EpCAM, a human tumor-associated antigen promotes Th2 development and tumor immune evasion

Epithelial cell adhesion molecule (EpCAM), a human cell-surface glycoprotein, was identified as a human tumor-associated antigen (TAA) in 1979 and was recently described as cancer stem cell marker. EpCAM is a panepithelial differentiation antigen that is restricted to normal epithelia in healthy subjects, but is typically overexpressed to various degrees in most human carcinomas, making it an attractive target for tumor therapy.¹  Several clinical trials with monoclonal²  or bispecific antibodies and vaccination strategies directed against EpCAM showed variable success in the treatment of carcinomas.¹  Development of EpCAM-specific T cell–based strategies were suggested by in vitro studies with either chimeric T-cell receptor (TCR)–transduced primary T lymphocytes³  or TAA-specific cytotoxic T lymphocytes (CTL),⁴  showing cytolytic activity against EpCAM-expressing tumor cells.

T cell–based therapies focus either on active immunization^5,6  or on adoptive transfer of in vitro–generated CD4⁺ or CD8⁺ T-cell lines.^7,8  As vaccination is easier to perform and has long been tested, most studies focus on active immunization. For active immunization antigens are administered either with different adjuvants or using tumor antigen–loaded dendritic cells (DCs) as a “natural adjuvant.”^9,10  Although such vaccination strategies are used in the clinic, the therapeutic efficiency is still limited.^5,6,11  One reason might be that immunization induces efficient T-cell expansion, but T cells fail to control tumor growth.¹²  Moreover, it is possible that tumors themselves actively paralyze tumor immune responses as shown in various models of tumor disease.^12-15  Thus, in mice that express the transgenic tumor promoter “T antigen” (Tag) and simultaneously the Tag-specific TCR, the growing tumors both paralyze and delete Tag-reactive T cells.¹⁴  Tag-specific T cells from mice with established tumors show severe defects in proliferation and cytokine production.¹⁴  Moreover, tumors can also actively induce regulatory T cells.¹⁶  Such negative regulators of protective immunity complicate active induction of therapeutically efficient tumor immune responses, even with the use of optimized DCs as an adjuvant.¹⁷ 

Adoptive transfer of in vitro–generated TAA-specific T cells can circumvent many of these problems. This approach has been extensively explored for the treatment of established experimental tumors in several mouse models^18-20  and even in humans.^21,22  Early studies used in vitro–primed spleen cells from immunized mice or CD8⁺ CTLs.^7,20  Subsequent studies with in vitro–generated T-cell subpopulations revealed that interferon-γ (IFN-γ)–producing CD4⁺ T-helper cells (Th1) are therapeutically more effective than either CD8⁺ CTLs or Th2 cells.^18,20,23,24  In contrast to initial reports,^25,26  we and others clearly showed that interleukin-4 (IL-4)–producing Th2 cells fail to provide protection against tumor cells, except if IL-4 is used for in vivo priming of DCs.²⁷  Importantly, the studies published until now were performed by priming CD4⁺ T cells either against major histocompatibility complex (MHC) class II–positive tumors²⁰  or against model antigens transfected into tumor cell lines such as ovalbumin (OVA).²⁸ 

To mimic a clinically closer situation, we immunized mice against one of the most important human TAA, EpCAM. The in vivo–primed T cells were used to generate EpCAM-reactive CD4⁺ T-cell lines in vitro for adoptive tumor therapy. The capacity of these T-cell lines to control the growth of EpCAM-transfected CT26 colon carcinoma cell lines (CT26-EpCAM) was analyzed. EpCAM-reactive CD4⁺ T cells from wild-type BALB/c mice inevitably developed an IL-4–producing Th0 or Th2 phenotype that provided no protection. Surprisingly, we were unable to generate EpCAM-reactive Th1 cells from wild-type BALB/c mice. In contrast, EpCAM-reactive Th1 cells were readily generated from IL-4.ko BALB/c mice. These Th1 cells established protective tumor immunity, even in naive BALB/c mice. Protective immunity induced intratumoral expression of antiangiogenic IFN-γ,^29,30  IL-17,³¹  and innate cytokines of the IL-12 family.^32-37  In consequence, tumor growth inhibition correlated with infiltration of CD11b⁺ monocytes/macrophages, T cells, and endothelial cell (EC) damage. Thus, EpCAM promotes tumor immune evasion through the induction of Th2-development.

Female wild-type BALB/c and DO11.10 TCR-transgenic mice were purchased from Charles River Laboratories (Wilmington, MA). IL-4.ko BALB/c mice were kindly provided by Dr M. Kopf (ETH, Zürich, Switzerland). The animals were maintained in the animal facilities of the Eberhard Karls University (Tübingen, Germany) under specific pathogen-free conditions. Animal experiments were approved by the Institutional Animal Care and Use Committee of the Regierungspräsidium Tübingen (HT01/03) and of the Regierung of Oberbayern (AZ 211-2531-63/98).

CT26-EpCAM cells were cultured in Dulbecco modified Eagle medium (DMEM; 4.5 g glucose) supplemented with 10% (vol/vol) fetal calf serum (FCS), 100 U/mL penicillin-G, and 100 μg/mL streptomycin. Isolated T cells were cultured in DMEM supplemented with 10% (vol/vol) FCS, 1% MEM amino acids (stock solution, 50×; Biochrom AG, Berlin, Germany), 10 μM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES) buffer, 1 mM sodium pyruvate, 100 U/mL penicillin-G, 100 μg/mL streptomycin, and 40 μM 2-mercaptoethanol. Cell lines were maintained at 37°C in a 7.5% CO₂ humidified incubator. CT26-EpCAM cells and the murine anti–IFN-γ (XMG-1.2) antibody for in vitro culture were from R.M. (Institut für Molekulare Immunologie [IMI], München, Germany). Murine anti–IL-4 (11B11) antibody was kindly supplied by Prof E. Schmidt (Institut für Molekulargenetik [IMGB], Mainz, Germany), and EpCAM protein (s17-1A) was friendly gifted by Prof E. Kopp (Institute of Immunology [IIM], Muenchen, Germany). OVA 323-339 peptide (ISQAVHAAHAEINEAGR) was purchased from EMC Microcollections (Tübingen, Germany). OVA protein was purchased from Sigma-Aldrich (St Louis, MO).

Either wild-type or IL-4.ko BALB/c mice were immunized twice subcutaneously with 10⁶ CT26-EpCAM cells dispended in 50 μL CpG-ODN1668 (200 μM, 5′-tccatgacgttcctgatgct-3′, phosphorothioate [PTO] oligonucleotide; MWG Biotech, Ebersberg, Germany). Wild-type BALB/c mice were immunized with 100 μg OVA protein suspended in water with 200 μM CpG-ODN1668. CD4⁺ T cells were isolated from spleen and lymph nodes (LN) 6 days after immunization using mouse anti-CD4 (L3T4) magnetic beads (MACS; Miltenyi Biotec, Gladbach, Germany). OVA- or EpCAM-reactive CD4⁺ T cells (1-5 × 10⁵) were stimulated with 10⁶ syngeneic and radiated (30 Gy) T cell–depleted splenocytes (antigen-presenting cells [APC]) 100 μg/mL OVA protein or 1 μg/mL EpCAM protein, and 0.2 μM CpG1668 and expanded for 14 days (short-term culture) with 50 U/mL IL-2 (Chiron Therapeutics, Emeryville, CA). For long-term culture, cells were restimulated with EpCAM, APCs, and CpG every 2 weeks.

For OVA T-cell culture, CD4⁺ cells were isolated from DO11.10 transgenic mice as described above. OVA CD4⁺ T cells (2.5 × 10⁵) were stimulated with 5 × 10⁵ APCs, 0.2 μM CpG1668, 5 μg/mL OVA peptide, and 2 μg/mL murine anti–IL-4 (11B11) antibody.

For survival studies, CT26-EpCAM tumor cells (10⁶) cell therapy was dispended in 100 μL PBS were injected intravenously in BALB/c mice. T-cell therapy was performed with intravenously injected (5 × 10⁵) in vitro–cultured CD4⁺ cells.

To analyze the mechanisms, CT26-EpCAM tumor cells (1.5 × 10⁶) dispended in 50 μL PBS were injected subcutaneously in the right and left flank of BALB/c mice (day 0). The mice were irradiated (2 Gy) on day 3 to induce a mild inflammatory response. On day 4, these mice received either cultured Th1 cells (5 × 10⁶) or PBS intraperitoneally. Tumors were measured with a sliding calliper on days 10, 13, and 17 after tumor cell injection, and tumor area was calculated as product of perpendicular tumor diameters. Mice were killed at the indicated time after injection, and tumors were harvested for further analysis.

In another approach, one group of mice received 500 μg murine anti–IFN-γ antibody (XMG 1.2) intraperitoneally after irradiation on day 3 and before adoptive transfer of T cells on day 4, as indicated.

For polyclonal T-cell stimulation, 10⁶ T cells were activated with 30 ng/mL phorbol-12-myristate-13-acetate (PMA), and 1.5 mg/mL ionomycin. Cells were treated with brefeldin A (concentration as instructed; BD Biosciences, San Jose, CA), fixed in 2% (vol/vol) formaldehyde in PBS, and permeabilized with 0.5% (wt/vol) saponin in flow cytometry buffer (0.5%, wt/vol, bovine serum albumin in PBS). Cells were stained for intracellular IFN-γ (XMG-1.2; BioLegend, San Diego, CA), IL-4 (11B11; BioLegend), IL-10 (JES5-16E3; BD Pharmingen, San Diego, CA), or IL-17 (TC11-18H10; BD Biosciences) in combination with surface CD4 (RM4-5; BioLegend) for fluorescence-activated cell-sorting (FACS) analysis. Cells were acquired on the FACSCalibur (BD Biosciences) and analyzed using CellQuest Pro software (BD Biosciences).

For characterization of tumor-infiltrating cells, tumors were excised from mice on the indicated days after injection, cut with Medimax (Günter Keul, Steinfurt, Germany), digested using collagenase D (Roche, Indianapolis, IN), and single cells were stained using the following anti–mouse antibodies: fluorescein isothiocyanate (FITC)–CD4 (RM4-5; BioLegend), phycoerythrin (PE)–CD8 (Ly.2; BD Pharmingen), FITC-CD11b (Mac-1; BD Pharmingen), allophycocyanin-CD11c (N418; Caltag Laboratories, Burlingame, CA), biotinylated anti–mouse F4/80 (BM8; BMA Biomedicals, Augst, Switzerland) with a secondary PE-labeled streptavidin antibody (BD Pharmingen), and the corresponding isotype rat IgG antibodies. Cells (30 000) were acquired on the LSRII (BD Biosciences) and analyzed using FCS Express software (De Novo Software, Los Angeles, CA). Total living cells were gated by forward/sideward scatter, and infiltrating cells were calculated on the basis of 10 000 living cells.

LN-derived T cells (0.1 × 10⁶ cells/well) were stimulated with: (1) APCs alone (0.5 × 10⁶ cells/well), (2) APCs and EpCAM (1 μg/mL), or (3) medium as negative control. Supernatants were then collected after 48 hours for further treatment. For detection of IFN-γ, IL-4, IL-10, and transforming growth factor-β (TGF-β) in supernatants of activated T cells and subsequent calculation of the protein content, sandwich enzyme-linked immunosorbent assays (ELISAs) were performed according to the manufacturers' instructions (IFN-γ and TGF-β; R&D Systems, Minneapolis, MN; and IL-4 and IL-10; BD Biosciences).

Total RNA was isolated from whole tumor tissue. Tissue was homogenized using the Mixer Mill (Retsch, Haan, Germany). RNA was purified according to the manufacturers' instructions (Macherey-Nagel, Düren, Germany) and converted to cDNA using Moloney murine leukemia virus (MMLV) reverse transcriptase (Promega, Madison, WI). Amplification was performed using the iCycler iQ instrument and the SYBR Green supermix (both from Bio-Rad Laboratories, Hercules, CA). For FoxP3, real-time polymerase chain reaction (PCR) was performed using the Light Cycler 405 and the SYBR Green mix (both from Roche). Specific primers (MWG Biotech) with the following sequences were used: hypoxanthine-guanine phosphoribosyltransferase (HPRT): 5′-GTTCTTTGCTGACCTGCTGGAT-3′ (forward), 5′-CTTAGGCTTTGTATTTGGCTTT-3′ (reverse); IL-23/p19: 5′-CATGGAGCAACTTCACACCTC-3′ (forward), 5′-GGTGATCCTCTGGCTGGA-3′ (reverse); IL-12/p35: 5′-ATGACCCTGTGCCTTGGTAG-3′ (forward), 5′-CAGATAGCCCATCACCCTGT-3′ (reverse); IFN-γ: 5′-TCAAGTGGCATAGATGTGGAAGAA-3′ (forward), 5′-TGGCTCTGCAGGATTTTCATG-3′ (reverse); intercellular adhesion molecule (ICAM): 5′-CCTGTTTCCTGCCTCTGAAG-3′ (forward), 5′-GTCTGCTGAGACCCCTCTTG-3′ (reverse); IL-6: 5′-CCGGAGAGGAGACTTCACAG-3′ (forward), 5′-CAGAATTGCCATTGCACAAC-3′ (reverse); IL-27/Ebi3: 5′-GCTCCCCTGGTTACACTGAA-3′ (forward), 5′-ACCGAGAAGCATGGCATT-3′ (reverse); IL-17: 5′-TCAGACTACCTCAACCGTTCC-3′ (forward), 5′-CTTTCCCTCCGCATTGACAC-3′ (reverse); CXCL10: 5′-TCCCTCTCGCAAGGACGGTC-3′ (forward), 5′-GTGTGTGCGTGGCTTCACTC-3′ (reverse); FoxP3: 5′-ACTCGCATGTTCGCCTACTT-3′ (forward), 5′-GTCCACACTGCTCCCTTCTC-3′ (reverse). Multiplex reactions were run in triplicates (for FoxP3, in duplicates), and samples were normalized to HPRT. The relative expression of genes of interest is represented as fold differences of the mean of treated tumor samples to the mean of control tumor samples. The expression of CXCL9, c-Myc, and CyclinD2 was measured semiquantitatively: CXCL9: 5′-TTTTCCTTTTGGGCATCATCTT-3′ (forward), 5′-AGCATCGTGCATTCCTTATCACT-3′ (reverse); c-Myc: 5′-AGCTGTTTGAAGGCTGGATT-3′ (forward), 5′-CCGCAACATAGGATGGAGAG-3′ (reverse); CyclinD2: 5′-ATTTCAAGTGCGTGCAGAAG-3′ (forward), 5′-ACACTTCTGTTCCTCACAGACCT-3′ (reverse).

Tumors obtained from mice at day 18 after injection were either fixed in 1% (wt/vol) paraformaldehyde (PFA), embedded in paraffin, and cut into 7-μm sections or were directly embedded in tissue-freezing medium (Tissue tec; Jung/Leica, Wetzlar, Germany), snap-frozen, and cut in 10-μm cryosections. Paraffin sections were immunostained with anti–mouse Ki-67 (0.3 mg/mL; Dako, Glostrup, Denmark) and acetone-fixed cryosections with anti–mouse CD4 (0.01 mg/mL; BD Pharmingen). Positive immunostaining was detected with a streptavidin-biotin immunoperoxidase system (Vector Laboratories, Burlingame, CA). Sections were counterstained with hematoxylin (Chem Mate; Dako). Immunofluorescence staining of acetone-fixed cryosections was performed with anti–platelet/endothelial cell adhesion molecule-1 (anti–PECAM-1; BD Pharmingen) or anti–vascular endothelial growth factor receptor-2 (anti–VEGFR-2) antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and a secondary Cy3-labeled antibody (Molecular Probes, Invitrogen, Carlsbad, CA). Sections were counterstained with Yopro1 (Molecular Probes, Invitrogen). Isotype-matched biotinylated rat IgG served as negative control in all stainings. For routine histology analysis, acetone-fixed cryosections were stained with hematoxylin and eosin (H&E). Digital images at the magnification indicated were obtained using an Axiovert 200 fluorescence microscope (Carl Zeiss, Jena, Germany).

Tumor samples for electron microscopy (EM) analysis were cut into small pieces and immediately fixed by immersion in 1.5% (wt/vol) glutaraldehyde in 0.1 M cacodylate buffer with 5% (wt/vol) sucrose for 2 to 4 hours. Specimens were postfixed in cacodylate buffer containing 1% (wt/vol) OsO₄ for 1 hour and dehydrated in ascending series of ethanol and propyleneoxide. For contrast enhancement, they were block-stained in uranyl-acetate in 70% ethanol for 4 hours and flat-embedded in Araldite (Serva, Heidelberg, Germany). Using an ultramicrotome (Ultracut R; Leica) semi-thin sections (1 μm) and ultrathin sections (50 nm) were cut. Ultrathin sections were stained with lead citrate, mounted on copper grids, and finally analyzed with a Zeiss EM 10 electron microscope (Carl Zeiss).

Statistical analysis was performed using an unpaired Mann-Whitney test using SSPS software (SSPS, Chicago, IL). P values less than .05 were considered statistically significant.

Adoptive transfer experiments with TCR-transgenic T cells that recognize artificial TAAs from transfected tumor cell lines show that CD4⁺ T cells frequently provide more efficient protection than CD8⁺ T cells.^20,23,24  We therefore tried to raise Th1 cells against EpCAM, a TAA relevant for tumor immune therapy in humans. We immunized BALB/c mice with CT26-EpCAM cells. Tumor cells were coinjected with CpG-ODN 1668 to obtain optimal in vivo priming for TAA-reactive Th1 cells. The concentration of CpG-ODN 1668 was sufficient to induce Th1 responses against Leishmania major in BALB/c mice^27,38  or conventional peptide antigens.³⁹  One week after the last boost, we isolated CD4⁺ T cells from spleen and LN and stimulated the T cells in vitro again with CpG-ODN 1668, 10 μg/mL 11B11 anti–IL-4 monoclonal antibody (mAb), APCs, and EpCAM protein. Surprisingly, developing CD4⁺ T cell lines generally resulted in either IL-4 producing Th2 or Th0 cells, with 27% of CD4⁺ cells producing IL-4 and 44% of CD4⁺ cells producing IFN-γ. Such Th2 or Th0 cells were unable to protect against intravenously injected CT26-EpCAM tumor cells (Figure 1A). Thus, the human TAA EpCAM did override the most stringent Th1 inducing signals, even the combination of anti–IL-4 mAb and CpG-ODN 1668. This strong Th2 driving force was unexpected and seems to be an unusual feature of this TAA. When T cells were primed under identical conditions with OVA, a protein that normally induces polarized Th2 responses in BALB/c mice,⁴⁰  we obtained classical Th1 cells that were again approximately 40% IFN-γ⁺, but less than 5% were positive for IL-4.

Interestingly, when we extended T-cell cultures directed against CT26-EpCAM for another 3 to 4 rounds of stimulation and expansion, cells progressively developed toward a Th1 phenotype with 86% of CD4⁺ cells producing IFN-γ and 1.5% of CD4⁺ cells producing IL-4. Simultaneously, these Th1 cells acquired the capacity of protecting against CT26-EpCAM tumors (Figure 1B) or EpCAM-expressing A20 lymphomas (Figure 1C). Yet, they failed to protect against A20 wild-type lymphomas, showing that the Th1 cell response was specific for the tumor antigen (Figure 1C).

To determine whether the IL-4 production was responsible for the lack of protection by short-term cultured T cells or whether it resulted from phenomena such as insufficient selection of T cells expressing TCR with high affinity for EpCAM, we generated Th1 cells from IL-4.ko BALB/c mice using the identical procedure. Priming and expansion of IL-4.ko T cells resulted in both Th1 and Th17 cells, 30% of the cells producing IFN-γ and 15% to 20% producing IL-17 (Figure 2A). In contrast to CD4 T cells from wild-type mice, such short-term cultured EpCAM-reactive Th1/Th17 cells provided solid protection against CT26-EpCAM tumors (Figure 2B), suggesting that IL-4 was responsible for the lack of protection. Therefore, Th1/Th17 cells were further generated from IL-4.ko mice. This also precludes that the tumor established an environment inaccessible to protective immune responses. Moreover, when analyzed for the release of immunosuppressive cytokines, CT26 cells produced no soluble TGF-β, and IL-10 production by CT26-EpCAM cells was lower than IL-10 production by A20-EpCAM lymphomas (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article).

Tumor rejection by Th1/Th17 cells may cause epitope spreading and thus newly prime naive T cells against glycoproteins that are expressed by the target tissue. To determine the effects of tumor rejection by EpCAM-reactive Th1/Th17 cells on the establishment of tumor immunity in wild-type mice, mice were rescued from CT26-EpCAM tumors by treatment with EpCAM-reactive Th1/Th17 cells from IL-4.ko mice. Subsequently, mice were challenged with a second tumor on day 100 after original tumor challenge. Surviving mice received one of the following tumors: CT26-EpCAM or A20-EpCAM to analyze reactivity against EpCAM, or parental CT26 tumors or A20 lymphomas to determine specificity. After rejection of the original CT26-EpCAM carcinoma, 100% of mice were protected against CT26-EpCAM carcinoma and 70% against A20-EpCAM lymphoma (Figure 2C). Thus, Th1 cells from IL-4.ko mice established a solid and persistent Th1 memory response against EpCAM-expressing tumors that resisted the Th2-driving properties of the EpCAM protein.

Importantly, the tumor responses were still highly specific, as all mice receiving the parental A20 lymphomas ultimately died (Figure 2C). In sharp contrast, mice challenged with the parental CT26 carcinoma were largely protected. These data suggest that rejection of CT26-EpCAM tumors resulted in functional relevant epitope spreading that may have contributed to the protective immunity against the parental CT26 carcinoma cell line.

To investigate the therapeutic mechanisms underlying the Th1 cell-mediated immunotherapy, we used a subcutaneous CT26-EpCAM tumor model, as this allows direct analysis of both the tumor and the inflammatory environment. We subcutaneously injected 1.5 × 10⁶ CT26-EpCAM cells into the right and left flank of syngeneic BALB/c mice on day 0. On day 4, mice received either 5 × 10⁶ EpCAM-reactive Th1 cells or PBS (control). To determine the impact of these EpCAM-reactive Th1 cells on CT26-EpCAM tumor growth, we measured the tumor size and calculated the tumor growth (Figure 3A). Until day 10, the Th1 cell–treated group and the PBS-treated group showed similar tumor growth dynamics. From day 13 on, the Th1 cell–treated group showed a significant inhibition of tumor growth (Figure 3A).

On day 18 or the indicated time, we determined tumor weight and analyzed in these tumors cytokine expression profiles, growth pattern by histology, and ultrastructure by EM.

Adoptive Th1-cell transfer inhibited tumor growth that was first detectable on day 13 (Figure 3A,B) and reached up to 84% reduction of tumor weight by day 18 (Figure 3C). Importantly, only EpCAM-reactive Th1 cells provided protection, while BALB/c mice receiving 5 × 10⁶ OVA-reactive Th1 cells remained fully susceptible to CT26-EpCAM cells (Figure 3D). As expected from these findings, quantitative mRNA analysis showed no significant increase of FoxP3 mRNA inside Th1 cell–treated tumors (Figure S2).

Quantitative mRNA analysis of tumor samples revealed that the onset of tumor growth inhibition on day 13 coincided with the strongest increase in cytokine expression in tumors of Th1 cell–treated mice as compared with controls (Figure 4A). Th1/Th17 cell–associated genes, namely IL-23/p19, IL-12/p35, IFN-γ, IL-27/Ebi3, IL-6, and ICAM were 3- to 10-fold up-regulated specifically in tumors of Th1 cell–treated mice (Figure 4A), with the most prominent increase for IL-23/p19. Surprisingly, we observed no obvious difference neither for the IFN-γ–induced chemokines, CXCL9 and CXCL10, nor for the IFN-γ–sensitive regulators of cell proliferation and differentiation, c-Myc or cyclin D2. Moreover, angiogenesis-associated genes such as VEGF, placenta growth factor (PIGF), angiopoietin-1 (Ang-1), and Ang-2 were poorly regulated (data not shown). As up-regulation of IL-23 and IL-6 in tumors of Th1 cell–treated mice should provide an environment favorable for the survival and proliferation of IL-17–producing cells,⁴¹  we analyzed the expression of IL-17 in these tumors. In agreement with this hypothesis, expression of IL-17A mRNA was highest (> 10-fold) again on day 13, the time when IL-23/p19 and IL-6 peaked (Figure 4B). These data suggest that infiltrating Th1 and Th17 cells might act in concert with cytokines of the IL-12 family to exert relevant tumor control, especially as IFN-γ and IL-12 are both well known to exert strong inhibition of tumor growth, in vitro and in vivo.^28-30,35 

To prevent activation-associated death of the transferred Th1 cells from IL-4.ko mice, we treated the tumor-bearing mice with 500 μg anti–IFN-γ mAb, 1 day before the adoptive transfer of EpCAM-reactive Th1 cells.⁴²  If the immune-mediated tumor rejection relies on the recipient's T cells, such a treatment should promote Th2 development and abrogate tumor protection by the Th1 cells. In contrast, if the therapy primarily relies on the transferred Th1 cells from IL-4.ko mice, the anti–IFN-γ mAb should protect those transferred Th1 cells from suicide and thus strengthen the protective effect against the tumor.⁴²  Indeed, on day 17, inhibition of tumor growth was even more efficient in mice where activation-associated cell death of Th1 cells was prevented by anti–IFN-γ mAb (Figure 5A) than in mice treated with Th1 cells alone. Reduced tumor growth in anti–IFN-γ–treated mice was associated with a 3-fold stronger up-regulation of intratumoral IFN-γ mRNA as compared with mice receiving Th1 cells only (Figure 5B). Most other cytokines remained comparable between the group receiving both, Th1 cells and anti–IFN-γ mAb, and the mice treated with Th1 cells only (Figure 5B).

As inflammatory cytokines were up-regulated in tumors of mice treated with Th1 cells, we investigated the potential source of these cytokines by characterizing the cellular infiltrate using FACS analysis and immunohistochemistry. Analyzing tumors of Th1 cell–treated mice showed a strong increase in homogeneously distributed CD4⁺ T cells (Figure 6Aii), peaking on day 18 with a 16-fold increase (Table 1). In contrast, tumors of the control group had only few CD4⁺ T cells (Figure 6Ai and Table 1). Besides CD4⁺ T cells, FACS analysis showed a relative 3-fold increase of CD11b⁺ cells in tumors of Th1 cell–treated mice at day 18. F4/80⁺ macrophages or CD11c⁺ infiltrating DCs were not detectable in any of the analyzed tumors (Table 1). These data suggest that infiltrating CD4⁺ T cells and cells of the monocyte/macrophage lineage were involved in arresting tumor growth. We found no evidence for infiltrating CD8⁺ T cells in the Th1 cell–treated mice. On day 18, we even detected 5 times more CD8⁺ T cells in tumors of untreated mice than in tumors of Th1 cell–treated mice (Table 1). Despite the significant difference in tumor size, we found only minor differences in tumor morphology (Figure 6Aiii,iv) or proliferation markers such as the proliferation-associated nuclear antigen Ki-67 (Figure 6Av,vi,B).

As the antiangiogenic cytokines IL-12, IFN-γ, and IL-27 were up-regulated in tumors of Th1 cell-treated mice and as tumor growth depends on neovessel formation, we investigated the effects of Th1 cell therapy on tumor angiogenesis. Immunohistologic staining with a mAb specific for the endothelial antigen PECAM-1 showed no gross differences between the 2 groups (Figure 6Ci-iv). However, changes in vessel morphology, such as clumpy and curved structures, appeared in tumors of Th1 cell–treated mice (Figure 6Cii,iv arrows). The expression of the most important receptor of angiogenesis, VEGFR-2, was not affected by the Th1-cell therapy (Figure 6Cv,vi).

We therefore performed EM to study ultrastructural changes. Tumors of Th1 cell–treated mice revealed major EC damage such as cell death, cytoplasma-enriched endothelia, and vessel obliteration (Figure 6Diii,v), demonstrating severe impairment of tumor vessels in Th1 cell–treated mice. This impaired vessel structure was associated with enhanced diapedesis as sign of inflammation (Figure 6Diii,iv). Tumors of PBS-treated mice showed a vascular bed typical for normal tumor tissue (Figure 6Di,ii).⁴³  Vessels of adjacent muscle tissue remained unchanged in Th1 cell–treated mice (Figure 6Dvi), showing that the Th1 cell–mediated effects on the vasculature were restricted to the tumor tissue.

Currently, various immune therapies are developed against solid human cancer. While antibody-based immune therapies show first success in humans, the therapeutic success of T cell–based clinical trials, such as active immunization or adoptive T-cell transfer, was only successful in single patients.²²  In some trials, tumor immune therapy did not only fail to establish protection but even promoted tumor development.^17,44  The mechanisms underlying these failures remain enigmatic. Indirect data from humans suggest that protective immunity was associated with the development of a Th1 immunity, while tumor promotion occurred in patients that developed a Th2 immunity.^17,44  These findings are in agreement with various experimental data from mice showing that only Th1 cells are capable of providing protective tumor immunity, while Th2 responses fail to protect.^20,24,29 

Using a clinically relevant model antigen, EpCAM, here we were unable to establish protective tumor immunity in normal BALB/c mice, even in the presence of the TLR9-ligand CpG-ODN and anti–IL-4 mAb.⁴⁵  Characterizing these inefficient immune responses, we found that EpCAM-reactive Th cells inevitably developed an IL-4–producing Th2 or Th0 phenotype, even under most stringent Th1 cell-inducing conditions. As IL-4 is required for the differentiation of naive Th cells toward a Th2 phenotype during the period of T-cell priming, the combination of anti–IL-4 mAb and CpG-ODN is most efficient in preventing Th-2 cell differentiation under most conditions, even in vivo.^38,45  Besides IL-4, specific sets of accessory molecules such as Jagged, certain C-type lectin receptors, parasite-derived proteoglycans, or eosinophil-derived neurotoxin⁴⁶  can promote Th2 differentiation. Such stimuli may either activate GATA3 through alternative activation of STAT6 or other signaling pathways that involve modulation of the TCR-signaling pathway.⁴⁷  The mechanisms underlying the potent Th2-inducing properties of EpCAM remain enigmatic.

To determine whether the failure of raising protective immune responses against CT26-EpCAM was secondary to the Th2/Th0 phenotype of the transferred T cells, we established EpCAM-reactive Th1 cells from syngeneic IL-4.ko BALB/c mice. Using the same protocol of in vivo priming followed by one round of in vitro stimulation, we generated EpCAM-reactive, IL-4–deficient Th1-cell lines. Surprisingly, these Th1 cells did not only specifically protect naive BALB/c mice against EpCAM-expressing tumors, but even established protective immunity against subsequent tumor challenges in normal BALB/c mice. This proves that EpCAM-induced IL-4 directly abrogated the capacity of CD4⁺ T cells to establish protective tumor immunity. Moreover, the data underlined the high therapeutic efficacy of adoptively transferred Th1 cells, even against tumors that normally induce tumor immune evasion through the induction of Th2 responses.

EpCAM-reactive Th1 cells protected against both intravenously and subcutaneously applied carcinomas. To analyze the mechanisms underlying this Th1 cell–mediated control of EpCAM-expressing, MHC class II–negative carcinoma cells, we focused on subcutaneously injected tumors. This model provides direct access to the tumor and thus allows to investigate the tumor-host interactions over time. In agreement with previous reports,^24,29  we found that the Th1-cell therapy induced the expression of antiangiogenic factors, while we found no evidence for a significant suppression of VEGF. The Th1-cell therapy strongly induced the expression of proinflammatory and antiangiogenic cytokines, namely IL-12, IL-23, IL-27, and IFN-γ. Expression of all 4 cytokines peaked exactly at the initial inhibition of tumor growth. Consistent with these findings, we found a 3-fold increase of CD11b⁺ monocytes/macrophages in tumors of Th1 cell–treated mice. Tumor-infiltrating macrophages that are activated by IFN-γ–producing Th1 cells (M1 phenotype) possess strong cytotoxic properties against tumor cells and produce proinflammatory cytokines such as IL-23 and IL-12.³⁵  While the antiangiogenic role of IFN-γ and IL-12 is well established,^29,30,32,33  the role of IL-23 in either promoting or inhibiting tumor development is ambiguous. While transfection of tumor cells with IL-23 or systemic application of IL-23 promote tumor rejection and establish protective immunity,^36,37  IL-23.ko mice are resistant to chemically induced tumor induction.⁴⁸  This role of IL-23 in antitumor immunity is reminiscent of the double-edged role of tumor necrosis factor (TNF).²⁴ 

Analyzing the tumor microenvironment at various times of tumor growth inhibition, we only found infiltrating CD4⁺ T cells but no evidence for CTL-mediated tumor cell killing or induction of regulatory T cells. In 2 additional model tumors, transplanted A20 lymphomas and endogenously growing islet carcinomas, we extensively investigated the role of CD8⁺ T cells and CTL. Both experiments showed, that the therapeutic outcome of the adoptive Th1-cell transfer was not attenuated by the depletion of CD8⁺ T cells.^20,24  Th1-cell transfer differs from conditions of active immunization.¹⁵ 

As tumors of either untreated or treated mice showed equal numbers of Ki-67–positive cells, it is unlikely that the Th1 cells directly inhibited the proliferation of tumor cells.²⁴  The intra-tumoral cytokine expression at the time of growth arrest together with the EC damage uncovered by EM strongly suggest that the antiangiogenic properties of Th1 cells significantly contributed to the T cell–mediated inhibition of tumor growth. EM revealed EC death, and cytoplasma-enriched ECs were frequent in tumors of Th1 cell–treated mice. Importantly, ECs in adjacent muscle tissue were not affected, showing that the antiangiogenic properties of the Th1 cell therapy were restricted to the tumor tissue. Together, the data suggest that the combined action of IFN-γ, IL-12, and IL-27 inhibited tumor growth by severe EC damage. Th1 cells can impair tumor angiogenesis either directly by inhibiting EC proliferation through IFN-γ or indirectly through induction of antiangiogenic chemokines.²⁴  Similar to the model described by Qin and Blankenstein,²⁹  IFN-γ seems to directly inhibit tumor angiogenesis in this model. Importantly, this may also be relevant for the human situation, where adoptive transfer of a Th1-cell clone resulted in the rapid involution of metastases without causing a cytokine release syndrome.⁴⁹ 

As prevention of activation-associated Th1-cell death further enhanced the efficiency of adoptive T-cell therapy with Th1 cells, the findings reported here strongly support the increasing importance of TAA-reactive CD4⁺ T cells in tumor therapy. The different tumor models using either injected tumor cell lines or endogenously growing tumors suggest that Th1 cells may inhibit tumor development and seeding of metastases through different modes of action. Yet, all studies unanimously show that proper Th1-cell differentiation, ie, a strong IFN-γ production and the absence of IL-4, is crucial for the establishment of protective tumor immune responses. Thus, human tumor vaccine approaches should focus on the use of TAA that allow appropriate induction of tumor-reactive Th1 responses.

The skillful help Gabi Frommer-Kästle with electron microscopy is gratefully acknowledged. We thank G. Riethmüller, P. Kufer, and E. Kopp for helpful discussions and providing EpCAM.

This work was supported by the Deutsche Forschungsgemeinschaft (SFB 685), the Wilhelm Sander-Stiftung (2005.043), the Deutsche Krebshilfe (10 7128), and the Comprehensive Cancer Center Tübingen.

Contribution: A.Z., R.H., and H.B. performed and analyzed the experiments and prepared the manuscript; S.W. technically supported the performance of the experiments; H.W. performed and analyzed the electron microscopy; R.M. and M.R. designed the experiments, interpreted the data, and prepared the manuscript; and all authors checked the final version of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Martin Röcken, Department of Dermatology, University Medical Center, Eberhard Karls University Tübingen, Liebermeisterstr 25, 72076 Tübingen, Germany; e-mail: mrocken@med.uni-tuebingen.de; or Ralph Mocikat, Institut für Molekulare Immunologie, Helmholtz-Zentrum München, Marcchioninistr 25, 81377 München, Germany; e-mail: mocikat@helmholtz-muenchen.de.