Induction of anti-leukemic responses by stimulation of leukemic CD3+ cells with allogeneic stimulator cells

Allogeneic cells are one of the more potent stimuli of the immune system, as 100–1000 fold more cells respond to an allogeneic major histocompatibility complex (MHC) molecules than to a foreign peptide presented by self MHC molecules [1]. Alloreactivity is the driving force for deleterious responses such as graft-versus-host disease (GVHD) and organ rejection seen in transplantation, but is also linked to beneficial graft versus leukemia (GVL) responses which has led to efforts to separate GVL responses from GVHD responses [2]. Within our group a nonengraftment donor leukocyte infusion protocol, in which infusion of large numbers of haploidentical cells at a dose of 1–2 × 10⁸ CD3+ cells/kg into patients with refractory hematological malignancies, resulted in responses of varying magnitude in over half of the patients tested (14/26 overall responses with 5 complete responses) including three complete responses in patients with refractory acute myeloid leukemia (AML) [3]. These responses occurred in the absence of durable engraftment as donor cells were no longer detected within 2 weeks of infusion. This protocol was characterized by the rapid development of a cytokine release syndrome (CRS) marked by high plasma levels of proinflammatory cytokines and fevers that remitted with corticosteroid administration. Anti-leukemic responses may potentially be mediated by donor lymphocytes for a limited period of time but, given the lack of durable chimerism, we postulate that extended responses were more likely mediated by recipient lymphocytes [4]. The intense alloreactive immune response generated by this protocol could facilitate anti-leukemic responses by a variety of mechanisms, such as the reactivation of memory anti-T cells from leukemic patients [5, 6] or the cross reactivity of patient alloreactive effector cells toward patient cancer cells [1, 7]. Herein we describe the ability of leukemic patient CD3+ lymphocytes to generate cytolytic responses against the syngeneic CD3-fraction, which includes leukemia cells, in vitro following stimulation with allogeneic donor peripheral blood mononuclear cells (PBMC).

The characteristics of 21 newly diagnosed leukemic patients studied are described in Table 1. Cytogenetic and molecular classifications were based on NCCN criteria [9]. The median age was 60.5 years (range, 29–86), the median number of absolute lymphocyte count collected was 2.7 × 10⁶/mL, and the median percentage of peripheral blasts in the CD3-cells was 23%. Disease types were (AML, 14 patients), chronic myeloid leukemia (CML, 1 patient), chronic myelomonocytic leukemia (CMML, 3 patients), and acute lymphocytic leukemia (ALL, 3 patients).

CD3+ cells were either incubated with mitomycin C-treated healthy unrelated donor allogeneic PBMC (allostimulated CD3+ cells) or syngeneic CD3-cells (autologous stimulated CD3+ cells). Proliferation was detected for all combinations with similar levels of proliferation between those cells stimulated with allogeneic cells (S.I. 13.2 ± 19.3) or autologous CD3-cells (S.I. 15.3 ± 21.1). On day 7, allostimulated CD3+ cells and CD3+ cells stimulated with syngeneic CD3-cells were examined for cytolytic killing of CD3-cells. Eleven of the twenty-one patient allostimulated CD3+ cells were able to lyse CD3-cells (Fig. 1a). CD3+ cells from seven of the responding leukemic patients lysed CD3-cells when stimulated with either allogeneic stimulator. CD3+ cells from four patients were able to lyse CD3-target cells when stimulated with syngeneic CD3-cells. These CD3+ cells also proliferated strongly to the syngeneic CD3-cells (S.I. of 3.9, 20.9, 24.2, 79.6) and three out of four also generated strong alloreactive CTL when stimulated with allogeneic cells. These findings contrasted to the lack of proliferation seen when CD3+ cells obtained from normal donors were stimulated with syngeneic CD3-cells (SI = 0.9 ± 0.3, n = 4). In addition no detectable lysis of normal syngeneic CD3-cells was seen when CD3+ cells from normal donors were stimulated with allogeneic PBMNC (n = 9), syngeneic PBMNC (n = 4) or syngeneic CD3-cells (n = 4).

T CD3-cell killing was induced via allogeneic stimulation for AML and CML/CMML but not in ALL samples (Fig. 1b). Patient samples with peripheral blasts greater than 50% had an increase in the stimulation index compared to those with less than 20% peripheral blasts. Additionally, lytic effector function in greater then 50% blasts had a P value of 0.08 compared to those with less than 20%. Together this suggests that candidates with more peripheral blasts may respond better to allostimulation than those with less peripheral blasts.

Analysis of factors that were associated with the ability or inability of the alloreactive CD3+ cells to generate cytolytic activity toward syngeneic CD3-cells containing leukemic cells identified several immunophenotypes associated with the CD3+ cells which did not generate cytolytic activity toward the CD3-cells. The first was an increased number of CD4+ cells expressing high levels of perforin, granzyme A and granzyme B (Fig. 1c). A second finding was an increased number of CD8+ cells expressing CD39, a phenotype found in terminally exhausted CD8+ cells. A nonsignificant increase of CD4+CD39+cells was also observed in this set of CD3+ cells. (Figure 1d). In addition the number of CD4+ cells expressing gamma delta T cell receptor (TCRγδ) was significantly upregulated on in the unresponsive CD3+ cells (Fig. 1d). No significant differences in the number of FoxP3+ cells or the expression of inhibitory receptors was detected between the patients that lysed leukemic cells and those that did not. Staining freshly isolated cells with antibodies detecting cytolytic activity or pathways including CD107a, TRAIL and Fas Ligand did not demonstrate any predictive value for identifying which CD3+ cells would generate anti-leukemic responses (data not shown). Interestingly, all three AML patients who harbored FLT-3 internal tandem duplication (ITD) mutations, a prognostic marker in AML with poor prognosis [10], were able to lyse leukemia containing CD3-cell fractions following allogeneic stimulation.

In summary, in our screening of 21 newly diagnosed leukemia samples approximately half of the patients’ CD3+ cells were able to generate lysis of syngeneic cells containing leukemic cells after allostimulation. The ability of alloreactive effector cells to lyse syngeneic leukemic cells was similar to the findings previously reported in mice [11, 12]. It was even reported that stimulation with syngeneic tumor cells induced alloreactive effector cells [13]. These findings would suggest that cancer cells express unique peptides on their MHC molecules that allow them to be recognized by alloreactive effector cells [14]. Previously, when the peptides bound to HLA B7 molecules isolated from normal cells and leukemic cells were analyzed and compared, it was observed that there were many more unique peptides present on the HLA B7 molecules isolated from the leukemic cells [15]. Many of these unique peptides were phosphopeptides. Stimulation of PBMC from normal donors with a pool of these unique peptides resulted in T cell responses while the same pool of peptides did not induce responses in PBMC from cancer patients. This suggested that those effector cells had been inhibited or eliminated in the cancer bearing patients.

While one possibility is that alloreactive T cells could cross react with tumor cells, another possibility is that the alloreactive responses activate T cells specific for tumor associated antigens. For example, T cells reactive with Wilms tumor-1 (WT1) antigen have been identified in many individuals [16‐20]. Guo et al. [17] found increased numbers of Wilm tumor-1 (WT1) antigen reactive CD8+ T cells in 33 out of 39 patients that received infusions of HLA-mismatched donor cells in their protocol. Although the source of these antigen reactive CD8+ cells was unknown in many of these cases, in 6 of these cases the WT1 reactive CD8+ cells could be determined to be of host origin. Several studies have also indicated that vaccination with WT1 vaccines facilitated or prolonged responses [21, 22]. A study in which patients received G-CSF mobilized haploidentical donor cells after receiving 100 cGy total body irradiation, a patient who went into complete remission was found to have high levels of CD8+ cells reactive with proteinase-3, a tumor associated antigen, with no detectable donor cells [23]. These examples indicate that the expansion of T cells reactive with antigens expressed by the leukemic cells could distinguish between those individuals whose cells responded in our study and those that did not.

Phenotypic studies identified an increased number of CD4+ cells expressing cytolytic effector molecules and increased number of recipient CD8+ cells expressing CD39, a marker of exhaustion as a characteristic of the CD3+ cells that did not generate anti-leukemic responses when stimulated with allogeneic cells. The appearance of cytolytic CD4+ cells can be linked to repeated antigen exposure in chronic infection and in cancer bearing individuals reflecting a senescent immune system [24]. It has been shown that these cells can mediate cytolytic activity either via the granzyme perforin pathway or via Fas/Fas ligand signaling. In patients with chronic hepatitis B infection and hepatocellular carcinoma it was found that the CD4+ CTL also produced IL-10 which inhibited the function of CD8+ CTL [25]. Adding anti-IL-10 antibody reversed the inhibition of CD8+ cell function mediated by the CD4+ CTL. This mechanism could provide a possible explanation for the association between increased numbers of CD4+ CTL and decreased CD8+ cell activity.

In our experiments, the increased numbers of CD4+ expressing the γδTCR was associated with lack of response. Usually γδ T cells have an advantage over αβ T cells in cancer because antigen processing and presentation is not required but can instead directly recognize cancer molecules [26] although some γδ subsets may facilitate cancer growth [27]. There is a recent report of CD39+ γδ T cells that were induced by TGF-β and served as potent immunoregulatory cells in colorectal cancer [28]. Although it cannot be determined directly from this data because antibodies to CD39 and γδ TCR were included in different antibody panels, increased numbers of both CD4+ γδ T cells and CD39+CD4+ cells were detected in the CD3+ cells from the unresponsive patients. These findings would be consistent with an increased number of these CD39+ γδ T regulatory cells in this group of patients that are inhibiting other T cell responses and this possibility will be examined more closely in future studies. The CD8+CD39+ phenotype is associated with the exhausted phenotype and increased numbers of these cells suggests that the cancer reactive CD8+ cells could be exhausted and unable to respond to allostimulators [29].

In future clinical trials of nonengraftment donor leukocyte infusion protocols, it will be interesting to test if the expression of these particular phenotypes will identify those patients unresponsive to this protocol. These in vivo trials will further define which factors allow for reproducible induction of an anti-leukemic response by allostimulated leukemic patient CD3+ cells.

Patient’s T lymphocytes are frequently able to attack syngeneic hematological malignancies when stimulated with allogeneic cells. Nonresponders can be defined using the phenotypic distinctions based on the presence of CD4+ CTL and CD8+CD39+ cells. Understanding the basis for the ability of alloreactive cells to lyse syngeneic cancer cells will allow for the development of protocols which harness these responses to achieve effective anti-cancer responses in all patients.