Vaccination with DNA Encoding a Single-Chain TCR Fusion Protein Induces Anticlonotypic Immunity and Protects against T-Cell Lymphoma¹

The structural diversity of the TCR³ α and β chains is generated by somatic recombination of V(D)J gene segments (1). The unique combination of these genetic elements gives rise to Id determinants that are expressed clonotypically by the T cell. When expressed by malignant or autoreactive T-cell clones, these Id determinants can serve as targets for active immunotherapy. The Id immunoglobulin of malignant B cells provides a similar molecular target. Vaccination strategies against B-cell Id have provided encouraging results in both preclinical models and human trials (2, 3, 4, 5). However, the difficulty of preparing individual Id proteins has driven the development of DNA vaccines that include the encoding V-region genes (6, 7). Although V-region sequences are weak tumor antigens, when fused to pathogen-derived sequences, they can elicit strong protective immunity-mediated primarily through anti-Id antibody (8, 9). Such fusion vaccines are now in Phase I/II trials for follicular lymphoma and multiple myeloma.

The greater difficulty of producing Id protein vaccines from the TCR has slowed investigation of this approach. However, recombinant TCR protein vaccines have been shown to prevent experimental autoimmune diseases and the development of a murine TCL (10, 11, 12). Here we extend the concept of DNA fusion gene vaccination to T-cell malignancies. We delineate two critical elements required for the vaccine, including a minimal VαVβCβ sequence able to encode a scTCR with conformational integrity as predicted from crystallographic studies (13, 14, 15). Our data reveal that products of VαVβ or VβCβ DNA constructs have conformational deficits and fail to induce effective immunity. The second element necessary for a successful vaccine is fusion of the VαVβCβ sequence to FrC, without which the scTCR fails to generate antibody or protective responses.

C6VL, a radiation-induced thymoma/lymphoma derived from the C57BL/Ka mouse strain (16), was kindly provided by Craig Okada (Stanford, CA). C6VL cells have a mature T-cell phenotype expressing TCRαβ, CD3, CD4, and H-2b. The TCR α chain is composed of AV9*04, AJ56, and the β-chain of BV19*01, BD1*01, and BJ1–3*01. TCL-1 arose spontaneously in a C57BL/Ka mouse bred in house and served as a control in tumor challenge experiments. 293 cells were purchased from the European Collection of Animal Cell Cultures (Salisbury, United Kingdom).

The TCR Vα and Vβ domains for C6VL were spliced together with a (Gly₄Ser)₃ peptide linker to generate the scTCR sequence VαVβ (Fig. 1,a) by primer overlap extension, using the paired primers: VA leader, 5′-AAGCTTAGCATGCTCCTGGCACTCCTCCC-3′; JA linker, 5′-AGAGCCACCTCCGCCTGAACCGCCTCCACCTGGTAT AACACTCAGAAC-3′; VB linker, 5′-GGCGGAGGTGGCTCTGGCGGTGGCGGATCGATCATTACTCAGACACCC-3′; and JB reverse, 5′-GCTAGCTACAACAATGAGCCGGCT-3′. To generate VαVβCβ (Fig. 1 a), the Vα domain was linked to the entire β-chain by extending the β-chain sequences to the residue right before the terminal cysteine, using the CB reverse primer, 5′-GCTAGCGTCTGCTCGGCCCCAGGCC-3′. For VβCβ, 5′-AAGCTTAGCATGAACAAGTGGGTTTTCTGC-3′ (encompassing the natural leader sequence for the Vβ-chain) was paired with the CB reverse primer. Assembled genes were subcloned into the expression vector pcDNA3 (Invitrogen, Leek, the Netherlands) as HindIII-NheI fragments in-frame and 5′ to the coding sequence for FrC (aa 865-1316 of tetanus toxin). Construction of p.FrC with a BCL1 V_H gene-derived leader sequence has been described previously (8). Plasmid DNA was purified for vaccination using a QIAfilter Giga kit (Qiagen, Hilden, Germany). All constructs were sequenced and checked for expression in vitro using the TNT T7 Coupled Reticulocyte Lysate System (Promega, Southampton, United Kingdom).

The ability of scTCR-FrC fusion constructs to synthesize and export protein was assessed after transient transfection into 293 cells by lipofection (Effectene; Qiagen). Supernatants were collected after 48 h, and the secreted protein captured onto ELISA plates with a monoclonal anti-FrC antibody (E6C, in house), coated at 5 μg/ml. Bound protein was detected with biotinylated mAb 124-40 (C6VL Vα specific; kindly provided by Ron Levy, Stanford, CA), RR4–7 (VB6 specific; Becton Dickinson, Oxford, United Kingdom), or polyclonal mouse anti-FrC antibody, followed by Streptavidin-horseradish peroxidase (Cambridge Bioscience, Cambridge, United Kingdom).

Recombinant VαVβCβ protein, biotinylated at the COOH terminus in vivo, was expressed in Drosophila melanogaster S2 cells. Two synthetic oligonucleotides encoding a biotin-acceptor sequence (17) were annealed and then ligated directly into pMT/Bip/V5-His (Invitrogen) as a NotI-AgeI fragment, generating the construct pMT-BA. VαVβCβ rDNA was then subcloned into pMT-BA as a SpeI-NotI fragment. A PCR product encompassing Escherichia coli biotin ligase was amplified from the plasmid pACYC184 (Avidity, Denver, CO) and subcloned into pMT/Bip/V5-His as an NcoI-NotI fragment. The VαVβCβ and biotin ligase constructs were cotransfected into S2 cells (Invitrogen) by calcium phosphate precipitation together with pCoHYGRO (Invitrogen). Stable transfectants, derived by selection in Schneider Drosophila expression medium containing 300 μg/ml hygromycin-B, were expanded in protein-free medium (Insect Xpress; Biowhittaker, Wokingham, United Kingdom), supplemented with 50 μm D-biotin. Protein expression was induced over a period of 72 h by the addition of copper sulfate to a final concentration of 0.7 μm. Supernatants were concentrated by centrifugal filtration (Millipore, Watford, United Kingdom) and then dialyzed against PBS. Biotin-tagged scTCR was then purified on a Softlink monomeric avidin column as recommended by the manufacturer (Promega).

C57BL/Ka mice, bred in house, were vaccinated at 8–10 weeks of age with a total of 50 μg of DNA in normal saline injected into two sites in the quadriceps. Injections were on days 0, 21, and 42. To assess antibody responses, tail bleeds were taken on days 35 and 56. Three weeks after the last immunization, mice were challenged with an i.v. injection of 750 C6VL cells. Cell depletion experiments were performed in vivo by i.p. injection of 100 μg of rat antimouse CD8α mAb (clone YTS 169.4.2.1, kindly provided by Dr. S. Cobbold, Sir William Dunn School of Pathology, Oxford, United Kingdom) or an isotype control. Mice were injected every 2–3 days over a fortnightly period beginning 1 week before tumor challenge.

Anti-FrC antibodies were measured by ELISA as described previously (8). For the measurement of antibody responses against the TCR of C6VL, biotin-tagged recombinant VαVβCβ protein was captured onto streptavidin-coated plates at 0.5 μg/ml. Mouse immune serum was titered over four wells in 4-fold dilutions. A pooled terminal serum was assigned an arbitrary value of 200 units/ml to act as a comparative standard. Bound mouse IgG was detected with horseradish peroxidase sheep antimouse Fcγ (The Binding Site, Birmingham, United Kingdom).

Mice were immunized with VαVβCβ-FrC or A31 scFv-FrC as described above. On day 56, mice were sacrificed, and serum was collected and pooled. Total IgG was purified on a protein G column (Amersham Pharmacia Biotech, Little Chalfont, United Kingdom) and dialyzed against PBS. Mice were injected i.p. with 100 μg of total IgG, followed by an i.v. challenge with 1000 C6VL cells.

To investigate DNA vaccination against T-cell malignancies, V-region sequences from the murine TCL cell line C6VL were assembled as scTCR by PCR and cloned in-frame and 5′ to the coding sequence for FrC. The Vα domain was spliced via a (Gly₄Ser)₃ flexible linker to either the Vβ domain alone or to the entire β-chain (minus the transmembrane/cytoplasmic regions), generating the constructs VαVβ-FrC and VαVβCβ-FrC, respectively (Fig. 1,a). To assess whether Vα was required to stabilize Vβ folding, a VβCβ-FrC fusion construct was also assembled (Fig. 1 a).

We initially evaluated the structural integrity of the scTCR DNA vaccine products by transfecting the constructs into 293 cells. All supernatants were found to contain levels of FrC protein similar to those produced from a plasmid containing the FrC gene alone (Fig. 1,b). Using a sandwich ELISA with an anti-FrC mAb as capturing antibody and the conformation-dependent, Vβ6-specific mAb RR4-7 for detection (18), we demonstrated clear reactivity with expressed VαVβCβ-FrC (Fig. 1,b). There was also partial reactivity with VβCβ-FrC, indicating limited Vβ-folding in the absence of the Vα domain. VαVβ-FrC protein failed to react with RR4–7 (Fig. 1,b), suggesting inappropriate folding of this molecule. Reactivity with mAb 124-40, which recognizes a clonotypic determinant on the C6VL Vα chain (16), could only be demonstrated to VαVβCβ-FrC (Fig. 1 b). These findings confirm the premise that Cβ is required to stabilize Vβ-folding (19, 20).

We next investigated whether the scTCR-FrC fusion constructs would induce antibody reactive with the C6VL lymphoma cells. Mice were vaccinated i.m. with 50 μg of plasmid DNA on days 0, 21, and 42. Serum samples were collected on day 56 and pooled and analyzed for anti-C6VL antibodies by flow cytometry (Fig. 2). Although all FrC-containing constructs generated antibodies against FrC (Table 1), only those animals vaccinated with VαVβCβ-FrC were able to induce antibody reactive with C6VL cells (Fig. 2). An absolute requirement for the scTCR to maintain a native conformation was highlighted by the failure of VαVβ-FrC or VβCβ-FrC to induce anti-C6VL antibodies. The essential role of the FrC sequence was shown by the failure of the VαVβCβ vaccine to induce anti-C6VL antibody (Fig. 2). Importantly, immune serum raised by vaccination with VαVβCβ-FrC was unable to stain peripheral lymphocytes from nonimmunized mice (data not shown), suggesting that the antibody response was directed against private Id determinants. Analysis of pooled sera for immunoglobulin subclasses showed comparable levels of IgG1 and IgG2a (0.8:1), consistent with a mixed T_H1/T_H2-dominated response.

The VαVβCβ-FrC vaccine induced antibody recognizing C6VL cells in only 42% of mice (Table 1). In contrast, sera from all mice recognized recombinant VαVβCβ protein by ELISA. Surprisingly, however, immune sera from mice vaccinated with VαVβ-FrC or VβCβ-FrC constructs, known to produce mainly misfolded protein, also recognized recombinant VαVβCβ protein (Table 1). The antibody response was specific for C6VL, with no reactivity against VαVβCβ protein prepared from an unrelated T-cell line (data not shown). This indicates that the recombinant VαVβCβ protein contains some misfolded molecules, possibly a common feature of recombinant proteins. It points to a requirement to confirm reactivity of apparent anti-Id antibodies with native cell-expressed Id determinants.

The relevance of conformational integrity for induction of protective immunity against C6VL lymphoma was then addressed. C6VL is a highly aggressive tumor, with as few as 50 cells killing 100% of recipient mice (data not shown). Vaccination with the DNA construct VαVβCβ-FrC induced strong protection against challenge with C6VL (Fig. 3 a), compared with nonvaccinated control mice and mice vaccinated with FrC alone (P < 0.001). Vaccinations with VαVβ-FrC, VβCβ-FrC, or VαVβCβ alone were also completely ineffective. The protective response generated was specific for C6VL, with no protection induced against TCL-1, an unrelated T-cell tumor (data not shown). The pattern of protection has been confirmed in two subsequent experiments, with survival rates of mice vaccinated with VαVβCβ-FrC ranging from 50 to 80%, whereas the other constructs were ineffective at prolonging survival.

One potential problem with this approach is that the DNA vaccine could raise an autoimmune response capable of eliminating normal T cells expressing the same Vα or Vβ framework determinants. To assess if vaccination with VαVβCβ-FrC caused a change in TCR usage, peripheral lymphocytes from immunized mice were processed 3 weeks after the final vaccination and stained with RR4–7. Cells expressing Vβ6 were still detectable at levels comparable with those found in nonvaccinated mice or mice vaccinated with FrC alone, indicating no autoimmune effect (data not shown).

Previous studies using a TCR protein vaccine against C6VL lymphoma had indicated a role for CD8⁺ T cells in mediating suppression of tumor growth (12, 21). We investigated this effector pathway by in vivo depletion of CD8⁺ cells with mAb YTS 169.4.2.1. Flow cytometric analysis of peripheral lymphocytes 1 day before tumor challenge confirmed >98% depletion of CD8⁺ T cells (data not shown). However, depletion did not compromise protection against tumor (Fig. 3 b). Thus, despite the known efficacy of DNA vaccines in inducing CD8⁺ T-cell responses, there was no evidence for involvement of this pathway with TCR as antigen. The critical requirement for native folding of the scTCR points to B cells being involved in the protective response. This was also suggested for TCR protein vaccines, which failed to induce immunity in B cell-deficient (J_H^−/−) mice (12, 21). However, B cells could be involved in antigen presentation, as well as antibody production. In our vaccinated mice, there was no clear correlation between anti-C6VL antibody levels and survival, and some mice with no detectable anti-Id were protected. This raises the possibility of a role for anti-Id CD4⁺ T cells, known to be capable of mediating protection against B-cell malignancies (22, 23). Parallel depletion of CD4⁺ T lymphocytes could not be performed because of expression of CD4 by C6VL cells.

The ability of immune serum to mediate tumor protection was assessed in vivo by passive antibody transfer. The efficacy of antibody was revealed by the fact that 90% of naive recipients given purified IgG from mice immunized with VαVβCβ-FrC were protected against a subsequent lethal challenge with C6VL cells (Fig. 3,c). Mice similarly treated with total IgG from mice immunized with an irrelevant fusion vaccine were not protected (Fig. 3 c). Therefore, it appears that antibody alone can mediate protection but that it may not be the only mechanism. Evidence from the use of recombinant scTCR as antigen to suppress autoimmune T lymphocytes has implicated CD4⁺ regulatory cells as effectors (11). Consistent with our findings, no antibody was produced by the VαVβ protein used in this study (11). Similarly, recombinant VαVβ scTCR protein from C6VL did not induce antibody or protective immunity (12), again supporting a requirement for conformational integrity. One interpretation is that anti-Id antibody is a highly effective mechanism for attacking T-cell clones and that recombinant protein may be an inefficient inducer because of inappropriate folding. For DNA vaccines where levels of protein expression are low, folding must be optimized, and the presence of the stabilizing Cβ domain is essential. There are likely to be additional cellular mechanisms capable of attacking T-cell targets, and these may be induced by less well-folded protein. However, our analysis reveals the power of anti-Id antibody and points to inclusion of this weapon in our attack against T-cell tumors.

These results reveal a strategy for selectively eliminating T-cell clones by DNA vaccination. A combination of optimally folded scTCR sequence with a pathogen-derived sequence induces antibody and generates protection against TCL. The range of application is wide but may be especially attractive for cutaneous TCLs, where clinical need and a feasible clinical setting are evident.

We thank Kerry Cox, Sam Martin, and Lisa Davis for invaluable technical assistance.