FLT3 is genetically essential for ITD-mutated leukemic stem cells but dispensable for human hematopoietic stem cells Open Access

Acute myeloid leukemia (AML) is a caricature of normal blood development¹ with leukemic stem cells (LSCs) residing at the top of the malignant hierarchy, resisting chemotherapy and serving as a reservoir for relapse.^2-5 As most patients with AML die of relapse,⁶ better targeting of LSCs is required, but LSCs and hematopoietic stem cells (HSCs) share numerous biological properties, raising the question of how to intensively target LSCs without harming normal HSCs.

FMS-like tyrosine kinase 3 (FLT3) is an appealing therapeutic target because it is a powerful oncogene and the most frequently mutated gene in AML.^7-9 The internal tandem duplication (ITD) of FLT3 is an activating mutation linked to increased chemoresistance, relapse, and mortality, even after allogeneic stem cell transplantation.^10-12 Detection of FLT3-ITD mutation in morphological remission is reported to be a stronger predictor of relapse and death than methods currently used to detect measurable residual disease.¹³ The human FLT3 gene was cloned in 1993,^14,15 activating mutations were identified in 1996,¹⁶ and clinical studies with FLT3 inhibitors started in 2000.¹⁷ However, many years passed before FLT3 inhibitors produced results significant enough to justify their widespread clinical use.¹⁸ Despite consistent progress,¹⁹ the use of FLT3 inhibitors in frontline or relapsed FLT3-mutant AML has only met with modest success.^18,20,21 Key issues over the years have been the lack of selectivity and potency of inhibitors, the limited duration of inhibition, occurrence of resistance-conferring FLT3 mutations, and clonal adaptation that drives resistance to FLT3 targeting. Although deeper inhibition or even deletion of FLT3 might overcome some of these issues, it is still unknown whether LSCs are dependent on FLT3. Moreover, clinical efficacy has 2 requirements: that LSCs are eradicated to prevent recurrence and that normal HSCs are able to tolerate full FLT3 inhibition. There have been no comprehensive mechanistic and functional studies to determine whether FLT3 is essential for human HSCs or LSCs. The few available human studies do suggest that FLT3 is expressed in HSCs^22-25; however, its functional importance has not been reported. In mice, it is well known that HSCs do not express FLT3.^26-31 This divergence in FLT3 expression on murine and human HSCs renders mouse models unsuitable for exploring the role of FLT3 in human hematopoiesis.

Here, we performed CRISPR/Cas9-mediated knockout (KO) on human hematopoietic populations to evaluate the effects of FLT3 deletion in human LSCs and normal HSCs across ontogeny. Using functional and transcriptomic studies, we demonstrate that FLT3 is genetically essential for LSCs to drive leukemogenesis in FLT3-ITD AML but is dispensable for both FLT3–wild-type (WT) AML and healthy hematopoiesis.

Detailed experimental methods are described in the supplemental Methods, available on the Blood website.

Fetal liver (FL) samples were collected from elective pregnancy terminations at 16 to 19 weeks of gestation at Mount Sinai Hospital, processed within 3 hours, and CD34⁺ cells were isolated (Miltenyi Biotec) and viably frozen at −150°C. Umbilical cord blood (CB) samples were collected at Trillium, William Osler, and Credit Valley Hospitals and processed within 24 to 48 hours after birth, lineage (Lin)-positive cell depletion was performed (StemCell Technologies) and Lin⁻ cells were viably frozen. Adult bone marrow (BM) samples were collected during elective hip replacement at Centro Hospitalar Universitário de São João, and mononuclear cells were viably frozen. Primary AML samples were obtained from patients’ peripheral blood at Princess Margaret Cancer Centre, and mononuclear cells were viably frozen. All human samples were collected after patient informed written consent and in accordance with guidelines approved by the ethics board from each mentioned health institution and University Health Network Research Ethics Board. After thawing, HSCs defined as CD34⁺CD38^–CD45RA^–CD90⁺CD49f⁺ were sorted from FL and CB, and hematopoietic stem and progenitor cells (HSPCs) defined as HSCs and multipotent progenitors (CD34⁺CD38^–CD45RA^–CD90^–CD49f^–) were sorted from BM. AML samples (Table 1) were sorted according to previously functionally validated LSC-enriched populations (see the supplemental Methods): CD34⁺ leukemic cells were sorted from all AML samples except for WT3, for which CD45⁺CD3^– population was sorted.

A 2-step staining was optimized to characterize FLT3 surface expression and to perform cell sorting according to FLT3 expression. Cells were primarily stained with FLT3 CD135 biotin-conjugated mouse antihuman CD135, clone 4G8 (BD, 1:50), washed twice, and subsequently stained with streptavidin PE (BD, 1:250), all steps at 4°C.

To edit FLT3, a pair of guide RNAs (gRNAs) deleting 88 base pairs at exon 20 of the human FLT3 gene was selected from 10 predicted by CRoatan algorithm.³² The Control gene (the olfactory receptor OR2W5) was edited by a pair of gRNAs validated in previous work.³² FL, CB HSCs, and BM HSPCs were cultured for 48 hours and AML cells for 24 hours, and then CRISPR/Cas9 ribonucleoprotein (RNP) electroporation was performed using the 4D-Nucleofector (Lonza), chemically synthesized gRNAs (Integrated DNA Technologies), and recombinant Cas9 nuclease (Integrated DNA Technologies).^33,34 After recovering overnight, cells were intrafemorally injected into sublethally irradiated NSG female mice, except for experiments where alternative recipients were used (see the supplemental Methods). Short-term (2 weeks), long term (12-20 weeks), primary and secondary xenotransplantation studies were performed. All mouse experiments were approved by the Animal Care Committee of the University Health Network. After euthanizing the mice, injected femur and noninjected bones (contralateral femur and tibias) were flushed separately so BM was independently collected and analyzed by flow cytometry to characterize human hematopoietic engraftment, and genotyping and cytomorphological analysis were performed. The spleens were collected, crushed, and similarly analyzed. In some instances, the spleens were fixed and embedded in paraffin for subsequent histological analysis. Percentage of gene-KO was determined in recovered DNA from cells collected before transplantation and from engrafted cells, by Sanger sequencing and indel analysis using the online tool ICE Synthego³⁵ and DECODR.³⁶ 

Characterization of the mutational composition of the samples and 12-week grafts was undertaken using single-molecule molecular inversion probe (smMIP) technology³⁷ followed by FLT3-ITD mutation calling using FLT3_ITD_ext³⁸ and Pindel³⁹ algorithms.

CD34⁺ cells were sorted from sample ITD3, and CD34⁺CD38^– cells were sorted from CB. FLT3-KO and OR2W5-KO were performed in each sample individually, and both cell populations were co-transplanted in a 1:1 proportion using 10 000 to 20 000 cells from each sample per mouse. Human engraftment was analyzed to determine the presence and proportion of multilineage hematopoiesis vs leukemia, using flow cytometry, cytomorphology, histology, and genotyping analyses in primary and secondary recipients.

Three samples from each condition, ITD-positive AML (ITD1-3), AML without ITD mutation (WT1-3), and CB (CB1-3), were sorted: CD34⁺ population from all AML samples, except WT3 in which CD45⁺CD3^– population was sorted, and CD34⁺CD38^– population from CB samples. After a 24- to 48-hour culture, RNA was isolated using the RNeasy Micro kit (Qiagen), and deep sequencing was performed on the NovaSeq 6000 system. Differential gene expression and pathway enrichment analysis were performed as described previously.⁴⁰ 

To assess the importance of FLT3 expression for LSC function, we induced CRISPR/Cas9-mediated FLT3-KO in LSC-enriched populations from primary AML samples with FLT3-ITD mutation (ITD 1-7) and without FLT3-ITD mutation (FLT3-WT; WT 1-3) and tested their capacity to generate leukemia in NSG mice (Figure 1A; Table 1). Most leukemic blasts and CD34⁺ leukemic cells in both ITD-positive and FLT3-WT AMLs expressed surface FLT3 at baseline (supplemental Figure 1A). FLT3-KO was designed to target exon 20, affecting both ITD-mutated and FLT3-WT alleles and resulting in an 88 base-pair frame-shift deletion near the catalytically active site (supplemental Figure 1B-C). FLT3-KO significantly reduced the sizes of leukemic grafts and spleens in 6 of 7 ITD-positive leukemias compared with control-gene KO (targeting olfactory receptor OR2W5) (Figure 1B; supplemental Figure 1D-E). By contrast, little to no effect on leukemic engraftment was found in FLT3-WT leukemias (Figure 1B; supplemental Figure 1D-E).

To assess the degree of FLT3-KO editing in leukemic grafts, we determined the KO percentage in cells collected before transplantation and at 12 weeks posttransplant. In pretransplantation cells, the percentage of cells with KO was >80% in control and FLT3-KO groups (Figure 1C), establishing that all mice were transplanted with cells containing efficiently edited DNA. In 12-week leukemic grafts from samples ITD 1 to 6, the percentage of cells with FLT3-KO dropped close to 0 (Figure 1C). FLT3-KO using a different pair of gRNAs in samples ITD 1 and 2 generated similar results (supplemental Figure 1F), independently confirming our findings. This drastic reduction of FLT3-KO percentage upon engraftment reflects a significant disadvantage of FLT3-KO cells, causing them to be outcompeted by the small number of cells that had escaped FLT3 gene editing (Figure 1C). The residual grafts were leukemic by immunophenotype and morphology (supplemental Figure 1G-H) and expressed FLT3 at levels similar to controls (ITD2, 3, 5, and 6 as representative examples, supplemental Figure 1I). Moreover, analysis of mutational composition (Tables 1 and 2; supplemental Table 1) revealed that most mutations present in the original patient samples were well represented in the grafts. This included FLT3-ITD mutation that was present at high variant allele frequencies (VAFs) and high ITD/WT ratios in both the original patient samples (Table 1; Figure 1D) and in 12-week grafts (Table 2; Figure 1E). Together, these data demonstrate that the residual grafts were leukemic and generated from ITD-positive LSCs that escaped FLT3-KO.

By contrast with the data from samples ITD1-6, in all FLT3-WT AML samples and sample ITD7, FLT3-KO was maintained in >80% of cells in 12-week grafts (Figure 1C), demonstrating functional independence from FLT3. FLT3 expression was drastically reduced in 12-week grafts (ITD7 and WT2 as representative examples, supplemental Figure 1I), confirming that FLT3-KO disrupted protein expression. Compared with all other ITD-positive samples, the original ITD7 patient sample had the lowest ITD/WT ratio (<0.5, Figure 1D) and the lowest FLT3-ITD VAF (32%, Table 2). Upon engraftment, the FLT3-ITD mutation became undetectable in ITD7 grafts (Figure 1E; Table 2), although the engrafted cells harbored a U2AF1 mutation present in the original patient sample. Thus, ITD7 grafts were generated by FLT3-WT (non-ITD) subclones, thereby explaining why ITD7 grafts phenocopied FLT3-WT AML grafts after FLT3-KO. Together, these findings demonstrate that ITD-positive LSCs require FLT3 expression to sustain leukemic engraftment, whereas FLT3-WT LSCs, either from FLT3-WT AMLs or present within ITD AMLs, do not depend on FLT3.

To test whether FLT3-KO prevents early engraftment or alternatively leads to the elimination of ITD-positive leukemic cells over time, we induced FLT3-KO in samples ITD 1, 2, and 6 and performed short-term xenotransplantation experiments. Grafts were readily detectable at 2 weeks (supplemental Figure 1J), although the proportion of FLT3-KO cells was lower than that of the input cells (Figure 1F) and the level of engraftment was lower than that of control-edited cells (supplemental Figure 1J). Nevertheless, taken together with the fact that FLT3-KO DNA was no longer detected in 12-week grafts generated by the same ITD samples (Figure 1C), the presence of FLT3-KO cells in 2-week grafts indicates that FLT3 is dispensable for short-term engraftment of ITD-positive leukemic progenitors but is required for the function of ITD-positive self-renewing LSCs that sustain leukemic engraftment over time.

To evaluate the requirement for FLT3 in normal human hematopoiesis, we first characterized FLT3 expression in human FL, CB, and adult BM by cell surface and transcriptomic analysis. FLT3 was expressed in most HSCs and immature progenitors and became downregulated upon differentiation (Figure 2A; supplemental Figure 2A). HSCs expressing high or low levels of FLT3 produced equivalent grafts at 20 weeks in NSG mice (supplemental Figure 2B).

To test the importance of FLT3 for human HSC function, we induced CRISPR/Cas9-mediated FLT3-KO in HSCs from FL, CB, and HSPCs from BM, followed by xenotransplantation (Figure 2B). The percentage of HSC with FLT3-KO pretransplantation was >80% in all cell sources (Figure 2C). Human engraftment was successfully generated by FLT3-KO HSCs from FL, CB at 20 weeks, and HSPCs from BM at 12 weeks (Figure 2D). Compared with controls, engraftment levels were similar in mice transplanted with FLT3-KO cells from CB and BM but significantly lower in mice transplanted with FLT3-KO cells from FL (Figure 2D; supplemental Figure 2C). The percentages of cells with KO in xenografts as determined by human-specific polymerase chain reaction (PCR) were on average >80% for both FLT3-KO and control groups in FL, CB, and BM (Figure 2E; supplemental Figure 2D), establishing that most engrafted cells arose from FLT3-KO HSC. In addition, FLT3 surface expression on phenotypically defined progenitors at 20 weeks was significantly reduced in the FLT3-KO group compared with control, confirming that the KO was efficacious (Figure 2F).

By limiting dilution analysis of 20-week primary grafts, the frequency of repopulating HSCs after KO was similar between FLT3-KO and control groups from FL and CB (Figure 2G). Differentiation into all blood lineages (erythroid, myeloid, T and B lymphoid and megakaryocytic) was not significantly altered by FLT3 KO (Figure 2H). We did not find differences in the percentages of engrafted human CD34⁺ cells, nor in the distribution of phenotypic HSCs, multipotent progenitors, and multi-lymphoid progenitors in mice transplanted with FLT3-KO cells from FL and CB compared with controls (supplemental Figure 2E). Limiting dilution analysis of secondary grafts demonstrated a similar frequency of repopulating cells between FLT3-KO and control groups in CB; however, in FL, there was a more than threefold lower frequency in the FLT3-KO group (Figure 2I). To ensure that our findings were not model specific, we also used NSGW41 recipient mice that permit human hematopoietic engraftment without the need for irradiation.⁴¹ FLT3-KO did not affect the capacity of FL and CB HSCs to engraft nonirradiated NSGW41 mice (supplemental Figure 2F). When we used an alternative pair of gRNAs targeting exon 6 of FLT3 gene (supplemental Figure 2G), we obtained results similar to those with our original pair of gRNAs. To provide a positive control for our study, we targeted KIT, a gene known to be highly relevant for the HSC function.^42-45 In contrast to FLT3-KO, we observed a drastic reduction in repopulation ability in KIT-KO HSC (supplemental Figure 2H).

Overall, our findings demonstrate that FLT3 is dispensable for the function of normal HSCs from CB and BM but may be required for maintenance of self-renewal in HSCs from FL.

To model the impact of FLT3-KO on a human hematopoietic system composed of normal and leukemic cells, such as found in patients with AML, we knocked out FLT3 or OR2W5 (control gene) from CB (CD34⁺CD38^–) and ITD3 AML (CD34⁺) cells and co-transplanted them into NSG mice (Figure 3A). In the resulting FLT3-KO grafts, it was possible to infer the origin (normal CB vs ITD-positive leukemia) of cells based on the presence of FLT3-ITD or WT alleles because pretransplant ITD3 leukemic cells harbored the FLT3-ITD allele with a VAF of 95% (Table 1), with no detectable FLT3-WT allele (supplemental Figure 3A). Thus, engrafted cells bearing the ITD allele originated from transplanted leukemic cells, whereas cells bearing the FLT3-WT allele originated from normal CB cells.

We performed this competitive repopulation assay twice (supplemental Figure 3B; Figure 3B). In both experiments, mice from control groups with OR2W5-KO were sick, with high levels of profoundly myeloid-biased human leukemic engraftment (Figure 3C; supplemental Figure 3C-D), high percentage of BM blasts (Figure 3D), and enlarged spleens also massively infiltrated with blasts (Figure 3E; supplemental Figure 3E). Engrafted cells were enriched for the ITD allele, indicating their origin from transplanted ITD-positive leukemic cells (supplemental Figure 3F). Secondary recipients also developed leukemia with high levels of human engraftment and CD34⁺CD33⁺ immature cells (Figure 3F), twofold higher frequency of repopulating cells compared with FLT3-KO grafts (supplemental Figure 3G), and only the ITD allele was detected, confirming their leukemic origin (Figure 3G).

In experiment 1, FLT3-KO frequency in transplanted cells was 71% (supplemental Figure 3B), leaving a significant percentage (29%) of unedited leukemic cells. The resulting FLT3-KO grafts contained variable percentages of FLT3-KO cells, with dominance of nonedited cells in some cases (Figure 3B). Compared with control-edited grafts, levels of human engraftment were similar (Figure 3C); grafts contained higher proportions of lymphoid cells but were still predominantly myeloid with expression of immature myeloid markers (supplemental Figure 3C-D) and spleens were smaller than controls but still enlarged (supplemental Figure 3E). Grafts with high frequency of FLT3-KO harbored only the WT allele, revealing their origin in CB cells, whereas grafts with low FLT3-KO harbored both ITD and WT alleles (supplemental Figure 3F, left), consistent with mixed engraftment by nonedited ITD-positive cells and edited CB cells.

By contrast, in experiment 2, where FLT3-KO frequency in the transplanted cells was 81%, FLT3-KO grafts uniformly contained high percentages of FLT3-KO cells (≥80%; Figure 3B) and engraftment levels were significantly lower compared with control-edited grafts (Figure 3C). FLT3-KO grafts were composed of normal multilineage populations (Figure 3D; supplemental Figure 3C) expressing B lymphoid (CD19) and mature myeloid (CD14 and CD15) markers (supplemental Figure 3D) with <5% blasts (Figure 3D), and mice presented with normal-sized spleens without blast infiltration (Figure 3E; supplemental Figure 3E). Secondary recipients transplanted with FLT3-KO cells (≥90% KO) remained healthy and presented with normal human grafts (Figure 3F; supplemental Figure 3G). Both primary (supplemental Figure 3F, right) and secondary (Figure 3G) grafts harbored only the FLT3-WT allele with no detection of ITD, demonstrating their origin from normal CB cells. Together, these results suggest that 100% FLT3-KO is not needed (and indeed is likely not feasible) as long as a high enough editing efficiency is achieved to enable normal hematopoietic cells, which are not affected by FLT3-KO, to outcompete any residual ITD-positive leukemic cells that escaped gene editing.

To investigate the molecular basis of why FLT3 is indispensable for ITD-positive leukemic engraftment, we performed gene expression analysis (supplemental Figure 3H). As both FLT3-WT leukemia and normal cells were not dependent on FLT3 to engraft, we were particularly interested in genes/pathways unique to ITD-positive leukemia. Following FLT3-KO, confirmed at the genomic (supplemental Figure 3I) and transcriptomic levels (supplemental Figure 3J), cells from ITD-positive AML, FLT3-WT AML, and CB samples were cultured short-term and collected for bulk RNA sequencing. We first compared FLT3-KO and control-edited groups to generate a ranked list of differentially expressed genes, from which we performed pathway enrichment analysis (supplemental Table 2). FLT3-WT AML and CB samples were then used as negative controls to identify pathways that were uniquely downregulated by FLT3-KO within ITD AMLs: the top 30 of these pathways were mainly involved in G1 cell cycle phase, G1/S transition in mitosis, cell cycle checkpoints, chromosome maintenance, and DNA damage response (Figure 3H).

To test whether FLT3-KO would affect these pathways in vivo, we knocked-out FLT3 in sample ITD2, performed 2-week transplants (a time point that allowed detection of FLT3-KO cells in ITD-AMLs), and assayed cell cycle, cell death, and DNA repair (supplemental Figure 3K). The proportion of cells in S/G2/M was higher in FLT3-KO grafts than in the control group, whereas the percentage of live cells was lower (Figure 3I), indicating that FLT3-KO cells were cycling more frequently, but dying more frequently as well. To test the capacity to repair damaged DNA, 2-week engrafted mice were irradiated 1 hour or 10 hours before euthanasia, followed by evaluation of γ-H2AX expression in the engrafted cells. At 10 hours post-irradiation, most control cells had repaired their DNA breaks, whereas 16% of FLT3-KO cells still displayed γ-H2AX expression (Figure 3I), indicating that FLT3-KO induced a defect in DNA repair in the context of FLT3–ITD-positive leukemia. Overall, our results suggest that FLT3 plays a role in DNA repair and cell cycle checkpoints that is critical for the function of the ITD-positive LSCs that drive FLT3-ITD leukemias.

Here, we have established that the ITD-mutant LSCs driving leukemogenesis and relapse have an absolute genetic requirement for FLT3, whereas it is biologically dispensable for both normal HSCs and FLT3-WT LSCs. FLT3 has long been a compelling therapeutic target, but after many years of optimization, selective inhibitors have only revealed modest clinical success. By deleting FLT3 on LSCs and HSCs using CRISPR/Cas9 genome-editing technology, we have demonstrated that FLT3 targeting is indeed an effective therapeutic strategy as it selectively eradicates ITD-positive LSCs responsible for relapse of FLT3-ITD–mutant AMLs without harming HSCs within a competitive transplantation environment. Our work provides a strong rationale for the development of more effective drugs or gene editing-based FLT3-targeting strategies to improve patient outcomes.

Our finding that FLT3 deletion does not affect the ability of neonatal and adult HSCs to reconstitute hematopoiesis provides reassurance of the overall safety of intensifying FLT3 therapeutic targeting in AML. The mechanisms underlying the myelosuppression observed with clinical FLT3 inhibitors have not been fully elucidated,^46-49 but our study strongly argues against a direct effect of FLT3 targeting in normal HSCs. FLT3 deletion did slightly reduce engraftment ability and self-renewal in FL-derived HSCs; however, these effects will not likely affect the management of FLT3-mutant AML, which is typically a postnatal disease. FLT3 may play a specialized role in FL HSCs, which are actively expanding^50,51 and thus are likely more dependent on growth factors and their receptors, including FLT3. In support of this idea, FLT3 is expressed in a transient fetal mouse HSC population that disappears in adult mice.³⁰ 

FLT3-ITD mutations were previously reported to be enriched in defined leukemia stem and progenitor populations using immunophenotypic^52-54 and transcriptional approaches^1,55; but the functional effects of FLT3 targeting in LSCs were unclear. The precise genetic deletion of FLT3 we undertook in this study avoids potentially confounding nonspecific effects of pharmacological inhibition. Our results demonstrate that FLT3-ITD but not FLT3-WT LSCs depend on FLT3 to generate persistent leukemic grafts. FLT3-KO effectively prevented leukemic engraftment of samples with high ITD/WT ratio, as most, if not all, cells harbored an ITD mutation. However, when a sample had low ITD/WT ratio and contained a large proportion of FLT3-WT cells, such as sample ITD7, then the FLT3-WT cells were still able to generate leukemic grafts despite losing FLT3 expression. These findings discourage the use of potent or precise FLT3-targeting as monotherapy, as WT clones will not be targeted. Our results also challenge the reported beneficial effects of FLT3 targeting in FLT3-WT AMLs,⁵⁶ suggesting that these effects might be attributed to multikinase inhibition or to inhibition of emerging FLT3-ITD clones at relapse rather than to specific effects of FLT3 targeting on WT LSCs. Future studies incorporating more samples with WT FLT3, samples with small ITD clones, and with non-ITD FLT3 mutations (TKD and noncanonical) would be of great interest.

Consistent with the selective targeting of ITD-positive LSCs on FLT3-KO, our study identified a signature comprising pathways involved in cell cycle checkpoints and DNA repair that were downregulated upon FLT3 deletion specifically in ITD-positive AMLs. We validated this signature in functional assays, where FLT3-KO increased cell division but also impaired DNA repair, increasing cell death and leading to the extinction of ITD-positive leukemic cells. DNA damage and repair has been linked to the FLT3-ITD mutation.⁵⁷ Several studies have suggested that some repair pathways work more efficiently in ITD AML than in FLT3-WT AML, whereas others are defective.⁵⁸ Cancer stem cells⁵⁹ and specifically LSCs⁶⁰ possess enhanced DNA repair mechanisms that contribute to resistance to genotoxic chemotherapy and radiation. Our data suggest that FLT3 plays a role in DNA repair mechanisms and cell cycle checkpoints in ITD-positive LSCs, contributing to the high relapse rates found in these patients.

Overall, our findings support the design of more potent and specific FLT3-inhibiting drugs and provide a strong rationale for the development of in vivo genetic engineering approaches, such as CRISPR/Cas9 genome engineering,⁶¹ to delete FLT3 within ITD-positive AML. The lack of demonstrated effects of FLT3-KO on FLT3-WT LSCs highlights the need for a multitargeted approach to treat AML, especially considering the genetic heterogeneity and clonal evolution that characterize this disease.

The authors thank the Orthopedics Department of Centro Hospitalar de São João for assistance with BM collection; the Obstetrics units of Trillium Health, Credit Valley, and William Osler Health for the CB samples; the Research Centre for Women's and Infants’ Health Biobank (Mount Sinai Hospital) for the FL samples and B. Chow at the Pathology and Laboratory Medicine (Mount Sinai Hospital) for assistance with pathology; M. Peixoto for assistance with BM samples and proofreading of the manuscript; M. D’Souza and R. Lopez at the Animal Resources Centre, University Health Network (UHN), for support with mouse work; N. Simard from SickKids-UHN Flow and Mass Cytometry Facility and Frances Tong from Princess Margaret Flow Cytometry Facility for assistance with antibody panel design and flow cytometry; Apresto at The Centre for Applied Genomics (SickKids) for sanger sequencing; K. Ho at the Centre for Applied Genomics (SickKids) for next-generation sequencing; K. Asoyan, C. Cimafranca, J. Mouatt, M. Peralta, and Y. Yang at the Pathology Research Program (UHN) and N. Law at the Sttarr Innovation Centre (UHN) for assistance with histology; L. Shultz at the Jackson Laboratory for providing NSGW41 mice; P.S. Coelho, N. Santos, A. Zeng, K. Kaufman, and S. Xie for their insight and suggestions; the laboratories of S. Chan and F. Notta for sharing equipment; N. Mbong, S. de Silva, and M. Anders for technical assistance; and all members of J.E.D.’s laboratory for critical review of the manuscript.

J.L.A. received funding from the Fundação para a Ciência e Tecnologia (SFRH/BD/136200/2018). Work in the laboratory of J.E.D. is supported by funds from the Princess Margaret Cancer Centre Foundation, the Canadian Institutes of Health Research (Foundation no. 154293 [J.E.D.]; operating grant nos. 154293 and 89932 [J.E.D.]; International Development Research Centre, Canadian Cancer Society grant no. 703212 [J.E.D.]), Terry Fox Research Institute Program Project grant, Ontario Institute for Cancer Research through funding provided by the Government of Ontario, a Canada Research Chair, and the Ontario Ministry of Health and Long Term Care. Work in the laboratory of P.P.-Ó. was funded by Fundo Europeu de Desenvolvimento regional funds through the COMPETE 2020, Operational Program for Competitiveness and Internationalization, Portugal 2020, and Portuguese funds through Fundação para a Ciência e Tecnologia do Ministerio da Ciência, Tecnologia e Ensino Superior in the framework of the project POCI-01-0145-FEDER-032656.

Contribution: J.L.A., J.E.D., and P.P.-Ó. conceived the study; J.L.A. and J.E.D. wrote the manuscript; E.L., E.W., O.I.G., J.C.Y.W., P.P.-Ó., and M.A.S.-S. edited the manuscript; J.L.A., E.L., E.W., and O.I.G. analyzed the experiments; J.L.A. and E.W. designed and performed the CRISPR/Cas9 experiments; J.L.A., E.W., E.L., O.I.G., and B.G. performed in vitro and in vivo experiments; J.M., L.J., A.M., and S.C. assisted with mouse work; J.M. and M.D. performed intrafemoral injections; V.V. and S.B. performed RNA sequencing analysis; A.A. and M.D.M. provided primary acute myeloid leukemia samples and coordinated patient consent and sample collection; J.-M.C.-C. performed the ITD/WT ratio assay; S.A. performed the single-molecule molecular inversion probe analysis; A.T. and D.K. performed FLT3-ITD analysis in next-generation sequencing data; J.C.Y.W., M.A.S.-S., and M.D.M. provided study consultation; and J.E.D. secured funding and supervised the study.

Conflict-of-interest disclosure: J.E.D. reports research funding from Bristol Myers Squibb and licensing of SIRPα to Trillium Therapeutics and Pfizer. The remaining authors declare no competing financial interests.

Correspondence: John E. Dick, Princess Margaret Cancer Centre, University Health Network, Toronto, ON M5G 1L7, Canada; email: john.dick@uhn.ca.