Dual Inhibition of FLT3 and AXL by Gilteritinib Overcomes Hematopoietic Niche-Driven Resistance Mechanisms in FLT3-ITD Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is a myeloid malignancy carrying a heterogeneous molecular panel of mutations. It participates in a blockade of differentiation and an increased proliferation of myeloid hematopoietic progenitor cells. The Fms-like tyrosine kinase 3 (FLT3) is a class-III receptor tyrosine kinase expressed in progenitor cells. Its ligand binds and activates FLT3, inducing a strong activation of PI3K/AKT and MAPK pathways (1). Twenty-five percent of patients with AML display internal tandem duplication (ITD) in the juxta-membrane domain of FLT3, leading to constitutive activation of its downstream signaling to which is added the activation of the transcription factor STAT5 (2). This mutation, which carries an adverse prognosis (3), has become an important therapeutic target (4), and its frequent presence at relapse suggests that FLT3-ITD AML-initiating cells are key targets for long-lasting remission.

FLT3 tyrosine kinase inhibitors (FLT3-TKI) have been developed as ATP-competitive inhibitors. First-generation FLT3-TKI have a slight single-agent activity but showed clinical efficacy in randomized placebo-controlled trials in association with an intensive chemotherapy (5, 6). Second-generation FLT3-TKI, such as quizartinib and gilteritinib, are more specific for FLT3 and have recently shown significant activity as single agents in relapsed/refractory (R/R) AML, with composite complete remissions ranging from 45% to 55%, associated with prolongation of overall survival (OS) compared with other standards of care (7, 8). Quizartinib is a FLT3-TKI that has been specifically optimized pharmacodynamically and pharmacokinetically to inhibit the FLT3-ITD mutant protein. This drug induces clearance of peripheral blasts through induction of apoptosis, whereas cell-cycle arrest and terminal myeloid differentiation have been shown in bone marrow (BM) blasts (9). Gilteritinib is a dual FLT3 and AXL inhibitor with in vitro activity against both FLT3-ITD and tyrosine kinase domain (TKD) mutations. Gilteritinib allows sustained FLT3 inhibition; response rates are close to those observed with quizartinib, but complete remissions are much more frequent. Gilteritinib induces two distinct BM responses in FLT3-mutated AML, with and without differentiation of leukemic cells (10).

The other RTK inhibited by gilteritinib, AXL, is a RTK that is activated by homodimerization upon binding of its ligand growth arrest–specific 6 (GAS6). The GAS6/AXL pathway contributes to cell growth, survival, chemotaxis, apoptotic body clearance, and immunity in physiologic conditions (11). Moreover, AXL is overexpressed in a broad range of cancers and is regularly associated with a poor prognosis. In AML, AXL and GAS6 levels of expression have been related to poor outcomes where paracrine AXL activation has been shown to induce AML resistance to various therapies (12, 13). Finally, we previously reported that several microenvironment-related mechanisms, such as hypoxia and STAT5-activating cytokines, converge to enhance AXL expression and activation, thereby sustaining prosurvival signals that selectively protect FLT3-ITD AML cells in the hematopoietic niche (14).

Here, we were interested in testing the efficacy of gilteritinib, as a dual FLT3 and AXL inhibitor, in AML cells within their specific microenvironment. We show that gilteritinib retains efficacy in vitro in FLT3-ITD AML despite bone marrow stromal cell (BMSC) coculture and hypoxia compared with quizartinib. Moreover, ex vivo primary AML blasts display the same profile of response and in various AML mouse models, we show that gilteritinib is particularly relevant in FLT3-ITD AML with high allelic ratio. Finally, we also studied the toxicity profile of gilteritinib and quizartinib on normal murine hematopoiesis in 2 mice strains.

MV4–11, MOLM-13, and MOLM-14 are 3 AML cell lines. MV4–11 has a 48, XY, t(4;11)(q21;q23), +8, +19 karyotype with a 30-bp FLT3-ITD mutation. MOLM-13 has the 51(48–52)<2n>XY, +8, +8, +8, +13, del(8)(p1?p2?), ins(11;9)(q23;p22p23) karyotype, carrying KMT2A-MLLT3 mutations and a 21-bp FLT3-ITD mutation. Finally, MOLM-14 has the -49(46–50)<2n>XY, +6, +8, +13, der(2)t(1;2)(q31;q35), ins(11;9)(q23;p22p23), del(14(q23q32.3), del(16)(q11.2q13.1) karotype, including also KMT2A rearrangement and a 21-bp FLT3-ITD mutation. All AML and BMSC lines (MV4–11, MOLM-13, MOLM-14, MS5) were cultured in an α–minimum essential medium (αMEM), supplemented with 10% FCS, 2 mmol/L l-glutamine, 50 U/mL penicillin, and 50 μg/mL streptomycin. Hypoxia at 1% was induced by incubating cells in a specific oxygen (O₂) chamber (BioSpherix). Stromal cells were implanted for 3 days to reach 80% confluence, then culture medium was removed, and leukemic cells were added to the flask. Coculture cells were incubated either in a specific O₂ chamber for 1% O₂ hypoxic culture or in an incubator at 5% CO₂ for normoxic culture. Cells were cultured with vehicle or gilteritinib (ASP2215; kindly provided by Astellas Pharma) or quizartinib (AC220; LC Laboratories), both dissolved in DMSO.

Cell proliferation assays were performed using MTS tetrazolium (Cell Titer96 Aqueous; Promega). Cells (10⁵) were plated and incubated in quadruplicate into microtiter-plate wells in 100 μL of culture medium with various doses of gilteritinib. For the analyses, plates were incubated for 3 hours with 20 μL of MTS and read in a microplate auto reader (Imark Biorad) at 490 nM. Results were expressed as the mean optical density of the 4-well set for each gilteritinib dose. All experiments were repeated at least 3 times during the following 4 days. The 50% IC₅₀ were determined by linear regression analysis.

Protein lysates were prepared with cell lysis buffer 10× (Cell Signaling Technology) and protein concentrations were determined by bicinchoninic acid (BCA; protein assay; Pierce). Equal amounts of proteins were resolved on 12% polyacrylamide gel and transferred to a nitrocellulose or polyvinylidene difluoride (PVDF) membrane (TurboBlot; Biorad). After membrane saturation, the blots were incubated with the respective antibody followed by an antirabbit or antimouse peroxidase-conjugated secondary IgG antibody (Supplementary Table S1) and revealed with an enhanced chemiluminescence-detection method (Western Lightning Chemiluminescence Reagent Plus; Perkin Elmer). Quantification of protein expression was performed by densitometry using Image Lab software (Biorad). Phosphorylation of protein expression was normalized using total protein expression by calculating the ratio of the phosphorylated protein level to their total protein level counterparts. The ratio of the control condition was normalized to 1.

Cell apoptosis was assessed using an APC-conjugated Annexin V–labeling detection kit coupled to flow cytometry and BD FACSDIVA^TM software (BD Bioscience). For normal hematopoiesis analyses, BM cells were obtained by crushing tibias, femurs, and hips with a pestle and mortar. Filtered single-cell suspensions were incubated in Fc block and then stained with antibodies. For NOD/SCID gamma (NSG) analyses, following incubation with Fc block, unfractionated BM cell suspensions were stained with lineage markers containing biotin-conjugated anti-CD4, anti-CD5, anti-CD8a, anti-CD11b, anti-B220, anti-Gr-1, and anti-Ter119 antibodies together with APC-conjugated anti-c-Kit. When applicable, to distinguish hCD45 AML xenotransplant cells from mCD45-recipient mice cells, FITC-conjugated anti-hCD45 and APC-Cy7–conjugated anti-mCD45 antibodies were included in the antibody cocktail described above. For C57BL/6J analysis, following incubation with Fc block, unfractionated BM cell suspensions were stained with lineage markers containing biotin-conjugated anti-CD4, anti-CD5, anti-CD8a, anti-CD11b, anti-B220, anti-Gr-1, and anti-Ter119 antibodies together with APC-conjugated anti-c-Kit; BV421-conjugated anti-Sca-1; AF700-conjugated anti-CD48; PE-Cy7–conjugated anti-CD150; FITC-conjugated anti-CD34; and PE-conjugated anti-CD135. Antibody clones are detailed Supplementary Table S2. Flow cytometry analyses were performed using a LSR Fortessa (BD Bioscience). Analysis was done using FlowJo v.X (BD Bioscience).

NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ immunodeficient mice (NSG) were bred at the University of Bordeaux's animal facility for experiments validated by the French ministry (authorization no. 00048.2). Female animals were included in protocols at 8 weeks old and were monitored weekly for body weight. For subcutaneous xenografts, prior to implantation, cells were resuspended in 25% Matrigel (BD Biosciences) and 75% culture medium without FCS. Female NSG mice received injections into their hind flanks with 1.75 × 10⁶ cells. Tumor volume was determined 3 times per week with a caliper and calculated by the formula (l²*L)/2. When tumors reached 400 to 500 mm³, treatment started with quizartinib (5 mg/kg/day body weight) or gilteritinib (5, 10, and 30 mg/kg/day body weight) by daily oral gavage. For MV4–11 hematopoietic engraftment, NSG mice were conditioned with intraperitoneal injections of busulfan (Pierre Fabre) at 20 mg/kg/day for 2 days, and then injected intravenously with 10⁶ MV4–11 luciferase-transduced cells (10⁶ cells/100 μL) at day 0. At day 7, engraftment was analyzed by whole bioluminescent imaging (BLI) before treating mice with vehicle or quizartinib (5 mg/kg/day body weight) or gilteritinib (30 mg/kg/day body weight) by daily oral gavage for 7 days. At day 14 and day 20, mice were injected intraperitoneally with firefly luciferase substrate D-luciferin (150 mg/kg) before imaging using a photon bioimager and M3Vision software (Biospace Lab).

Patients at the University Hospital of Bordeaux gave written informed consent for the use of biological samples for research, in accordance with the Declaration of Helsinki. AML cells were from a 57-year-old woman diagnosed with a de novo AML 4 in French–American–British classification (FAB) with normal karyotype FLT3-ITD with an ITD/wild-type (WT) ratio at 0.82 at diagnosis and a NPM1 mutation. Flow cytometry profile was CD34⁻, HLA-DR⁻, CD56⁻, CD13⁺, CD33⁺, CD117⁺, MPO⁺, CD7⁻, and CD19⁻. Patient-derived xenografts (PDX) were generated by retro-orbital injections of 10⁶ viable primary AML cells in 8-week-old female NSG mice, previously conditioned over 2 days by intraperitoneal injections of busulfan (Pierre Fabre; 20 mg/kg/injection/day). BM engraftment of primary human AML cells in each mouse was followed monthly by intrafemoral aspiration and flow cytometry analyses. When levels reached at least 10% human AML cells (CD33⁺CD45⁺; see Supplementary Table S2 for the antibodies used), the mice were sacrificed and BM cells were collected, pooled, and cryopreserved (Cryostor CS10; Stem Cell Technologies). The FLT3 status of human AML cells (hereafter AML PDX cells) in BM was assessed as described previously (15). Two batches of AML PDX cells were generated. The first batch was obtained from 2 mice sacrificed 6 weeks posttransplantation with an ITD length at 45 bp and an FLT3-ITD/WT ratio of 0.44 and FLT3-ITD/FLT3 total ratio of 0.30, which demonstrated the presence of FLT3-ITD and WT AML clones. The second batch was obtained from 5 mice sacrificed 12 weeks posttransplantation. All human cells in this second batch carried just a FLT3-ITD of 45 bp, which demonstrated a LOH.

AML PDX cells were injected by retro-orbital route in nonconditioned 8-week-old female NSG mice (10⁶ viable PDX cells/mouse). After AML-PDX engraftment (> 1% of human cells in BM), 4 to 5 mice/group were treated with vehicle (10% DMSO in water) or quizartinib (5 mg/kg/day body weight) or gilteritinib (30 mg/kg/day body weight) by daily oral gavage for 2 weeks. Mice were sacrificed 3 days after the last gavage and BM collected from the whole rear limb (hip, femur, and tibia). AML PDX cells and murine normal hematopoiesis were analyzed by flow cytometry as described previously (16).

For both NSG and C57Bl/J models, 5 mice/group were treated with vehicle (10% DMSO in water) or quizartinib (5 mg/kg/day body weight) or gilteritinib (30 mg/kg/day body weight) by daily oral gavage for 2 weeks. Animals were either sacrificed for BM analysis 3 days posttreatment or kept alive to assess complete blood count over time for a month. Complete blood count was measured using a Scil Vet ABC Plus (Horiba). Colony-forming cell (CFC) assays were performed using MethoCultTM M3434 (STEMCELL Technologies). Two replicates were used per group in each experiment. Colonies were tallied at day 12.

Live AML PDX cells (10⁶ DAPI-negative cells) of each mouse recipient were collected and transplanted for secondary xenografts in nonconditioned 8-week-old female NSG mice. At 9 to 10 weeks posttransplantation, the mice were sacrificed and AML PDX cells and murine normal hematopoiesis were analyzed by flow cytometry as described previously (16).

All analyses were performed using GraphPad Prism software. Unless otherwise indicated, data are presented as the mean ± SEM. Statistical significance was calculated using the Student t test unless otherwise stated. Differences with P values < 0.05 were considered statistically significant with *, P < 0.05; **, P < 0.01; ***, P < 0.001; NS, nonsignificant.

We first validated the inhibitory effect of increasing doses of gilteritinib on 3 human FLT3-ITD AML cell lines also expressing AXL. The drug inhibited proliferation of the 3 cell lines in a dose-dependent manner with IC₅₀ values at 3.3 ± 0.6 nmol/L for MV4–11, 19.0 ± 3.2 nmol/L for MOLM-13, and 25.0 ± 1.0 nmol/L for MOLM-14 (Fig. 1A). In addition, gilteritinib induced a proapoptotic effect, as shown by an increased frequency of annexin V⁺ cells: 38.9 ± 0.8% for MV4–11, 54.9 ± 2.0% for MOLM-13, and 90.2 ± 0.8% for MOLM-14 at 90 nmol/L for 48 hours (Fig. 1B). Similar results were obtained at 72 hours in all conditions tested (data not shown). These antiproliferative and proapoptotic effects were correlated with decreased Y₅₉₁FLT3 and Y₇₇₉AXL phosphorylation as well as decreased activation of its downstream signaling. Indeed, densitometry showed dose-dependent decrease in Y₇₇₉AXL phosphorylation in 3 AML cell lines [1 vs. 0.3, 0.1, 0.2, 0.0 (MV4–11); 1 vs. 1, 0.7, 0.3, 0.2 (MOLM-13); and 1 vs. 0.5, 0.3, 0.2, 0.1 (MOLM-14)], for treatment with gilteritinib at 5 nmol/L, 20 nmol/L, 50 nmol/L, and 100 nmol/L, respectively, compared with vehicle treatment. The decrease in FLT3 phosphorylation was also confirmed [1 vs. 0.7, 0.5, 0.6, 0.4 (MV4–11); 1 vs. 0.7, 0.4, 0.6, 0.4 (MOLM-13); and 1 vs. 0.6, 0.4, 0.4, 0.3 (MOLM-14)], for treatment with gilteritinib at 5 nmol/L, 20 nmol/L, 50 nmol/L, and 100 nmol/L, respectively, compared with vehicle treatment. Finally, data quantifications in Fig. 1C also showed a dose-dependent decrease in phosphorylation levels for pY₆₉₄STAT5, pSer₄₇₃AKT, and pT₂₀₂/Y₂₀₄ERK for all 3 cell lines.

The role of AXL in resistance to FLT3 inhibition has been demonstrated previously (13, 14). We were interested in whether dual FLT3/AXL inhibition provided by gilteritinib could be more relevant than single FLT3 inhibition by a highly selective drug such as quizartinib, particularly in the context of the hematopoietic niche. First, we determined that 3 nmol/L quizartinib and 100 nmol/L gilteritinib for 48 hours induced close levels of apoptosis in normoxia and without stromal cells for various FLT3-ITD cell lines. This strategy allowed us to compare the specific effect of such doses considering culture conditions mimicking the BM environment. In the absence of BMSC, gilteritinib and quizartinib induced quite similar apoptosis in the 3 FLT3-ITD AML cell lines in normoxia or in hypoxia at 1% O₂ (Fig. 2A, left panel). We previously demonstrated that without BMSC, the level of AXL expression in AML and the level of autocrine secretion of its ligand GAS6 are very low. In contrast, and as demonstrated in a previous work (14), we observed an upregulation of AXL in hypoxia and in BMSC coculture condition (Fig. 2B). Finally, we observed a significant increase in apoptosis induced by gilteritinib in hypoxia with BMSC compared with normoxia: 17.7 ± 5.3% vs. 3.4 ± 2.3% (MV4–11), 15.6 ± 7.9% vs. 3.7 ± 1.2 (MOLM-13), and 44.9 ± 5.4% vs. 15.8 ± 1.7% (MOLM-14), respectively (Fig. 2A, right panel).

Densitometry exhibited a strong dose-dependent inhibition of Y₇₇₉ AXL phosphorylation by gilteritinib compared with quizartinib, particularly in culture conditions associating hypoxia and stromal cells (Fig. 2B). However, this decrease of Y₇₇₉ AXL phosphorylation was not correlated with a strong inhibition of ERK and AKT which remained similar for quizartinib and gilteritinib suggesting that MAPK and PI3K pathways were mainly dependent of FLT3 signaling (Fig. 2B). Next, we investigated whether such effects were observed ex vivo in primary AML blasts. Results are shown in Fig. 2C and patients are described in Supplementary Table S3. These data showed that gilteritinib 100 nmol/L allows a trend (P = 0.06) to induce a higher apoptotic response in AML blasts in coculture as compared with quizartinib 3 nmol/L. Taken together, these results indicate that gilteritinib is more efficient in conditions mimicking the hematopoietic niche, suggesting that it could overcome protective mechanisms mediated by the BM microenvironment in vivo.

Encouraged by the in vitro data suggesting a therapeutic benefit of dual FLT3/AXL inhibition in the BM microenvironment, we were interested in whether gilteritinib could be more effective than quizartinib in vivo. The objective was to specifically examine the role of the BM microenvironment in the efficacy of both TKIs, so we used two settings: one excluding and one including BM microenvironment. In this context, the subcutaneous model appeared relevant to study in vivo the efficacy of TKIs outside the BM microenvironment. As previously performed in vitro, we determined dosing regimens in mice for quizartinib and gilteritinib that induced close levels of response in a subcutaneous model of FLT3-ITD AML cell line. We subcutaneously xenografted MV4–11 cells into immunodeficient NSG mice. When the tumor size reached 400 to 500 mm³, the mice were treated by gilteritinib 5 times per week for 2 weeks with 5 mg/kg/day, 10 mg/kg/day, and 30 mg/kg/day, or 5 mg/kg/day of quizartinib, which was previously determined to be an effective dose in vivo (14). According to guidelines for dose conversion between animals and humans (17) and the quantitative determination of quizartinib in mouse plasma (18), we expected that 5 mg/kg/day corresponds to a dosing regimen ranging from 40 to 50 mg/day in humans (19, 20). At day 28, the median tumor volume was increased to 777.1 mm³ with 5 mg/kg/day of gilteritinib, whereas it decreased to 84.2 mm³ with 10 mg/kg/day of gilteritinib. Mice were in complete remission at day 28 with quizartinib 5 mg/kg/day and gilteritinib 30 mg/kg/day with the same response profile, and at day 40 with gilteritinib 10 mg/kg/day. Treatment by FLT3-TKI was discontinued when mice were in complete remission, except for mice in the gilteritinib–5 mg/kg/day arm, which were sacrificed at day 28 for ethical reasons (Fig. 3A). Finally, median time to relapse was 13 days for quizartinib–5 mg/kg/day arm and 14 days for gilteritinib–30 mg/kg/day arm (P = NS), whereas it was at 2 days for gilteritinib–10 mg/kg/day arm (P < 0.001 compared with the two other arms; Fig. 3B). Based on the quantitative determination of the gilteritinib in mouse plasma (21), we expected that 30 mg/kg/day corresponds to a dosing regimen at 120 to 130 mg/day in humans (22). Finally, both dosing regimens coincided with clinically relevant posology used as single agents in clinical trials.

This strategy allowed us to compare the specific effect of such doses considering engraftment in the BM environment in the next experiment. We first used FLT3-ITD MV4–11 cells that were xenografted by vein injection. Leukemic cell engraftment was analyzed by bioluminescence imaging (BLI) as described previously (14). At day 7, one group of the mouse cohort was treated by daily oral gavage for 14 days with quizartinib (5 mg/kg/day), one group by gilteritinib (30 mg/kg/day), and a control group with vehicle. Response to treatment was assessed by BLI at day 14, and then at day 20 (Fig. 3C). At day 14, a significant decrease in total bioluminescence signal was observed in response to quizartinib compared with the control group (4.3 × 10⁷ Ph/s/sr vs. 1.5 × 10⁶ Ph/s/sr; P < 0.001) and in response to gilteritinib compared with the control group (4.3 × 10⁷ Ph/s/sr vs. 0.8 × 10⁶ Ph/s/sr; P < 0.001). This leukemic burden decrease was stronger at day 14 (0.8 × 10⁶ Ph/s/sr vs. 1.5 × 10⁶ Ph/s/sr; P < 0.05) and at day 20 (0.5 × 10⁶ Ph/s/sr vs. 1.3 × 10⁶ Ph/s/sr; P < 0.05) with gilteritinib compared with quizartinib, respectively (Fig. 3C).

Encouraged by these in vivo results with the FLT3-ITD AML cell line MV4–11, we explored the effects of these two FLT3-TKI in vivo on two different batches of AML PDX cells generated from a patient with FLT3-ITD AML using the process presented in Fig. 4A. The first batch contained FLT3-ITD and WT AML cells with an ITD/WT ratio of 0.44. The second batch contained FLT3-ITD AML cells with LOH. We transplanted AML PDX cells of these two batches into NSG mice and monitored engraftment in BM using flow cytometry (CD45/CD33 double staining). Then, mice were treated by daily oral gavage with quizartinib (5 mg/kg/day) and gilteritinib (30 mg/kg/day) for 2 weeks. Response to treatment was assessed by flow cytometric analysis of human CD45⁺/CD33⁺ blasts from crushed femurs, tibias, and hips (Fig. 4B). Analysis of the results based on molecular characterization of the transplanted batches confirmed that mean percentage of human CD45⁺/CD33⁺ blast cells in the BM at week 9 posttransplantation was significantly higher in quizartinib-treated mice (27.8% ± 5.8) than in gilteritinib-treated mice (15.4% ± 2.5) in FLT3-ITD AML PDX with LOH (Fig. 4C). Conversely, mean percentage of human CD45⁺/CD33⁺ blast cells in the BM at week 9 posttransplantation was similar in quizartinib-treated mice and in gilteritinib-treated mice in FLT3-ITD AML PDX with an ITD/WT ratio of 0.44 (Fig. 4D). These results suggested that gilteritinib displayed a higher level of efficacy in BM microenvironment for AML with high FLT3-ITD/WT ratio.

After demonstrating the potential for dual FLT3/AXL inhibition, we were interested in the tolerance of normal murine hematopoiesis to gilteritinib treatment compared with quizartinib as it also impacts C-Kit signaling. Three days post–FLT3-TKI treatment, there were no differences in BM cellularity between vehicle-treated mice, quizartinib-treated mice, and gilteritinib-treated mice in a nontransplanted NSG group (Fig. 5A, left panel). In contrast, in AML-PDX NSG group, murine leukocyte cellularity was significantly higher in gilteritinib-treated mice [median 89.3 × 10⁶, interquartile range (IQR) 83.7–95.3] than in quizartinib-treated mice (median 73.0 × 10⁶, IQR 45.1–78.6; Fig. 5A, right panel). Finally, comparing across transplanted and nontransplanted groups, we showed that murine BM cellularity was returned to normal only with gilteritinib. When separating murine leukocytes between differentiated precursor cells defined as mCD45⁺ Lin⁺ cells (Fig. 5B) and a broader immature population defined as mCD45⁺ Lin⁻ cells (Fig. 5C), the same conclusions can be drawn. We next decided to investigate the more primitive compartments (Supplementary Fig. S3A) such as hematopoietic stem/progenitor cells (HSPC) using mCD45⁺ Lin⁻ c-Kit^High immunophenotyping approach; there were here some differences between vehicle-treated mice, quizartinib-treated mice, and gilteritinib-treated mice in the nontransplanted NSG group suggesting a slight impairment of this compartment induced by FLT3-TKI, particularly gilteritinib (median 5.7 × 10⁶, IQR 5.3–6.3 vs. 4.6 × 10⁶, IQR 4.4–5.3 vs. 4.2 × 10⁶, IQR 3.9–4.4 in vehicle, quizartinib, and gilteritinib, in nontransplanted NSG, respectively). However, the absolute numbers of murine mCD45⁺ Lin⁻ c-Kit^High cells were similar in both gilteritinib-treated mice (median 3.5 × 10⁶, IQR 3.3–4.6) and quizartinib-treated mice (median 3.2 × 10⁶, IQR 1.8–4.3) in AML-PDX NSG group. Finally, as observed for mature compartments, gilteritinib was able to maintain the level of mCD45⁺ Lin⁻ c-Kit^High cells in AML-PDX NSG group close to the one observed in gilteritinib-treated nontransplanted NSG group whereas quizartinib did not (Fig. 5D). It should nevertheless be noted that outside the leukemic context, gilteritinib impairs normal primitive murine hematopoiesis more than quizartinib (Fig. 5C and 5D, left panels). Taken together, these data could suggest either a better efficacy of gilteritinib in BM clearance of AML blasts, leading to a better murine hematopoiesis, or a toxicity of quizartinib on normal murine hematopoiesis specifically in the context of AML, since this point is not observed in nontransplanted NSG mice. To discriminate between these two hypotheses, we analyzed murine hematopoiesis according to BM clearance as previously shown in Fig. 4C and 4D. In “LOH” group, BM clearance was higher with gilteritinib, whereas in “ratio 0.44”, BM clearance was the same with both treatments. In both FLT3-ITD AML-PDX group “LOH” and “ratio 0.44”, independently of AML blasts clearance, murine leukocyte numbers were higher upon gilteritinib treatment compared with quizartinib treatment (Fig. 5E). This rules in favor of a lower toxicity for gilteritinib for maintaining normal murine hematopoiesis in a BM microenvironnement impaired by AML.

In front of these results, we next assessed HSPC functions. First, we performed CFC assay at sacrifice of nontransplanted NSG mice and found that none of the FLT3-TKI impaired progenitors' ability to generate colonies (Fig. 5F). Then, we analyzed complete blood count (CBC) at day 4, day 18, and day 31 post–TKI treatment in a nontransplanted mice group. We did not find any clinically relevant effects of either TKI, quizartinib, or gilteritinib, supporting the fact that these treatments are well tolerated at theses doses on a steady-state hematopoiesis whether on white blood cells (WBC), red blood cells, platelets, lymphocytes, monocytes, or granulocytes (Supplementary Fig. S1). To confirm our conclusions, we performed a new set of experiments on C57Bl/6J mice. During treatment, CBC was followed and did not show any differences between the three groups (Supplementary Fig. S2). As observed in NSG counterparts, 3 days post–FLT3-TKI treatment, there were no significant differences between vehicle-treated mice, quizartinib-treated mice, and gilteritinib-treated mice, for the following cell subsets: Lin⁻ cells (Fig. 5G), Lin⁻ Sca1⁺ c-Kit^High cells (Fig. 5H), and hematopoietic stem cell (HSC) immunophenotypically defined as LSK CD48⁻ CD150⁺ CD135⁻ CD34⁻ (Fig. 5I; Supplementary Fig. S3B). We also analyzed CFU-GEMM, -GM, -G, -M and Red/Mk (Fig. 5J) without noticing any significant differences.

Taken together, these data support the fact that gilteritinib and quizartinib display close toxicity profile on normal murine hematopoiesis at steady state, even if gilteritinib displays a slight impairment of primitive compartment without impact on CBC. Conversely, quizartinib displays higher toxicity on murine hematopoiesis in BM impaired by AML.

To further explore the ability of both FLT3-TKI to target leukemic initiating cells in their microenvironment, we performed secondary transplants from BM of gilteritinib-treated and quizartinib-treated AML PDX cells (10⁶ AML-PDX alive cells/secondary recipient) as described in Fig. 4B. Once again, analysis of the results (Supplementary Fig. S4) based on molecular characterization of the engrafted batch revealed that gilteritinib was more effective than quizartinib in controlling leukemic stem–cell (LSC) engraftment and leukemic burden in FLT3-ITD AML PDX with LOH (Fig. 6A) compared with FLT3-ITD AML PDX with an ITD/WT ratio at 0.44 (Fig. 6B), in which quizartinib and gilteritinib displayed the same levels of efficacy.

Here, we demonstrated that dual AXL and FLT3 inhibition should target FLT3-ITD leukemic cells protected from cytotoxic treatments by various mechanisms including low O₂ concentration, BMSC, and soluble factors in the BM microenvironment. AXL is ectopically- or over-expressed in a wide variety of cancers and has always been associated with a poor prognosis (23). We have reported resistance mechanisms involving AXL in chronic myeloid leukemia (24). In AML, AXL and GAS6 levels of expression have been related to poor outcomes (25, 26). Paracrine AXL activation has been shown to induce AML resistance to conventional chemotherapies but also to FLT3-targeted therapy (12, 13, 27, 28). We previously demonstrated that AXL was particularly relevant in the setting of the hematopoietic niche. These data highlighted that the BM microenvironnement hampers the response to quizartinib through combined upregulation of AXL expression and activation. Accordingly, inactivation of AXL by knock-down, pharmacologic inhibition or ligand trapping, sensitized leukemic cells to quizartinib (14), suggesting that gilteritinib could be a candidate to improve FLT3-ITD AML-initiating cell eradication.

We confirmed that gilteritinib induced dose-dependent antiproliferative and proapoptotic effects associated with an inhibition of FLT3 activation and its downstream signaling (21). We demonstrated the specific interest of gilteritinib in the setting of hypoxia and BMSC coculture, linked to a better efficacy compared with quizartinib, another FLT3-TKI much more specific of FLT3. Gilteritinib also provided better response to treatment in BLI with a mouse model xenografted with a FLT3-ITD AML cell line. Then, in a FLT3-ITD AML-PDX model, we found that gilteritinib as single agent was more effective in controlling FLT3-ITD AML with a high allelic ratio, whereas its efficacy was similar to that of quizartinib for AML with FLT3-ITD and FLT3 WT clones. Finally, by performing a secondary xenograft, we reproduced identical results but neither quizartinib nor gilteritinib fully eradicated leukemic initiating cells. Nevertheless, gilteritinib-treated AML PDX displayed extended time in terms of leukemic burden, specifically in the FLT3-ITD AML with LOH batch. We could speculate that such efficacy by continuous treatment longer than 2 weeks could perhaps enable deeper impairment of FLT3-ITD AML clone. Finally, we could also speculate that other kinases inhibited by gilteritinib could also play a role in the BM microenvironment such as LTK or ALK (21). Indeed, effects of mesenchymal stromal cells and endothelial cells on AML cells have been investigated through contact factors and cytokine-mediated crosstalk. Several soluble mediators are known to be linked with remodeling of the HSC niche into a leukemic-permissive niche such as angiogenic growth factors like VEGF, Ang-1, proinflammatory cytokines such as IL-6, and homing-disrupting cytokines such as CXCL12. Other studies also detected variation according to osteoblastic or adipocytic BMSC differentiation, but also immune cells secreting FLT3 ligand or IL-1 (29).

Using two FLT3-ITD AML-PDX models, we showed that quizartinib was as effective as gilteritinib when PDX displayed both FLT3-ITD and WT clones. These experiments reproduced outcomes of phase II clinical trials in which FLT3-TKI as single agent allowed 12% and 45% of overall response in FLT3 WT AML for gilteritinib (30) and quizartinib (31), respectively. Quizartinib is known to inhibit KIT phosphorylation, leading to an impaired hematopoiesis (7). If gilteritinib does not inhibit KIT, its use in routine clinical practice unveils also a certain impairment of normal hematopoiesis, independently of its ability to induce blast differenciation (10). Here, we provide some data regarding murine hematopoiesis, demonstrating that gilteritinib and quizartinib display close toxicity profile on normal murine hematopoiesis at steady state without impact on CBC and potentially different toxicity profiles on hematopoiesis impaired by AML involvement.

These different endpoints raise an issue with clinical implications. In this article, quizartinib was equally effective as gilteritinib as a single agent in AML carrying both FLT3-ITD and FLT3 WT clones. Both FLT3-TKI have been developed in the R/R situation and the phase III outcomes showed that gilteritinib was better for complete remission but just slightly better for OS. In a front-line situation, FLT3-ITD AML also carries FLT3 WT clones and is treated with intensive chemotherapy, which by itself induces deep toxicity in normal hematopoiesis. Quizartinib could be an effective treatment in this case regardless of its action on normal hematopoiesis. Moreover, priming with quizartinib has recently been shown to inhibit chemotherapy-induced myelosuppression by inducing a transient quiescence of multipotent progenitors, protecting these cells from chemotherapy (32). In contrast, gilteritinib as single agent is more effective and less toxic than standard chemotherapy, establishing a new standard of care for management of R/R FLT3-mutated AML. Our results suggest that its intrinsic characteristics make it most effective as single agent in high allelic ratio FLT3-ITD AML.

As single agents, FLT3-TKI such as gilteritinib or others are unlikely to result in a high cure rate for FLT3-ITD AML. Gilteritinib, quizartinib, or crenolanib could improve the cure rates for patients with FLT3-mutated AML, but only when incorporated into a treatment regimen that includes other chemotherapies or drugs. In that setting, ability of gilteritinib to inhibit AXL could be a leverage to eradicate FLT3-mutated AML clones but could also help to target FLT3 WT AML clones shielded by their microenvironment since AXL resistance to chemotherapy is encountered beyond FLT3-ITD AML cells. These data strongly support current clinical trials associating chemotherapy and gilteritinib.

P.-Y. Dumas reports grants and personal fees from Astellas and Daiichi-Sankyo during the conduct of the study, as well as personal fees from Janssen, BMS Celgene, and AbbVie outside the submitted work. A. Pigneux reports grants from Astellas during the conduct of the study. No disclosures were reported by the other authors.

P.-Y. Dumas: Conceptualization, resources, data curation, formal analysis, funding acquisition, methodology, writing–original draft. A. Villacreces: Investigation, writing–original draft. A.V. Guitart: Investigation. A. El-habhab: Investigation. L. Massara: Investigation. O. Mansier: Investigation. A. Bidet: Resources. D. Martineau: Investigation. S. Fernandez: Investigation. T. Leguay: Resources. A. Pigneux: Writing–review and editing. I. Vigon: Writing–review and editing. J.-M. Pasquet: Conceptualization, supervision. V. Desplat: Formal analysis, supervision, funding acquisition, methodology, writing–original draft.

We thank Mrs. Bernadette De Buhan and her family for their generous support. This work was supported by grants from Ligue contre le cancer (Comité des Pyrénées-Atlantiques). We thank the Cytometry Facility, Vectorology Facility, Cell'Oxia, and Animals Facility A2 of the TBMCore, Bordeaux University. P.-Y. Dumas was a recipient of the SIRIC Brio for research fellowship for one year, then the MD-PhD program from the University Hospital of Bordeaux for 2 years. We also acknowledge the Centre de Ressources Biologiques Cancer, Bordeaux Biothèques Santé (BB-0033–00036) at Bordeaux University Hospital for providing biological material.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.