In vitro Stimulation with WT1 Peptide-Loaded Epstein-Barr Virus-Positive B Cells Elicits High Frequencies of WT1 Peptide-Specific T Cells with In vitro and In vivo Tumoricidal Activity Free

Adoptive transfer of donor-derived lymphocytes can induce durable remissions in >80% of marrow allograft recipients relapsing with chronic myelogenous leukemia, and in 25 to 30% of patients relapsing with acute myelogenous leukemia (AML) or myeloma (1, 2). In patients treated for chronic myelogenous leukemia, this antileukemic activity has been ascribed to allospecific T cells reacting against clonogenic Ph⁺ hematopoietic precursors (3, 4). Indeed, the transfer of clones of such cells has induced remissions in isolated cases (3). Unfortunately, minor alloantigen-specific T cells can also induce graft-versus-host disease (5). An alternative strategy for adoptive therapy would be to use T cells specific for antigens differentially expressed by leukemic cells, such as T-cell receptor (TCR) and immunoglobulin idiotypes specific to leukemic clones (6, 7), the protein products of fusion genes resulting from leukemia-specific translocations such as bcr/abl (8), or oncofetal proteins differentially expressed by leukemic blasts. Two oncofetal proteins [the Wilms tumor gene product, WT1 (9, 10, 11), and proteinase 3 (12)] have recently been proposed for evaluation.

WT1, a gene mutated in Wilms tumor, encodes a zinc finger transcription factor that binds and represses promoters of several important growth factor genes (13, 14). Expression of WT1 is high during fetal development of the genitourinary system. Postnatally, WT1 can be detected only at low levels in kidneys, splenic capsule, Sertoli cells of the testis, and granulosa cells of the ovary (14), as well as in normal CD34⁺ hematopoietic progenitors (15, 16). In contrast, WT1 is overexpressed in a high proportion of patients with acute lymphoblastic leukemia, AML, myelodysplastic syndrome, and chronic myelogenous leukemia (17), and has been specifically correlated with a poor outcome in patients with AML and myelodysplastic syndrome (18, 19). WT1 is also highly expressed in several solid tumors (20, 21, 22, 23).

Recently, immunogenic peptides of WT1 with high affinity for HLA-A2402 and HLA-A0201 have been identified. T-cell clones (9, 10) specific for these peptides have been isolated and expanded from single donors that specifically lysed WT1⁺ HLA-A2402⁺ or WT1⁺ HLA-A0201⁺ leukemia cell lines in vitro (9, 10) and inhibited the growth of lung cancer xenografts in nude mice in vivo (24). Recently, Gaiger et al. (25) and Eliseeva et al. (26)detected antibodies to WT1 in up to 29% of patients with AML. Scheibenbogen et al. (27) also recorded low but detectable frequencies of WT1-peptide-specific IFNγ-secreting T cells in the blood of >30% of patients with AML but not in normal donors. However, Rezvani et al. (28) detected IFNγ mRNA in T cells stimulated with WT1-peptide in 10 of 18 healthy individuals and WT1-peptide-HLA-A0201-tetramer-binding T cells in a small subset of these individuals. Recently, Gao et al. (11) have also generated alloreactive T cells from HLA-A0201⁻ donors specific for HLA-A0201⁺ cells expressing WT1, which could lyse WT1⁺ HLA-A0201⁺ leukemic cells and deplete clonogenic cells capable of transferring leukemia in nonobese diabetic-severe combined immunodeficient (NOD-SCID) mice (29). However, to date, no study has evaluated the capacity of HLA-binding peptides derived from this potentially tolerogenic “self” protein to stimulate the generation of WT1-peptide-specific HLA-restricted T cells from random HLA-A0201- or HLA-A2402-bearing normal donors in comparison with the responses to known immunogenic peptides derived from viral proteins. Furthermore, the in vivo leukemocidal activity of the WT1-specific T cells generated from normal individuals has not been evaluated.

In this study, we demonstrate that HLA-restricted CD8⁺ cytotoxic T-cell lines that are specific for WT1 immunogenic peptides can be regularly generated in vitro from HLA-A0201⁺ and HLA-A2402⁺ normal human donors after sensitization with autologous peptide-loaded cytokine-activated monocytes (CAMs) or EBV-transformed B cells (EBV-BLCLs). Strikingly, the frequencies of WT1-peptide-specific T cells, generated in response to peptide-loaded EBV-BLCLs, were similar to those generated against highly immunogenic EBV peptides. These T cells specifically lysed WT1⁺ leukemias and selected WT1⁺ solid tumors expressing the restricting HLA allele. When infused into tumor-bearing NOD-SCID mice, WT1-peptide-specific T cells selectively migrate to, and induce regressions of, established WT1⁺ human leukemic xenografts expressing the restricting HLA allele as demonstrated by imaging, immunohistology, and the clinical response of disease.

WT1 peptides (Research Genetics, Huntsville, AL) included two previously described HLA-A0201-binding peptides, _126–134RMFPNAPYL (RMF) and _187–195SLGEQQYSV (SLG; ref. 10), and one heretofore-not-described WT1 peptide, _301–310RVPGVAPTL (RVP), that was predicted to have high affinity for HLA-A2402 by using the predictive algorithm, (SYFPEITHI; ref. 30).

Blood and tissue samples were obtained from consenting patients and normal donors according to protocols approved by the Institutional Review Board at Memorial Sloan-Kettering Cancer Center (MSKCC).

EBV-BLCLs were generated by infecting peripheral blood mononuclear cells (PBMCs) with the EBV isolate B95.8, as described previously (31). To generate CAMs, we cultured PBMC-derived plastic-adherent monocytes in RPMI 1640, containing 5% human AB serum (HS; Gemini, Calabasas, CA). Interleukin (IL)-4(1,000 IU/mL) and granulocyte-macrophage colony-stimulating factor (GM-CSF, 1,000 IU/mL; both from Sigma, St. Louis, MO) were added on days 0, 2, 4, and 6 of culture. Beginning on day 6, tumor necrosis factor α (10 ng/mL; Sigma) was added to the medium. CAMs harvested at day 9 were used for T-cell sensitizations.

To prepare peptide-loaded antigen-presenting cells (APCs), we incubated 1 to 2 × 10⁶ cells/mL CAMs or EBV-BLCLs from the same donor for 3 hours with a 50 μg/mL of one of the synthetic WT1 nonapeptides in serum-free medium; we then washed and irradiated the cells before sensitization of autologous T cells.

T-cell-enriched fractions of PBMCs were obtained by depletion of monocytes by adhesion, followed by immunoadsorption of cells binding anti-CD56/CD16 monoclonal antibodies (Immunotech, Marseille, France) with antimouse IgG-coated immunomagnetic beads (Nexell Therapeutics, Irvine, CA). These cells were then stimulated with WT1-negative EBV-BLCLs or CAMs (EBV-CTLs or CAM-CTLs) or WT1-peptide-loaded autologous EBV-BLCLs or CAMs (WT1/EBV-CTLs or WT1/CAM-CTLs) at a 20:1 responder-to-stimulator (R:S) ratio and cultured in Yssel’s medium, containing 5% HS(YH5; Gemini), supplemented every 2 to 3 days with 10 to 30 units/mL IL-2 (Collaborative Biomedical Products, Bedford, MA) beginning from day 7 and restimulated weekly with the same APCs at a 4:1 R:S ratio.

T cells recognizing HLA-A0201-binding WT1-peptides were quantitated by fluorescence-activated cell sorting (FACS) analysis of CD8⁺ CTLs binding RMF/HLA-A0201- or SLG/HLA-A0201-PE-labeled tetramers. EBV-specific T cells were also quantitated by using EBNA3C/A0201, LMP1/A0201, LMP2/A0201, BZLF1/A0201, BMLF1/A0201 PE-labeled tetramers. Tetramers were generated in the MSKCC Tetramer Core Facility, as described previously (32).

T cells producing intracellular IFNγ in response to secondary stimulation with peptides were quantitated as described previously (33).

The WT1-peptide- and EBV-specific cytotoxic activity of sensitized T cells was measured with a standard ⁵¹Cr-release assay, as previously described (31).

Expression of WT1 protein by target cells was measured by FACS analysis after permeabilization and staining with mouse antihuman-WT1 (Neomarkers, Fremont, CA) and goat-antimouse FITC-conjugated antibodies (Becton Dickinson, San Jose, CA). Surface expression of HLA-class I and II was assessed by FACS analysis with FITC-labeled monoclonal antibodies (Becton Dickinson). HLA typing was performed as previously described (34).

Isolated CD34⁺ cells (5 × 10²), preincubated for 4 hours at 37°C 5% CO₂ with or without 2.5 × 10⁴ of WT1/EBV-CTLs or EBV-CTLs, were cocultured with the same effector cells in triplicate in 35-mm tissue culture dishes containing 1 mL of IMDM supplemented with 1.2% methylcellulose (Fisher Scientific, Fairlawn, NJ), 30% fetal bovine serum, 5 × 10⁻⁵ mol/L 2-mercaptoethanol, 2 mmol/L l-glutamine (Life Technologies, Inc., Rockville, MD), 0.5 mmol/L hemin (Sigma Aldrich, St. Louis, MO), 20 ng/mL recombinant human (rh) kit ligand, 20 ng/mL rhIL-3, 20 ng/mL rhG-CSF, and 6 units/mL erythropoietin (Amgen, Thousand Oaks, CA) at 37°C in 5% CO₂ for 14 days. Colony-forming unit (CFU)-GM, burst-forming unit erythroid, and mixed colonies (CFU-Mix) were scored, as described previously (35).

For analyses of the antileukemic activity of T cells in vivo, an adoptive transfer model previously described by our group was used (31, 36, 37). Briefly, up to four separate human tumor xenografts were established in 4-to-6-week-old male NOD-SCID mice (Taconic Farms, Germantown, NJ) by subcutaneous injection of 5 × 10⁶ cells of each tumor type suspended in Matrigel into the thighs or shoulders. After the development of measurable tumors, specified test and control groups of mice (n = 5 per group) were treated intravenously with WT1/EBV-CTLs or EBV-CTLs. All of the mice were also treated with 2,000 IU of rhIL-2 three times a week beginning from the day of the CTL injection.

Mice were monitored daily, and tumor volumes were measured weekly, as described previously (31). At specified intervals after the T-cell infusion, mice were sacrificed, and tumors and organs were examined for WT1⁺ tumor cells and infiltrating CD3⁺ T cells with standard immunohistochemical techniques.

WT1/EBV-CTLs or EBV-CTLs were transduced with a modified retroviral vector encoding a herpes simplex virus thymidine kinase/green fluorescence protein (HSV-TK/GFP)-fusion protein containing CpG-reduced HSV-TK cDNA, as described previously (31, 38, 39). Transduced T cells expressing GFP were isolated with a MoFlo FACS Sorter (Cytonation, Fort Collins, CO) and were expanded in the presence of OKT3 antibodies (Ortho Biotech, Raritan, NJ), autologous irradiated PBMCs, and autologous irradiated RMF-loaded (RMF-BLCL) or unloaded EBV-BLCL. Before adoptive transfer, the transduced T cells were tested for binding to RMF and EBV-peptide/tetramer and for cytotoxicity against WT1⁺ and EBV⁺ targets. HSV-TK expression was evaluated as described previously (40).

For in vivo imaging of antigen-specific migration and targeting of EBV- and WT1-reactive T cells, 28 days after tumor inoculation, NOD-SCID mice, bearing xenografts of two HLA-A0201⁺ EBV-BLCLs and HLA-A0201⁺ WT1⁺ B-lineage acute lymphoblastic leukemia (B-ALL) and an HLA-A0201⁺ WT1⁻ B-ALL were given infusions of 16 × 10⁶ cells of HSV-TK/(enhanced)GFP-transduced EBV-CTLs (EBV-TG-CTLs; n = 5) or WT1/EBV-CTLs (WT1/EBV-TG-CTLs; n = 5) per mouse. An additional control group bearing the same tumors xenografts (n = 5) received no CTLs.

1-(2-Deoxy-2-fluoro-β-d-arabinofuranosyl)-5-[¹³¹I]iodouracil ([¹³¹I]FIAU) was administered intravenously at a dose of 100 μCi per animal in two animals from each group 24 hours after T-cell injection and in the other three animals of each group on day 7 after CTL administration. Thyroid accumulation of ¹³¹I was partially blocked by intraperitoneal administration of 500 μL of 0.9% NaI solution. Images were obtained with a Forte gamma camera (Philips Medical Systems, Bothell, WA) and a high-energy high-resolution collimator at 4 and 24 hours after the administration of [¹³¹I]FIAU.

To selectively define the migration and accumulation of WT1-reactive T cells, on day 21 after tumor inoculations, we injected HLA-A0201-restricted EBV-TG-CTLs or WT1/EBV-TG-CTLs intravenously into groups of NOD-SCID mice (n = 8 in each group) that were concurrently bearing four human xenografts that differed in expression of WT1 and HLA-A0201. An additional control group bearing the same tumor xenografts received no CTLs. 1-(2′-Deoxy-2′-[¹⁸F]fluoro-β-d-arabinofuranosyl)-5-ethyluridine ([¹⁸F]FEAU) was administered intravenously at a dose of 100 μCi per mouse per dose on days 1 and 7 after the T-cell injection. Images were obtained with microPET (Concorde, Knoxville, TN) 4 hours after each radiotracer administration.

All of the animals were treated with 2000 units of IL-2 per mouse, three times a week, throughout the course of the experiment from the day of T-cell injection. Animals were sacrificed after imaging. Tumors and tissue samples were assessed for radiotracer accumulation, and selected single-cell suspensions thereof were analyzed by FACS for infiltration with enhanced-GFP (eGFP)-expressing T cells.

A permutation test was used to determine differences in tumor volume between groups in the mouse studies. The test statistic was based on the average of differences in tumor volume between groups stratified over time. Kendall’s Tau statistic was used to assess the level of monotone association between the proportion of target cells that expressed WT1 and cytolysis by WT1-peptide-sensitized T cells.

We initially assessed WT1-peptide-specific responses by quantitating RMF and SLG/HLA-A0201-tetramer-binding T cells in T-cell cultures stimulated with autologous EBV-BLCLs or CAMs alone (EBV-CTLs or CAM-CTLs), or with EBV-BLCLs or CAMs loaded with either RMF (RMF-BLCLs or RMF-CAMs) or SLG (SLG-BLCLs or SLG-CAMs) WT1-peptides. Cumulative results are presented in Fig. 1 A.

RMF/HLA-A0201-tetramer-binding cells detected in EBV-CTLs or CAM-CTLs ranged from 0.01 to 0.19%. In contrast, significant percentages of RMF/HLA-A0201-tetramer-binding T cells were detected in cultures sensitized with RMF-BLCLs (RMF/EBV-CTLs) or RMF-CAMs (RMF/CAM-CTLs), ranging between 1 and 32% and 0.8 and 5%, respectively. In the donors tested for responses to both RMF and SLG, the percentages of T cells in the SLG-stimulated cultures (SLG/EBV-CTLs or SLG/CAM-CTLs) binding SLG/HLA-A0201-tetramers were similar to those in RMF/EBV-CTLs or RMF/CAM-CTLs binding RMF/HLA-A0201-tetramers.

Strikingly, for each of 13 donors tested, the percentages of RMF- or SLG/HLA-A0201-tetramer-binding T cells detected in WT1/EBV-CTLs were equivalent to the frequencies of the EBV-specific T cells detected in the same cultures binding EBNA3C-, BZLF-1-, BMLF-1-, or LMP-2-EBV-peptide/HLA-A0201 tetramers (Fig. 1,A and B). Furthermore, the percentages of EBV-peptide-reactive T cells detected in WT1/EBV-CTLs did not differ from those detected in EBV-CTLs. Moreover, the levels of tetramer-binding exhibited by WT1/EBV-CTLs, when exposed to either RMF/HLA-A0201- or EBNA3C/HLA-A0201-tetramers, were also similar, as reflected by mean fluorescence of the positive cells (Fig. 1 C). The percentages of WT1/EBV-CTLs or EBV-CTLs that bound HLA-A0201-LMP1-tetramers were less than 1% in 11 of 13 donors tested, a finding consistent with the known poor immunogenicity of LMP1 peptides.

We also compared yields of WT1-reactive T cells in cultures sensitized with either RMF-BLCLs or RMF-CAMs. At each week tested, the yield of CD8⁺ T cells binding RMF/HLA-A0201-tetramers was significantly greater in WT1/EBV-CTLs (Fig. 1 D).

CTLs from 7 of the 13 HLA-A0201⁺ donors, contemporaneously sensitized for 35 to 42 days with RMF-BLCLs, RMF-CAMs, and EBV-BLCLs, were also tested for their content of T cells that generated IFNγ in response to restimulation with autologous RMF-loaded PBMCs and for their cytotoxic activity against a panel of targets. A representative analysis of one of these donors is shown in Fig. 2.

EBV-CTLs contained negligible levels of tetramer-binding or IFNγ-producing RMF-specific T cells and lysed only autologous EBV-BLCLs (Fig. 2). In contrast, RMF/EBV-CTLs contained 14% RMF/HLA-A0201-tetramer⁺ T cells and 5.8% IFNγ⁺ T cells. In addition, RMF/CAM-CTLs contained 4.42% RMF-tetramer⁺ CD8⁺ T cells and 1.02% IFNγ⁺ T cells. In cytotoxicity assays, RMF/EBV-CTLs and RMF/CAM-CTLs exhibited similar capacity to lyse autologous and HLA-A0201⁺ allogeneic RMF-loaded phytohemagglutinin stimulated (PHA) blasts and HLA-A0201⁺ WT1⁺ primary human AML, T-lineage acute lymphoblastic leukemia (T-ALL), and B-ALL cells but did not lyse unmodified autologous or HLA-A0201⁺ allogeneic PHA blasts, autologous CD34⁺ peripheral-blood stem-cells (PBSCs), HLA-mismatched EBV-BLCLs unmodified or loaded with RMF, or a WT1⁻ HLA-A0201⁺ AML (Fig. 2,C). As expected, RMF/EBV-CTLs lysed autologous EBV-BLCLs, whereas RMF/CAM-CTLs did not. CAM/CTL did not bind WT1-peptide-/HLA-A0201-tetramers (Fig. 1 A). Cell yields of these T-cell cultures, however, were insufficient to permit analysis of cytokine generation or cytotoxic activity.

Five donors expressing HLA-A2402 were tested for responses to a newly identified HLA-A2402-binding WT1-peptide, RVP, when loaded on autologous CAMs (n = 3; RVP-CAMs) and/or EBV-BLCLs (RVP-BLCLs; n = 5; Table 1). Unfortunately, stable RVP/HLA-A2402 tetramers could not be generated. However, the RVP-peptide-sensitized T cells generated from each donor specifically lysed RVP-loaded autologous PHA blasts as well as unmodified HLA-A2402⁺ WT1⁺ leukemic cells. Furthermore, T cells from each of two donors, coexpressing HLA-A2402 and HLA-A0201 sensitized with RVP-loaded-APCs, exhibited equal or greater cytotoxicity against WT1⁺ B-ALLs coexpressing HLA-A2402 and HLA-A0201 than the same T cells sensitized with RMF-loaded APCs (Table 1).

The proportion of HLA-A0201⁺ CD34⁺ PBSCs and cord blood mononuclear cells lysed by the WT1/CAM-CTLs and WT1/EBV-CTLs was extremely low, and similar to the proportion of WT1⁻ PHA blasts lysed by these cells (Fig. 3,A). In contrast, these T cells exhibited a significant cytolytic activity against 6 of 9 leukemias and 4 of 11 solid tumor cell lines tested. The level of cytotoxicity observed was correlated with the percentage of cells in the target cell population that expressed WT1 (Kendall’s Tau statistic = 0.57; P < 0.001; Fig. 3 B, top panel).

In normal or malignant target cell populations containing either high or low proportions of WT1⁺ cells, WT1⁺ cells exhibited a uniformly high-intensity fluorescence that was easily distinguishable from negative cell populations (Fig. 3,B, bottom panel). To ascertain the limits of the cytotoxicity assay for discriminating activity against these WT1⁺ cells, we assessed the cytotoxic activity of RMF/EBV-CTLs from three donors against a WT1⁺ HLA-A0201⁺ B-ALL serially diluted with WT1⁻ CD34⁺ PBSCs (Fig. 3C). Again, cytolysis was significantly correlated with the percentage of WT1⁺ cells in the target pool (Kendall’s Tau statistic = 0.56, P < 0.001). However, at concentrations of WT1⁺ cells <20%, cytotoxicity could not be distinguished from that against WT1⁻ controls.

Because clonogenic progenitors could be included in the small fraction of WT1⁺ CD34⁺ PBSCs, we also compared RMF/EBV-CTLs and EBV-CTLs for their potential to inhibit the growth of hematopoietic progenitors from CD34⁺ PBSCs. The growth of CFU-GM, burst-forming unit erythroid, and CFU-mix was not affected by either of the T-cell populations when compared with the growth of these progenitors in the absence of CTLs (Fig. 3 D).

In initial experiments, adoptive transfer of RMF/EBV-CTLs or SLG/EBV-CTLs inhibited growth of WT1⁺ HLA-A0201⁺ T-ALL xenografts to an equivalent degree (Fig. 4,A; P < 0.01). Single infusions of RMF/EBV-CTLs also induced regressions of fresh HLA-A0201⁺ WT1⁺ AML xenografts(P < 0.01; Fig. 4 B). When regrowth of the leukemic xenografts was detected at week 9, a second infusion of these T cells induced a complete regression of the leukemia in each of the five mice treated.

To further assess the target specificity of T cells in vivo, RMF/EBV-CTLs were adoptively transferred into NOD-SCID mice simultaneously bearing subcutaneous xenografts of human WT1⁺ HLA-A0201⁺ B-ALL and WT1⁻ HLA-A0201⁺ B-ALL (63% versus 3% WT1⁺ cells). Sequential intravenous infusions of RMF/EBV-CTLs significantly inhibited the growth of WT1⁺ leukemia (P < 0.01; Fig. 4,C, top panel). In contrast, the growth of WT1⁻leukemia (Fig. 4,D, top panel) was not affected. EBV-CTLs had no effect on the growth of either WT1⁺ or WT1⁻ leukemic xenografts (Fig. 4,C and D, top panel). Immunohistopathologic analyses of tumors harvested from mice treated with RMF/EBV-CTLs also demonstrated infiltration of T cells and necrosis in the WT1⁺ leukemic xenograft (Fig. 4,C, bottom panel). In contrast, the WT1⁻ leukemic xenografts from the same animals exhibited viable leukemic cells without evidence of T-cell infiltration or necrosis (Fig. 4, D, bottom panel). Neither the WT1⁺ nor the WT1⁻ leukemic xenografts exhibited T-cell infiltration or evidence of necrosis in mice treated with EBV-CTLs or with IL-2 alone (data not shown).

To evaluate the in vivo distribution of the WT1/EBV-CTLs in xenografted mice, RMF-sensitized WT1/EBV-CTLs and EBV-CTLs that were contemporaneously expanded from the same donor and transduced with the HSV-TK/eGFP vector were then sorted by FACS to 98 and 95% purity, respectively. The sorted WT1/EBV-TG-CTLs contained 3.1% of RMF/HLA-A0201-tetramer⁺ T cells compared with <0.1% in sorted EBV-TG-CTLs. WT1/EBV-TG-CTLs lysed autologous RMF-PHA blasts and the HLA-A0201⁺ WT1⁺ leukemia. EBV-TG-CTLs did not lyse these targets. The cumulative proportions of T cells binding HLA-A0201- and HLA-B0701-tetramers that contained peptides of EBNA3C, BZLF1, BMLF1, and LMP2 were 50 and 52% in the WT1/EBV-TG-CTLs and EBV-TG-CTLs, respectively. These cells exhibited comparable HLA-A0201-restricted EBV-specific cytotoxicity (data not shown).

The K1 accumulation rate of the HSV-TK-substrate FIAU that characterized the HSV-TK-expression in sorted GFP⁺ T cells, measured by an in vitro radiotracer accumulation assay, was 0.234 mL/g/min and 0.192 mL/g/min for WT1/EBV-TG-CTLs and EBV-TG-CTLs, respectively, FIAU-to-thymidine ratios were 0.0933 and 0.11, respectively, which satisfied the in vivo imaging criteria (38).

Initially, we wanted to compare the ability of T cells in the WT1/EBV-TG-CTLs and EBV-TG-CTLs to migrate to WT1⁺ or EBV⁺ tumors expressing the HLA-A0201 restricting allele. Scintigraphic images (Fig. 5,A), obtained 24 hours after [¹³¹I]FIAU infusion into NOD-SCID mice bearing two B-ALL and two EBV-BLCL xenografts demonstrated accumulation of [¹³¹I]FIAU in the autologous EBV-BLCL tumor (Fig. 5,A, T2) and allogeneic HLA-A0201⁺ EBV-BLCL tumor (Fig. 5,A, T3) in mice treated with either WT1/EBV-TG-CTLs or EBV-TG-CTLs. In contrast, the WT1⁺ HLA-A0201⁺ B-ALL tumor (Fig. 5,A, T1) was visualized only in the animals treated with WT1/EBV-TG-CTLs. No signal was observed in the HLA-A0201⁺ WT1⁻leukemia tumor (Fig. 5 A, T4) in any of the treated groups of animals. Furthermore, none of these tumors were visualized after [¹³¹I]FIAU infusion in the animals treated with IL-2 alone (control).

Images obtained 8 days after T-cell injections (24 hours after a second dose of [¹³¹I]FIAU) demonstrated additional radiotracer accumulation in the HLA-A0201⁺ EBV-BLCL xenografts (Fig. 5,A, T2 and T3). Increased signal was also seen in the WT1⁺ B-ALL (Fig. 5,A, T1) in animals treated with WT1/EBV-TG-CTLs but not in the animals treated with EBV-TG-CTLs. Again, [¹³¹I]FIAU was not detected in the WT1⁻ leukemia. Images of mice treated with either WT1/EBV-TG or EBV-TG-CTLs also demonstrated signal in the spleen both at day 2 and at day 8 after the CTL injection. Direct measurements of radioactivity in tissue samples (Fig. 5,B) confirmed the images, demonstrating a concentration of [¹³¹I]FIAU in the EBV⁺ tumors (Fig. 5,B, T2 and T3) in animals treated with either of the T-cell populations, and a selective accumulation of radiolabel in WT1⁺ leukemia (Fig. 5,B, T1) in the animals treated with WT1/EBV-TG-CTLs. Only low levels of the radioisotope, comparable with those in stomach, colon, and muscles, were retained in the WT1⁻ leukemia. FACS analysis of the tumor cell suspensions also demonstrated infiltration of T1, T2, and T3 tumors with GFP⁺ T cells in the animals treated with WT1/EBV-TG-CTLs, whereas only the EBV⁺ tumors T2 and T3 contained a proportion of GFP⁺ T cells in the animals treated with EBV-TG-CTLs (Fig. 5 C).

In subsequent experiments, we used positron emission tomography (PET) imaging to examine with greater precision the migration of TK/eGFP-transduced T cells into tumor xenografts varying only in their expression of WT1 or the restricting HLA-A0201. The axial cross-sectional PET images (Fig. 5,D), through tumors in the shoulders (Fig. 5,D, T1 and T2) and thighs (Fig. 5,D, T3 and T4), were obtained on days 1 and 8 after the injection of WT1/EBV-TG-CTLs or EBV-TG-CTLs into NOD-SCID mice, bearing a WT1⁺ HLA-A0201⁺ rhabdomyosarcoma (Fig. 5,D, T1), a WT1⁺ HLA-A0201⁺ B-ALL (Fig. 5,D, T2), a WT1⁺ HLA-A0201⁻ rhabdomyosarcoma (Fig. 5,D, T3), and a WT1⁻ HLA-A0201⁺ B-ALL (Fig. 5,D, T4). [¹⁸F]FEAU was accumulated in the WT1⁺ HLA-A0201⁺ tumors (Fig. 5 D, T1 and T2) only in the animals treated with WT1/EBV-TG-CTLs. Accumulation of [¹⁸F]-FEAU was greater in the WT1⁺ HLA-A0201⁺ B-ALL(T2). By FACS analysis, the proportion of T1 and T2 cells that expressed surface HLA-class-I was the same (99%). However, more T2 cells expressed WT1 (60% versus 44%), and they were more sensitive to RMF/EBV-CTL-mediated killing in vitro [60% versus 25% at an effector-to-target (E:T) ratio of 50:1]. This characteristic might explain the differences observed in T-cell accumulations.

WT1/EBV-TG-CTLs did not accumulate in tumor T3. This rhabdomyosarcoma contained a high proportion of WT⁺ cells (73%) and was genotypically HLA-A0201⁺. However, it did not express the HLA-A0201 allele on the cell surface (1% HLA-class-I⁺ cells) and was not lysed by the HLA-A0201-restricted RMF/EBV-TG-CTLs in vitro.

As expected, WT1/EBV-TG-CTLs, also, did not accumulate in the WT1⁻ HLA-A0201⁺ B-ALL (Fig. 5 D, T4). These results, thus, demonstrate a requirement for the coexpression of the specific antigen and the T-cell-restricting HLA allele on tumor cells for T-cell targeting and tumor-selective accumulation in vivo.

Differential expression of WT1 in most leukemias (17, 18, 19) and several solid tumors (20, 21, 22, 23) has stimulated interest in WT1 as a potential target for immunotherapy. Initially, Ohminami et al. (9) isolated a CD8⁺ T-cell clone, termed TAK-1, which recognized the HLA-A2402⁺-binding WT1 peptide _235–243CMTWNQMNL. TAK-1 exhibited HLA-A2402-restricted cytotoxic activity against WT1⁺ leukemias and lymphomas, lysed WT1⁺ HLA-A2402⁺ lung cancer cell lines, and inhibited the growth of HLA-A2402⁺ WT1⁺ lung cancer cell-line xenografts in nude mice (24). Oka et al. (10) subsequently sensitized T cells from one HLA-A0201⁺ donor with T2 cells loaded with WT1 nonapeptides that contained major anchoring motifs for HLA-A0201, including RMF and SLG. T cells, sensitized with RMF, lysed HLA-A0201⁺ EBV-BLCLs loaded with peptide and WT1⁺ HLA-A0201⁺ leukemic cell lines. In a different approach, Gao et al. (11) used Drosophila cells transduced to express human HLA-A0201 and/or T2 cells loaded with RMF-peptide to sensitize T cells from an unspecified number of HLA-A0201⁻ donors. Selected clones lysed targets coexpressing both HLA-A0201 and WT1, including acute leukemia cell lines and Ph⁺ CD34⁺ chronic myelogenous leukemia progenitors expressing high levels of WT1. These HLA-A0201⁻ T cells, presumably specific for the HLA-A0201/WT1-peptide complex, did not lyse WT1⁻ HLA-A0201⁺ EBV-BLCLs or WT1⁺ HLA-A0201⁻ leukemic lines, nor did they lyse normal HLA-A0201⁺ CD34⁺ cells. Although this approach was reported to consistently yield WT1-reactive T cells, the relative immunogenicity of the WT1 peptides themselves cannot be assessed because the T-cell response is not specific for the WT1 peptide alone.

In our study, sensitization of T cells from 16 normal donors with the HLA-A0201-binding RMF and SLG WT1 peptides or the HLA-A2402-binding RVP-peptide loaded on autologous APCs regularly yielded CD8⁺ peptide-specific cytotoxic T cells. Among HLA-A0201⁺ donors, the percentages of SLG and RMF/HLA-A0201-tetramer-binding T cells closely correlated with the percentage of T cells generating intracellular IFNγ in response to restimulation with peptide in vitro. In addition, these T cells exhibited HLA-A0201-restricted peptide-specific cytotoxicity against both peptide-loaded targets and WT1⁺ fresh leukemic cells. T cells from each of five HLA-A2402⁺ donors, who were sensitized with autologous EBV-BLCL loaded with the RVP-peptide, also specifically lysed peptide-loaded HLA-A2402⁺ PHA blasts and unmodified HLA-A2402⁺ leukemias. In two individuals expressing both HLA-A0201 and HLA-A2402, the responses to RMF and RVP peptides were equivalent. The activity of these peptide-specific T cells against WT1⁺ HLA-A2402⁺ leukemia indicates that the RVP peptide is naturally presented after intracellular processing of WT1 protein.

The in vitro generation of T cells for adoptive cell therapy can often be thwarted by limited or inconsistent availability of donor-derived APCs, required for repeated in vitro restimulations of T cells during selection and expansion of antigen-specific CTL. EBV-BLCLs have advantages as APCs because they can be grown indefinitely and can express those determinants essential for antigen presentation. Indeed, transformation by EBV leads to enhanced APC function (through the up-regulation of the peptide transporters TAP-1 and TAP-2) as well as to enhanced expression of MHC class I (41). EBV-BLCLs, transduced to express cytomegalovirus pp65, are highly effective in inducing cytomegalovirus pp65-specific T cells at levels equal to, or exceeding, those cogenerated against peptides of immunodominant EBV antigens such as EBNA-3 (42). However, from prior reports, it was unclear whether EBV-BLCLs, loaded with immunogenic peptides that were derived from a self antigen such as WT1 protein, would concomitantly elicit a significant WT1-specific T-cell response. Indeed, we expected that responses to WT1 peptides might be diluted out by responses to dominant EBV epitopes. Strikingly, however, in cultures sensitized with peptide-loaded autologous EBV-BLCLs, the percentages of T cells specific for two different HLA-A0201-binding WT1 epitopes, RMF and SLG, were comparable with or exceeded those detected in populations of the same T cells contemporaneously sensitized with peptide-loaded CAMs. Furthermore, the total yield of WT1-peptide-specific T cells was consistently higher in cultures sensitized with the WT1-peptide-loaded EBV-BLCL.

It has been speculated that, because WT1 is a self protein, T-cell responses to WT1 peptides would be low and T cells, responding to self antigens, are subject to early deletion in the thymus and, if exported, are suppressed by regulatory T cells in normal individuals (43, 44). For these reasons, we expected that the frequencies of WT1-specific T cells in cultures sensitized with peptide-loaded EBV-BLCLs would be low, and markedly lower than those generated against immunodominant EBV antigens. However, in the T cells sensitized with WT1-peptide-loaded EBV-BLCL, the proportions of RMF- and SLG-HLA-A0201-tetramer-binding T cells were actually equal to those in T cells that bound tetramers of HLA-A0201 and immunogenic peptides of EBV. Thus, although the cumulative response to the spectrum of EBV antigens detected in the T-cell lines was much greater, responses to the WT1 peptide were surprisingly comparable with those generated against individual viral epitopes known to be highly immunogenic.

The T cells generated from HLA-A0201⁺ and HLA-A2402⁺ donors, sensitized with RMF and SLG(HLA-A0201) or RVP(HLA-A2402), were specifically cytotoxic for primary AML, B-ALL, and T-ALL, and leukemic cell lines, as well as for solid tumor cells expressing WT1 and their restricting HLA allele. In our studies, there was a significant correlation between the proportion of WT1⁺ cells detected in each target population and the proportion of cells in that population lysed by the WT1-peptide-specific T cells. However, the mean fluorescence intensity of WT1 in the WT1⁺ cells was strikingly high and very similar within each of the target cell populations. Because detection of WT1 in target cells necessitated permeabilization and fixation, we could not determine whether the WT1⁺ cells that were detected in high proportions in tumor cell fractions resembled those in the small fraction of WT1⁺ cells detected in cord blood or isolated CD34⁺ cells, in their sensitivity to WT1-peptide-specific cytotoxic T cells. Inoue et al. (45) have reported that the concentration of WT1 RNA detected in isolated populations of normal CD34⁺ marrow cells is 10- to 100-fold lower than that detected in populations of leukemic blasts. On the basis of these data and similar data from Gao et al. (11), who measured WT1 RNA and protein in CD34⁺ cell fractions, it could be hypothesized that normal CD34⁺ cells express less WT1 and are, therefore, less susceptible to lysis by WT1-peptide-specific T cells. However, when Hosen et al. (46) subsequently measured the WT1 RNA content of single CD34⁺ cells derived from the marrows of a series of normal donors, they demonstrated that, in fact, only 1 to 2% of the CD34⁺ Lin⁻ cells express any WT1 RNA at a given time and that, in this small fraction of cells, the level of WT1 was comparable with that detected in leukemic cells (46). These findings are consistent with our findings that cytolysis is proportional to the fraction of cells expressing WT1 in the target population at the time of exposure to the WT1-peptide-specific T cells.

The fact that WT1 protein can be detected in a small proportion of early CD34⁺ hematopoietic progenitors has raised concerns regarding the potential myelosuppressive effects of WT1-specific T cells. In our studies, WT1-peptide-specific T cells neither lysed normal cord blood cells or CD34⁺ cells nor inhibited the growth of CFU-GM, burst-forming unit erythroid, or mixed colonies from the isolated CD34⁺ cells. Ohminami et al. (9) also demonstrated that TAK-1 did not inhibit the growth of CFU-GM from the marrow of normal HLA-A2401⁺ donors. Similarly, the WT1-HLA-A0201-specific T-cell line developed by Gao et al. (11, 29) lysed WT1⁺ CD34⁺ chronic myelogenous leukemia cells, inhibited their colony-forming activity in vitro and prevented their growth in NOD-SCID mice but had no effect on normal HLA-A0201⁺ CD34⁺ cells (29). In murine models, WT1 plasmid vaccines, inducing responses sufficient to prevent the growth of WT1⁺ tumors, have also not affected normal hematopoietic function (47). Similarly, in leukemic patients who achieve remissions and who have detectable antibody or T-cell responses to WT1, impaired recovery of blood counts has not been detected (26, 27). These findings suggest that, either normal CD34⁺ hematopoietic cells contain clonogenic progenitors in both WT1⁺ and WT1⁻ fractions, or, as has been suggested by experiments with WT1 antisense oligonucleotides (48, 49), expression of specific isoforms of WT1 may be essential to the growth of clonogenic leukemic cells.

Although Gao et al. (29) have reported that preincubation of WT1⁺ human leukemic blasts with T-cell lines can deplete clonogenic cells capable of transferring leukemia into NOD-SCID mice, the effects of WT1-peptide-specific T cells on established leukemic grafts have not heretofore been evaluated. Our studies demonstrate that after adoptive transfer into NOD-SCID mice that are bearing established leukemic xenografts, the WT1-peptide-specific T cells selectively infiltrate WT1⁺ leukemias bearing the restricting HLA-allele and significantly inhibit their growth.

In mice treated with WT1/EBV-CTLs transduced with the HSV-TK/eGFP vector, selective accumulation of T cells in WT1⁺ leukemia or rhabdomyosarcoma grafts was also documented in vivo by either scintigraphic or PET imaging after sequential intravenous infusions of [¹³¹I]FIAU or [¹⁸F]FEAU. We have previously used this approach to demonstrate that HLA-restricted EBV-specific T cells selectively accumulate in and induce regressions of EBV-lymphoma xenografts bearing the T cells restricting HLA-allele (31, 36), a finding confirmed here in studies of mice bearing HLA-A0201⁺ EBV-BLCL as well as WT1⁺ and WT1⁻ leukemia xenografts, treated with EBV-TG-CTLs alone. However, in mice bearing the same group of tumors treated with HLA-A0201-restricted dual-specific WT1/EBV-TG-CTLs, the T cells also accumulated in the HLA-A0201⁺ WT1⁺ leukemia xenografts. In these animals, accumulations in the HLA-A0201⁺ EBV-lymphomas were markedly higher then those in the HLA-A0201⁺ WT1⁺ leukemias, reflecting the 10-fold higher cumulative number of infused T cells specific for EBV peptides in the transduced T-cell population. Similarly, in the mice bearing leukemia and rhabdomyosarcoma xenografts that varied in the expression of WT1 or the HLA-A0201 restricting allele, WT1/EBV-TG-CTLs selectively accumulated only in tumors coexpressing WT1 and HLA-A0201.

Immunohistologic analysis of tumors from patients with several forms of cancer has documented the absence of HLA-class-I expression in a substantial proportion of cases (50). Such tumors are resistant to cytotoxic T cells in vitro. Thus, modulation of HLA expression has been hypothesized to be a major mechanism whereby tumors can elude the immune system. In our studies, a WT1⁺ rhabdomyosarcoma that failed to express its inherited HLA-A0201 allele and was resistant to lysis by WT1-peptide-specific HLA-A0201--restricted T cells in vitro also could not be targeted by the WT1-specific T cells in vivo (Fig. 5 D). Thus, tumors, or metastases thereof, that fail to express HLA-class-I in vivo may not only be resistant to lysis by specific T cells but may also be untargetable by such T cells in the circulation.

In conclusion, our studies demonstrate that T cells isolated from normal HLA-A0201⁺ and HLA-A2402⁺ individuals can be regularly sensitized with immunogenic peptides of the WT1 protein loaded on autologous APCs, and that the frequencies of the WT1-peptide-specific T cells that are generated are comparable with those generated against single immunogenic peptides of EBV. These WT1-peptide-specific HLA-restricted T cells selectively lyse WT1⁺ leukemias and solid tumors in vitro and selectively migrate to, accumulate in, and induce regressions of WT1⁺ leukemic xenografts bearing the appropriate restricting HLA allele in vivo. These studies, thus, describe a clinically applicable approach to the generation of WT1-specific tumor-reactive T cells and provide further support for the potential use of WT1-peptide-specific T cells as effectors for adoptive immunotherapy in humans.

We thank Drs. Malcolm A. S. Moore and Juri Gelovani for their advice in this study; Elena Kanaeva, Anna Bierwiaczonek, and Judith Guerrero for excellent technical support; Pat Zanzonico, for assistance in imaging studies; Julius Balatoni for [¹³¹I]FIAU synthesis; Nagavarakishore Pillarsetty for [¹⁸F]FEAU synthesis; Ronald Ghossein, for immunohistochemistry; and Ingrid Leiner for the generation of tetramers.