In acute myeloid leukemia (AML), chemotherapy resistance remains prevalent and poorly understood. Using functional proteomics of patient AML specimens, we identified MEF2C S222 phosphorylation as a specific marker of primary chemoresistance. We found that Mef2cS222A/S222A knock-in mutant mice engineered to block MEF2C phosphorylation exhibited normal hematopoiesis, but were resistant to leukemogenesis induced by MLL–AF9. MEF2C phosphorylation was required for leukemia stem cell maintenance and induced by MARK kinases in cells. Treatment with the selective MARK/SIK inhibitor MRT199665 caused apoptosis and conferred chemosensitivity in MEF2C-activated human AML cell lines and primary patient specimens, but not those lacking MEF2C phosphorylation. These findings identify kinase-dependent dysregulation of transcription factor control as a determinant of therapy response in AML, with immediate potential for improved diagnosis and therapy for this disease.
Significance: Functional proteomics identifies phosphorylation of MEF2C in the majority of primary chemotherapy-resistant AML. Kinase-dependent dysregulation of this transcription factor confers susceptibility to MARK/SIK kinase inhibition in preclinical models, substantiating its clinical investigation for improved diagnosis and therapy of AML. Cancer Discov; 8(4); 478–97. ©2018 AACR.
This article is highlighted in the In This Issue feature, p. 371
Acute myeloid leukemias (AML) are cancers of the blood that originate in the hematopoietic progenitor cells as a result of the accumulation of genetic mutations that lead to cell transformation. Recent advances in genomic profiling have revealed distinct genetic subsets of AML, including specific mutational classes of cytogenetically normal and chromosomally rearranged leukemias (1–3). Overall, AML is characterized by a prevalence of mutations of genes encoding regulators of gene expression, such as the MLL–AF9 (KMT2A–MLLT3) gene fusion that dysregulates the expression of genes controlling self-renewal, differentiation, and cell survival (4). Recent studies have begun to reveal specific molecular dependencies that can be used for improved targeted therapies in AML (5). In spite of this knowledge, intensive chemotherapy and stem cell transplantation continue to be essential means to achieve cure in the treatment of AML. However, current chemotherapy regimens remain inadequate and fail to induce or sustain remissions in more than 50% of adults and 30% of children with AML (6–8). Thus, improved therapeutic strategies to overcome or block chemotherapy resistance are needed.
Patients with distinct molecularly identifiable types of leukemia have been found to exhibit varying degrees of response to chemotherapy, leading to therapy intensification for patients with high-risk disease (9–12). However, human leukemias also exhibit large heterogeneity of chemotherapy response, and its molecular determinants continue to remain poorly understood. For example, mutations of DNMT3A, as well as gene fusions such as MLL–AF9 and NUP98–NSD1, have been found to contribute to chemotherapy resistance (13–15). However, patients with leukemias with these and other mutations can nonetheless achieve durable remissions (3). Likewise, leukemias that exhibit primary chemotherapy resistance that is refractory to induction chemotherapy have not been found to show significant enrichment for these or other single-gene mutations (9, 11, 16). Thus, additional genetic or molecular mechanisms must cause chemotherapy resistance in AML.
In support of this idea, human leukemias and their genetically engineered mouse models are characterized by distinct cell populations, comprising defined functional compartments such as leukemia stem or initiating cells that exhibit unique phenotypic properties, including self-renewal and enhanced cell survival (17, 18). In part, these behaviors are caused by co-option of developmental programs that regulate normal hematopoiesis, such as for example self-renewal and therapy resistance of MLL–AF9 leukemias (4, 19). In particular, MEF2C, a member of the MADS family of transcription factors, normally regulates hematopoietic self-renewal and differentiation, supports the proliferation of MLL–AF9 leukemias, and is associated with the increased risk of relapse when highly expressed in multiple subtypes of AML in patients (19–25). However, the precise molecular mechanisms by which MEF2C is dysregulated in AML are not known. Recurrent mutations of MEF2 family members, including MEF2B, MEF2C, and MEF2D in refractory lymphoblastic leukemias and lymphomas (26–28), suggest that MEF2C and its orthologs may regulate a general mechanism of therapy resistance.
Here, using recently developed high-accuracy mass spectrometry techniques, we determined phospho-signaling profiles of human AML specimens collected at diagnosis from patients with primary chemotherapy resistance and failure of induction chemotherapy. Analysis of these profiles revealed high levels of phosphorylation of S222 of MEF2C, which was found to be significantly associated with primary chemotherapy resistance in an independent cohort of cytogenetically normal and MLL-rearranged leukemias. By integrating genome editing and biochemical and cell biological approaches, we tested the hypothesis that MEF2C phosphorylation promotes chemotherapy resistance and that its blockade can be leveraged for improved AML therapy. These studies have revealed an unexpected dependence on kinase-dependent dysregulation of transcription factor control as a determinant of therapy response in AML, with immediate potential for translation into improved diagnosis and therapy for this disease.
Phosphorylation of S222 in MEF2C Is a Specific Marker of AML Chemotherapy Resistance
Previously, we assembled a cohort of primary AML specimens matched for AML subtypes and therapy and collected at diagnosis from patients with failure of induction chemotherapy and those who achieved remission after two cycles of cytarabine and daunorubicin-based induction chemotherapy (16). In this analysis, we found that defined gene mutations were associated with primary chemotherapy resistance only in a minority of cases. Thus, we sought to investigate alternative molecular mechanisms that may explain primary chemotherapy resistance in AML.
We focused on phospho-signaling because kinase activation is one of the hallmarks of AML pathogenesis (29, 30). Recent advances in quantitative proteomics, particularly in high-efficiency, multidimensional fractionation platforms (31), enable in-depth analysis of signaling molecules from rare cell populations (32). Leukemia cells purified from a discovery cohort of eight diagnostic adult AML bone marrow aspirate specimens with normal karyotypes (Supplementary Table S1) were analyzed by metal affinity chromatography (IMAC; ref. 33) and isobaric tagging (iTRAQ) mass spectrometry (34). This yielded 2,553 unique phosphopeptides, 34 of which were significantly enriched in induction failure specimens (Supplementary Data S1; Supplementary Fig. S1A and S1B). We identified phosphorylation of serine 222 (pS222) in MEF2C among the top 20 most highly abundant phosphoproteins in induction failure specimens as compared with age, therapy, and disease-matched remission specimens (P = 5.0 × 10−3, t test, Fig. 1A and B; Supplementary Fig. S1B).
MEF2C was the preferred candidate to study given its known function as an oncogene in lymphoid malignancies (35) and cooperation with MLL-rearranged AML (19). The observed MEF2C pS222 peptide was distinct from the related MEF2A, MEF2B, and MEF2D peptides (Supplementary Fig. S1C). We also observed phosphorylation of S396 MEF2C (Supplementary Data S1), but this modification was present in substoichiometric amounts, as measured using quantitative targeted mass spectrometry (36). In addition, this analysis revealed phosphoproteins previously implicated in therapy resistance, such as HSBP1 pS15 (37), as well as those not previously observed but likely functional, such as HGF pT503 (29).
To investigate the diagnostic significance of MEF2C pS222 in primary AML chemotherapy resistance, we assessed the prevalence of MEF2C pS222 in an independent cohort of 47 pediatric and adult primary AML specimens matched for age, disease biology, and therapy (Supplementary Table S2; ref. 16). As previously observed, this cohort included specimens with gene mutations associated with poor prognosis such as those with cryptic rearrangements of MLL/KMT2A and combined DNMT3A and NPM1 mutations (Supplementary Table S2). We developed an affinity-purified antibody against MEF2C pS222 and validated its specificity using synthetic peptide competition and phosphatase assays (Supplementary Fig. S2A–S2C). We found that expression of MEF2C pS222 and total MEF2C were significantly associated with induction failure and primary chemotherapy resistance (P = 6.5 × 10−4 and 6.0 × 10−3, respectively; Fig. 1C and D; Supplementary Fig. S2D–S2G), remaining statistically significant in multivariate analyses (Supplementary Table S3). Overall, pS222 and total MEF2C expression levels were high in the induction failure as compared with complete remission specimens (Supplementary Fig. S2H and S2I) and significantly associated with induction failure as compared with disease relapse (P = 2.7 × 10−2; Fig. 1E). In addition, presence of MEF2C pS222 above its median expression level correlated with reduced overall event-free survival (EFS; P = 3.8 × 10−2; Fig. 1F) and was a significant binary predictor of poor outcome as assessed by the ROC analysis (P = 3.2 × 10−2; Fig. 1G). In addition, we analyzed MEF2C expression data from the ECOG E1900 cohort of young adult AML and found that high MEF2C expression correlated with reduced EFS (P = 3.8 × 10−2; Supplementary Fig. S2K). In contrast, the highly homologous MEF2D did not exhibit this association (P = 0.98; Fig. 1D; Supplementary Fig. S2J). Post hoc review of the top 10 specimens with moderate MEF2C pS222 expression from remission cases at diagnosis revealed that 6 of these cases had in fact developed early relapse after initial induction therapy treatment (Fig. 1C), corroborating the specific association of MEF2C pS222 with AML chemotherapy resistance and suggesting that MEF2C phosphorylation is necessary but not sufficient to cause chemotherapy resistance. Overall, these findings indicate that MEF2C pS222 is a specific marker of primary chemoresistance and failure of induction therapy in AML.
MEF2C pS222 Is Dispensable for Normal Hematopoiesis
The activity of MEF2C is known to be regulated by posttranslational modifications, including acetylation, sumoylation, and phosphorylation, that can affect the recruitment of transcriptional corepressors and coactivators (38–40). Analysis of existing phospho-signaling data revealed the presence of MEF2C S222 phosphorylation in human K562 leukemia cells and cytokine-stimulated hematopoietic progenitor cells (41, 42). The specific association of MEF2C pS222 with failure of induction chemotherapy raises the possibility that MEF2C S222 phosphorylation promotes chemotherapy resistance in AML. To test this hypothesis, we first sought to investigate the potential function of MEF2C S222 phosphorylation in normal hematopoiesis. Thus, we engineered knock-in mice harboring a loss-of-function Mef2c S222A allele that cannot be phosphorylated, and a gain-of-function S222D allele that mimics phospho-serine by using CRISPR/Cas9 genome editing (Fig. 2A; Supplementary Fig. S3A). Genotyping of founder animals identified Mef2cS222A and Mef2cS222D mutant alleles in 10% and 7% of the mice born, respectively. We confirmed the absence of apparent off-target mutations of the Mef2c locus by genomic DNA sequencing of each founder animal. To control for possible off-target effects, we obtained two independent founder strains for both Mef2cS222A and Mef2cS222D alleles, and back-crossed them to wild-type C57BL/6J mice. Subsequently, Mef2cS222A/S222A and Mef2cS222D/S222D mice obtained from heterozygous intercrosses were detected at expected Mendelian ratios (Supplementary Fig. S3B) and exhibited approximately equal MEF2C protein expression, as measured by western immunoblotting of B220+ and Gr-1+/CD11b+ bone marrow cells, which are known to have relatively high and low MEF2C expression, respectively (Fig. 2B). Mef2cS222A/S222A and Mef2cS222D/S222D animals exhibited normal growth as compared with wild-type littermates (Supplementary Fig. S3C and S3D) and had no discernible defects on histologic analysis of tissues where physiologic activities of MEF2C are required for normal development (43, 44), including brain, heart, skeletal muscle, and spleen.
Consistent with the dispensability of MEF2C S222 phosphorylation for normal hematopoiesis, Mef2cS222A/S222A and Mef2cS222D/S222D mice exhibited normal peripheral blood cell counts and morphologies (Supplementary Fig. S3E and S3F). Likewise, we observed no significant differences in the absolute numbers of mature blood cells from myeloid and lymphoid lineages in the bone marrow of Mef2cS222A/S222A and Mef2cS222D/S222D mice, as compared with wild-type littermates (Fig. 2C and D; Supplementary Fig. S3G). Similarly, Mef2cS222A/S222A and Mef2cS222D/S222D bone marrow granulocyte–macrophage progenitor (GMP) cells exhibited preserved clonogenic capacities as compared with wild-type Mef2c GMP cells from littermate control mice (Supplementary Fig. S3H). Likewise, CD45.2+ Mef2cS222A/S222A and Mef2cS222D/S222D hematopoietic stem cells displayed preserved repopulation capacities in competitive bone marrow transplants with wild-type CD45.1+ cells in lethally irradiated mice, as monitored by CD45.2+/CD45.1+ chimerism of peripheral blood cells in recipient mice as compared with wild-type controls over 16 weeks after transplantation (Fig. 2E and F; Supplementary Fig. S3I–S3J). Thus, MEF2C S222 phosphorylation is dispensable for normal steady-state and stress mouse hematopoiesis.
MEF2C Phosphorylation Is Required for MLL–AF9 Leukemogenesis
Overexpression of Mef2c itself is not sufficient to cause leukemia, but potently cooperates with other leukemia oncogenes (45). In particular, Mef2c is required for the maintenance of MLL-rearranged mouse leukemias (19, 23) and high MEF2C expression is associated with MLL-rearranged primary AML (23), as supported by the finding of three cryptic MLL rearrangements, including MLL–AF9 fusions, in our cohort (Supplementary Table S2). Thus, we reasoned that MEF2C phosphorylation may be required for MLL–AF9-induced leukemogenesis.
To test this hypothesis, we assessed leukemia induction by retroviral MLL–AF9 in bone marrow GMP cells of Mef2cS222A/S222A, Mef2cS222D/S222D, or wild-type littermates. Bone marrow GMP cells were transduced with the MSCV-IRES-GFP retrovirus encoding MLL–AF9, and GFP-expressing cells were purified by FACS and plated in methylcellulose for clonogenic assays (Fig. 2G). We observed a significant reduction of clonogenic efficiency in serial replating of the Mef2cS222A/S222A progenitor cells transformed with MLL–AF9, as compared to wild-type controls (P = 3.3 × 10−3, t test, Fig. 2G). We observed the same phenotype in clonogenic assays of bone marrow GMP cells from two independent Mef2cS222A/S222A founder strains (Supplementary Fig. S3K), consistent with the specific effects of Mef2c S222A.
To test the effects of MEF2C S222 phosphorylation on leukemia initiation in vivo, MLL–AF9 transduced cells were transplanted into lethally irradiated C57BL/6J mice by intravenous tail-vein injection (Supplementary Fig. S3L). Recipient mice transplanted with MLL–AF9-transduced cells derived from wild-type littermate control and Mef2cS222D/S222D mice became moribund at approximately 80 days after transplant with more than 90% of the bone marrow replaced with GFP+ leukemia cells (Fig. 2H; Supplementary Fig. S3M). In contrast, mice transplanted with MLL–AF9-transduced Mef2cS222A/S222A cells survived for more than 170 days after transplant without physical signs of leukemia development (P = 6.8 × 10−9, log-rank test; Fig. 2H). We observed no GFP-expressing cells in the peripheral blood and bone marrow of these mice (Supplementary Fig. S3M), consistent with the essential function for MEF2C pS222 in leukemia stem cell homing or engraftment, as supported by prior studies (23). Thus, MEF2C S222 phosphorylation is required for leukemia initiation by the MLL–AF9 oncogene.
MEF2C Phosphorylation Cooperates with Distinct Leukemia Oncogenes and Is Required for Leukemia Stem Cell Maintenance
Protection of Mef2cS222A/S222A hematopoietic progenitor cells from MLL–AF9-induced leukemogenesis suggests that MEF2C S222 phosphorylation may be necessary for the survival of MLL–AF9-transformed leukemia cells. To investigate the function of MEF2C S222 phosphorylation in MLL–AF9-induced leukemia maintenance, we transformed wild-type bone marrow GMP cells by retroviral transduction of MSCV-IRES-GFP MLL–AF9 and coexpressed wild-type MEF2C, S222A, or S222D mutants using the MSCV-IRES-tdTomato (MIT) retrovirus (Supplementary Fig. S4A). This allowed us to purify GFP/tdTomato double-positive MLL–AF9-transformed cells using FACS where MEF2C transgenes can function as dominant mutants because of their dimerization with endogenous MEF2C (46, 47). We confirmed that all MEF2C transgenes were expressed at approximately equal levels by western immunoblotting, with MEF2C S222A–transduced cells exhibiting substantially reduced but not completely eliminated levels of pS222 as compared with wild-type controls (Supplementary Fig. S4B). We found that MLL–AF9 leukemia cells expressing MEF2C S222A exhibited transient reduction in clonogenic efficiency in vitro (Supplementary Fig. S4C) and displayed increased apoptosis (Supplementary Fig. S4D) as compared with cells expressing wild-type MEF2C. MEF2C S222D–expressing cells behaved similarly to MEF2C wild-type–expressing cells, similar to the phenotype of MLL–AF9-transduced Mef2cS222D/S222D leukemias, indicating that MEF2C phosphorylation is necessary but insufficient to cause enhanced leukemia cell survival. Notably, when allowed to develop leukemia (Supplementary Fig. S4E) and secondarily transplanted into sublethally irradiated mice to assess leukemia maintenance in vivo, MLL–AF9;MEF2C S222A leukemias were significantly impaired in leukemia development as compared with MLL–AF9;MEF2C wild-type leukemias (P = 2.2 × 10−10, log-rank test, Fig. 3A). Using limiting dilution analysis, we found that the expression of wild-type phosphorylated MEF2C increased the leukemia initiating cell frequency as compared with MIT control (1/16 vs. 1/85) and consistent with previous reports (48), whereas the expression of MEF2C S222A significantly blocked this effect (1/139, P = 6.7 × 10−8, χ2 test, Fig. 3B; Supplementary Fig. S4F–S4H). These results indicate that MEF2C S222 phosphorylation is required for the enhanced maintenance of leukemia stem cells and its cooperation with MLL–AF9 leukemogenesis in vivo.
To investigate MEF2C S222 phosphorylation in non–MLL-rearranged leukemias, we assessed its function in leukemias with inactivating mutations of RUNX1 and internal tandem duplication (ITD) mutations of FLT3, a leukemia subtype associated with high rates of therapy failure that also exhibits high levels of pS222 MEF2C. First, we obtained genetically engineered mouse Runx1−/−;Flt3ITD leukemias (49), which we transduced with wild-type MEF2C and transgenes modeling loss and gain of S222 phosphorylation using the MIT retrovirus (Supplementary Fig. S5A–S5C). We found that Runx1−/−;Flt3ITD leukemias expressing MEF2C S222A displayed significantly reduced clonogenic efficiency in serial replating as compared with cells transduced with wild-type or S222D MEF2C (P = 1.9 × 10−5, t test, Fig. 3C). Similar to the MLL–AF9 leukemia model, we observed that mice transplanted with secondary Runx1−/−;Flt3ITD;MEF2C wild-type leukemia cells exhibited accelerated leukemia development as compared with Runx1−/−;Flt3ITD MIT control leukemias, which was blocked in Runx1−/−;Flt3ITD;MEF2C S222A–expressing leukemias (P = 3.1 × 10−3, log-rank test; Fig. 3D). In contrast, we found that MEF2C S222 phosphorylation was dispensable for leukemias induced by ectopic retroviral expression of Hoxa9 and Meis1 (Supplementary Fig. S5D–S5H), consistent with the epistatic function of aberrant Hox gene expression in AML (50, 51). Thus, MEF2C S222 phosphorylation is required for maintenance of Runx1−/−;Flt3ITD leukemias.
MEF2C Phosphorylation Is Required for Leukemia Maintenance in MLL-Rearranged Human AML Cells
To assess the function of MEF2C S222 phosphorylation in human AML, we analyzed its expression in a panel of human AML cell lines and identified both MLL-rearranged and nonrearranged leukemias with activated levels of MEF2C pS222, including OCI-AML2, MOLM-13, K562, and Kasumi-1 (Supplementary Fig. S6A and S6G). Thus, we used a doxycycline-inducible lentivirus vector to express wild-type and dominant-negative MEF2C S222A and S222D mutants and confirmed near-physiologic transgene expression and pS222 levels by western immunoblotting (Fig. 3E and F; Supplementary Fig. S6B). We observed that OCI-AML2 and MOLM-13 AML cell lines with endogenous MEF2C S222 phosphorylation had significantly reduced viability upon the doxycycline-induced expression of MEF2C S222A as compared with wild-type MEF2C (P = 8.4 × 10−3 and 3.4 × 10−3, t test, respectively; Fig. 3G) and S222D MEF2C, at least in part due to the induction of apoptosis (Supplementary Fig. S6C and S6D) and reduction of cells in S phase (Supplementary Fig. S6E and S6F), in agreement with prior reports of cell cycle–dependent regulation of MEF2C (52, 53). In contrast, HL-60 and U937 AML cells that lack MEF2C expression exhibited no significant effects upon expression of wild-type or mutant MEF2C S222A (Fig. 3G). Overall, these results indicate that MEF2C S222 phosphorylation is required specifically for the survival of MEF2C-activated AML cells but not those lacking MEF2C expression.
To investigate the function of MEF2C phosphorylation in human leukemias in vivo, we next assessed the effects of inhibiting MEF2C S222 phosphorylation by expression of MEF2C S222A in OCI-AML2 cells transplanted orthotopically by tail-vein injection in immunodeficient mice (Fig. 3H). Cells were transplanted into NOD-SCID-IL2Rcγnull (NSG) mice, and transgene expression was induced 3 days after transplant using doxycycline chow. Mice transplanted with OCI-AML2 cells expressing dominant-negative MEF2C S222A had significantly prolonged survival and impaired leukemia development as compared with those transplanted with cells expressing wild-type MEF2C or nontransduced control OCI-AML2 cells (P = 3.5 × 10−6, log-rank test; Fig. 3H). These findings indicate that MEF2C S222 phosphorylation is required for the survival of human AML cells in vitro and in mouse xenografts in vivo.
MEF2C Phosphorylation Is Required for MEF2C-Mediated Gene Repression Program
MEF2C functions as a transcription factor by sequence-specific recognition of MEF2 response elements (54), through regulated recruitment of transcriptional coactivators and corepressors via its transactivation (residues 118–473) domain (38, 55). To test the hypothesis that MEF2C S222 phosphorylation regulates MEF2C transcriptional activity, we assessed the transactivation of MEF2 response element–driven firefly luciferase as compared with a cytomegalovirus promoter–driven Renilla luciferase control introduced by electroporation in K562 cells. To model the effects of MEF2C S222 phosphorylation, we used lentivirus transduction to generate K562 cells expressing wild-type or dominant-negative S222A MEF2C upon doxycycline treatment (Supplementary Fig. S6G–S6H). As expected, expression of wild-type MEF2C led to an 8-fold increase in MEF2C transcriptional activity as compared with unmodified control K562 cells, which was rescued by mutant S222A protein expression (P = 3.1 × 10−3, t test; Supplementary Fig. S6I). We confirmed these findings in HEK 293T cells stably expressing wild-type and mutant MEF2C S222A (Fig. 4A). Expression of the wild-type MEF2C in HEK293T cells that lack endogenous MEF2C led to an 18-fold increase in MEF2C transcriptional activity as compared with the vector control (P = 6.3 × 103, t test, Fig. 4A). Despite equal levels of wild-type MEF2C and mutant S222A protein expression (Fig. 4B), MEF2C transcriptional activity was significantly reduced due to the effects of MEF2C S222 phosphorylation in the S222A mutant (P = 4.0 × 10−2, t test; Fig. 4A and B). This suggests that MEF2C S222 phosphorylation is required for maximal activation of MEF2C-dependent gene transcription.
To elucidate the gene expression program controlled by MEF2C S222 phosphorylation in AML cell survival, we analyzed gene expression profiles of OCI-AML2 cells expressing wild-type as compared with dominant-negative MEF2C S222A using RNA sequencing (RNA-seq; Fig. 4C). Consistent with the near-physiologic expression of MEF2C transgenes (Fig. 3E), we found essentially no significant differences in gene expression profiles of cells expressing wild-type MEF2C upon doxycycline-induced transgene expression (Fig. 4C). In contrast, we observed a significant change in gene expression of cells upon doxycycline-induced expression of MEF2C S222A, as compared with both uninduced cells and cells expressing wild-type MEF2C (Fig. 4C). In particular, we found 276 genes that were significantly altered in expression in cells expressing MEF2C S222A as compared with uninduced cells (Fig. 4C; Supplementary Data S2). Using qRT-PCR, we found that none of the previously reported canonical MEF2C target genes, such as Nr4a1/Nur77, Hdac7, Jun, and Cebpa, were significantly changed in expression upon expression of MEF2C S222A (Supplementary Fig. S4I–S4L). Instead, gene set enrichment analyses (GSEA) revealed that loss of MEF2C S222 phosphorylation led to downregulation of distinct gene expression programs, including E2F and MYC target genes (Fig. 4D; Supplementary Data S2). In addition, we observed altered expression of genes regulating apoptosis and cell cycle (Fig. 4D and E; Supplementary Data S2), consistent with the phenotypic induction of apoptosis and cell-cycle defects upon the blockade of MEF2C S222 phosphorylation in mouse and human leukemia cells (Supplementary Fig. S4D and S6C–S6F).
To identify genes specifically regulated by S222 phosphorylated MEF2C, we profiled chromatin accessibility using the assay for transposase accessible chromatin with sequencing (ATAC-seq), annotated for the presence of canonical MEF2 binding sites, in OCI-AML2 cells expressing wild-type and S222A-mutant MEF2C upon 48 hours of doxycycline-induced transgene expression. This approach was necessary, because mapping MEF2C genome localization using chromatin immunoprecipitation with sequencing (ChIP-seq) was not possible due to the inadequate chromatin enrichment achieved with currently available MEF2C-specific antibodies. Using ATAC-seq analysis annotated for the presence of MEF2 binding sites as a surrogate of MEF2C binding, we identified genes that exhibited significant changes in chromatin accessibility of MEF2 binding sites and concomitantly had significant changes in their expression, as measured using RNA-seq (Fig. 4F; Supplementary Fig. S6J and Data S2). This analysis identified a total of 262 genes, including 18 genes repressed by the expression of MEF2C S222A, consistent with their aberrant induction by S222 phosphorylated MEF2C, such as MYC and LYL1 (marked in blue, Fig. 4F). Observed gene expression changes in LYL1 repression were verified by protein immunoblotting (Fig. 4G), potentially explaining previous findings of MEF2C-dependent LYL1 expression and its aberrant expression in AML (56–58). Similarly, we also observed 159 genes that were upregulated by MEF2C S222A expression, consistent with their repression by phosphorylated MEF2C in AML cells. These genes included BCOR, BCL6, and NCOA3, as well as apoptotic regulators BMF, CASP8, and FOXO3 (marked in red, Fig. 4F). In agreement with the MEF2C S222 phosphorylation–induced suppression of these proapoptotic factors, we found that OCI-AML2 cells expressing MEF2C S222A exhibited enhanced apoptotic priming compared to MEF2C wild-type cells as assessed by BH3 profiling flow cytometry, with particular sensitivity to BIM and BID (Fig. 4H), consistent with their reported interactions with BMF, CASP3, and FOXO3 (59–62).
Homodimerization or heterodimerization of the MEF2 family of transcription factors is a key mechanism of their regulation, which can be modulated by posttranslational modifications (46, 63). To investigate whether MEF2C S222 phosphorylation can regulate its oligomerization, we treated OCI-AML2 cells induced to express MEF2C wild-type or S222A with disuccinimidyl suberate (DSS), a cell-permeable amine-reactive cross-linking agent, to label and capture oligomeric MEF2C complexes. Analysis of DSS-treated cells revealed the formation of a reduced mobility complex, consistent with dimerization of MEF2C, which was significantly upregulated upon the loss of MEF2C S222 phosphorylation in MEF2C S222A–expressing cells as compared with wild-type controls (P = 7.9 × 10−4, t test; Fig. 4I; Supplementary Fig. S6K), consistent with previous studies of dominant-negative MEF2 mutants (46). Thus, MEF2C S222 phosphorylation controls the assembly of its transcriptionally active complex, which is associated with induction of both activating and repressive gene expression programs, leading to the dysregulation of genes controlling leukemia cell survival such as LYL1, MYC, BMF, CASP8, and FOXO3.
MARK-Mediated Phosphorylation of MEF2C S222 Confers Susceptibility to MARK/SIK Inhibitors in AML
The functional requirements of MEF2C S222 phosphorylation for leukemia initiation and maintenance, combined with its dispensability for normal hematopoiesis, suggest that blockade of MEF2C phosphorylation may be exploited for antileukemic therapy. Given that direct inhibition of transcription factors is pharmacologically challenging, we reasoned that inhibition of the upstream kinases phosphorylating MEF2C at S222 could downregulate MEF2C signaling to impair aberrant AML cell survival. To identify candidate kinases that can phosphorylate S222 of MEF2C, we screened a library of 172 recombinant human serine kinases in their ability to phosphorylate a model substrate in the presence of synthetic pS222-containing peptide as a putative competitive product inhibitor (Fig. 5A; Supplementary Data S3). We found that only five protein kinases scored significantly in this screen, of which four belonged to a closely related microtubule-associated protein/microtubule affinity regulating kinase (MARK) 1, 2, 3, 4 family (Fig. 5A). Analysis of primary AML specimens revealed that MARK3 and MARK2 genes are highly expressed in most subtypes of AML (Supplementary Fig. S7A–S7B), with high MARK3 expression in particular significantly associated with inferior survival among adults with AML (P = 1.7 × 10−3, log-rank test; Supplementary Fig. S7C). We validated the specific kinase activity of recombinant MARK3 with synthetic MEF2C pS222 peptide, as compared with the closely related CAMK1α kinase as specificity control (Fig. 5B). These results indicate that MARK3 and related MARK kinase family members can specifically phosphorylate S222 MEF2C in vitro.
To evaluate whether MARK3 could phosphorylate MEF2C S222 and regulate its transcriptional activity in cells, we coexpressed MEF2C and MARK3 in HEK293T cells and assessed MEF2C phosphorylation and transcriptional activity by western immunoblotting and transcriptional reporter assays, respectively (Fig. 5C). We found that MEF2C transcriptional activity was significantly increased upon coexpression of MEF2C and MARK3 as compared with cells expressing MEF2C alone (P = 3.4 × 10−5, t test; Fig. 5C). This increased transcriptional reporter activity was associated with increased levels of pS222, which was blocked by the coexpression of the enzymatically impaired MARK3 T211A activation loop mutant (refs. 64, 65; P = 5.0 × 10−5, Fig. 5C). Importantly, expression of MARK3 failed to induce S222 phosphorylation or activate the transcriptional activity of the MEF2C S222A mutant that cannot be phosphorylated (Fig. 5C). Taken together, these data indicate that MARK kinase signaling can regulate MEF2C transcriptional activity specifically in an S222 phosphorylation–dependent manner.
Next, we tested the hypothesis that inhibition of MARK-mediated phosphorylation of MEF2C may have antileukemic efficacy. Genetic depletion of MARK3 using shRNA interference was not sufficient to reduce MEF2C pS222 phosphorylation (Supplementary Fig. S7D), consistent with potentially redundant functions of MARK3, MARK2, and MARK4. Therefore, we identified the ATP-competitive kinase inhibitor MRT199665, which exhibits high selectivity and potency against MARK and the structurally similar SIK and MELK kinases as compared with other AMP kinase family members (66). We found that SIK, NUAK, and MELK kinases, which can be inhibited by MRT199665 at nanomolar concentrations, scored in the bottom 45% of the recombinant kinase screen for MEF2C phosphorylation (Supplementary Data S3). Thus, we reasoned that MRT199665 can be used as a selective inhibitor to block MEF2C S222 phosphorylation. To test this directly, we first investigated whether MRT199665 treatment can phenocopy the suppression of MEF2C phosphorylation–induced gene expression program by the dominant-negative MEF2C S222A mutant. Transcriptome profiling of OCI-AML2 cells treated with 100 nmol/L MRT199665 for 48 hours revealed significant gene expression changes as compared with vehicle control-treated cells, where 191 and 128 genes were upregulated and downregulated, respectively (Supplementary Fig. S7E and Data S4). GSEA revealed that MRT199665-treated cells and MEF2C S222A–expressing cells exhibit similar changes in gene expression (r = 0.64, Fig. 5D), including downregulation of cell-cycle and MYC target genes and upregulation of genes mediating apoptosis (Fig. 5E; Supplementary Data S4). Furthermore, treatment of OCI-AML2 and MOLM-13 cells with increasing concentrations of MRT199665 led to a dose-dependent reduction in total and pS222 MEF2C (Fig. 5F; Supplementary Fig. S7F), causing more than 40% reduction in MEF2C phosphorylation at 10 nmol/L as compared with untreated cells, as assessed by quantitative fluorescent western immunoblotting (Fig. 5G; Supplementary Fig. S7G). MRT199665 treatment also caused a decrease of total MEF2C protein expression, suggesting that its phosphorylation can affect MEF2C stability or degradation, similar to previous reports (53). Notably, we observed that MRT199665 treatment led to dose-dependent loss of MEF2C S222 phosphorylation in OCI-AML2 cells expressing wild-type MEF2C, but not MEF2C S222A, consistent with its specific effects on MEF2C S222 phosphorylation (Supplementary Fig. S7H–S7I). Thus, MRT199665 treatment can block MEF2C S222 phosphorylation in AML cells.
Consistent with this mechanism, we found that human AML cell lines with endogenous MEF2C phosphorylation (OCI-AML2, MV4-11, MOLM-13, and Kasumi-1) were more sensitive to MRT199665 as compared with cell lines lacking MEF2C (NB-4, HEL, HL-60, and U937), with mean 50% inhibitory concentrations (IC50) of 26 ± 13 versus 990 ± 29 nmol/L, respectively (P = 5.6 × 10−5, t test, Fig. 5H), and displayed reduced leukemia growth in clonogenic assays (Supplementary Fig. S7J). Importantly, expression of MEF2C in OCI-AML2 cells conferred enhanced susceptibility to MRT199665, which was partially rescued by the expression of MEF2C S222D that models constitutive MEF2C S222 phosphorylation (Fig. 5I). Likewise, we found that primary patient AML specimens with MEF2C activation including MLL-rearranged and non–MLL-rearranged leukemias (Supplementary Table S4 and Supplementary Fig. S8A) treated in short-term cultures ex vivo exhibited enhanced susceptibility and apoptosis in response to MRT199665 treatment as compared with AML cell lines lacking MEF2C expression (IC50 = 280 ± 136 vs. 1300 ± 360 nmol/L, respectively, P = 1.8 × 10−5, t test, Fig. 5J; Supplementary Fig. S8B–S8D). Thus, activation of MEF2C in human AML confers susceptibility to MARK/SIK kinase inhibition.
MEF2C Phosphorylation Is Required for Chemotherapy Resistance in AML
Because MEF2C phosphorylation was specifically observed in diagnostic AML specimens in patients with failure of induction therapy (Fig. 1), we hypothesized that MEF2C S222 phosphorylation induces chemotherapy resistance. In agreement with this prediction, OCI-AML2 cells induced to express dominant-negative MEF2C S222A were significantly more sensitive to both cytarabine and doxorubicin treatment as compared with cells expressing wild-type MEF2C, MEF2C S222D, or untransduced control OCI-AML2 cells (P = 8.8 × 10−8 and P = 3.4 × 10−9, respectively, χ2 test; Fig. 6A and B). To investigate this therapeutic effect in vivo, we transplanted OCI-AML2 cells induced to express MEF2C S222A as compared with wild-type MEF2C into NSG mice, induced expression of transgenes by doxycycline chow, and treated leukemic mice with cytarabine or PBS control (Fig. 6C and D). We found that mice with leukemias expressing MEF2C S222A were significantly sensitized to cytarabine treatment in vivo as compared with MEF2C wild-type expressing controls (P = 6.8 × 10−5, log-rank test; Fig. 6D).
Because MEF2C phosphorylation is required for AML chemotherapy resistance, we investigated the antileukemic efficacy of MARK/SIK kinase inhibition in combination with chemotherapy. Because MRT199665 exhibits rapid metabolism and elimination in mice and consequently is not suitable for animal studies (67), its effects on chemotherapy resistance could only be assessed in vitro. We found that human AML cell lines with endogenous MEF2C phosphorylation (OCI-AML2, MV4-11, MOLM-13, and Kasumi-1) were significantly sensitized to cytarabine in the presence of 100 nmol/L MRT199665, as compared with cell lines lacking MEF2C (NB-4, HEL, HL-60, and U937) that remained unaffected by MRT199665 (P = 0.024, paired t test; Fig. 6E). Similarly, patient-derived primary AML specimens with MEF2C activation (2 MLL-rearranged and 2 non–MLL-rearranged) exhibited enhanced susceptibility to cytarabine treatment in the presence of MRT199665 in vitro (IC50 = 58 ± 76 vs. 230 ± 280 nmol/L in the presence vs. absence of MRT199665, respectively, P = 0.036, t test, Fig. 6F), in contrast to healthy CD34+ human blood umbilical cord progenitor cells, which showed no sensitization (Fig. 6F). Taken together, these results indicate that MEF2C phosphorylation is required for AML chemotherapy resistance, which can be blocked by MARK/SIK kinase inhibition.
Our current findings indicate that MEF2C S222 phosphorylation is a specific marker of primary chemotherapy resistance and failure of induction chemotherapy in patients with both cytogenetically normal and chromosomally rearranged AMLs (Fig. 1). Activity of S222-phosphorylated MEF2C appears to be aberrant, insofar as mice genetically engineered to block its phosphorylation exhibit normal steady-state and stress hematopoiesis, but are resistant to leukemogenesis induced by the MLL–AF9 oncogene in vivo (Fig. 2). At least in part, this effect is due to the requirement for MEF2C phosphorylation for the survival of leukemia stem or initiating cells (Fig. 3) and enhanced transcriptional activity and induction of gene expression programs regulating cell survival and apoptosis (Fig. 4). Finally, MARK kinases can specifically phosphorylate MEF2C, potentiating its transcriptional activity (Fig. 5), and inhibition of MEF2C phosphorylation can overcome chemotherapy resistance of MEF2C-activated human AML cell lines and patient leukemias (Fig. 6).
Resistance to chemotherapy remains the major barrier to improving clinical outcomes for patients with AML, yet our current concepts of chemoresistance lack sufficient explanatory power. For example, inactivation of TP53 and overexpression of xenobiotic transporters have been found to cause chemoresistance, but the majority of patients with primary refractory or relapsed AML lack identifiable mutations of TP53 or known components of TP53-dependent DNA damage response, and do not exhibit xenobiotic transporter overexpression (1–3). Likewise, we do not yet understand how recently identified groups of genetic mutations associated with inferior clinical outcomes cause chemotherapy resistance (3). Recently, impaired susceptibility to mitochondrial apoptosis has been associated with chemoresistance in AML (68, 69). However, the molecular mechanisms responsible for these differences also remain poorly understood. Our findings identify kinase-dependent dysregulation of transcription factor control as a determinant of therapy response in AML, at least in part mediated by aberrant survival and apoptosis resistance of leukemia stem cells (Fig. 7). This suggests that varied mechanisms of chemotherapy resistance and disease relapse in patients may be ultimately caused by aberrant gene expression programming of privileged leukemia cell subsets that provide a chemoresistant disease reservoir (70).
The observation of MEF2C S222 phosphorylation in genetically diverse AML subtypes, including cytogenetically normal and MLL-rearranged leukemias, raises the possibility that common gene expression programs and regulatory mechanisms may be engaged by distinct genetic and molecular classes of AML, such as for example distinct groups of mutations and/or their induction in specific leukemia-initiating cell populations. Likewise, recurrent mutations of MEF2B, MEF2C, and MEF2D in refractory lymphoid cancers (26–28) suggest that MEF2 family members may regulate essential survival or homeostatic mechanisms in hematopoietic cells, which cause therapy resistance in myeloid and lymphoid malignancies. However, whereas mutational activation of MEF2C in refractory T-cell acute lymphoblastic leukemias appears to inhibit apoptosis via NR4A1/NUR77-mediated effects on BCL2 (35), apoptosis resistance induced by MEF2C phosphorylation in AML appears to involve BMF, CASP8, and FOXO3 dysregulation of BIM and BID (Fig. 4). Additionally, MEF2C phosphorylation–dependent regulation of LYL1 expression suggests that its oncogenic functions in AML may contribute to the co-option of hematopoietic stem cell programs (71), at least in part mediated by aberrant transcription factor complexes (72), which remains an important question for future studies.
We observe that MEF2C is both phosphorylated and highly expressed in AML, and its leukemogenic activities may therefore be due to both its high abundance and S222 phosphorylation. Changes in MEF2C abundance and S222 phosphorylation may affect its activity by (i) recruitment of distinct coactivators or corepressors, (ii) altered dimerization with endogenous MEF2 family members, and/or (iii) altered target gene localization due to secondary changes in the posttranslational modifications and sequence binding preferences of its DNA binding domain (47, 73–75). We found that MEF2C S222 phosphorylation can regulate the assembly of its transcriptionally active complex, which is associated with its induction of both activating and repressive gene expression programs. This leads to the dysregulation of genes controlling cell survival such as LYL1, MYC, BMF, CASP8, and FOXO3 (Fig. 4), consistent with the previously reported activities of MEF2C as both transcriptional activator and repressor (76). This dysregulation is likely important for both AML pathogenesis and chemotherapy resistance, and it will be important to determine whether shared mechanisms regulate leukemia-initiating cell populations and those resistant to chemotherapy. Because our functional studies were largely based on MLL–AF9-rearranged and RUNX1-mutant leukemias, it is possible that MEF2C phosphorylation may involve distinct molecular mechanisms in different leukemia subtypes. It is also possible that the marked phenotype of nonphosphorylatable MEF2C mutants may be due to dominant-negative effects on other MEF2 family members, given their apparent dimerization. Similarly, although MARK3 can specifically phosphorylate MEF2C S222, additional serine kinases inhibited by MRT199665, or those which were not included in our screen, such as mTOR, may induce MEF2C phosphorylation in distinct types of AML.
The MARK family was originally discovered based on its functions in controlling cell polarity and microtubule dynamics as part of the basolateral polarity complex (77, 78), but their functions in hematopoietic cells are not yet defined (79). Our finding that MARKs can regulate MEF2C suggests that the basolateral polarity complex may have distinct functions in normal hematopoiesis, which may be dysregulated in leukemia cells. Similarly, the mechanisms of MARK3-mediated phosphorylation of MEF2C in chemotherapy-resistant AML remain to be defined. For example, MARK kinases can be phosphorylated and activated by GSK-3β (80), which can in turn by activated by integrin signaling, including reports of its activation in AML (81). Thus, activation of MARK signaling and MEF2C phosphorylation may be linked to autocrine or paracrine signaling by AML cells and the bone marrow niche (29, 82, 83).
Finally, the lack of apparent phenotypes in knock-in mutant mice homozygous for Mef2cS222A/S222A or Mef2cS222D/S222D suggests that phosphorylation of MEF2C S222 is dispensable for normal development, establishing a compelling therapeutic window and substantiating its therapeutic targeting. However, additional studies of MEF2C phosphorylation in specific physiologic states (24, 25), and further preclinical development of selective non-pyrrolopyrimidinone MARK inhibitors, will be needed to advance their use as MEF2C-targeted therapeutics. Future functional profiling studies of large AML cohorts should establish the prevalence of MEF2C phosphorylation and other mechanisms of primary chemotherapy resistance.
Collection of Patient Samples
Written informed consent and approval by the Institutional Review Boards of participating institutions was obtained in accordance with the Declaration of Helsinki for all subjects. Primary chemotherapy resistance was defined based on the presence of at least 5% of abnormal blasts by morphologic and immunophenotypic assessment of bone marrow aspirates obtained after two cycles of induction chemotherapy, as assessed by the respective institutional or central pathologic reviews. Specimens were collected and leukemia cells purified as previously described (16). Mononuclear cells were purified using Ficoll gradient centrifugation, and leukemia cells were purified by negative immunomagnetic selection against CD3, CD14, CD19, and CD235a, based on the absence of their expression by the majority of AML specimens.
Production and Purification of pS222 MEF2C Antibody
A phospho-specific antibody against MEF2C S222 (RefSeq ID: NM_002397.4) was generated by PhosphoSolutions (Catalog p1208-222, RRID:AB_2572427; PhosphoSolutions). Rabbits were immunized with a synthetic phosphopeptide corresponding to amino acids surrounding S222 of human MEF2C, conjugated with Keyhole limpet hemocyanin. Serum isolated from peripheral blood of immunized rabbits was screened for phospho-specificity using enzyme-linked immunosorbent assays (ELISA). ELISA-positive sera were pooled and sequentially affinity purified using both the phosphopeptide and non-phosphopeptide columns to isolate the affinity-purified pS222 MEF2C antibody.
Phosphoproteomic profiling was performed as described previously (31). Briefly, purified leukemia cells were lysed using guanidine hydrochloride, and proteins were reduced with dithiothreitol, alkylated with iodoacetamide, and digested using trypsin. Tryptic peptides were purified by solid phase extraction, and purified peptides were labeled using iTRAQ reagents. Phosphopeptides were enriched using iron affinity chromatography, and separated using three-dimensional RP-SAX-RP chromatography coupled to nanoelectrospray ion source. Spectra were recorded using Orbitrap Velos mass spectrometer (ThermoFisher Scientific) in data-dependent mode. Data files were analyzed using multiplierz (84). Analysis of phosphoproteomics data is detailed in Supplementary Methods.
Human AML cell lines were obtained from DSMZ and HEK293T cells from ATCC in 2013. Cell line authentication testing by STR genotyping (Genetica DNA Laboratories) for all cell lines was performed in May 2014 and February 2017. The absence of Mycoplasma sp. contamination was verified most recently in August 2017 using Lonza MycoAlert (Lonza Walkersville, Inc.). Cell lines were used for experiments in this article within 20 passages from thawing.
All mouse studies were conducted with approval from the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee. C57BL/6J, NOD-SCID IL2Rγ-null (NSG), and B6.SJL-Ptprca Pepcb/BoyJ mice were all obtained from The Jackson Laboratories. Knock-in Mef2cS222A and Mef2cS222D mutant mice were generated using CRISPR/Cas9 genome editing and homologous recombination. A CRISPR gRNA targeting exon 5 of the Mef2c gene was designed (GGAATGGATACGGCAACCCC) and expressed in vitro using the approach previously described (85). The donor templates for recombination were designed according to Wefers and colleagues (86), which produced in-frame substitutions and the addition of restriction sites for genotyping. Mouse zygotes from C57BL/6J:CBA F2 hybrids were injected with 100 ng/μL of gRNA, 50 ng/μL of Cas9 mRNA, and 50 ng/μL of donor oligonucleotide into the pronucleus. Founder mice were first screened using restriction endonuclease digestion of genomic DNA isolated from tail tissues and positive mice analyzed by PCR. For additional details, see the Supplementary Methods. For all experiments using Mef2c knock-in mutant animals, wild-type littermates were used as controls.
Murine Bone Marrow Transplantation Models
For primary MLL–AF9 and Hoxa9/Meis1 mouse leukemia transplants, 150,000 and 500,000 transduced cells, respectively, were transplanted by intravenous injection into lethally irradiated (900 rad) C57BL/6J animals with 125,000 support wild-type bone marrow cells and monitored for the development of leukemia. For Runx1−/−;Flt3ITD mouse leukemia transplants, 750,000 and 50,000 secondary leukemia bone marrow cells harvested from moribund primary recipients were transplanted by intravenous injection into sublethally irradiated (600 rad) C57BL/6J mice. For competitive bone marrow transplantation studies, 1 million CD45.2 mononuclear bone marrow cells from Mef2C transgenic mice were mixed with 1 million CD45.1 mononuclear bone marrow cells from B6.SJL-Ptprca Pepcb/BoyJ mice and transplanted via tail-vein injection in lethally irradiated C57BL/6J recipient animals. Transplanted mice were monitored by CD45.1/CD45.2 peripheral blood chimerism using FACS with the surface markers CD45.2 and CD45.1. Antibodies used in flow cytometry are detailed in Supplementary Table S5.
Human AML Xenograft Model
NSG mice (8–10 weeks old) were sublethally irradiated (200 rad) and transplanted with 500,000 OCI-AML2 cells via tail-vein injection. Doxycycline-inducible transgene expression was induced using doxycycline chow 3 days after transplantation (Harlan). In xenotransplants involving cytarabine treatment, 1.2 million OCI-AML2 cells were transplanted into nonirradiated NSG mice to account for the toxicity effects of cytarabine. Cytarabine in PBS (50 mg/kg/day) was administered 7 days after transplantation intraperitoneally for a total of 5 doses.
Recombinant Kinase Screen
The protein kinase screen was performed using a recombinant serine kinase library as previously described (87). Briefly, 172 serine/threonine kinases (Supplementary Data S3) were individually expressed as N-terminal GST-fusion proteins in insect cells, and purified using glutathione sepharose chromatography. A synthetic peptide corresponding to phosphoserine 222 for human MEF2C (RefSeq ID: NM_002397.4) Ac-GNPRN[pS]PGLLVC-NH2 was synthesized (Tufts University), purified by HPLC, and confirmed by mass spectrometry. For the screen, the MEF2C peptide was dissolved in DMSO at 10, 1, and 0.1 μmol/L, and specific kinase activity on respective substrates was determined by the off-chip mobility shift assay using the LabChipTM3000 instrument (Caliper Life Sciences). Methods for analysis are detailed in Supplementary Methods.
All in vitro experiments were performed a minimum of three independent times with a minimum of three biological replicates. Statistical significance was determined using a two-tailed Student t test for continuous variables, Pearson χ2 test for limiting dilution assays, and log-rank test for survival analysis. For ROC curve analysis, MEF2C S222 phosphorylation expression above the median was considered to be positive. Data were plotted as mean values with error bars representing standard deviation.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: F.C. Brown, W. Mark, M. Gonen, G. Pouliot, S.A. Armstrong, A. Kentsis
Development of methodology: F.C. Brown, E. Still, P. Cifani, P. Romanienko, C. McCarthy, C. O'Donnell, A.V. Krivtsov, J.A. Marto, A. Kentsis
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F.C. Brown, E. Still, C.Y. Yim, S. Takao, P. Cifani, C. Reed, S. Gunasekera, S.B. Ficarro, P. Romanienko, W. Mark, C. McCarthy, E. de Stanchina, C. O'Donnell, C. Stutzke, J. Hébert, A. Melnick, E.M. Paietta, M.S. Tallman, G. Pouliot
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.C. Brown, R.P. Koche, C.Y. Yim, S. Takao, P. Cifani, S.B. Ficarro, P. Romanienko, C. McCarthy, M. Gonen, V. Seshan, B. Spitzer, V.-P. Lavallée, A. Letai, G. Sauvageau, R. Levine, S.A. Armstrong, A. Kentsis
Writing, review, and/or revision of the manuscript: F.C. Brown, C.Y. Yim, C. Reed, S.B. Ficarro, M. Gonen, V. Seshan, E.M. Paietta, M.S. Tallman, G. Pouliot, R. Levine, J.A. Marto, S.A. Armstrong, A. Kentsis
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F.C. Brown, E. Still, P. Cifani, S. Gunasekera, P. Bhola, A.V. Krivtsov, A. Melnick, A. Kentsis
Study supervision: F.C. Brown, A. Kentsis
Other (created the model organism for the study): W. Mark
We are grateful to Nathanael Gray, Alejandro Gutierrez, Marc Mansour, Leo Wang, and Michael Kharas for critical discussions. We thank John Schwarz and Hanna Mikkola for Mef2c conditional mice, Jeffrey Magee and Shaina Porter for Runx1−/−;Flt3ITD mice, Philip Cohen for MRT199665, and Ralph Garippa, Antoine Gruet, Matthew Witkin, and Yang Li for technical assistance. This work was supported by NIH R21 CA188881, R01 CA204396, P30 CA008748, U10 CA180827, and U24 CA196172, Burroughs Wellcome Fund, Josie Robertson Investigator Program, Rita Allen Foundation, Alex's Lemonade Stand Foundation, American Society of Hematology, and Gabrielle's Angel Foundation. A. Kentsis is the Damon Runyon-Richard Lumsden Foundation Clinical Investigator.
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.