Primary CD33-targeting CAR-NK cells for the treatment of acute myeloid leukemia

Acute myeloid leukemia (AML) is a malignant disorder derived from neoplastic myeloid progenitor cells characterized by abnormal proliferation and differentiation. Although novel therapeutics have recently been introduced, AML remains a therapeutic challenge with insufficient cure rates. In the last years, immune-directed therapies such as chimeric antigen receptor (CAR)-T cells were introduced, which showed outstanding clinical activity against B-cell malignancies including acute lymphoblastic leukemia (ALL). However, the application of CAR-T cells appears to be challenging due to the enormous molecular heterogeneity of the disease and potential long-term suppression of hematopoiesis. Here we report on the generation of CD33-targeted CAR-modified natural killer (NK) cells by transduction of blood-derived primary NK cells using baboon envelope pseudotyped lentiviral vectors (BaEV-LVs). Transduced cells displayed stable CAR-expression, unimpeded proliferation, and increased cytotoxic activity against CD33-positive OCI-AML2 and primary AML cells in vitro. Furthermore, CD33-CAR-NK cells strongly reduced leukemic burden and prevented bone marrow engraftment of leukemic cells in OCI-AML2 xenograft mouse models without observable side effects.


INTRODUCTION
Acute myeloid leukemia (AML) represents a devastating disease for which only limited therapeutic progress has been made in the last decades. Besides high-dose chemotherapy, allogeneic bone marrow transplantation currently constitutes the most effective therapy, which is not applicable for all patients, often not fully eliminates tumor cells, and harbors the significant risk of inducing a graft-versus-host disease. Overall, this results in poor prognosis especially for elderly patients [1][2][3][4]. In other hematological malignancies such as B-cell acute lymphoblastic leukemia (B-ALL) novel cellular therapies with chimeric antigen receptor (CAR)modified T cells lead to striking results which culminated in the approval of multiple drugs by the U.S. Food & Drug Administration (FDA) and the European Medicines Agency (EMA) [5][6][7][8][9]. Nevertheless, treatment of AML with CAR-T cells appears to be complicated due to the lack of AML-specific antigens and the tremendous molecular heterogeneity of the disease [1,10]. While for B-ALL CD19 is well known as a highly specific antigen for leukemic blasts [11,12], in AML such an exclusive antigen still needs to be identified. As a promising target CD33 was shown to be expressed on leukemic blasts and leukemia-inducing cells in the majority (88%) of AML patients, but also on normal hematopoietic stem cells [10,13]. Furthermore, preclinical results demonstrated increased anti-AML activity of CD33-specific CAR-T cells [14,15]. However, clinical application of long-persistent CAR-T cells appears to be challenging due to potential long-term suppression of hematopoiesis. Furthermore, CAR-T cells need to be generated from autologous sources, which is cost and time intensive and often not possible from heavily pre-treated patients. As adoptively transferred natural killer (NK) cells possess a shorter lifetime, are associated with fewer side effects, and hold an intrinsic CAR-independent killing capacity against AML, CD33-CAR-NK cells constitute a promising alternative [16][17][18][19]. The use of peripheral-blood-derived NK cells (PB-NK) also appears advantageous, since NK cell lines such as NK-92 need to be irradiated prior to infusion, which can limit the proliferation and killing capacity of applied cells [17]. In addition, NK cells do not induce graft-versushost diseases. As a perspective, the allogeneic transfer of CAR-NK cells in combination with freezing of cells harbors the potential for a true "off-the-shelf" product, which could dramatically improve the availability and reduce the costs of such an immune-cell based therapy.
Here, we report on the generation of CD33-specific CAR-NK cells from peripheral blood using baboon envelope pseudotyped lentiviral vectors (BaEV-LV). CD33-CAR-NK cells showed stable CAR-expression, functional proliferation, and significantly increased killing capacity against CD33-positive AML cell lines as well as primary AML cells in vitro. Furthermore, CD33-CAR-NK cells effectively eliminated all bone marrow-and spleenengrafted, as well as the majority of peripheral AML cells in an OCI-AML2 xenograft mouse model without detectable side effects.

MATERIALS AND METHODS Primary AML cells
Frozen primary AML cells were thawed and cultivated in DMEM medium supplemented with 15% FBS, 50 µM ß-Mercaptoethanol, 1% Pen/Strep, 100 ng/mL stem cell factor (SCF), 10 ng/mL IL-3, 20 ng/mL IL-6, 10 ng/ml thrombopoietin (TPO) and 10 ng/ml FMS-like tyrosine kinase 3 ligand (FLT3L)). Cells were seeded with 700,000 cells in 3 ml per well of a six-well plate and 1 µg/ml f.c. DNAse was added. The usage of primary patient material was approved by the Ethics Commission of the University Hospital Frankfurt, Germany (approval no. 274/18). All participants gave written informed consent in accordance with the Declaration of Helsinki.

CAR construction and lentiviral vector production
Second-generation CD33-CAR incorporating the My96 scFv sequence was constructed as described earlier [22]. Briefly, the My96 scFv was combined in frame with CD8 hinge and transmembrane domain, 4-1BB/CD137 costimulatory domain, and CD3ζ activation domain. A leader peptide derived from GM-CSFRα was included to facilitate CAR cell surface expression. Third-generation self-inactivating baboon envelope-pseudotyped lentiviral vectors (BaEV-LVs) were produced by transient transfection into HEK293T cells using MACSfectin (Miltenyi Biotec) or polyethylenimine (PEI) [23].

Transduction of NK cells
On day two post NK cell isolation NK cells were transduced with lentiviral particles. Therefore, 0.5 × 10 6 cells per well were seeded in a flat bottomed 48-well plate. Subsequently, lentiviral particles and Vectofusin-1 (Miltenyi Biotec; 2.5 µg/ml final concentration per well) were mixed in identical volumes, incubated at room temperature for 7 min, and added to the cells, to reach a final cell concentration of 1 × 10 6 cells/ml. All previous steps were performed in serum-free NK-MACS ® medium supplemented with 1% NK-MACS ® Supplements, with or without 1% Pen/Strep and 80 ng/ml IL-1β (Miltenyi Biotec), 500 IU/ml IL-2 (Miltenyi Biotec or Novartis). Finally, 10 ng/ ml IL-15 (Miltenyi Biotec or Peprotech) or 50 ng/ml IL-15 (CellGenix) was added. Subsequently, the plate was centrifuged at 400 × g for 2 h at 32°C. Twenty-four hours post-transduction half of the medium was replaced by fresh medium containing 5% human plasma and a combination of IL-2 and IL-15 [24].

CD33-expression analysis
CD33 expression on AML cell lines as well as primary AML and NK cells were analyzed by flow cytometry using anti-CD33-PE (clone REA775, Miltenyi Biotec) or -BV421 (clone WM53, BD Bioscience).

CAR-NK cell functional assay
Four hours or 24 h endpoint cytotoxicity of CAR-and UTD-NK cells was analyzed by flow cytometry. Target cells either expressing GFP or stained with Cell Trace CFSE proliferation kit (Invitrogen) were cocultured with effector cells at various effector-target (E:T)-ratios for the indicated time period at 37°C and 5% CO 2 . Afterwards, cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) (AppliChem), or propidium iodide (PI) (Miltenyi Biotec), and viability of target cells was analyzed using a BD FACSCelesta (BD Biosciences) or MACSQuant Analyzer 10 (Miltenyi Biotec).
To analyze cytotoxicity over time, real-time monitoring using the IncuCyte S3 system (Sartorius, Goettingen, Germany) was performed. Assays were set up by co-culturing GFP-transgenic target cells with CAR-NK cells at an E:T-ratio of 1:1 or 1:5. Phase contrast and green fluorescence images were captured with 10× magnification every two hours for 4 days. Analysis of images was performed using the following settings for GFP: Top-hat (110 μm) and Threshold (0.5 GCU).
Cytotoxicity upon repetitive tumor challenge was analyzed by coincubating effector cells with GFP-expressing target cells at an E:T-ratio of 1:1 and rechallenging every 2 days with fresh target cells and 50% of new culture medium. Viable cell count was quantified by flow cytometry 24 h after each target cell addition. Cytokine secretion assays were conducted by co-culturing UTD-and CAR-NK cells with target cells at a ratio of 1:1 for 24 h. Harvested supernatants were then analyzed using the MACSPlex Cytokine 12 Kit for human analytes (Miltenyi Biotec) following manufacturer's protocol.

Chimerism analysis
Mouse whole blood was collected in EDTA microvette tubes after cheek vein penetration and DNA was extracted using the QIAamp Mini spin column kit (Qiagen, Hilden, Germany). DNA content of samples was determined with a NP80 nanophotometer (Implen, Munich, Germany). The proportion of human DNA was assessed by a quantitative real-time PCR approach specific for the human Albumin gen [26]. The ratio of human NK and AML-target cells was calculated by chimerism analyses of informative human STR-markers using the Powerplex16 kit marker panel (Promega, Madison, WI, USA) [27].

Bone embedding and sectioning
Bones were harvested and incubated in a series of 4% PFA, EDTA-and sucrose-solutions before they were embedded in gelatin. 7 µm-sections of bones were generated on the CryoStarTM NX70 cryostat (Thermo Fisher Scientific, Waltham, MA, USA). NK cells were stained with a Nkp46 antibody (R&D Systems, Minneapolis, MN, USA). For staining of CD33-CAR-NK cells, bone sections were incubated with CD33-biotin over night at 4°C followed by the addition of secondary streptavidin AF647 antibody (BD Biosciences). For nuclear counterstaining DAPI (Sigma-Aldrich, St. Louis, MO, USA) was used.

Statistical analysis
For statistical analysis, a normal distribution for NK cell functionality was assumed due to the deployment of healthy donors as NK cell source. Concerning the mouse in vivo-experiments also a normal distribution can be assumed. Accordingly, data were analyzed by two-tailed, unpaired Student's t-test or Mann-Whitney-test and defined as significant when p < 0.05. Statistical analysis was performed using GraphPad PRISM version 6-8 (GraphPad Software, Inc., San Diego, CA, USA) for experiments with three or more donors/animals.

CD33-CAR-NK cells demonstrate potent in vitro activity against AML
Aiming to enhance the anti-leukemic activity of NK cells towards AML, primary NK cells derived from peripheral blood (PB) were genetically modified by lentiviral transduction to express a second-generation CD33-CAR using a single-chain variable fragment (scFv) derived from the clinically tested antibody-conjugate AVE9633 (Fig. 1A) [28]. Transgene integration by BaEV-LVs resulted in transduction rates between 35 and 60% and no severe impairment of expansion was observed during ex vivo cultivation, although CD33 has been reported to be upregulated on activated lymphocytes (Fig. 1A, B) [29]. For functional in vitro analysis, OCI-AML2 was selected as target-AML cell line, which showed relative insensitivity to natural cytotoxicity of NK cells and expressed CD33 (Fig. 1C, D). CD33-CAR-NK cells were more potent in eliminating OCI-AML2 and CD33-positive primary AML cells in a short-term killing assay compared to UTD-NK cells (Fig. 1E, F; Supplementary  Fig. 1A), while killing of the CD33-dim KG1a cell line was not improved (Supplementary Fig. 1B). When the cytolytic activity was evaluated longitudinally by dynamic monitoring using IncuCyte, UTD-NK cells exhibited AML control at best, while CD33-CAR-NK cells efficiently eradicated OCI-AML2 cells even at the unfavorable E:T-ratio of 1:5 (Fig. 1G). Importantly, the sustained serial killing capacity of CD33-CAR-NK cells was also observed when the immune cells were re-challenged with AML cells supplied with fresh medium, indicating that the tumor cell death was not mediated by secondary mechanisms (for example increased levels of IFN-γ or TNF-α) but by sustained CAR-directed cytotoxic activity (Fig. 1H). Cytokine secretion and phenotypical analysis revealed no major differences between UTD and CD33-CAR-NK cells, again pointing towards a CAR-dependent killing mechanism (Supplementary Fig. 1C, D).
A single dose of CD33-CAR-NK cells shows effective clearance of leukemic cells in the majority of OCI-AML2-engrafted NSG-SGM3 mice With robust in vitro data in hand, the anti-leukemic effect of a single dose of CD33-CAR-NK cells was further analyzed in a humanized OCI-AML2 xenograft NOD.Cg-Prkdc scid Il2rg tm1Wjl Tg (CMV-IL3,CSF2,KITLG)1Eav/MloySzJ (NSG-SGM3) mouse model ( Fig. 2A). This model was chosen, as NSG-SGM3 mice constitutively produce IL-3, GM-CSF, and stem cell factor (SCF), which reflects physiological conditions of a human host and promotes the engraftment of AML cells [25,30]. NSG-SGM3  Fig.  2A-D). Analysis of peripheral blood from CD33-CAR-NK treated mice revealed significant increase of human INF-γ serum levels on day 1 after the first NK cell-application compared to mice that had received UTD-NK cells, while serum levels of MIP-1α, IL-6, IL-10, and TNF-α remained below physiological relevant concentrations ( Fig. 2C; Supplementary Fig. 2E). BLI analysis of femur, tibia, and spleen of representative mice on day 7, 14, and 21 revealed impeded AML engraftment in CD33-CAR-NK treated animals which was confirmed by flow cytometry analysis of cells isolated from BM and spleen (Fig. 2D). Furthermore, CD33-CAR-NK cells showed increased BM-and spleen-infiltration compared to UTD-NK cells. Thereof, the majority of NK cells could be identified as CAR-positive (BM: 89-95%; spleen: 78-90%; Fig. 2E, F). Consequently, BM-located CD33-CAR-NK cells prevented engraftment of leukemic cells and the successive alteration of the BM, which became evident not only by flow cytometry but also through histologic analysis of BM via confocal microscopy. Thereby, BM of untreated mice or mice which had received UTD-NK cells displayed an altered, pitted structure together with high GFP-positive AML cell infiltration, while a normal BM structure and no GFP signal could be observed for mice treated with CD33-CAR-NK cells (Fig. 2G).
Enhanced clearance of AML in OCI-AML2-engrafted NSG-SGM3 mice by repetitive injections of CD33-CAR-NK cells With the aim of improving outcomes, following CAR-NKproduct characterization ( Supplementary Fig. 3A-C), in total three doses of 1 × 10 7 CD33-CAR-NK cells were administered to a group of seven OCI-AML2-engrafted NSG-SGM3 mice (female, 9-13 weeks old) (Fig. 3A). This treatment resulted in strong reduction of AML burden by day 21 in all animals (Fig. 3B). BLI analysis of femurs, tibiae, and spleens on day 22 revealed impeded AML engraftment in CD33-CAR-NK-treated mice while the administration of UTD-NK cells did not have any effect. This was confirmed by flow cytometry analysis of cells isolated from BM and spleen. Here only a reduction of BM-and no changes for spleen-infiltrating AML cells could by observed in mice treated with UTD-NK cells, whereas CD33-CAR-NK cells were able to completely eradicate AML cells in both organs (Fig. 3C). Additionally, NK cell-infiltration in the BM or spleen on day 22 was significantly increased in mice treated with CD33-CAR-NK compared to UTD-NK cells, and the majority of NK cells were identified as CAR-positive (78-95%; Fig. 3F). Chimerism analysis of PB revealed high amounts of DNA from human NK cells (10-30%) without detectable DNA of human AML cells in mice, which were treated with CD33-CAR-NK cells. In contrast, only small amounts of human DNA were found in the blood samples of mice treated with UTD-NK cells which partially comprised AML cell-DNA (Fig. 3D). Auxiliary, significant increases in human INF-γ and MIP-1α serum levels on day 1 after the first NK cellapplication could be observed in mice, which had received CD33-CAR-NK compared to UTD-NK cells, while serum levels of IL-6, IL-10, and TNF-α remained below physiological relevant concentrations ( Fig. 3E; Supplementary Fig. 3E). Histologic analysis of BM by confocal microscopy of mice treated with CD33-CAR-NK cells revealed the presence of CAR-NK cells in an intact, AML-free BM, while UTD-NK-treated or untreated mice displayed an altered, pitted BM structure together with high GFP-positive AML cell infiltration ( Fig. 3G; Supplementary Fig.  3D). Of note, following both single and repetitive CD33-CAR-NK cell injections, the animals did not show any clinical signs of cytokine release syndrome (CRS), weight reduction, altered appearance or behavior, or graft-versus-host-disease (GvHD), in line with histological analysis of lung, liver, and colon.

DISCUSSION
In the rapidly moving field of CAR-immune cell therapies, recent studies have shown that primary CAR-NK cells are safe and can be highly effective [31]. With the aim of developing NK cell therapy concepts that address malignant diseases beyond CD19-expressing B cell neoplasia, which are still difficult to treat, research on primary CAR-NK cells targeting AML is urgently required. Our study showed that CD33-CAR-NK cells exhibit stable CAR expression and maintained a high proliferation rate, although fratricide of CD33-positive NK cells was observed in different donors ( Supplementary Fig. 1E, F). Importantly, CD33-CAR-NK cells efficiently eliminated AML cells in vitro and eradicated all BM-and spleen-infiltrating as well as the majority of peripheral AML cells in NSG-SGM3 mice without observable side effects. Significantly higher numbers of BM-, spleen-as well as PB-NK cells were detected in mice treated with CD33-CAR-NK cells. Additionally, while the CAR-expression rates of the applied NK cells reached between 30 and 50%, the Fig. 3 Repetitive administration of CD33-CAR-NK cells displays improved anti-tumor efficacy in OCI-AML2 engrafted NSG-SGM3 mice. A Scheme of the in vivo evaluation of a repetitive treatment with CD33-CAR-NK cells (1 × 10 7 intravenously) combined with subcutaneous IL-2 treatment in OCI-AML2 (Luc + ) xenograft NSG-SGM3 mice. B Total flux analysis, as well as representative BLI images of differently treated OCI-AML2 (Luc+), engrafted NSG-SGM3 mice over time (n = 7 per group). Mice received a total of three weekly doses of 1 × 10 7 NK cells. Mice that were treated with CD33-CAR-NK cells show severely reduced leukemic burden compared to untreated mice (UT) or mice which received untransduced (UTD)-NK cells. C Total flux analysis of femurs/tibiae and spleens, as well as flow cytometry analysis of isolated cells from BMs or spleens at day 22 post AML-injection, revealed the absence of GFP-positive leukemic cells in CD33-CAR-NK treated mice as well as increased NK cell infiltration (n = 6-7 per group). Values of zero were set to 1 for total flux analysis. Median ± range. D Chimerism analysis d22 post AMLinjection revealed high amounts of DNA from human NK cells without detectable DNA of AML in blood of mice that were treated with CD33-CAR-NK cells (n = 6-7 per group). Mean ± SD. E Serum analysis of blood day 3 before AML injection and day 1 post first NK cell application showed significantly increased pro-inflammatory human cytokines for mice that received CD33-CAR-NK cells (n = 6-7 per group). F Flow cytometry-based CAR-expression analysis of BM-or spleen-infiltrating NK cells in CD33-CAR-NK treated mice revealed the presence of mainly CAR-positive cells (n = 6-7 per group). Mean ± SD. G Confocal microscopy imaging demonstrated GFP-positive leukemia cells in BM of UTD-NK treated NSG-SGM3 mice while absent in mice that received CD33-CAR-NK cells. Additionally, CAR-NK cells could be detected in the BM of CD33-CAR-NK treated mice. One representative image from a total of four are shown. Statistical analysis was performed by Mann-Whitneytest (for total flux analysis) or Student's t test (for the rest) (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). majority (≥80%) of BM-and spleen-infiltrating NK cells could be identified as CAR-positive. This indicates increased survival or improved BM homing of CAR-NK cells. Latter constitutes a particularly desirable feature as the density of BM located NK cells following adoptive NK cell-transfer was shown to correlate with clinical response [32]. The increased persistence of CD33-CAR-NK cells compared to UTD-NK cells without endogenous cytokine-production might also constitute a clinical advantage, as recent studies showed that CAR-NK cells constitutively secreting IL-15 can cause severe toxicities in animal models [33]. Additionally, increased but still moderate values of proinflammatory cytokines in mice treated with CD33-CAR-NK cells could fortify a systemic immune response against AML, which might further improve outcomes. While donor-dependent heterogeneity in terms of CAR-expression and proliferation constitutes a known hurdle in vitro, we could also observe strong donor-dependent differences in anti-leukemic effects of UTD-NK cells in vivo, which was mostly abrogated by introducing a CAR-dependent killing mechanism. Nevertheless, donordependent characteristics of allogeneic applied NK cells need to be carefully monitored and require further investigations to maximize outcomes and avoid unwanted side effects. While in recent studies with CD33-CAR-T cells, certain constructs such as Gemtuzumab-based CARs were shown to induce in vivo toxicities, the AVE9633-derived CAR deployed in this study might pose a promising option for AML therapy [34]. Overall, these data clearly suggest that CD33-CAR-NK cells might be suitable for the treatment of AML, due to their strong antitumor efficacy and highly improved presence in BM, spleen, and PB compared to normal NK cells. In particular, for cases in which the generation of autologous CAR-immune cells from heavily pre-treated AML patients appears to be difficult, CAR-NK products from third-party donors may be considered a clinically important advantage.