “patchy” red/white appearance. The change in color evolved gradually and was most prominent after the second adoptive transfer, which suggests a potential role of infiltrating lymphocytes. Histological analysis of tumors, recovered 6 weeks after irradiation alone, showed a mild T-cell infiltration (Fig. 4, B and C) and a Tag expression pattern similar to untreated tumors (Fig. 4,D). In contrast, white tumor nodules displayed an extremely high degree of lymphocytic infiltration with colocalization of CD4⁺ and CD8⁺ T cells (Fig. 4, E and F). Lymphocyte penetration into tumor tissue correlated with a decreased number of Tag-expressing tumor cells (Fig. 4 G), most likely because of immune-mediated tumor cell destruction. After the second adoptive transfer, the number of intrapancreatic lymph nodes was increased by a factor of 2. Thus, de novo formation of lymph node tissue in close vicinity to tumors may contribute to therapeutic efficacy.

Induction of angiogenesis is commonly observed in growing tumors with a high demand for nutrients and oxygen. Solid tumors in untreated RIP1-Tag5 mice are highly angiogenic. In contrast, the majority of tumors recovered from mice after irradiation and repeated transfers were homogeneously white, with minimal hemorrhaging. Radiation-induced apoptosis of normal endothelial cells has been shown in vitro and in vivo (28, 29). To investigate whether ionizing radiation damages tumor endothelium, the presence of apoptotic endothelial cells was assessed in untreated tumors and in tumors after irradiation and two adoptive transfers (Fig. 5, A and C). Because tumor-cell apoptosis has been reported to be an integral part of the RIP-Tag tumorigenesis (30), we infer that the TUNEL assay detects apoptotic tumor cells (Fig. 5,A). A moderate increase in the apoptotic index was observed in irradiated tumors (data not shown) and in tumors after irradiation and two adoptive transfers (Fig. 5,C). Surprisingly, combined irradiation/transfer therapy did not induce endothelial cell death, but enhanced vessel density in the tumors of treated mice (Fig. 5,C). Neither irradiation nor adoptive transfers alone induced vessel death or any morphological changes to the vasculature (data not shown). We then used intravital microscopy to analyze tumor vessel morphology and hemodynamic parameters in vivo. Untreated RIP1-Tag5 tumors displayed an irregular vascular network with large vessels juxtaposed to small vessels (Fig. 5,B), loss of vessel hierarchy, and zones of neovascularization next to large hemorrhagic areas. However, tumors of white appearance obtained after therapy (n = 3), have a uniform vascularity (Fig. 5 D). The majority of vessels (83%) displays diameters of 10 μm, thus resembling capillaries of the normal circulation. In addition, structurally defined arterioles, capillaries, and venules become apparent, whereas vascular sinusoids and lacunas, typical markers of high angiogenic activity, were absent (data not shown). Erythrocyte velocity was not affected by alteration in vessel morphology (data not shown). Collectively, these data show that irradiation and subsequent massive lymphocytic infiltration transform the aberrant, highly angiogenic vasculature into a network of small capillaries of normal appearance, most likely because of the inhibition of ongoing angiogenesis.

An inflammatory response involves tissue localization of leukocytes and recruitment of reactive cell types in response to chemoattractant signals. We analyzed mRNA expression of several prominent immune modulators in tumors of untreated mice, mice receiving irradiation alone, and mice treated with irradiation followed by two adoptive transfers. mRNA from solid tumor masses, including nontumor cells such as stroma, endothelial cells, and infiltrating effector cells was quantified by real-time PCR. Irradiation alone caused a 4-to 6-fold up-regulation of IFN-γ mRNA in RIP1-Tag5 tumors, and a 2-fold up-regulation of TNF-α mRNA, when compared with untreated mice or mice receiving adoptive transfer alone (Fig. 6,A). Elevated IFN-γ expression correlated with enhanced expression of two distinct members of the CXC-chemokine family, namely Mig and IP10 (Fig. 6,A). These findings imply that irradiation alone has a moderate, but distinct effect on the tumor microenvironment that is not sufficient to prevent tumor growth but that facilitates lymphocyte extravasation. Subsequently, migration of adoptively transferred lymphocytes into tumor tissue led to a dramatic increase in IL-12 (data not shown), IFN-γ, TNF-α, Mig, and IP10 expression (Fig. 6,A). Although IFN-γ and TNF-α in themselves enhance immunological recognition of tumor cells (31, 32), they are potent inducers of Mig and IP10 and can also act synergistically (33). The antitumor effects of IL-12 have been attributed to an up-regulation of IFN-γ, which in turn induces Mig and IP10 expression (34). In addition, Mig and IP10 are shown to have antiangiogenic effects in in vivo Matrigel assays (35). Because reduced angiogenesis is a feature of regressing RIP1-Tag5 tumors (Figs. 4,A and 5, C and D), we analyzed chemokine expression and localization in tumor tissue by in situ hybridization. A few Mig- or IP10-positive cells were present in irradiated tumors (Fig. 6, B and C), but a dramatic increase became apparent after combined irradiation/transfer therapy (Fig. 6, E and F). Tumor cells did not secrete Mig or IP10. On the basis of the expression patterns shown by in situ hybridization and immunohistochemistry (Fig. 6, D and G), we infer that a subset of infiltrating CD11b⁺ monocytes/macrophages display a high chemokine expression. The strong correlation between RIP1-Tag5 tumor regression and high levels of Mig and IP10, together with their documented antiangiogenic and chemotactic activities (35, 36), implies a crucial role for these factors in reestablishing a uniform vascularization and in recruiting activated effector cells.

In this study, we have shown that irradiation and repetitive adoptive transfers lead to complete macroscopic regression of established tumors. Therapeutic efficacy correlates with proinflammatory events in the tumor, which subsequently trigger antigen-dependent effector functions. Our data provide convincing evidence that activated antitumor lymphocytes can lead only to immune-mediated tumor rejection when used in combination with a permissive microenvironment, a concept yet to be applied in human malignancy.

In humans and mice, immunogenic tumors can coexist with antitumor lymphocytes without any impact on tumor progression (1, 21). Furthermore, there is evidence for tolerization of self-specific T cells in the presence of a growing tumor (1, 37). Indeed, in RIP1-Tag5 mice, we cannot rule out the presence of regulatory mechanisms that interfere with the initial priming of the endogenous immune response. In the present study, we focused on the effector arm of antitumor immunity by performing adoptive transfers with a mixture of activated lymphocytes, including MHC class II-restricted T cells with anti-Tag specificity and CD8⁺ T cells of the normal repertoire. It is well accepted that CD4⁺ T cells have the capacity for tumor rejection, but tumor immunity may require a collaborative effort of multiple immunological effectors (38). Although transferred anti-Tag TCR cells proliferate in vivo after encountering the antigen, their life span is short. In the autochthonous RIP1-Tag5 tumor model, we show that the elimination of established tumors by immune cells occurs over a period of weeks, not days, which emphasizes the need for effector cells to persist in the host. Repetitive infusions are required to maintain a consistently high level of antitumor effector cells. In a clinical setting, vaccination strategies may provide a more efficient way to generate and restimulate sufficient numbers of antitumor lymphocytes.

Because we had previously hypothesized a crucial role for the microenvironment in tumor escape from immune destruction (21), the failure of adoptive transfers alone to prolong the life of transgenic mice was not unexpected. Indeed, there is precedence for progressive tumor growth even in the presence of highly activated T cells specific for a tumor antigen (4, 14, 19). Viral infections, for instance, have been shown to slightly increase the efficacy of antitumor responses, most likely because of bystander effects (4, 14). Similarly, a locally applied infectious agent, such as bacillus Calmette-Guérin (BCG), triggers mononuclear cell infiltration into the superficial layers of the bladder and is used in the treatment of bladder cancer (39). Moreover, genetically engineered tumor cells, which locally increase cytokine concentrations, interfere with tumor growth in mouse transplantation models (11, 31, 32). Thus, the concept that inflammatory mediators enhance tumor regression is not new, but has been attributed mainly to a general boost in effector function. Similarly, irradiation of the tumor-bearing host before the administration of ex vivo activated tumor-infiltrating lymphocytes was shown to be more effective than adoptive transfers alone (6, 40). However, in those experiments, the effect of irradiation was mainly attributed to a survival advantage of adoptively transferred cells and the elimination of a postulated, intrinsic, cellular suppressor activity (41).

In this study, we show that irradiation of tumor-bearing RIP1-Tag5 mice with a dose that does not affect tumor growth, “conditions” solid tumors and renders them susceptible to lymphocyte infiltration. Subsequent tumor rejection is observed only when irradiation is combined with the action of tumor-antigen-specific T cells. Infiltration of nonspecifically primed T cells has no impact on tumor growth. Strikingly similar observations were made in a transgenic mouse model for liver autoimmunity, in which “conditioning” of the organ by irradiation or bacterial infection, is a prerequisite for tissue infiltration, and only self-reactive lymphocytes cause liver damage (42). Thus, our data present evidence for the hypothesis that, in a noninflammatory environment, effector and target cells can coexist without initiating autoimmunity or tumor rejection (43).

We also demonstrate that irradiation changes the tumor microenvironment by increasing the expression of adhesion molecules, cytokines, and chemokines. The effect is pleiotropic and surprisingly sustained, such that even 3 weeks after irradiation, an infusion of activated, self-specific T cells causes tissue damage. We have evidence that tumor escape in RIP1-Tag5 mice is linked to the sequential transformation of the vasculature during angiogenesis.⁵ Solid tumors display an aberrant vasculature with a loss of vessel hierarchy and leukocyte adhesion. Reduced leukocyte adherence to tumor endothelium has also been demonstrated in other animal models (25, 44), and angiogenic stimuli are known to directly suppress expression of multiple adhesion molecules on tumor-associated vessels (45). It is, therefore, reasonable to speculate that radiation-induced changes to the endothelium reverse the nonadhesive phenotype of tumor endothelium and, thus, directly facilitate leukocyte-endothelium interactions. Although radiation therapy kills normal endothelial cells (22), we did not observe apoptosis in tumor vessels of irradiated RIP1-Tag5 mice. It is possible that a proangiogenic environment protects tumor endothelium from apoptosis. For instance, in vitro, high levels of basic fibroblast growth factor (bFGF) and vascular endothelial growth factor (VEGF) confer radiation-resistance on normal endothelial cells (28, 29). Nevertheless, combined irradiation/adoptive transfer therapy showed profound effects on RIP1-Tag5 tumors such as a dramatic transformation of the vasculature into a uniform network of capillaries with almost normal appearance. Again, no endothelial cell death or necrosis was obvious, but we observed a reduction in Tag⁺ cells and dramatically enhanced expression of IFN-γ, TNF-α, and the angiostatic chemokines Mig and IP10. In immunocompromised mice, intratumoral Mig treatment reduces the tumor burden either by decreasing vessel density (46) or by inducing endothelial cell death and necrosis (47). Our data imply that elevated expression of endogenous Mig and IP10 in RIP1-Tag5 tumors plays a crucial role in reestablishing a more normal, functional vasculature, which clearly argues for an angiogstatic effect of Mig and IP10 under physiological conditions. Mig and IP10 also display chemotactic properties and enhance leukocyte adhesion in vitro (33, 36). Because we observed not only decreased angiogenic activity in RIP1-Tag5 tumors but also massive leukocytic infiltration, we propose a dual role for Mig and IP10 in irradiation-triggered tumor rejection, namely, inhibiting angiogenesis and promoting lymphocyte chemotaxis. Although it is not clear which cell types secrete high amounts of Mig and IP10 during antitumor therapy, we show that a subset of infiltrating CD11b⁺ cells, but not tumor cells, are likely to express high levels of both chemokines. Contribution by other cell types, e.g., endothelial cells, cannot be excluded.

As a potential sequence of events, we postulate that a proinflammatory stimulus activates the host immune system and enables monocytes/macrophages to gain access to tumors. The permissive environment, including an altered vessel morphology, provides all of the necessary components for a localized, antigen-driven immune response, in which CD4⁺ T cells mediate specificity and recruit CD8⁺ T cells and other effector cells, like monocytes/macrophages. Then, convergence and collaboration of leukocytes within the tumor ultimately result in its destruction. In the present study, inflammation was induced by irradiation, but more specific mediators such as immunostimulatory bacterial DNA should also be effective. Irradiation alone induces proinflammatory events, which do not reach a critical threshold to activate the endogenous immune system derived from bone marrow reconstitution. It is only by subsequently infusing activated effector cells that tumor rejection occurs. However, the infused cells have a short life span, necessitating repeated transfers for a sustained antitumor response.

Although we have increasing knowledge about molecules such as Mig and IP10, and their functions during an antitumor response, no single agent has thus far been identified that is able to trigger the entire complex cascade of events needed to eradicate solid tumors. In human tumors, the degree of infiltration depends on the tumor type and is variable and unpredictable (21). We envision that more promising immunotherapeutic strategies against cancer will use a combination of nonspecific inflammatory stimuli to optimize T-cell infiltration, with continuous effector priming at the tumor site to elicit a sustained antitumor inflammatory response.

We thank C. Schumann and L. Umansky for excellent technical assistance, I. Förster (Technichal University, Munich, Germany) and D. Hanahan (UCSF, San Francisco, CA) for providing transgenic mice and reagents, T. Sacher for artwork, H-J. Gröne for advice on histology, and H. Ee and G. Moldenhauer for critical reading of the manuscript.