Reversal of Acquired Drug Resistance in FLT3-Mutated Acute Myeloid Leukemia Cells via Distinct Drug Combination Strategies

One third of the patients with acute myeloid leukemia (AML) have gain-of-function mutations in the fms-like tyrosine kinase 3 (FLT3) gene. These mutations include FLT3 internal tandem duplication (FLT3-ITD) mutations in the juxtamembrane (JM) domain and point mutations in the tyrosine kinase domains (TKD, e.g., mutations at D835/836; refs. 1 and 2). Leukemias with FLT3 mutations are associated with a higher risk of relapse following therapy and, as a result, shorter overall patient survival (1, 3). A number of small molecule inhibitors targeting constitutive FLT3 activation have been developed and are categorized as type I and type II TKIs (4, 5). Type I TKIs (e.g., CEP701 and PKC412) bind in and around the region occupied by the adenine ring of ATP and do not require the “DFG-out” conformation (6). Type II TKIs (e.g., sorafenib, AC220, and MLN518) occupy an allosteric site that is directly adjacent to the ATP binding pocket created by the activation-loop that adopts a “DFG-out” inactive conformation (7, 8).

We have identified sorafenib as a potent, clinically effective inhibitor of FLT3-ITD mutations and demonstrated an encouraging and effective reduction in leukemia blasts in peripheral blood and bone marrow. However, only 10% of the FLT3-ITD-mutated patients with AML achieved complete remission, or a complete remission with incomplete platelet recovery (CRp), in our phase I clinical trial (9, 10). Furthermore, relapse routinely occurs after prolonged therapy with sorafenib as well as other small-molecule TKIs including type II inhibitors AC220 and MLN518, and the type I inhibitor CEP701 (10–19). These findings would suggest that preventing, or surmounting, the resistance mechanisms associated with FLT3-targeted therapies of AML is of paramount importance for the successful therapy of this disease.

The mechanism(s) of secondary AML resistance to FLT3 inhibitors are poorly understood. Acquired point mutations in the FLT3 gene have been reported in a cell line that exhibited resistance to the type 1 TKI PKC412 (20). Secondary FLT3 point mutations that conveyed resistance to FLT3 inhibitors were also detected in cultured leukemic cells exposed to the mutagen N-ethyl-N-nitrosourea (21, 22), or the type II TKI MLN518 and AC220 (13, 23). Notably, Smith and colleagues recently identified secondary point mutations in either TKD1 or TKD2 region of FLT3 gene in all 8 relapsed ITD-mutated patients with AML who received AC220 monotherapy, which include D835 and F691 mutations. These mutations have also showed cross-resistance to sorafenib (13). Conversely, sorafenib resistance in 2 patients whose AML-harbored FLT3-ITD mutations was reportedly not linked to secondary point mutations in the FLT3 gene (24). Consequently, further investigation is needed to determine the biologic significance and potential mechanistic underpinnings of acquired FLT3 point mutations in the development of resistance to FLT3-targeted therapy. This tenet served as the basis for the assessments reported in this study.

Sorafenib and PKC412 were purchased from American Custom Chemicals Corporation and Sigma-Aldrich, respectively. Crenolanib (CP-868–596) was provided by AROG Pharmaceuticals LLC. MLN518, AC220, CI-1040, and CCI-779 were purchased from Skelleck Chemicals. Rabbit polyclonal antibodies were purchased from the following sources: phospho-FLT3 (Tyr589/591), phospho-ERK (Thr202/Tyr204), phospho-AKT (Ser473), phospho-S6K (Ser240/Ser244), phospho-PI3K, S6K, and cleaved caspase-3 all from Cell Signaling Technology; Bim from Calbiochem Co.; FLT3 from Santa Cruz Biotechnology; and Stat5A/B from R&D Systems. Mouse monoclonal antibodies were purchased from the following sources: phospho-Stat5 (Tyr694/699) from Upstate, Inc. and ERK2 from Santa Cruz Biotechnology.

Ba/F3-ITD cells, a murine pro-B lymphocyte line created by transfecting Ba/F3 cells with lentivirus containing human FLT3-ITD mutations were isolated and characterized as described previously (25) and kindly provided by Dr. J. Donald Small (Department of Pediatric Oncology, Johns Hopkins University, Baltimore, MD). The Ba/F3-ITD-Res cells were isolated by exposing Ba/F3-ITD to increasing low doses (i.e., 1–5 nmol/L) of sorafenib for up to 3 months, and further exposed for an additional 3 weeks to serial dilutions (5–50 nmol/L) of sorafenib. The resistant clones were collected for mutation analysis. The engineered cell lines with point mutations Ba/F3-ITD+651, Ba/F3-ITD+687, Ba/F3-ITD+676, Ba/F3-ITD+842, and Ba/F3-ITD+676/842 were developed by individually introducing D651G, I687F, N676D, and/or Y842C mutations, respectively, in the Ba/F3-ITD cells using lentiviral infection. Briefly, the human FLT3-ITD sequence was amplified by reverse transcription-PCR. The restriction enzymes were introduced into the cDNA by PCR, and the product was inserted into a CD510B-1 lentiviral vector using a ligation reaction. In each point mutation experiment, pairs of vector primers were prepared and used to amplify all of the vector sequences except for the region containing the sites to be mutated. For PCRs, a QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies, Inc.) was used following the manufacturer's instructions. The PCR products were transformed into XL10-Gold ultra-competent cells, and the vector clones were analyzed by Sanger-based DNA sequencing. The mutation-containing DNA vectors were produced by transient transfection of 293T cells using FuGENE 6 (Roche), and supernatants were harvested after 48 hours. Lentiviral infection was performed in Ba/F3-ITD cells, and GFP-positive cell populations were collected using fluorescence-activated cell sorting (FACSAria II, BD Bioscience). All cells were maintained in RPMI medium supplemented with 10% FBS.

Four patients with AML who lost response to sorafenib were selected from patients enrolled in phase I/II clinical trials conducted at the University of Texas M.D. Anderson Cancer Center from 2006 to 2009 (10, 11, 26). Bone marrow or peripheral blood samples were collected when resistance to the therapy was first observed. The patient characteristics are presented in Supplementary Table S1. The mononuclear cells in these samples were purified by Ficoll-Hypaque (Sigma-Aldrich) density–gradient centrifugation, and the cells were cultured as described above before treatment.

Total mRNA from the patient samples was isolated and point mutations were screened for cDNA-based sequencing. Briefly, total mRNA was prepared using TRizol (Life Technologies) reagent as described by the manufacturer. After reverse transcription with random hexamers (SuperScript III First-Strand Synthesis SuperMix; Life Technologies), a PCR-based mutation screen was performed by screening full length FLT3 gene using the following primer pairs: 5′-GGAGTTGTTTCCATGGTCATT-3′ and 5′-CGTTTCTTGCCACTGATGAT-3′; 5′-GCTTTTGTATCATCAGTGGCA-3′ and 5′-AAGAGGAGACAAACACCAATTG-3′; 5′-GCTTTTGTATCATCAGTGGCA-3′ and 5′-GAAGTTAGCATCAACCGGAAT-3′; and 5′-TGTCGAGCAGTACTCTAAACA-3′ and 5′-CTGAAAGAAAAAGCAGACAGCT-3′.

AML cells were plated in 24-well plates at 2 × 10⁵ cells/mL in 2 mL of medium. The indicated drugs, dissolved in dimethyl sulfoxide (DMSO, control; Sigma-Aldrich), were added to the culture medium (i.e., as single-agent or multiple agents in combination treatments) for 48- to 72-hour incubations. The Ba/F3-Flt3 cells were treated with the compounds in the presence of IL-3 (2 ng/mL). Cell viability was assessed using the Trypan blue dye exclusion method, and apoptosis was determined via FACS by annexin V positivity as described previously (9). The 50% inhibitory concentration (IC₅₀) for cell growth inhibition and the 50% effective concentration (EC₅₀) for apoptosis were calculated using CalcuSyn software (BioSoft). The Ba/F3-ITD or Ba/F3-ITD+842 cells were treated with indicated compounds in the presence of FL (25 or 50 ng/mL) for 48 hours, and apoptosis induction was determined as previously described above.

Phosphorylation and total protein levels were determined using Odyssey Infrared Imaging System (LI-COR Biosciences); semiquantitative immunoblotting data was generated using Beta 4.03 imaging software (Scion Corporation) by comparing the ratios of phosphoproteins and total proteins, or the loading controls tubulin (or GAPDH) as described previously (27).

Kinase assay was performed in vitro to examine if the inhibitors directly suppress phosphorylation of mutated FLT3 protein. ITD plus Y842C mutated FLT3 protein was isolated from Ba/F3-ITD+842 cells and phosphorylation enzyme reaction was performed in magnesium/ATP-containing reaction buffer in vitro for 30 minutes at 30°C in the presence/absence of crenolanib (0.5 μmol/L) and/or sorafenib (0.5 μmol/L) as described previously (9). Phosphorylation level of FLT3 protein was measured using immunoblotting and the ratio of phospho-FLT3 to total FLT3 was determined using Beta 4.03 imaging software as described above.

The data are presented as the means ± SD of triplicate samples or assays. The statistical analyses were performed using unpaired Student t test. A P ≤ 0.05 was considered statistically significant. Isobologram and combination index (CI) analyses were performed using CalcuSyn software (Biosoft; refs. 28 and 29). A CI value of 1 indicates an additive effect, a value of less than 1 indicates synergy, and a value of greater than 1 indicates antagonism. The average CI values were calculated at different effect levels (50% effective concentration ED₅₀, ED₇₅, and ED₉₀; ref. 30). A 2-sided Fisher exact test was used to determine statistical significance between different groups.

We screened for acquired mutations of the FLT3 gene in the sorafenib-resistant cell line Ba/F3-ITD-Res developed by long-term exposure of Ba/F3-ITD cells to low doses of sorafenib in vitro, and identified N676D and Y842C mutations in TKD1 and TKD2, respectively (Fig. 1A). Assuming that the point mutations in TKDs might be responsible for the resistance, we further studied samples from FLT3-ITD–mutated patients with AML who relapsed while on sorafenib. Indeed, in addition to the ITD mutations we identified acquired FLT3 point mutations in 2 of 4 patients, which were also located in the TKDs (i.e., G619C, D651G, and I687F; Supplementary Table S1). We introduced these mutations individually, or tandem mutations in both TKD1 and TKD2, into the Ba/F3-ITD cells and tested their in vitro sensitivity to sorafenib. Varying degrees of resistance were observed in the sublines with single point mutations, and the effective concentration in 50% of the treatment population (EC₅₀, mean ± SD) for apoptosis induction was 0.69 ± 0.18, 0.61 ± 0.13, and 0.17 ± 0.02 μmol/L for the cells containing mutations D651G, N676D, and I687F, respectively, in TKD1, and 2.6 ± 0.81 μmol/L for the cells containing Y842C mutation in TKD2. By comparison, the EC₅₀ for the parental cells Ba/F3-ITD was 0.006 ± 0.002 μmol/L (Fig. 1B). These results suggested that a mutation in either TKD was sufficient to provide resistance to sorafenib-induced apoptosis, and that a mutational alteration in TKD2 was more important to the acquired resistance compared to an analogous alteration in the TKD1. Furthermore, cells with compound mutations in both TKDs (e.g., the Ba/F3-ITD-Res and the engineered Ba/F3-ITD+676/842 cells) in addition to the ITD mutations displayed even greater resistance (EC₅₀ of 4.2 ± 1.50 and 6.6 ± 0.53 μmol/L, respectively; Fig. 1B), indicating a pivotal role for the integrity of both TKDs in maintaining sensitivity of FLT3-ITD AML cells to sorafenib.

Relevant biomarkers were investigated in the aforementioned cells to facilitate the understanding of the biologic consequences of these mutations. The results illustrated that point mutations in a single TKD augmented the FLT3 activation caused by the ITD mutation. This was accompanied by a modest upregulation of phosphorylated FLT3 that persisted to varying degrees despite sorafenib treatment (Fig. 1C). The cells harboring TKD2 mutations (e.g., Ba/F3-ITD+842 and Ba/F3-ITD+676/842) were least sensitive to the inhibitory effects of sorafenib, maintaining persistent phosphorylation of FLT3 and its downstream effectors ERK, AKT, Stat5, and S6K after exposure to sorafenib. However, the suppression of phosphorylated FLT3 itself did not parallel those of its downstream effectors, whereby an increased level of phosphorylated ERK was observed in Ba/F3-ITD+842 cells, and persistent phosphorylated-AKT, -Stat5, and -S6K in Ba/F3-ITD+676/842 cells (Fig. 1C). These findings implied that a conformational change of FLT3 protein, resulting from the gene point mutations, may interfere with a binding ability (i.e., accessibility) of sorafenib, and that mutations in TKD2 are more critical for impairing the binding of sorafenib than those occurring in TKD1. Moreover, these results also suggested that a bypass/feedback regulation could be an effective approach in the modulation of FLT3 downstream signaling.

We next assessed the sensitivity of either TKD1 and/or TKD2 point mutations to various small molecule FLT3 inhibitors. AC220 (also known as quizartinib), a potent apoptogenic agent in leukemia cells (31), was first tested in our resistant cell lines. AC220 demonstrated a similar pattern of resistance as observed with sorafenib treatment (Fig. 2A). Because both sorafenib and AC220 are type II TKIs, we further tested cell sensitivity by treating these cells with a type I TKI PKC412 as well as another type II TKI MLN518. The results demonstrated that all resistant cells retained sensitivity to PKC412 at submicromolar concentrations. However, MLN518 was only active in cells with TKD1 mutations, not in the cells with TKD2 or compound TKDs mutations (Fig. 2B). This further suggested that mutations in TKD2 (such as Y842C) might directly interfere with the binding ability of type II TKIs while retaining sensitivity to type I TKIs. A similar mechanism has been suggested in cells with acquired AC220 resistance, but retaining sensitivity to PKC412 (32).

The novel type I TKI crenolanib (CP-868,596) has recently been reported to retain activity against the imatinib-resistant PDGFR-α D842V mutation by targeting the active conformation of a kinase ATP pocket in gastrointestinal stromal tumor (GIST; ref. 33). In this study, crenolanib demonstrated impressive apoptosis induction in all tested sorafenib-resistant cell lines at submicromolar concentrations and, furthermore, the concentrations required for cell growth inhibition were 10-fold lower than that required for apoptosis induction (Fig. 2C). We also tested crenolanib in AML cells harboring FLT3-D835 mutation, which is commonly found in patients with acquired resistance to FLT3 inhibitors. Interestingly, crenolanib induced profound proapoptotic effects in Ba/F3-D835G cells (Supplementary Fig. S2).

With the exception of the Ba/F3-D835Y cells, immunoblot analysis revealed that very low concentrations (i.e., 0.05 and 0.2 μmol/L) of crenolanib exerted a marked suppression of FLT3 phosphorylation, as well as that of its downstream effectors ERK and AKT, and induced caspase-3 cleavage in the sorafenib resistant cells (Fig. 2D). Interestingly, Ba/F3-D835Y displayed very low basal levels of phosphorylated FLT3 at the Tyr589/591 sites, which may limit the effectiveness of crenolanib in targeting FLT3. These findings indicated a striking potency of crenolanib in targeting FLT3-ITD cells with acquired point mutations of the TKDs, especially in those with TKD2 mutations.

We next assessed the impact of combining the type I TKI crenolanib with the type II TKI sorafenib in sorafenib-resistant cells. This combination showed profound synergistic proapoptotic effects even at 10-fold lower concentrations than the growth inhibitory concentrations of crenolanib as single-agent treatment. The CIs were 0.70 ± 0.01, 0.67 ± 0.03, 0.56 ± 0.12, and 0.82 ± 0.08 in Ba/F3-ITD-RES, Ba/F3-ITD+842, Ba/F3-ITD+676/842, and Ba/F3-ITD cells, respectively (Fig. 3A and Supplementary Fig. S3). Immunoblot analysis demonstrated that the combination enhanced the suppression of FLT3 downstream proteins ERK and AKT, especially in the Ba/F3-ITD-RES cells after a 1-hour treatment, but showed modest suppression of phosphorylated FLT3 itself (Fig. 3B). This data imply that the synergistic effects resulted not only from FLT3 suppression but also from blockade of other signaling events involved in the modulation of these pathways, at least in more resistant cell Ba/F3-ITD-RES. In addition, our in vitro kinase assay further demonstrated that these agents were capable of directly targeting the FLT3 protein and suppressing its phosphorylation (Fig. 3C).

To further evaluate if this synergistic effect was exclusive for the combination of sorafenib and crenolanib, the Ba/F3-ITD+676/842 cells were exposed to different combinations of type I and type II TKIs. The results indicated that each type I/type II combination (e.g., crenolanib plus sorafenib, crenolanib plus MLN518, or PKC412 plus sorafenib) exerted identical synergistic proapoptotic effects in the resistant cells (Fig. 3D and Supplementary Fig. S4). Interestingly, the combination of 2 type I TKIs (i.e., crenolanib and PKC412) also triggered a synergistic effect in cells that were resistant to the type II TKI sorafenib. However, the synergistic effect was not apparent in a combination of 2 type II TKIs (i.e., sorafenib and MLN518; Fig. 3D), implying that cells resistant to type II TKIs might still retain sensitivity to a type I TKI. Of note, these TKI combinations did not show significant antileukemic effects in the Ba/F3-FLT3 wide-type cells, even when increasing the drug concentration up to micromolar levels (Supplementary Fig. S5), suggesting that the cytotoxic effect of the combinations was chiefly in the FLT3-mutated AML cells.

Our data demonstrate that sorafenib exhibited impaired ability to suppress phosphorylated ERK or AKT/S6K in the resistant Ba/F3-ITD+842 or Ba/F3-ITD+676/842 cells, but still retained effectiveness, to certain degree, in the suppression of phosphorylated FLT3. This would imply a possible feedback upmodulation of MEK/ERK and/or AKT/S6K signaling pathways in resistant cells. Therefore, we examined the concomitant blockade of FLT3 and ERK/AKT signaling pathways. When we combined sorafenib with the MEK inhibitor CI-1040, there was a synergistic proapoptotic effect in the sorafenib-resistant cell lines. Less synergistic effect, although statistically significant, could be observed in combination of sorafenib with the mTOR antagonist CCI-779 by comparison to the single agent(s) treatments. Triple combination of sorafenib/CI-1040/CCI-779 triggered marked synergistic proapoptotic effect (Fig. 4A and B). This was accompanied by a profound suppression of phospho-ERK, -AKT, and -S6K and the upregulation of the proapoptotic BH3-only protein Bim (Fig. 4C). These results illustrated that the concomitant inhibition of FLT3, and its downstream targets MEK/ERK, could produce a potentially clinical beneficial result by the killing of sorafenib-resistant cells.

Next, we examined the combination effects of sorafenib and crenolanib in FLT3 ITD/D835 mutated primary human AML cells. Two of 4 the patients received TKIs therapy before current ex vivo study and became to relapsed/refractory (Supplementary Table S1). Synergistic or additive proapoptotic effects could be achieved in the AML samples at physiologically achievable drug concentrations (refs. 34 and 35 and Fig. 5A). Interestingly, this combination had very limited cytotoxicity in normal bone marrow samples or FLT3 wild-type AML patient samples (Supplementary Figs. S6 and S7), implying specificity for targeting FLT3-ITD–mutated patients with AML with the combination strategy. Immunoblotting showed the suppression of FLT3 phosphorylation, as well as that of its downstream effectors ERK and AKT, which was observed in either the crenolanib or sorafenib single-agent treatments. However, a substantial suppression of these phosphorylated proteins was also observed in the combination treatment (Fig. 5B), suggesting the synergistic suppression of FLT3 and its downstream targets.

An increase in FLT3 ligand as a cellular response to chemotherapy has been reported as a cause of resistance to FLT3-targeted therapy (36). We assumed that FLT3 ligand might activate FLT3 signaling and its downstream cascade signaling MEK/ERK or mTOR/S6K in our AML cell systems. Thus, we first tested the effects of the sorafenib and/or crenolanib in FLT3-ITD–mutated cells in the presence or absence of FLT3 ligand. The results showed that apoptosis induction by low doses of either single agent (i.e., sorafenib or crenolanib), or their combination could be abrogated by the presence of 25 ng/mL FLT3 ligand (Fig. 6A). However, apoptosis induction could be partly retained by increasing the drug concentrations to 0.01 μmol/L sorafenib or 0.1 μmol/L crenolanib for a 48-hour time period. Impressively, the combination of sorafenib and crenolanib at these concentrations demonstrated marked apoptosis induction even in the presence of FLT3 ligand at high levels (i.e., 50 ng/mL; Fig. 6B).

We further tested the effects of combining sorafenib with the MEK inhibitor CI-1040 and/or the mTOR inhibitor CCI-779. The results showed that cotargeting FLT3-ITD/MEK with sorafenib/CI-1040 had a significant synergistic proapoptotic effect in the Ba/F3-ITD and Ba/F3-ITD+842–resistant cells despite presence of FLT3 ligand. In turn, cotargeting FLT3-ITD/mTOR with sorafenib/CCI-779 showed only limited apoptosis induction in the cells in the presence of FLT3 ligand at 25 ng/mL concentration (Fig. 6C). Immunoblot analysis showed that high phosphorylation levels of FLT3, ERK, and Stat5, but not AKT, could be retained in the presence of FL; in turn, the combination of sorafenib with crenolanib, CI-1040 or CI-1040/CCI-779 profound suppressed phospho-FLT3, its downstream proteins, and induced caspase-3 cleavage after a 6-hour treatment (Fig. 6D). Our findings suggest that these combinations, using the type I/type II inhibitors or cotargeting FLT3 and MEK signaling, could effectively overcome FLT3 ligand–mediated resistance.

The efficacy of AML therapy with FLT3 inhibitors is limited by the development of resistance even with next-generation FLT3 inhibitors AC220 (11, 12, 37). Many point mutations located at either TKD1 (such as N676, F691, and G697) or TKD2 (such as D835 and Y842 in activation loop) have been reported after treatment with various FLT3 inhibitors (excluding sorafenib) in vitro and in vivo and were related to leukemic cell resistance (13, 20, 22, 32). Conversely, no point mutations were detected in 2 patients with AML who reportedly relapsed after an initial response to sorafenib therapy (24), and one study reported that D835Y mutations were detected in 4 of 7 patients with AML with FLT3-ITD mutations after sorafenib treatment (38). Although the mutation sites identified in this study differ from D835Y, they all are located in the FLT3 TKD and several mutations of this type have been observed in patients relapsed during treatment with other FLT3 inhibitors. This would imply instability of these TKDs that are invariably under “selective pressure” by FLT3 inhibition in leukemic cells. Alternatively, these point mutations are too rare to be detected before therapy, and after suppressing the FLT3-ITD allele by FLT3 inhibition, the resistant alleles could conceivably become dominant in the surviving AML clones.

Our engineered cell lines harboring point mutations of TKDs demonstrated varying degrees of resistance to sorafenib, suggesting that mutations of the TKDs are critical for sorafenib resistance. Furthermore, AML cells with TKD2 mutations (i.e., Y842C) showed more resistance to sorafenib compared with those with TKD1 mutations. Cells harboring compound TKD mutations demonstrated even greater resistance to sorafenib implying that conformational changes in the TKD2 might alter sorafenib's binding affinity more than those associated with TKD1, and that mutations of both TKDs might result in further anomalous conformational changes and increased resistance.

Sorafenib-resistant patients with acquired AML point mutations also had a longer length of ITD mutations than those without point mutations (Supplementary Table S1), suggesting instability of the ITD-mutated gene may be associated with the length of ITD mutations. Schnittger's group reported that the patients with longer FLT3-ITD mutations had a markedly shorter event-free survival (7.4 vs. 12.6 months; ref. 39). This would suggest the possibility that the patients with longer ITD mutations might easily acquire additional point mutations and become resistant to therapy. Because this investigation examined relatively few patient samples, further studies are required to validate this hypothesis.

It was highly likely that sorafenib-resistant cells may display resistance to other type II TKIs. Indeed, our results demonstrate that sorafenib-resistant cells have very similar resistance patterns compared with other type II inhibitors such as C220 and MLN518. However, they retain sensitivity to type I inhibitors such as PKC412 and crenolanib. Furthermore, a combination of 2 type I inhibitors (i.e., crenolanib and PKC412), but not of 2 type II inhibitors (i.e., sorafenib and MLN518) showed impressive synergistic proapoptotic effects in sorafenib-resistant cells Ba/F3-ITD+676/842. Because our in vitro screening for resistant mutations was performed in cells exposed to the type II inhibitor sorafenib, it is reasonable to assume that the mutation-mediated conformational change in FLT3 might impair the accessibility of other type II inhibitors, but retains binding ability with type I inhibitors. Albers and his colleagues also reported an patient with AML with acquired point mutation at F691L who developed resistance to type II kinase inhibitor AC220, but retained sensitivity to type I inhibitor PKC412 (32). This result supports our findings, that is, the cells resistant to type II kinase inhibitor may remain sensitive to type I kinase inhibitors. However, Weisberg and colleagues reported that the point mutations of D835Y (or N676K) which developed after therapy with type I inhibitor PKC412, retained sensitivity to highly selective second generation type II inhibitor NVP-AST487 (40), although these mutant cells exhibited resistance to less-specific type II inhibitor sorafenib (41). Altogether, these findings strongly suggest that resistance developed by one type kinase inhibitor could be overcome by another type of kinase inhibitor.

The combination of the type I inhibitor crenolanib with a type II inhibitor (i.e., either sorafenib or MLN518) triggered a profound proapoptotic effect in the sorafenib-resistant cell line Ba/F3-ITD+676/842, and showed synergistic/additive proapoptotic effects in relapsed AML patient samples with FLT3-ITD and/or D835 mutations, although some of these patients showed relapsed or refractory to other single TKI therapy (such as PKC412 or AC220). Mechanistically, this seemingly occurred by the suppression of phosphorylation levels of FLT3 and its downstream signaling intermediates. During the revision of this manuscript, Zimmerman and her colleagues also reported a synergistic antileukemia effect by combining type I inhibitor crenolanib with type II inhibitor sorafenib in their murine leukemia model, which was engrafted with FLT3-ITD–mutated cells MV4–11 (42). Taken together, these findings suggest that combination of type I and type II tyrosine kinase may provide a promising strategy for targeting relapsed/refractory patient with AML and also be beneficial for preventing resistance of FLT3-targeted therapy by reducing doses and enhancing therapeutic efficacy.

Our data also demonstrate that the suppression of FLT3 activation was not sufficient for inducing apoptosis of leukemia cells, at least in our sorafenib-resistant cells Ba/F3-ITD+842 and Ba/F3-ITD+676/842, which demonstrated a FLT3-independent high activation of ERK and/or AKT (Fig. 1C). This would suggest that these downstream proteins themselves could also be modulated by other signaling pathways and play a pivotal role in maintaining the leukemia cell survival. In fact, aberrant activation of ERK or AKT could be frequently observed in co-culture with bone marrow stromal cells (43) or in hypoxia environment in our resistant cells (Supplementary Fig. S8). Therefore, a strategy for cotargeting FLT3 and MEK/ERK or AKT/mTOR signaling pathways simultaneously could enhance antileukemic effects. However, the combination of sorafenib with MEK inhibitor CI-1040 showed more impressive antileukemic effect than that of sorafenib with mTOR antagonist CCI-779 in all tested sorafenib-resistant cells, even in the presence of high concentrations of FLT3 ligand, suggesting that concomitant inhibition of FLT3 and ERK activation might be a potent strategy for overcoming FLT3-inhibition resistance.

In summary, our findings provide an improved understanding of the development of resistance of FLT3-ITD–mutated AML cells. They also extend our knowledge of the role of acquired point mutations in TKDs that are associated with resistance to FLT3 inhibition. The use of combinatorial strategies in this scenario produced effective antileukemia effects in most types of sorafenib-resistant cells, which may benefit the design of future clinical trials for the therapy of AML.

J.E. Cortes has commercial research grants from Ambit, Arog, Astellas, and Novartis. F. Ravandi-Kashani has received a commercial research grant from Celgene, is a consultant/advisory board member for and has received speakers bureau honoraria from and Bayer/Onyx. No potential conflicts of interest were disclosed by the other authors.

Conception and design: W. Zhang, A. Ramachandran, M. Andreeff

Development of methodology: W. Zhang, Y. Chen, R.O. Jacamo, M. Andreeff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Zhang, C. Gao, M. Konopleva, Y. Chen, G. Borthakur, J.E. Cortes, F. Ravandi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Zhang, C. Gao, J.E. Cortes, A. Ramachandran

Writing, review, and/or revision of the manuscript: W. Zhang, M. Konopleva, G. Borthakur, J.E. Cortes, F. Ravandi, M. Andreeff

Study supervision: M. Andreeff

The authors thank W. Chen and S.V. Mathews who provided follow-up data for this study. S. Patterson and N. Hail, Jr. provided critical reviews and editorial assistance in preparation of this manuscript.

This work was supported in part by grants from NIH/NCI (CA143805, CA100632, CA055164, and CA049639) and a Cancer Center Support Grant (CA16672; to M. Andreeff), and a Leukemia SPORE Career Development Award (CA100632-05; to W. Zhang).

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.