Early Detection and Rapid Isolation of Leukemia-Reactive Donor T Cells for Adoptive Transfer Using the IFN-γ Secretion Assay

Allogeneic stem cell transplantation is successfully applied in the treatment of hematologic malignancies such as acute myeloid leukemia (AML), chronic myeloid leukemia (CML), acute lymphoblastic leukemia (ALL), and chronic lymphocytic leukemia (CLL). A higher incidence of recurrent disease is observed when the grafts are depleted of T cells, illustrating the importance of the graft-versus-leukemia effect exerted by donor-derived T cells (1–3). Stable engraftment of donor cells after transplantation allows the potential application of further cellular immunotherapeutic approaches like infusions of unmodified donor T cells (donor lymphocyte infusion). Because after allogeneic stem cell transplantation, normal hematopoiesis in these patients is of donor origin, alloreactive T cells recognizing polymorphic antigens expressed by hematopoietic cells of the patient will have a relative specificity against the malignant cells. However, the induction of chronic and acute phase graft-versus-host disease remains a major cause of morbidity after donor lymphocyte infusion, probably due to the lack of specificity of nonselected donor lymphocytes for the various tissues from the patient.

In patients responding to donor lymphocyte infusion in the absence of severe graft-versus-host disease, we have previously shown the development of immune responses directed against hematopoiesis or leukemia-associated minor histocompatibility antigens, which are differentially expressed in donors and patients (4–6). Especially in patients with CML in chronic phase, treatment with unmodified donor lymphocyte infusion seems to be effective, resulting in complete responses in 70% to 80% of the patients (7, 8). However, in patients with more rapidly progressing diseases like AML or CML in accelerated phase or blast crisis, and in the B-cell malignancies, CLL, and ALL, complete response rates of only 10% to 30% can be achieved (9, 10). The poor immunogenicity of these malignancies is likely to play an important role in the lack of induction of a proper antileukemic immune response by unmodified T cells (11–14). This limitation may be overcome by the in vitro generation of leukemia-reactive T cells that can be adoptively transferred into patients with relapsed disease after transplantation (15–21). Repetitive in vitro exposure of donor T cells to in vitro generated leukemic antigen-presenting cells (APC) has been shown to result in increased specificity for the leukemia (21, 22). Therefore, in vitro enrichment of these leukemia-reactive T cells and the relative depletion of other alloreactive T cells from the adoptively transferred T-cell population will probably decrease the risk of inducing graft-versus-host disease.

The feasibility and toxicity of this approach has been investigated in a phase I/II clinical study (21, 22). Although a potential efficacy of the adoptive transfer of in vitro generated leukemia-reactive T cells was seen in 4 of 13 patients, the long-term in vitro and in vivo survival of the T cells seemed to be limited, and the development of sustained immunologic memory could not be shown in most cases. The repetitive stimulations and extensive culture period needed for the enrichment, selection, and expansion of leukemia-reactive T cells using the previously described limiting dilution assay method, may have led to the selection of a population of T cells with a limited residual capacity to survive and expand in vivo (23, 24).

In previous studies, we have shown that it is possible to directly isolate leukemia-reactive T cells from peripheral blood of patients treated with donor lymphocyte infusion using selection and isolation based on the production of IFN-γ in response to stimulation with leukemic cells (4, 5). This procedure using the IFN-γ secretion assay (25, 26) seemed to select for a population of T cells containing a high frequency of leukemia-reactive T cells with direct cytolytic activity against the hematopoietic cells from the patient. In the present study, we adapted this procedure for the early detection and isolation of leukemia-reactive T cells from primary immune responses. Immunogenic APC were generated from primary leukemia, and HLA-matched donor T cells were cocultured with these leukemic APC for 2 weeks under mild stimulatory conditions to enrich for the leukemia-reactive T cells. After this enrichment period, the T cells were specifically restimulated with the leukemic APC resulting in a synchronized IFN-γ production by the responding T cells and efficient isolation of leukemia-reactive T cells using the IFN-γ secretion assay. Using this procedure, we could reproducibly isolate leukemia-reactive T cells with high proliferative potential and with apparent specificity for cells of the hematopoietic system of the patient. Therefore, these results are likely to improve the perspectives for adoptive immunotherapy in the context of allogeneic stem cell transplantation in patients with relapsed CML, AML, CLL, and ALL by increasing both the effectiveness and the specificity.

Donor-patient pairs, cell collection, and preparation. After informed consent, peripheral blood and/or bone marrow was obtained from three patients with chronic-phase CML, two patients with AML (French-American-British type M1), three patients with B-cell CLL, and three patients with B-cell ALL before treatment. In addition, peripheral blood was obtained from the HLA-identical stem cell donors of these patients. Mononuclear cells were isolated by Ficoll density separation and cryopreserved for further use. Standard flow cytometric phenotyping showed that >90% of the peripheral blood mononuclear cells from AML, CLL, and ALL patients were of malignant origin. CD34⁺ CML precursor cells comprising 20%, 7%, and 10% of peripheral blood mononuclear cells in patients 1 to 3, respectively, were isolated from the peripheral blood mononuclear cells using clinical grade CD34 Clinimacs selection (Miltenyi Biotech, Bergisch Gladbach, Germany) as described previously (27). Mesenchymal stem cells (MSC) were derived from the bone marrow of two patients with CML by culturing the adherent cells for several weeks on DMEM supplemented with l-alanyl-l-glutamine, sodium pyruvate, 1 mg/mL glucose, and pyridoxine (Gibco, Invitrogen, Oslo, Norway), and 10% fetal bovine serum.

Modification of leukemic cells into malignant APC. Leukemic cells were modified into malignant APC using various disease-specific strategies.

AML blasts were seeded at a concentration of 1 × 10 E6 cells/mL in Iscove's modified Dulbecco's medium (IMDM, BioWhittaker, Verviers, Belgium) containing 10% irradiated fetal bovine serum (Cambrex BioScience, Verviers, Belgium; approved for use under good manufacturing practice conditions), and cultured for 4 days in the presence of 100 ng/mL granulocyte macrophage colony-stimulating factor (Novartis, Basel, Switzerland), 20 ng/mL stem cell factor (Amgen, Thousand Oaks, CA), 10 ng/mL tumor necrosis factor-α (Cellgro 1006; Cellgenix, Freiburg, Germany), and 500 IU/mL IL-4 (Schering-Plough, Innishammon, Cork, Ireland). Subsequently, the cells were matured for 2 days using monocyte conditioned medium–mimic (28, 29), containing 10 ng/mL IL-1β (Cellgenix), 10 ng/mL IL-6 (Cellgenix), 10 ng/mL tumor necrosis factor-α, 500 IU/mL IFN-γ (Immukine; Boehringer Ingelheim, Alkmaar, the Netherlands), and 1 μg/mL prostaglandin E₂ (Sigma-Aldrich, Zwijndrecht, the Netherlands).

Purified CD34⁺ CML precursor cells were cultured for 7 days at a concentration of 0.5 × 10 E6/mL in IMDM containing 10% FCS, 100 ng/mL granulocyte macrophage colony-stimulating factor, 20 ng/mL stem cell factor, and 2 ng/mL tumor necrosis factor-α. At day 4, the medium was refreshed and IL-4 (500 IU/mL) was added. At day 7, the cells were washed to remove the FCS, resuspended in IMDM containing 10% prescreened human serum, followed by 2 days of maturation using the monocyte conditioned medium–mimic as described above.

As described previously, CLL and ALL were modified into malignant APC by triggering of CD40 (30). To induce CD40 expression, ALL cells were first cultured in IMDM, containing 10% prescreened human serum, 500 IU/mL IL-4, and 10 μg/mL synthetic CpG oligonucleotides (5′-TCGTCGTTTTGTCGTTTTGTCGTT-3′, Eurogentec, Seraing, Belgium) at a concentration of 1 × 10 E6 cells/mL for 1 day. Subsequently, both CLL and ALL cells were cultured for 3 days at a concentration of 1 × 10 E6 cells/mL in six-well plates (Costar, Cambridge, MA) in IMDM with 10% FCS, and 500 IU/mL IL-4 on a monolayer of 4 × 10 E5 irradiated ltk murine fibroblast cells transfected with human CD40 ligand (tC40L; kindly provided by Dr. C. van Kooten, Department of Nephrology, Leiden University Medical Center).

Immunophenotyping of malignant APC and cytokine measurements. Immunophenotype analysis of the leukemic APC was done by staining the cells with FITC-labeled CD1a (BD PharMingen, San Diego, CA), CD40 (Serotec, Oxford, United Kingdom), CD86 (BD PharMingen), and HLA-DR (BD Biosciences, San Jose, CA) monoclonal antibodies, and phycoerythrin-labeled CD11c (BD), CD19 (BD), CD33 (BD), CD34 (BD), CD80 (BD), and CD83 (Sanbio, Uden, the Netherlands) monoclonal antibody. Expression was analyzed using a FACSCalibur and Cellquest software (BD).

Cytokine measurements were done using commercial IL-10 and IL-12 p40/p70 (U-CYTech, Utrecht, the Netherlands) ELISA kits according to the manufacturer's instructions.

Activation and isolation of leukemia-reactive T cells. Monocytes were depleted from peripheral blood mononuclear cells of related or unrelated HLA-matched donors using Clinimacs CD14 beads (Miltenyi Biotech) according to the manufacturer's instructions. CD14-depleted responder cells were cocultured for 2 weeks with irradiated (25 Gy) malignant APC at a 10:1 responder to stimulator ratio in IMDM containing 10% prescreened human serum. At days 7 and 11, a low concentration of interleukin 2 (IL-2; 10 IU/mL; Chiron, Amsterdam, the Netherlands) was added to the culture. At day 14, the T cells were restimulated with irradiated (25 Gy) malignant APC in a 10:1 responder to stimulator ratio. Sixteen hours after this restimulation, the T cells responding to the malignant APC by producing IFN-γ were stained using the IFN-γ secretion assay (Miltenyi Biotech) and isolated by MACS isolation (bulk) or by single cell per well or bulk fluorescence-activated cell sorting (FACS) as described previously (4). Briefly, cells were incubated with IFN-γ catch reagent and cultured for 45 min at 37°C. Cells were counterstained with phycoerythrin-labeled IFN-γ antibody. If IFN-γ–producing cells were isolated via magnetic bead separation, the IFN-γ phycoerythrin-positive cells were bound to antiphycoerythrin microbeads (Miltenyi Biotech) and isolated using the midi-MACS system (Miltenyi Biotech). The purity of the isolated fractions was analyzed on a BD FACSCalibur after counterstaining with FITC-labeled CD4 (BD) or CD8 (Caltag, Invitrogen, Oslo, Norway) antibodies, and APC-labeled CD3 antibodies (BD). Low-dose propidium iodide (0.2 μg/mL) was added to exclude death cells from the analysis. For FACS, the T cells were counterstained using FITC-labeled CD4 and CD8 monoclonal antibodies, and propidium iodide (0.2 μg/mL) was added. Viable (propidium iodide–negative) CD4/8–positive, IFN-γ–positive cells were sorted either in bulk cultures or single cell per well using a FACSVantage (BD). Viable IFN-γ negative, CD4/8–positive cells were sorted as enrichment control. After isolation, bulk cultures were cultured for 5 days in culture medium containing 25 IU/mL IL-2 before they were tested for their cytotoxic capacity.

Analysis of cytotoxicity and cytokine production. Cytotoxicity of the IFN-γ–positive and negative bulk populations against unmodified leukemic cells, malignant APC, and donor-derived phytohemagglutinin blasts was analyzed using the carboxyfluorescein diacetate succinimidyl ester (CFSE)–based cytotoxicity assay (31). In short, effector cell populations were added at low effector to target ratios (3:1 to 0.1:1) to 5,000 CFSE–labeled (2.5 μmol/L) target cells in 96-well plates in IMDM, containing 10% human serum and 25 IU/mL IL-2. After 24 h of coculture, the cells were stained with an APC-labeled leukemia marker (CD19, CD33, CD13, or CD34; BD) and CD3-phycoerythrin (BD). The wells were harvested, a fixed amount of fluorescent microspheres was added (Flowcount beads; Beckman Coulter, Mijdrecht, the Netherlands) to allow quantitative analysis, and death cell exclusion was done using low-dose propidium iodide (0.2 μg/mL). The cells were directly analyzed using a FACSCalibur and Cellquest software (BD). Absolute numbers of surviving leukemic cells were calculated, and the percentage of specific lysis of the leukemic (precursor) cells was calculated compared with the absolute number of leukemic cells at time point 0 using the following formula:

Cytotoxicity of T-cell clones was determined in standard 4 and 16 h ⁵¹Cr release assays (31). As target cells, unmodified leukemic cells and malignant APCs were used. Target cells were incubated with effector cells at a 5:1 effector to target ratio. HLA class I and II blocking studies were done by incubating target cells with saturating concentrations of anti–HLA class I (W6.32) or anti–HLA class II (PdV5.2) antibodies for 30 min at room temperature before effector and target cells were cocultured.

For analysis of IFN-γ production, 5,000 T cells were cocultured with 30,000 target cells. After 24 h, supernatants were harvested and the concentration of IFN-γ was measured by ELISA (CLB, Amsterdam, the Netherlands). When MSCs were used as target cells 5,000 T cells were cocultured on a monolayer of 5,000 plated MSCs in 96-well flat-bottomed plates.

Statistical analysis. Statistical evaluation of the data was done using the paired Student's t test.

Generation of malignant APC. As shown in Table 1, the expression of markers associated with the APC function of primary AML (n = 2), CML CD34⁺ (n = 3), ALL (n = 3), and CLL (n = 3) cells was generally low. Almost no primary leukemic cells showed expression of CD80 (median 2%, range 0-14%) and CD83 (median 5%, range 1-24%). CD86 expression was present on three of six B-cell malignancies, but was absent on most myeloid malignancies (median 6%, range 1-99%). The number of HLA-DR/CD11c coexpressing cells was variable between the patients (median 13%, range 1-68%) and did not correlate with the expression of the other markers. To improve the APC phenotype, the leukemic blasts were matured using different disease-specific protocols as outlined in Materials and Methods section. As summarized in Table 1, these maturation protocols significantly (all P < 0.01) increased the expression of CD80 (median 73%, range 13-94%), CD86 (median 69%, range 31-99%), HLA-DR/CD11c (median 64%, range 2-91%), and CD83 (median 77%, range 27-99%), resulting in the development of phenotypically appropriate APC with a relatively mature phenotype from all four types of leukemic cells. In Fig. 1, the intensity of the expression of CD80 and CD86 on the leukemic APC is shown. For the myeloid malignancies AML and CML, only part (30-70%) of the cells acquired a professional APC phenotype, whereas for the B-cell malignancies ALL and CLL, almost 100% of the cells expressed a professional APC phenotype. Whereas none of the leukemic APC produced the T-cell inhibitory cytokine IL-10, only the malignant APC generated from the B-cell malignancies ALL and CLL produced IL-12 (median 800 pg/mL, range 230-1,510 pg/mL).

Activation and isolation of leukemia-reactive T cells from primary immune responses using malignant APC and HLA-matched donor T cells. Initially, we investigated whether it was feasible to use the IFN-γ secretion assay as described previously (4, 5, 32) for the detection and isolation of leukemia-reactive T cells from primary antileukemia immune responses after a single stimulation of donor T cells with the leukemic APC. In three pilot experiments, we clonally isolated the IFN-γ–producing donor T cells (0.1-0.3% of total) after overnight stimulation with leukemic APC. However, none of the isolated T-cell clones exerted reactivity against the leukemic cells of the patient (data not shown). Due to the technical makeup of the IFN-γ secretion assay, only cells producing IFN-γ within the fixed 45-min secretion period can be isolated. Because it was already observed that the IFN-γ production of donor T cells after stimulation with either anti–CD3/CD28 or HLA-mismatched stimulator cells occurs in a nonsynchronized way during the first days of the immune response (33), we hypothesized that this same phenomenon hampered the isolation of leukemia-reactive T cells during the first days of the immune response.

Therefore, we developed a strategy to synchronize the production of IFN-γ by the leukemia-reactive T cells. Primary immune responses were initiated from the 11 HLA-matched donor-recipient pairs using the leukemic APC as stimulator cells. CD14-depleted donor PBMCs were stimulated with the APC in a 10:1 responder to stimulator ratio. The cells were cocultured for 2 weeks in the presence of a very low concentration of IL-2 (10 IU/mL) sufficient for the T cells to survive, but not inducing large-scale proliferation (median 79% recovery, range 42-124%). After this 2-week culture period, the activation status of the T cells was measured by expression of HLA-DR and CD25 and the production of IFN-γ. Although part of the T cells (median 51%, range 15-69%) still expressed the activation markers HLA-DR and CD25, the number of IFN-γ–producing T cells at day 14 was low (median 1.2%, range 0-3.8%, n = 11; Fig. 2, white columns). After 16 h of specific restimulation with the leukemic APC in a 10:1 responder to stimulator ratio, the number of IFN-γ–producing T cells was significantly increased to 3.9% (range 1.2-7.5%, n = 11, P = 0.0026; Fig. 2, black columns). As shown in Table 2, the number of CD4-positive T cells responding to the leukemic APC by producing IFN-γ was significantly higher (median 2.2%, range 0.8-7.0%) than the number of IFN-γ–producing, CD8-positive T cells (median 0.6%, range 0.2-3.1%, P = 0.00029). IFN-γ–producing cells were then isolated using FACS or magnetic bead separation. A representative example of the latter is shown in Fig. 3. The IFN-γ–positive fractions contained 81% to 98% T cells (median 94%; Table 2).

To test the residual proliferative capacity of the leukemia-reactive T cells, T-cell expansion was induced by the addition of CD3/28 T-cell expander beads (Dynal, Invitrogen, Paisley, United Kingdom). Vigorous expansion of the leukemia-reactive T cells was observed, which was comparable with the proliferative capacity of normal unmodified donor T cells (data not shown).

Antileukemic activity of T-cell populations isolated on the basis of specific IFN-γ production. The isolated T-cell fractions were cultured for an additional 5 days in the presence of low-dose IL-2 (25 IU/mL). Subsequently, functional antileukemic activity of the isolated T-cell fractions was analyzed against unmodified primary leukemic cells and/or leukemic APC using the carboxyfluorescein diacetate succinimidyl ester–based cytotoxicity assay (31). As shown in Fig. 4A, the IFN-γ positively selected T-cell fractions added at an effector to target ratio of only 1:1 lysed 15% to 70% (median 32%) of unmodified primary leukemic cells, whereas the addition of the IFN-γ–negative T-cell fractions resulted in only 0% to 20% lysis (median 5%, P < 0.001). Because CD40 triggering does not only result in activation but also in proliferation of B cell malignancies, the anti-ALL and anti-CLL responses were also tested against CD40-triggered leukemic APC (Fig. 4B). CLL or ALL APC showed 10% to 73% lysis (median 35%) after exposure to the IFN-γ–positive T-cell fractions and 0% to 14% lysis (median 2.5%, P = 0.004) after exposure to the IFN-γ–negative T-cell fractions. CD3/28 expansion of the IFN-γ–positive fractions did not affect the antileukemic activity of the T cells (data not shown).

From two CML patients, we were able to culture MSCs from the bone marrow of the patients. These cells were used as target cells to determine the possible reactivity of the leukemia-reactive T cells against nonhematopoietic tissues (Fig. 5). Whereas the leukemia-reactive T cells produced high amounts of IFN-γ (median 515 pg/mL) against the CD34⁺ CML cells, no IFN-γ was produced after coculture on the MSCs. In accordance, microscopic analysis showed that the monolayer of MSCs was still intact after coculture with the leukemia-reactive T cells (data not shown), whereas coculture with a positive alloreactive control CTL clone resulted in kill and detachment of the MSCs from the culture plate, coincided by IFN-γ production by the T-cell clone (median 42 pg/mL).

Frequency determination by single-cell cloning. To determine the frequency and recognition pattern of leukemia-reactive T cells, we additionally sorted the IFN-γ–producing T cells, single cell per well from six responses (CML 2 and 3, AML 1, CLL 2, and ALL 1 and 2). The proliferating T-cell clones were tested for their capacity to recognize unmodified primary leukemic cells and leukemic APC in conventional ⁵¹Cr release assays and/or by IFN-γ ELISA. As summarized in Table 3, mainly CD4-positive T-cell clones and a limited number of CD8-positive T-cell clones were isolated. Most T-cell clones that were cytotoxic against the leukemia also produced IFN-γ in response to stimulation with the leukemia. Some clones capable of IFN-γ production upon recognition of the leukemia were not cytotoxic. In addition, from both anti-CML responses, some proliferating natural killer cell clones were generated. Although these NK clones were cytotoxic against the leukemia, no specific IFN-γ production in response to stimulation by the leukemic cells was observed. The overall frequency of clones reactive against the leukemia ranged from 8% to 53%. A substantial number of clones from each response were tested in HLA-blocking experiments. The lysis and recognition could be specifically blocked by the relevant anti-HLA antibodies (median 95% inhibition, range 62-100%, n = 16), illustrating normal HLA-restricted lysis by these clones. In addition, we tested the clones derived from several responses for their capacity to kill nonleukemic hematopoietic targets of donor and patient origin. Low recognition of phytohemagglutinin blasts from the patients was seen (median 8% lysis, range 0-15%), whereas phytohemagglutinin blasts and EBV-transformed B cells of the donor were not recognized, indicating recognition of minor histocompatibility antigens (mHag) differentially expressed between donor and recipient. Differential recognition of several partially HLA-matched third-party EBV-LCL further showed mHag-specific recognition by these T-cell clones (data not shown).

The aim of this study was to develop a new strategy to isolate leukemia-reactive T cells with a better probability to survive and expand in vivo. In a previous phase I/II feasibility study, we treated patients with relapsed leukemia after allogeneic stem cell transplantation with leukemia-reactive donor T cells that were repetitively stimulated with relatively immature malignant APCs (21, 22). Although proof-of-principle was shown, no sustained immunologic memory could be achieved in most patients. In addition, the limiting dilution assay–based method used to induce the cultures and to select for leukemia-reactive T cells is complex and time-consuming. The extensive in vitro culture period needed for the selection and expansion of sufficient numbers of leukemia-reactive T cells might have led to the infusion of T-cell lines containing cells with a limited capacity to survive and expand in vivo. In mice, it has been shown that the length of the in vitro culture of T cells, especially in the presence of high-dose IL-2, inversely correlated with their in vivo survival (23, 24). Furthermore, selection based solely on the cytolytic effector function of T cells might have led to enrichment of a population of effector end cells. Therefore, in the development of a new strategy, we aimed at a reduced in vitro expansion period and selection of the T cells on basis of their specific cytokine production rather than their cytolytic effector function.

Virus-specific T cells isolated on the basis of their specific production of IFN-γ have been reported to be capable of vigorous proliferation and exerted antigen-specific cytotoxicity in vitro and in vivo (25, 34, 35). Several reports have been published on the beneficial role of IFN-γ–producing Th1 type CD4⁺ T cells in combination with cytotoxic CD8⁺ T cells in the control of viral infections and in antitumor responses (36–39). We previously showed that isolation of leukemia-reactive mHag-specific T cells from patients with leukemia responding to donor lymphocyte infusion using the IFN-γ secretion assay was feasible (4, 5). These T cells isolated on basis of their specific IFN-γ production were capable of vigorous expansion and exerted mHag-specific HLA-restricted lysis of hematopoietic cells of the patient. Based on these results, we aimed at developing a strategy to isolate leukemia-reactive CD4⁺ and CD8⁺ T cells from primary immune responses based on their specific secretion of IFN-γ to be used for adoptive transfer.

To induce leukemia-reactive T cells, we first generated professional APC from leukemic (progenitor) cells, because the use of nonprofessional immature APC may lead to the induction of tolerance (40–42). The immunogenicity of leukemic blasts was increased by transforming these cells into phenotypically mature APC with good expression of the costimulatory molecules CD40, CD80, CD86, CD83, and HLA-DR/11c. The myeloid malignancies CML and AML were matured using a cytokine cocktail previously described as monocyte conditioned medium–mimic (28). All components needed for this maturation protocol are available as clinical grade reagents, allowing direct implementation of this procedure for the generation under good manufacturing practice conditions. However, a clinical grade protocol needs to be developed allowing CD40 ligation under good manufacturing practice conditions, a step that is necessary for the B-cell malignancies. The leukemic APC were used as stimulator cells to activate HLA-matched donor T cells. In pilot studies (data not shown), we analyzed the IFN-γ production of these T cells during the first 3 days after stimulation. Due to the technical makeup of the cytokine secretion assay, only the T cells that produce IFN-γ within the fixed 45-min culture period between the labeling with the catch reagent and the staining with the detection antibody can be isolated using the IFN-γ secretion assay. Extension of this period leads to aspecific binding of IFN-γ to neighboring cells. Similar to previous observations (33), we found that the nonsynchronic IFN-γ production at initiation of the culture hampered the isolation in the first 3 days of these primary immune responses. This was in contrast to the studies we did on secondary stimulations of in vivo primed leukemia-responsive donor T cells isolated from patients responding to donor lymphocyte infusion (4, 5), indicating a more synchronized IFN-γ production in these responses.

Therefore, we developed a culture protocol aiming at more synchronized production of IFN-γ by the leukemia-reactive T cells during primary immune responses, allowing a more efficient isolation using the IFN-γ secretion assay. To achieve this, we first specifically stimulated the donor T cells with the leukemic APC followed by culture for 14 days without further stimulation in the absence of other proliferation-inducing agents. The low dose of IL-2 (10 IU/mL) added to the culture was sufficient for the T cells to survive, but did not induce expansion of the total cell population, resulting in a 0% to 58% decline of the initial cell numbers. After this culture period, only few T cells still produced IFN-γ despite the expression of the activation markers HLA-DR and CD25. After the second specific stimulation of the T cells with the leukemic APC, a significant percentage of the T cells responded by the production of IFN-γ. Isolation of these IFN-γ–producing T cells resulted in a clear enrichment for leukemia-reactive T cells in the IFN-γ–positive fraction, whereas only marginal cytotoxic activity resided in the IFN-γ–negative fraction, illustrating the relatively synchronized production of IFN-γ within the fixed 45-min secretion period by the leukemia-reactive T cells after the second stimulation. In a limited number of patients, we were able to culture MSCs from the bone marrow of the patient. Using these MSCs as nonhematopoietic control cells in a T-cell activation assay, we were able to show the relative hematopoiesis specificity of the isolated leukemia-reactive T cells. In contrast to other studies in which extensive cloning and expansion protocols were needed to select for the right T cells (43, 44), we were able to select the leukemia-reactive T cells early in the immune response. CD3/28 activation experiments showed the high residual proliferative capacity of the isolated T cells. This expansion procedure did not affect the antileukemic reactivity of the T cells.

To determine the frequency of leukemia-reactive T cells within the positively isolated fraction, we clonally expanded 6 of 11 of the responses. These analyses revealed that 8% to 53% of the T-cell clones was capable of exerting specific reactivity against the leukemia either by classic HLA-restricted lysis or specific production of IFN-γ. No toxicity or recognition was found against phytohemagglutinin blasts of donor origin, illustrating that the reactivity of the T cells is probably directed against mHags that are differentially expressed on cells of donor and recipient origin. Differential recognition of third-party EBV-LCL expressing the relevant HLA molecules further strengthened this assumption. From some responses, a number of NK cell clones were expanded. Although capable of lysing the leukemic cells, no IFN-γ production was seen, indicating that these NK cells were either aspecifically coisolated with the IFN-γ secretion assay or that they lost their IFN-γ–productive capacity due to the in vitro expansion and culture conditions.

In conclusion, we here present a new approach for the isolation of leukemia-reactive T cells to be used for the treatment of patients with relapsed leukemia after allogeneic stem cell transplantation. In this study, we developed one general stimulation and isolation protocol that can be used for all different types of leukemia. Using this procedure, we were able to reproducibly enrich for leukemia-reactive T cells with apparent hematopoiesis specificity and high residual proliferative capacity. The increased immunogenicity of the leukemic APCs made it possible to circumvent the problem of anergy induction, an entity probably underlying the ineffectiveness of donor lymphocyte infusion by the majority of patients with B-cell malignancies. In addition, the isolation at an early stage in the immune response and the short in vitro culture period under mild stimulatory conditions will probably result in an increased capacity of the T cells to survive and expand in vivo.