PP2A-activating Drugs Enhance FLT3 Inhibitor Efficacy through AKT Inhibition–Dependent GSK-3β–Mediated c-Myc and Pim-1 Proteasomal Degradation

Internal tandem duplication of the fms-like tyrosine kinase 3 receptor tyrosine kinase (FLT3-ITD), resulting in constitutive and aberrant FLT3 signaling (1), is present in acute myeloid leukemia (AML) cells of 30% of patients (2), and patients with FLT3-ITD AML have short disease-free survival (2). FLT3 tyrosine kinase inhibitors (TKIs) are cytotoxic toward FLT3-ITD–expressing AML cells in vitro and in vivo, but clinical responses are generally limited and transient (3). Efforts have focused on identifying drug combinations to improve responses to FLT3 inhibitors, and thus patient outcomes (3). FLT3-ITD oncogenic tyrosine kinase activity promotes leukemogenesis through both STAT5-dependent (1) overexpression of the oncogenic serine/threonine kinase Pim-1 (4) and inactivation of the tumor-suppressor multimeric serine/threonine protein phosphatase 2A (PP2A; refs. 5, 6), providing additional therapeutic targets.

FLT3-ITD both constitutively activates FLT3 and causes aberrant signaling through STAT5 and downstream Pim-1, in addition to signaling through PI3K/AKT and MEK/ERK (1). Pim-1 contributes to FLT3-ITD proliferative and antiapoptotic effects through phosphorylation-dependent stabilization of regulators of cell growth and survival, including the c-Myc oncogene (7). It also phosphorylates and stabilizes FLT3 in a positive feedback loop in FLT3-ITD–expressing cells (8, 9). Inhibition of Pim activity enhances FLT3 inhibitor cytotoxicity in FLT3-ITD–expressing AML cells in vitro and in vivo (8–11).

PP2A is a heterotrimeric protein composed of 65 and 36 kDa structural/scaffold A and catalytic C subunits and a diverse repertoire of structurally distinct regulatory B subunits that dictate subcellular localization and substrate specificity (12). PP2A enzymatic activity is negatively regulated through binding of inhibitory proteins including the cancerous inhibitor of PP2A (CIP2A), SET, and the binding protein for SET (SETBP1; ref. 12). CIP2A upregulation has been reported as a mechanism of PP2A inactivation in FLT3-ITD–expressing cells (13). Cotreatment with the PP2A-activating drugs (PADs) FTY720 or OP449 was shown to enhance FLT3 inhibitor cytotoxicity in FLT3-ITD–expressing cells in vitro (5, 6).

Pim-1 kinase is a PP2A substrate, and its PP2A-dependent dephosphorylation decreases its expression through decreased protein stability (14). Mechanistically, Pim-1 interacts with the PP2A regulatory B subunit B56β; B56β knockdown decreases Pim-1 ubiquitination and increases Pim-1 protein half-life, increasing expression of both of the Pim-1 isoforms, 44 kDa Pim-1L and 33 kDa Pim-1S (15).

FLT3-ITD, PP2A, and Pim-1 all regulate expression of the transcription factor c-Myc (7, 16, 17). c-Myc is transcriptionally upregulated downstream of FLT3-ITD, and c-Myc or Pim-1 knockdown has antiproliferative effects in FLT3-ITD–expressing cells (16). In contrast, PP2A and Pim-1 both regulate c-Myc expression posttranslationally (7, 17, 18). Two N-terminus phosphorylation sites, serine 62 (S62) and threonine 58 (T58), are important in regulating c-Myc protein stability (18). Pim-1 increases S62 phosphorylation, stabilizing c-Myc protein, in association with decreased T58 phosphorylation (7), whereas the PP2A regulatory subunit B56α selectively associates with the c-Myc N-terminus and PP2A dephosphorylates c-Myc at S62, resulting in reduced c-Myc protein stability (17). Thus, FLT3 inhibition decreases c-Myc transcription, whereas both Pim-1 inhibition and PP2A activation decrease c-Myc protein stability.

PADs are in development (19). Here, we studied mechanisms by which they enhance FLT3 inhibitor efficacy in cells with FLT3-ITD.

Mouse Ba/F3 cells transfected with human FLT3-ITD (Ba/F3-ITD) and wild-type (WT) FLT3 (FLT3-WT; Ba/F3-WT) and the FLT3-ITD–expressing MV4-11 human AML cell line were obtained and cultured as previously described (20). The FLT3-WT cell lines OCI-AML2 and THP-1 were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen. Mycoplasma testing was performed every 6 months with the Mycoplasma PCR Detection Kit (GeneCopoeia).

Ba/F3-ITD cells were infected with a pMX-puro retroviral vector encoding Flag-K67M kinase-dead (KD) Pim-1, from Dr. Tomasz Skorski (Temple University, Philadelphia, PA), or empty vector control, as described (20). Pim-1 overexpression was confirmed by immunoblotting.

The myc-estrogen receptor (ER) expression vector pBABE-puro-myc-ER (plasmid No. 19128; ref. 21) and pBABE-puro empty vector control were from Addgene. Approximately 80% confluent Phoenix-AMPHO packaging cells (ATCC CRL-3213) were incubated in 25 μmol/L chloroquine for 1 hour and then transfected with 20-μg retroviral plasmid DNA by the calcium phosphate method. Ba/F3-ITD cells were infected with virus-containing medium collected after 24 hours in the presence of polybrene (4 μg/mL). Cells were seeded in 2-mL virus-containing medium, centrifuged at 1,800 rpm at 32°C for 45 minutes, then incubated at 32°C for 4 hours, centrifuged at 1,800 rpm at 32°C for 45 minutes, and then incubated in fresh virus-containing medium at 32°C for 2 hours. The cells were then incubated at 37°C for 24 hours, infected with virus-containing medium, and incubated at 32°C for 5 hours. Infected cells were incubated in fresh virus-free medium overnight and then cultured with 1 mg/mL puromycin for 14 days. Myc overexpression was confirmed by immunoblotting. Myc-ER–expressing Ba/F3-ITD cells were treated with 300 nmol/L 4-hydroxytamoxifen (4-OHT; Sigma-Aldrich) to activate the myc-ER fusion protein via translocation from cytoplasm to nucleus.

Ba/F3-ITD cells were also infected with constitutively active myristoylated AKT (Myr-AKT), pBabe-Puro-Myr-Flag-AKT1 (Addgene; plasmid No. 15294; ref. 22), or pBABE-puro empty vector control, as above.

Finally, Ba/F3-ITD cells were infected with pMSCVpuro-Flag-cMyc-T58A plasmid (Addgene; plasmid No. 20076; ref. 18) containing c-Myc with a T58 mutation changing threonine to alanine, inhibiting phosphorylation, or pMSCVpuro empty vector control (Takara Bio USA).

Blood and bone marrow samples were obtained from patients with FLT3-ITD or FLT3-WT AML (Supplementary Table S1) on a University of Maryland Baltimore Institutional Review Board–approved protocol, following written informed consent, in accordance with the Declaration of Helsinki. Mononuclear cells isolated by density centrifugation over Ficoll-Paque (Sigma-Aldrich) were cultured in RPMI 1640 with 20% FBS, without cytokine supplementation, or with and without FLT3 inhibitors and/or PADs. Cells were studied fresh, without prior cryopreservation.

Gilteritinib (ASP2215; Active Biochem; ref. 4) and quizartinib (AC220; Selleck Chemicals; ref. 4), type I and II FLT3 inhibitors, respectively, clinically active in FLT3-ITD AML, were used at pharmacologically relevant concentrations (23, 24). The SET-sequestering PAD FTY720 (ref. 25; Fingolimod; Cayman Chemical Company) was also used at pharmacologically relevant concentrations. DT-061, an orally bioavailable small-molecule activator of PP2A developed by reengineering tricyclic neuroleptics and proposed to directly bind the PP2A Aα subunit (26–29), was provided by Dr. G. Narla. The pan-Pim inhibitor AZD1208 and GSK-3β inhibitors TC-G 24 and TWS119 were from Tocris Bioscience, the c-Myc inhibitor 10058-F4 from Sigma-Aldrich, and the pan-AKT inhibitor MK-2206 from Selleck Chemicals.

Cytotoxicity was measured using the WST-1 assay after 48-hour culture (20). IC₅₀ values were determined by nonlinear curve fitting to a dose–response curve using Prism V software (GraphPad).

Cells cultured in RPMI 1640 medium with 10% FBS with and without drug(s) were collected at serial time points, and live cells were counted after trypan blue dye exclusion, as described previously (20).

Apoptosis was measured by Annexin V/propidium iodide (PI) staining, as described (20).

Cells plated in triplicate on 96-well plates were treated with drugs at various concentrations alone and in combinations. Assays were terminated after 48 hours, and combination indexes were determined according to the Chou–Talalay method using CompuSyn software, as previously described (11). Values less than 1, equal to 1, and greater than 1 are synergistic, additive, and antagonistic, respectively.

NOD-scid IL2Rg^null (NSG) mice age 6 to 8 weeks were purchased from The Jackson Laboratory. Exponentially growing MV4-11-luc cells (1 × 10⁶), from Dr. Sharyn Baker (The Ohio State University, Columbus, OH), were injected intravenously (i.v.) into the lateral tail veins of restrained mice (11). Engraftment was assessed 7 days later using the Xenogen IVIS Spectrum Imaging System (Caliper Life Sciences) after intraperitoneal injection of 150 mg/kg D-luciferin, and mice were sorted into four treatment groups (n = 5 each) with equal mean signal intensity. Gilteritinib (Chemietek) and DT-061(MedChemExpress USA) were formulated via loading into preformed liposomes by transmembrane ammonium sulfate gradient or passive equilibration techniques, respectively. Gilteritinib (2 mg/kg) and/or DT-061 (5 mg/kg) were administered i.v. to MV4-11-luc–bearing mice every other day beginning on day 7 after inoculation. Leukemia burden was measured weekly by bioluminescence imaging, as above. Bioluminescent image data were analyzed with Living Image software (PerkinElmer). Total bioluminescent signal was obtained as photons/second. Endpoints were 20% body weight loss, hind limb paralysis, or lack of mobility to eat/drink. The University of Maryland Institutional Animal Care and Use Committee approved the study.

Cells were lysed in 150 mmol/L NaCl lysis buffer with protease and phosphatase inhibitors (Sigma-Aldrich). Protein concentration was measured using the Dye Reagent Concentrate (Bio-Rad Protein Assay Kit; Bio-Rad Laboratories), and 20 μg from each sample was immunoblotted (11, 20). Immunoblots were incubated with antibodies (catalog numbers in parentheses) including polyclonal antibodies to c-Myc (9402), Pim-1 (2907), BAD (9292), p-BAD (S112; 9291), p-p44/42 MAPK (ERK1/2; T202/Y204; 9101), and ubiquitin (3933) and mAbs to p44/42 MAPK (ERK1/2; 9107), GSK-3α/β (5676), p-GSK-3α/β (S9/S21; 8566), AKT (9272), p-AKT (S473; 4060), p-AKT (T308; 2965), and p44/42MAPK (ERK1/2; 9107; Cell Signaling Technology), β-actin (clone C-11, sc-1615 HRP; Santa Cruz Biotechnology), vinculin (V9264; Sigma-Aldrich), p-Myc (S62; ab78318), and p-Myc (T58; ab185655; Abcam) overnight at 4°C, and then with horseradish peroxidase–conjugated secondary antibodies for 1 hour at room temperature. Band intensities measured by densitometry (VisionWorksLS, UVP) at serial time points, were compared with pretreatment intensities, defined as 100%.

To study protein turnover, cells were treated with 100 μg/mL cycloheximide (CHX; Sigma-Aldrich) for 60 minutes to block new protein translation before addition of FLT3 inhibitor and/or PAD or DMSO control. Protein expression was measured at serial time points by immunoblotting. Band intensities measured by densitometry at serial time points were compared with pretreatment intensity, defined as 100%. Protein half-life was calculated using the algorithm t_1/2 = 0.693t_n/ln(C₀/C_n), where n represents the last time point (30).

To study the effect of proteasomal degradation, cells were treated with CHX as above, with and without addition of the proteasome inhibitor carbobenzoxy-L-leucyl-L-leucyl-L-leucinal (MG-132; Calbiochem; 20 μmol/L) 30 minutes after adding CHX and 30 minutes before adding FLT3 inhibitor and/or PAD.

Ba/F3-ITD cells were treated with 15 nmol/L gilteritinib and 2 μmol/L FTY720 or DMSO control for 4 hours. Cell lysates were pulled down with c-Myc antibody (Cell Signaling Technology), and immunoprecipitated protein was probed with ubiquitin antibody (Cell Signaling Technology). Ba/F3-ITD cells were also treated with 1 nmol/L quizartinib and 2 μmol/L FTY720, or DMSO control, for 1 hour. Cell lysates were pulled down with Pim-1 antibody (Santa Cruz Biotechnology), and immunoprecipitated protein was probed with ubiquitin antibody.

RNA was isolated using Trizol Reagents (Thermo Fisher Scientific), and cDNA was created using Superscript IV Reverse Transcriptase (Thermo Fisher Scientific). RT-qPCR was performed using SYBER Green (MilliporeSigma). All reactions were performed in triplicate. Primers are shown in Supplementary Table S2. The ΔCt method for relative quantification of gene expression was used to determine mRNA expression levels.

The Pim-1 amino acid sequence (https://www.uniprot.org/uniprot/P11309 No. P11309-1) was searched for potential GSK-3β phosphorylation sites using the Group-based Prediction System algorithm (GPS 5.0, http://gps.biocuckoo.cn), with cutoff set to a medium threshold with a false prediction rate below 6% for serine/threonine kinases. Sites identified included S95/S4 on Pim-1L/Pim-1S, which is present in a sequence (MLLSKINSL) highly conserved among mammalian species. WT peptide with S95/S4 (EVGMLLSKINSL) and synthetic peptide with S95A/S4A mutation (EVGMLLAKINSL) preventing phosphorylation at this site were obtained from LifeTein. GSK-3β phosphorylation was measured using the nonradioactive ADP-Glo GSK-3β kinase assay (Promega), according to the manufacturer's instructions. WT and mutant peptides (10 μg) were added to 0.2 ng GSK-3β kinase enzyme, 1 μg GSK-3β kinase substrate as a positive control, and 50 μmol/L ATP, and the reaction was incubated at room temperature for 60 minutes in the dark. ADP-Glo reagent (5 μL) was then added to each well, followed by incubation for 30 minutes in the dark (ATP depletion phase). Kinase Glo detection reagent (10 μL) was then added, followed by incubation for 30 minutes at room temperature in the dark (ADP detection phase). The luminescent signal was detected using the FlexStation 3 multi-mode microplate reader (Molecular Devices), and data were analyzed by SoftMax Pro 5.4.6. software (Molecular Devices). Luminescent signal for mutant peptide was normalized to luminescent signal for WT peptide.

All in vitro data were derived from at least three independent experiments, with error bars representing SEM. Statistical analysis was performed by two-way ANOVA with post hoc Bonferroni testing, using Prism 5 (GraphPad). In the in vivo model, photon intensity in mice treated with DT-061 and gilteritinib combination versus gilteritinib alone was compared by two-way ANOVA with Sidak's multiple comparison test.

To demonstrate the effect of cotreatment on cell growth in vitro, MV4-11 and Ba/F3-ITD cells, with FLT3-ITD, were grown in RPMI 1640 medium with 10% FBS with FLT3 inhibitors (gilteritinib or quizartinib) and/or PADs (FTY720 or DT-061) at their IC₅₀ concentrations for these cell lines (Supplementary Table S3), or DMSO control, and viable cells were counted at serial time points. While FLT3 inhibitors and PADs as single agents decreased growth of both cell lines relative to DMSO control, concurrent treatment with PAD and FLT3 inhibitor markedly decreased growth relative to single-drug treatments in MV4-11 (Fig. 1A) and Ba/F3-ITD (Supplementary Fig. S1A) cells.

To investigate the cytotoxic effects of cotreatment, MV4-11 and Ba/F3-ITD cells were cultured for 48 hours with the FLT3 inhibitors gilteritinib or quizartinib at their IC₅₀ concentrations for these cell lines (Supplementary Table S3) with FTY720 or DT-061 at increasing concentrations, and apoptosis was measured by Annexin V/PI labeling, detected by flow cytometry. PAD and FLT3 inhibitor combinations significantly increased apoptosis, relative to single-drug treatments in MV4-11 (Fig. 1B) and Ba/F3-ITD (Supplementary Fig. S1B) cells.

Synergy of PAD and FLT3 inhibitor combinations was also demonstrated by Chou–Talalay analysis in the FLT3-ITD cell lines MV4-11 (Fig. 1C) and Ba/F3-ITD (Supplementary Fig. S1C), whereas synergy was not seen in the FLT3-WT AML cell lines THP-1 and OCI-AML2 (Supplementary Fig. S2).

To study PAD and/or FLT3 inhibitor effects in vivo, NSG mice with MV4-11-luc cells injected i.v. and allowed to engraft were treated with gilteritinib and/or DT-061 or vehicle control, five mice per treatment group, as described in Materials and Methods, and leukemia burden was measured by luciferin imaging. By day 35, leukemia burden was significantly lower in mice treated with DT-061 and gilteritinib combination, compared with gilteritinib alone (P = 0.0018 by two-way ANOVA with Sidak's multiple comparison test; Fig. 1D), and also compared with DT-061 alone (P < 0.0001). Images of all mice are shown in Supplementary Fig. S3.

Finally, blasts of patients with FLT3-ITD AML were also cultured for 48 hours with gilteritinib (15 nmol/L) or quizartinib (1 nmol/L) with and without 4 μmol/L FTY720 or 10 μmol/L DT-061. PAD treatment significantly increased apoptosis induction by FLT3 inhibitors (Fig. 1E). In contrast, PAD and FLT3 inhibitor combinations at these concentrations were not cytotoxic in blasts from patients with AML with FLT3-WT or marrow cells from patients with AML in complete remission (CR) or CR with incomplete platelet recovery (Supplementary Fig. S4).

To study effects of PAD and FLT3 inhibitor treatment on c-Myc and Pim-1, expression of these proteins was measured by immunoblotting at serial time points in Ba/F3-ITD and MV4-11 cells and blasts from patients with FLT3-ITD AML treated with gilteritinib and/or FTY720 or DT-061, or DMSO control. Cotreatment with gilteritinib and FTY720 or DT-061 markedly downregulated both c-Myc and Pim-1 protein expression, relative to treatment with single drugs or DMSO control (Fig. 2A–D). Expression of p-T58- and p-S62-c-Myc relative to total c-Myc was also studied, demonstrating that p-T58-c-Myc expression was increased and more sustained than p-S62- and total c-Myc expression in both Ba/F3-ITD and MV4-11 cells treated with gilteritinib and FTY720 combination, though expression was more sustained in Ba/F3-ITD (Supplementary Fig. S6).

mRNA expression was also studied (Supplementary Fig. S7). c-Myc mRNA expression was stable in Ba/F3-ITD cells treated with gilteritinib and FTY720, but decreased and then stabilized in similarly treated MV4-11 cells, as did Pim-1 mRNA expression in both cell lines. Progressive mRNA downregulation, as seen for protein expression, was not observed.

Finally, effects of PAD and FLT3 inhibitor treatment on c-Myc and Pim-1 in blasts from patient 7, with FLT3-WT AML, are shown in Supplementary Fig. S8. c-Myc and Pim-1 were not downregulated by combination treatment.

As an initial approach to testing whether c-Myc and Pim-1 downregulation by PAD and FLT3 inhibitor cotreatment is posttranslational, c-Myc and Pim-1 protein expression was measured by immunoblotting at serial time points in Ba/F3-ITD cells treated with quizartinib and/or FTY720, or DMSO control, with and without pretreatment with the proteasome inhibitor MG-132. c-Myc and Pim-1 protein expression was downregulated in Ba/F3-ITD cells cotreated with quizartinib and FTY720, and pretreatment with MG-132 inhibited this downregulation (Supplementary Fig. S9). This suggested increased proteasomal degradation as a mechanism for the observed downregulation of c-Myc and Pim-1 protein expression in FLT3-ITD–expressing cells cotreated with PAD and FLT3 inhibitor.

To confirm that c-Myc and Pim-1 protein downregulation was a result of increased proteasomal degradation, Ba/F3-ITD (Fig. 3A) and MV4-11 (Fig. 3B) cells were pretreated with CHX for 60 minutes to block new protein translation, with and without addition of MG-132 after 30 minutes, and then treated with gilteritinib and FTY720, or DMSO control. c-Myc and Pim-1 protein turnover increased with gilteritinib and FTY720 cotreatment, compared with DMSO control, with a marked reduction in the half-lives of c-Myc and Pim-1 proteins (Supplementary Fig. S10). MG-132 treatment decreased both c-Myc and Pim-1 protein turnover (Fig. 3A and B; Supplementary Fig. S10), consistent with increased proteasomal degradation as the mechanism for the c-Myc and Pim-1 protein downregulation in FLT3-ITD–expressing cells cotreated with PAD and FLT3 inhibitor.

To further confirm this mechanism, ubiquitinated c-Myc and Pim-1 were measured in Ba/F3-ITD cells treated with gilteritinib and FTY720, or DMSO control, for 4 hours and 1 hour, respectively. A marked increase in ubiquitinated c-Myc (Fig. 3C) and Pim-1 (Fig. 3D) was seen with gilteritinib and FTY720 treatment, compared with control.

To determine whether c-Myc downregulation by PAD and FLT3 inhibitor cotreatment occurs via effects on Pim-1, Ba/F3-ITD cells treated with the pan-Pim kinase inhibitor AZD1208 and Ba/F3-ITD cells expressing KD Pim-1 were studied. Ba/F3-ITD cells were treated with FTY720 and gilteritinib with and without pretreatment with the pan-Pim kinase inhibitor AZD1208, and c-Myc expression was measured at serial time points. c-Myc was similarly downregulated in cells with and without AZD1208 pretreatment (Fig. 4A). c-Myc expression was also studied at serial time points in parental Ba/F3-ITD cells and Ba/F3-ITD cells expressing KD Pim-1 or empty vector control treated with gilteritinib and FTY720. c-Myc was similarly downregulated in all three (Fig. 4B). Decreased p-BAD (S112) demonstrates inhibition of Pim-1 activity in Fig. 4A and B.

To determine whether Pim-1 downregulation by FTY720 and quizartinib cotreatment is dependent on c-Myc downregulation, Pim-1 expression was measured at serial time points in Ba/F3-ITD cells and Ba/F3-ITD cells infected with ER-Myc plasmid or empty vector control. Pim-1 was similarly downregulated in all three (Fig. 4C). In addition, treatment of Ba/F3-ITD cells with the c-Myc inhibitor 10058-F4 at 100 μmol/L, a concentration at which it induced apoptosis (Fig. 4D, inset), did not downregulate Pim-1 expression (Fig. 4D).

Because CIP2A overexpression has been associated with significantly higher c-Myc, STAT5, and p-AKT (S473) (active) levels in FLT3-ITD–expressing cells (13) and p-AKT (S473) regulates c-Myc stability in prostate cancer cells (31), we hypothesized that p-AKT (S473) might regulate c-Myc protein turnover in FLT3-ITD–expressing cells. To test this hypothesis, we studied p-AKT (S473) expression in FLT3-ITD–expressing cells treated with PAD and/or FLT3 inhibitor.

p-AKT (S473) decreased rapidly in Ba/F3-ITD and MV4-11 cells and blasts from patients with FLT3-ITD AML cotreated with quizartinib or gilteritinib and FTY720 or DT-061, but not in cells treated with single drugs or with DMSO control (Fig. 5A–C). p-AKT (S473) decrease preceded Pim-1 and c-Myc downregulation (Fig. 5A–C). p-AKT (T308) also decreased with drug cotreatment (Fig. 5B).

To confirm that AKT inactivation occurs upstream of c-Myc downregulation, Ba/F3-ITD cells were treated with the Myc inhibitor 10058-F4. Myc inhibition neither decreased p-AKT (S473) nor inhibited the decrease in p-AKT induced by gilteritinib and FTY720 cotreatment (Fig. 5D).

To determine whether AKT inactivation is necessary and sufficient for c-Myc and Pim-1 downregulation and apoptosis induction by PAD and FLT3 inhibitor cotreatment, Ba/F3-ITD cells were infected with Myr-AKT1, causing constitutive AKT activation, and were also treated with the pan-AKT inhibitor MK-2206.

In Ba/F3-ITD cells infected with pBabe-Puro-Myr-Flag-AKT1, FTY720 and quizartinib cotreatment did not downregulate c-Myc or Pim-1 expression, in contrast to findings in parental Ba/F3-ITD and Ba/F3-ITD infected with empty vector control (Fig. 6A). pBabe-Puro-Myr-Flag-AKT1 expression also inhibited induction of increased c-Myc and Pim-1 protein turnover by FTY720 and quizartinib cotreatment (Fig. 6B). Moreover, FTY720 and quizartinib cotreatment did not induce apoptosis in Ba/F3-ITD cells infected with pBabe-Puro-Myr-Flag-AKT1, in contrast to findings in parental Ba/F3-ITD and Ba/F3-ITD infected with empty vector control (Fig. 6C). Therefore, AKT inactivation is necessary for downregulation of c-Myc and Pim via increased protein turnover and induction of apoptosis by PAD and FLT3 inhibitor cotreatment.

In addition, treatment with the AKT inhibitor MK-2206 downregulated c-Myc and Pim-1 protein expression similarly to FTY720 and gilteritinib cotreatment (Fig. 6D). Moreover, MK-2206 increased c-Myc and Pim-1 protein turnover, and this effect was inhibited by pretreatment with MG-132 (Fig. 6E). Finally, MK-2206 induced apoptosis of Ba/F3-ITD and MV4-11 cells similarly to FTY720 and gilteritinib cotreatment (Fig. 6F). Thus, AKT inactivation is sufficient for c-Myc and Pim-1 downregulation via increased protein turnover and for induction of apoptosis in FLT3-ITD–expressing cells.

We also studied p-ERK1/2 expression in Ba/F3-ITD and MV4-11 cells treated with gilteritinib and/or FTY720. p-ERK1/2 (T202/Y204) was downregulated in cells treated with drug combination, but later than, and to a lesser extent than, p-AKT (S473; Supplementary Fig. S11).

Based on the fact that c-Myc inhibition potently induces apoptosis of cells with FLT3-ITD (Fig. 4D), c-Myc is likely responsible for the apoptosis observed in Fig. 6F, but it cannot be stated with certainty that it is solely responsible.

Because GSK-3β phosphorylates c-Myc at T58 to enhance its proteasomal degradation (18) and GSK-3β is an AKT substrate (32), we hypothesized that AKT inactivation by PAD and FLT3 inhibitor cotreatment decreases GSK-3β phosphorylation, thereby activating it, resulting in c-Myc T58 phosphorylation and thus enhanced c-Myc proteasomal degradation.

Gilteritinib and DT-061 cotreatment caused rapid GSK-3α/β dephosphorylation (activation) in MV4-11 cells (Fig. 7A). Moreover, cotreatment with the GSK3-β inhibitor TC-G 24 prevented c-Myc and Pim-1 downregulation in Ba/F3-ITD cells (Fig. 7B) and blasts from patients with FLT3-ITD AML (Fig. 7C) treated with gilteritinib and DT-061. Treatment with TC-G 24 also inhibited apoptosis induction by PAD and FLT3 inhibitor combinations (Fig. 7D). A second GSK3-β inhibitor, TWS119, also inhibited c-Myc and Pim-1 downregulation in Ba/F3-ITD cells (Supplementary Fig. S12A) and inhibited apoptosis induction (Supplementary Fig. S12B) by PAD and FLT3 inhibitor combinations.

We next demonstrated that AKT inactivation is necessary and sufficient for GSK-3β activation by PAD and FLT3 inhibitor cotreatment. FTY720 and quizartinib cotreatment caused GSK-3α and -β dephosphorylation (activation) in parental Ba/F3-ITD cells and Ba/F3-ITD cells infected with empty vector, but not with myristoylated (constitutively activated) AKT (Fig. 7E). In addition, treatment with the AKT inhibitor MK-2206 resulted in GSK-3α and -β dephosphorylation, similarly to FTY720 and gilteritinib cotreatment (Fig. 7F).

To demonstrate that AKT-mediated c-Myc proteasomal degradation occurs through c-Myc phosphorylation at T58, Ba/F3-ITD cells infected with pMSCVpuro-Flag-cMyc-T58A plasmid with c-Myc mutation at T58, inhibiting phosphorylation, or empty vector control were cotreated with FTY720 and gilteritinib. FTY720 and gilteritinib cotreatment did not increase c-Myc protein turnover in cells expressing c-Myc with the T58A mutation (Fig. 8A), demonstrating that FLT3 inhibitor and PAD cotreatment increases c-Myc proteasomal degradation through c-Myc T58 phosphorylation. pMCVpuro-FLAG-cMYC-WT is unavailable as a control, but it should be noted that Pim-1 kinase is similarly downregulated by combination treatment in parental Ba/F3-ITD and Ba/F3-ITD overexpressing c-Myc or empty vector control (Fig. 4C). In addition, expression of c-Myc with the T58A mutation protected against apoptosis induction by quizartinib and FTY720 combination (Fig. 8B).

We next sought to determine the mechanism by which GSK-3β regulates Pim-1 protein turnover. To investigate potential GSK-3β phosphorylation sites on Pim-1L and Pim-1S, bioinformatic analysis of their protein sequences (GPS 5.0, http://gps.biocuckoo.cn) was performed. Putative phosphorylation sites identified included S65, S69, S74, and S95 on Pim-1L and S4 on Pim-1S. The S95 sequence in Pim-1L (MLLAKINSL) is homologous to the S4 sequence in Pim-1S (Fig. 8C, top) and is also conserved in multiple mammalian species, supporting essential biological functions of these residues. We therefore tested the peptide sequence EVGMLLSKINSL, which includes the putative phosphorylation sites S95 in Pim-1L and S4 in Pim-1S, as potential substrates for GSK-3β kinase activity in an in vitro kinase assay. Commercially available GSK-3β active kinase was incubated with this peptide or its corresponding mutant version in which the predicted phosphorylation serine (S) was substituted by alanine (A), which cannot be phosphorylated. GSK-3β phosphorylated the WT peptide, but not the mutant peptide with A substituted for S (Fig. 8C, bottom). We therefore tested this site, S95/S4 on Pim-1/Pim-1L, with peptides EVGMLLSKINSL and EVGMLLAKINSL, the latter with a mutated putative GSK-3β phosphorylation site, in a GSK-3β kinase assay, described in Materials and Methods. GSK-3β kinase activity was markedly decreased when EVGMLLAKINSL, rather than EVGMLLSKINSL, was present in the phosphorylation reaction (Fig. 8C, bottom), consistent with GSK-3β phosphorylation of S95/S4 on Pim-1L/Pim-1S.

The pathway affected by the PAD and FLT3 inhibitor combination treatment is shown in Fig. 8D.

FLT3-ITD is present in AML cells of 30% of patients (2) and results in constitutive and aberrant FLT3 signaling (1). Patients with FLT3-ITD AML have short disease-free survival (2). FLT3 inhibitors are cytotoxic toward FLT3-ITD–expressing AML cells and produce clinical benefit, with the FLT3/multikinase inhibitor midostaurin approved for use after chemotherapy for newly diagnosed AML with FLT3 mutations (33), and the more potent and specific FLT3 inhibitor gilteritinib approved for single-agent treatment of refractory or relapsed AML with FLT3 mutations (34). Nevertheless, clinical responses to FLT3 inhibitors are generally limited and transient, with rapid development of resistance (3). Combinations with drugs targeting other molecules in FLT3-ITD signaling pathways may improve responses to FLT3 inhibitors (3). Drugs that have been shown to enhance efficacy of FLT3 inhibitors when given in combination have included Pim kinase inhibitors (8–11, 35), AKT inhibitors (36, 37), the Bcl-2 inhibitor venetoclax (38), and PADs (5, 6).

PADs and FLT3 inhibitors have been previously shown to produce synergistic cytotoxicity in FLT3-ITD–expressing AML cells in vitro (5, 6). The PAD OP449, which activates PP2A by antagonizing its inhibition by SET, is cytotoxic toward the FLT3-ITD–expressing cell line MOLM-14 and results in decreased STAT5, AKT, and S6 ribosomal protein phosphorylation, and OP449 and quizartinib cotreatment synergistically reduces MOLM-14 cell growth (5). Treatment with the PAD FTY720 and its analogue AAL(S) decreases ERK and AKT phosphorylation, induces death of FLT3-ITD–expressing cells, and inhibits colony formation, and FTY720 and AAL(S) has synergistic cytotoxicity with FLT3 inhibitors, including in the presence of bone marrow stroma, and synergistic effects on colony formation by FLT3-ITD–expressing cells, but not normal CD34⁺ bone marrow cells (6). It was hypothesized that because PP2A induces degradation and inactivation of Pim-1 (14), FTY720 or AAL(S) might inhibit Pim-1 as the mechanism for synergy with FLT3 inhibitors (6). Here, we demonstrate that PAD and FLT3 inhibitor cotreatment of FLT3-ITD–expressing cells results in marked downregulation of both c-Myc and Pim-1 protein expression through increased proteasomal degradation of both proteins, resulting from AKT inactivation, through a GSK-3β–dependent mechanism. Our work adds to the growing literature on effectiveness of the dual therapeutic strategy of activating phosphatases while inhibiting kinases to enhance efficacy of kinase inhibitors (39), and elucidates for the first time the mechanism of action of this dual therapeutic strategy in FLT3-ITD AML.

Combined PAD and kinase inhibitor treatment also has therapeutic efficacy in other malignant cell types. Cotreatment with OP449 and dovitinib, an orally active multireceptor TKI, synergistically reduced T-acute lymphoblastic leukemia cell viability, via c-Myc, p-c-Myc (S62), p-ERK1/2, p-AKT, and p70S6 kinase downregulation (40). DT-061 cotreatment sensitized KRAS-mutant lung cancer cells to MEK inhibitor via p-AKT and c-Myc downregulation (41). Correspondingly, PP2A inactivation by recurrent mutation in the scaffolding subunit resulted in resistance to MEK inhibitors (42). In another study, combination of PADs and the TKI afatinib was synergistic in TKI-sensitive and -resistant lung adenocarcinoma cells, with TKI resensitization of resistant cells thought to be due to AKT and MAP kinase (MAPK) pathway inactivation, as acquired TKI resistance commonly occurs through reactivation of PI3K and MAPK pathways downstream of EGFR (43). Finally, PAD and mTOR inhibitor cotreatment decreased pancreatic ductal adenocarcinoma growth through suppression of AKT/mTOR signaling and posttranslational c-Myc protein downregulation (44).

AKT is activated downstream of FLT3-ITD by phosphorylation on S473 and T308 (1). We found that combined PAD and FLT3 inhibitor treatment rapidly inactivates AKT in FLT3-ITD–expressing cells through dephosphorylation of the S473 and T308 residues. The dual PI3K/PDK-1 inhibitor BAG956 was previously shown to be cytotoxic toward cell lines and patient-derived FL3-ITD–expressing AML cells, including Ba/F3-ITD cells resistant to the FLT3 inhibitor midostaurin (PKC412; ref. 36). In a subsequent study using a combinatorial high-throughput drug screen, selective AKT inhibitors, including AT7867, GSK690693, and MK-2206, were found to synergize with FLT3 inhibitors in FLT3-ITD–expressing AML cells in either the absence or the presence of stroma (37). In addition, in two cell lines with induced resistance to sorafenib without acquired FLT3 mutations, and in sorafenib-resistant FLT3-ITD AML patient blasts, FLT3 signaling was inhibited by sorafenib, but the PI3K/mTOR pathway remained activated, and cells were sensitive to the selective PI3K/mTOR inhibitor gedatolisib (45). In other work, STAT5 activation by FLT3-ITD was found to protect cells treated with PI3K/Akt pathway inhibitors from apoptosis by maintaining Mcl-1 expression through the mTORC1/4EBP1/eIF4E pathway (46), potentially supporting efficacy of coinhibition of FLT3 and the PI3K/Akt pathway in FLT3-ITD–expressing cells.

We further found that AKT inactivation by PAD and FLT3 inhibitor cotreatment resulted in posttranslational downregulation of Pim-1 and c-Myc protein expression. We found that posttranslational c-Myc downregulation occurred through T58 phosphorylation, in turn caused by GSK-3β activation, and that treatment with GSK-3β inhibitors replicated the effects of PAD and FLT3 inhibitor cotreatment. In addition, whereas GSK-3β has been found to be a Pim-1 substrate (47), we found here that Pim-1 is a potential GSK-3β substrate, elucidating a potential negative feedback loop, in which Pim-1 inhibits GSK-3β by phosphorylating it on S9 (47), and GSK-3β phosphorylates Pim-1, resulting in its posttranslational upregulation, shown here.

In addition to elucidating a novel regulatory pathway, our findings have potential clinical applicability, supporting the use of PAD and FLT3 inhibitor combinations to treat FLT3-ITD AML. PADs are in development (19). The PAD FTY720 (Fingolimod) is approved by the FDA for treatment of multiple sclerosis (MS) and is thought to have an acceptable risk to benefit ratio for treatment of AML, given its lack of bone marrow toxicity and low rate of serious toxicities in healthy subjects (48) and in patients with MS (49). Novel FTY720 analogs are also in development (50, 51). The orally bioavailable phenothiazine derivative DT-061 is in preclinical development (19). Our findings support potential therapeutic efficacy of PAD and FLT3 inhibitor combinations in AML with FLT3-ITD. They also support potential efficacy of AKT inhibitors and GSK-3β inhibitors in this disease subtype.

FLT3 inhibitor resistance mechanisms include FLT3 tyrosine kinase domain mutations, RAS mutations, and Pim overexpression (3). Dual targeting of FLT3-ITD signaling pathways may overcome FLT3 inhibitor resistance (3). Future work will address PAD cotreatment efficacy in preventing and overcoming FLT3 inhibitor resistance in FT3-ITD AML. Notably, a genome-wide CRISPR screen identified GSK-3 as a gene critical for resistance to quizartinib (52). Expression of GSK-3α was markedly reduced in quizartinib-resistant FLT3-ITD cells, and GSK-3 knockout resulted in quizartinib resistance, through reactivation of FLT3-ITD downstream signaling pathways (52). In addition, GSK-3 inhibits Wnt signaling, and GSK-3 deletion restores it (52). Cotreatment with PAD and FLT3 inhibitor inactivates AKT, which activates GSK-3, and might overcome this mechanism of quizartinib resistance. PAD cotreatment warrants further study as a strategy for preventing or reversing FLT3 inhibitor resistance.

R.G. Lapidus reports being a founder of a start-up called KinaRx, which develops drugs (kinase inhibitors) for AML. J. Sangodkar reports personal fees from RAPPTA Therapeutics outside the submitted work and has no other conflict of interest if the patented use of PADs in the treatment of leukemias and on their killing effect on leukemia progenitors and drug-resistant quiescent leukemia stem cells is excluded. G. Narla reports grants and personal fees from RAPPTA Therapeutics during the conduct of the study; and personal fees and other from RAPPTA Therapeutics outside the submitted work; in addition, G. Narla has a patent related to the development of PP2A activators issued and licensed to RAPPTA Therapeutics. No potential conflicts of interest were disclosed by the other authors.

M. Scarpa: Investigation, methodology, writing–review and editing. P. Singh: Investigation, writing–review and editing. C.M. Bailey: Investigation, writing–review and editing. J.K. Lee: Investigation, writing–review and editing. S. Kapoor: Investigation, writing–review and editing. R.G. Lapidus: Methodology, writing–review and editing. S. Niyongere: Investigation, writing–review and editing. J. Sangodkar: Resources, writing–review and editing. Y. Wang: Investigation, methodology, writing–review and editing. D. Perrotti: Methodology, writing–review and editing. G. Narla: Resources, writing–review and editing. M.R. Baer: Conceptualization, supervision, funding acquisition, writing–original draft, project administration.

The authors gratefully acknowledge Nicole Glynn-Cunningham, University of Maryland Greenebaum Comprehensive Cancer Center, for tissue procurement, Pedram Azimzadeh and Nariman Balenga, Department of Surgery, and Rossana Trotta, Department of Microbiology and Immunology, University of Maryland School of Medicine and Greenebaum Comprehensive Cancer Center, for technical advice.

This study was supported by Merit Review Award BX002184 from the U.S. Department of Veterans Affairs Biomedical Laboratory Research and Development Service (M.R. Baer); NIH-NCI grants RO1 CA163800 (D. Perrotti), R01 CA181654 (G. Narla), R01 CA240993 (G. Narla), and P30 CA134274; NIH-NHLBI grant HL144741R01 (G. Narla); DOD grant W81XWH-19-BCRP-BTA12 (G. Narla); University of Maryland, Baltimore UMMG Cancer Research Grant No. CH 649 CRF issued by the State of Maryland Department of Health and Mental Hygiene under the Cigarette Restitution Fund Program; Rogel Cancer Gift Funds (G. Narla), the Valanda Wilson Leukemia Research Fund; and Mary Ellen's Angelic Fund for Leukemia Research.

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.