Decreased Tumor Surveillance after Adoptive T-Cell Therapy

The effect of cancer immunotherapy on the endogenous immune response against tumors is largely unknown. Therefore, we studied immune responses against murine tumors expressing the glycoprotein (GP) and/or nucleoprotein of lymphocytic choriomeningitis virus (LCMV) with or without adoptive T-cell therapy. In nontreated animals, CTLs specific for different epitopes as well as LCMV-GP–specific antibodies contributed to tumor surveillance. Adoptive immunotherapy with monoclonal CTLs specific for LCMV-gp33 impaired the endogenous tumor-specific antibody and CTL response by targeting antigen cross-presenting cells. As a consequence and in contrast to expectations, immunotherapy enhanced tumor growth. Thus, for certain immunogenic tumors, a reduction of tumor-specific Band T-cell responses and enhanced tumor growth may be an unwanted consequence of adoptive immunotherapy. [Cancer Res 2007;67(15):7467–76]


Introduction
Adoptive transfer of large numbers of tumor-specific CTLs (10 9 -10 11 cells per infusion) is currently being developed to treat cancer by changing the relative balance between tumor load and the immune response (1,2).Clinically relevant responses after adoptive immunotherapy have resulted in disease stabilization or even in regression of solid tumors (1)(2)(3).These tumor responses are the proof of principle that adoptive immunotherapy in cancer patients may be efficacious.However, in a large proportion of patients, tumors progress despite adoptive immunotherapy (2)(3)(4)(5).Loss of antigen or MHC class I expression on tumor cells, induction of T-cell anergy, and a limited survival of the transferred CTLs are well defined mechanisms that explain the lack of therapeutic efficacy in some situations (6)(7)(8)(9).In addition, the adoptive transfer of high numbers of tumor-specific CTLs may also influence the endogenous tumor-specific immune response, but this has not been analyzed yet.
The idea that the immune system is involved in the control of tumors dates back to the 1950s when Burnet and Thomson formulated the immunosurveillance hypothesis (10).More recent experiments analyzing the spontaneous tumor development showed a role of the adaptive immune system and of IFNg in the control of solid tumor formation (11).Although various effector mechanisms of the immune system are involved in tumor control, most experimental data support the notion that eradication of solid tumors rests predominantly on CD8 + T lymphocyte activity (3,12).Tumor-specific CTLs are induced directly either by tumor cells that migrate to secondary lymphoid organs or by dendritic cells cross-presenting tumor antigens on MHC class I molecules (13)(14)(15).Both mechanisms may be inefficient during early tumor development; therefore, antigenic solid tumors may grow outside secondary lymphoid organs due to immunologic ignorance (16,17).
In the present study, we analyzed the immunologic control of lymphocytic choriomeningitis virus (LCMV)-glycoprotein (GP)expressing and/or LCMV-nucleoprotein (NP)-expressing tumor cells and found that antibodies and CTLs are involved in the control of tumors transplanted as solid fragments.We therefore studied the effects of adoptively transferred monoclonal LCMV-gp33-specific CTLs (P14) on tumor-specific immune responses and tumor growth.Adoptive immunotherapy using P14 cells was therapeutically efficacious against tumors when applied to tumorbearing immunodeficient RAG-1 À/À mice.In contrast, adoptive immunotherapy increased tumor growth in immunocompetent C57BL/6 (BL/6) mice by reducing the endogenous tumor-specific immune response.Our results suggest that in situations where a polyspecific endogenous immune response is necessary for tumor control, adoptive immunotherapy against a single epitope of a tumor antigen may enhance tumor progression.
Cell lines and constructs.B16-gp33 cells have been described (21).MC57G (MC; ref. 13), B16.F10 (B16; ref. 21), and D2 cell lines (14) were transfected either with a plasmid expressing a codon-optimized LCMV-GP open reading frame (ORF) under control of elongation factor 1a promoter (M369, -GP i+e ).The codon usage of the LCMV-GP was optimized for highlevel expression in mammalian cells and the corresponding cDNA synthesized by the company GeneArt.The construct has been deposited under DQ886924 at Genbank.Alternatively, an otherwise identical construct was used containing a leucine to proline mutation at position 110 of the LCMV-GP ORF (pDP12,-GP i ).pDP12 was generated with standard two-way PCR-based site-directed mutagenesis using M369 as a template.Gene transfection was carried out using Superfect (Qiagen).
Tumor growth in vivo.Tumor cells (2 Â 10 6 ) were injected s.c. in the flank of RAG-1 À/À mice.Solid tumor fragments were explanted, cut into pieces of 2 Â 2 Â 2 mm, and transplanted s.c. in the flank of recipient mice.Growth was measured with a caliper and the volume was calculated by the formula V = p Â abc / 6, where a, b, and c are orthogonal diameters.Singlecell suspensions ex vivo were prepared by cutting solid tumors into small pieces and smashing through a stainless steel grid.The suspension was then depleted from cell clumps by a quick-spin centrifugation.Single-cell suspensions of in vitro cultures were generated after incubation in nonenzymatic cell dissociation solution (Sigma-Aldrich).
Activation of P14 cells.In vitro activated P14 cells were generated as described (21) and purified for CD8 + T cells using MACS (Miltenyi Biotec).P14 CTLs were activated in vivo by injecting 10 6 naive P14 splenocytes to BL/6 mice i.v.followed by an infection with 10 4 plaque-forming units LCMV-WE.Eight days after, infection-activated P14 CTLs were used directly in a 51 Cr release assay.
Dendritic cells.Dendritic cells were generated as described (22).For immunization, matured dendritic cells were labeled with gp33 and/or np396 peptides (10 À6 mol/L) for 1.5 h at 37jC.Pulsed dendritic cells (2 Â 10 5 ) were injected i.p. or into foot pad of BL/6 mice.In certain experiments, peptide-pulsed dendritic cells were labeled with 1 Amol/L CFSE in PBS for 10 min at 37jC at a cell concentration of 2 Â 10 6 cells/mL.

51
Cr release assay and intracellular IFN; staining assay.Primary 51 Cr release assays without restimulation or secondary 51 Cr release assays after in vitro restimulation for 5 days were done as described earlier (21,23).For intracellular staining, lymphocytes (10 6 per well) were stimulated for 5 h with the relevant peptide (10 À6 mol/L/well) in the presence of 5 Ag/mL brefeldin A (Sigma-Aldrich) and recombinant 25 units/mL interleukin-2 (IL-2; Sigma-Aldrich) in 96-well round-bottomed plates (Carl Roth).Cells were stained for surface molecules and fixed with 4% paraformaldehyde and cell membranes were permeabilized with Perm buffer (PBS, 2% FCS, 5 mmol/L EDTA, 0.1% saponin, 0.2% NaN 3 ) and stained with aIFNg-FITC.Relative fluorescence intensities were measured using a flow cytometer.Nonpeptide-pulsed cultures or nontransfected tumor cell lines served as controls in all experiments.These background values are indicated in each plot in brackets.
Antibodies and flow cytometry.Antibodies were from eBioscience except goat anti-mouse IgG, which was from Caltag.Mouse anti-LCMV-GP monoclonal antibody (mAb; KL25) was produced as described (24).Relative fluorescence intensities were measured on a FACScan or BD LSR II (BD Bioscience) and analyzed using FlowJo software (Tree Star).
Depletion of CD8 T cells.Mice were treated i.p. on days À3 and À1 before transplantation and on days 7 and 14 after transplantation with 100 Ag aCD8 mAb (YTS 169.4).Efficiency of depletion was verified by flow cytometry.
GP-1-specific serum IgG ELISA.Detection of LCMV-WE GP-1-specific serum Ig was carried out using recombinant GP-1-IgG fusion protein in a standard ELISA (25).
Statistical analysis.Statistical significance was determined by Students t test, unpaired.P < 0.05 was considered significant.

Results
Endogenous immune response to LCMV-GP-expressing tumors.Fibrosarcoma (MC and D2) and melanoma (B16) cell lines were transfected with a construct expressing the entire LCMV-GP (MC-GP i+e , B16-GP i+e , and D2-GP i+e ).These tumor cell transfectants express the LCMV-GP intracellularly and extracellularly on the cell surface (i+e).A codon-optimized ORF under control of elongation factor 1a was used for high protein expression on the cell surface and peptide presentation on MHC class I molecule.It has been shown before that a leucine to proline substitution at amino acid 110 of the LCMV-GP prevents expression of LCMV-GP at the cell surface and the protein is only expressed intracellularly (i) without altering the amino acid sequence of any of the known CTL and T helper epitopes of LCMV-GP (26,27).Therefore, site-directed mutagenesis was used to change the leucin to proline at position 110 in a second construct.This mutated construct was similarly used to generate stable MC and B16 tumor cell transfectants (MC-GP I and B16-GP i ).LCMV-GP protein expression at the cell surface was tested by flow cytometry with the mAb KL25 (Fig. 1A).LCMV-GP peptide presentation on MHC class I molecule was tested by analyzing intracellular IFNg production of LCMV-immune splenocytes after in vitro stimulation with the relevant tumor cell line (Fig. 1A).
We transplanted MC-GP i+e and B16-GP i+e tumor fragments of f2 Â 10 6 to 5 Â 10 6 tumor cells s.c.into the flank of BL/6 mice.MC cells transfected with the empty puromycin resistance vector (MC-vector) or parental B16 tumor cells served as controls.We then analyzed tumor growth, LCMV-GP-specific CTL, and antibody induction.MC-GP i+e tumor fragments did not grow in immunocompetent BL/6 mice, whereas fragments of MC-vector transfectants grew (Fig. 1B).Similarly, growth of B16-GP i+e tumors was delayed when compared with nontransfected parental B16 tumors (Fig. 1C).Both MC-GP i+e and B16-GP i+e tumors efficiently induced LCMV-gp33-specific CTLs in BL/6 mice (Fig. 1B and C).Moreover, CTL responses against LCMV-gp276 were induced (Fig. 1B).In addition, LCMV-GP-binding antibodies were mounted in MC-GP i+e and B16-GP i+e tumor-bearing mice (Fig. 1B and C).
Earlier experiments using MC LCMV-GP transfectants with a nonoptimized plasmid under a cytomegalovirus promoter indicated that solid tumors grew in the absence of a CTL response due to inefficient cross-presentation of the antigen by professional antigen-presenting cells (APC; refs.13,14,16).Our experiments now show that MC and B16 transfectants with our novel optimized LCMV-GP construct efficiently primed CTLs, even if tumors were transplanted as solid fragments (Fig. 1B and C).We therefore analyzed if tumor cell transfectants with the optimized LCMV-GP construct are able to induce LCMV-GP-specific CTLs via crosspriming.We tested cross-priming in (BALB/c Â BL/6) F 1 mice immunized with H-2 d -positive D2 fibrosarcoma cells transfected with LCMV-GP.D2-GP i+e (H-2 d ) cells injected as single-cell suspension i.p. induced a potent H-2 b -gp33-restricted CTL response in (H-2 d Â H-2 b ) F 1 mice, indicating that LCMV-gp33 is efficiently cross-presented on MHC class I (Fig. 1D).Similarly, the transplantation of D2-GP i+e tumor fragments to (BALB/c Â BL/6) F 1 recipient mice induced gp33-restricted CTLs (data not shown).
Taken together, our results indicate that LCMV-GP-transfected B16 and MC tumors efficiently induce specific antibodies and CTLs that may be involved in the control of tumor development.
Immunologic control of MC-GP i+e and B16-GP i+e tumor formation.The immunologic effector mechanisms involved in tumor control of MC-GP i+e tumors were analyzed after transplantation to RAG-1 À/À mice, BL/6 mice either untreated or depleted of CD8 + T cells, and B-cell-deficient J H À/À mice either left untreated or depleted of CD8 + T cells (Supplementary Fig. S1A).MC-GP i+e tumors regularly grew in the absence of B and T cells in RAG-1 À/À mice but not in immunocompetent BL/6 mice.Fifty percent of tumors grew in CD8 + T-cell-depleted BL/6 mice but tumor growth kinetics were slower than in RAG-1 À/À mice.MC-GP i+e tumors did not grow in J H À/À mice.However, 100% of tumors grew in CD8 + T-cell-depleted J H À/À mice with growth kinetics comparable with tumors transplanted to RAG-1 À/À mice.Similarly, growth of B16-GP i+e tumors was enhanced in RAG-1 À/À mice (Supplementary Fig. S1B).Growth of B16-GP i+e tumors was slightly enhanced in J H À/À mice when compared with BL/6 mice.B16-GP i+e tumors grew faster in CD8 + T-cell-depleted BL/6 mice and the most rapid tumor formation was observed in CD8 + T-cell-depleted J H À/À mice (Supplementary Fig. S1B ).
Therefore, specific CTLs and antibody-producing B cells are involved in the control of MC-GP i+e and B16-GP i+e tumors.To analyze if tumor rejection is predominantly mediated by CTL responses against the immunodominant epitope gp33 or if other CTL specificities are also involved in tumor control, tumors were transplanted to H8 mice.These mice ubiquitously express a COOHterminal truncated version of LCMV-GP (amino acids 1-50) and gp33-specific CTLs are absent due to thymic deletion (18).All MC-GP i+e tumor fragments were rejected when transplanted to H8 mice, indicating that other CTL specificities against LCMV-GP than gp33 were sufficient for tumor rejection (Supplementary Fig. S1A).
Adoptive immunotherapy in the presence or absence of endogenous immune control of tumors.Recent experiments have suggested that due to more efficient in vivo activation, proliferation, and survival, transferred naive, early effector, or central memory CTLs are most efficient in tumor control (28)(29)(30).We therefore first studied the consequences of adoptively transferred naive P14 CTLs on tumor growth.MC-GP i+e tumors regularly grew when transplanted to RAG-1 À/À mice.The adoptive transfer of 10 7 splenocytes from naive P14 mice containing f10 6 LCMV-gp33specific CTLs at the day of tumor transplantation to RAG-1 À/À mice delayed or even prevented tumor growth (Fig. 2A).Surprisingly, if naive P14 splenocytes were transferred to BL/6 mice at the day of MC-GP i+e tumor transplantation, most transplanted tumors grew (Fig. 2A).Titration experiments revealed that MC-GP i+e tumor growth was observed in 7 of 10 tumors when 10 7 naive P14 splenocytes were transferred, in 3 of 6 tumors with 10 6 P14 splenocytes, and in 0 of 6 tumors when 10 5 P14 splenocytes were transferred (Fig. 2A).Ten of 12 transplanted MC-GP i+e tumors grew when transplanted s.c. to naive P14 mice (Fig. 2B).In contrast, if a comparable number of MC-GP i+e tumor cells (2 Â 10 6 ) was injected s.c. as single-cell suspensions in BL/6 or P14 mice, tumor cells were regularly controlled by the immune system and no tumor formation was observed (Fig. 2B; data not shown).Adoptive immunotherapy with tumor-specific CTLs has been reported to be most efficient in lymphopenic hosts, mainly due to homeostatic proliferation of tumor-specific CTLs (28).To test if P14 CTLs are able to control MC-GP i+e tumors in the absence of homeostatic proliferation, adoptive transfer of naive P14 CTLs was done after transplantation of MC-GP i+e tumor fragments to RAG-1 À/À Â OT-1 mice.In these mice, the CD8 + T-cell compartment is filled by ovalbumin-specific CTLs.It has been shown before that homeostatic proliferation of adoptively transferred TCR transgenic T cells is abrogated in RAG-1 À/À Â OT-1 mice (31).Because RAG-1 À/À Â OT-1 mice possess only ovalbumin-specific CTLs, transplanted MC-GP i+e tumors were not controlled by the endogenous ovalbumin-specific immune system (Fig. 2C).Adoptive transfer of 10 7 naive P14 splenocytes at the day of tumor transplantation reduced but did not control tumors in RAG-1 À/À Â OT-1 mice.This indicates that the transfer of tumor-specific P14 CTLs was also effective in a nonlymphopenic host, which lacks tumor-specific B and T cells.
Growing MC-GP i+e tumors were explanted at the end of the experiment and tumor cells were analyzed for expression of LCMV-GP epitopes.MC-GP i+e tumor cells isolated from RAG-1 À/À mice efficiently stimulated IFNg production by splenocytes of LCMV-infected memory mice (Fig. 2D).In contrast, tumor cells isolated from RAG-1 À/À mice after transfer of naive P14 CTLs (Fig. 2D) or isolated from P14 mice (data not shown) restimulated LCMV memory responder cells less efficiently.This indicated that CTL escape variants developed in the presence of P14 T cells.
Compared with the results obtained in the fibrosarcoma model, adoptive immunotherapy with naive 10 7 P14 splenocytes at the day of tumor transplantation delayed B16-GP i+e tumor formation in RAG-1 À/À mice but enhanced tumor growth in BL/6 mice (Fig. 3A).B16-GP i+e tumor cells isolated from growing tumors of untreated BL/6 mice efficiently stimulated IFNg production by splenocytes of LCMV-immune mice.Tumor cells isolated from BL/6 mice, treated with naive P14 CTLs, restimulated LCMV-immune splenocytes less efficiently (Fig. 3A).Importantly, LCMV-GP epitopes were still expressed by the tumor cells (although at a lower level) at the day of explantation.B16-GP i+e tumors transplanted s.c.grew faster in naive P14 mice than in naive BL/6 mice.In contrast, if tumor cells were injected as single-cell suspensions to BL/6 or P14 mice, no tumor formation was observed (Fig. 3A; data not shown).
In clinical protocols, adoptive immunotherapy is often done with in vitro activated CTL clones or lines (2,3).We therefore activated P14 CTLs in vitro in the presence of gp33 peptide and IL-2 for 3 days.To mimic a clinically more relevant situation, treatment with activated P14 CTLs was done in mice with established tumors.Mice with s.c.transplanted B16-GP i+e tumor fragments were treated with 10 6 activated P14 CTLs when tumor diameter reached f0.3 cm in RAG-1 À/À (day 22 after transplantation) and BL/6 mice (day 32 after transplantation).Again, activated P14 CTLs efficiently reduced growth of established tumors in RAG-1 À/À mice but enhanced tumor growth in BL/6 mice (Fig. 3B).Similarly to the results observed after transfer of naive P14 CTLs (Fig. 2C), activated P14 CTLs efficiently controlled B16-GP i+e tumors in the absence of homeostatic proliferation in RAG-1 À/À Â OT-1 mice (Fig. 3B).
Most known tumor-associated antigens (TAA) are intracellular proteins and are not expressed on the cell surface (32).Therefore, we analyzed the influence of adoptive transfer of P14 CTLs on growth of tumors with intracellular LCMV-GP expression (B16-GP i ).Similarly to the results obtained with tumor cells expressing LCMV-GP on the cell surface, tumor growth was enhanced after transfer of activated P14 CTLs to B16-GP I -tumor bearing BL/6 mice (Fig. 3C).
As shown above, LCMV-GP transfectants induced CTLs specific for various epitopes (Fig. 1B) as well as antibodies against LCMV-GP (Fig. 1B and C).To analyze if adoptive immunotherapy may be effective if targeted against a tumor that expresses only a single immunogenic epitope, B16 tumors transfected with LCMV-gp33 epitope were analyzed (21).The parental B16 tumor cells are poorly immunogenic and usually few cells injected s.c. as single-cell suspension are sufficient for tumor formation in immunocompetent hosts (33).The adoptive transfer of naive 10 7 P14 splenocytes delayed B16-gp33 tumor formation in immunodeficient RAG-1 À/À Figure 3. Melanoma growth after adoptive T-cell therapy.A, B16-GP i+e tumor fragments were transplanted s.c. to RAG-1 À/À and BL/6 mice.Mice received 10 7 naive P14 splenocytes i.v.where indicated.Black arrows, day of adoptive immunotherapy.B16-GP i+e tumor cells were explanted 36 d after transplantation from untreated BL/6 mice and from BL/6 mice treated with 10 7 naive P14 splenocytes.Single-cell suspensions of these explanted tumors were used as stimulators in an IFNg secretion assay of LCMV-immune splenocytes.Numbers F SE indicate percentage of CD8 + IFNg + T cells.Stimulation with gp33 peptidepulsed splenocytes was taken as 100%.Values in parentheses show percentage of IFNg + CD8 + T cells after stimulation with nontransfected B16 cells.B16-GP i+e tumors were either transplanted as tumor fragments or injected as single-cell suspension (2 Â 10 6 ) s.c. to P14 mice (A).B, B16-GP i+e tumor fragments were transplanted s.c. to RAG-1 À/À , BL/6, and RAG-1 mice but also in immunocompetent BL/6 mice (Fig. 3D).Importantly, B16-gp33 and B16-GP i+e transfectants were lysed similarly in vitro, suggesting that the observed differences in tumor growth are not due to differences in P14-mediated tumor cell lysis (Supplementary Fig. S2B).
Adoptive immunotherapy impairs endogenous tumor-specific immune responses.Our results revealed that the same immunotherapy protocol controlled tumor growth in the absence of a tumor-specific immune response but enhanced tumor growth in the presence of an endogenous immunosurveillance.This prompted us to analyze the effect of the adoptive transfer of gp33specific P14 T cells on endogenous CTL and antibody responses.To reach high frequencies of specific CTLs after immunization, MC-GP i+e tumor cells were injected i.p. and the resulting CTL response was analyzed in an intracellular IFNg assay (Fig. 4A).MC-GP i+e tumor cells induced an LCMV-gp33-and gp276-specific immune response.If 10 7 naive P14 splenocytes were transferred i.v.before tumor cell injection, gp33-specific CTLs were detectable at a frequency comparable with control mice.In contrast, adoptive transfer of P14 CTLs impaired the CD8 + T-cell response to LCMV-gp276.In an additional experiment, P14 Â Ly5.1 + CTLs were adoptively transferred to BL/6 mice (Ly5.1 À ) before immunization with MC-GP i+e tumor cells.This experiment revealed that after transfer of P14 CTLs, the majority of IFNg-secreting gp33-specific CTLs derived from the transferred monoclonal P14 CTLs (61%), whereas the endogenous gp33-specific immune response was reduced (31%; Fig. 4B).P14 Â Ly5.1 + CTLs did not differ in the expression of activation markers CD25, CD69, and CD44 when compared with endogenous gp33-specific CTLs (data not shown).In addition, TCR transgenic and nontransgenic CTLs did not differ in IFNg production (mean fluorescence P14 Â Ly5.1 + , 38.5 F 3.5; endogenous gp33-specific CTLs, 38.2 F 2.2) and lysis of peptidepulsed target cells was identical (Fig. 4C).The effect of adoptive immunotherapy on tumor-specific antibody responses was analyzed after transplantation of B16-GP i+e tumor fragments to BL/6 mice and treatment with 10 6 naive P14 splenocytes i.v. at the day of transplantation.In analogy to the impaired CTL response, mice treated with P14 splenocytes mounted a reduced LCMV-specific antibody response (Fig. 4D).
In summary, adoptive transfer of tumor-specific CTLs directed against a single epitope of a model tumor antigen reduced tumor antigen-specific antibody titers and impaired CTL responses against other epitopes of the model tumor antigen.
Adoptive immunotherapy against tumors expressing different model tumor antigens.To analyze if and how adoptive immunotherapy against one tumor antigen influences the endogenous immune response to a second nonrelated tumor antigen, MC-GP i+e tumor cells were cotransfected with the NP of LCMV (MC-GP i+e /NP).MC-GP i+e and MC-GP i+e /NP tumor cells were lysed comparably in vitro by activated P14 CTLs (Supplementary Fig. S2A).MC cells transfected only with LCMV-NP served as controls (MC-NP).MC-GP i+e /NP tumor fragments did not grow after s.c.transplantation to BL/6 mice (Fig. 5A).Surprisingly, growth of MC-GP i+e /NP tumors was enhanced in BL/6 mice if naive P14 splenocytes were transferred at the day of tumor transplantation (Fig. 5A).A, BL/6 mice were treated with 10 7 naive P14 splenocytes or left untreated.All BL/6 mice were then immunized thrice (days 0, 7, and 14) i.p. with 2 Â 10 6 MC-GP i+e tumor cells.Splenocytes were analyzed 21 d later for IFNg + CD8 + T cells after in vitro stimulation with gp33 and gp276.Mean F SE of four mice.Values in parentheses show percentage of IFNg + CD8 + T cells of nonpeptidepulsed cultures; percentage of IFNg + CD8 + T cells of cultures pulsed with irrelevant LCMV-np396 was similar to nonpeptide-pulsed cultures.BL/6 mice were immunized similarly with MC cells i.p.Percentage of IFNg + CD8 + T cells after in vitro restimulation with gp33 was 0.2 F 0.02%.B, mice were treated with Ly5.1 + P14 CTLs as described for (A ).The percentage of adoptively transferred Ly5.1 + and Ly5.1 À IFNg + CTLs is shown.Mean F SE of four mice.C, splenocytes of mice immunized with MC-GP i+e or with MC either treated with P14 CTLs or left untreated were tested in a 51 Cr release assay 21 days after tumor injection.Filled symbols, gp33-pulsed target cells; empty symbols, unpulsed target cells.D, B16-GP i+e tumor fragments were transplanted s.c. to BL/6 mice and were either treated with 10 6 naive P14 splenocytes at the day of transplantation or left untreated.Thirty days after transplantation, LCMV-GP binding antibodies in the serum were measured by ELISA.BL/6 naive mice served as negative control.Points, mean of three mice per group; bars, SE.
The influence of adoptively transferred P14 CTLs on the endogenous immune response against LCMV-GP and LCMV-NP was analyzed in an intracellular IFNg assay after i.p. injection of MC-GP i+e /NP tumor cells as single-cell suspension.BL/6 control mice mounted CTL responses against LCMV-gp33, LCMV-gp276, and LCMV-np396 (Fig. 5D).In contrast, CTL responses to LCMV-gp276 and LCMV-np396 were markedly reduced in mice treated with naive (data not shown) or activated P14 CTLs (Fig. 5D).Similarly, the transfer of naive P14 CTLs to mice transplanted with solid MC-GP i+e /NP tumor fragments reduced the np396-specific CTL response when compared with nontransfected controls (data not shown).In contrast, adoptive immunotherapy with LCMV-gp33specific CTLs did neither enhance growth of MC-NP single transfected control tumors (Fig. 5B) nor reduced np396-specific CTL response (Fig. 5C).This excludes a nonantigen-specific immunosuppression after transfer of monoclonal CTLs.Thus, adoptive immunotherapy with CTLs specific for one epitope of a tumor antigen impaired the induction of CTL responses directed against other epitopes of the same antigen but also against other nonrelated antigens.
Adoptive immunotherapy targets antigen-presenting dendritic cells.To discriminate between effects of adoptively transferred P14 CTLs on the tumor itself and on antigen crosspresenting cells, D2-GP i+e tumor cells were transplanted to (H-2 d Â H-2 b ) F1 mice and then treated with CTLs from (BALB/c Â P14) F1 mice.In this experimental system, H-2 b gp33-restricted P14 CTLs will not recognize LCMV-GP epitopes presented on D2-GP i+e tumors (H-2 d ) but they will recognize gp33 epitopes crosspresented on host APCs (H-2 d Â H-2 b ).The adoptive transfer of naive 10 7 (BALB/c Â P14) F1 CTLs markedly enhanced growth kinetics of D2-GP i+e tumors (Fig. 6A).This experiment indicates that adoptively transferred tumor-specific CTLs enhance tumor growth by influencing antigen cross-presenting dendritic cells rather than the tumor itself.
We next analyzed the influence of adoptively transferred naive P14 T cells on induction of CTLs after immunization with LCMV-gp33 and LCMV-np396 peptide-pulsed dendritic cells.Peptidepulsed dendritic cells (2 Â 10 5 ) were injected i.p., and 8 days later, CTL responses were analyzed in an intracellular IFNg assay.Dendritic cells efficiently induced LCMV-gp33-and LCMV-np396specific immune responses in BL/6 control mice (Fig. 6B).Adoptive transfer of naive P14 CTLs slightly enhanced the frequency of LCMV-gp33-specific CTLs but reduced the frequency of LCMV-np396-specific CTLs (Fig. 6B).We further analyzed the effect of adoptively transferred P14 CTLs on the presence of gp33expressing dendritic cells in local lymph nodes.Dendritic cells were derived from BL/6 mice and were pulsed with gp33 and np396 or only np396, labeled with the fluorescent dye CFSE, and injected s.c.into the foot pad of the hind legs of BL/6 mice.At the same time, one group of mice received 2 Â 10 6 in vitro activated P14 CTLs.Twenty hours later, the number of CFSE + dendritic cells in the draining popliteal lymph node was analyzed (Fig. 6C).The number of CFSE + dendritic cells in lymph nodes of control mice was significantly higher (absolute number, 725 F 104) than in mice treated with P14 CTLs (absolute number, 261 F 66).No reduction of CFSE + np396-pulsed dendritic cells could be observed in BL/6 mice treated with P14 CTLs.Therefore, adoptive immunotherapy with LCMV-gp33-specific CTLs reduced the number of gp33expressing dendritic cells.This abrogates the presentation of other peptides of TAA by the same dendritic cells and results in an impaired CTL induction against peptides that are not the target of the adoptively transferred CTLs.

Discussion
To characterize the immunosurveillance of fibrosarcomas and melanomas and to analyze the effect of adoptively transferred tumor-specific CTLs, we established novel tumor transfectants with high expression of LCMV-GP as model tumor antigen.Transfected tumor cells efficiently induced CTLs against different epitopes of LCMV-GP as well as antibodies binding to LCMV-GP.Both effector arms contributed to immunosurveillance of transplanted melanomas and fibrosarcomas.
The surprising and unexpected finding of our study was that adoptive immunotherapy targeting a single epitope of an antigen enhanced tumor growth of solid immunogenic tumors by reducing the endogenous immune response to antigens or epitopes not targeted by the adoptive immunotherapy.The adoptively transferred P14 CTLs acquired effector function comparable with the endogenous gp33-specific CTLs, as indicated by IFNg production and lysis of peptide-pulsed target cells ex vivo.Adoptive immunotherapy of H-2 d tumors with BALB/c Â P14 CTLs in (H-2 d Â H-2 b ) F1 mice showed that tumor growth was enhanced due to a direct effect of adoptively transferred T cells on antigen cross-presenting cells.Earlier experiments using minor histocompatibility antigens or epitopes of ovalbumin and LCMV have shown that CD8 + T cells can cross-compete for different epitopes if presented by the same APC (34)(35)(36).We confirmed these earlier results using dendritic cells pulsed with gp33 and np396.In addition, we showed that adoptive transfer of monoclonal CTLs reduced the polyspecific endogenous immune response against a tumor to the epitope specificity of the transferred CTLs.The nature of this cross-competition has been proposed to be either of steric nature, competition for cytokines in the local environment, or inactivation of the APC via cell killing (37).Similarly to our study, it has been shown before that the CD8 + T-cell-dependent elimination of dendritic cells injected for vaccination limited the induction of antitumor immunity (35,38,39).Here, we report that after adoptive transfer of monoclonal CTLs, mechanisms of cross-competition on antigenpresenting dendritic cells may in some situations abrogate the induction of an endogenous tumor-specific CTL response and, as a consequence, enhance tumor growth.
It is currently ill defined how therapeutically manipulating one arm of the immune system affects the anticancer or cancerpromoting properties of the other.However, it has been reported that antibodies and/or B cells may in some situations reduce CTLmediated tumor control and enhance tumor growth (40).In addition, Siegel et al. (41) have shown that active immunization with an oncoprotein mutant in cancer-prone mice resulted in enhanced tumor growth, probably by inducing oncoproteinspecific antibodies.An enhanced tumor growth after adoptive immunotherapy has not been reported thus far.Earlier studies have documented that adoptive immunotherapy is effective when applied before or simultaneously with a tumor challenge with tumor cells injected as single-cell suspension (21,42,43).Similarly, in our study, MC-GP i+e and B16-GP i+e tumor cells injected in suspension s.c. were regularly rejected by tumor-specific P14 T cells.In contrast, eradication of established tumors by adoptive T-cell therapy is difficult and dependent on tumor size and the number of transferred CTLs (42,44).Therefore, cancer cells in solid tissue may be much more difficult to reject by the immune system.
Poorly immunogenic tumors expressing only weak CTL epitopes and tumors that are MHC class I negative may grow largely uncontrolled by the endogenous immune system.For example, poorly immunogenic B16 tumors grew similarly in pmel-1 TCR transgenic and control mice (8).If poorly immunogenic tumors are transfected with a single immunodominant epitope, targeting this epitope by monospecific CTLs results in tumor control.We showed in the present study that adoptive transfer of P14 CTLs improved the immunologic control of solid tumor fragments of B16 tumor cells transfected with the single immunodominant epitope of LCMV-GP (B16-gp33) in both immunodeficient RAG-1 À/À and immunoproficient BL/6 mice.Similarly, P14 T cells have been shown to cause regression of 3LL-A9-gp33 and B16-gp33 tumors (21,45).However, even cancer cells transfected with a strong antigen as entire protein have been shown to grow in immunocompetent hosts without evidence of T-cell activation (14,46,47).Consequently, transplantation of these tumors directly in TCR transgenic mice or the transfer of large numbers of tumor-specific CTLs will neither reduce immunosurveillance nor enhance tumor growth (46).Therefore, earlier studies analyzing adoptive T-cell therapy focused on tumors that grow largely independent of the endogenous immune system.In none of these earlier experimental models, an enhanced tumor growth after adoptive immunotherapy has been reported.These poorly immunogenic tumor models may be representative for a large group of human tumors expressing low immunogenic differentiation and tissue-specific TAA (32).In contrast, in the present study, we analyzed adoptive immunotherapy against a highly immunogenic tumor that is controlled by endogenous immunosurveillance.This experimental system may be more comparable with human tumors expressing highly immunogenic TAA, such as cancer-testis antigens or viral-tumor antigens (32).
In summary, the net therapeutic effect of adoptively transferred tumor-specific CTLs against the tumor is determined by effects not only on the tumor itself but also on tumor antigen cross-presenting cells.A therapeutic benefit is seen in situations where transferred CTLs are sufficient to eliminate the tumor completely.This situation has rarely been reached in clinically manifested tumors thus far, especially not in more advanced disease.However, protocols combining adoptive T-cell therapy with nonmyeloablative lymphodepletion, antigen-specific vaccination, and cytokine therapy or the repetitive transfer of large numbers of monospecific CTLs have been shown to cause regression of established tumors (8,48).If the solid tumor is not eliminated completely by the adoptively transferred CTLs, the consequences of the transferred CTLs on antigen cross-presenting dendritic cells gain importance.Here, the positive therapeutic effect on the tumor may be reduced or in some situations even reversed by the reduction of the endogenous immunosurveillance to nontargeted tumor antigens.Therefore, adoptive transfer of tumor-specific CTLs may influence the delicate equilibrium between the immune control and tumor growth, depending on the effect of the therapy on the endogenous tumor immunosurveillance.

Figure 1 .
Figure 1.MC-GP i+e and B16-GP i+e tumor cells elicit B-and T-cell responses.A, MC-GP i+e (filled curve ) and MC-GP i (empty curve) or B16-GP i+e (filled curve ) and B16-GP i (empty curve) tumor cells were analyzed for surface expression of LCMV-GP by flow cytometry.Isotype control (dotted line ).Splenocytes of LCMVimmune BL/6 mice were stimulated for 5 h in vitro with different numbers of MC-GP i+e or MC-GP i and B16-GP i+e or B16-GP i tumor cells.Results are shown as percentage of IFNg + CD8 + T cells after stimulation with relevant tumor cell lines.Stimulation with gp33 peptide-pulsed splenocytes was taken as 100%.B, fragments of MC-GP i+e -or empty vector-transfected MC tumors were transplanted s.c. to BL/6 mice and tumor growth was followed.C, similarly, B16-GP i+e or parental B16 fragments were transplanted s.c. to BL/6 mice and tumor growth was followed.Fourteen days after transplantation (MC-GP i+e in B ; B16-GP i+e in C), splenocytes were analyzed in an intracellular IFNg assay after restimulation with the relevant peptide.Results are given as numbers F SE of three to four mice per group.Values in parentheses show percentage of IFNg + CD8 + T cells of nonpeptide-pulsed cultures.Thirty days after transplantation, LCMV-GP binding antibodies in the serum were measured in an ELISA (BL/6 mice with MC-GP i fragments or naive BL/6 mice served as controls; B and C ). D, (BALB/c Â BL/6) F1 mice were immunized twice (days 0 and 7) i.p. with 2 Â 10 6 D2-GP i+e tumor cells.Splenocytes were tested in a 51 Cr release assay against gp33-labeled EL-4 target cells 15 d later.LCMV-immune and naive (BALB/c Â BL/6) F1 mice served as controls.Points, mean of six transplanted tumors per group; bars, SE (B and C ). Numbers in (B and C ) indicate total number of transplanted tumors at this experimental condition, summarized from two to six experiments.*, P < 0.05.

Figure 2 .
Figure 2. Fibrosarcoma growth after adoptive T-cell therapy.A to C, MC-GP i+e tumor fragments were transplanted s.c. to RAG-1 À/À (A ), BL/6 (A ), P14 (B ), and RAG-1 À/À Â OT-1 (C) mice.Mice either received naive P14 splenocytes i.v. at the day of transplantation or were left untreated (A and C ). MC-GP i+e tumors were either transplanted as tumor fragments or injected as single-cell suspensions s.c.(2 Â 10 6 cells) to P14 mice (B).Black arrows, day of adoptive immunotherapy.Tumor growth was measured at the indicated days after transplantation.Points, mean of six to eight transplanted tumors; bars, SE.Numbers indicate total number of transplanted tumors at this experimental condition, summarized from one to two experiments.D, MC-GP i+e tumor cells were explanted 31 d after transplantation from untreated RAG-1 À/À mice and from RAG-1 À/À mice treated with naive P14 CTLs at the day of transplantation.Single-cell suspensions of these explanted tumors were used as stimulators in an IFNg secretion assay of LCMV-immune splenocytes.Numbers F SE indicate percentage of CD8 + T cells producing IFNg.Stimulation with gp33 peptide-pulsed splenocytes was taken as 100%.Values in parentheses indicate percentage of IFNg + CD8 + T cells after stimulation with nontransfected MC tumor cells.*, P < 0.05; **, P < 0.01.NS, not significant.

Figure 4 .
Figure 4. Effect of adoptive immunotherapy on the tumor-specific endogenous immune response.A, BL/6 mice were treated with 10 7 naive P14 splenocytes or left untreated.All BL/6 mice were then immunized thrice (days 0, 7, and 14) i.p. with 2 Â 10 6 MC-GP i+e tumor cells.Splenocytes were analyzed 21 d later for IFNg + CD8 + T cells after in vitro stimulation with gp33 and gp276.Mean F SE of four mice.Values in parentheses show percentage of IFNg + CD8 + T cells of nonpeptidepulsed cultures; percentage of IFNg + CD8 + T cells of cultures pulsed with irrelevant LCMV-np396 was similar to nonpeptide-pulsed cultures.BL/6 mice were immunized similarly with MC cells i.p.Percentage of IFNg + CD8 + T cells after in vitro restimulation with gp33 was 0.2 F 0.02%.B, mice were treated with Ly5.1 + P14 CTLs as described for (A ).The percentage of adoptively transferred Ly5.1 + and Ly5.1 À IFNg + CTLs is shown.Mean F SE of four mice.C, splenocytes of mice immunized with MC-GP i+e or with MC either treated with P14 CTLs or left untreated were tested in a 51 Cr release assay 21 days after tumor injection.Filled symbols, gp33-pulsed target cells; empty symbols, unpulsed target cells.D, B16-GP i+e tumor fragments were transplanted s.c. to BL/6 mice and were either treated with 10 6 naive P14 splenocytes at the day of transplantation or left untreated.Thirty days after transplantation, LCMV-GP binding antibodies in the serum were measured by ELISA.BL/6 naive mice served as negative control.Points, mean of three mice per group; bars, SE.

Figure 5 .
Figure 5. Adoptive immunotherapy against a tumor expressing two model tumor antigens.A and B, MC-GP i+e /NP (A) or MC-NP (B ) tumor fragments were transplanted s.c. to BL/6 mice.Mice were either treated with 107 naive P14 splenocytes at the day of transplantation or left untreated.Black arrows, day of adoptive immunotherapy.Tumor growth was measured at the indicated days after transplantation.C, BL/6 mice were immunized once with 2 Â 10 6 MC-NP tumor cells i.p. and were either treated with 10 7 naive P14 splenocytes at the same day or left untreated.Eight days later, the CTL response was assessed in a 51 Cr release assay.Filled symbols, np396-labeled EL-4 target cells; empty symbols, unpulsed target cells.Points, mean of six to eight transplanted tumors per group; bars, SE (A and B ). Numbers in (A and B) indicate total number of transplanted tumors at this experimental condition, summarized from two experiments.D, BL/6 mice were immunized once with 2 Â 10 6 MC-GP i+e / NP tumor cells i.p. and were either left untreated or treated with 2 Â 10 6 activated P14 T cells on the same day.Splenocytes were analyzed 8 d later for CD8 + T cells producing IFNg after in vitro stimulation with gp33, gp276, and np396.Mean F SE of six to nine mice per group summarized from two experiments indicates percentage of CD8 + T cells producing IFNg within total CD8 + T cells.Values in parentheses show percentage of IFNg + CD8 + T cells of nonpeptide-pulsed cultures.*, P < 0.05.
Figure 6.Adoptive immunotherapy reduces antigen presentation by dendritic cells (DC ) in local lymph nodes.A, D2-GP i+e tumor fragments were transplanted s.c. to (BALB/c Â BL/6) F1 mice.Mice were either treated i.v. with 10 7 naive P14 Â BALB/c splenocytes at the day of transplantation or left untreated.Black arrow, day of adoptive immunotherapy.SE of seven mice per group summarized from two experiments.Values in parentheses show percentage of IFNg + CD8 + T cells of nonpeptide-pulsed cultures.C, 2 Â 10 5 CFSE-labeled dendritic cells pulsed with gp33 and np396 or only np396 were injected s.c.into the foot pad of the hind legs of BL/6 mice.Then, BL/6 mice were either treated with 2 Â 10 6 activated P14 CTLs or left untreated.Twenty hours later, the number of CFSE + CD11c + dendritic cells in the draining popliteal lymph node (popl.LN ) was analyzed by flow cytometry.Columns, mean of four mice per group; bars, SE. **, P < 0.01.