Allogeneic human double negative T cells as a novel immunotherapy for acute myeloid leukemia and its underlying mechanisms.

Abstract

Purpose:

To explore the potential of ex vivo expanded healthy donor derived allogeneic CD4 and CD8 double negative cells (DNTs) as a novel cellular immunotherapy for leukemia patients.

Experimental Design:

Clinical grade DNTs from peripheral blood of healthy donors were expanded and their anti-leukemic activity and safety were examined using flow-cytometry based in vitro killing assays and xenograft models against AML patient blasts and healthy donor-derived hematopoietic cells. Mechanism of action was investigated using antibody-mediated blocking assays and recombinant protein treatment assays.

Results:

Expanded DNTs from healthy donors target a majority(36/46) of primary AML cells including 9 chemotherapy-resistant patient samples in vitro and significantly reduce the leukemia load in patient-derived xenograft models in a DNT-donor unrestricted manner. Importantly, allogeneic DNTs do not attack normal hematopoietic cells or affect hematopoietic stem/progenitor cell engraftment and differentiation, nor cause xenogeneic graft-versus-host disease in recipients. Mechanistically, DNTs express high levels of NKG2D and DNAM-1 that bind to cognate ligands preferentially expressed on AML cells. Upon recognition of AML cells, DNTs rapidly release IFNγ which further increases NKG2D and DNAM-1 ligands expression on AML cells. IFNγ pretreatment enhances the susceptibility of AML cells to DNT-mediated cytotoxicity, including primary AML samples that are otherwise resistant to DNTs, and the effect of IFNγ treatment is abrogated by NKG2D and DNAM-1 blocking antibodies.

Conclusion:

This study supports healthy donor-derived allogeneic DNTs as a therapy to treat patients with chemotherapy-resistant AML and also reveals interrelated roles of NKG2D, DNAM-1, and IFNγ in selective targeting of AML by DNTs.

Introduction

Acute myeloid leukemia (AML) is the most common form of adult acute leukemia with 5-year survival rates of ~5% and ~35% for elderly and younger patients, respectively.^1–4 Despite decades of using chemotherapy to treat AML patients, a high relapse rate and refractoriness to chemotherapy are major challenges to patient survival.^1–3 Allogeneic hematopoietic stem cell tran splantation (allo-HSCT) is a potential curative cellular therapy for chemotherapy-resistant leuke mia^5–7, but limited donor availability, the risk of graft-versus-host-disease (GvHD) and other transplantation-related toxicities restrict its wide application in elderly and debilitated patients.^4,7–10

Recently, chimeric antigen receptor (CAR)-T cell therapies have shown excellent clinical responses against lymphocytic leukemia^11–13 and promising results in pre-clinical AML studies.^14–17 However, in clinical trials, CAR-T cells specific for CD33 and LeY antigen caused severe toxicity in some of the treated AML patients.^18,19 Also, genetic modification of infused cells increases the treatment cost and potential risk to patients.¹⁷ Multiple clinical trials using allogeneic NK cell therapy to treat AML patients demonstrated the safety and clinical benefits, but a noticeable number of patients either did not achieve complete remission or experienced disease relapse.^20–25 Some studies highlighted the importance of precise HLA-KIR matching for improved efficacy of NK therapy^24,25, but this can limit treatment availability to patients. Furthermore, long-term efficacy of NK cell therapy has yet to be demonstrated. Therefore, the need for safer and more effective cellular therapies with a broader and easier clinical applicability to target chemotherapy-resistant AML and prevent disease relapse remains.

In this study, we explored a non-conventional T cell subset named double negative T cells (DNTs) as a novel therapy to target AML, including leukemic cells that are resistant to the conventional chemotherapy. DNTs are mature peripheral T lymphocytes expressing CD3-T cell receptor (TCR) complex but not CD4, CD8, nor invariant NK T cell markers. They represent 1–3% of peripheral blood mononuclear cells (PBMCs).²⁶ The function of human DNTs was largely unknown due to their low frequency in vivo and a lack of an effective method to expand them to sufficient numbers for in vivo studies. We have shown previously that DNTs expanded from AML patients were cytotoxic to autologous AML cells in vitro.²⁶ However, attempts to expand autologous DNTs for treating AML patients have failed perhaps due to the defect of DNTs in patients. Whether allogeneic DNTs expanded from healthy individuals can effectively target AML cells in vitro and in vivo while sparing non-malignant cells and tissues of recipients, and the mechanisms involved, have not been explored previously.

Here we demonstrate, for the first time, that therapeutic quality and quantity of DNTs can be expanded ex vivo from healthy donors (HDs), and that these cells can selectively target a large array of primary AML samples including those from chemotherapy-resistant patients without observed toxicity towards normal cells and tissues. Furthermore, we identified a positive feedback loop of NKG2D, DNAM-1, and IFNγ, which helps to explain the ability of DNTs to selectively recognize and target AML but not normal cells. Collectively, our findings open a new window of cellular therapy using DNTs expanded from healthy volunteers as a potential off-the-shelf product to treat patients with high-risk AML, and perhaps other cancers.

Materials and Methods

DNTs and leukemic cell lines

DNTs were expanded ex vivo as previously described²⁶. Briefly, DNTs enriched by depleting CD4⁺ and CD8⁺ cells from PBMCs using CD4- and CD8- depletion cocktail (Stemcell Tech) were cultured in anti-CD3 antibody coated plates (OKT3; 5 μg/ml) for 3 days in RPMI-1640 supplemented with 10% FBS and 250 IU/ml of IL2 (Proleukin, Novartis Pharmaceuticals, Canada); soluble anti-CD3 (0.1 ug/ml) was added on day 7, 10, and 14. On days 3, 7 and 10, fresh media and IL-2 were added. DNTs were harvested 10–20 days post expansion for subsequent experiments. The leukemic cell lines AML3/OCI (AML3), KG1a, and MV4–11 were obtained from ATCC.

Flow cytometry based in vitro killing assay

DNTs stained with PKH-26 (Sigma) were co-cultured with target cells for 2–4 hours, cells were then stained with anti-human CD3 (HIT3a), CD33 (WM53), CD45 (HI30) FITC-Annexin V and 7AAD (all from BioLegend), and analyzed using flow cytometry. Specific killing was calculated by: $\frac{{\text{\%}AnnexinV}_{withDNT}^{+}-\text{\%}{AnnexinV}_{withoutDNT}^{+}}{100-\text{\%}{AnnexinV}_{withoutDNT}^{+}}\times 100$. For blocking assays, DNTs were incubated with neutralizing antibodies for 0.5–1 hour prior to co-incubation with target cells at 4 to 1 effector to target ratio for 2 hours or 2 to 1 effector to target ratio for 4 hours. % Inhibition of killing was calculated by $\frac{{\text{\%}Specifickilling}_{withoutAb}-{\text{\%}Specifickilling}_{withAb}}{{\text{\%}Specifickilling}_{withoutAb}}\times 100\frac{\left({\text{\%SpecificKilling}}_{\text{withoutAb}}-\text{\%}{\text{SpecificKilling}}_{\text{withAb}}\right)}{{\text{\%SpecificKilling}}_{\text{withoutAb}}}\times \text{d}$. For IFNγ pretreatment assays, DNTs or AML cells were pre-treated with 50ng/ml of recombinant human IFNγ (BioLegend) for 1 hour or overnight. % increase in killing was calculated by $\frac{{\text{\%}Specifickilling}_{IFN\gamma treated}-{\text{\%}Specifickilling}_{IFN\gamma untreated}}{{\text{\%}Specifickilling}_{IFN\gamma untreated}}\times 100$.

Xenograft models

${\text{NOD.Cg-}Prkdc}^{scid}{Il2rg}^{tm\mathit{1}Wjl}/\text{SzJ}$ (NSG) mice (Jackson Laboratories, Bar Harbor, ME) were maintained at UHN animal facility. 8–12 week old females were irradiated (250 cGy) 24 hours prior to intrafemoral or tail vein injection of the 2–5×10⁶ primary AML blasts. 2×10⁷ DNTs were injected intravenously at the indicated time points. rIL2 (Proleukin) was administered (10⁴ IU/mouse) intraperitoneally concordant with the DNT injections on days 1, 2, 4, 7 and weekly thereafter until euthanized. 2–4 weeks after the last DNT injection, mice were sacrificed and spleen and BM cells were harvested and the frequency of AML was analyzed by flow cytometry. For GvHD-related experiments, 5–20×10⁷ DNT, PBS, or 5×10⁶ PBMCs were intravenously injected into irradiated naïve NSG mice and % body weight change was calculated as $\frac{{bodyweight}_{day0}{-bodyweight}_{dayx}}{{bodyweight}_{day0}}\times 100$. Mice with weight loss greater than 25% were euthanized via cervical dislocation. 14 days post injection, liver, lung, and small intestine were obtained, fixed in 10% Formalin for 24–48hrs, and then in 70% ethanol until samples were sent for hematoxylin and eosin histological analysis. Histological staining and picture acquisition was done by the Advanced Optical Microscopy Facility at UHN. For safety-related experiments, irradiated NSG mice were injected with 3×10⁵ CD34⁺CD133⁺ HSPCs before treated with DNTs.

Statistical Analysis

All graphs and statistical analysis were generated using GraphPad Prism 5. Student’s t test and linear regression test were used. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001 indicate significance between experimental and control values. Error bars represent ± SEM.

Human samples and study approval

Human blood, BM, and CD34+ cells were collected from healthy adult donors and AML patients after obtaining written informed consent and used according to University Health Network (UHN) Research Ethics Board (05–0221-T) and NHLBI approved protocols. PBMCs from HD or AML patients were separated by Ficoll (GE Healthcare) density gradient. AML patient samples were viably frozen in the Princess Margaret Leukemia Bank and stored in liquid nitrogen until used. Animal studies were approved by the institutional Animal Care Committee of the UHN (Permit Number: 741.22) and carried out in accordance with the Canadian Council on Animal Care Guidelines.

Results

Anti-leukemic function of healthy donor-derived DNTs

To explore the potential of using allogeneic human DNTs to treat leukemia, we developed a protocol allowing for a large scale GMP expansion of DNTs from HDs. So far we have expanded DNTs from more than 60 HDs of which 6 expansions were done under GMP conditions. On average, 1.41±0.51×10⁸ DNTs were obtained from each milliliter of peripheral blood after 17–20 days ex vivo expansion with >90% purity (Figure 1A). The ex vivo expanded DNTs expressed CD3 and αβ- or δγ- TCR but not CD4, CD8, or iNKT cell markers (Figure 1B) and exhibited a central/effector memory T cell phenotype (Supplementary Figure S1). Since αβ-TCR⁺ and δγ-TCR⁺ DNT subsets display comparable anti-leukemic function²⁶, for the simplicity of clinical applications, no further selection for specific TCRs was performed for the subsequent studies when more than 90% of cells in expanded products were DNTs.

Figure 1. Anti-leukemic activity of allogeneic DNTs against primary AML blasts.

A) Expansion profile of DNTs expanded under GMP conditions from peripheral blood of healthy donor over 17 days. Result shown is representative of three separate experiments. B) PBMCs (top panels) and DNTs (bottom panels) expanded as described in Methods were stained with immune cell subset markers: CD3, CD4, CD8 for T cells, and α-Galactosylceramide loaded-CD1d for invariant NKT cells. Result shown is representative of 11 separate expansions for DNTs and 3 separate experiments for PBMC and NKT cells. C) Susceptibility of primary AML blasts obtained from 46 patients to DNT-mediated cytotoxicity was determined using a 2-hour flow cytometry-based killing assay at 4:1 DNT-to-AML ratio. Effector DNTs were labeled with PKH-26 and AML blasts were defined as PKH-26^-CD3^-CD45^low/CD33⁺/CD34⁺ population. The level of target cell death was determined by Annexin V and 7AAD staining. Percent specific killing was determined according to the calculation in the Methods section. Dashed lines indicate 10% specific killing, where AML samples with % specific killing lower than 10% were considered as non-sensitive targets. Each number represents an ID for each patient. DNTs expanded from 11 different HDs were used. D) Specific killing mediated by DNTs from 2 HDs or NK92 against K562, AML3/OCI, and 6 primary AML samples was determined as described above. Experiments were done in triplicate using six different HD DNTs in three separate experiments. E) Schematic diagram of PDX model used. Sublethally irradiated (225cGy) NSG mice were injected with 2.5–5×10⁶ primary AML blasts followed by single or triple infusion of 2×10⁷ DNTs or PBS between days 3–14 of tumor inoculation. Bone marrow or spleen harvested on day 24–50 post blast injection and AML engraftment analyzed by flow cytometry. F) Representative dot plot of human AML cells in mouse BM. Numbers represent the frequency of AML cells. G) Four AML patient samples (090392, 090240, 5786, and 090543) engrafted NSG mice were treated with a single infusion of DNTs expanded from three different donors (HD1, HD2 and HD3). AML sample IDs and HD identification numbers are indicated at the bottom. Each symbol represents % reduction in AML load in each of DNT-treated mouse (n=41) compared to the average AML load in PBS treated group (n=20). Horizontal bars represent the mean ± SEM. H) NSG mice engrafted with primary blasts were injected with PBS (n=5), a single injection of DNTs on day 3 (n=8) or three doses of DNTs on day 3, 7 and 11 post blast injection (n=9). AML engraftment in BM was analyzed on day 32.

Expanded DNTs effectively killed 36 out of 46 primary AML patient samples (patient information shown in Supplementary Table S1) (Figure 1C) in a dose-dependent manner (Supplementary Figure S2), albeit with heterogeneity in the level of cytotoxicity to each patient sample. Higher effector-to-target ratio (Supplementary Figure S2A) or longer incubation time (Supplementary Figure S2B) showed increased killing. The threshold of 10% was used to distinguish susceptible and resistant AML samples, as samples with specific killing greater than 10% showed a dose-dependent cytotoxicity, whereas those with lower than 10% killing did not (Supplementary Figure S2C). No correlation was observed between the susceptibility to DNTs and the patient’s age at diagnosis, WBC count, or MRC cytogenetic risk groups (Supplementary Figure S3A–S3C). However, significantly lower cytolysis was observed in samples obtained from patients with lower percentages of AML cells in the bone marrow (BM), in male patients, and patients with AML secondary to prior myelodysplastic syndrome (MDS) (Supplementary Figure S3D–S3F). On the other hand, samples obtained from M5 FAB classified AML patients showed a higher level of susceptibility to DNT-mediated cytotoxicity than the rest of the patient samples tested (Supplementary Figure S3G).

To determine the relative potency of DNTs in comparison to other cytotoxic cells, DNTs from two HDs and NK92, an NK cell line used in a clinical trial for AML patient treatment (NCT00900809)²⁷, were used as effectors against eight leukemia targets. While K562, a natural NK cell target, was effectively killed at a comparable level by both DNTs and NK92, all six primary AML blasts and an AML3/OCI cell line were more susceptible to DNT- than NK92-mediated cytotoxicity (Figure 1D). Notably, DNTs effectively killed OCI-AML-3 and three primary AML samples (130783, 130794 and 090239) that were resistant to NK92 (Figure 1D). Similarly, DNTs were more effective at lysing leukemic targets than ex vivo expanded primary CD8⁺ T cells (Supplementary Figure S4A) and primary NK cells (Supplementary Figure S4B).

The ability to generate large numbers of human DNTs allowed us, for the first time, to study their anti-leukemic activity in vivo. Using a patient-derived xenograft (PDX) model (Figure 1E), we found that a single allogeneic DNT infusion significantly reduced the frequency of human CD45⁺CD33⁺ leukemic cells in the BM of NOD scid gamma (${\text{NOD.Cg-}Prkdc}^{scid}{Il2rg}^{tm\mathit{1}Wjl}/\text{SzJ}$, NSG) mice inoculated with primary AML samples obtained from 4 different patients (Figure 1F and 1G). Two additional DNT treatments further reduced leukemia load (Figure 1H). Overall, 69 mice inoculated with primary AML blasts from 8 different patients, of which 42 received one dose of DNTs and 27 received three doses, had a 52.43±30.31% and 74.35±23.94% reduction in leukemia load, respectively, compared to the PBS-treated group. As inoculation of the patient AML samples that we have tested so far did not cause death in recipient mice even when engraftment reaches >80%, so we were not able to study the effect of DNTs on recipient survival. Collectively, these results demonstrate that allogeneic DNTs can effectively target a broad range of primary AML cells both in vitro and in vivo in a dose-dependent manner.

Allogeneic DNTs can effectively target chemotherapy-resistant AML in vitro and in vivo.

As chemotherapy-resistance is the major cause of low survival rates in AML patients, we studied the effect of DNTs on chemotherapy-resistant AML cells. We found that 69.2% (9/13) of chemotherapy-resistant AML cells, obtained from chemotherapy refractory or relapsing patients, were effectively killed by DNTs in vitro (Figure 2A), and that their level of sensitivity to DNT-mediated cytotoxicity was comparable to those obtained from chemotherapy-susceptible patients (Figure 2B). Consistent with this, significant reductions in the leukemia load were observed in mice inoculated with chemotherapy-resistant primary AML cells from 4 different patients after DNT cell treatment (Figure 2C). Together, these results indicate that allogeneic DNTs are effective at targeting the majority of chemotherapy-resistant AML cells in vitro and in PDX models, supporting the use of these cells to eliminate AML cells that are not cleared by standard chemotherapy.

Figure 2. DNTs can effectively target chemotherapy-resistant primary AML blasts.

A) Cytotoxicity of DNTs expanded from HDs against primary chemotherapy-resistant AML cells obtained from 13 patients was determined using the in vitro flow cytometry-based killing assay as described in Figure 1A. B) The level of in vitro susceptibility of chemotherapy-susceptible (n=20) and -resistant (n=13) primary AML samples to DNT-mediated cytotoxicity was compared. AML samples from chemotherapy-susceptible and -resistant patients show a similar level of average sensitivity to anti-leukemic activity mediated by DNTs. (C) NSG mice engrafted with primary AML cells from three chemotherapy-resistant patients were treated with three injections of DNTs cells (n=6) or PBS (n=7). AML cells in the BM were detected 23 days post blast injection. Result shown is representative of five separate experiments done with different AML patient samples. In total, 71 mice were engrafted with primary AML samples, and of those, 27 were treated with PBS and 44 were treated with DNTs. Each dot represents result from one mouse and horizontal bars represent the mean values, and the error bars represent SEM of each group. *, p<0.05; **, p<0.01; ***, p<0.001;****, p<0.0001, using unpaired, two-tailed Student’s t test.

Infusion of DNTs does not cause GvHD nor kill normal allogeneic PBMCs and CD34⁺ hematopoietic stem/progenitors cells.

Allogeneic HSCT induces curative graft-versus-leukemia effects¹⁰, but is associated with morbidity and mortality due to donor-derived immune cells attacking normal host cells and tissue^8,9. To determine the potential toxicity of allogeneic DNTs toward normal cells, the susceptibility of normal PBMCs and CD34⁺ hematopoietic stem/progenitor cells (HSPCs) to allogeneic DNT-mediated cytotoxicity was compared to that of primary AML patient samples and AML cell lines with similar maturation status (CD33⁺CD34^- for PBMCs and CD33^-CD34⁺ for HSPCs). DNTs displayed potent cytotoxicity against primary AML samples and AML cell lines but had virtually no cytotoxicity towards normal allogeneic PBMCs (Figure 3A), peripheral blood CD33⁺ myeloid cells (Supplementary Figure S5) or HSPCs (Figure 3B).

Figure 3. Allogeneic DNTs do not kill normal cells in vitro or in vivo.

(A and B) Cytotoxicity of allogeneic DNTs expanded from 3 HDs against CD33⁺CD34^- AML: AML3-OCI, 2 primary AML patient blasts (110164 and 090596), and normal allogeneic PBMCs (PBMC-1 and PBMC-2) from 2 HDs (A) or 3 CD33^-CD34⁺ AMLs (130723, 090240, and 130624) and HSPCs (HSPC-1 and HSPC-2) from 2 HDs (B) was determined in vitro as described in Figure 1B. Experiments were done in triplicates in three separate experiments for PBMC and two separate experiments for HSPCs using DNTs from three HDs. (C and D) NSG mice were intravenously injected with PBS, 2×10⁷ ex vivo expanded DNTs, or 5×10⁶ human PBMCs obtained from 4 different HDs (n=5 per group). C) Mouse body weight was measured on days indicated, and % weight loss was calculated as described in Methods. D) On day 14, liver, lung, and small intestine were examined histologically via hematoxylin and eosin staining (20x magnification for liver and lung, 10x magnification for small intestine; n=3). Data shown are representative of four independent experiments done using DNTs and PBMCs from 4 different HDs. (E and F) CD133⁺CD34⁺ human HSPCs were infused into NSG mice (3×10⁵ cells/mouse, n=13). Six to eight weeks post HSPC injection, mice were treated with ex vivo expanded allogeneic DNTs (n=7) or PBS (n=6). The percentage of human leukocytes E) and its subsets F) in BM, spleen, and peripheral blood was determined 8 weeks after DNT treatment. Each dot represents % chimerism in one mouse and horizontal bars represent the mean ± SEM of each group. The graphs shown are a representative of 3 independent experiments done with HSPCs from 2 HDs and allogeneic-DNTs expanded from 4 HDs. *, p<0.05; **, p<0.01; ***, p<0.001;****, p<0.0001, using unpaired, two-tailed Student’s t test.

To study whether DNTs have toxicities against normal tissues in vivo, DNTs or bulk human PBMCs were intravenously infused into NSG mice and monitored for associated morbidities. As expected from prior literature, infusion of human PBMCs caused severe xenogeneic acute GvHD²⁸ as evidenced by weight loss (Figure 3C). Histologic analysis of PBMC-transfused mice revealed acute GvHD pathology in multiple organs, including portal inflammation in the liver, mononuclear infiltrates extending into the alveolar septa, endotheliitis in the lungs, and lamina propria expansion with some architectural distortion in the small intestine (Figure 3D). However, despite tissue infiltration by DNTs, neither weight loss (Figure 3C) nor tissue damage was observed in the liver, lungs, and intestines (Figure 3D) even when equal or four-fold higher numbers of DNTs compared to PBMCs were infused.

To further assess potential detrimental effects of allogeneic DNTs on normal HSPC engraftment and differentiation, NSG mice were engrafted with normal CD34⁺ HSPCs and subsequently treated with DNTs from two HDs that were allogeneic to the HSPC donors. Similar to reports by others^29,30, we consistently observed a high level of chimerism from the HSPC donors within the spleen and BM (~70–80%), and ~15% in the peripheral blood of engrafted mice. Also, no differences in the frequency or differentiation of hematopoietic cells derived from transplanted HSPC cells between DNT-treated and PBS-treated mice were observed (Figure 3E and 3F), including the CD34⁺ HSPC population in BM. These findings show that DNTs do not target allogeneic HSPCs and their progeny, nor interfere with differentiation of HSPCs into hematopoietic lineages. Together, these results demonstrate that ex vivo expanded allogeneic DNTs have potent anti-leukemic effects but are non-cytotoxic to normal tissues and hematopoietic cells in xenograft models.

NKG2D and DNAM-1 contribute to DNT-mediated donor-unrestricted and leukemia-specific cytotoxic activity.

Along with no associated toxicity, being able to use allogeneic cellular therapy without donor restriction will significantly increase its clinical applications. We found that DNTs from a single donor targeted blasts from different AML patients (Figure 4A) and DNTs obtained from different donors showed comparable cytotoxicity to the same AML target (Figure 4B), suggesting that the efficacy of DNT therapy does not depend on the donor origin and that the susceptibility of AML cells to DNTs is mainly determined by intrinsic characteristics of leukemic cells.

Figure 4. DNTs kill AML cells in NKG2D- DNAM-1-dependent, but TCR-independent manner.

A) Killing assay performed using DNTs expanded from 2 different HDs against 6 primary AML samples (shown in different symbols), demonstrating that DNTs from a single HD can target an array of AML samples. Experiments were done in triplicates. Result shown is representative of two separate experiments. B) DNTs expanded from different HDs show similar levels of cytotoxicity against the same AML blasts. Killing assays were done by using DNTs expanded from 3 HDs as effectors against 4 primary AML samples (100857, 090239, 110164, and 090517). Experiments were done in triplicates and a summary of pooled results from three separate experiments is shown. C) DNTs were pre-incubated with IgG2a isotype control or anti-TCRαβ and TCRγδ antibodies (10μg/ml for each antibody) for 30 min before co-culture with Jurkat, AML3/OCI cells or primary AML cells (140012, 080009, and 110164) at 4-to-1 DNT-to-target ratio and % specific killing was determined. Experiments were done in triplicates and the graph represents the results of 4 independent experiments. D) Ex vivo expanded DNTs were stained with DNAM-1 and NKG2D antibodies. Filled histograms represent FMO controls. The graphs shown are representative of DNTs expanded from 3 different HDs. E) Primary AML patient blasts (solid line) and normal PBMCs from HDs (dotted line) were stained for NKG2D ligands ULBP-1, ULBP-2/5/6, ULBP-3, ULBP-4, and MIC-A/B, and DNAM-1 ligands CD155 and CD112. Filled histograms represent FMO controls. Numbers shown are % of cells that expressed corresponding ligands by AML blasts (top) or normal PBMCs (bottom). F) DNTs were pre-incubated with IgG1 isotype control, anti-NKG2D, DNAM-1 or NKG2D + DNAM-1 blocking antibodies for 1 hour before co-culture with primary AML blasts (090239 and 110164) or AML3-OCI. % inhibition of killing was determined as described in Methods section. Experiments were done in triplicates, and representative data from 4 separate experiments are shown. **, p<0.01; ***, p<0.001;****, p<0.0001, using unpaired, two-tailed Student’s t test.

To dissect the mechanisms by which DNTs selectively recognize AML cells over normal cells in a DNT-donor unrestricted manner, we first studied the involvement of TCR. We found that the addition of αβ- and γδ-TCR blocking antibodies significantly reduced the level of cytotoxicity towards Jurkat cells, a T cell lymphoma cell line, but did not affect DNT-mediated cytotoxicity against AML cells (Figure 4C), supporting a TCR-independent recognition of AML cells by DNTs.

Next, we focused on innate receptor-ligand molecules involved in anti-cancer immunity. High expression of activating receptors NKG2D and DNAM-1 was observed on DNTs (Figure 4D) and significantly higher expression of NKG2D ligands (ULBP1 and ULBP3) and DNAM-1 ligands (CD112 and CD155) were found on primary AML cells over normal PBMCs (Figure 4E). Blocking NKG2D, DNAM-1, or both significantly reduced the ability of DNTs to kill primary AML cells and AML3/OCI cell line (Figure 4F). In contrast, blocking other activating receptors, NKp30, NKp44, and NKp46, used by other immune cells to recognize cancer cells^31,32 showed no effect on DNT-AML interactions (Supplementary Figure S6). Collectively, these results demonstrate that DNTs can preferentially recognize and kill AML cells but not normal cells in a donor-independent manner, partially, through NKG2D and DNAM-1.

DNTs produce IFNγ upon encountering AML cells, which augments their cytotoxicity toward AML cells but not to normal PBMCs.

Ex vivo expanded DNTs express a high level of intracellular IFNγ (Figure 5A), but minimal IFNγ levels were detected in the supernatant from co-cultures of allogeneic DNTs with normal PBMCs (0.50±0.054ng/ml) and DNT-resistant primary AML cells (0.28±0.10 ng/ml). Interestingly, significantly higher levels of IFNγ (3.29±0.58 ng/ml; Figure 5B) were released when DNTs were co-cultured with DNT-susceptible AML cells, which corresponded with the degree of cytotoxicity (Supplementary Figure S7). The level of IFNγ release was significantly reduced in the presence of NKG2D and DNAM-1 blocking antibodies, further supporting the involvement of these molecules in the recognition of AML cells by DNTs (Supplementary Figure S8). The addition of an IFNγ-neutralizing antibody significantly reduced AML cell death induced by DNTs (Figure 5C), whereas addition of exogenous recombinant IFNγ (rIFNγ) resulted in a higher level of DNT-mediated cytotoxicity without direct toxicity toward AML cells (Figure 5D–i and 5D–ii).

Figure 5. DNTs release IFNγ upon recognizing susceptible AML cells, and IFNγ sensitizes the AML targets to DNT-mediated cytotoxicity.

A) Intracellular staining of IFNγ in activated DNTs. The filled graph is the FMO control. The result shown is a representative of two separate experiments done with 4 HD DNTs. (B) Ex vivo expanded DNTs were co-incubated with allogeneic PBMCs, AML3/OCI, DNT-resistant (n=5), or DNT-susceptible (n=6) primary AML samples at 4:1 effector to target ratio, and the level of IFNγ in the culture supernatants was determined by ELISA. The number represents the average amount ± SEM of IFNγ produced from each co-culture. The data are representative of 3 independent experiments each with triplicates. C) Ex vivo expanded DNTs were pre-treated with 10μg/ml anti-IFNγ antibody or isotype control antibody for 30 mins before co-incubation with AML3/OCI or primary blasts. The results represent 3 independent experiments each with triplicates. (D-F) AML3/OCI, DNTs (D), primary AML samples, or allogeneic HD PBMCs (E and F) were pretreated or untreated with recombinant IFNγ (rIFNγ, 50ng/ml) for 1 hour and co-cultured with DNTs and % specific killing was determined. The graphs shown are representative of 3, 4, 3, and 6 separate experiments for PBMC, AML3/OCI, DNT, and primary AML samples, respectively, and each experiments were done in triplicates. F) Percent increase in DNT-mediated killing was determined as described in Methods section. Number above represents the average % increase in killing ± SEM for DNT-resistant (n=6) or -susceptible (n=14) primary AML samples upon rIFNγ pretreatment. The result is the summary of 6 separate experiments, each with triplicates. n.s. not significant. *, p<0.05; **, p<0.01; ***, p<0.001;****, p<0.0001, using unpaired, two-tailed Student’s t test or linear regression test.

Pre-treatment of DNTs with rIFNγ did not significantly affect their function (Figure 5D–v), but incubation of AML cells with rIFNγ rendered them more susceptible to DNT-induced cytotoxicity (18.4% vs. 31.9% for untreated versus rIFNγ pre-treated; Figure 5D–iii and 5D–iv), demonstrating that IFNγ sensitizes the AML cells rather than augmenting the DNTs’ cytotoxic activity. Consistent with this, the level of DNT-mediated cytotoxicity significantly increased in 10 out of 20 primary AML samples after rIFNγ pretreatment, including 4 out of 6 otherwise DNT-resistant AML samples (Figure 5E) and DNT-resistant AML cells were sensitized to a greater degree (Figure 5F). Importantly, rIFNγ pretreatment did not affect the susceptibility of normal allogeneic PBMCs to DNTs (Figure 5E and 5F).

IFNγ upregulates NKG2D and DNAM-1 ligand expression on AML cells

To understand the mechanism of how IFNγ sensitizes AML targets (Figure 5D–5F), we first tested the effect of IFNγ on NKG2D and DNAM-1 ligand expression given that these pathways contribute to DNT-mediated anti-AML activity. While it has been reported that IFNγ downregulates expression of NKG2D ligands on solid tumours^33,34, rIFNγ pretreatment upregulated the expression of NKG2D ligands ULBP1, ULBP2/5/6, ULBP3 and MICA/B, as well as DNAM-1 ligands CD112 and CD155 on AML cells (Figure 6A). In contrast, IFNγ treatment did not affect the expression levels of NKG2D and DNAM-1 ligands on normal PBMCs (Supplementary Figure S9), which is in agreement with IFNγ treatment lacking its effect on the susceptibility of PBMCs to DNTs (Figure 5F and 5G). Furthermore, the effect of rIFNγ on AML susceptibility to DNTs, as shown in Figure 5F, was neutralized by the blocking of NKG2D and DNAM-1 (Figure 6B), confirming that IFNγ can exert its role through NKG2D and DNAM-1 pathways. Further, the level of DNT-mediated cytotoxicity was inhibited by NKG2D and DNAM-1 antibodies at a significantly greater level for IFNγ-pretreated AML targets than untreated ones (22.69±1.86% vs. 13.65±0.68%; Figure 6C). These data indicate that IFNγ increases the sensitivity of AML cells, but not normal PBMCs, to DNT-mediated cytotoxicity in part by upregulating NKG2D and DNAM-1 ligand expression on leukemic cells. It also revealed, for the first time, uniquely interconnected roles of NKG2D, DNAM-1, and IFNγ in AML cells, which form a positive feedback loop to facilitate DNT cell recognition and elimination of AML cells.

Figure 6. IFNγ increases NKG2D and DNAM-1 ligand expression on AML cells and enhances their susceptibility to DNT-induced cytolysis.

(A) AML3/OCI cells were incubated with (solid lines) or without (dotted lines) 50 ng/ml rIFNγ overnight and their expression of NKG2D and DNAM-1 ligands is shown. Filled histograms represent FMO controls. Graphs are representative of 4 separate experiments done with 3 AML cell lines AML3/OCI, KG1a, and MV4–11, each experiment with triplicates. (B and C) AML3/OCI were pretreated or untreated with rIFNγ (50 ng/ml) then co-cultured with DNTs in the presence of 10μg/ml anti-NKG2D and DNAM-1 blocking antibodies or isotype control antibody. Percent specific killing of targets from each treatment are shown (B). Percent inhibition of DNT-mediated cytotoxicity by anti-NKG2D and DNAM-1 antibodies in a killing assay conducted against IFNγ-pretreated and untreated targets was calculated as described in Methods section (C). Results represent 4 separate experiments each with triplicates. ***, p<0.001;****, p<0.0001, using unpaired, two-tailed Student’s t test.

Discussion

Previously, we showed that DNTs could be expanded from peripheral blood of a small number of AML patients during chemotherapy-induced complete remission and were able to kill autologous AML cells in vitro ²⁶. Due to the low yield and purity of DNTs obtained from AML patients, it was not possible to study their function in vivo nor use them therapeutically. Herein, we developed a simple, cost-effective, highly reproducible method allowing for ex vivo expansion of DNTs from healthy individuals under GMP condition to therapeutic numbers with a high purity, in less than 20 days without feeder cells or genetic modifications. We demonstrated that the expanded DNTs selectively targeted a wide spectrum of primary AML samples both in vitro and in PDX models in a donor-unrestricted manner. Broad, yet cancer-specific cytotoxic activity of DNTs was also seen against cell lines derived from other forms of leukemia and lymphoma (Supplementary Figure S10), such as Burkitt’s B cell lymphoma (Daudi), T-cell lymphoma (Jurkat), histiocytic lymphoma (U937), chronic myeloid leukemia (K562), and myeloma (H929, 8226, and LDI), suggesting that our findings from this study may be translatable to other types of leukemia and lymphoma.

Importantly, while DNTs were cytotoxic to CD34⁺ AML cells in vitro and inhibited their engraftment in PDX models, they did not affect normal stem cell engraftment and differentiation. Contrary to the infusion of human CD4/CD8 T cells or PBMCs, infusion of human DNTs did not cause GvHD in recipients, demonstrating that DNTs selectively target leukemic cells while sparing normal cells and tissues. Moreover, we found that cryopreserved DNTs maintained comparable viability (Supplementary Figure S11A) and anti-leukemia activity (Supplementary Figure S11B) to those cultured without cryopreservation and thawing procedures. Given that ex vivo expanded allogeneic DNTs from HDs have no observed toxicity, can target broad range of leukemia cells in a donor-unrestricted manner and are cryopreservable, these cells can potentially be used as a new “off-the-shelf” cellular therapy for treating leukemia. Also, given their superior expansion profile, potent cytotoxic function, and lack of allo-response by endogenous TCRs, DNTs would be a good vehicle for CAR or transgenic TCR technology.

Mechanistically, we found that DNTs preferentially recognize and target AML cells in a TCR-independent and HLA-unrestricted manner that is partially mediated through the innate receptors, NKG2D and DNAM-1. Natural cytotoxic receptors, such as NKp30, NKp44, and NKp46, and other activating receptors such as NKG2D and DNAM-1 are expressed by NK cells and subsets of activated T cells; a role for these molecules in cancer immunity has been shown previously.^31,35–38 The expression at least one of NKG2D³⁹ and DNAM-1³⁵ ligands has been detected in the majority of AML patients in large cohort studies. We found that blocking of NKG2D and DNAM-1, but not natural cytotoxic receptors, reduced the level of AML cell apoptosis induced by DNTs, indicating the role of these molecules in AML-DNT interaction. While others have shown the role of NKG2D and DNAM-1 pathways in targeting leukemic cells, we are the first to show their unique positive feedback interaction with IFNγ. Notably, blocking these pathways did not completely abrogate the killing activity of DNTs, thus there are likely other molecules involved in the selective recognition and targeting of AMLs by DNTs, which are being currently explored.

IFNγ is a well-known inflammatory cytokine with a pleiotropic function that can elicit both pro- and anti-tumorigenic effects.^33,40–43 A previous clinical trial using IFNγ as a monotherapy failed to achieve a clinical response in AML patients,⁴⁴ which is in agreement of our finding that IFNγ alone does not induce AML cell death directly. However, we found that IFNγ renders AML blasts more susceptible to DNTs. Six primary AML blast samples obtained from 10 patients (60%) that were initially resistant to DNT-mediated cytotoxicity became susceptible after pretreatment with exogenous IFNγ. In combination with IFNγ, >85% (40/46) of tested AML samples were effectively killed by DNTs. Unlike solid tumors^33,34, IFNγ induced higher expression of NKG2D and DNAM-1 ligands on AML cells, which rendered them more sensitive to DNTs and theoretically to other cytotoxic cells such as NK cells. Notably, normal PBMCs express low levels of NKG2D and DNAM-1 ligands, which were not upregulated by IFNγ, and rIFNγ treatment did not render normal cells sensitive to DNTs. These findings help to explain why DNTs can selectively target AML but not normal cells. Furthermore, our data suggest that rIFNγ and DNT combination therapy may result in a synergistic effect leading to a greater efficacy against AML via sensitizing leukemic cells to DNT-mediated cytotoxicity and other modes of anti-leukemic activity of IFNγ^{41–43,45,46}.

Resistance to chemotherapy results in refractory AML and disease relapse, both of which significantly hamper clinical outcomes in patients. Alternative forms of treatment to target these cells are in urgent need. Here we demonstrate that a majority of primary AML cells obtained from chemotherapy-resistant and relapsing patients are susceptible to DNT-mediated cytotoxicity both in vitro and in vivo. Expression of NKG2D and DNAM-1 ligands is regulated by the DNA-damage-repair pathway^47,48, explaining higher levels of ligand expression on transformed cells^31,37,38,49. The majority of chemotherapy drugs cause DNA damage and interrupt the cell cycle, hence, treatment of myeloma with doxorubicin and bortezomib has been shown to increase expression of NKG2D and DNAM-1 ligands^48,50. These findings support the potential application of DNT therapy after conventional chemotherapy, which may yield synergistic effects via NKG2D and DNAM-1 pathways to target chemotherapy-resistant residual disease to prevent disease relapse.

We show in PDX models that 1 or 3 DNT infusions significantly reduced leukemia load but complete eradication of AML cells may be needed to prevent relapse. A higher dose or more DNT treatments or DNT in combination with other forms of therapies maybe needed to eradicate the disease. In this study, DNTs were infused 3–14 days post AML infusion when the leukemia load is relatively low. Whether DNTs given at later time points where recipients have a higher leukemia load are effective needs to be determined.

In summary, we demonstrate that allogeneic human DNTs have potent anti-leukemic activity against primary AML cells, including chemotherapy-resistant cells, both in vitro and in vivo in PDX models, without observed toxicity to normal cells and tissues and elucidate the underlying mechanism of the selective toxicity of DNTs toward AML. Therapeutically, our findings support the use of DNTs expanded from HDs as a new “off-the-shelf” non-toxic cellular immunotherapy to target chemotherapy-resistant AML populations following conventional chemotherapy to improve patient survival.

Statement of translational relevance:

Chemotherapy resistance represents a significant barrier to acute myeloid leukemia (AML) therapy. Different forms of cellular therapy have been developed to overcome this barrier, but low efficacy and associated toxicity have hampered their wide use in clinic. Here, we demonstrate that double negative T cells (DNTs) from healthy individuals can be expanded ex vivo to therapeutic levels under GMP conditions and be cryopreserved. Expanded human DNTs target a large array of primary AML cells including chemotherapy-resistant patient samples in vitro and significantly reduce the leukemia load in patient-derived xenograft models in a DNT-donor unrestricted manner. Importantly, allogeneic DNTs do not attack normal human cells nor cause xenogeneic graft-versus-host disease. Collectively, healthy donor-derived allogeneic DNTs provide a potential new off-the-shelf cellular therapy that is safe and effective to treat patients with chemotherapy-resistant AML. A first-in-human phase I clinical trial (NCT03027102) using allogeneic DNTs to treat patients with high-risk AML has been initiated.