Combining menin and MEK inhibition to target poor prognosis KMT2A-rearranged RAS pathway–mutant acute myeloid leukemia

Key Points

• •KMT2A-r AML often harbor RAS pathway mutations that, together, drive an aggressive leukemia with a poor prognosis.
• •Dual menin and MEK inhibition shows synergy in vitro and a decrease in leukemia burden in vivo.

Visual Abstract

Abstract

KMT2A-rearranged (KMT2A-r) acute leukemias are especially prevalent in the pediatric population. KMT2A-fusion proteins drive leukemogenic gene expression through an interaction with a chromatin complex that includes the scaffold protein menin, giving rise to aggressive acute leukemias. RAS pathway mutations are also common in pediatric leukemia. In a cohort of 1750 patients enrolled on Children’s Oncology Group (COG) trials, we identified RAS pathway mutations in 43% of acute myeloid leukemia (AML) cases. The presence of RAS pathway mutations in KMT2A-r AML was associated with a lower complete remission rate, poor event-free survival and overall survival (OS), and early relapses. Given the inferior outcome observed for children with dual mutations, we next sought to identify efficacious targeted drug combinations for this subset of childhood leukemia. We evaluated RAS/MAPK targeting using the MEK1/2 inhibitor selumetinib in combination with the menin inhibitor revumenib. Treatment of AML cell lines and cultured leukemia cells from patient-derived xenograft (PDX) models resulted in a synergistic decrease in viability and promoted cell cycle arrest, apoptosis, and downregulation of Myc targets in the combination compared with each drug alone. In vivo, the combination treatment of AML pediatric PDX models harboring KMT2A-r and RAS mutations reduced leukemia burden compared with single-drug treatments but without improving OS compared with menin inhibition alone. Our preclinical study suggests a potential targeted treatment combination for KMT2A-r and RAS pathway–mutant leukemia but one that will require further optimization. These trials were registered at www.clinicaltrials.gov as NCT00070174, NCT00372593, NCT01371981, and NCT00002798.

Introduction

Chromosomal translocations involving the gene encoding the lysinemethyltransferase KMT2A give rise to aggressive acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL). KMT2A rearrangements (KMT2A-r) are present in ∼20% of pediatric AML.^1,2 The KMT2A fusion complex consists of the N-terminal of KMT2A, responsible for protein-protein interactions, and >100 fusion partners.³ Wild-type KMT2A and the aberrant fusion build a chromatin binding complex with the protein menin, which leads to epigenetic changes and a distinct gene expression profile characterized by increased stem cell programs and impaired hematopoietic differentiation.¹

Co-occurring mutations with KMT2A-r involve signaling pathways, including MAPK, with mutations in NRAS, KRAS, or PTPN11.⁴ These mutations increase signaling via the MEK/extracellular signal-regulated kinase (ERK) pathway, promoting cell survival and proliferation. The prognostic significance of these mutations is under investigation. Co-occurrence of KMT2A-r and KRAS G12 mutations in AML has been associated with an adverse outcome.⁵ Furthermore, subclonal RAS pathway mutations present at diagnosis can contribute to relapse of KMT2A-r acute leukemia.6, 7, 8

The presence of RAS pathway mutations increases sensitivity to MEK inhibitors, such as selumetinib or trametinib, in vitro and in vivo, in both AML and ALL.8, 9, 10, 11, 12, 13, 14 MEK inhibition reduces proliferation and induces cell cycle arrest in AML, and promotes apoptosis in AML and ALL cell lines harboring RAS mutations.^12,15 Despite this association, clinical responses to MEK inhibitors in acute leukemia have been modest. In adult AML, unselected for RAS mutations, selumetinib was well tolerated but demonstrated limited activity.¹⁶

The combination of an epigenetic modulator targeting KMT2A with an inhibitor of the RAS/MAPK pathway could potentially improve rate and duration of response in leukemias with alterations in both targets. Menin inhibitors disrupt the KMT2A fusion complex and showed preclinical efficacy in KMT2A-r acute leukemias.^17,18 NPM1c, among the most common mutant proteins in adult AML, also interacts with KMT2A, to enhance a transcriptional program that is similar to that observed with KMT2A fusions.^19,20 The menin inhibitors revumenib (SNDX-5613) and ziftomenib (KO-539) reverse oncogenic transcriptional programs driven by KMT2A fusion oncoproteins or NPM1c, such as HOX cluster genes and cofactors MEIS1 and PBX3, leading to cellular differentiation and loss of stem cell properties in these leukemia subtypes.^21,22 Several menin inhibitors have shown efficacy in early clinical trials, including complete molecular remissions,^17,18,23, 24, 25 with revumenib approved by the US Food and Drug Administration for pediatric and adult patients with relapsed or refractory acute leukemias with KMT2A-r and NPM1c and ziftomenib recently approved for patients with relapsed or refractory NPM1c AML. The rapid development of MEN1 mutations, however, necessitates effective, rational combination treatments for full efficacy of this targeted therapy.²⁶

Given the prevalence of RAS mutations in pediatric patients with KMT2A-r AML, we set out to study the prognostic significance of these mutations. With the recent availability of menin inhibitors, we studied the combination of menin with MEK inhibition for KMT2A-r and RAS pathway–mutant AML.

Methods

Patient data analysis

We analyzed data from 1750 pediatric and young adult patients with de novo AML enrolled on Children’s Oncology Group (COG) trials AAML03P1, CCG2961, AAML0531, and AAML1031.^2,27, 28, 29 NRAS, KRAS, and PTPN11 mutations with a variant allele frequency of ≥1% were included in the RAS pathway–mutant cohort. All studies were conducted in accordance with the Declaration of Helsinki.

Drug sensitivity testing

VTP-50469 was purchased from MedChemExpress; and selumetinib (AZD6244), trametinib, and ziftomenib from Selleck. Cells were treated with increasing concentrations of these drugs or the indicated combination, and viability was measured after 6 days using a CellTiter-Glo luminescent assay (Promega) following the manufacturer’s protocol. Luminescence was measured using FLUOstar OPTIMA (BMG Labtech). The 50% inhibitory concentration values were determined using Prism GraphPad version 9.1.1 software.

Synergy calculations

The expected dose-inhibitory fraction relationships for the indicated combinations of VTP-50469, ziftomenib, trametinib, and selumetinib were assessed based on the Chou-Talalay Combination Index for Loewe additivity, which uses a dose-effect strategy.30, 31, 32, 33

Sequencing

RNA sequencing was performed on OCI-AML3 and NOMO1 cells treated with dimethyl sulfoxide, or VTP-50469, selumetinib, or the combination. Three samples were collected per treatment condition per time point. RNA was extracted using a RNeasy kit (Qiagen) and library preparation and sequencing performed by Novogene. Patient-derived xenograft (PDX) samples were verified for mutations using the rapid heme panel assay³⁴ and RNA sequencing for fusion assessment.

In vivo experiments

All animal experiments were performed with approval from the Dana-Farber Cancer Institute (DFCI) institutional animal care and use committee and were conducted by the DFCI Experimental Therapeutics core. PDXs were established under protocols approved by the DFCI (CPCT-0007¹⁴) and Cincinnati Children’s Hospital Medical Center (PDX17-14) institutional review boards. Primary patient samples were acquired after informed consent. Detailed information on the PDX models is in supplemental Table 1. Chow containing 0.033% (abbreviated to 0.03% in the text and figures) or 0.1% revumenib was provided by Syndax Pharmaceuticals. Selumetinib was purchased from LC Laboratories. Details on design of the in vivo studies are in the supplemental Methods.

Additional methods are within the supplemental Methods.

Results

RAS mutations are associated with a poor prognosis in pediatric KMT2A-r AML

The clinical impact of RAS mutations in KMT2A-r infant ALL is established and is associated with poor outcome.^35,36 The prognostic impact of RAS pathway mutations in pediatric AML is less clear. We therefore analyzed data from 1750 pediatric patients with AML treated on the COG AAML03P1, CCG2961, AAML0531, and AAML1031 studies (Table 1, supplemental Table 2). In this cohort, 767 (43.8%) of 1750 patients had a mutation in the RAS pathway (NRAS, KRAS, or PTPN11). Patients within the genetic subgroups of KMT2A-r, inv(16), or monosomy 7 showed an enrichment of RAS pathway mutations, whereas patients with a normal karyotype were less likely to harbor RAS pathway mutations. Patients with Fms-related receptor tyrosine kinase 3 internal tandem duplication (FLT3-ITD), including FLT3-ITD with high allelic ratio (allelic ratio >0.4), were less likely to have a co-occurring RAS pathway mutation. The presence of RAS pathway mutations did not have any prognostic significance in the full cohort of patients (Figure 1A-B).

Table 1.

AML cohort (N = 1750)

Characteristics	RAS pathway wild type		RAS pathway mutation		P
	n	%	n	%
Total	983	56.2	767	43.8
Sex
Male	497	50.6	410	53.5	.229
Female	486	49.4	357	46.5
Study
AAML03P1	47	4.8	37	4.8	.967
AAML0531	389	39.6	323	42.1	.283
AAML1031	519	52.8	374	48.8	.094
CCG2961	28	2.8	33	4.3	.100
Study treatment
AAML03P1	47	4.8	37	4.8	.967
AAML0531, arm A	180	18.3	170	22.2	.046
AAML0531, arm B	209	21.3	153	19.9	.501
AAML1031, arm A	222	22.6	184	24.0	.489
AAML1031, arm B	249	25.3	185	24.1	.561
AAML1031, arm C	48	4.9	5	0.7	<.001
CCG2961	28	2.8	33	4.3	.100
Age group, y
0-1	152	15.5	122	15.9	.800
2-10	344	35.0	276	36.0	.668
>11	487	49.5	369	48.1	.552
Age, y
Median (range)	10.9	(0-29.84)	10.4	(0.15-29.55)	.289
Race
American Indian or Alaska Native	7	0.4	6	0.4	.850
Asian	42	2.3	34	2.4	.831
Native Hawaiian or other Pacific Islander	4	0.2	4	0.3	.734
Black or African American	118	6.6	91	6.5	.996
White	717	39.9	551	39.4	.869
Multiple races	1	0.1	0	0.0	1.000
Unknown	94		81
Ethnicity
Hispanic or Latino	172	18.3	138	18.5	.914
Not Hispanic or Latino	767	81.7	607	81.5
Unknown	44		22
WBC, ×103 μL
Median (range)	23.7	(0.2-827.2)	37.8	(0.5-550)	<.001
BM blasts, %
Median (range)	73	(0-100)	64	(0-100)	<.001
Cytogenetics
Normal	285	21.9	156	15.9	<.001
t(8;21)	143	11.0	85	8.7	.029
inv(16)	69	5.3	131	13.4	<.001
t(9;11)/11q23	167	12.8	183	18.7	<.001
t(6;9)	16	1.2	19	1.9	.215
Monosomy 7	15	1.2	28	2.9	.005
Del(7q)	18	1.4	7	0.7	.105
Monosomy 5/del(5q)	11	0.8	11	1.1	.567
+8	53	4.1	34	3.5	.345
Other abnormalities	177	13.6	95	9.7	.001
Unknown	29		18
CEBPA status
Negative	891	91.7	727	95.9	<.001
Positive	81	8.3	31	4.1
Unknown	11		9
NPM1 status
Negative	863	89.0	691	91.2	.133
Positive	107	11.0	67	8.8
Unknown
Cyto fusion risk group
Standard	243	25.2	192	25.5	.708
Low	394	40.8	314	41.6	.489
High	303	31.4	214	28.4	.273
Risk group (cyto/molecular) ITD high AR excluded
Standard	437	45.2	389	51.6	.009
Low	397	41.1	313	41.5	.862
High	132	13.7	52	6.9	<.001
Response by end of course 1
CR	728	76.2	577	77.0	.697
Not CR	227	23.8	172	23.0
Not evaluable	28		18
ITD
No	742	76.3	698	91.4	<.001
Yes	230	23.7	66	8.6
MRD1
No	589	69.1	469	72.9	.102
Yes	264	30.9	174	27.1

Bold values indicate statistical significance of P < .05.

AR, allelic ratio; BM, bone marrow; Cyto, cytogenetic; CEBPA, CCAAT/enhander-binding protein alpha; ITD, internal tandem duplication; MRD1, minimal residual disease after induction 1; NPM1, nucleophosmin; WBC, white blood cell count.

Figure 1.

Pediatric KMT2A-r AML frequently harbors RAS mutations, which are associated with a prognostic disadvantage compared with pediatric KMT2A-r AML with wild-type RAS. (A-F) Kaplan-Meier survival curves for pediatric patients with AML from the COG studies AAML03P1, CCG2961, AAML0531, and AAML1031. (A) EFS after study entry comparing all patients with RASwt (n = 983) in blue with RAS+ (n = 767) in red, P = .557. (B) OS after study entry comparing all patients with RASwt (n = 983) in blue with those with RAS+ (n = 767) in red, P = .737. (C) EFS after study entry comparing patients with KMT2A-r and RASwt (n = 182) in blue with those with RAS+ (n = 191) in red, P = .011. (D) OS after study entry comparing patients with KMT2A-r and RASwt (n = 182) in blue with those with RAS+ (n = 191) in red, P = .047. (E) Curves showing relapse risk (RR) at 2 years after end of induction 1 chemotherapy cycle comparing all patients with RASwt (n = 728) in blue with those with RAS+ (n = 577) in red, P = .702. (F) RR curves at 2 years after end of induction 1 chemotherapy comparing patients with KMT2A-r and RASwt (n = 142) in blue with those with RAS+ (n = 132) in red, P = .004 at 0.5 years and P = .030 at 1 year from the end of induction 1. P value estimated by the log-rank test. RAS+, RAS pathway mutant; RASwt, RAS wild type.

Overall, 373 (21%) patients analyzed for a RAS pathway mutation also had a KMT2A-r. Of 373 patients with KMT2A-r, 191 (51.2%) had mutations in NRAS, KRAS, or PTPN11 (supplemental Table 2). Children with both a KMT2A-r and RAS mutation had inferior disease-related outcomes to that of children with KMT2A-r without RAS pathway abnormalities. Specifically, the former were less likely to achieve complete remission (CR) after induction 1 (71.7% vs 80.7%; P = .047) and their 5-year EFS was inferior at 28.9% compared with 36% (P = .011) in children with a KMT2A-r without a RAS mutation (Figure 1C). Five-year disease-free survival was 33.6% vs 40.2% (P = .042) respectively (supplemental Table 3). The 5-year OS was 48.9% vs 56.1% (P = .047) in the RAS mutant compared with the wild-type RAS group (Figure 1D; supplemental Table 3). Patients with KMT2A-r and RAS pathway mutations had a significantly increased rate of early relapse within 1 year of end of induction compared with the patients with KMT2A-r and wild-type RAS AML (Figure 1E-F; supplemental Table 3). Presence of KMT2A-r was associated with worse EFS (hazard ratio, 1.54; P < .001) and OS (hazard ratio, 1.56; P < .001) in univariate and multivariable analysis of the full cohort (supplemental Table 3). Within the KMT2A-r cohort, RAS mutations were associated with poor EFS and OS in a univariate model, and poor EFS in the multivariable analysis (supplemental Table 3).

Menin inhibition and MEK inhibition are synergistic in RAS pathway–mutated AML in vitro

We investigated the efficacy of combining menin inhibitors and MEK inhibitors, targeting the activated RAS signaling pathway, in KMT2A-r (NOMO1 and PDX 17-14) and NPM1c-mutant AML (OCI-AML3). We first determined the 50% inhibitory concentrations for VTP-50469, a preclinical precursor of revumenib, and selumetinib in these AML cell lines and PDX models with different genetic characteristics cultured in vitro (supplemental Figure 1A). The presence of RAS mutations was associated with increased sensitivity to selumetinib consistent with previous publications (supplemental Figure 1A).^13,37,38

Next, we expanded our investigation to test multiple combinations of menin and MEK inhibitors in these AML cells cultured in vitro. In addition to VTP-50469 and selumetinib, we included the menin inhibitor ziftomenib and the MEK inhibitor trametinib. Cells were treated with a range of concentrations of each drug combination simultaneously and viability determined. Using the Chou-Talalay Combination Index and isobolograms, additivity or synergy were observed across all 4 combinations tested: VTP-50469 with selumetinib, VTP-50469 with trametinib, ziftomenib with selumetinib, and ziftomenib with trametinib (Figure 2A-C; supplemental Figure 1B). These findings demonstrate that the additive/synergistic effect is a class effect of combined menin and MEK inhibition rather than being specific to individual drug pairs.

Figure 2.

Impact of the combination of MEK and menin inhibition on cell viability, cell death, cell cycle, and colony forming in AML cell lines and PDX cells in vitro. (A-C) Synergy assays between menin and MEK inhibitors. CI analysis over a range of concentrations for the combinations of VTP-50469 and SEL, Zifto, and SEL; VTP-50469 and Tram; and Zifto and Tram treated for 6 days in PDX17-14 (A), NOMO1 (B), and OCI-AML3 (C) cells. (D) Cell cycle analysis in indicated cell lines and PDX cells treated with VTP-50469 (100nM), SEL (100nM), or the combination for 72 hours. (E) Effect of VTP-50469 (20nM), SEL (100nM), or the combination on the number of total colonies in NOMO1 and OCI-AML3 cell lines after 7 and 8 days, respectively, and VTP-50469 (100nM), SEL (100nM), or the combination in PDX17-14 after 10 days, compared with DMSO control. Bar plots show mean ± standard error of the mean, n = 3. ∗∗∗∗P < .0001, ∗∗∗P < .001, ∗∗P < .01, ∗P < .05, using 1-way analysis of variance (ANOVA) with the Tukey multiple comparison test. CI, combination index; Combo, combination; DMSO, dimethyl sulfoxide; ns, not significant; SEL, selumetinib; Tram, trametinib; VTP, VTP-50469; Zifto, ziftomenib.

The combination of VTP-50469 and selumetinib resulted in an increase in G1 arrest in PDX17-14 and the AML cell lines NOMO1 and OCI-AML3 (Figure 2D). We next tested the impact of the combination on colony formation in methylcellulose. The combination led to a decrease in colony number and size in PDX17-14, compared with either drug alone (Figure 2E; supplemental Figure 2A). In the cell lines NOMO1 and OCI-AML3, the combination of VTP-50469 with selumetinib led to smaller colonies but total number of colonies with the combination was not significantly different compared with VTP-50469 treatment alone (Figure 2E; supplemental Figure 2A).

Combination treatment with VTP-50469 and selumetinib leads to apoptosis and is associated with alterations in BCL-2 family member protein levels

The combination of menin inhibitors VTP-50469 or ziftomenib and MEK inhibitors selumetinib or trametinib resulted in an increase in apoptosis as measured by increased annexin V staining and induction of cleaved poly-(adenosine diphosphate [ADP]-ribose) polymerase compared with single-drug treatments (Figure 3A-B; supplemental Figure 3A). To further elucidate the mechanisms underlying combination-induced cell death, we examined the expression of key B-cell lymphoma 2 (BCL-2) family members in PDX17-14, NOMO1, and OCI-AML3 cells after treatment with VTP-50469 and selumetinib. We observed variability in the changes in protein levels of BCL-2 family members, including downregulation of antiapoptotic proteins BCL-2 and/or myeloid cell leukemia-1 (MCL-1) (most prominent in NOMO-1 and OCI-AML3) and upregulation of proapoptotic BIM (most prominent in PDX17-14; Figure 3C). To functionally assess mitochondrial priming for apoptosis, we performed dynamic BH3 profiling on OCI-AML3 cells treated with VTP-50469, selumetinib, or the combination. This assay revealed priming toward apoptosis with single-agent treatment, which increased with the combination treatment (Figure 3D; supplemental Figure 3B-C). Collectively, these data support apoptosis as a mechanism of cell death induced by the combination therapy, shifting the balance of BCL-2 family protein levels toward apoptosis across all 3 AML models, albeit through distinct patterns.

Figure 3.

Menin and MEK inhibition induces apoptosis and is associated with changes in BCL-2 family member protein levels. (A) Percentage of early apoptotic (annexin V⁺/PI⁻), late apoptotic (annexin V⁺/PI⁺), and necrotic (annexin V⁻/PI⁺) cells treated with VTP-50469 (100nM), selumetinib (100nM), or the combination for 96 hours in the indicated cell lines and PDX model. Representative graphs of 2 independent experiments are shown. Bar plots show mean ± standard deviation (SD) for n = 3. ∗∗∗∗P < .0001, ∗∗∗P < .001, ∗∗P < .01, ∗P < .05, using 1-way ANOVA with the Tukey multiple comparison test. (B) Immunoblot of indicated proteins after single-agent or combination treatment with VTP-50469 (100nM) and selumetinib (100nM) for 96 hours. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is used as a control. (C) Immunoblot of indicated proteins after single-agent or combination treatment with VTP-50469 (100nM, 3 days) and selumetinib (100nM, 24 hours). Vinculin is used as a control. (D) Dynamic BH3 profiling in OCI-AML3 cells treated with DMSO, 1μM VTP-50469, 0.25μM selumetinib, or the combination for 24 hours followed by dynamic BH3 profiling. Overall priming was quantified in response to 50nM BIM BH3 peptide. Change in cytochrome C release (Δ) was calculated as (percentage release in treated cells) minus (percentage release in control). Shown are mean ± SD; n = 9; 1-way ANOVA with the Tukey multiple-comparisons test, ∗∗P < .01, ∗∗∗P < .001. BAX, BCL-2-associated X protein; CI, combination index; Combo, combination; DMSO, dimethyl sulfoxide; MCL1, myeloid cell leukemia-1; ns, not significant; PARP, poly-(ADP-ribose) polymerase; PI, propidium iodide; SEL, selumetinib; VTP, VTP-50469.

Impact of combined menin and MEK inhibition on MAPK signaling and gene expression

Activating RAS pathway mutations lead to an increase in MEK/ERK signaling, which results in increased cell proliferation.³⁹ RAS pathway activation has also been reported in KMT2A::MLLT3 and KMT2A::AFF1 fusion leukemia without RAS mutations, suggesting a general increased activity of the RAS signaling pathway in KMT2A-r leukemia.⁷ We thus tested the effect of VTP-50469 alone and in combination with selumetinib on phosphorylated ERK (P-ERK) in AML cell lines. VTP-50469 led to a dose-dependent decrease in P-ERK (supplemental Figure 4A) although P-ERK repression was stronger with selumetinib or treatment with both VTP-50469 and selumetinib (supplemental Figure 4B).

KMT2A-r and NPM1c⁺ leukemias share a common gene expression signature, characterized by an overexpression of HOX cluster genes and MEIS1.⁴⁰ These genes are highly expressed in early lineage progenitor cells and anticorrelated with myeloid differentiation.^41,42 Menin inhibition has been reported to decrease MEIS1 expression in vitro and in AML cells in patients with KMT2A-r or NPM1c⁺ leukemia treated in clinical trials with revumenib.^17,25,43 In our study, VTP-50469 alone or in combination with selumetinib decreased MEIS1 expression (supplemental Figure 4C). Selumetinib alone led to an increase in MEIS1 RNA levels although this did not result in increased MEIS1 protein levels (supplemental Figure 4D). Treatment with VTP-50469 or the 2 drugs together resulted in a decrease in MEIS1 protein levels (supplemental Figure 4D).

To further define the mechanisms of synergy of VTP-50469 and selumetinib, we analyzed gene expression changes in 2 AML cell lines, NOMO1 (KMT2A-r) and OCI-AML3 (NPM1c⁺), treated with VTP-50469, selumetinib, or the combination, for 3 and 6 days. In NOMO1, there were 6021 and 5569 differentially expressed genes at days 3 and 6, respectively, in the combination treatment compared with the dimethyl sulfoxide (DMSO) control. For OCI-AML3, gene expression changes were less dynamic, with 4260 and 3762 differentially expressed genes at days 3 and 6, respectively (supplemental Figure 5A; supplemental Table 4). Pathway analysis showed a decrease in MYC signatures as the most significant alteration (Figure 4A; supplemental Figure 5B), as well as signatures associated with cell cycle control and mechanistic target of rapamycin (mTOR) signaling. We identified leading edge genes within the Hallmark MYC targets V1 and V2 signatures that were decreased more with the combination treatment than either single-drug treatment (Figure 4B; supplemental Figure 6A-B). c-Myc expression itself was decreased with each drug alone, and a further decrease was observed with combination treatment (Figure 4C; supplemental Figure 6C). We confirmed this RNA-sequencing finding by real-time polymerase chain reaction and western blotting in which we observed a decrease in c-Myc mRNA and protein levels in 2 cell lines and a PDX model (Figure 4D).

Figure 4.

Changes in Myc pathway upon menin and MEK inhibition in AML cell lines and PDX cells. (A) Heat map showing the single-sample gene set enrichment analysis scores across MSigDB Hallmarks for the transcriptional changes induced by NOMO1 treatment with VTP-50469 (100nM, 3 days), selumetinib (100nM, 24 hours), or the combination (VTP-50469 100nM for 3 days, and selumetinib 100nM 24 hours) vs control. (B) Heat map showing the log2-fold change in expression of individual genes in the HALLMARK_MYC Targets V1 and V2 after single or combination treatment of NOMO1 and OCI-AML3 cell lines as described in panel A. Genes are ranked based on expression changes induced by the combination treatment compared to DMSO. Top 30 leading edge genes are shown. (C) Bar plots showing mean ± SD for c-Myc mRNA log2(TPM+1) expression in NOMO1 and OCI-AML3 cell lines treated with VTP-50469 (100nM, 6 days) and selumetinib (100nM, 24 hours), or the combination (VTP-50469 100nM for 6 days, and selumetinib 100nM 24 hours). Transcriptional changes between replicates in treatment groups compared to DMSO are estimated by eBayes limma test (∗∗∗∗P < .0001, ∗∗P < .01, ∗P < .5). (D) c-Myc mRNA expression after treatment with VTP-50469 (100nM, 6 days), selumetinib (100nM, 24 hours), or the combination (VTP-50469 100nM for 6 days, and selumetinib 100nM for 24 hours). Quantitative reverse transcription polymerase chain reaction reads normalized to GAPDH and reported as relative to DMSO. Bar graphs represent the mean ± SD of 2 independent experiments, each performed in 3 technical replicates. ∗∗∗∗P < .0001, ∗∗∗P < .001, ∗∗P < .01, ∗P < .05, using 1-way ANOVA with the Tukey multiple comparison test. Immunoblots show c-Myc protein levels from 1 of the 2 experiments. Combo, combination; DMSO, dimethyl sulfoxide; ns, not significant; SEL, selumetinib; VTP, VTP-50469.

Menin inhibitors are well reported to induce AML differentiation in preclinical studies.^17,18,44 Indeed, treatment with menin inhibitors has been associated with differentiation syndrome in patients, even resulting in a temporary hold on the phase 1b clinical trial KOMET-001. In the phase 1 trial investigating revumenib (ClinicalTrials.gov identifier: NCT04065399), 16.2% of patients experienced differentiation syndrome.²⁵ Although the combination with selumetinib induced less CD11b, a marker of myeloid differentiation, than the menin inhibitor alone, the combination led to an increase in the Blood Atlas monocyte differentiation signature based on RNA sequencing as well as an increase in inflammatory signatures (supplemental Figure 7A-C). Similarly, by evaluation of cell morphology using May-Grünwald-Giemsa staining, differentiated cells were observed in the combination (supplemental Figure 7D).

The combination of revumenib with selumetinib in AML decreases disease burden in vivo

Next, we tested the combination of revumenib, a clinical grade menin inhibitor structurally related to VTP-50469, and selumetinib in AML PDX models harboring different KMT2A-r and RAS mutations in vivo (supplemental Table 1). A tolerability study was conducted, and the selumetinib dose of 50 mg/kg per day was determined tolerable in combination with revumenib (supplemental Figure 8A). The first study was conducted in PDX17-14, a model with KMT2A::AF10 and a KRAS G13D mutation. As observed in our in vitro studies, protein levels of c-Myc, P-ERK, and MEIS1 in PDX17-14 bone marrow cells were decreased after the combination treatment, as well as P-ERK expression in the peripheral blood by flow cytometry (Figure 5A-B), consistent with target inhibition by the drugs. After 18 days of treatment, a statistically significant reduction in disease burden was observed with the combination in the bone marrow compared with either single-drug treatment alone and in the peripheral blood and by spleen weight compared with revumenib. There was a trend toward greater decline in leukemia burden with combination therapy compared with selumetinib in the peripheral blood and spleen (Figure 5C-D). Overall, target inhibition and decreased leukemia burden were noted, but disease eradication after 18 days of treatment was not observed. We therefore repeated the in vivo study and extended treatment duration to 56 days, to mirror 2 28-day cycles of therapy as has been delivered in clinical trials, to discern how prolonged drug exposure affected leukemia burden and survival.

Figure 5.

Effects of menin and MEK inhibition on leukemia burden and survival in in vivo studies with pediatric PDXs. (A-C) Mice injected with PDX17-14 were dosed with REV 0.03% chow; SEL 50 mg/kg per day, 5 days a week by oral gavage; or Combo for a total of 18 days. (A) Western blot showing protein levels of c-Myc, ERK, P-ERK, MEIS1, and vinculin as a loading control in bone marrow (BM) cells sorted for human CD45 (hCD45). (B) Graph showing intracellular P-ERK MFI by flow cytometry in hCD45⁺ cells in mouse peripheral blood at day 18. (C) Graph showing percent of hCD45 cells (%) by flow cytometry in the BM, spleen, and peripheral blood (PB). (D) Mouse spleen weight after treatment with indicated drugs for 18 days. (E-H) Mice were treated for a total of 56 days with REV 0.03% chow; SEL 50 mg/kg per day, 5 days a week by oral gavage; or Combo. (E) Graph showing percent of hCD45 cells by flow cytometry in the BM, spleen, and PB after treatment with indicated drugs for 43 days. For control mice that died before day 43, hCD45 is shown at the time of death. (F) Kaplan-Meier survival curve of NSG mice engrafted with PDX17-14 treated with the indicated drugs. (G) Percent of hCD45⁺ cells by flow cytometry in the BM, spleen, and PB at time of mouse death. (H) Graph showing percent of hCD45 in the PB from start of treatment to the time of death. For panels B-G, ∗∗∗∗P < .0001, ∗∗∗P < .001, ∗∗P < .01, ∗P < .05, using 1-way ANOVA with the Tukey multiple comparison test and log-rank test for panel F. Combo, combination of REV and SEL; MFI, mean fluorescence intensity; ns, not significant; REV, revumenib; SEL, selumetinib.

In this second study, we observed a decrease in P-ERK with the selumetinib or combination compared with the vehicle (supplemental Figure 8B). The vehicle treated mice did not survive the planned treatment course, necessitating evaluation of disease burden at day 43, before treatment completion. Leukemic burden decreased in all drug-treated cohorts in the bone marrow, spleen, and peripheral blood and there was resolution of splenomegaly after 43 days of treatment (Figure 5E; supplemental Figure 8C). After a total of 56 days of treatment, each drug individually was associated with increased survival compared with vehicle, and there was a trend toward improved survival in the combination arm compared with either drug alone (Figure 5F). At the time of death, the mice in the combination arm had lower leukemia burden measured by flow cytometry in bone marrow, spleen, and peripheral blood. There was also less leukemia infiltration in the liver, as well as decreased spleen weights compared with revumenib treated mice (Figure 5G-H; supplemental Figure 8D-E). There was no difference in body weight and no indication of liver toxicity in the different treatment groups (supplemental Figure 8E-F). Assessment of the most frequent MEN1 hot spot mutations, S160, M327, G331, and T349, after exposure to revumenib in mice with available DNA revealed T349M in 1 of 5 mice in the revumenib arm and 1 of 4 mice in the combination treatment arm (supplemental Table 5).

Next, we tested the combination of revumenib and selumetinib in the CPCT-0007 model with KMT2A::PICALM and NRAS Q61L mutation (supplemental Table 1). Mice were treated for 24 days, and leukemia burden was assessed at the end of treatment. Revumenib and selumetinib alone were efficacious at decreasing disease burden at the measured time point (supplemental Figure 8G-H). In this model, mice treated with either the menin inhibitor alone or the combination were cured (supplemental Figure 8I).

Discussion

KMT2A-r results in aggressive infant, childhood, and adult acute leukemias, characterized by activation of a HOX/MEIS1 transcription program, a primitive differentiation state, and lineage promiscuity. In pediatric patients with AML, RAS pathway mutations are present in >40% of these leukemias, with NRAS mutations the most common followed by KRAS mutations.^4,45 Here, using data from COG trials, AAML03P1, CCG2961, AAML0531, and AAML1031, we show RAS pathway mutations to be present in >50% of KMT2A-r AML. In our analysis, RAS pathway mutations were associated with decreased CR rates after first induction chemotherapy, poor EFS, early relapse, and poor OS within the KMT2A-r AML subset. Larger cohorts of patients are needed to further distinguish the prognostic significance of RAS mutations based on KMT2A fusion partners and other cytogenetic abnormalities and require sharing and integration of data across the international community to achieve statistical power.⁴⁶ The COG trials that we analyzed span different treatment eras, risk-stratification, and therapeutic approaches compared with more recent international studies. The prognostic significance of RAS mutations in the more recent trials remains to be studied.^2,46, 47, 48

Menin inhibitors are a promising therapy for KMT2A-r leukemia. There are several inhibitors in clinical trials, and revumenib was recently approved by the US Food and Drug Administration for patients with relapsed or refractory acute leukemia with KMT2A-r or AML with NPM1c mutation, and ziftomenib was approved for adults with relapsed or refractory AML with NPM1c mutation. Preclinical data demonstrated striking efficacy of these drugs in KMT2A-r leukemia including cures in PDX models.^17,22,49,50 Treatment with revumenib in the phase 1 AUGMENT-101 clinical trial of heavily pretreated patients was associated with a CR or CR with partial hematologic recovery rate of 30% and an overall response rate of 53%, but resistance and subsequent relapse remain challenges.²⁵ Similarly, treatment with the menin inhibitor ziftomenib on the KOMET-001 trial at the recommended phase 2 dose was associated with a 25% CR/CR with partial hematologic recovery in patients with KMT2A-r or NPM1 mutations.⁵¹ Of patients treated on AUGMENT-101, with available DNA samples, 38.7% developed mutations in MEN1, the gene encoding menin, as a mechanism of resistance, likely leading to subsequent disease progression.²⁶ Next-generation menin inhibitors will be necessary to overcome this mechanism of resistance. Another strategy to prevent the emergence of resistant clones is combination therapy. Revumenib and ziftomenib were associated with AML differentiation; combination therapies that increase cell death need to be identified. We found that the combination of menin inhibitors with MEK inhibitors to target RAS pathway signaling was synergistic in AML cell lines in vitro, with the combination leading to decreased cell growth, increased G1 cell cycle arrest, and apoptosis.

The combination of revumenib and selumetinib converged on c-Myc as a potential mechanism for the synergy observed in vitro. C-MYC is a proto-oncogene frequently overexpressed in cancer, driving proliferation and tumor progression. In AML, the ERK/c-Myc axis is reported to play a critical role in drug resistance of leukemic stem cells through the activation of survivin.⁵² c-Myc facilitates the formation of transcriptional complexes, including with menin^53,54, and is also a direct transcriptional target gene of KMT2A fusion proteins.^55,56 c-Myc protein levels are particularly high in a broad range of cytogenetically distinct AML, including with KMT2A-r, and are associated with poor prognosis.^7,57, 58, 59, 60 We found that revumenib, or the drug combination, decreased c-Myc both in AML cell lines in vitro and in AML cells in the bone marrow, with the combination having a greater effect.

Beyond c-Myc downregulation, our mechanistic investigations revealed that apoptosis is a key pathway induced with the selumetinib and revumenib combination therapy. We observed decreased expression of the antiapoptotic BCL-2 and MCL-1 proteins along with functional BH3 profiling, demonstrating enhanced mitochondrial priming for apoptosis with the combination treatment compared with either single agent alone. These findings align with previous reports showing that MEK inhibition can prime cells for apoptosis through modulation of BCL-2 family proteins,^61,62 and suggest that the combination of menin and MEK inhibition enhances apoptotic susceptibility.

The combination of revumenib with selumetinib was more efficacious in decreasing leukemia burden in vivo compared with single drug treatments in AML with 1 or 2 months of treatment. Combination treatment, however, did not improve mouse survival in the combination arm compared with revumenib alone. Both revumenib and selumetinib are dosed continuously in humans (ClinicalTrials.gov identifier: NCT03190915, NCT04487106, NCT03326310, and NCT04065399). The median time to achieving a cytogenetic response in patients treated on AUGMENT-101 was 1.9 months, consistent with the differentiation mechanism of action for this drug.²⁵ Indeed, extending treatment duration in our mouse studies did lead to improved efficacy. Finally, although there was no difference in long-term survival in the study with AML model PDX17-14 in the combination vs revumenib arms, there was a difference in disease burden at time of death (lower burden with combination treatment), suggesting that factors other than disease burden contributed to the death of mice in the combination arm. Although we did not observe overt evidence of toxicity, including in evaluation of liver hematoxylin and eosin stains, future studies to optimize the dose, schedule, and duration of treatment in mice will need to be performed to more comprehensively assess for impact of combination treatment on survival and toxicity, including normal hematopoiesis.³⁷

In summary, RAS mutations in the context of KMT2A-r pediatric AML portend a worse EFS. We observed a synergistic effect of the combination treatment of menin and MEK inhibition in leukemia cell lines and PDX models in vitro and evidence of a greater impact on disease burden with combination therapy in vivo. OS, however, was not improved in the preclinical study, suggesting further evaluation of this combination or other related combinations, such as with new direct RAS inhibitors63, 64, 65, 66, will be needed before translation to patients with KMT2A-r and RAS-mutant AML.

Conflict-of-interest disclosure: K.S. previously received grant funding from the Dana-Farber Cancer Institute/Novartis Drug Discovery Program; and is a member of the scientific advisory board of, and has stock options with, Auron Therapeutics, on topics unrelated to this work. S.A.A. has been a consultant and/or shareholder for Neomorph Inc., C4 Therapeutics, Accent Therapeutics, Stelexis Biosciences, AstraZeneca, Novartis, and Hyku Therapeutics. S.A.A. has received research support from Janssen and Syndax. S.A.A. is an inventor on a patent application related to MENIN inhibition WO/2017/132398A1. The remaining authors declare no competing financial interest.