Adaptation to TKI treatment reactivates ERK signaling in tyrosine kinase-driven leukemias and other malignancies

Abstract

FLT3 tyrosine kinase inhibitors (TKI) have been tested extensively to limited benefit in acute myeloid leukemia. We hypothesized that FLT3/ITD leukemia cells exhibit mechanisms of intrinsic signaling adaptation to TKI treatment that are associated with an incomplete response. Here we identified reactivation of ERK signaling within hours following treatment of FLT3/ITD AML cells with selective inhibitors of FLT3. When these cells were treated with inhibitors of both FLT3 and MEK in combination, ERK reactivation was abrogated and anti-leukemia effects were more pronounced compared to either drug alone. ERK reactivation was also observed following inhibition of other tyrosine kinase-driven cancer cells, including EGFR-mutant lung cancer, HER2-amplified breast cancer and BCR-ABL leukemia. These studies reveal an adaptive feedback mechanism in tyrosine kinase-driven cancers associated with reactivation of ERK signaling in response to targeted inhibition.

Introduction

FMS-like tyrosine kinase-3 (FLT3) is one of the most commonly mutated genes in acute myeloid leukemia (AML). The most frequently observed genetic alteration, the internal tandem duplication (FLT3/ITD), occurs in approximately 23% of patients with AML and is associated with an inferior prognosis (1). FLT3/ITD is an established driver mutation in AML (2) and results in constitutive dimerization and activation of the receptor, thereby activating downstream signaling pathways including STAT5, PI3K/AKT, and RAF/MEK/ERK (3, 4).

FLT3 tyrosine kinase inhibitors (TKIs), including first generation (sorafenib, midostaurin) and second generation (quizartinib, crenolanib) inhibitors, are being actively pursued as a promising therapeutic strategy in patients with FLT3/ITD AML. However, the extent of clinical responses to these therapies has remained limited and is usually transient (5–7). Factors contributing to the limited efficacy of these drugs include acquired resistance mutations in the FLT3 kinase domain (2), acquired resistance via activation of parallel pathways (8), upregulation of FLT3 ligand (9), and bone marrow stromal cell-mediated activation of ERK signaling (10).

In some cancers driven by mutated oncogenes, it has been shown that small molecule inhibition of the target results in relief of feedback inhibition and resultant reactivation of signaling pathways. This response results in attenuation of the antitumor effects of targeted therapy and can be abrogated with combinatorial treatment strategies. This so-called “adaptive resistance” has been described in the context of PI3K/AKT/mTOR (11–13) and BRAFV600E (14) inhibition. However, it remains unclear whether this phenomenon extends to cancers driven by receptor tyrosine kinase (RTK) activation, such as FLT3/ITD AML.

Here we describe, in FLT3/ITD leukemia cells, intrinsic mechanisms of signaling adaptation in response to FLT3 inhibitor treatment, which result in reactivation of ERK signaling following maximal inhibition. The addition of submaximal concentrations of a MEK inhibitor abrogates this ERK reactivation and sensitizes cells to FLT3 inhibitor treatment, resulting in a synergistic combination. A similar adaptive response also occurs in response to TKI treatment in the context of BCR-ABL, EGFR-mutant, HER2-amplified cancer. These findings suggest that the addition of low-dose MEK inhibitor to standard TKI treatment may improve outcomes for patients with FLT3/ITD AML, and possibly other tyrosine kinase-driven malignancies.

Materials and Methods

Cell Lines and Reagents

MV4;11, SKBR3, K562, and HL60 cells were purchased from ATCC. Molm14 cells were obtained from the DSMZ. PC-9 cells were obtained from C. Hann (Johns Hopkins University). Parental cell lines were purchased within the past 10 years and authenticated by ATCC or DSMZ prior to purchase by the short tandem repeat method. Cell lines were not authenticated after purchase. 32D/ITD cells were generated as previously described (15). All cells except SKBR3 were maintained in RPMI 1640 supplemented with 10% FBS. SKBR3 cells were maintained in DMEM/F12 medium supplemented with the same. All cells were tested for mycoplasma contamination with the SouthernBiotech Mycoplasma Detection Kit less than 3 years prior to use. Sorafenib, quizartinib, and lestaurtinib, and crenolanib were obtained from LC Labs. PD0325901, trametinib, imatinib, and lapatinib were obtained from Selleckchem. TTT-3002 was a generous gift of H. Roder (Tautatis Inc.) Drugs for in vitro studies were dissolved in DMSO to yield a 10mM or 1mM stock solution and stored at −80°C.

Immunoblot Analysis and Ras-GTP Assay

Cellular lysis was performed using Cell Lysis Buffer (Cell Signaling Technology) following the manufacturer’s protocol. Lysates were quantified by BCA assay (Pierce) and equal amounts of lysates were resolved by SDS-PAGE, transferred to a nitrocellulose membrane, and proteins probed with antibodies against the protein of interest. Antibodies against FLT3 were obtained from Santa Cruz. All other antibodies were obtained from Cell Signaling Technology. After incubation with horseradish peroxidase-conjugated secondary antibodies, proteins were detected using a Chemidoc Touch Imaging System (Bio-Rad). GTP-bound Ras was measured using the Active Ras Detection Kit (Cell Signaling Technology) following the manufacturer’s instructions. Briefly, GTP-bound Ras was captured using a GST tagged Raf1-RBD peptide, followed by immunoblot for total Ras.

Flow Cytometry Analysis

Flow cytometry analysis was performed using a BD FACS-Calibur machine (BD Biosciences). Apoptosis was measured by incubating treated cells with Annexin V-APC antibody for 10 minutes according to the manufacturer’s instruction (BD Biosciences). To measure leukemia burden, peripheral blood and bone marrow were isolated and, following red blood cell lysis, stained with anti-mouse CD45 and anti-human CD45 for 30 minutes. All data were analyzed by FlowJo analysis software (Tree Star).

Proliferation Analysis

Cells were plated in 96-well plates at a density of 3,000–4,000 cells per well. Cells were analyzed 96 hours after drug treatment using Cell Proliferation Kit I (MTT, Roche) or Cell Counting Kit - 8 (Sigma-Aldrich) following the manufacturer’s instructions and plates were read using an absorbance spectrophotometer.

Animal Studies

Eight to ten-week-old NOD/SCID Gamma (NSG) female mice were obtained from the Johns Hopkins Animal Core Facility. Experiments were carried out under an IACUC-approved protocol and institutional guidelines for the proper and humane use of animals in research were followed. For transplant studies, 0.5 million Molm14 cells were injected into each mouse via tail vein injection. Treatment was initiated three days after injection. Mice were randomized to receive DMSO, 8mg/kg sorafenib, 5mg/kg PD0325901, or the combination once daily for three weeks. All drugs were formulated in corn oil (Sigma-Aldrich) and administered by oral gavage. Mice were sacrificed by CO₂ euthanasia.

Human Samples

Human AML and plasma samples were collected under a JHH IRB-approved protocol with patient informed consent in accordance with the Declaration of Helsinki. Mononuclear cells were isolated by Ficoll centrifugation and cryopreserved in liquid nitrogen until use. Cells were cultured in RPMI 1640 supplemented with 10% FBS for drug treatments and lysed and analyzed via immunoblot as described. Plasma was isolated following Ficoll centrifugation and frozen at −20°C until use. Plasma samples obtained during sorafenib therapy were collected on study NCT01578109 and analyzed per previously described methods (16).

Statistical Analysis

Statistical analysis was performed with Student t-test using the GraphPad Prism software analysis program. Combination index (CI) values were calculated using the Chou-Talalay method (17). All data are presented as the mean ± standard deviation (SD).

Results

FLT3 inhibition in FLT3/ITD AML cell is associated with reactivation of ERK signaling

In studying the efficacy of FLT3 inhibitors, we and others have generally analyzed short-term drug exposures (in the range of 30 min – 2 hrs) via immunoblot to assess maximal target and downstream inhibition, either via addition of drug to in vitro culture (2, 18), or direct application of patient plasma via the plasma inhibitory activity (PIA) assay (19–21). We sought to investigate whether signaling downstream of FLT3 is effectively inhibited over a longer treatment course in the context of FLT3-selective TKI treatment. We treated two FLT3/ITD AML cell lines, Molm14 and MV4;11, with five different inhibitors of FLT3 (18, 20, 22–24) for up to 24 hours at concentrations effective for >95% FLT3 and downstream target inhibition (25 nM sorafenib, 10 nM quizartinib, 1 nM TTT-3002, 25 nM lestaurtinib, and 25 nM crenolanib). Despite inhibition following one hour of treatment, rebound of ERK phosphorylation (pERK) was observed after 24 hours, while no such rebound was observed in the phosphorylation of downstream targets AKT or STAT5 (Fig. 1A and B). The same phenomenon was also observed in murine 32D cells stably transfected with FLT3/ITD (Supplementary Fig. S1). The rebound in pERK began as early as 16 hours after treatment, and was accompanied by a time-dependent increase in phosphorylation of MEK and p90RSK. An increase in levels of active RAS was also prominently observed in the MV4;11 cell line, suggesting a reactivation of the ERK signaling cascade (Fig. 1C and D). A minimal rebound in AKT phosphorylation was observed in the Molm14 cell line starting 8 hours after treatment, but was reduced by 24 hours. An increase in total FLT3 protein and, in some cases, FLT3 phosphorylation was also observed, consistent with previous reports which demonstrate that FLT3 inhibition significantly decreases the rate of its proteasomal degradation (25).

Fig. 1. ERK signaling is reactivated over time in response to FLT3 inhibition.

(A–B) FLT3/ITD AML cell lines Molm14 (A) and MV4;11 (B) were treated with the indicated FLT3 inhibitor for the indicated times. (C–D) Molm14 (C) and MV4;11 (D) cells were treated with 25 nM sorafenib for the indicated times. Protein lysates were analyzed by immunoblot. Whole cell lysates (WCL) were subjected to immunoprecipitation assays with GST-bound RAF1 Ras-binding domain (RBD). WCL and pull down (PD) products were immunoblotted with a pan-Ras antibody to detect levels of total and GTP-bound RAS, respectively.

To investigate the possibility that the observed adaptive response observed may simply be a result of loss of drug exposure (due either to half-life or to drug efflux), cells were treated with sorafenib for 24 hours, followed either by no change, washout and re-addition of drug, or replacement with culture medium alone. Despite washout and re-addition of fresh drug, ERK phosphorylation remained at its 24 hour rebound level and continued to increase over time, reaching pre-treatment levels by 48 hours after initial treatment (Fig. 2A). The rebound in pERK was dose-dependent, as the degree of rebound was reduced with increasing concentrations of sorafenib, crenolanib, and TTT-3002 (Fig. 2B and Supplementary Fig. S2A–B). This finding suggests a role for FLT3 signaling in the observed rebound; quizartinib and lestaurtinib treatment, however, elicited a low, persistent level of pERK rebound even at high drug concentrations (Fig. 2C and Supplementary Fig. S2C). Evidence of persistent adaptive increase in ERK phosphorylation despite effective inhibition of FLT3 suggests at least a partial contribution from signaling via alternate receptors and highlights the possibility that off-target effects may explain the dose-dependent nature of pERK rebound.

Fig. 2. Characterization of FLT3 inhibitor-mediated signaling adaptation.

(A) Molm14 cells were treated with 25 nM sorafenib for 24 hours, followed by no change, re-addition of drug, or replacement with culture medium alone for an additional one or 24 hours. (B–C) Molm14 cells were treated with the indicated concentrations of sorafenib (B) or quizartinib (C) for up to 24 hours. (D) Molm14 cells were treated with 25 nM sorafenib cultured in medium with the indicated concentrations of fetal bovine serum (FBS) for up to 24 hours. (E) Molm14 cells were initially treated with 25 nM sorafenib for up to 24 hours and cultured in 1% FBS. After 24 hours of treatment, FBS concentration was either increased to 10% or left at 1% for an additional 1 and 24 hours as indicated. (F) Molm14 cells were treated for up to 24 hours in 10% FBS. After 24 hours of treatment, FBS concentration was decreased to 5%, 1%, or 0% for an additional hour as indicated. Protein lysates were analyzed by immunoblot.

Rebound was also dependent on serum components, as a reduction in serum concentration resulted in sustained ERK inhibition, and re-introduction of serum following treatment in low serum conditions restored pERK rebound (Fig. 2D and E). As expected, 48 hours of persistent low-serum conditions resulted in reduced cell viability and increased protein degradation. Reducing serum concentration for one hour following 24 hour sorafenib treatment resulted in a profound decrease in MEK and ERK phosphorylation (Fig. 2F), supporting the presence of critical signaling molecules in serum which support pERK rebound. Importantly, FLT3 signaling was sustained over 24 hours in low-serum conditions in the absence of drug (Supplementary Fig. S3).

Together, these data demonstrate that FLT3/ITD leukemia cells exhibit a mechanism of intrinsic signaling adaptation in response to FLT3 inhibitor treatment, resulting in a reactivation of ERK signaling despite persistent inhibition of FLT3 and other effector pathways.

MEK inhibition overcomes rebound in phospho-ERK and sensitizes cells to FLT3 inhibition

We next sought to investigate whether FLT3 inhibitor-mediated pERK rebound could be overcome with the addition of a MEK inhibitor. The addition of allosteric MEK inhibitors PD0325901 (26) or trametinib (27) at submaximal concentrations significantly reduced the degree of pERK rebound 24 hours following sorafenib or crenolanib treatment in both Molm14 and MV4;11 cells (Fig. 3A and B and Supplementary Fig. S4A–D). This reduction was accompanied by decreased levels of the ERK target cMyc. Combination treatment also resulted in a slight increase in PARP cleavage, indicating increased apoptosis as compared to FLT3 inhibition alone. Interestingly, MEK inhibitor treatment alone resulted in an increase in pMEK and concurrent rebound in pERK in these cells, likely due to relief of negative feedback as has been described in the context of other RTK-driven cancers (28).

Fig. 3. MEK inhibitor treatment abrogates phospho-ERK rebound and sensitizes FLT3/ITD cells to FLT3 inhibitor treatment.

(A–B) Molm14 (A) and MV4;11 (B) cells were treated with 25 nM sorafenib, 5 nM PD0325901, or both, for the indicated times. Protein lysates were analyzed by immunoblot. Densitometric quantification of phospho-ERK relative to total ERK is shown. (C–D) Molm14 (C) and MV4;11 (D) cells were treated for 48 hours with 0–80 nM sorafenib and 0–80 nM PD0325901 in combinations. Annexin V positive cells were measured by flow cytometry as a readout of apoptosis. (E–F) Molm14 (E) and MV4;11 (F) cells were treated for 96 hours with 0–80 nM sorafenib and 0, 2.5, 5, or 10 nM PD0325901 as indicated. Proliferation was measured by MTT assay. Bar graphs represent sorafenib IC₅₀ as calculated using GraphPad Prism analysis software. Data are presented as the mean ± standard deviation (SD).

To evaluate whether MEK inhibition enhances the sensitivity of FLT3/ITD cells to FLT3 inhibition, cells were treated with various combinations for 48 or 96 hours and analyzed for apoptosis and cell proliferation. While FLT3/ITD cells were insensitive to MEK inhibition alone, the simultaneous addition of even very low concentrations of PD0325901 sensitized cells to sorafenib treatment, synergistically increasing apoptosis (Fig. 3C and D and Supplementary Fig. S5A to D) and reducing cell proliferation (Fig. 3E–F). The clinically approved MEK inhibitor trametinib was also evaluated in combination with sorafenib in FLT3/ITD cells. As with PD0325901, a low concentration of trametinib sensitized cells to sorafenib treatment (Supplementary Fig. S5E to H).

One hypothesis for the efficacy of such a combination is that sorafenib treatment together with MEK inhibition is a generally synergistic combination in cells with high ERK signaling, irrespective of FLT3 status. We therefore tested the effects of the sorafenib/MEK inhibitor combination in the FLT3-negative AML cell line HL60, which harbors the NRAS activating mutation Q61L. These cells showed intermediate sensitivity to MEK inhibition, yet complete insensitivity to sorafenib treatment (Supplementary Fig. S6A and B). Notably, the addition of sorafenib to PD0325901 treatment failed to impact the extent of apoptosis or proliferation of these cells. Additionally, K562 cells, which harbor wild type FLT3 and RAS, were insensitive to sorafenib, PD0325901, and the combination (Supplementary Fig. S6C).

These data support the notion that MEK inhibition, even at submaximal concentrations, is effective in overcoming the intrinsic signaling adaptation in FLT3/ITD leukemia cells, thereby sensitizing these cells to FLT3 inhibition and enhancing the biological response to drug treatment.

The addition of a MEK inhibitor to sorafenib treatment reduces leukemia burden in vivo

MEK inhibition has been previously proposed as a means to overcome bone marrow stromal cell-mediated activation of ERK signaling in FLT3/ITD leukemia (10). This work provides evidence suggesting that MEK inhibition may increase efficacy of FLT3 inhibitors in vivo via a distinct mechanism. To evaluate the efficacy of such a combination treatment, NOD/SCID gamma (NSG) mice were engrafted with Molm14 cells and treated with 8 mg/kg sorafenib, 5 mg/kg PD0325901, or the combination once daily for three weeks. Notably, the dose of PD0325901 used in these experiments is well below the reported maximum tolerated dose in murine models (26). Compared to either treatment alone, the combination of sorafenib and PD0325901 significantly reduced the leukemia cells in both the peripheral blood (Fig. 4A) and the protective environment of the bone marrow (Fig. 4B). These data provide further evidence for the in vivo efficacy of combining a low dose of MEK inhibitor with a FLT3 inhibitor in a FLT3/ITD leukemia model.

Fig. 4. MEK inhibition improves efficacy of FLT3 inhibition in vivo.

(A–B) 0.5 million Molm14 cells were injected via tail vein injection into Nod scid gamma mice. Mice were treated with 8 mg/kg sorafenib, 5 mg/kg PD0325901, or both via oral gavage once daily for three weeks. Mice were sacrificed and peripheral blood (A) and bone marrow (B) cells were analyzed for human CD45 (hCD45) positive leukemia cells by flow cytometry. n ≥ 4 for each group. Error bars represent standard deviation (SD).

TKI-mediated signaling adaptation occurs in primary blasts and is elicited by human plasma

It is well recognized that several aspects of standard in vitro culture conditions do not purely reflect the kinetics and pharmacodynamics of drug targeting when dosing patients with small molecule inhibitors of kinase signaling. This may be particularly relevant to our work given the dose-dependent nature of pERK rebound we observed in cultured FLT3/ITD leukemia cells (Fig. 2B). We therefore sought to determine whether rebound in ERK signaling following FLT3 inhibition occurs in patients as well. We first examined whether primary leukemic cells from patients with FLT3/ITD AML (Supplementary Table S1) also exhibit the same phenomenon observed in FLT3/ITD cell lines. After 24 hours of sorafenib treatment, the majority of primary blasts revealed a similar rebound in ERK phosphorylation, despite sustained inhibition of phospho-STAT5 (Fig. 5A). As observed with the cell lines, the addition of PD0325901 to sorafenib treatment resulted in sustained ERK inhibition after 24 hours, despite profound pERK rebound when treated with either drug alone (Fig. 5B).

Fig. 5. Phospho-ERK rebound is observed in patient blasts and elicited by human plasma.

(A) AML blasts from four patients (Table S1) were treated with 25 nM sorafenib in culture for the indicated times. (B) AML blasts from one patient were treated in culture with 25 nM sorafenib, 5 nM PD0325901, or both for the indicated times. Densitometric quantification of phospho-ERK relative to total ERK is shown. (C) Molm14 cells were cultured with either healthy donor plasma for one hour (ctrl) or plasma derived from patients on active sorafenib treatment (Table S2) for the indicated times. Protein lysates were analyzed by immunoblot.

Serum concentrations of orally dosed drugs are dependent on multiple variables including absorption, metabolism and protein binding. Clinical trials have revealed extremely high interpatient variability associated with sorafenib treatment, with maximum concentrations ranging from 0.89 to 26.2 µM (29, 30). Therefore, we next sought to determine whether signaling adaptation observed in culture conditions of 10% fetal bovine serum would be recapitulated by the concentrations of sorafenib present in human plasma from patients being treated with this FLT3 TKI. To replicate these conditions, we cultured Molm14 cells in plasma samples obtained from patients treated with sorafenib for a prolonged period (Supplementary Table S2). Using a variant of the PIA assay (19), Molm14 cells were treated in either normal human plasma for one hour or plasma from patients treated with sorafenib for either one or 24 hours. In all cases of cells exposed to plasma from these patients, pERK rebound was observed by 24 hours after treatment (Fig. 5C). These data support the notion that FLT3 inhibitor-mediated signaling adaptation is a phenomenon that occurs in patients, is elicited at clinically relevant concentrations of sorafenib in human plasma, and can be abrogated with the addition of a MEK inhibitor.

TKI-mediated pERK rebound is observed in other tyrosine kinase-driven cancers

Finally, we sought to determine whether TKI-mediated rebound in ERK activity was restricted to FLT3-mutant AML or whether it also occurs in the context of other tyrosine kinase-driven cancers, such as those driven by EGFR mutation or HER2 amplification. We treated the EGFR-mutant lung cancer cell line PC-9 with the EGFR/HER2 inhibitor lapatinib at concentrations sufficient for >95% EGFR inhibition. As was seen for FLT3/ITD-driven AML, pERK rebound was observed following 24 hours of TKI treatment, despite persistent inhibition of the receptor and downstream target AKT (Fig. 6A). Again, the rebound could be overcome with the addition of a submaximal concentration of PD0325901 (Fig. 6B). A similar trend was observed for the HER2-amplified breast cancer cell line SKBR3 (Fig. 6C and D) and the BCR-ABL cell line K562 (Fig. 6E and F). As with the FLT3/ITD cells, MEK inhibition increased sensitivity of K562 cells to imatinib treatment (Supplementary Fig. S7A). Unlike what was observed for FLT3/ITD cells however, the addition of PD0325901 failed to markedly increase sensitivity to TKI treatment in PC9 and SKBR3 cells (Supplementary Fig. S7B and C), consistent with a low dependency on ERK output in these cell types (28). These data demonstrate that TKI treatment is associated with signaling adaptation in a variety of tyrosine kinase-driven malignancies, including leukemia, lung cancer, and breast cancer.

Fig. 6. Phospho-ERK rebound is observed in other tyrosine kinase-driven cancers following exposure to tyrosine kinase inhibitors.

(A) PC-9 (EGFR Del E746-A750) cells were treated with 500 nM lapatinib for the indicated times. (B) PC-9 cells were treated with 500 nM lapatinib, 5 nM PD0325901, or both drugs for the indicated times. (C) SKBR3 (HER2 amplified) cells were treated with 75 nM lapatinib for the indicated times. (D) SKBR3 cells were treated with 75 nM lapatinib, 5 nM PD0325901, or both drugs for the indicated times. (E) K562 (BCR-ABL) cells were treated with 750 nM imatinib for the indicated times. (F) K562 cells were treated with 750 nM imatinib, 5 nM PD0325901, or both drugs for the indicated times. Protein lysates were analyzed by immunoblot. Densitometric quantification of phospho-ERK relative to total ERK is shown.

Discussion

The FLT3 internal tandem duplication (FLT3/ITD) is an oncogenic driver in AML, and while selective inhibitors of FLT3 that maximally and specifically inhibit the target have now been studied (18, 20), clinical responses remain limited (5–7). Several factors that may contribute to this limited efficacy have been postulated, and include both cell intrinsic and extrinsic processes, including acquisition of resistance mutations (2, 8), upregulation of FLT3 ligand (9), and stromal cell-mediated protection in the bone marrow microenvironment (10). We sought to explore, however, whether the biochemical response to FLT3/ITD inhibition might involve changes in signaling networks that also contribute to their limited efficacy.

Previous work has characterized either the short-term signaling response (<2 hours) or the long term generation of resistance (>1 month) following FLT3 inhibition in FLT3/ITD AML (8, 31). Here, we explored the signaling response to FLT3 inhibition within 48 hours of treatment, finding reactivation of ERK signaling beginning as early as 16 hours after treatment despite persistent target inhibition. This effect was observed in standard culture conditions as well as a modified plasma inhibitory activity (PIA) assay (19). The difference in timing of interrogation of the signaling pathways likely explains why this adaptive rebound was not previously observed. This work reveals a limitation of the PIA assay in its current form and highlights the need to analyze signaling over a greater treatment duration to fully understand the biologic response.

The concept of signaling adaptation in response to small molecule inhibitors of tyrosine kinases has been recently described in the context of other targeted therapies (11–14). In several cases the adaptive changes in signaling suggest reactivation of other nodes in critical mitogenic pathways that are therefore targets for combined therapies. In line with these findings, this report reveals a phenomenon by which FLT3/ITD AML selectively reactivates ERK signaling within hours in response to targeted inhibition. These data contribute to the growing understanding of short-term signaling adaptation to oncoprotein inhibitors, which is accompanied by attenuation of the antitumor effects of these compounds (11–14). The reactivation of ERK signaling described here may represent an extension of the previously described phenomenon whereby attenuation of ERK feedback inhibition allows for ligand-dependent signaling via wild type receptors present on the cell surface (Fig. 7A–C). This idea is supported by the dependence of pERK rebound on a sufficient concentration of serum, suggesting that signaling through receptors to mitogenic pathways is an essential component of the adaptation (Fig. 2B), though further work is needed to fully characterize the mechanism of ERK reactivation and to determine what signaling molecule(s) may be responsible. However, regardless of which receptors, if any, are involved in the enhanced signaling as an adaptive response, the measurement of elevated levels of ERK phosphorylation suggest that addition of a MEK inhibitor, rather than specific inhibitors of cell surface receptors, is an alternative approach for combination therapy that will likely overcome multiple mechanisms of adaptation through inhibition of a shared terminal event.

Fig. 7. Proposed model of TKI-mediated phospho-ERK rebound.

(A) At baseline, mutant tyrosine kinase (eg. FLT3/ITD) signaling results in high ERK activity and therefore high ERK-mediated feedback inhibition. (B) Following treatment with an inhibitor (TKI) of the mutant RTK, feedback inhibition is reduced allowing for signaling via alternate, wild type receptors to effector pathways. (C) MEK inhibition (MEKi) abrogates the rebound by blocking ERK phosphorylation downstream of alternate receptors.

RTK-driven cancers are, for the most part, insensitive to MEK inhibitors (28, 32), and clinical trials of single agent MEK inhibitors in AML have shown little efficacy, with FLT3/ITD patients achieving no responses (33). Combining FLT3 and MEK inhibitors as a treatment strategy for patients with FLT3/ITD AML has been proposed by others as a means to overcome bone marrow stromal cell-mediated resistance to FLT3 inhibition (10). Indeed, a dual FLT3/MEK inhibitor was recently described to be effective in preclinical models of FLT3/ITD AML, and a phase 1 clinical trial is ongoing (34). In this study, we have determined that MEK inhibition is an effective means to overcome FLT3 inhibitor-mediated pERK rebound, and that MEK inhibitor treatment sensitizes FLT3/ITD cells to FLT3 inhibition even in the absence of stromal cells. These data support the combination of MEK and FLT3 inhibitor treatment as a means to improve treatment for patients with FLT3/ITD AML, and further suggest that low-dose MEK inhibitor may be sufficient for improved outcomes. This is supported by the low concentrations necessary to observe synergistic effects in vitro (Fig. 3) and by the reduced leukemia burden observed in vivo using a dose of PD0325901 well below the maximal tolerated dose previously reported for this compound in mice (26) (Fig. 4). The advantage of this approach includes avoidance of the toxicities associated with full dose MEK inhibitor while maintaining sufficient activity to allow for enhanced response to receptor tyrosine kinase inhibition. Further, this combinatorial strategy could be readily applied to the clinic given the availability of inhibitors of FLT3 and MEK, including clinically approved compounds such as sorafenib and trametinib. Indeed, a phase I clinical trial exploring sorafenib and trametinib in combination for the treatment of advanced hepatocellular cancer is ongoing (https://clinicaltrials.gov/show/NCT02292173).

We also explored the possibility that this phenomenon may be generalizable beyond FLT3/ITD AML. We demonstrate that a similar phenomenon occurs in models of BCR-ABL leukemia, EGFR-driven lung cancer and HER2-amplified breast cancer in response to TKI treatment. As in FLT3/ITD AML, pERK rebound can be overcome with the addition of a MEK inhibitor (Fig. 5). While further work is needed to validate the feasibility of such a combination, these data suggest that the use of a MEK inhibitor may enhance the response of tyrosine kinase-driven cancers to TKI therapy, thereby improving outcomes for patients with cancers carrying these genetic alterations.