FMS-like tyrosine kinase 3 (FLT3) is a transmembrane protein expressed on normal hematopoietic stem and progenitor cells (HSC) and retained on malignant blasts in acute myeloid leukemia (AML). We engineered CD8+ and CD4+ T-cells expressing a FLT3-specific chimeric antigen receptor (CAR) and demonstrate they confer potent reactivity against AML cell lines and primary AML blasts that express either wild-type FLT3 or FLT3 with internal tandem duplication (FLT3-ITD). We also show that treatment with the FLT3-inhibitor crenolanib leads to increased surface expression of FLT3 specifically on FLT3-ITD+ AML cells and consecutively, enhanced recognition by FLT3-CAR T-cells in vitro and in vivo. As anticipated, we found that FLT3-CAR T-cells recognize normal HSCs in vitro and in vivo, and disrupt normal hematopoiesis in colony-formation assays, suggesting that adoptive therapy with FLT3-CAR T-cells will require subsequent CAR T-cell depletion and allogeneic HSC transplantation to reconstitute the hematopoietic system. Collectively, our data establish FLT3 as a novel CAR target in AML with particular relevance in high-risk FLT3-ITD+ AML. Further, our data provide the first proof-of-concept that CAR T-cell immunotherapy and small molecule inhibition can be used synergistically, as exemplified by our data showing superior antileukemia efficacy of FLT3-CAR T-cells in combination with crenolanib.
FMS-like tyrosine kinase 3 (FLT3) is a type I transmembrane protein that plays an essential role in normal hematopoiesis and is physiologically expressed on normal hematopoietic stem cells (HSCs), as well as lymphoid, myeloid and granulocyte/macrophage progenitor cells in humans [1,2,3,4]. In mature hematopoietic cells, FLT3-expression has been reported in subsets of dendritic cells and natural killer cells [5,6,7]. FLT3 is also uniformly present on malignant blasts in acute myeloid leukemia (AML), providing a target for antibody and cellular immunotherapy [1, 4, 8,9,10,11]. The antigen density of FLT3 protein on the cell surface of AML blasts is in the range of several hundreds to several thousand molecules per cell, which is optimal for recognition by engineered T-cells that are equipped with a synthetic chimeric antigen receptor (CAR) [12, 13].
At the molecular level, FLT3 transcripts are universally detectable in AML blasts, with graded expression levels in distinct FAB (French–American–British) subtypes [9, 14]. Higher FLT3 transcript levels correlate with higher leukocyte counts and higher degrees of bone marrow infiltration by leukemic cells, independent from the presence of FLT3 mutations . FLT3 is important for survival and proliferation of AML blasts and of particular pathophysiologic relevance in AML cases that carry activating mutations in the FLT3 intracellular domain [1, 11]. Of these, internal tandem duplications (ITDs) in the juxtamembrane domain and mutations in the intracellular tyrosine kinase domain (TKD) are the most common aberrations that collectively occur in approx. 30% of AML cases [1, 11, 14, 15]. Both aberrations cause constitutive FLT3 activation in a ligand-independent manner and act as gain-of-function ‘driver mutations’ that contribute to sustaining the malignant disease [16, 17]. These attributes suggest FLT3-ITD+ AML is particularly susceptible and a preferred AML subset for anti-FLT3 immunotherapy because the risk to incur FLT3-/low antigen-loss variants is likely low. Indeed, the presence of an FLT3-ITD is associated with an inferior clinical outcome after induction/consolidation chemotherapy and allogeneic hematopoietic stem cell transplantation (HSCT), and defines a subset of high-risk AML patients that require novel, innovative treatment strategies [18, 19].
FLT3 is being pursued as a target for tyrosine kinase inhibitors (TKIs) and numerous substances are at advanced stages of clinical development. However, the clinical efficacy of single agent therapy with type-II-inhibitors, e.g., midostaurin that target the inactive kinase conformation has been rather limited, owing at least, in part, to the development of resistance through novel mutations in the FLT3 intracellular domain, or FLT3 overexpression in AML blasts [20,21,22,23]. Crenolanib is a specific type-I-inhibitor that targets the active FLT3 kinase conformation and is effective against FLT3 with ITD and TKD mutations that confer resistance to midostaurin [24, 25]. Crenolanib is also active against platelet-derived growth factor receptor alpha/beta and is being evaluated in patients with gastrointestinal stromal tumors and gliomas [26, 27]. In AML, crenolanib is effective against AML with FLT3-ITD and TKD mutations [28, 29].
FLT3 has also been pursued as a target for antibody immunotherapy, even though the antigen density of FLT3 on AML blasts is much lower compared to e.g. CD20 on lymphoma cells and not presumed to be optimal for inducing potent antibody-mediated effector functions . A mouse anti-human FLT3 monoclonal antibody (mAb) 4G8 has been shown to specifically bind to AML blasts and to a lesser extent to normal HSCs. After Fc-optimization, 4G8 conferred specific reactivity against AML blasts with high FLT3 antigen density in preclinical models . Here we engineered T-cells to express a FLT3-specific CAR with a targeting domain derived from the 4G8 mAb and analyze the antileukemia reactivity of FLT3-CAR T-cells against FLT3 wild-type (wt) and FLT3-ITD+ AML cells, alone and in combination with crenolanib. Further, we evaluate recognition of normal HSC as an anticipated side effect of effectively targeting FLT3 to instruct the clinical use of FLT3-CAR T-cells.
Materials and Methods
Peripheral blood was obtained from healthy donors and AML patients after written informed consent to participate in research protocols approved by the Institutional Review Boards of the University of Würzburg and the University of Regensburg.
A codon optimized targeting domain comprising the VH and VL segments of the FLT3-specific 4G8 mAb  was synthesized (GeneArt ThermoFisher, Regensburg, Germany) and fused to a CAR backbone comprising a short IgG4-Fc hinge spacer, a CD28 transmembrane domain, CD28 or 4-1BB costimulatory moiety, and CD3z, in-frame with a T2A element and EGFRt transduction marker (Supplementary Figure 1a) [30,31,32]. The entire transgene was encoded in a lentiviral vector epHIV7 and expressed under control of an EF1/HTLV hybrid promotor . Similarly, targeting domains specific for CD19 (clone FMC63) and CD123 (clone 32716) were used to generate CD19 and CD123-CARs with CD28 costimulatory moiety, respectively [30, 31, 33, 34].
In vivo experiments
All experiments were approved by the competent Institutional Animal Care and Use Committees. NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (female, 6–8 week old) were purchased from Charles River (Sulzfeld, Germany) or bred in-house. Mice were inoculated with 1 × 106 ffluc_GFP+ MOLM-13 AML cells by tail vein injection on day 0 and randomly allocated to treatment and control groups. On day 7, mice received a single dose of 5 × 106 T-cells (i.e., 2.5 × 106 CD4+ and 2.5 × 106 CD8+ in 200 µL of PBS/0.5% FCS) by tail vein injection. Crenolanib (Selleck Chemicals, Houston, TX) was formulated in 30% glycerol formal (SigmaAldrich, Munich, Germany) and administered intraperitoneally (i.p.) at a dose of 15 mg/kg from Monday to Friday for 3 consecutive weeks. AML progression/regression was assessed by serial bioluminescence imaging following i.p. administration of D−luciferin substrate (0.3 mg/g body weight) (Biosynth, Staad, Switzerland) using an IVIS Lumina imaging system (PerkinElmer, Waltham, Massachusetts). The data were analyzed using LivingImage software (PerkinElmer). In vivo models with primary AML and normal HSCs are described in the Supplement.
Crenolanib-treatment of MOLM-13 AML cells
MOLM-13 cells were maintained in RPMI-1640 medium, supplemented with 10% fetal calf serum, 2 mM glutamine, 100 U/mL penicillin/streptomycin, and 10 nM crenolanib. Every 7 days, MOLM-13 cells were adjusted to 1 × 106/mL in fresh medium and 1 mL of this cell suspension plated per well in 48-well plates (Costar, Corning, NJ). In some experiments, MOLM-13 cells were labeled with efluoro670 according to the manufacturer’s instructions to assess proliferation by flow cytometry.
Statistical analyses were performed using Prism software v6.07 (GraphPad, San Diego, California). Unpaired Student’s t tests were used to analyze the data obtained in in vitro and in vivo experiments. Log-rank (Mantel–Cox) testing was performed to analyze differences in survival observed in in vivo experiments. Differences with a p value < .05 were considered statistically significant.
FLT3-CAR T-cells eliminate FLT3 wt and FLT3-ITD+ AML cells
We prepared CD4+ and CD8+ FLT3-CAR-modified T-cell lines from healthy donors and AML patients (n = 6). FLT3-CAR-expressing T-cells were enriched to > 90% purity using the EGFRt selection marker prior to expansion and functional testing (Supplementary Figure 1a,b). First, we confirmed specific recognition of FLT3 surface protein by CD4+ and CD8+ FLT3-CAR T-cells using native K562 (phenotype: FLT3-) and K562 target cells that had been transduced to stably express wt FLT3 (K562/FLT3) (Supplementary Figure 2). Then, we included the AML cell lines THP-1 (FLT3 wt), MOLM-13 (FLT3-ITD+/-) and MV4;11 (FLT3-ITD+, loss of heterozygosity)  into our analyses and confirmed specific high-level cytolytic activity of CD8+ FLT3-CAR T-cells against each of the cell lines at multiple effector to target cell ratios (E:T, range 10:1–2.5:1) (Fig. 1a,b). Further, CD4+ and CD8+ FLT3-CAR T-cells produced effector cytokines including IFN-γ and IL-2, and underwent productive proliferation after stimulation with each of the AML cell lines, whereas control T-cells derived from the same donors only showed background reactivity (Fig. 1d, e; Supplementary Figure 3a, b). We found that our prototypic FLT3-CAR with CD28 costimulatory domain conferred higher levels of IL-2 production and superior proliferation in T-cells compared to a corresponding receptor with 4–1BB costimulatory domain and therefore used the FLT3-CAR/CD28 in all subsequent experiments (Supplementary Figure 4). Because the FLT3-CAR binds to an epitope in the extracellular domain of FLT3, recognition of AML cells was independent from the mutation status of the intracellular tyrosine kinase domain, but rather correlated with the antigen density of FLT3 protein on the surface of target cells, as assessed by mean fluorescence intensity (MFI) (THP-1 ~ MOLM-13 > MV4;11) (Fig. 1a).
We also confirmed potent activity of CD8+ and CD4+ FLT3-CAR T-cells against primary AML blasts (patients #1 and #2: FLT3-ITD+; patient #3: unknown) (Fig. 1a, c; Supplementary Figure 3c, d). In particular, CD8+ FLT3-CAR T-cells conferred strong cytolytic activity, leading to eradication of >80% AML blasts within as short as 4 h at an E:T ratio of 1:1 (Fig. 1a, c). Notably, the antileukemia activity of FLT3-CAR T-cells against primary AML blasts was equivalent to T-cells expressing a CAR specific for the alternative AML antigen CD123 (Fig. 1c; Supplementary Figure 3c, d).
FLT3-CAR T-cells induce durable remission of AML in a xenograft model in vivo
We performed experiments in xenograft AML models in immunodeficient NSG mice to analyze the function of FLT3-CAR T-cells in vivo. First, we inoculated mice with ffLuc_GFP-transduced MOLM-13 AML cells that rapidly expanded to systemic leukemia in peripheral blood, and heavily infiltrated bone marrow and spleen (Fig. 2a, b). Leukemia-bearing mice were treated with a single dose of 5 × 106 FLT3-CAR-modified or untransduced T-cells (CD4+:CD8+ ratio = 1:1), or received no treatment (Supplementary Figure 5a). We observed a strong antileukemia effect in all mice, where FLT3-CAR T-cells engrafted. In these mice, FLT3-CAR T-cells could be readily detected in peripheral blood, increased in number during the antileukemia response, and were present in bone marrow and spleen at the end of the experiment, confirming persistence for >3 weeks after adoptive transfer (Fig. 2b; Supplementary Figure 5b, c). Complete elimination of leukemia cells from peripheral blood occurred within 3 days after adoptive transfer (Fig. 2b), and bioluminescence imaging confirmed strong systemic antileukemia activity (Fig. 2a, c). Flow cytometric analyses in bone marrow and spleen confirmed that sustained complete remission of AML was accomplished after treatment with FLT3-CAR T-cells, which translated into superior overall survival compared with mice that received control T-cells and no treatment (p < .05) (Fig. 2d, Supplementary Figure 6a, b). In all mice that responded to FLT3-CAR T-cell therapy, we observed recurrence of MOLM-13 cells in anatomical sanctuary sites (e.g., subcutaneous tissue and peritoneum) (Fig. 2a). FLT3 was expressed at similar levels on native and recurring MOLM-13 cells. We could not detect infiltrating FLT3-CAR T-cells in specimen obtained from these sanctuary sites, even though they were present in peripheral blood and bone marrow (Supplementary Figure 5c, 7a, b).
Then, we inoculated NSG mice with primary AML blasts and confirmed development of leukemia within a 4-week engraftment period (Supplementary Figure 8a), consistent with prior work . Also, in this model, FLT3-CAR T-cells could be detected at multiple time points following adoptive transfer, increased in number during the antileukemia response, and induced complete remission from AML in all treated mice (Supplementary Figure 8b–e). We did not observe recurrence of primary AML blasts in this model. In aggregate, these data show that FLT3-CAR T-cells confer potent antileukemia activity against FLT3 wt and FLT3-ITD+ AML cell lines and primary AML cells in vitro and in vivo.
Crenolanib induces increased FLT3 surface protein expression in FLT3-ITD+ AML cells
We hypothesized that upregulation of FLT3 as a compensatory mechanism of AML blasts to counteract the effect of FLT3-inhibitors could be exploited to enhance the antileukemia efficacy of FLT3-CAR T-cells [22, 23]. We cultured native MOLM-13 AML cells (MOLM-13Native) (FLT3-ITD+/-) in the presence of the FLT3-inhibitor crenolanib (MOLM-13Creno) using a 10 nM dose, which is a clinically achievable serum level [25, 37]. We analyzed FLT3-expression on MOLM-13Creno by flow cytometry after 7 days of exposure to the drug and observed significantly higher levels of FLT3 surface protein by MFI compared to MOLM-13Native cells (n = 3 experiments, p < .05) (Fig. 3a, Supplementary Figure 9a). Interestingly, withdrawal of crenolanib led to a decrease in FLT3-expression on MOLM-13 cells to baseline levels within 2 days, but increased again upon re-exposure to the drug (Fig. 3b, Supplementary Figure 9b). After primary exposure to crenolanib, we observed a moderate cytotoxic effect and slower expansion of efluor670 labeled MOLM-13Creno cells compared to MOLM-13Native cells for approx. 7 days (Supplementary Figure 9a, c). However, despite continuous supplementation to the culture medium, the cytotoxic effect of crenolanib subsequently ceased and the expansion of MOLM-13Creno cells accelerated, suggesting they had acquired resistance to the drug.
An increase in FLT3-expression upon exposure to crenolanib was also observed with MV4;11 AML cells (FLT3-ITD+), but did not occur in several cell lines expressing wt FLT3, i.e., THP-1 AML cells, JeKo-1 mantle cell lymphoma, and K562 erythro-myeloid leukemia, suggesting upregulation of FLT3-expression in response to crenolanib treatment specifically occurred in FLT3-ITD+ AML cells (Fig. 3a, Supplementary Figure 9d). In contrast to FLT3, CD33 and CD123-expression on both MOLM-13 and MV4;11 was not affected by crenolanib and did not increase (Supplementary Figure 9e).
Higher FLT3-expression on crenolanib-treated MOLM-13 AML cells leads to enhanced antileukemia reactivity of FLT3-CAR T-cells in vitro
We sought to determine whether the higher antigen density of FLT3 on MOLM-13Creno would enhance recognition by FLT3-CAR T-cells. Our earlier data showed rapid modulation of FLT3-expression upon exposure to and withdrawal of crenolanib (Fig. 3a,b), suggesting maximum reactivity of FLT3-CAR T-cells against MOLM-13Creno would be accomplished in the presence of the drug. It is known that TKI may interfere with T-cell activation and function [38, 39], and we therefore confirmed that crenolanib per se did not affect the effector function of FLT3-CAR T-cells (Supplementary Figure 10). In subsequent functional analyses, we observed superior cytolytic activity of CD8+ FLT3-CAR T-cells against MOLM-13Creno compared to MOLM-13Native cells (p < .05), enhanced production of IFN-γ and IL-2, as well as enhanced proliferation of FLT3-CAR T-cells after stimulation with MOLM-13Creno compared to MOLM-13Native (Fig. 3c). In contrast to FLT3-CAR T-cells, the antileukemia reactivity of CD123-CAR T-cells against MOLM-13Creno and MOLM-13Native was not significantly different (Fig. 3c).
We confirmed that upregulation of FLT3 after treatment with crenolanib also occurred on primary FLT3-ITD+ AML blasts and led to increased cytolysis by FLT3-CAR T-cells (Supplementary Figure 11). Collectively, these data show that treatment with crenolanib leads to increased expression of FLT3 specifically in FLT3-ITD+ AML cells and consecutively, enhanced recognition by FLT3-CAR T-cells.
FLT3-CAR T-cells and the FLT3-inhibitor crenolanib act synergistically in mediating regression of AML in vivo
This encouraged us to examine the antileukemia effect of FLT3-CAR T-cells in combination with crenolanib in the MOLM-13/NSG xenograft model. Mice were inoculated with MOLM-13Native AML cells on day 0, and on day 7 were treated with either FLT3-CAR T-cells alone, crenolanib alone, the combination treatment with FLT3-CAR T-cells and crenolanib, or were left untreated. We observed potent antileukemia efficacy in mice receiving the combination treatment with FLT3-CAR T-cells and crenolanib (Fig. 4a). There was superior engraftment and in vivo expansion of FLT3-CAR T-cells (Fig. 4b); a higher overall response rate (combination: n = 8/8, 100% vs. FLT3-CAR T-cells alone n = 6/8, 75% vs. crenolanib alone n = 0/8, 0% vs. no treatment n = 0/8, 0%); faster and deeper remissions as assessed by bioluminescence imaging (Fig. 4a,c); as well as improved overall survival of mice receiving the FLT3-CAR T-cell and crenolanib combination, compared to monotherapy with either FLT3-CAR T-cells or crenolanib alone, or no treatment (p < .05) (Supplementary Figure 12a). Crenolanib monotherapy had only a minuscule antileukemia effect and MOLM-13 cells that we recovered from peripheral blood and bone marrow at the experiment endpoint had uniformly and strongly upregulated FLT3, which is consistent with our earlier observation in vitro (Fig. 4d). At the experiment endpoint, peripheral blood, bone marrow and spleen in mice treated with the FLT3-CAR T-cell/crenolanib combination and FLT3-CAR T-cell alone were completely free from MOLM-13 AML cells, whereas mice receiving crenolanib monotherapy and untreated mice showed a high degree of infiltration with MOLM-13 cells (Fig. 4e; Supplementary Figure 12b). Also, with the FLT3-CAR T-cell/crenolanib combination treatment, mice experienced recurrences of FLT3+ MOLM-13 cells in anatomical sanctuary sites (Supplementary Figure 12c). Collectively, our data show that FLT3-CAR T-cells and crenolanib can be used synergistically in combination therapy to confer a potent antileukemia effect against FLT3-ITD+ AML cells in vitro and in vivo.
FLT3-CAR T-cells recognize normal HSCs and compromise hematopoiesis in colony-formation assays
We sought to determine the collateral damage of FLT3-CAR T-cells on normal hematopoietic stem and progenitor cells. We confirmed expression of FLT3 on normal GM-CSF-mobilized peripheral blood CD34+CD38- HSCs and CD34+CD38+ progenitor cells, although the antigen density of FLT3 was lower compared to AML cells by MFI (Fig. 5a, Supplementary Figure 13a). We performed a flow cytometry-based cytotoxicity assay in vitro and found that FLT3-CAR T-cells lysed ~50 and 80% of normal HSCs within 4 and 24 h, respectively (E:T ratio = 5:1) (Fig. 5b). As a reference, we included T-cells expressing a CD123-specific CAR, that has been reported to completely eliminate normal HSCs and induce myeloablation . CD123-CAR T-cells exerted a faster and stronger cytolytic effect on normal HSCs compared to FLT3-CAR T-cells and lysed > 95% of HSCs within 24 h (Fig. 5b). Next, we performed in vitro colony-formation assays from residual HSCs at the end of the 24-hour co-culture with FLT3-CAR T-cells. After 14 days, we only detected a small number of erythroid colonies, whereas formation of myeloid colonies was completely abrogated (Fig. 5c). To corroborate our data, we administered normal HSCs to NSG-3GS mice and after 8-week engraftment, treated them with FLT3-CAR T-cells. We found that also in vivo, normal HSCs and progenitor cells were depleted from bone marrow after treatment with FLT3-CAR T-cells, similar to mice that received CD123-CAR T-cells (Supplementary Figure 13b–d). To assess the influence of crenolanib on normal HSCs, we analyzed FLT3 expression that, is consistent with their wt FLT3 genotype, did not increase and assessed viability, which in contrast to MOLM-13 AML cells, did not decrease within 7 days of exposure to 10 nM of the drug (Supplementary Figure 14; Supplementary Figure 9c).
Collectively, these data show that normal HSCs are recognized and eliminated by FLT3-CAR T-cells. These data suggest that the clinical use of FLT3-CAR T-cells against AML would be restricted to a defined window of time prior to allogeneic HSCT that allows subsequent CAR T-cell depletion and reconstitution of the hematopoietic system.
We are pursuing FLT3 as a novel target antigen for CAR T-cells in AML. Clinical trials with CD19-CAR-modified T-cells in acute lymphoblastic leukemia (ALL) and non-Hodgkin lymphoma have provided evidence for the curative potential of CAR T-cells in hematologic malignancies. However, they also exposed several challenges that affect safety and limit efficacy of CAR T-cell therapy, including on-target recognition and long-term elimination of non-tumor cells that express the targeted antigen [40, 41]. We detected FLT3 surface protein on normal HSCs, which is consistent with previous work that demonstrated uniform expression of FLT3 on human hematopoietic stem and progenitor cells obtained from bone marrow and cord blood of healthy donors [1, 2, 4]. Our data also confirm the previous notion that FLT3 density is lower on normal HSCs compared to AML blasts by flow cytometry (MFI) . Even though the precise antigen density on target cells that is required to induce CAR T-cell triggering is unknown, it is presumed that this threshold is in the order of (few) hundreds of molecules per cell , which is the range that has been estimated for FLT3-expression on HSCs . Indeed, our data show that FLT3-CAR T-cells eliminate the majority of HSCs within a 24 h co-culture assay, leading to qualitatively and quantitatively impaired hematopoiesis in colony-formation assays in vitro. Overall, recognition of normal HSCs by FLT3-CAR T-cells is an anticipated finding, in line with the previous demonstration that mAb 4G8—from which we derived the targeting domain of our FLT3-CAR—shows uniform binding to normal HSCs . Our data suggest that the clinical use of FLT3-CAR T-cells may be limited to a defined therapeutic window of time and restricted to a clinical context that permits subsequent reconstitution of the hematopoietic system. Such a window of opportunity is provided in the context of allogeneic HSCT with adoptive transfer of FLT3-CAR T-cells prior to HSCT to reduce leukemia burden and/or induce minimal residual disease (MRD)-negativity, followed by FLT3-CAR T-cell depletion and engraftment of donor-derived normal HSCs. This strategy requires the ability to rapidly and completely remove CAR T-cells to protect incoming normal HSCs. The FLT3-CAR T-cells employed in our study are equipped with an EGFRt depletion marker. We have recently shown in immunocompetent mice that administration of an anti-EGFR mAb can mediate CAR T-cell depletion and reversal of CAR T-cell induced systemic toxicity . An alternative strategy is the inclusion of the iCasp9 suicide gene that has been demonstrated to accomplish (near)-complete removal of CAR T-cells in the order of minutes to few hours following administration of an inducer drug . Operational models of myeloablative CAR T-cell therapy have been established  and will aid in defining the optimal timing of FLT3-CAR T-cell administration and elimination, and subsequent reconstitution with normal HSCs.
It is uncertain, whether a small proportion of FLT3low HSCs may escape elimination by FLT3-CAR T-cells, and whether this diminished pool of HSCs would be capable of replenishing a quantitatively and qualitatively normal hematopoietic system. A recent study even suggested that FLT3-CAR T-cells would not deplete HSCs and preserve HSC differentiation in vivo . However, the only experiment to substantiate this statement had been performed in immunodeficient NSG-3GS mice that received simultaneous injections of human HSCs and FLT3-CAR T-cells, and it is unclear whether this application mode actually leads to any interactions of HSC and FLT3-CAR T-cells, especially in the bone marrow. Rather, FLT3-CAR T-cells should have been administered after HSC engraftment and hematopoietic differentiation is established, which takes several weeks in this model . Indeed, our data in the NSG-3GS/HSC model show that normal HSCs are eliminated after adoptive transfer of FLT3-CAR T-cells. Several alternative CAR target antigens are being pursued in AML, including CD33 and CD123 that have advanced to clinical evaluation, but share with FLT3 the challenge of being expressed on normal HSCs [34, 46]. CD123-CAR T-cells have been shown to induce myeloablation in vivo due to recognition of normal HSCs, consistent with our data that showed rapid and complete elimination of normal HSCs by CD123-CAR T-cells in our co-culture assay in vitro and in HSC-engrafted NSG-3GS mice in vivo .
Another observation from the clinical use of CD19-CAR T-cells in ALL is relapse with leukemia cell variants that have lost CD19-expression under therapeutic pressure either through lymphoid to myeloid trans differentiation; alternative splicing, resulting in loss of the epitope targeted by the CAR; or selection of pre-existing antigen-negative leukemia cell clones [47,48,49]. These data underscore the need to select CAR target antigens that are uniformly expressed and ideally of pathophysiologic relevance for leukemia cells. Encouragingly, FLT3 is uniformly expressed in AML blasts, including leukemia stem/initiating cells (LSC/LIC), suggesting the potential to confer a curative treatment for AML with FLT3-CAR T-cell therapy [14, 17]. There is no clinical experience with FLT3 as an immune target to aid in assessing the risk that mutations in the extracellular FLT3 domain may arise during therapy that delete the epitope recognized by our FLT3-CAR. We observed disease recurrence in anatomical niches in our xenograft model with MOLM-13 AML cells. We confirmed that FLT3-expression and the epitope recognized by the FLT3-CAR had been retained on recurring MOLM-13 cells, but could not detect FLT3-CAR T-cells in sanctuary lesions. This is in line with the prior notion that human T-cells are not readily able to migrate through murine endothelial and epithelial barriers .
There is increasing clinical data from the use of small molecule FLT3-inhibitors [20,21,22] and intriguingly, an observation that has been made with the FLT3-inhibitor lestaurtinib is upregulation of FLT3-expression in AML blasts after repeated exposure to this drug . The conceptual appeal of using FLT3-inhibitors and FLT3-CAR T-cells in combination is that AML blasts that acquire resistance to FLT3-inhibitors and upregulate FLT3, expose themselves to recognition and elimination by FLT3-CAR T-cells. Indeed, our data show strong upregulation of FLT3 in FLT3-ITD+ AML cells after treatment with the FLT3-inhibitor crenolanib, and enhanced antileukemia reactivity of FLT3-CAR T-cells against crenolanib-treated FLT3-ITD+ AML cells in vitro and in vivo. We show that upregulation of FLT3 occurs specifically in FLT3-ITD+mutated AML cells, supporting the critical role of FLT3-ITD in AML pathogenesis. These data demonstrate for the first time that CAR T-cell immunotherapy and small molecule inhibition can be used synergistically in a hematologic malignancy, and provide proof-of-concept with the FLT3-CAR T-cell/crenolanib combination. Experiments that evaluate synergy between FLT3-CAR T-cells and other FLT3-inhibitors are ongoing. Upregulation of FLT3 has been previously reported in FLT3-ITD+ MOLM-13 AML cells that acquired resistance to the FLT3-inhibitor midostaurin , providing another attractive combination partner for FLT3-CAR T-cells. Clonal heterogeneity is a challenge for treating AML, and even though FLT3 is uniformly expressed on AML blasts, there is a potential risk that FLT3-/low AML relapse may develop from FLT3 wt subclones even in patients that are classified as FLT3-ITD+. We did not observe FLT3-/low AML escape variants in our in vivo model with primary AML blasts. However, it has been demonstrated that FLT3-ITD+ AML subclones show preferential engraftment in NSG mice compared to FLT3 wt subclones and therefore, this model may underestimate this risk .
It has recently been shown that FLT3 acts as a cytoprotective kinase in cardiomyocytes . It is unknown whether FLT3 surface expression on cardiomyocytes is sufficient for CAR recognition and hence, particular caution must be taken when clinically translating FLT3-CAR T-cell therapy, especially in combination with FLT3-inhibitors. We did not detect FLT3 upregulation after crenolanib treatment in AML and non-AML cells that express wt FLT3 (including HSCs), suggesting the concomitant use of crenolanib enhances selectivity of FLT3-CAR T-cells for FLT3-ITD+ AML blasts compared to HSC and non-AML cells.
Our data suggest that FLT3-ITD+ AML cases are particularly susceptible and have a high likelihood to benefit from FLT3-CAR T-cell therapy. In particular, the risk for encountering FLT3 loss under therapeutic pressure with FLT3-CAR T-cells is likely lower in FLT3-ITD+ compared to FLT3 wt cases, and may be further lowered through concomitant treatment with crenolanib and other FLT3-inhibitors. Further, our data suggest that FLT3 in FLT3-ITD+ AML cases is a preferred CAR target, advantageous to the alternative antigens CD33 and CD123, receptors for sialic acid and interleukin-3, respectively, that are of less pathophysiologic relevance in AML compared to FLT3. Clinical trials have already demonstrated that rapid antigen downregulation occurs in AML blasts after treatment with anti-CD33 and anti-CD123 mAbs [52, 53]. FLT3-expression has also been demonstrated in pre-B-ALL and T-ALL, mixed-lineage leukemia and myelodysplastic syndrome [1, 4, 8], expanding the spectrum of hematologic diseases amenable to adoptive immunotherapy with T-cells expressing our novel FLT3-CAR.
We thank Silke Frenz and Elke Spirk for their expertise in performing the mouse experiments. H.J. was supported by a grant from the German Excellence Initiative awarded to the Graduate School of Life Sciences (GSLS), University of Würzburg. I.G.G. was supported by a grant from Fundación Alfonso Martin Escudero, Spain. M.H. is a member of the Young Scholar Program (Junges Kolleg) and Extraordinary Member of the Bavarian Academy of Sciences (Bayerische Akademie der Wissenschaften). The research was supported by the Deutsche Forschungsgemeinschaft via SFB/TR 221‚ modulation of graft-versus-host and graft-versus leukemia immune responses after allogeneic stem cell transplantation.This work was also supported by German Cancer Aid (Max Eder Program Award 110313 [M.H.].Author Contributions
HJ designed and performed the experiments, analyzed the data and wrote the manuscript. IG-C, TN, ST and JR designed and performed the experiments, and analyzed the data. WH, JBM and JS analyzed the data. HB provided biologic material and analyzed the data. MH and HE designed experiments, analyzed the data, wrote the manuscript and supervised the project.