Strategies that generate tumor Ag-specific effector cells do not necessarily cure established tumors. We hypothesized that the relative efficiency with which tumor-specific effector cells reach the tumor is critical for therapy. We demonstrate in this study that activated T cells respond to the chemokine CCL3, both in vitro and in vivo, and we further demonstrate that expression of CCL3 within tumors increases the effector T cell infiltrate in those tumors. Importantly, we show that adenoviral gene transfer to cause expression of CCL3 within B16ova tumors in vivo increases the efficacy of adoptive transfer of tumor-specific effector OT1 T cells. We additionally demonstrate that such therapies result in endogenous immune responses to tumor Ags that are capable of protecting animals against subsequent tumor challenge. Strategies that modify the “visibility” of tumors have the potential to significantly enhance the efficacy of both vaccine and adoptive transfer therapies currently in development.
Tumors can progressively grow in individuals despite the presence of tumor-specific effector cells. A useful concept to explain these observations is that there is an “ignorance” of the developing antigenic tumor. Various models have demonstrated that efficient priming of effector cells does not necessarily mean that Ag-specific T cells will eliminate tumors (1, 2, 3). For example, in a model where mice were transgenic for a TCR that recognizes self-MHC presenting an Ag transgenically expressed in normal liver cells, Ag-specific vaccination failed to cause liver pathology (1). In this model, administration of a liver-specific pathogen was required to cause T cell-mediated liver destruction (1). Similar results have been seen in many models in which large numbers of transgenic or in vitro-derived activated CTLs do not consistently cause regression of tumors (2, 3). The ability of T cells to mount an effector response to peripheral Ag is distinct from their specificity for that Ag. Activated effector T cells, rather than naive cells, are capable of trafficking to peripheral inflammatory sites (4, 5, 6). This ability to traffic or home to sites experiencing inflammation is not related to Ag specificity, since effector cells simultaneously generated to Ags not present in an ongoing infection are equally present at an infection site (7). Both naive and activated effectors have demonstrated that they are unable to traffic to tumor sites expressing their specific Ag; however, delivery of proinflammatory factors that change the tumor microenvironment allows infiltration of activated T cells and tumor destruction (8). In human trials, low numbers of adoptively transferred cells accumulate in tumors despite Ag expression (9, 10, 11), strengthening the distinction between Ag specificity and trafficking. However, the overall trafficking of effector cells that do reach a specific site will subsequently be influenced by the presence or absence of specific Ag through retention of Ag-specific cells and release of additional chemoattractants by Ag-specific cells (12). The role of chemokines in directing the peripheral trafficking of effector T cells is suggested by the phenotypic remodeling that occurs following T cell activation (4, 5, 6), and treatment of mice with pertussis toxin, an inhibitor of chemokine signaling, has been shown to prevent systemic therapy of tumors by effector T cells (13).
We hypothesize that increasing the trafficking of effector T cells to tumors will significantly enhance immunotherapy approaches that adoptively transfer or endogenously generate tumor-specific effector T cells. In this study, we demonstrate that T cells express the chemokine receptor CCR5 on activation and that these cells are functionally responsive to its cognate ligand CCL3 both in vitro and in vivo. Expression of CCL3 in tumors enhances trafficking of effector T cells to tumor sites, resulting in enhanced tumor destruction. Finally, we combine intratumoral gene delivery of CCL3 with adoptive transfer of activated effector T cells and demonstrate that enhancing the “visibility” of tumor sites to activated T cells significantly enhances adoptive immunotherapy of tumors.
Materials and Methods
Cell lines, reagents, and mice
The B16ova cell line was kindly provided by Dr. E. Celis (Mayo Clinic, Rochester, MN). The B16pCR3.1 and B16CCL3 cell lines have been previously described (14). Briefly, B16 cells were transfected with pCR3.1 empty vector or pCR3.1 incorporating murine CCL3 and clones selected based on resistance to 5 mg/ml neomycin. Individual clones were screened for secretion of CCL3 by specific ELISA, and a clone secreting high levels of CCL3 was designated B16CCL3. This clone and a CCL3-negative empty vector-transfected clone designated B16pCR3.1 showed identical in vitro growth rates to parental B16 cells. The directly conjugated Abs CCR5-PE, CD8-PE, CD69-FITC, and CD62 ligand (CD62L) 3-allophycocyanin were purchased from BD Biosciences. SIINFEKL peptide was synthesized by the Mayo Clinic Protein Core Facility and oligonucleotide primers were synthesized by the Mayo Clinic Oligonucleotide Core Facility. Six-week-old C57BL/6 mice were age and sex matched for individual experiments. OT1 mice, transgenic for a TCR that recognizes the SIINFEKL peptide of OVA presented on H-2Kb, have been previously described (15).
Preparation of primary murine cells
For preparation of activated OT1 CTLs, spleen and lymph nodes from OT1-transgenic mice were combined and crushed through a 100-μm filter to prepare a single-cell suspension. RBC were removed by a 2-min incubation in ACK buffer (sterile dH2O containing 0.15 M NH4Cl, 1.0 mM KHCO3, and 0.1 mM EDTA adjusted to pH 7.2–7.4). Remaining cells were adjusted to 2.5 × 106 cells/ml in IMDM plus 5% FCS, 10−5 M 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin and stimulated with 1 μg/ml SIINFEKL peptide and 50 IU/ml human IL-2. Every 2–3 days one-half of the medium was removed and replaced with fresh medium containing 50 IU/ml IL-2. For use in vivo, nonadherent and loosely adherent cells were harvested following one activation cycle of 3–5 days and viable cells were purified by density gradient centrifugation using Lympholyte-M (Cedarlane Laboratories) according to the manufacturer’s instructions.
Cell labeling and analysis
For analysis of phenotype, 1 × 106 cells were washed in PBS containing 0.05% BSA (wash buffer), resuspended in 50 μl of wash buffer, and exposed to directly conjugated primary Abs for 30 min at 4°C. Cells were then washed and resuspended in 500 μl of PBS containing 4% formaldehyde. Cells were analyzed by flow cytometry and data were analyzed using CellQuest software (BD Bioscences).
For in vivo tracking, cells were labeled with the viable cell dye chloromethyl benzamido/1,1′-dioctadecyl-3,3,3′,3,′-tetramethylindocarbocyanine perchlorate (DiI) (Molecular Probes) according to the manufacturer’s instructions. Briefly, viable cells were resuspended in PBS at 1 × 106 cells/ml and incubated for 30 min with 2.5 μM DiI. Cells were washed, resuspended in IMDM plus 5% FCS, 10−5 M 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin and incubated for an additional 15 min on ice. Finally, cells were washed another three times in PBS before in vivo injection of 2 × 106 cells. For analysis of labeled cell infiltrate, tumors were harvested and fixed in 10% Formalin in PBS, then paraffin embedded and sectioned. Unstained sections were visualized for accumulation of DiI-labeled cells. H&E-stained sections were prepared for analysis of tissue destruction and gross infiltrate. A pathologist examining H&E sections, while blinded to the experiment design, scored the degree of necrosis.
In vitro response of CD8 T cells to chemokines
For functional responses to chemokines, viable cells were washed and resuspended at 1 × 107 cells/ml in HBSS plus 10 mM HEPES. Cells were mixed with an equal volume of HBSS plus 10 mM HEPES containing 20 μM indomethacin 1 acetyl ester (Indo-1AM; Molecular Probes) and incubated for 30 min at 37°C. Cells were washed two times in HBSS plus 10 mM HEPES plus 0.05% BSA (wash buffer) and resuspended at 5 × 106 cells/ml in wash buffer for analysis. Where appropriate, following Indo-1AM labeling, cells were costained with directly conjugated primary Abs for 15 min at room temperature, then washed one time in wash buffer and stored on ice until immediately before analysis. For analysis, samples were individually warmed to 37°C and analyzed on a FACStar+ flow cytometer (BD Biosciences). Baseline readings were performed for 1 min, then agonist was added and the ratio of FL4 (390 nm):FL5 (500 nm) Indo-1am fluorescence was determined to calculate calcium flux over a 4-min interval following addition of agonist. The response of cells to agonists was analyzed using FlowJo software (TreeStar).
For in vitro chemotaxis assays, activated OT1 T cells were washed and 5 × 105 cells added per well of a ChemoTx chemotaxis chamber (NeuroProbe) separated by membrane from 48-h supernatants from B16 cells infected with 1000 multiplicity of infection of AdCCL3, control adenovirus, or uninfected B16. Further controls included medium alone and medium containing 100 μg/ml rCCL3. ChemoTx chambers were incubated at 37°C for 90 min, and cells entering the lower chamber were quantified by the MTT assay (Roche). Values were compared with a standard curve of cell number to determine cellular infiltration for each well.
RT-PCR of chemokine injection site
C57BL/6 mice were injected with three daily injections of CCL3 (200 ng/mouse per injection) or vehicle control (PBS) and the injection site was harvested after 4 days. Total RNA was prepared from the injection site (RNeasy; Qiagen), 1 μg total RNA reverse transcribed using an AMV Reverse Transcriptase Kit (Roche) according to the manufacturer’s instructions, and PCR was performed with primers specific for a range of chemokine receptors: CCR1 (forward, CTAATGATTCTGGTGCTCATGC; reverse, CCACTGCTTCAGGCTCTTGTAG); CCR5 (forward, CATGATGGTCTTCCTCATCTTG; reverse, AACAGGGTGTGGAGAATTCCTG); CCR6 (forward, ATGGTGGTGATGACCTTTGC; reverse, CCAAAGAACAGCTCCAGTCC); CCR7 (forward, CAAACAGGAGCTGATGTCCA; reverse, ATGACAAGGAGAGCCACCAC); and GAPDH (forward, GTGGGCCGCTCTAGGCACCAA; reverse, CTCTTTGATGTCACGCACGATTTC). PCR was performed in a 50-μl reaction mixture with 250 μM of each dNTP, 100 nM of primers, 2.5 U of AmpliTaq DNA polymerase (Applied Biosystems), and 0.1-μg equivalent of reverse transcribed mRNA in a 1× final concentration of reaction buffer (Applied Biosystems). PCR was performed for 40 cycles of 95°C for 1 min; 55°C for 1 min, and 72°C for 2 min.
In vivo chemoattraction assay
To establish a defined in vivo environment to measure chemoattraction, sterile Gelfoam sponges (Pharmacia and Upjohn) were inserted s.c. into mice following the protocol of Buchanan and Murphy (16) as previously described (14). Once established for 72 h in vivo, sponges were injected with recombinant chemokine or PBS vehicle, and sponges were harvested over a time course. Sponges were finely chopped and then incubated in 15 ml of collagenase enzyme mixture (20 mg/ml BSA and 400 U/ml collagenase in saline G, 1.1 g/L glucose, 8 g/L NaCl, 0.4 g/L KCl, 0.29 g/L Na2HPO4.7H2O, 0.15 g/L KH2PO4, 0.15 g/L MgSO4.7H2O, and 0.016 g/L CaCl2.H2O in endotoxin-free water) for 3 h at 37°C with agitation. The suspension was passed through a 100-μm filter, washed with HBSS, and infiltrating cells from three mice per group were pooled, and total cells were counted by trypan blue exclusion on a hemocytometer.
Preparation of adenovirus
Replication-defective AdGFP has previously been described (17). We have previously described the cloning and characterization of the murine CCL3 cDNA (14). For construction of an adenovirus expressing CCL3, the CCL3 cDNA was subcloned into the AdEasy shuttle vector (Qbiogene). This shuttle vector was cotransformed with transfer vector into bacteria according to the manufacturer’s protocol, and full-length recombinant adenoviral genomes incorporating CCL3 were identified by restriction enzyme digest. Recombinant adenoviral DNA was transfected into 293 cells and adenoviral plaques were screened by ELISA for CCL3 in supernatants. A selected positive plaque was amplified and cesium chloride purified to concentrations in excess of 1 × 1010 PFU/ml.
In vivo experiments
All experiments involving animals were performed in accordance with the Mayo Clinic Institutional Animal Care and Use Committee. For survival analysis, all groups contained 10 mice and statistical analysis of significance was performed using the log rank test.
In vitro T cell restimulation
The spleen was isolated from those animals surviving >60 days posttumor challenge, treated with intratumoral AdCCL3 and adoptive transfer of activated OT1 T cells. Cells were isolated by crushing through a 100-μm nylon filter and seeded at 5 × 106 cells/ml in IMDM containing 5% FCS, 10−5 M 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin supplemented with 50 IU/ml human IL-2. Cells were stimulated with 5 μM specific peptide for OVA257–264 (SIINFEKL), the H2Kb-binding peptide recognized by the 2C TCR (SIYRYYGL), murine gp10025–33 (EGSRNQDWL), or Trp2180–188 (SVYDFFVWL). Cell-free supernatants were collected after 48 h and tested by specific ELISA for IFN-γ (BD OptEIA IFN-γ; BD Biosciences).
Expression of CCR5 by activated T cells
We have observed that activated OT1 T cells begin to lose the ability to cure B16ova tumors in C57BL/6 mice if tumors are allowed to establish for >7 days before adoptive transfer (data not shown). Increasing the number of T cells transferred into tumor-bearing mice has a limited ability to improve the therapy of established tumors (data not shown). We hypothesized that this inability to eradicate the tumor was attributable to lack of visibility of an established tumor to activated T cells. To test this hypothesis, we examined chemokine receptors on activated T cells. In vitro activation of OT1 splenocyte and lymph node cells with SIINFEKL peptide plus IL-2 leads to an expansion of CD8 T cells, and an emergence of a CCR5+ T cell population (Fig. 1,a). Gating on CD8+ cells in the spleen and lymph node from naive OT1 mice confirmed that naive CD8 T cells have central trafficking properties, with high expression of CD62L and no expression of the early activation marker CD69 (Fig. 1,b). There are few CCR5+ cells in the naive lymph node and spleen, and these cells likely represent non-T cell populations. Following activation, CD8+ cells up-regulate CD69 and there are both centrally trafficking (CD62Lhigh) and peripheral trafficking (CD62Llow) populations (Fig. 1,b). Not all of the CD8+ OT1 cells express CCR5 at this stage of culture, with ∼5% of CD8 T cells CCR5+, and this proportion decreases with continued in vitro culture. We hypothesize that CCR5 expression on CD8 T cells relates to cytokine exposure on activation as has been described for CD4 T cells (18, 19), and that it would be possible to design in vitro activation protocols that generate a much higher proportion of CCR5+ CD8 T cells for improved trafficking to inflammatory sites. Gating on the CCR5+ cells in this population indicates that they are activated T cells, since they uniformly express CD3, the activation marker CD69, and these cells are exclusively CD62Llow (Fig. 1 b). These data suggest that the CCR5+ population that appears on activation represent T cells that are capable of peripheral trafficking to sites of inflammation.
CD8 T cells respond to inflammatory chemokines following activation
To confirm that the expression of CCR5 on CD8 T cells translated into a functional effect, the ability of CD8 T cells to respond to recombinant chemokine was tested. Activated T cells were labeled with Indo-1, which has distinct emission spectra depending on whether it is calcium bound or unbound, permitting real-time flow cytometric measurements of calcium flux in live cells, an indicator of chemokine receptor signaling. Activated OT1 cells were loaded with Indo1 and counterstained with directly conjugated Abs to confirm the identity of CCL3-responsive cells. During flow cytometry, cells were treated with recombinant chemokines and calcium flux in stained cell populations was determined. Figure 2 demonstrates that the CD8 population of activated OT1 cells functionally responds to rCCL3. These cells do not respond to the control chemokine CCL21, the receptor for which is primarily expressed on naive T cells or the unrelated chemokine CX3CL1. Thus, activated T cells both express CCR5 and respond to its cognate ligand CCL3.
To confirm expression of functional CCR5 in ligand-directed chemotaxis, OT1 cells were tested in a chemotaxis assay against rCCL3 or supernatants of B16 melanoma cells engineered to express CCL3. Figure 2 b demonstrates that compared with supernatants from unmodified B16 cells, activated OT1 cells show significant chemotaxis toward supernatants from B16 cells stably expressing CCL3 (p < 0.05), B16 cells infected with AdCCL3 (p < 0.05), or rCCL3 alone (p < 0.01), but not toward control supernatants from B16 infected with AdGFP (p = 0.25). These results demonstrate a specific chemotaxis of activated OT1 cells toward CCL3 that is not influenced by the adenoviral vector and supports the hypothesis that CCL3 is a suitable chemokine to render target sites more “visible” to activated OT1 cells.
CCL3 attracts CCR5-expressing cells in vivo
To confirm that CCL3 functions to attract CCR5+ cells in vivo, we injected s.c. sites with three daily doses of CCL3 or PBS and characterized expression of chemokine receptors in the injection site 24 h after the last injection. RT-PCR of RNA prepared from the injection site demonstrated the presence of CCR5 only in sites injected with CCL3 (Fig. 3). Expression of CCR1, CCR6 and CCR7 were not seen in PBS or CCL3 injection sites. The lack of CCR1 expression is interesting since CCL3 has been described to signal via both CCR1 and CCR5 (20). These data suggest that CCL3 displays an in vivo selectivity for cells expressing CCR5; however, we cannot exclude the possibility that expression of CCR1 is below detection in this system.
RT-PCR identifies changes in receptor RNA transcripts and is thus not affected by ligand-mediated receptor modulation that may occur in the injection site. However, to exclude the possibility that chemokine injection increased receptor expression on resident cells rather than caused influx of new cells, we tested the effect of chemokine treatment on activated splenocytes in vitro. High doses (25 ng/ml) of CCL3, CCL20, or CCL21 did not alter surface expression of CCR5 on activated splenocytes (data not shown), supporting the hypothesis that injection of recombinant chemokine causes influx of new CCR5-expressing cells to the injection site.
To confirm that CCL3 injection can cause chemoattraction in vivo, we used a Gelfoam sponge model to measure cell influx to a defined environment. Sponges were established s.c. in mice and injected with rCCL3 or PBS vehicle. Sponges were harvested over a time course, and dissociated and infiltrating cells were counted. Figure 3 b shows that the addition of rCCL3 caused a transient influx of cells into the sponge. This influx is over and above the continuing background population of the sponge environment that is seen with the PBS control. These data demonstrate that rCCL3 injection causes an increase in CCR5-expressing cells in the injection site and a transient in vivo chemoattraction.
Activated effector cells are attracted to sites of Ag and chemokine in vivo
To test our hypothesis that tumor control by specific effector cells is limited by the restricted access of effector cells to the tumor site, we studied in vivo tracking of effector T cells to the tumor site. In vitro-activated OT1 T cells were labeled with DiI and adoptively transferred into mice bearing tumors on opposite flanks. In mice bearing 10-day B16ova tumors on one flank and control B16pCR3.1 tumors on the opposite flank, more DiI-labeled cells were visible in the Ag-expressing tumor (Fig. 4, a and b). To study the effect of chemokine expression, tumors were established where 90% of the cells were B16ova and 10% of the cells on one flank were B16CCL3 and on the other flank were B16pCR3.1. Similar numbers of DiI-labeled cells were visible in the B16ova/B16pCR3.1 to those seen in 100% B16ova tumors (Fig. 4, a and d). However, expression of CCL3 in the tumor enhanced the infiltrate of activated T cells (Fig. 4,c). These data are summarized in Table I. That expression of CCL3 enhances the infiltrate of adoptively transferred activated T cells supports the hypothesis that the infiltrate is limited in tumors that do not express CCL3.
|Tumor .||Infiltrate .|
|Tumor .||Infiltrate .|
Mice were injected with tumor cells on opposite flanks, and once established to 0.3 cm in diameter, mice were injected i.v. with 2 × 106 in vitro-activated OT1 T cells that had been labeled with the viable cell dye DiI. Four days after injection, tumors were harvested, Formalin fixed, and paraffin sections taken for analysis of infiltrate of fluorescent cells. Multiple fields were assessed and scored on a scale from − (no infiltrating cells stain with DiI) to +++++ (100% of infiltrating cells stain with DiI).
Expression of CCL3 in tumors enhances antitumor efficacy of adoptively transferred effector T cells
Serial sections from tumors shown in Fig. 4 were H&E stained, and tissue destruction and cellular infiltrate were studied. B16pCR3.1 tumors were essentially undisturbed, with limited necrosis and limited infiltration (Fig. 5,a). Tumors expressing OVA demonstrated infiltration and small areas of necrosis (Fig. 5, a and d). In contrast, tumors expressing OVA plus CCL3 demonstrated significantly larger infiltrates and larger areas of necrosis than tumors expressing OVA alone (Fig. 5, c and d). Pathology scores of necrosis are shown in Table II.
|Tumor .||Area of Necrosis .|
|Tumor .||Area of Necrosis .|
Mice were injected with tumor cells on opposite flanks, and once established to 0.3 cm in diameter, mice were injected i.v. with 2 × 106 in vitro-activated OT1 T cells. Four days after injection, tumors were harvested, Formalin fixed, and paraffin sections taken and HCE stained. Stained sections were assessed for degree of necrosis by pathologists blinded to the experiment design. The degree of necrosis was scored on a scale from − (no necrosis) to +++++ (total necrosis).
To test whether this enhanced tumor destruction had a significant impact on tumor growth, we established B16ova tumors in mice and allowed the tumors to grow to 0.3 cm in diameter. Tumors were then treated with three daily injections of recombinant adenoviral vector expressing CCL3 or control adenovirus expressing GFP. On the second day of adenovirus injection, one-half were treated with adoptive transfer of in vitro-activated OT1. AdCCL3 significantly delayed tumor growth (median survival 30 days) compared with control AdGFP injection (median survival 23 days, p < 0.001; Table III). Adoptive transfer of OT1 cells to AdGFP-treated mice significantly delayed tumor growth (median survival 39 days, p < 0.001); however, all mice eventually succumbed to tumor. Adoptive transfer of these numbers of effectors alone, though causing significant delays in tumor growth, consistently fails to cure mice of tumors once established for >1 wk. The combination of AdCCL3 with adoptive transfer of OT1 significantly delayed tumor growth (median survival 42 days) compared with AdCCL3 alone (p < 0.05; Table IV). Expression of CCL3 combined with adoptive transfer of OT1 did not significantly enhance survival of mice over AdGFP in combination with OT1 (p = 0.06), although 40% of mice survived with AdCCL3 plus OT1 compared with 0% with AdGFP plus OT1 (Table III).
|Tumor .||Adenovirus Vector .||Adoptive Transfer .||Median Survival (days) .||Survivors .||Rechallenge .||Survivors .|
|Tumor .||Adenovirus Vector .||Adoptive Transfer .||Median Survival (days) .||Survivors .||Rechallenge .||Survivors .|
Mice were injected s.c. with B16ova cells, and once established to 0.3 cm in diameter, three daily intratumoral injections of 1 × 109 PFu of AdGFP or AdCCL3 were performed. On the second day, mice were injected i.v. with 2 × 106 in vitro activated OT1 cells. Tumor growth was monitored and mice were sacrificed if tumors exceeded 1.2 cm in diameter. Mice surviving for >80 days were rechallenged on the opposite flank with 2 × 105 parental B16 cells. Tumor growth was monitored and mice were sacrificed if tumors exceeded 1.2 cm in diameter.
|90% Tumor .||10% Tumor .||Adoptive Transfer .||Median Survival (days) .||Survivors .|
|90% Tumor .||10% Tumor .||Adoptive Transfer .||Median Survival (days) .||Survivors .|
Mice were injected s.c. with 2 × 105 B16 cells, consisting of 90% B16ova cells and 2 × 105 of either B16CCL3 or control B16pCR3.1 cells. On the same day, mice were injected i.v. with 2 × 106 in vitro-activated OT1 cells. Tumor growth was monitored and mice were sacrificed if tumors exceeded 1.2 cm in diameter.
To exclude the potential effect of adenoviral infection on the trafficking and cytotoxicity of adoptive T cell therapy, the effect of chemokine expression by tumors on adoptive T cell therapy was tested in a constitutive rather than gene therapy model. Mice were injected s.c. with B16 tumor cells consisting of 90% B16ova and 10% of either B16CCL3 or control B16pCR3.1 cells. On the same day, mice were treated i.v. with adoptive transfer of activated OT1 T cells and tumor development was monitored. Only a combination of adoptive therapy and chemokine expression led to a significant survival benefit (p < 0.05) in tumor-bearing animals (summarized in Table IV). Although adenoviral vector components could be playing a role in the gene therapy model described above, these data suggest that chemokine expression in the absence of adenoviral vector delivery can significantly increase the efficacy of adoptive T cell therapy.
Mice rejecting tumors following intratumoral chemokine expression demonstrate both adoptively transferred and endogenous T cell responses
We were surprised to observe that adoptive transfer of SIINFEKL-specific OT1 cells cleared tumors in four of five mice in which 10% of the tumor cells did not express the ova Ag (Table IV). These data suggest that an endogenous immune response may play some role in the tumor clearance initiated through adoptive transfer and chemokine expression. To confirm this result and identify whether these endogenous effectors could sustain a functional antitumor response, we rechallenged the survivor mice indicated in Table III on the opposite flank with parental B16 tumors. These tumors will present melanoma epitopes recognized by endogenous T cells, for example gp100 and Trp2, but not recognized by the adoptively transferred OT1 T cells that are restricted to the SIINFEKL epitope of OVA. The last two columns of Table III summarize these experiments, showing that mice rejecting 0.3-cm B16ova tumors, via intratumoral AdCCL3 alone or combined with adoptive transfer of OT1 cells, subsequently rejected parental B16 tumors. To monitor the specificity of the response to the secondary tumor challenge, splenocytes from mice rejecting both B16ova and subsequent B16 tumors were stimulated in vitro with a panel of peptides and IFN-γ ELISA was performed on cell supernatants. Figure 6 demonstrates that these mice retain a specific response to SIINFEKL, with IFN-γ secretion significantly greater than in unstimulated cells (p < 0.05) or in the presence of the irrelevant 2C peptide (p < 0.05). Importantly, in one-half of the survivor mice there is a significant IFN-γ response to Trp2 peptide compared with untreated or 2C peptide-treated cells (p < 0.0001); however, when considered as a group there is no significant difference between Trp2-treated and untreated or 2C-treated cells. Similarly, one-half of the mice secreted significantly more IFN-γ in response to gp100 peptide than untreated cells or to irrelevant 2C peptide, although there is no significant difference when analyzed as a group. As expected, naive age-matched control mice did not secrete IFN-γ in response to SIINFEKL or melanoma differentiation Ags. Thus, these intratumoral therapies both enhance adoptive immunotherapy and stimulate functional endogenous tumor-specific immune responses that in some cases can be identified as targeting known melanoma-specific differentiation Ags.
We demonstrate that a subpopulation of T cells expresses the inflammatory chemokine receptor CCR5 on activation and that these T cells represent a peripheral trafficking effector population. These chemokine receptors are functional, and injection of the cognate ligand CCL3 causes attraction of receptor-expressing cells to the injection site. Expression of CCL3 in tumors in vivo enhances the trafficking of adoptively transferred effector T cells to the tumor site and results in enhanced tumor destruction. We further demonstrate that these data can be translated into therapy via adenovirus-mediated gene transfer of CCL3 into tumors, which enhances the survival of tumor-bearing mice following adoptive transfer. Importantly, the intratumoral immunotherapy also resulted in endogenous tumor-specific T cell responses that were capable of rejecting subsequent tumor challenge.
Ag-specific vaccination strategies generally result in accumulation of specific cells at the vaccination site (12, 21, 22). Local inflammation is a critical component of vaccination, such that vaccination in the absence of inflammation inducers like adjuvant, limit Ag-specific immune responses (23). However, the presence of adjuvant also influences the ability of effector cells to reach the Ag site (12). Inflammation in peripheral tissue sites has multiple roles in initiating local immune responses, including maturation of APCs (24, 25, 26, 27), influencing the cytokine milieu in draining lymphoid organs (28, 29, 30), and activation of local endothelia (31, 32, 33).
Chemokines play a critical role in the distribution of effector cells following Ag-specific priming. A range of chemokines induced during inflammation enhances lymphocyte binding to endothelia (31), and experiments applying pertussis toxin to block chemokine-receptor signal transduction inhibits this tight adhesion (34, 35), correlating with deficient in vivo trafficking and function (36). Thus, chemokines are a valid therapeutic target to manipulate the distribution of selected cell populations in vivo. The data presented in this article studying CCR5 and CCL3 are consistent with similar experiments performed with the chemokines CXCL10 and XCL1, which also enhance effector T cell trafficking to tumors and synergistically enhance adoptive T cell therapy (37, 38). Together these data emphasize that immunotherapies that can generate large numbers of effector T cells in vivo and that simultaneously recruit these effector T cells to the tumor site will significantly enhance immune control of tumors.
CCL3 represents an attractive molecule for intratumoral expression, since it is commonly associated with IFN-γ expression and acts in concert with the cytokine IFN-γ and other chemokines to drive Th1 responses in vivo (39). CCR5 is expressed on the effector-effector/memory population (6, 40, 41) and this interaction is known to be critical for control of T cell responses in a number of models (4, 5, 6, 31). Interestingly, the presence of IL-12 during T cell activation has been described to enhance the differentiation of T cells into a phenotype expressing CCR5 (42). The expression of IL-12 is closely tied to expression of IFN-γ (43) and CCR5 is responsible for T cell migration to tumors in IL-12-treated mice (44).
That CCL3 expression alone significantly extends survival, but does not itself abrogate growth of s.c. B16 tumors is in agreement with published studies (14, 45). It has previously been demonstrated that constitutive expression of CCL3 in tumors causes influx of CD8 T cells (14, 45); however, this influx was transient (45). This could be a result of a changed cytokine environment caused by the new cells, adaptation of the response to a stable chemokine gradient, or reflect a stable balance between new cell influx and existing cell exit or death. We demonstrate that the chemotactic response to recombinant chemokine is, as expected, transient and limits the usefulness of recombinant protein for tumor modification. For these reasons we developed a gene therapy model for treatment of established tumors, where adenoviral vectors were used to deliver CCL3 or control genes into the tumor site. In our hands, adenoviral vector delivery of CCL3 is more effective in delaying tumor growth than constitutive expression of CCL3 in selected B16 clones (data not shown). Similarly adoptive transfer of OT1 T cells to an AdGFP-treated tumor enhanced survival more than adoptive transfer to mice bearing untreated tumors (data not shown). Replication-defective adenoviral vectors have intrinsic immunogenicity (46, 47) and this may explain the enhanced immunogenicity of adenovirus-infected tumors. In this regard, tumors excised from mice within 2–3 wk of adenoviral vector delivery still contained small areas of GFP expression where treated with AdGFP, and culture of tumor cells followed by specific ELISA demonstrated continued expression of CCL3 in tumors treated with AdCCL3 (data not shown). However, we also show that adoptive transfer of OT1 T cells is also enhanced by constitutive expression of chemokines in the absence of adenoviral delivery. Thus, although vector components may play a role in the gene therapy model, chemokine expression significantly enhances the efficacy of Ag-specific adoptive therapy.
In these experiments, animals developed endogenous T cell responses not encoded by the adoptively transferred T cells. The mechanism by which this occurs requires further study. Although the data suggest that epitope spreading has occurred, we have previously demonstrated that expression of CCL3 in tumors, when combined with Ag release, can generate tumor-specific T cell responses that are capable of clearing a subsequent tumor challenge (14). Alternatively the involvement of additional T cell epitopes could be required for the success of the therapy (48), particularly since in our hands Ag-directed therapies for B16 melanoma commonly result in the emergence of Ag escape variants (T. Kottke and R. Vile, manuscript in preparation).
In summary, these data suggest that strategies that aim to generate immunotherapies to treat tumors must take into account the accessibility of the tumor site to immune cells. Vaccination distant from the site of tumor growth may efficiently activate Ag-specific effector cells that may not adequately traffic to the tumor site. Therefore, it may be necessary to incorporate strategies to render effector cells better able to traffic to tumors (49) or to modify the tumor site to increase its accessibility to effectors.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the Mayo Foundation and National Institutes of Health Grant R01 CA94180.
Abbreviations used in this paper: Indo-1AM, indomethacin 1 acetyl ester; CD62L, CD62 ligand; DiI, 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate.