Targeting of GD2-positive tumor cells by human T lymphocytes engineered to express chimeric T-cell receptor genes



Genetic engineering of human T lymphocytes to express tumor antigen-specific chimeric immune receptors is an attractive means for providing large numbers of effector cells for adoptive immunotherapy while bypassing major mechanisms of tumor escape from immune recognition. We have applied this strategy to the targeting of a GD2-positive tumor, neuroblastoma, which is the commonest extracranial solid tumor of childhood. Chimeric immune receptors were generated by joining an extracellular antigen-binding domain derived from either of the 2 ganglioside GD2-specific antibodies sc7A4 and sc14.G2a to a cytoplasmic signaling domain. The variable domains of hybridoma antibody 14.G2a were cloned and selected using a phage display approach. Upon coincubation with GD2-expressing tumor cell targets, human T lymphocytes transduced with recombinant retroviruses encoding chimeric receptors based on sc14.G2a, but not sc7A4, secreted significant levels of cytokines in a pattern comparable to the cytokine response obtained by engagement of the CD3 receptor. T cells transduced with the sc14.G2a-based chimeric T-cell receptors also displayed specific lysis of GD2-positive neuroblastoma cells, which was blocked in the presence of monoclonal antibody 14.G2a. In the absence of nonspecific stimulation of transduced cells, their functionality declined over time and antigenic stimulation of the chimeric receptor alone did not induce commitment to proliferation. These results support the feasibility of redirecting human T lymphocytes to a tumor-associated ganglioside epitope but emphasize that successful chimeric receptor-mediated adoptive immunotherapy will require additional strategies that overcome functional inactivation of gene-modified primary T lymphocytes. © 2001 Wiley-Liss, Inc.

Neuroblastoma is one of the most common solid tumors of childhood. The overall survival for the highest-risk population has remained essentially unchanged at <15% over the last 3 decades. Although dose intensification of chemotherapy has increased initial response rates, this effect has not translated into durable remissions in patients with disseminated disease. Eradication of microscopic foci of disease following cytoreductive chemotherapy by immunotherapeutic measures is an alternative approach to prevent relapses and provide long-term disease control.

Recent experience with immunotherapy for neuroblastoma has shown that this tumor is indeed amenable to immune-mediated cytotoxicity. Treatment of neuroblastoma with an IL-2 gene-transduced autologous tumor cell vaccine has produced clinically effective antitumor immune responses in children with advanced neuroblastoma.1 However, the clinical use of this highly individualized strategy is limited by the requirements for in vitro culture and transduction of each patient's tumor cells. Use of an allogeneic cell vaccine to overcome these obstacles appears to be considerably less effective at raising specific immune responses,2 most likely due to inadequate representation of an individual tumor's antigenic repertoire by the immunizing tumor cell line. Antigen-specific T-cell targeting of neuroblastoma is hampered by the lack of identified immunodominant tumor-specific protein antigens and by the requirement for antigen processing and MHC-restricted antigen presentation.

A gene-therapy strategy has been developed that allows the recognition specificity of T lymphocytes to be extended beyond classical T-cell epitopes.3 T cells are genetically modified to express chimeric receptor genes encoding the variable domains of a tumor-specific monoclonal antibody (MAb) joined to a cytoplasmic signaling domain. Engagement of the extracellular component of the chimeric receptor results in tyrosine phosphorylation of immune-receptor activation motifs present in the cytoplasmic domain, initiating T-cell signaling to the nucleus and exertion of tumor cell-directed effector functions. This strategy is applicable to every malignancy that expresses a tumor-associated antigen for which a MAb exists.4–9 Critically, therefore, tumor targeting by chimeric T-cell receptors (TCRs) is not confined to protein antigen.

As a derivative from embryonic neuroectoderm, neuroblastoma is characterized by abundant expression of the ganglioside antigen GD2.10 Although lacking tumor specificity, the highly restricted expression of GD2 on normal tissue or in an immunoprivileged site of the brain allows an operational window of specificity.

The hybridoma MAbs 3F811 and 14.G2a,12 as well as its human/mouse chimeric variant ch14.18,13 are currently being tested in clinical studies and have shown limited success in the treatment of relapsed neuroblastoma. To enhance antibody-mediated antitumor activity by recruiting T-cell effector mechanisms, we constructed GD2-specific chimeric TCRs from 2 different hybridoma antibodies. Human T lymphocytes transduced to express 14.G2a-based recombinant receptor genes produced specific lysis and cytokine secretion upon exposure to GD2-expressing neuroblastoma cells. However, transduced T cells did not have the capacity to proliferate and expand in response to stimulation with the ganglioside antigen. Furthermore, during prolonged in vitro culture in the absence of nonspecific stimulation, tumor-specific functionality progressively declined, indicating failure of chimeric receptors alone to provide a complete activation stimulus in resting T lymphocytes.


Cell lines and antibodies

The neuroblastoma cell lines LAN-1 and LAN-5 were provided by Dr. Seeger′s laboratory (UCLA, Los Angeles, CA) and JF was established in our laboratory. Other tumor cell lines used as targets were obtained from the ATCC (Rockville, MD) and included IMR-32 and SK-N-SH (neuroblastoma), A-204 (rhabdomyosarcoma), Y-79 (retinoblastoma) and Jurkat (T-cell leukemia). The ecotropic packaging cell line Phoenix14 was provided by Dr. G.P. Nolan (Stanford, CA). The packaging cell line PG13 was obtained from the ATCC. Virus-producing cell lines were cultured in DMEM (BioWhittaker, Walkersville, MD), supplemented with 10% heat-inactivated FBS (Summit, Fort Collins, CO) and 2 mM L-glutamine. The hybridoma cell line 14.G2a (mouse IgG2a;κ)15 was provided by Dr. R.A. Reisfeld (La Jolla, CA). The single-chain antibody scFv 7A416 was provided by Dr. J.-L. Teillaud (Institute Curie, Paris, France). MAbs were purified 14.G2a (Pharmingen, San Diego, CA) and antiidiotypic antibody 1A717 (TriGem; Titan, South San Francisco, CA).

Cloning and phage display selection of 14.G2a variable domains

The genes encoding the VH and the VL domains of MAb 14.G2a were amplified from cDNA prepared from 14.G2a hybridoma cells using a set of murine variable domain-specific primers.18 Primers were modified to generate SfiI restriction sites at one end and complementary linker fragments at the other end of the amplified VH and VL molecules. Combinatorial scFv genes were generated by splicing-by-overlap PCR. scFv fragments were ligated into SfiI sites of the replicative form of fUSE5 vector phage DNA.19 Ninety-three randomly selected clones were screened for specific binding to GD2-expressing target cells by whole-cell ELISA. The cellular specificity of selected clones was confirmed by flow cytometry and plasmid DNA was analyzed by nucleic acid sequencing using insert-flanking primers.

Construction of chimeric receptor genes

scFv 14.G2a and scFv 7A4 DNA sequences were amplified by PCR using primers introducing XbaI and StuI restriction sites at the 5′ and 3′ ends, respectively. The PCR product was ligated into the XbaI and blunted BamI sites of pRSV-γ DNA3 (provided by Dr. Z. Eshhar, Rehovot, Israel) in frame with the DNA coding for the transmembrane and intracellular domains of the human FcϵRI γ-chain. To replace the murine signal peptide preceding the chimeric receptor gene in pRSV-γ, a double-stranded oligonucleotide encoding for the human IgG1 leader peptide was introduced into the SnaBI and XbaI restriction sites. The transmembrane and cytoplasmic portions of the human ζ-chain were amplified from pGEM3zζ.20 5′-BamHI and 3′-XhoI restriction sites introduced by specific primers were used to clone into the respective sites of pRSV-14.G2a-γ and pRSV-7A4-γ after cutting out the human FcϵRI γ-chain component. The truncated variant pRSV-14.G2a-Δγ was engineered by PCR, inserting a stop codon after the first 3 cytoplasmic amino acids of the FcϵRI γ-chain. The DNA sequences and reading frames of all molecules generated by PCR were confirmed by sequence analysis. The chimeric scFv-γ and ζ genes were subsequently subcloned into the BamHI and NcoI sites of the retroviral vector SFG21 (provided by Dr. R.C. Mulligan, Cambridge, MA).

Production of recombinant retrovirus

Cells of the ecotropic packaging cell line Phoenix-eco were transiently transfected with vector DNA using FuGENE6 transfection reagent (Roche, Indianapolis, IN). Retroviral supernatants were collected within 24 hr of addition of IMDM supplemented with 20% FCS at 36 hr following transfection and filtered through a 0.45 μm pore size filter. Fresh virus was used to infect the packaging cell line PG13 in the presence of polybrene (8 μg/ml) for 48 hr at 32°C. After replacing the virus with fresh culture medium, infected cells were incubated overnight at 37°C and then subjected to a second round of infection under the same conditions using freshly generated Phoenix-eco cell supernatants. GALV-pseudotyped viral supernatants were generated on the resulting bulk producer cell lines for 24 hr at 32°C.

Isolation and transduction of lymphocytes

Fresh PBMCs from healthy donors and patients with neuroblastoma were separated on Lymphoprep gradients (Nycomed, Oslo, Norway), depleted of adherent cells and resuspended in culture medium consisting of equal amounts of AIM-V (GIBCO BRL, Gaithersburg, MD) and RPMI-1640 (BioWhittaker, Walkersville, MD) containing recombinant human IL-2 (R&D Systems, Minneapolis, MN) at 100 IU/ml and supplemented with 10% FBS and 2 mM L-glutamine. Cells were prestimulated on a 24-well plate precoated with OKT-3 (1 μg/ml; Ortho, Raritan, NJ) and anti-CD28 antibody (1 μg/ml, Pharmingen) at 1 × 106 cells per well for 72 hr. Transductions were carried out in 24-well non-tissue culture–treated plates (Becton Dickinson, Franklin Lakes, NJ) coated with recombinant FN CH-296 (Retronectin; Takara Shuzo, Otsu, Japan) at a concentration of 4 μg/cm2. Prestimulated T lymphocytes were resuspended at 1 × 106 cells/ml in culture medium containing IL-2 (200 IU/ml) and incubated with equal volumes of freshly generated viral supernatant for 36 hr at 37°C and 10% CO2. No positive selection of transduced cells was performed. Cell cultures were expanded in culture medium and IL-2 (100 IU/ml) by maintaining cell densities of 2 × 106/ml.


Total RNA was extracted from 1 to 5 × 106 transduced cells using TRIZOL reagent (GIBCO BRL). For each sample, 2.5 μg of RNA were treated with amplification-grade DNase (GIBCO BRL) and first-strand cDNA was synthesized using oligo-dT primers (Superscript kit, GIBCO BRL). Amplification was performed using primer 5′14.G2a (5′-TTTCTGTTCTC AAAGTACAC) or 5′7A4 (5′-ATAGATACGACGGGGGCTA), annealing to hypervariable regions within the scFv heavy chain domain, in combination with primer 3′Hinge (5′-GGGCATGTGTGAGTTTTG), annealing to an 18 bp region within the spacer domain. As a housekeeping gene, β2-microglobulin was amplified using primers 5′β2MG (5′-CTTAGCTGTGCTCGCGCTAC) and 3′β2MG (5′-ATGGTTCACACGGCAGGCAT).

Flow cytometry

T lymphocytes transduced with chimeric receptor genes containing scFv 14.G2a were incubated with 14.G2a antiidiotypic antibody 1A7 (200 ng/5 × 105 cells) in the presence of normal goat serum for 20 min on ice, followed by incubation with FITC-labeled goat antimouse antibody (Becton Dickinson, San Jose, CA) for 20 min on ice. For each sample, 10,000 cells were analyzed by FACSCalibur with Cell Quest software (Becton Dickinson). For immunophenotyping, cells were stained with fluorescein-conjugated MAbs (Becton Dickinson) directed against CD3, CD4, CD8, CD16, CD56 and CD25 surface proteins, then analyzed as above.

Measurement of cytokine production

To determine target-specific cytokine release by transduced T lymphocytes, duplicate samples of transduced effector cells (5 × 104/well) were cocultured with various tumor cell targets at E:T ratios of 6:1, 3:1, 1:1 and 1:3 in 96-well round-bottomed plates in 200 μl culture medium per well, containing 100 IU/ml IL-2. As a positive control, transduced cells were incubated in wells of a 96-well flat-bottomed plate precoated with anti-CD3 antibody (OKT-3, 10 μg/ml). After 24 hr, supernatants were harvested and analyzed for human IFN-γ, GM-CSF, IL-4, IL-10 or TNF-α by ELISA (R&D Systems) according to the manufacturer.

Cytotoxicity assays

Cytotoxic specificity was determined in a standard 51Cr-release assay. Various numbers of T-lymphocyte effector cells were coincubated in triplicate with 5,000 target cells labeled with 100 μCi 51Cr/0.5 × 106 cells in a total volume of 200 μl in a V-bottomed 96-well plate. At the end of a 5 hr incubation period at 37°C and 5% CO2, supernatants were harvested and radioactivity was counted in a gamma counter. Maximum release was determined by lysis of target cells with Triton X. Percent specific lysis was calculated as follows: [(test counts – spontaneous counts)/maximum counts – spontaneous counts)] × 100%. For antibody-blocking experiments, labeled target cells were preincubated with various concentrations of 14.G2a MAb 30 min prior to addition of effector cells. For blocking assays with soluble ganglioside antigen, vacuum-dried disialoganglioside GD2 (Sigma-Aldrich, St. Louis, MO) was resuspended in 95% ethanol as described and further diluted in culture medium. Control experiments with equal amounts of ethanol alone were performed in parallel. Activity of the ganglioside was verified by a sensitive capture ELISA, which detected 14.G2a at 1/50,000 MAb dilution (modified from Chu et al.22 and Ravindranath et al.23).

Proliferation assays

Transduced T lymphocytes were coincubated in triplicate at 5 × 104 cells/well with various tumor cell targets at a 3:1 stimulator-to-responder ratio. Following a 72 hr coincubation period, wells were pulsed with 2.5 μCi of [3H]-thymidine for 18 hr and samples were harvested onto glass fiber filter paper for β scintillation counting.


Generation of GD2-specific chimeric T lymphocytes and expression in human T lymphocytes

To generate gene-modified T lymphocytes specific for neuroblastoma, we constructed chimeric TCR genes using extracellular recognition domains derived from either of the 2 anti-GD2 MAbs, 7A4 or 14.G2a. 7A4 was used as a precloned single-chain antibody that had been shown by others to preserve the specificity and most of the affinity of the parental MAb.24 To ensure cloning of the correct genes that code for the variable domains of MAb 14.G2a, we functionally evaluated single clones of recombinant fusion phage fUSE5 displaying linked VH and VL fragments derived from 14.G2a hybridoma transcripts. Among 93 phage clones tested for binding to GD2-expressing cells of the JF neuroblastoma cell line in a whole-cell ELISA, 3 yielded ELISA signals of >0.3 optical density units above background (data not shown). The nucleotide sequences of the 2 strongest binders varied by 4 amino acids, 2 of which were located in the framework portions of the VH and VL domains. Clone sc14.G2aC4 maintained the highest degree of target cell binding related to MAb 14.G2a by flow cytometry (Fig. 1) and was chosen as the recognition portion of the chimeric receptor. Both single-chain domains, sc7A4 and sc14.G2aC4, were linked to the transmembrane and cytoplasmic signaling domains of the TCR ζ-chain or the FcϵRI γ-chain via a human IgG1 hinge-derived spacer domain. Chimeric receptor sequences were preceded by a human IgG signal peptide.

Figure 1.

Binding of MAb 14.G2a and fusion phage clone 14.G2aC4 to JF neuroblastoma cells. Cells were stained with MAb 14.G2a (dotted line), fUSE5/14.G2aC4 and secondary antibody anti-M13 (dashed line) or an IgG2a isotype control antibody (solid line), followed by incubation with a FITC-labeled goat antimouse antibody and flow-cytometric analysis.

Retroviral gene transfer into in vitro activated human T lymphocytes was achieved using the vector SFG and a GALV-pseudotyped producer cell line. Average T-cell transduction efficiencies, as assessed by quantification of cells stained with anti-14.G2a idiotypic antibody 1A7, were 35% (range 32–39%) for 14.G2a-γ (Fig. 2a) and 30% (range 24–38%) for 14.G2a-ζ. With both constructs, chimeric receptor expression was maintained for extended culture periods of up to 45 days without any observable downregulation (Fig. 2b). Staining of nontransduced T lymphocytes with antibody 1A7 did not yield fluorescence intensities above the levels obtained with an isotype control antibody. CD4+ and CD8+ T lymphocytes within the cultured population were transduced equally well (data not shown). The growth characteristics, as assessed by weekly counting of viable cells and the immunophenotype of the transduced T cells during in vitro culture in the presence of IL-2 did not differ from nontransduced T-cell cultures (data not shown).

Figure 2.

(a) Surface immunofluorescence of in vitro expanded peripheral blood lymphocytes 8 days after retroviral transduction with 14.G2a-γ chimeric receptor genes. Cells were stained with MAb 1A7 (dashed line) or IgG1 isotype antibody (solid line), followed by incubation with FITC-labeled goat antimouse antibody. (b) Surface immunofluorescence of in vitro expanded peripheral blood lymphocytes at different time points after retroviral transduction with 14.G2a-γ or 14.G2a-ζ chimeric receptor genes. At left is the percentage of fluorescence-positive cells and at right is the median fluorescence absorbance.

The presence of the chimeric receptor constructs in the genome of transduced T lymphocytes and the expression of chimeric receptor RNA were demonstrated by PCR (not shown) and RT-PCR (Fig. 3) using specific primers. Chimeric receptor proteins were detected in lysates of transduced cells by immunoblotting and probing with TCR ζ- and FcϵRI γ-chain-specific antibodies (data not shown).

Figure 3.

Expression of chimeric receptor gene RNA in transduced cells. RNA was extracted from equivalent numbers of transduced T lymphocytes, transcribed into cDNA and amplified using primers annealing to hypervariable regions within scFv 7A4 (a) or 14.G2a (b) and to the IgG1 hinge region of the chimeric receptor cDNA. Lanes 1 and 14, 100 bp DNA ladder; lanes 2 and 9, water controls; lane 3, nontransduced cells; lane 5, cells transduced with SFG/7A4-γ; lane 7, cells transduced with SFG/7A4-ζ; lane 10, cells transduced with SFG/14.G2a-γ; lane 12, cells transduced with SFG/14.G2a-ζ. Lanes 4, 6, 8, 11, 13, nontranscribed RNA controls.

Cytokine secretion by chimeric receptor-transduced T cells upon stimulation with GD2-expressing tumor cells

The specific functional activity of chimeric receptor-transduced T cells was determined by measuring cytokine release by modified T lymphocytes from 4 healthy donors following coincubation with tumor target cells. 14.G2a-ζ- and 14.G2a-γ-transduced T lymphocytes released IFN-γ (up to 11,900 pg/ml for 106 cells over 24 hr) (Fig. 4), GM-CSF (up to 5,340 pg/ml for 106 cells over 24 hr), TNF-α (up to 420 pg/ml for 106 cells over 24 hr) and IL-10 (up to 840 pg/ml for 106 cells over 24 hr) upon stimulation with GD2-positive target cells. No IL-4 was detected in the supernatants of stimulated cells. A similar pattern of cytokine secretion was seen after stimulation of transduced and nontransduced PBMCs by incubation on immobilized anti-CD3 antibody. Chimeric receptor-modified cells did not differ from nontransduced T lymphocytes in their ability to respond to stimulation by anti-CD3 antibody. The amounts of secreted cytokines by the modified effector cells correlated with the GD2 expression level on target cells, quantified by flow-cytometric analysis of cells stained with MAb 14.G2a (Table I). No specific cytokine secretion was found in the supernatants of nontransduced T cells or T cells transduced with EGF protein or the truncated receptor variant (not shown). T lymphocytes transduced with 14.G2a-based chimeric receptor genes containing the TCR ζ-chain yielded significantly higher cytokine release compared to chimeric receptors in which the signaling domain was provided by FcϵR γ (p < 0.005) (Table II). In contrast, expression of chimeric receptor genes based on MAb 7A4 did not provide transduced T lymphocytes with any specific antitumor activity, demonstrated by the lack of tumor-specific cytokine release (not shown).

Figure 4.

IFN-γ release by T lymphocytes transduced with SFG/14.G2a-ζ (a) and SFG/14.G2a-γ (b) in response to coincubation with tumor target cells. On day 8 after transduction, 14.G2a-γ and 14.G2a-ζ–transduced T lymphocytes were cocultured with tumor cells at the indicated E:T ratios for 24 hr. LAN-5, JF and IMR-32 are GD2+ neuroblastoma cell lines, Y79 is a GD2+ retinoblastoma line, SK-N-SH is a GD2low neuroblastoma, A-204 (rhabdomyosarcoma) and Jurkat (T-cell leukemia) express only very low levels of surface GD2 (Table I). Data are representative of independent experiments performed with transduced effector cells from 4 different donors.

Table I. Secretion of Cytokines by T Lymphocytes Transduced with SFG/14.G2a-ζ in Response to Tumor Cells Expressing Different Levels of Surface GD2
Target cellsCytokine concentrations (pg/ml × 106 cells)GD2 surface expression of target cells
IFN-γTNF-αGM-CSFIL-10IL-4% Gmath imageMedian fluorescence absorbance
  1. On day 8 following transduction, 14.G2a-ζ-transduced T lymphocytes were coincubated with tumor cells at a 3:1 E:T ratio, and cytokine concentrations in the supernatant were quantified after 24 hr. The percentage of GD2-expressing cells and the mean fluorescence absorbance were determined by flow-cytometric analysis after staining with FITC-labeled MAb 14.G2a.

SK-N-SH171<84346Not tested<16844.6517.15
Jurkat220<84357Not tested<1684.1189.77
A20474<84Not tested<42<1684.1157.77
Effector alone45<84<42<42<168
Table II. Comparison of IFN-γ Secretion by T Lymphocytes Transduced with 14.G2a-γ or 14.G2a-ζ Chimeric Receptor Genes
DonorTarget cellsIFN-γ secretion (pg/ml × 106 cells over 24 hr) by transduced T cellsFold IFN-γ secretion by T cells transduced with 14.G2a-γ vs. 14.G2a-ζ
  1. IFN-γ concentrations in supernatants were measured following 24 hr coincubation with GD2-expressing tumor cell lines at an E:T ratio of 3:1.


Specific cytolytic activity of chimeric receptor-transduced T cells

The ability to exert specific cytotoxicity against GD2-positive target cells was evaluated in chimeric receptor-transduced T cells from 4 donors using a standard 5-hr Cr-release assay. 14.G2a-γ- and 14.G2a-ζ-transduced T cells reproducibly showed specific cytolysis of all GD2-positive target cell lines in the absence of significant cytotoxicity against GD2-negative cells (Fig. 5a). Background cytolysis of Jurkat cells was observed by others and explained by the low but detectable level of GD2 expression. We also tested various Ewing's sarcoma (RD-ES, A673, SK-ES-1), PNET (SK-N-MC) and small-cell lung-cancer (NCIH345) cell lines that have low levels of GD2 surface expression. None of these cell lines was specifically lysed by the modified effector cells (data not shown). The construct endowed with the TCR ζ-chain cytoplasmic domain consistently yielded superior cytotoxicity compared to the FcϵRI γ-chain–containing receptor (Fig. 5b). No specific Cr release was found by T-lymphocyte effector cells that expressed a truncated variant of the 14.G2a-specific chimeric receptor (Fig. 5b). In accordance with the lack of specific cytokine release mediated by sc7A4-based receptor constructs, neither of the scFv 7A4-containing chimeric receptors mediated any detectable target cell lysis when expressed on T-effector cells (Fig. 5c). Patient-derived peripheral blood lymphocytes, obtained at diagnosis from 2 patients with neuroblastoma and transduced to express the 14.G2a-ζ chimeric receptor gene, specifically lysed not only continuous neuroblastoma cell lines but also a neuroblastoma cell primary culture at passage 2 (Fig. 5d).

Figure 5.

(a) GD2-specific lysis of tumor cell targets by 14.G2a-ζ transduced T lymphocytes. JF, LAN-5, IMR-32 and Y79 are GD2+ tumor targets; A-204 was used as a GD2 control target. Jurkat cells express very low levels of surface GD2 (Table I). Cells were tested on day 22 (top) and on day 14 (bottom) after transduction in a 5 hr 51Cr-release assay. Results were reproduced multiple times in independent experiments with transduced cells from this and 2 other healthy donors. (b) Lysis of LAN-5 GD2high neuroblastoma cells by T lymphocytes transduced with 14.G2a-γ or 14.G2a-ζ genes in a 5 hr 51Cr-release assay. Cells transduced with a truncated variant of the 14.G2a-ζ receptor (14.G2a-Δζ) were used as negative control effector cells. Similar results were obtained with transduced cells from another donor. (c) T lymphocytes transduced with 7A4-γ or 7A4-ζ genes fail to lyse GD2+ JF target cells. Cells transduced with the chimeric receptor genes or SFG/EGF protein–transduced control cells were tested for specific lysis of JF target cells in a 5 hr 51Cr-release assay on day 7 following transduction. Results were independently reproduced with transduced T lymphocytes from 3 different donors. (d) GD2-specific lysis of tumor cell targets by 14.G2a-ζ–transduced T lymphocytes (top) obtained at diagnosis from a patient with neuroblastoma. LAN-1 is a GD2high tumor cell line; NB-1 is a GD2+ neuroblastoma cell primary culture. Autologous blasts were obtained by stimulation of PBMCs with phytohemagglutinin (1 μg/ml) for 72 hr and subsequent expansion in culture medium containing IL-2 (100 IU/ml). Cells transduced with the truncated receptor 14.G2a-Δζ (bottom) served as negative controls and did not lyse any of the cellular targets. Cells were tested on day 7 following transduction in a 5 hr 51Cr-release assay. Results were reproduced with transduced cells from 3 other neuroblastoma patients.

Preincubation of GD2-positive target cells with 14.G2a MAb resulted in up to 89% inhibition of target cell lysis at 20 μg/ml (Fig. 6), supporting specificity of the chimeric receptor for GD2-expressing target cells.

Figure 6.

Specific inhibition of 14.G2a-ζ–mediated target cell lysis by preincubation with MAb 14.G2a. 51Cr-labeled GD2+ LAN-1 and JF neuroblastoma cells were preincubated with the indicated concentrations of MAb 14.G2a for 30 min, then coincubated for 5 hr with transduced T lymphocytes from a patient with neuroblastoma at a 20:1 E:T ratio. Similar results were obtained with transduced cells from another neuroblastoma patient.

Soluble GD2 does not significantly affect the function of 14.G2a-ζ–transduced T cells

The presence of free GD2 ganglioside antigen during the interaction of modified T cells with tumor targets may inhibit 14.G2a-ζ–mediated tumor cell lysis by blocking chimeric receptor-binding sites. To evaluate this possibility, the tumor-specific cytotoxicity of chimeric receptor-transduced T cells was measured after preincubation of target cells with soluble GD2. Target cell lysis was not inhibited by the presence of soluble GD2 in concentrations exceeding the highest serum levels of 2,160 pmol/ml found in patients with advanced neuroblastoma (Fig. 7). 25

Figure 7.

Lysis of GD2+ target cells by 14.G2a-ζ–transduced T lymphocytes in the presence of soluble GD2. 51Cr-labeled GD2+ LAN-1 and JF neuroblastoma and Y-79 retinoblastoma cells were preincubated with soluble GD2 at the indicated concentrations, then coincubated for 5 hr with transduced T lymphocytes from a healthy donor at a 40:1 E:T ratio. Similar results were obtained with transduced cells from 2 other healthy donors.

Chimeric receptor stimulation and functionality

To investigate whether signaling through the chimeric receptor alone provides a proliferation stimulus sufficient for prolonged in vitro expansion of cells, we cultured 14.G2a-ζ–transduced T cells in the presence of irradiated GD2+ and GD2 tumor cell targets and rhIL-2 (100 IU/ml). Cell cultures could not be maintained for longer than 8 weeks upon stimulation with GD2+ tumor cells and showed a growth pattern comparable to T cells expanded in IL-2–supplemented growth medium or in the presence of GD2 tumor targets (data not shown). The inability of chimeric receptor targets to induce commitment to proliferation was further investigated by [3H]-thymidine incorporation assays. Chimeric receptor-transduced T cells stimulated with GD2+ irradiated tumor target cells did not proliferate above background levels obtained with GD2 tumor cell targets or by testing T cells of the same donor transduced with EGF protein (Fig. 8a). Repeated assessment of tumor-specific IFN-γ release by cultured T cells revealed a substantial decrease in their functional activity over time (Fig. 8b) in the absence of chimeric receptor downregulation, as assessed by continued cell-surface staining with anti-14.G2a idiotypic antibody (Fig. 2b).

Figure 8.

(a) [3H]-Thymidine incorporation by T lymphocytes transduced with SFG/14.G2a-γ or EGF protein in response to coincubation with tumor target cells. On day 14 following transduction, T lymphoctes were cocultured with GD2+ (LAN-1, JF) or GD2 (A-204) tumor cells at a 3:1 E:T ratio for 72 hr before [3H]-thymidine was added to wells. Data represent independent experiments performed with effector cells from 3 different donors transduced with either 14.G2a-γ or 14.G2a-ζ. (b) IFN-γ release by T lymphocytes transduced with SFG/14.G2a-γ in response to coincubation with tumor target cells. On days 8, 12 and 21 following transduction, T lymphoctes were cocultured with tumor cells at a 3:1 E:T ratio for 24 hr. Data represent independent experiments performed with effector cells from 3 different donors transduced with either 14.G2a-γ or 14.G2a-ζ.


To retarget cytotoxic T lymphocytes to human neuroblastoma cells while bypassing most of the limitations of classical immunotherapeutic strategies, we generated a chimeric TCR that provides T lymphocytes with specificity for the ganglioside antigen GD2. 14.G2a-γ- and ζ-transduced T lymphocytes from both normal donors and neuroblastoma patients recognize and lyse GD2-expressing tumor cell targets in an antigen-specific manner. Conflicting data exist as to whether chimeric TCR signaling is more effective when mediated by either the TCR ζ-chain26–28 or the Fc receptor γ subunit.29 In our system, TCR ζ provided the chimeric receptor construct with superior tumor-specific functionality compared to the FcϵRI γ-chain.

The successful clinical use of a specific tumor-associated antigen as a target for chimeric receptor-mediated cytolysis requires tumor selectivity, intermediate to high cell-surface antigen density and stability of antigen expression. GD2 is a ganglioside antigen expressed in abundance on tumors originating from a neuroectoderm-derived lineage,30 including neuroblastoma melanoma, glioblastoma and a proportion of small-cell lung cancer. Although expression of GD2 in normal tissues is highly restricted,30 the clinical use of MAbs directed against GD2 resulted in significant dose-related neurologic side effects.31–33 In contrast to the high-affinity interaction of MAbs with their antigen, T-cell immunity is characterized by the ability to discriminate subtle differences in target antigen affinity.26 The dynamics of interaction between chimeric receptor-transduced T lymphocytes and the corresponding B-cell epitope have not been investigated in detail. However, engineered T cells may selectively respond to target cells that express at least intermediate levels of target antigen.34, 35 Our data support the relative safety of GD2 targeting by modified T lymphocytes by showing a lack of specific targeting of tumor cells that have only low GD2 surface density. Furthermore, the immune privileged status of the CNS, which limits access of systemic immune cells through the blood–brain barrier, should prevent an immune assault of ganglioside-expressing brain cells by adoptively transferred GD2-specific T lymphocytes.

Soluble GD2, as a result of ganglioside shedding from tumor cells, is detected in the serum of neuroblastoma patients with advanced disease in median concentrations of 237 pmol/ml, ranging from ≤50 to 2,160 pmol/ml.25 Shed antigen may inhibit chimeric receptor binding to the intended target cell. However, we did not see any specific inhibition of tumor cell lysis by preincubation of target cells with soluble GD2 at clinically relevant concentrations, suggesting preferential engagement of the chimeric receptors by cellular antigen even in the presence of competing soluble antigen.

Another potential escape mechanism for tumor cells during treatment with chimeric receptor-expressing T lymphocytes is downregulation of antigen expression or selection of an antigen-negative subclone. GD2, however, is stably expressed on most tumors throughout the stages of disease and following relapse.36 Furthermore, GD2-specific immunotherapy has not resulted in downregulation of GD2 expression in the vast majority of patients.37

A critical requirement for the effective targeting of B-cell epitopes via modified T cells is the availability of a single-chain antibody that retains optimal affinity and high specificity of antigen recognition when expressed as the recognition portion of a chimeric TCR. Anti-GD2 antibody 7A438 retains the targeting properties of the MAb after cloning as an isolated single chain.16 However, despite full-length expression of 7A4-γ and 7A4-ζ RNA and protein dimers in transduced T lymphocytes, the transgene failed to provide the cells with the ability to specifically interact with GD2-expressing target cells. Similar results have been reported for other single-chain antibodies.5 The failure of selected scFvs to serve as potent recognition domains may be explained by suboptimal ligand binding due to conformational changes when expressed as the extracellular portion of a T cell-anchored receptor. Under the hypothesis that the pattern of reactivity and affinity of a phage-presented scFv may come closer to its performance as the recognition domain of a chimeric receptor, we cloned the variable domains of MAb 14.G2a via phage display. Phage display is an attractive system to assure cloning of the relevant V domains in a pool of hybridoma immunoglobulin transcripts39 and has been used successfully to clone several single-chain antibodies, including an anti-TAG72 scFv, which was subsequently used for T-cell engineering.40 Only a small proportion of phage clones containing a full-length insert had the specificity provided by the 14.G2a-derived hypervariable domains, further supporting the rationale for using phage display as a rapid screening procedure during hybridoma antibody cloning. Although target cell binding by the phage antibody was significantly reduced when compared to the parent antibody, the resulting chimeric receptor mediated highly efficient GD2-specific T-cell activation and potent cytolysis by nonselected transduced T-cell populations.

An issue with high clinical relevance is the potential immunogenicity of the modified cells that may result in rapid clearance from the circulation, mediated by human antimouse immune responses. Thus, we attempted to largely reduce possible targets of immunogenicity in our constructs. With our retroviral gene-transfer system, we consistently transduce 20% to 30% of T lymphocytes with the chimeric receptor gene, sufficient to obtain efficient target cell recognition and lysis even without positive selection, overcoming the need for coexpression of immunogenic selectable markers. To prevent presentation of signal peptide-derived fragments on MHC to cytotoxic T lymphocytes, we substituted a human IgG1 leader fragment. Immunogenicity of the humanized MAb 14.G2a is largely restricted to the idiotypes.12 Generation of clinical antiidiotypic immune responses to anti-GD2 MAbs is beneficial, correlating with increased survival rates.41 Based on the concept of an antiidiotypic network,42 this phenomenon has been explained by a tumor-specific response elicited by antiidiotypic antibodies and anti-14.G2a idiotypic antibody 1A7 is currently being clinically tested as an alternative approach to tumor vaccination. Thus, although we cannot rule out a possibly deleterious effect on the life span of adoptively transferred cells, triggering of a host antitumor response by autologous T lymphocytes transduced to express 14.G2a-γ or ζ may actually contribute to an antitumor effect.

While the chimeric receptor-transduced cells described here may be of clinical value, past experience indicates that cells of this type rapidly lose functionality in vivo.43, 44 This effect is manifest in our own system by our data showing a substantial decrease in tumor-specific cytokine release during in vitro culture. We also show here that stimulation of transduced T cells with GD2-expressing tumor cells alone does not provide a sufficient stimulus for commitment to proliferation, continued T-cell expansion and long-term in vitro persistence. To favor survival and activation of tumor-reactive T cells, a CD28-like chimeric receptor specific for GD2 was expressed in T lymphocytes and shown to induce a functional costimulatory response.45 While combining chimeric receptor-mediated T-cell activation and costimulation by coexpression of both receptors may result in prolonged in vivo functionality of the modified cells, transgenic mouse models have shown that chimeric receptor-mediated signaling alone is insufficient to trigger activation of resting primary T cells even in the presence of adequate costimulation.46 Instead, the extended in vivo survival and functionality of chimeric receptor-modified cells likely requires repeated T-cell stimulation via the native TCR. We are currently investigating the consequences of expressing the chimeric TCR described here in cytotoxic T lymphocytes with native specificity for Epstein-Barr virus.


This work was supported by National Institutes of Health grants CA 75014 (to MB) and CA 78456 (to JN) and by a grant from Deutsche Krebshilfe (to CR).