A novel molecular mechanism of primary resistance to FLT3-kinase inhibitors in AML

In 30% to 40% of adult patients with acute myelogenous leukemia (AML), activating mutations in FMS-like tyrosine kinase 3 (FLT3) can be detected, and the majority of these mutations belong to the class of internal tandem duplications (ITDs). In AML, internal tandem duplication (ITD) mutations of FLT3 are associated with higher white blood cell counts, an increased relapse rate, and decreased overall survival in response to standard chemotherapy.^1-4  Therefore, treatment using FLT3 tyrosine kinase inhibitors (TKIs) is a novel and promising approach to overcome the dismal prognosis of FLT3_ITD-positive AML. Several new small-molecule TKIs targeting FLT3 kinase (eg, PKC412 [midostaurin], CEP-701, SU11248) are currently being investigated in phase 1/2 clinical trials.^5-7  These studies have shown thus far that monotherapy using FLT3 TKI may result in measurable clinical responses including significant reductions of peripheral blood (PB) and bone marrow (BM) blasts. However, in most cases these responses are transient and patients either become resistant to TKI treatment after a short period of response (secondary resistance) or show up-front resistance to FLT3 TKIs (primary resistance).^5-8  The kinase inhibitor PKC412 (midostaurin) is a derivative of the alkaloid staurosporine and is described as a potent inhibitor of mutant FLT3 receptors with an IC₅₀ below 10 nM.⁹  In a clinical phase 2 trial in AML patients using PKC412 monotherapy, of the 28 of 86 patients with FLT3 mutations identified in the original screening, 20 of 28 patients (71%) had a blast response, which was defined as a decrease of more than 50% in PB or BM blasts.¹⁰  Thus, approximately 30% of patients harboring FLT3 mutations may exhibit primary resistance to PKC412 treatment.^7,10 

Up until now, the etiology of primary and secondary resistance to FLT3 TKIs in AML has been poorly understood but is of major importance for development of future therapeutic strategies using these compounds. Acquired resistance mutations may prevent TKI binding to the FLT3 receptor as described in resistance to imatinib mesylate in BCR-ABL–positive chronic myelogenous leukemia.¹¹  Indeed, a limited number of potential resistance mutations surrounding the drug binding site of FLT3 have been predicted by random mutagenesis in vitro.¹²  In line with these results, we recently identified a mutation (N676K) in the FLT3 tyrosine kinase domain conferring clinical resistance to PKC412 in FLT3_ITD-positive AML.¹³  Apart from FLT3 overexpression leading to reduced efficacy of TKIs, activation of compensatory pathways, rendering cells independent of FLT3, has also been proposed as a possible scenario. This idea was further supported by observations that, despite inhibition of the FLT3 kinase by TKIs, downstream pathways remained activated and leukemic blasts proved to be resistant to FLT3 TKIs.^8,14,15 

In this report, we describe a novel molecular mechanism leading to primary clinical resistance to TKI therapy (PKC412) in AML. An atypical ITD mutation integrating in the first tyrosine kinase domain (TKD1) at amino acid position 627 (ITD627E), generating a nonjuxtamembrane (JM) ITD,¹⁶  conferred resistance to a panel of FLT3 TKIs in vitro and in vivo. Up-regulated expression of myeloid cell leukemia 1 protein (MCL-1), an anti-apoptotic member of the BCL-2 family, was shown to act as the major resistance mechanism increasing the threshold for induction of apoptosis in response to TKIs.

Heparin-treated PB samples (20 mL) were obtained from a patient enrolled in a phase 2 study investigating the efficacy and toxicity of PKC412 (PKC412A-2104 trial^7,10 ). The samples were obtained before the start of PKC412 treatment, at the time point of documented primary resistance (day 35), and 2 days after PKC412 was discontinued. Informed consent was obtained in accordance with the Declaration of Helsinki. Laboratory experiments on ITD variants isolated from patient material were performed with approval from the Ethikkommission Mainz institutional review board.

Mononuclear cells (MNCs) enriched in AML leukemic blasts were isolated as described.¹⁷  Genomic DNA from PB MNCs was extracted using the QIAamp DNA Blood Mini Kit (QIAGEN, Hilden, Germany). ITD mutation screening by polymerase chain reaction (PCR) and subcloning of PCR products into pCR4-TOPO vectors (Invitrogen, Groningen, The Netherlands) was performed as described.¹⁶ 

A human FLT3_ITD construct, subcloned into the pAL expression vector, was used and has been previously described.¹⁷  This ITD allele (36 bp/12 amino acids [aa]) integrates between codons 598 and 599 in the JM domain of FLT3. Subcloning of the ITD627E allele into the pAL vector was performed as described.¹⁶  The ITD627E allele (93 bp/31 aa) integrates at codon 627 in the β2 sheet of the TKD1 of FLT3 and leads to an amino acid exchange at codon 627 (alanine to glutamate).¹⁶  The FLT3_ITD627A mutant was generated by Medigenomix (Martinsried, Germany) by site-directed mutagenesis. This ITD mutant is identical in length and position of integration to ITD627E (93 bp/31 aa) with the amino acid at codon 627 reverted to wild type (alanine).¹⁶  All vector constructs were verified by nucleotide sequencing. The amino acid sequences of the ITD alleles are shown in Figure S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article).

Transfection of 32D cells with different FLT3 DNA constructs was performed as described¹³  and polyclonal cell lines were used for further experiments.

All mock-transfected control cells died upon withdrawal of IL-3 from the culture medium. Clonal cell lines from early passages of polyclonal cell lines were generated by limiting dilution in 96-well plates.

32D_ITD mMCL-1 cells, ectopically overexpressing MCL-1, were generated by transfection of 32D_ITD cells with 20 μg p3 × FLAG-CMV10 vector containing the coding sequence of murine MCL-1 (kindly provided by Dr H. Schulze-Bergkamen, University of Mainz, Mainz, Germany) by electroporation. Cells were selected with 0.5 mg/mL G418 and polyclonal cell lines were used for experiments.

Isolated AML leukemic blasts and/or 32D transfectants were treated with different concentrations of PKC412 (kindly provided by Novartis, Basel, Switzerland), or U0126 (Cell Signaling Technology, Frankfurt, Germany).

For Western blot analysis,¹⁷  the following antibodies were used: anti–phospho-FLT3 (Y591), anti–phospho-AKT (S473), anti–phospho–extracellular signal–related kinase 1/2 (ERK1/2; T202/Y204), AKT, ERK1/2, anti–phospho-S6 protein (S240/S244), S6 protein, and anti–phospho-STAT3 (S727 and Y705), all from Cell Signaling Technology; FLT3, STAT-5, MCL-1, GRB-2, and STAT3 (Santa Cruz, Heidelberg, Germany); anti–phosphotyrosine (4G10) and anti–phospho-STAT5 (Y694/Y699), from Upstate, Lake Placid, NY; anti–BCL-X_L (BD Biosciences, Heidelberg, Germany); actin (MP Biomedicals, Aurora, OH), GAPDH (Biodesign International, Saco, ME), and tubulin (Sigma-Aldrich, Munich, Germany). The FLT3 phosphoepitope-specific antibodies FLT3pY589, FLT3pY599, FLT3pY768, FLT3pY955, and FLT3pY969 were generated in the laboratory of Dr Lars Rönnstrand (Experimental Clinical Chemistry, Department of Laboratory Medicine, Lund University, Malmö University Hospital, Malmö, Sweden). Immunoprecipitation of FLT3 was performed as described with modifications.¹⁷ 

The percentage of apoptotic cells was determined by measuring the sub-G₁ fraction upon propidium iodide (PI) incorporation using flow cytometry as described.^13,18  Further, cellular apoptosis was measured by annexin V labeling according to the manufacturer's protocol (BD Biosciences). Briefly, 2 × 10⁵ cells/mL were treated with different concentrations of PKC412 for 24 hours, and early and late cellular apoptosis was determined by staining with annexin V–FITC antibody/PI and measuring flow cytometry.

Intact mitochondrial membrane potential was visualized by staining with tetramethylrhodamine ether (TMRE) upon incubation of cells (2.5 × 10⁴ cells/mL) with 50 nM TMRE at 37°C for 30 minutes, followed by flow cytometry.¹⁹ 

32D_ITD627E cells (2 × 10⁶) were transfected by electroporation with h-FLT3 (siGENOME SMARTpool, M-003137-01; Dharmacon, Lafayette, CO), m-MCL-1 (sc-35 878), m-STAT3 (sc-29 494), m-BCL-2 (sc-29 215) or m-GRB-2 (sc-29 335; all from Santa Cruz) specific siRNAs. siRNAs were dissolved to a final concentration of 20 μM stock solution and for each transfection 10 μL siRNA stock solution was used. As a negative control, an equivalent concentration of AllStars Negative Control siRNA (QIAGEN) was used. For assessment of apoptosis, cells were left to recover from the transfection for 6 hours and then treated with or without PKC412. Percentage of apoptotic cells was assessed after 24- and 48-hour incubation as described in “Apoptosis and viability assays.”

Cellular protein lysates of FLT3_ITD and FLT3_ITD627E 32D cells were prepared as described in “Protein extract preparation, immunoprecipitation, and Western blot analysis” using lysis buffer, supplemented with protease and phosphatase inhibitors, supplied by Kinexus Bioinformatics (Vancouver, BC). Dye labeling, hybridization, and analysis of the Kinex antibody array were done by Kinexus.

32D cells expressing either FLT3_ITD or FLT3_ITD627E were extracted with lysis buffer. Anti-FLT3 antibody (10 μg; C-20, Santa Cruz) was added to each extract and incubated at 4°C overnight. A mixture of 30 μL protein-G sepharose (Thermo Fisher Scientific, Rockford, IL) and 120 μL Sepharose CL4B (Sigma-Aldrich)—each 1:1 suspension with lysis buffer—were added. The beads were sedimented and washed 3 times. Then, the suspension was divided into 8 aliquots, and the beads sedimented again and washed once with kinase buffer (20 mM HEPES [pH 7.5], 5 mM MnCl₂, 0.1 mM sodium orthovanadate). The liquid was removed carefully, 25 μL kinase buffer and PKC412 in different concentrations were added (final DMSO concentration 1%), and the samples incubated on ice for 15 minutes. The kinase reaction was performed by adding 3 to 5 μCi [³²Pγ] ATP per reaction followed by incubation at 30°C for 30 minutes. The reactions were stopped by adding sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer, and heating at 95°C for 5 minutes. SDS-PAGE was performed, and gels were subjected to autoradiography and quantitated using a GS250 Molecular Imager (Bio-Rad, Hercules, CA).

Clinical material from an AML patient treated with PKC412 within a phase 2 clinical trial¹⁰  was subjected to molecular analysis. Patient characteristics, schedule and dose of PKC412, and hematologic responses have been described previously.^7,10  The patient had not experienced a blast response (ie, > 50% decrease in leukemic blasts) (Figure 1A) but showed primary resistance to PKC412 (200 mg/day).¹⁰  RNA and DNA from leukemic blasts were isolated before the start of PKC412 therapy and at the time point of documented resistance (day 35; Figure 1A). At baseline, FLT3 mutation analysis revealed 2 different ITD alleles (ITD1 and ITD2; Figure 1B), while at the time point of resistance, only 1 ITD allele was detectable (ITD1; Figure 1B). ITD1 (31 amino acids) integrated at nucleotide 1880 of FLT3, thereby generating a single nucleotide change from GCA to GAA.¹⁶  This resulted in an amino acid substitution from alanine to glutamate at amino acid position 627 (ITD627E). An additional mutation in FLT3_ITD627E was excluded. Thus, FLT3_ITD627E represents an atypical ITD of FLT3 not integrating into the JM domain (amino acids 572-603) but into the β2-sheet (amino acids 624-630) of TKD1 of FLT3.¹⁶ 

Because the ITD627E allele was selected in vivo during PKC412 treatment, we next assessed the status of FLT3_ITD627E signaling ex vivo. Primary AML blasts from the patient were isolated at the time point of resistance. FLT3 protein was found to be present and phosphorylated in untreated cells (Figure 1C). Upon incubation with PKC412 at doses of 10 nM and 100 nM, inhibition of FLT3 phosphorylation was 64% and 91%, respectively, as evaluated by densitometric scanning (Figure 1C). No residual tyrosine phosphorylation of STAT5 was detectable when cells were treated with 10 nM or 100 nM PKC412 (Figure 1C). The degree of PKC412-induced inhibition of FLT3 phosphorylation illustrated in Figure 1C was comparable to the inhibition observed in patients responding to PKC412 (Heidel et al¹³  and data not shown). Trough levels of PKC412 and its active metabolite, CGP62221, were obtained at the time point of resistance (day 35). The actual sum of the levels of PKC412 and CGP62221, corrected for 99% protein binding,⁷  was similar (37 nM) to trough levels observed in other FLT3_ITD-positive AML patients at the time point of complete blast clearance.¹³  Together, these data suggest adequate inhibition of FLT3_ITD627E receptors in vivo at the time of primary resistance.

To assess the biological properties of the FLT3_ITD627E allele in vitro, stably transfected hematopoietic 32D cells expressing the mutant receptor were treated with PKC412, and the percentage of apoptotic cells was determined. Sensitivity of 32D_ITD627E cells was assessed in comparison to 32D cells expressing a standard FLT3_ITD integrating in the JM domain between codons 598 and 599.¹⁷  In these experiments, 32D_ITD627E proved to exhibit resistance to PKC412 treatment up to concentrations of 130 nM (Figure 2A). As a control, annexin V/PI assays were also performed with comparable results (data not shown). Analogous results were obtained when TMRE staining and flow cytometry was applied (Figure 2B). This assay visualizes intact mitochondrial outer membrane potential, and its loss is considered a hallmark for apoptosis.¹⁹  A newly transfected polyclonal cell line harboring the ITD627E allele (32D_ITD627Ep4) and 2 independent clonal cell lines established from early passages of the first polyclonal 32D_ITD627E cell line (32D_ITD627E c1 and 32D_ITD627E c2) showed a similar PKC412-resistant phenotype (Figure S2A,B).

Next, we addressed the question whether resistance to PKC412 is solely dependent on the ITD integration site at this particular position or whether the amino acid exchange from alanine to glutamate induced by ITD integration at position 627 is involved. To test this hypothesis we employed 32D cells stably transfected with a mutant in which the amino acid exchange at position 627 has been reverted to alanine by site directed mutagenesis (32D_ITD627A). 32D_ITD627A cells, 32D_ITD cells, and 32D_ITD627E cells were treated in parallel with PKC412. Again, 32D_ITD627E proved to be resistant to PKC412 treatment, whereas 32D_ITD627A cells, like 32D_ITD, were sensitive to PKC412 (Figure 2C). This result indicates that the point mutation A627E is essential for the resistance phenotype of FLT3_ITD627E.

To determine sensitivity to other FLT3 TKIs, cells were treated with various concentrations of K252a, a compound that is structurally similar to CEP-701^12,20  and with SU5614, a structurally more distant tyrosine kinase inhibitor.²¹  These experiments demonstrated cross-resistance of 32D_ITD627E cells to these compounds as compared with 32D cells expressing the standard JM-ITD (32D_ITD; Figure 2D).

To assess the effects of PKC412 on FLT3 signal transduction, 32D_ITD and 32D_ITD627E cells were exposed to a range of PKC412 concentrations and the activation status of key signaling molecules was determined by Western blotting. Surprisingly, we detected comparable dephosphorylation of the FLT3 receptor at tyrosine 591 (Y591) in both cell lines (Figure 3A). FLT3-Y591 has been shown to be a major phosphorylation site of activated FLT3 kinase and indicates activated downstream signal transduction.²²  Furthermore, we found activation of key signaling nodes as STAT5, AKT, and S6 protein were equally down-regulated by PKC412. Interestingly, this was not the case for ERK1/2 phosphorylated at T202 and Y204, which was up-regulated and remained activated/phosphorylated in 32D_ITD627E cells, as compared with 32D_ITD cells, despite full suppression of FLT3 receptor phosphorylation at Y591 (Figure 3A).

As a next step, immunoprecipitation of the FLT3 receptor followed by immunoblotting was performed using anti-phosphotyrosine–specific antibodies. Consistent with the results described in Figure 3A, FLT3 was found equally dephosphorylated in 32D_ITD and in 32D_ITD627E cells upon treatment with PKC412 (Figure 3B). To directly determine the kinase activity of FLT3_ITD and FLT3_ITD627E receptors, in vitro autophosphorylation kinase assays were performed. Figure 3C illustrates equal inhibition of kinase activity from both FLT3 receptors upon treatment with PKC412. To analyze the phosphorylation status of the mutant FLT3 receptors in detail, we used a battery of antibodies specific for 6 different phosphorylation sites of FLT3. Again, we detected equal dephosphorylation at all the different phosphorylation sites upon PKC412 treatment (Figure 3D).

To rule out that ERK1/2 phosphorylation and IL-3–independent growth of 32D_ITD627E cells had become independent from FLT3 signaling by acquisition of cryptic genetic events, we used RNA interference (RNAi) technology. Down-regulation of FLT3_ITD627E by transient transfection of FLT3-specific siRNA led to substantial inhibition of ERK1/2 phosphorylation (31% of control; Figure 3E left panel) and induced apoptosis in a major fraction of 32D_ITD627E cells (Figure 3E right panel). These results show that signaling from the mutant FLT3_ITD627E receptor is essential for ERK1/2 phosphorylation and for IL-3–independent survival of 32D_ITD627E cells.

To assess the functional contribution of ERK1/2 activation to the FLT3-TKI–resistant phenotype, we applied the highly selective MEK/ERK inhibitor U0126.²³  In these experiments, U0126 completely suppressed ERK1/2 activation in 32D_ITD and in 32D_ITD627E cells (Figure 4A). Next we tested whether ERK1/2 inhibition restores PKC412 sensitivity in 32D_ITD627E cells. In comparison to 32D_ITD cells, a small but consistently detectable increase in apoptosis was observed when 32D_ITD627E cells were treated with a combination of PKC412 and the ERK1/2 inhibitor U0126 for 48 hours (Figure 4B). This suggests that ERK activation may make a minor contribution to the TKI-resistance phenotype of FLT3_ITD627E but does not play an essential role.

To screen 32D_ITD627E cells for differentially activated signaling pathways, we performed protein expression and phosphorylation profiling. The Kinex antibody microarray allows comparison of protein expression and the phosphorylation status of 615 different proteins. Using this platform, we could identify 62 differentially expressed/phosphorylated proteins in protein lysates from FLT3_ITD627E cells compared with FLT3_ITD cells (Figure 5A and Table 1). A major proportion of these proteins are known to be involved in signal transduction. Our next step was to validate these findings and determine the functional role of these proteins. Table 1 shows that serine phosphorylation of STAT3 at position S727 was found to be up-regulated by 50% in comparison to the control. Immunoblot analysis indeed demonstrated an increased level of STAT3 S727 phosphorylation as compared with that in 32D_ITD cells (Figure 5B top panel). Full activation of STAT3 is known to require serine phosphorylation at position 727 in addition to tyrosine phosphorylation.²⁴  RNA interference experiments using FLT3-specific siRNA confirmed that the FLT3_ITD627E receptor is essential for up-regulation of serine phosphorylation of STAT3 at position 727 (Figure 5B bottom panel). To determine whether increased P-S727-STAT3 signaling underlies the resistance phenotype of 32D_ITD627E cells, STAT3 protein expression was down-regulated by STAT3-specific siRNA and sensitivity to PKC412 was measured. Figure 5C shows that suppression of STAT3 restored sensitivity to PKC412 in a minor proportion of cells. This result suggests that signaling via STAT3 contributes to the resistance phenotype of 32D_ITD627E cells but does not appear to play a major role. This result is also in line with the observation that phosphorylation of STAT3 at S727 and also Y705 was down-regulated upon PKC412 treatment (Figure S4A,B).

Strikingly, by antibody microarray analysis, MCL-1 was found to be the most up-regulated protein in 32D_ITD627E cells (+675%; Table 1). Western blotting confirmed this finding (Figure 5D top panel and Figure S2A,B). Interestingly, in 32D_ITD627E cells, MCL-1 expression was unchanged upon PKC412-treatment, while in 32D_ITD cells PKC412 treatment resulted in a marked decrease of MCL-1 expression with regard to the higher molecular weight subspecies (Figure 5D top panel). Appearance of MCL-1 as a doublet in SDS-PAGE analysis has been described and has been attributed to differentially phosphorylated isoforms.²⁵ 

Confirmation that MCL-1 expression is indeed regulated by the FLT3_ITD627E receptor was achieved by siRNA experiments. Figure 5D (bottom panel) shows that knockdown of FLT3_ITD627E using FLT3-specific siRNA resulted in significant suppression of MCL-1 protein levels. However, FLT3_ITD627E regulated MCL-1 expression does not appear to involve the ERK1/2 or STAT3 pathways because neither suppression of phosphorylated ERK1/2 by the selective MEK/ERK inhibitor U0126 nor suppression of STAT3 by RNA interference resulted in decreased MCL-1 protein levels (data not shown). Interestingly, stable MCL-1 protein expression, despite FLT3 receptor dephosphorylation, could also be demonstrated in primary AML blasts isolated from the patient and treated ex vivo with PKC412 (Figure S3B).

Finally, we tested whether MCL-1 up-regulation contributes to the resistance phenotype of 32D_ITD627E cells. By decreasing MCL-1 protein levels using MCL-1–specific siRNA, the majority of 32D_ITD627E cells proved to restore sensitivity to PKC412 (10 nM) treatment (Figure 5E right panel). Induction of apoptosis was further enhanced by increasing the dose of PKC412 (20 nM), indicating rescue of dose-dependent sensitivity to PKC412. From these data it appears that deregulated MCL-1 expression acts a major resistance factor in 32D_ITD627E cells. To exclude off-target effects on related anti-apoptotic proteins, we controlled for expression of BCL-2 and BCL-X_L protein in MCL-1–siRNA transfected cells and found both protein levels to be unchanged (Figure 5E left panel). As differences in siRNA knockdown efficiencies in favor of MCL-1 might account for the pronounced effect observed upon MCL-1 down-regulation but not STAT3 knockdown, we determined the level of suppression using densitometric scanning. Mean suppression levels of STAT3 were consistently higher (40.4%) than those of MCL-1 (27.5%; data not shown).

The unique role of MCL-1 in mediating PKC412 resistance is further supported by the fact that down-regulation of the anti-apoptotic BCL-2 protein only had a minor effect on PKC412 resistance. Transient transfection of 32D_ITD627E cells with BCL-2–specific siRNA caused a moderate increase in apoptotic cell death (Figure 5F), comparable to that seen in STAT3 siRNA knockdown experiments (Figure 5C). Finally, ectopic overexpression of murine MCL-1 in 32D_ITD cells alone was sufficient to confer a PKC412-resistant phenotype, similar to that observed in 32D_ITD 627E cells (Figure 5G). Expression levels of other anti-apoptotic proteins, like BCL-2 and BCL-X_L, were unaffected by overexpression of murine MCL-1 in 32D_ITD cells (Figure S4C).

In contrast to 32D_ITD627E cells, only minimal STAT3 S727 phosphorylation and MCL-1 expression was detected in 32D_JM_ITD cells (Figure 5B top panel and 5D, respectively). In addition, MCL-1 expression is clearly dependent on FLT3-autophosphorylation in 32D_JM_ITD cells, as shown in Figure 5D. To further explore these differences in signal transduction, we transfected 32D_JM_ITD cells with siRNA targeting either STAT3 or MCL-1 and treated the cells with PKC412. Whereas knockdown of MCL-1 causes increased sensitivity to PKC412 treatment, similar to FLT3_ITD627E-expressing cells, the same effect was observed upon suppression of STAT3 (Figure S3A) and GRB-2 (data not shown). Surprisingly, no effect on MCL-1 expression was seen upon knockdown of FLT3 using specific siRNA (Figure S3C). However, as FLT3 knockdown was not perfect, residual FLT3 receptors are likely still constitutively phosphorylated and mediate persistent MCL-1 expression.

We next aimed to identify a signaling intermediate which persisted in an active state despite PKC412-induced dephosphorylation of the FLT3_ITD627E receptor and which may represent a link to MCL-1 overexpression. GRB-2 appeared to be an interesting candidate since it is known to participate in FLT3 signaling and in regulation of the MEK/ERK pathway.^26,27  To determine binding of GRB-2 to the FLT3-receptor, we treated 32D_ITD, 32D_ITD627E, and 32D_ITD627A cells, with PKC412 (10 nM), immunoprecipitated FLT3 from cellular lysates, and analyzed co-immunoprecipitated GRB-2 by Western blotting. In untreated cells, significantly higher levels of GRB-2 protein were bound to the FLT3_ITD627E receptor than to the FLT3_ITD and FLT3_ITD627A receptors (Figure 6A top panel). This difference was not due to up-regulation of total GRB-2 protein expression, as equal amounts of GRB-2 protein were detected in all cell lysates (Figure 6A bottom panel). Therefore, the interaction of GRB-2 with the FLT3 receptor was disrupted in PKC412-sensitive 32D_ITD and 32D_ITD627A cells, but persisted in PKC412-resistant 32D_FLT3_ITD627E cells (Figure 6A). Furthermore, sustained association of GRB-2 with the FLT3_ITD627E receptor was still detected in cells treated with PKC412 concentrations as high as 50 nM (Figure S3D). Binding of GRB-2 to the FLT3 receptor has been described to serve as a platform for activation of a variety of signaling pathways. Thus, it is conceivable that persistent binding of GRB-2 to the FLT3_ITD627E receptor accounts for tyrosine phosphorylation-independent activation of a signaling node involved in up-regulation of MCL-1.

The functional relationship between GRB-2, MCL-1, and the PKC412-resistance phenotype was assessed by down-regulation of GRB-2 using GRB-2–specific siRNA. Figure 6B illustrates that knockdown of GRB-2 results in decreased MCL-1 expression. When GRB-2 siRNA-transfected cells were treated with PKC412, sensitivity to PKC412 was restored in the majority of previously resistant 32D_ITD627E cells (Figure 6B). These data are consistent with results presented (Figure 5E) and together suggest that in 32D_ITD627E cells, persistent binding of GRB-2 mediates MCL-1 overexpression and resistance to FLT3 TKIs.

In this report, we describe a novel mechanism of resistance to FLT3 TKIs. In clinical material derived from an AML patient showing primary resistance to PKC412, we identified a unique non-JM ITD FLT3 receptor featuring ITD integration in the TKD1 at amino acid position 627 (FLT3_ITD627E). While 2 different FLT3_ITD alleles were present before the start of PKC412, reevaluation at the time of clinical resistance showed clonal evolution of FLT3_ITD627E-positive blasts only. Thus, it appears that upon start of PKC412 treatment, rapid selection of a drug-resistant leukemic clone harboring FLT3_ITD627E had occurred. This hypothesis was confirmed by testing FLT3_ITD627E in a reconstitution model. These experiments demonstrated that FLT3_ITD627E is sufficient to confer resistance to a panel of FLT3 TKIs in vitro. To systematically examine the molecular basis of TKI resistance, an antibody microarray screen was applied. This screen identified the anti-apoptotic protein MCL-1²⁸  to be dramatically up-regulated in FLT3_ITD627E cells. Functional validation using RNA interference showed that MCL-1 up-regulation was strictly dependent on expression of FLT3_ITD627E receptors and that suppression of MCL-1 levels rescued sensitivity to PKC412. The anti-apoptotic effects of MCL-1 up-regulation in FLT3_ITD627E cells are currently unknown and are subjects for further studies. However, MCL-1 may prevent cytochrome c–release from mitochondria by blocking pro-apoptotic members of the BCL-2 protein family, thereby impeding their activation and mitochondrial outer membrane permeabilization (MOMP). Our data using TMRE staining are consistent with the concept that up-regulated MCL-1 expression protected against PKC412-induced MOMP. In line with these results, elevated MCL-1 expression has been identified in various human cancers and is associated with poor prognosis and drug resistance.^29-33 

Kinase measurements demonstrated that FLT3 tyrosine kinase activity of both the FLT3_ITD627E receptor and the FLT3_JM_ITD receptor was equally inhibited upon incubation with PKC412. FLT3 tyrosine phosphorylation was investigated applying 4 different approaches and, as compared with 32D_JM_ITD, no significant differences were detected upon PKC412 treatment. Accordingly, with respect to the mechanism, this result shows that resistance to PKC412 is not associated with failure to effectively inhibit FLT3 kinase activity. Interestingly, this is consistent with results obtained in clinical material: Analysis of FLT3 signaling in primary AML blasts isolated at the time point of resistance exhibited effective inhibition of FLT3 kinase upon ex vivo treatment with PKC412. These results point to a FLT3-independent mode of resistance. However, our data show that downstream signaling nodes of FLT3_ITD627E are not uniformly inhibited by PKC412. Using a panel of Western blot analyses, phosphorylation of ERK1/2 and expression of MCL-1 protein proved to be independent of PKC412-induced dephosphorylation of the FLT3_ITD627E receptor. To investigate this discrepancy, we studied the interaction of the FLT3_ITD627E receptor with the adaptor GRB-2, situated upstream of MEK/ERK and MCL-1, with surprising results. Although PKC412 induced dephosphorylation of the FLT3_ITD627E receptor, GRB-2 binding was maintained at high levels. Phosphorylation-independent binding of GRB-2 to FLT3_ITD627E may be accomplished by the N-terminal SH3 domain or the C-terminal SH3 domain of GRB-2. A possible alternative mechanism would be phosphorylation-independent binding of GRB-2 via the SH2 domain. This hypothesis is supported by published data demonstrating that phosphorylation is required for binding of most SH2 domains; however, SH2 domains, including GRB-2 SH2 domains, may bind to their peptide recognition domains in the absence of phosphorylation, albeit with lower affinity.^34-37  Interestingly, the analysis of single FLT3 phosphoepitopes (Figure 3D) included 3 potential GRB-2 binding sites (Y955, Y969, and Y768), also suggesting a phosphorylation-independent association of FLT3_ITD627E with GRB-2.

Although the precise mechanism involved in GRB-2/FLT3_ITD627E interaction is currently unknown, and residual phosphorylation at yet undetected amino acid residues cannot be completely ruled out, our data are consistent with the concept that a neo–binding motif is generated by integration of ITD627E. This neo-motif appears to critically depend on the amino acid exchange of alanine to glutamate at position 627, because reverting glutamate to alanine abrogated both increased and phosphorylation-independent binding of GRB-2. It also seems to be important that this amino acid exchange was caused by the ITD integration at position 627 in the TKD1 of FLT3. We have recently reported that 28.7% of all FLT3_ITD-positive patients harbor an integration in the TKD1 of FLT3 and not the JM domain of the receptor.¹⁶  It is tempting to speculate that amino acid exchanges induced by ITD integrations in this highly conserved region are more likely to have dramatic consequences for the overall structure of the FLT3 receptor than integrations in the JM domain.

Comparable FLT3 tyrosine kinase activity was noted in extracts from ITD627E and JM_ITD cells. Accordingly, as to the mechanism of MCL-1 up-regulation, a difference in kinase activity is unlikely. However, increased association of GRB-2 with the FLT3_ITD627E receptor has been detected, and GRB-2 was shown to be essential for MCL-1 expression and for the TKI-resistance phenotype. Thus, we propose the following molecular mechanism: (1) generation/exposure of a neo-epitope by integration of ITD627E into the FLT3 receptor serves as a platform for enhanced binding of GRB-2 to the receptor; (2) signals originating from differential binding of GRB-2 promote up-regulation of MCL-1 expression; and (3) GRB-2 binding to the FLT3_ITD627E receptor and MCL-1 expression are sustained upon inhibition of the FLT3-kinase by TKIs and this results in an increased threshold for induction of apoptosis in response to FLT3 TKIs.

In addition to MCL-1, a number of signaling molecules, such as ERK1/2 and STAT3, were also shown to be deregulated in FLT3_ITD627E cells. However, functional analysis employing selective inhibitors and RNA interference indicated that ERK1/2 and STAT3 may play a minor role in TKI resistance and suggests that up-regulation of MCL-1 accounts for most of the FLT3_ITD627E-induced resistance phenotype.

Interestingly, in primary AML blasts from FLT3 TKI resistant cases, it has been observed previously that mitogen-activated protein (MAP) kinase activation may persist despite complete inhibition of FLT3.³⁸  This has been attributed to FLT3 independent survival pathways.^15,38  However, in our model, sustained activation of ERK1/2 was shown to strictly depend on FLT3 signaling, as revealed by RNA interference assays. Thus, our data suggest that in TKI-resistant primary AML blasts, both FLT3-independent and FLT3-dependent mechanisms may underlie persistent MAP kinase activation.

In conclusion, in this report we have uncovered the functional role of an atypical ITD integrating into the β2-sheet of FLT3 for drug resistance to TKIs in vivo and in vitro. The data presented suggest that particular ITDs of FLT3 may be associated with rewired signaling and differential responsiveness to TKIs. We believe it will be interesting to relate this information to clinical trials employing FLT3 TKIs.

We thank J. Roesel (Novartis Pharma, Basel, Switzerland) for the provision of PKC412, and M.H. Thiede (THANARES Analytik und Research GmbH, Dresden, Germany) for PKC412 dose level analysis in patient material.

We are thankful to M. Schuler for his support during the process of this work.

This work was supported by grants from the Deutsche Krebshilfe Foundation (Bonn, Germany, nos. 108218 and 108702) and the Wilhelm Sander-Stiftung (Munich, Germany) to T.F.

Contribution: F.B. designed experiments, performed research, analyzed data, and participated in writing the manuscript; B.M. designed experiments, performed research, and analyzed data; S.K. performed experiments and analyzed data; B.C. and T.S. performed research; F.D.B. designed experiments, performed research, and analyzed data; K.M. and L.R. performed research and analyzed data; C.H. contributed to the interpretation of experimental data; T.K. analyzed data and wrote the manuscript; and T.F. designed and supervised experimental work, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: T.F. received honoraria from Novartis. The remaining authors declare no competing financial interests.

The current address for Dr Breitenbuecher, Dr Markova, and Dr Kasper is Department of Medicine (Cancer Research), West German Cancer Center, University Hospital Essen, Essen, Germany. The current address for Dr Fischer is Department of Hematology/Oncology, Medical Center, Otto-von-Guericke University Magdeburg, Magdeburg, Germany.

Correspondence: Thomas Fischer, Department of Hematology/Oncology, Medical Center, Otto-von-Guericke University Magdeburg, Leipzigerstr 44, 39120 Magdeburg, Germany; e-mail: thomas.fischer@med.ovgu.de.