Peripheral blood lymphocytes genetically modified to express the self/tumor antigen MAGE-A3 induce antitumor immune responses in cancer patients

Since tumor-associated antigens (TAAs)¹  have been identified, several small-scale therapeutic vaccination and adoptive transfer studies^2,3  have been performed. These trials have explored different vaccine formulations such as antigenic peptides, proteins, or peptide-pulsed dendritic cells.^2,4,5  Although immune responses have been observed repeatedly, overall, these treatments have induced tumor regression in only 15% to 20% of the vaccinated patients.^2,3  Several factors could be responsible for this limited efficacy, some of them related to the lack of induction of the antivaccine immune response, others related to the large variety of immune escape mechanisms operating in advanced tumors.^6–8 

In recent years, vaccination approaches based on dendritic cell (DC) targeting in vivo have successfully been investigated in tumor murine models.⁹  Most of them are based on the use of model TAA-bound antibodies recognizing phagocytic receptors on DC subsets.⁹  We have recently developed an approach of active vaccination based on a novel in vivo DC targeting strategy by the use of genetically modified lymphocytes (GMLs) expressing a defined TAA.¹⁰  Indeed, murine GMLs expressing the TRP-2 tumor antigen proved able to efficaciously load CD11c⁺CD8α⁺ dendritic cells within secondary lymphoid organs,¹⁰  thus eliciting TRP-2–specific effectors that could control the growth of B16F1 melanomas. The rationale of this strategy was based on clinical observations made on patients who were able to mount a strong transgene-specific response, following the infusion of genetically modified T cells.^11–14  Here, we present the first results of the clinical translation of this approach in 10 patients affected by stage IIIc/IV melanoma expressing gene MAGE-A3.¹  The infusion of autologous GMLs expressing MAGE-A3 and TK elicited circulating anti–MAGE-A3 T lymphocytes in 3 patients. Moreover, we demonstrate that these anti–MAGE-A3 T cells are able to migrate to the tumor site and to contribute to the inflammatory response associated with recognition of the tumor antigen in peripheral tissues (ie, delayed-type hypersensitivity [DTH]). These results, if confirmed on a larger group of patients, will allow treatment of patients affected by hematologic and solid tumors expressing MAGE-A3.

This is a pilot study of active vaccination for patients affected by malignant melanoma (stage IIIc/IV AJCC) with expression of MAGE-A3 tumor antigen, evaluated by reverse-transcription–polymerase chain reaction (RT-PCR) analysis, who have already received one line of chemotherapy and/or biochemotherapy for metastatic disease. The study was approved by the institutional review board (IRB) of the Scientific Institute S. Raffaele and by the Italian Institute of Health (ISS; Rome, Italy) and patient informed consent was obtained in accordance with the Declaration of Helsinki. Treatment consists of 5 intravenous infusions (at biweekly intervals) of escalating numbers (15, 30, and 50 × 10⁶ × 3) of autologous lymphocytes transduced to express MAGE-A3 alone (CIP-18 and CIP-19) or in combination with TK (all the others). The 5 infusions correspond to a cycle. After the completion of the 10-week treatment, the investigator could administer (if clinically indicated) as maintenance therapy additional infusions at the maximum dose level (50 × 10⁶) (at monthly intervals; exceptions to this rule are indicated). The primary end point of the study was safety. Secondary end points were in vivo (ie, DTH) and in vitro immune responses to the vaccine (HSV-TK and MAGE-A3). Based on the results described in this paper, a phase 1/2 clinical trial sponsored by MolMed has recently started.

All patients had progressive disease at study entry and had failed standard therapies, with the exception of patient CIP-19, who was clinically disease-free following biochemotherapy and surgery (Table 1). All patients received the first cycle of treatment. Some patients received additional injections, as indicated in Table 1. Before and after treatment, the patients were tested for their ability to mount a DTH reaction to the MAGE-A3 and TK gene products, and for the presence of circulating anti–MAGE-A3 and anti-TK T cells. Four patients (CIP-6, -13, -21, and -23) received only the first cycle of treatment, whereas the others (CIP-5, -11, -18, -19, -25, and -26) were given additional injections. Patients CIP-5, -6, -11, -13, -21, -23, -25, and 26 were treated with M3TN-transduced lymphocytes. Patients CIP-18 and CIP-19 were treated only with M3-CSM–transduced lymphocytes. DTH was carried out by subcutaneous injection of 1 to 2 × 10⁶ autologous mock-GMLs or M3-GMLs. Twenty-four and 48 hours after injections, erythema and/or induration of at least 5 mm was scored as positive in the absence of reaction at the control injection site. Tumor sites were evaluated by physical examinations and scans performed before and after treatments. Standard definitions of response according to World Health Organization (WHO) criteria were used.¹⁵  Adverse effects were recorded using WHO toxicity criteria. Patient HLA typing was performed to identify the restriction elements involved in the elicited anti-TK and anti–MAGE-A3 immune responses (Table S1, available on the Blood website; see the Supplemental Materials link at the top of the online article).

The human melanoma cell lines CIP-5, CIP-26, and MSR3-mel were established in our laboratory and cultured in IMDM supplemented with 10% FBS. CIP-5-mel expresses HLA-A and -C alleles but not the HLA-B alleles (data not shown) that were restored by IFN-γ treatment (data not shown). Patient peripheral blood mononuclear cells (PBMCs) for immunomonitoring were harvested from blood samples collected 15 days after each infusion during the first cycle of treatment, and 1 month after each infusion for the additional treatments. Exceptions to this rule are indicated. MAGE-3.A26 and MAGE-1.A26 tetramers were done as described.¹⁶  TK.B7 pentamer and MelanA/MART-1.A2 tetramer were from Proimmune (Oxford, United Kingdom) and Beckman Coulter (Hialeah, FL), respectively. All tetramers were coupled to PE-conjugated monoclonal Abs against human CD3, CD4, and CD8, and isotype controls were from BD Biosciences Pharmingen (San Jose, CA). MAGE-3_250-258, MAGE-1_273-282, TK.B7,¹⁴  and MelanA/MART-1_26-35 analog¹⁷  peptides were synthesized by Espikem (Florence, Italy) and by PRIMM (Milan, Italy) and used at 20 mM. Lymphocytes were cultured in IMDM 10% HS. The retroviral vectors SFCMM3 and M3-CSM have previously been described.^18,19  The M3TN retroviral vector codes for MAGE-3 and for the HSV-TK/neo (TN) fusion protein. Retroviral transduction of lymphocytes and lymphoblastoid B-cell lines was performed as previously described.^18,19 

Tetramer staining was performed incubating PBMCs or restimulated lymphocytes for 15 minutes at room temperature. Staining with anti-CD8, anti-CD3, and anti-CD4 mAbs was performed for 20 minutes at 4°C. Tetramer⁺ cells were sorted by anti-PE microbeads and restimulated with irradiated (100 Gy) autologous Epstein-Barr virus (EBV) pulsed with the relevant peptide and allogeneic EBV in the presence of IL-2 (100 U/mL) and IL-7 (10 ng/mL).

PBMCs before and after treatment (2 × 10⁵/well) and targets expressing TK or MAGE-3 or mock-transduced (0.75-1 × 10⁵/well) were added in several replicates to nitrocellulose-bottomed 96-well plates (Millipore, Billerica, MA) coated with 10 mg/mL anti–human IFN-γ mAb (1-D1K; Mabtech, Nacka Strand, Sweden). Biotinylated secondary mAb to IFN-γ (7-B6-1; Mabtech), HRP-streptavidin (1:2000), and AEC-substrate solution (Zymed Laboratories, South San Francisco, CA) were then used according to the manufacturers' recommendations. Spots were evaluated and counted using a computer-assisted video imaging analysis system (Axioplan 2; Zeiss, Heidelberg, Germany).²⁰  Spots released by stimulators and responders alone were subtracted. We scored as positive the samples with several spots greater than 5.

Multiple microcultures were performed for a semiquantitative assessment of TK- and MAGE-A3–specific T-cell precursors before and after treatment and among TILs. Peripheral blood leukocytes (PBLs) before and after treatment (10⁵) or TILs were cultured with irradiated (30 Gy) autologous target cells (10⁵) expressing either HSV-TK or MAGE-A3 in 96-well round-bottomed plates. Forty-eight hours later, rhIL-2 at 10 U/mL was added. For TILs, rhIL-2 (20 U/mL) was added from the beginning of the stimulation. On days 12 to 14, each microwell was divided into 4 equal samples and challenged in duplicate against HSV-TK–transduced, MAGE-A3–transduced, or mock-transduced autologous targets (0.5 × 10⁵). Twenty-four hours later, 100 μL supernatant from each microculture was harvested and the IFN-γ content was measured by enzyme-linked immunosorbent assay (ELISA) assay (BD Biosciences Pharmingen) according to the manufacturer's recommendations. Microcultures were scored positive if the release of IFN-γ on target cells expressing HSV-TK or MAGE-A3 was greater than 50% the IFN-γ release on mock-transduced target cells. The spontaneous release of IFN-γ by stimulators and responders alone was subtracted.

Both enzyme-linked immunospot assay (ELISPOT) and microculture experiments were done twice on the same blood sample to confirm the results. The frequencies of anti–MAGE-A3 T cells were determined by the Poisson distribution analysis: frequency of anti–MAGE-A3 precursors =−ln [(number of wells tested − number of positive wells)/number of wells tested]/number of CD3⁺ T cells/well.²¹ 

Mixed lymphocyte peptide culture (MLPC) was assessed as previously described.¹⁶  Briefly, PBMCs were thawed and incubated for 60 minutes at room temperature at 10⁷ cells/mL in the presence of the relevant peptide (20 μM). Then, the cells were washed and plated (1-2 × 10⁵ cells/well) in the presence of rhIL-2 (20 U/mL), and rhIL-7 (10 ng/mL). On day 7, 50% of the medium was replaced by fresh medium containing peptide. Tetramer labeling was performed on day 14. Frequencies of anti–MAGE-3.A26 and antitumor MLPCs were determined by the Poisson distribution analysis.

PBMCs (5000-20 000 cells/well) were stimulated with irradiated (100 Gy) autologous tumor cells (1000 cells/well) in 96-well round-bottomed plates in the presence of 5 ng/mL rhIL-7.²²  rhIL-2 (10 U/mL) was added 72 hours later. Mixed lymphocyte tumor cell cultures (MLTCs) were restimulated every week. At day 21, aliquots of them were used in an IFN-γ assay in duplicate against autologous tumor, mock-transduced autologous target cells (0.5 × 10⁵), K562 (0.5 × 10⁵), or Cos-7 transfected with HLA-A2603 and MelanA/MART-1. A second IFN-γ assay was performed at day 27. Transfection of Cos-7 cells was performed by the DEAE-dextran-chloroquine method as previously described.²³  Frequencies of antitumor and anti-MelanA/MART-1.A26 T cells were determined by the Poisson distribution analysis. For blocking experiments the following mAbs were used: anti–MHC-I (W6/32), anti–HLA-DR (L243), anti–HLA-DP (B7/21 clone), and anti-NGFr (negative control).

Peripheral and tumor-infiltrating MAGE-3.A26 tetramer⁺ lymphocytes were analyzed for the expression of TCR Vβ chain using the TCR Vβ repertoire Kit (IOTest Beta Mark; Beckman Coulter) and by RT-PCR. Total RNA was extracted by TRIzol (Invitrogen, Frederick, MD). cDNA was amplified using upstream primer specific for Vβ7 and a downstream C primer of TCR V region. PCR products were sequenced to obtain the complete identification of the CDR3 region. Alignment was performed using the International Immunogenetics Database (http://imgt.cines.fr; accession no. X61444).^24–25 

Immunohistochemistry was performed on 5-μm tissue sections previously fixed in buffered formalin and embedded in paraffin. Immunostainings were performed with a nonbiotin detection system (Refine polymer; Novocastra, Newcastle upon Tyne, United Kingdom) and diaminobenzidine development. The following Abs were used: anti–HLA-I (clone HC10; gift of Soldano Ferrone, Cancer Institute, University of Pittsburgh, Pittsburgh, PA), anti-CD8 (clone 1A5), anti–β2-microglobulin (polyclonal; Dako, Carpinteria, CA), pan anti–HLA-I (W6/32), and anti–HLA-A30 (One Lambda). Staining was then performed with the indicated antibodies for 30 minutes and analyzed with a Nikon DS-5M CCD camera (Calenzano, Italy), using a 20×/0.75 NA objective on a Nikon Eclipse 80i microscope. Images were taken and deconvoluted with the software provided by Nikon.

Statistical significance (P < .05) was determined by performing ANOVA test followed by Dunnett as posttest, or by 2-tailed Student t test.

No toxicity or severe side effects attributable to the infusion of transduced lymphocytes were observed. Although this study was not designed to assess clinical end points, patients' progress was carefully monitored. In 3 patients, we observed indication of possible clinical benefit (Table 1). Patient CIP-5 experienced a long-lasting stable disease (SD; 50 months); patient CIP-26 achieved 2 complete responses (a first CR of 6 and a second CR of 3 months, following relapse); patient CIP-19, disease free at study entry, remained disease free after 61 months of follow-up (NED, 61 months⁺). Importantly, patients did not receive other therapies during the vaccination course.

During the first cycle of vaccination, 7 of 8 patients developed TK-specific T cells detectable either with an ex vivo IFN-γ ELISPOT assay or with a semiquantitative recall assay consisting of an IFN-γ release assay performed by ELISA after in vitro stimulation of multiple groups of 10⁵ PBLs with TK-GML (Figure 1A,B). Only one patient failed to respond to the TK antigen (CIP-11; Figure 1B). These TK-specific responses were characterized further for some patients, according to prevaccination and postvaccination PBL availability.

TK recognition was usually restricted by HLA class I (HLA-I) molecules, such as HLA-A26, -B44, and -B51 in patient CIP-5 (Figure 1C), HLA-B51 and Cw6 in patient CIP-21 (Figure 1D), or HLA-B7 in patient CIP-25 with the TK_279-288 epitope, which we described previously¹⁴  (Figure 1E). In some patients, we also detected anti-TK CD4⁺ T cells (data not shown), as previously described by others.¹² 

In CIP-23 and CIP-25 patients, we estimated prevaccination and postvaccination frequencies of anti-TK T cells using limiting dilution analysis. In patient CIP-23, the frequency rose from 10⁻⁶ of the CD3 cells before vaccination to 3.5 and 3.7 × 10⁻⁴ after the third and fifth infusion, respectively (Figure 1F). A similar 100-fold increase was observed in patient CIP-25 (from 8.2 × 10⁻⁶ before treatment to 4.6 × 10⁻⁴ and 2.8 × 10⁻⁴ in the postfourth and postfifth infusions, respectively; Figure 1F). In agreement with the results of the microcultures assay (ie, 0.046% of anti-TK T cells), in the postfourth infusion blood sample of this patient, 0.05% of the CD3⁺ cells were stained by a pentamer specific for the HLA-B7–restricted TK_279-288 peptide (data not shown), indicating that all TK-specific effectors recognized this epitope.

Overall, these results also confirm in metastatic melanoma patients the immunogenicity of TK-expressing GMLs previously observed in leukemic patients.¹⁴ 

We measured the number of circulating anti–MAGE-A3 T cells in patients before and after treatments. In 2 of 10 patients (CIP-6 and -18) MAGE-A3–specific T cells were detected by ex vivo IFN-γ ELISPOT assay, whereas for the other patients we applied the semiquantitative recall assay also used for the detection of TK-specific T cells. We estimated the frequencies of anti–MAGE-A3 T cells only for the latter patients (Table 2).

Patients CIP-6, -11, -13, -21, -23, and -25, had no detectable anti–MAGE-A3 immune response before and after therapy (Table 2). Patient CIP-18 had a transient 4.5-fold increase of MAGE-A3–specific T cells after 2 infusions, which was no longer detected after the subsequent rounds of vaccination (data not shown). Patients CIP-5, -19, and -26 had an increase of circulating anti–MAGE-A3 T cells (Table 2). The clinical and immunologic follow-up of these 3 patients is reported in the subsequent subheadings.

Patient CIP-19, treated when disease-free but at high risk of relapse, showed an increase of blood anti–MAGE-A3 T cells detectable between the fifth and ninth infusion of GMLs (Figure 2A,B). The frequency of anti–MAGE-A3 increased further, reaching 1.7 × 10⁻⁵ of the CD3⁺ cells after the 12th vaccination. This represented a 50-fold increase over the prevaccination frequency (Figure 2B). Microcultures displaying reactivity to MAGE-A3 contained CD4⁺ and CD8⁺ T cells (data not shown). One example is the CD4⁺ microculture no. 8, with an IFN-γ response to MAGE-A3–transduced autologous EBV-B cells, inhibited by anti-DP but not anti-DR or an irrelevant control mAbs (Figure 2C). As patient CIP-19 was typed HLA-DP*1001/DP*0401, effector cells were stimulated with either autologous EBV-B cells pulsed with a known M3.DP*0401 peptide²⁶  or with allogeneic MAGE-A3–transduced EBV-B cells sharing the HLA-DP*1001 allele. The cells recognized the MAGE-A3⁺ HLA-DP*1001 EBV-B cells, but not the M3.DP*0401 peptide (Figure 2C). Furthermore, the recognition of the MAGE-A3⁺ HLA-DP*1001 EBV-B cells was abolished by the addition of anti-DP mAb (data not shown). We conclude that these T cells recognized a yet unknown MAGE-A3 epitope presented by HLA-DP*1001. The increase of circulating MAGE-A3–specific effectors paralleled the development of DTH reactivity to MAGE-A3 (Figure 2B). Indeed, a strong DTH reactivity was detected after 5 infusions, which further increased after subsequent injections (10 mm erythema and induration after fifth, 15 mm after 12th, and 25 mm 9 months after 14th; Figure 3B and data not shown). Limiting dilution analysis of T cells isolated from a skin biopsy of the last DTH reaction provided CD4⁺ T-cell clones (data not shown) that proliferated (Figure 2D) and released IFN-γ (Figure 2E) when challenged with MAGE-A3–expressing cells. The detection of DTH reactivity against MAGE-A3 9 months after the last infusion indicates that this approach is able to establish long-term memory T cells, in agreement with the results obtained in the mouse models.¹⁰ 

Patient CIP-26 (Figure 3A) had a decrease of circulating anti–MAGE-A3 T cells during the first cycle of treatment (2.2 × 10⁻⁶ in the pretreatment vs 1.8 × 10⁻⁷ in the postfifth samples; Figure 3B), followed by a progressive increase reaching 1.8 × 10⁻⁵ in the post-10th blood sample (Figure 3B). Concomitantly, the patient displayed a complete clinical response (Figure 3A) documented with 2 consecutive positron emission tomography (PET)/computed tomography (CT) scans (Figure S1) and associated with vitiligo on the limb previously showing subcutaneous metastases (data not shown). Analysis of microcultures revealed that the MAGE-A3–specific effectors were HLA-A restricted as IFN-γ release was partly blocked by mAb W6/32, but not by mAb 4E, which recognizes only HLA-B and -C alleles (Figure 3C). The patient developed a positive DTH reaction after the seventh infusion concurrently with the increased level of circulating MAGE-A3–specific effectors (5 and 6 mm erythema and induration after the seventh and tenth infusion, respectively) (Figure 3D). Thereafter, the anti–MAGE-A3 T cells persisted at the level of approximately 1.5 × 10⁻⁵, even though the patient relapsed (Figure 3A,B). Subcutaneous and lymph node tumor lesions were resected in June 2005 and in March 2006 (Figure 3A). To enumerate anti–MAGE-A3 T cells among the tumor infiltrating lymphocytes (TILs), we applied the same approach as for the blood lymphocytes. MAGE-A3–specific T cells were 5 to 10 times more frequent in the 2 tumor lesions (0.02% and 0.009%, respectively) than in the blood (0.002%). Interestingly, we compared the specificity of TILs from the 2005 regressing subcutaneous tumor with that of lymphocytes of a reactive inguinal lymph node resected at the same time. Only the TILs released IFN-γ in the presence of M3-GMLs (Figure 3E). Among those TILs, CD8⁺ T cells were the main MAGE-A3 effectors, as shown with purified CD4⁺ and CD8⁺ T-cell populations (Figure 3F). TILs recognized M3-GMLs in HLA-A–restricted manner (data not shown), as already observed in the periphery. We evaluated by immunohistochemistry the expression of HLA-I molecules on the 2 tumor lesions (ie, 2005 and 2006). The first tumor was poorly stained by mAb HC-10 (Figure 4A), which recognizes HLA-B and -C free heavy chains,²⁷  but was stained by anti–β2-microglobulin mAb (Figure 4B), W6/32 (pan HLA-I mAb; Figure 4C), and an anti–HLA-A30 mAb (Figure 4D), suggesting that it was mainly expressing the HLA-A molecules. Tumor cells from the second lesion (2006) were not stained by mAb HC-10 (Figure 4E) and by the anti–β2-microglobulin mAb (Figure 4F). Differently from the first lesion, this tumor displayed some areas not stained by both W6/32 and anti–HLA-A30 mAbs (Figure 4G,H). These results are compatible with the hypothesis that the induced HLA-A–restricted MAGE-A3–specific T cells could have exerted an in vivo antitumor activity (Figure 3E), and selected an HLA-I–loss tumor variant.

This patient received the first cycle of 5 vaccinations with M3-TK-GML, and the blood frequency of anti–MAGE-A3 T cells increased from 4 × 10⁻⁷ to 1.2 × 10⁻⁶ of the CD3⁺ cells (Figure 5B). Since the DTH reactivity observed after this first cycle of treatment was directed mainly against TK (data not shown), the patient was further vaccinated with GMLs expressing only MAGE-A3 (Figure 5A). After 10 additional infusions of M3-GMLs (after 15th, July 2002), the frequency of MAGE-A3–specific T cells increased 100-fold compared with the pretreatment blood sample (Figure 5B). Both purified CD8⁺ and CD4⁺ recognized MAGE-A3 (data not shown). This response paralleled the development of a MAGE-A3–specific DTH reaction (5 mm erythema and induration; Figure 5B). The CD8⁺ effectors recognized the autologous tumor and were directed against epitopes of MAGE-A3 presented by different autologous HLA-I alleles (Figure 5C). Some microcultures, such as microculture no. 10, released IFN-γ when stimulated with Cos-7 cells transfected with MAGE-A3 and HLA-A26, -B44, or -Cw1 but not -A2 constructs (Figure 5C). To identify the peptides presented by HLA-A26 and -B44 molecules, 3 MAGE-A3 cDNA fragments (Figure S2A) were cotransfected with the HLA alleles in Cos-7 cells (Figure S2B,C) and incubated with the different T-cell populations (ie, MAGE-A3.A26 and MAGE-A3.B44). The pattern of recognition of the different fragments indicated that a region between BamHI and EcoRI restriction sites encoded the antigenic peptides. By screening this region for peptides carrying HLA-A*2603 and HLA-B*4402 binding motifs,²⁸  one nonapeptide was identified, FVQENYLEY (amino acids 250-258). This peptide, referred to as M3_250-258, sensitized Cos-7 cells transfected with either HLA-A*2603 or HLA-B*4402 to the recognition by MAGE-A3.A26 and MAGE-A3.B44 T cells, respectively (Figure S2B,C). The identification of this epitope allowed evaluation of whether one of these populations had undergone a predominant expansion. Multiple cultures of PBMCs (10⁵) were stimulated with the peptide M3_250-258 and tested by tetramer staining or IFN-γ release assay on Cos-7 cells transfected with MAGE-A3 and HLA-A*2603 or HLA-B*4402. There was a close match between the blood frequencies of anti–MAGE-A3 T cells and those of the anti-M3_250-258 T cells restricted by HLA-A26 (Figure 5B), whereas the anti–MAGE-A3.B44 T cells were detected at lower levels (between 1.9 × 10⁻⁷ and 6.3 × 10⁻⁷ T cells; Figure 5B), thus indicating that in this patient most of the anti–MAGE-A3 effectors recognized this peptide presented by HLA-A26. The MAGE-A3–specific immune response of patient CIP-5 was long-lasting, as the induced anti–MAGE-A3 memory T cells were still detectable in vitro 1 year after the 15th infusion, performed in 2002 (Figure 5B). From July 2002 to June 2004 patient CIP-5 discontinued the therapy (Figure 5A). Nevertheless, we were still able to detect circulating MAGE-A3–specific effectors (ranging from 5.2 × 10⁻⁶ to 2.07 × 10⁻⁶; Figure 5B) and the patient remained clinically stable for 50 months (Figure 5A). Then the frequency of the anti–MAGE-A3 T cells declined and the disease progressed (Figure 5A). To boost the anti–MAGE-A3 immune response, the patient received a new cycle of 15 infusions and the anti–MAGE-A3 T cells rose up again to 3.76 × 10⁻⁵ (Figure 5B). These results further support the concept that M3-GMLs are able to induce anti–MAGE-A3 T cells and strongly suggest that the maintenance of such levels (10⁻⁵) of circulating anti–MAGE-A3 T cells requires multiple and frequent booster injections. To estimate the blood frequencies of T cells against tumor antigens other than MAGE-A3 we set up MLTCs, taking advantage of an autologous tumor cell line (CIP-5-mel) established from a metastasis resected prior to vaccination. Tumor-specific T cells were already present at a frequency of 10⁻⁴ before vaccination (Figure S3), and this frequency did not increase after vaccination (Figure S3). We identified some of the target antigens: the cancer germline protein MAGE-A1¹  and the melanocytic differentiation antigen MelanA/MART-1.^23,29  We characterized CD8⁺ T cells recognizing a MAGE-A1 peptide presented by HLA-A26 (C.T., S. Tanzarella, R.F., and V.R., manuscript in preparation) and MelanA/MART-1 peptides presented by either HLA-A2 or HLA-A26. Using multiple MLTCs or multiple mixed lymphocyte-peptide cultures, we estimated the frequencies of blood T cells recognizing MAGE-A1.A26, MelanA/MART-1.A2, and MelanA/MART-1.A26 (Figure S3), with the latter showing an increase 1 year after the 15th infusion (up to 10⁻⁴; Figure S3). The patient had an abdominal lymph node metastasis resected in 1998, before study, and other abdominal metastases were resected in 2005, after a long-lasting stable disease status. This allowed us to perform a comparative analysis of TILs collected before and after treatment and to evaluate the number and function of anti–MAGE-A3 (ie, antivaccine) and other antitumor effectors inside the lesions (Figure 5A; Table 2). Histologic analysis of the 2005 lesions showed approximately 90% of necrotic tumor cells, with living cells maintaining the expression of MAGE-A3, as evaluated by RT-PCR analysis (data not shown). TILs from the pretreatment lesion did not contain detectable anti–MAGE-A3.A26 and anti–MAGE-A3.B44 T cells (Table 3; Figure S4A). But the postvaccination metastasis contained 0.04% of anti–MAGE-A3.A26 T cells (Table 3; Figure S4B) and only 0.00074% of anti–MAGE-A3.B44 T cells (Table 3). Most of the anti–MAGE-A3.A26 TILs were functional, producing IFN-γ when stimulated with Cos-7 cells transfected with HLA-A26 and MAGE-A3 cDNAs or with the autologous tumor cells (Figure S4C). The frequencies of the other antitumor T cells were similar (0.02%) within the metastases removed before and after vaccination (Table 3), with the exception of anti–MelanA/MART-1.A26–specific T cells increasing from less than 0.00075% to 0.0041% (Table 3). To prove that at least some of the blood anti–MAGE-A3 T cells were recruited at the tumor site, we used tetramers to sort anti–MAGE-A3.A26 CD8⁺ T cells from blood and TILs (Figure S5A), and identified their TCRβ sequences. The 2 types of cells expressed the same Vβ7.1 rearrangement, as demonstrated by spectratyping analysis (data not shown) and by cDNA sequencing (Figure S5B).

Altogether these data show that GML vaccination induces functional vaccine-specific T cells capable of migrating to inflamed skin and, more importantly, to infiltrate tumor lesions.

In this study we have analyzed vaccine- and tumor-specific immune responses of tumor-bearing patients treated with a novel vaccination strategy based on the infusion of autologous lymphocytes genetically modified to express the cancer germline gene MAGE-A3 and the viral antigen TK. The use of such a strategy relies on clinical^12,14  and preclinical¹⁰  in vivo studies, demonstrating that genetically engineered lymphocytes, which are currently used as effector cells in adoptive immunotherapy protocols,^13,30  can act as efficient vehicles for in vivo loading of DCs and consequently can behave as a vaccine, eliciting TAA-specific T cells.

This vaccination strategy may account for different advantages compared with other approaches. Some are related to the previously identified ability of GMLs to target DCs in vivo,¹⁰  thus exploiting a physiologic route of immunization and “selecting” the most appropriate subset of lymphoid organ-resident DCs.⁹  Others are associated with the unexpected finding that contacts between GMLs and DCs triggers DC activation,¹⁰  without further addition of adjuvants. Lastly, there are advantages over treatments based on subcutaneous/intradermal injection of single tumor-derived peptides which include the intravenous route of administration, allowing antigens delivery not only in the draining lymph nodes³¹  but in the whole lymphoid compartment, and the expression of the entire TAA protein in vivo, favoring the presentation of an array of epitopes by the autologous HLA-I and -II alleles, as observed in patients CIP-5 and -19, thus leading to the spontaneous selection of the most frequent and effective CD8⁺ and CD4⁺ T-cell repertoire, and reducing the risk of tumor escape and the induction of anergy/tolerance of TAA-specific T cells.^6,7,32 

The infusion of GMLs expressing MAGE-A3 and TK antigens was effective as it induced vaccine-specific CD4⁺ and CD8⁺ T cells and established antigen-specific long-term memory T cells. Indeed, all but one of the vaccinated patients (CIP-11) showed the increase of circulating TK-specific effectors (Figure 1A,B; Table 1), thus confirming that this population of patients was immunocompetent and, therefore, appropriate for testing TAA-specific responses in vivo. Of this immunocompetent population, only 3 of 9 patients were able to mount a response against the TAA, detectable in the blood. Of note, these were the only patients with limited tumor burden. The reason why the other 6 patients completely failed to show any response is currently not known. However, the careful analysis of the 3 positive patients and its correlation with the clinical course may shed some light on the possible mechanisms for tumor escape in the whole patient population.

The observed rate of immunization and the frequencies of anti–MAGE-A3 T cells in the responding patients are in the lower range of frequencies reported by others³³  (ie, from > 10-2 to 10-6). However, in the majority of these studies, the induction of vaccine-specific T-cell effectors did not correlate with the clinical outcome.³³  In contrast, in our study, the 3 patients showing a sustained increase of anti–MAGE-A3 T cells (ie, CIP-5, -19, and -26) experienced a more favorable clinical outcome. Based on these observations, we hypothesize that GML-elicited effector T cells could be qualitatively different compared with those elicited by other strategies. This hypothesis is supported by 2 observations. First, the 3 patients showing a sustained increase of anti–MAGE-A3 T cells also developed vaccine-specific DTH reactions, a phenomenon that has been correlated to a favorable clinical outcome³⁴ ; second, the circulating anti–MAGE-A3 T cells were also able to infiltrate tumor lesions. In particular, in 2 patients (CIP-26 and CIP-5) we observed an enrichment of anti–MAGE-A3 T cells among TILs collected after treatment. In patient CIP-26, the anti–MAGE-A3 T cells infiltrating the tumor were 10 and 5 times more frequent than in the blood. Furthermore, the ex vivo analysis for the presence of MAGE-A3–specific effectors in both a regressing tumor lesion and a reactive inguinal lymph node clearly showed the presence of these cells only in the tumor (Figure 3E). In patient CIP-5, we identified functional MAGE-A3.A26–specific effectors only among TILs of the tumor collected after treatment. The analysis of the TCRβ sequences indicated that these effectors expressed the same Vβ7.1 rearrangement as those isolated from the blood, therefore demonstrating that the elicited anti–MAGE-A3 T cells had migrated to tumors. Based on these observations, we started collecting tumor samples of patients who did not show circulating anti–MAGE-A3 T cells to evaluate their possible presence inside the tumor. Preliminary results in patient CIP-25 indicated that some MAGE-A3–specific effectors inside the tumor were detectable after one round of in vitro stimulation, however they failed to expand after subsequent restimulations (data not shown). On the contrary, we were consistently able to expand the MAGE-A3–specific effectors from tumor lesions of patients with high frequency of anti–MAGE-A3 T cells in the blood (ie, CIP-5 and CIP-26). These results suggest the presence, in patient CIP-25, of antitumor T cells undergoing anergy within tumor. These data, if confirmed on a larger group of patients, would further extend the notion that the tumor microenvironment may be detrimental for a sustained activity of antitumor T-cell effectors.

In patient CIP-5, we observed a strong increase of the antivaccine T cells (MAGE-A3.A26; 0.04%) but only a modest amplification of the anti–MelanA/MART-1.A26 T cells (0.0041%), in a metastasis collected after treatment. An opposite scenario has recently been described in metastases of MAGE-A3–vaccinated patients, with strong increase of antitumor specificities, but a modest amplification of the MAGE-A3–specific effectors.^35,36  The differences between these studies may rely on the method used to quantify antitumor immune responses, by the fitness of the elicited MAGE-A3–specific effectors or may depend on the status of tumor. Indeed, the tumor metastasis of patient CIP-5 was made up mainly of necrotic tumor cells (approximately 90%, data not shown), thus suggesting that most effector cells could have already left the tumor following extensive tumor cell killing. In this regard, it is worth remembering that these analyses “photograph” a static process occurring in the tumor. Ex vivo dynamic studies could, in the future, also contribute to unravel some important aspects related to kinetics of the immune response within the tumor microenvironment. Nevertheless, our data would suggest that different situations may take place in tumor lesions of patients treated with immunotherapeutic approaches and that the identification of the underlying mechanisms may further improve understanding the biology of the antitumor immune response and therefore the development of more effective treatments.

This study demonstrates that GMLs administered intravenously behave as a vaccine eliciting immune responses against the self-/tumor antigen MAGE-A3. Moreover, the induced anti–MAGE-A3 T cells are able to traffic through inflamed tissues and to infiltrate tumor sites. The results described in this report deserve to be investigated on a larger group of patients, particularly those affected by early-stage diseases, to evaluate the clinical efficacy of such a vaccination strategy.

We thank M. P. Protti (Cancer Immunotherapy and Gene Therapy Program) for discussions and critical comments. In addition, we thank P. van der Bruggen for providing us with the MAGE-3.DP*0401 peptide and K. Fleishhauer and B. Mazzi for HLA typing.

This study was supported by the Italian Association for Cancer Research (AIRC; Milan, Italy) and by European Commission (EC; Brussels, Belgium) Project “Cancerimmunotherapy” LSHC-CT-2006-518334.

Contribution: R.F. designed and performed research and analyzed data; M.B., C.B., and C.T. designed research, analyzed data, and wrote the paper; A.C., L.R., C.R., D.M., F.L., and F.C. performed research and analyzed data; S.M. and D.C. contributed vital new reagents; C.D. contributed vital new reagents and analytical tools; P.G.C. contributed vital new reagents, analyzed data, and wrote the paper; and V.R. designed and performed research, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: F.C. has declared financial interest in MolMed Spa, whose potential product was studied in the present work. C.T. and C.B. are employees of MolMed Spa, whose potential product was studied in the present work. The remaining authors declare no competing financial interests.

Correspondence: Claudio Bordignon, Scientific Institute S Raffaele, via Olgettina 58, 20132 Milan, Italy; e-mail: claudio.bordignon@hsr.it.