RAS mutations drive proliferative chronic myelomonocytic leukemia via a KMT2A-PLK1 axis

Abstract

Proliferative chronic myelomonocytic leukemia (pCMML), an aggressive CMML subtype, is associated with dismal outcomes. RAS pathway mutations, mainly NRAS^G12D, define the pCMML phenotype as demonstrated by our exome sequencing, progenitor colony assays and a Vav-Cre-Nras^G12D mouse model. Further, these mutations promote CMML transformation to acute myeloid leukemia. Using a multiomics platform and biochemical and molecular studies we show that in pCMML RAS pathway mutations are associated with a unique gene expression profile enriched in mitotic kinases such as polo-like kinase 1 (PLK1). PLK1 transcript levels are shown to be regulated by an unmutated lysine methyl-transferase (KMT2A) resulting in increased promoter monomethylation of lysine 4 of histone 3. Pharmacologic inhibition of PLK1 in RAS mutant patient-derived xenografts, demonstrates the utility of personalized biomarker-driven therapeutics in pCMML.

Fig. 1. RAS pathway mutations correlate with WHO-defined proliferative chronic myelomonocytic leukemia (pCMML).

A. Kaplan–Meier curve depicting overall survival (OS) of CMML patients from the Mayo-GFM-Austrian cohort stratified by dCMML and pCMML subtypes. B Kaplan–Meier curve depicting AML-free survival. C Prevalence of NRAS and RAS pathway mutations in dCMML and pCMML by next generation sequencing. D Kaplan–Meier curve depicting OS of CMML patients stratified by NRAS wild type and NRAS mutant cases. E Kaplan–Meier curve depicting OS in pCMML patients with NRAS mutations relative to other CMML genotypes. F Kaplan–Meier curve depicting AML-free survival in pCMML patients with NRAS mutations relative to other CMML genotypes. G Violin plots representing variant allele frequencies (VAFs) of the most frequent RAS pathway mutations in dCMML and pCMML. VAF is depicted on the y-axis. Width of horizontal hatches correlates to number of samples with the indicated VAF. H, I Odds ratios of genetic factors selected by L1-regularized logistic regression model that have the greatest impact on clinical parameters. H Impact on dysplastic/proliferative CMML categorization and leukemic transformation (LT) risk. The x-axis represents odds ratios of LT risk. The y-axis represents categorization of low or high white blood cell (WBC) count, or pCMML vs dCMML classification. I Impact on AML-free survival (LFS) and OS. The x- and y-axes represent odds ratios of binarized OS and LFS values, respectively (by the median). The horizontal and vertical size of ellipses reflects corresponding p-values. Confidence intervals (ɑ = 0.05) of odds ratio values are presented. p-values are indicated with each comparison.

Fig. 2. Alterations in driver mutation profiles delineate stages of CMML progression. Whole exome analysis of 48 paired CMML and secondary AML (sAML) samples.

A Prevalence of driver mutations and somatic copy number alterations (SCNAs) in CMML and sAML. RAS pathway represents all mutations in MAPK pathway. B Frequency of driver mutations and SCNAs in CMML and sAML. C Violin plots depicting variant allele frequencies (VAFs) for the most frequent driver mutations in CMML and sAML. D Frequency of driver mutations and SCNAs in CMML and sAML and their categorization as 1d, 2d, or 3d. E Genetic profiling of CMML evolution in eighteen patients at two time points by whole exome sequencing. F Odds ratios of genetic factors selected by the L1-regularized logistic regressions that have the greatest impact on dysplastic/proliferative CMML categorization and leukemic transformation (LT) risk. The x-axis represents odds ratios of LT risk (high or low). The y-axis represents categorization of low or high white blood cell (WBC) count, or pCMML vs dCMML classification. The horizontal and vertical size of ellipses reflects corresponding p-values calculated by two-tailed Fisher’s exact test. Additionally, confidence intervals (ɑ = 0.05) of odds ratio values are presented. The mean is the measure of center. G Stratification of parameters associated with aggressive disease phenotype by number of driver mutations. H Fish plots derived from patient-derived single colony assays of representative cases of pCMML (left) and dCMML (right). Scatter plot to the left of each fish plot represents the genotype of individual colonies as determined by Sanger sequencing of the indicated genes. RAS mutation status is depicted on the x-axis. The arrow indicates inferred evolutionary trajectory from which the fish plot was derived.

Fig. 3. NRAS^G12D mutation drives a proliferative CMML phenotype.

A. Representative hematopoietic progenitor colony formation assay using NRAS-wild-type CMML patient-derived mononuclear cells (MNCs) after transduction with either a null adenoviral construct (control) or NRAS-expressing vector (NRAS). Red circles indicate individual colonies. Scatter plot depicts colony counts at day 12 after inoculation in five individual cases. B Daily cell counts of CMML patient-derived MNCs after siRNA depletion of NRAS in NRAS^G12D cells (above) and transfection of NRAS^G12D in NRAS wild-type cells (below). Representative western blots of three experiments are depicted for validation by assessing phosphorylated ERK (pERK) relative to total ERK (tERK) levels with Vinculin as a loading control. Knockdown of NRAS by qPCR was ≥85%. Indicated p-value in panels A and B by two-tailed Student’s t test. C–F Receiver operating characteristic curve (ROC) analyses illustrating the diagnostic ability of variant allele frequency (VAF) of NRAS mutations (C) or oncogenic RAS pathway mutations (NRAS, KRAS, CBL, PTPN11) (D) at discrimination between pCMML and dCMML phenotypes. If >1 mutation was present in the sample, maximal VAF was used. Impact of VAF_RAS as a predictor of pCMML vs dCMML phenotype, using binary RAS mutation status (E), or a continuous VAF_RAS value (F). Characterization of the Vav-Cre-Nras^G12D mouse model of CMML is depicted at 6 weeks of age (panels G–I) and when moribund (panels J and K). G Differences in peripheral blood (PB, left) and bone marrow (BM, right) monocytes and neutrophils in Nras^G12D mice relative to wild-type controls. H BM cells were serum- and cytokine-starved for 2 h at 37 °C. Cells were then stimulated with or without 2 ng/ml of mGM-CSF for 10 min at 37 °C. Levels of phosphorylated ERK1/2 (pERK) were measured using phosphor-flow cytometry. Myeloid progenitors are enriched in Lin^-/low c-Kit⁺ cells. Indicated p-value in panels G and H by two-tailed Student’s t test. I Histopathologic comparisons between wild-type control (top) and Nras^G12D (bottom) mouse PB, BM, and spleen. PB smears depict monocytes (arrows). BM shows normal hematopoiesis (arrows, above) and megakaryocytic atypia and hyperplasia (arrows, below). Lower power spleen reveals normal white and red pulp (black and red arrows respectively, above) with effacement of this architecture (black and red arrows, below). High power spleen reveals normal white and red pulp architecture (black and red arrows respectively, above) and dysplastic megakaryocytes (arrows, below). J Representative hematopoietic progenitor colony formation assay using MNCs from wild-type control (top) and Nras^G12D (bottom) mice. Bar graph depicts colony counts at day 12 after inoculation. Indicated p-value by two-tailed Student’s t test. K Kaplan–Meier curve demonstrating overall survival of Nras^G12D mice relative to wild-type controls. Data in panels A, B, G, H and J are presented as mean ± SEM. The indicated n represents the number of biologic replicates. P-values are indicated with each comparison. Source data are provided as a Source Data file.

Fig. 4. RAS pathway mutations drive expression of mitotic checkpoint kinase PLK1.

A Unsupervised hierarchical clustering of RNA-seq performed on peripheral blood samples from 35 CMML patients. Cluster 1 (black) with RAS wild-type dCMML cases and Cluster 2 (red) with RAS mutant pCMML cases. B Volcano plot demonstrating differentially upregulated (red) and downregulated (green) genes in RAS mutant pCMML (vs RAS wild-type dCMML) expressed as log2-fold change. Indicated p-values on the y-axis is calculated by the Wald test and false discovery rate-corrected. C Table of most upregulated therapeutically actionable genes in pCMML relative to dCMML. D Quantitative PCR (qPCR) validation of PLK1 expression in patient-derived MNCs from NRAS wild-type (wt) dCMML, NRAS mutant (mt) dCMML, JAK2 mt pCMML, and NRAS mt pCMML cases. E Scatter plot comparing relative PLK1 expression to variant allele frequency (VAF) of NRAS mutation. F qPCR for NRAS and PLK1 in NRAS mutant pCMML MNCs after transfection with non-targeting siRNA (siNT), or siRNA against NRAS (siNRAS). G qPCR for NRAS and PLK1 in NRAS wild-type dCMML patient-derived MNCs after transfection with an empty vector (Control) or NRAS^G12D. H qPCR for KRAS and PLK1 in KRAS mutant pCMML MNCs after transfection with siNT or siKRAS. I qPCR for KRAS and PLK1 in KRAS wild-type dCMML MNCs after transfection with an empty vector (Control), or KRAS^G12D. Western blots in panels F–I are representative of three experiments. Data in panels D and F–I are presented as mean ± SEM. Indicated p-value in panels D–I by two-tailed Student’s t test. The indicated n represents the number of biologic replicates. Source data are provided as a Source Data file.

Fig. 5. Genome-wide and sequence-specific enrichment of the H3K4me1 histone mark in RAS mutant pCMML.

A ChIP-seq on BM MNC from CMML patients assessing relative global enrichment of histone 3 lysine 4 monomethylation (H3K4me1), trimetylation (H3K4me3), and histone 3 lysine 27 trimethylation (H3K27me3). Indicated p-value by two-tailed Student’s t test. B MA plot of differential binding analysis of ChIP-seq H3K4me1 peak data. Each point represents a binding site with each red data point representing sites identified as differentially enriched. C Volcano plot integrating differential enrichment of H3K4me1 at gene promoters in RAS mutant pCMML (vs RAS wild-type dCMML) with differentially upregulated gene expression (purple) in RAS mutant pCMML expressed as log2-fold change. PLK1 is indicated as a red data point. Indicated p-values on the y-axis is calculated by the Wald test and false discovery rate-corrected. D Enrichment and consensus peak calling for H3K4me1 occupancy at the PLK1 locus. Three individual traces from RAS wild-type dCMML (blue, above) and RAS mutant pCMML (red, below) patients are depicted. Relative localization along the PLK1 gene body is depicted. The broad consensus peaks are indicated below each set of patient samples. The accompanying box plot indicates differences in H3K4me1 consensus peak height between pCMML and dCMML samples. Indicated p-value by two-tailed Student’s t test. E–J ChIP-PCR of CMML MNCs assessing PLK1 promoter occupancy of H3K4me1. This was done in NRAS mutant (E) or KRAS mutant (F) MNCs after transfection with siNT or siNRAS/siKRAS. Western blots are representative validation of RAS knockdown. It was also performed in NRAS (G) or KRAS (H) wild-type MNCs after transfection with a control vector or NRAS^G12D/KRAS^G12D. Western blots are representative validation of RAS knockdown. I ChIP-PCR was performed in JAK2 mutant MNCs after transfection with siNT or siJAK2. Validation of JAK2 knockdown is done using phosphorylated STAT3 (pSTAT3) levels. J ChIP-PCR was also performed in NRAS mutant dCMML MNCs after transfection with siNT or siNRAS. Western blots in panels E–J are representative of three experiments. Indicated p-value in panels E–J by two-tailed Student’s t test. Data in panels A, D, and E–J are presented as mean ± SEM. The indicated n represents the number of biologic replicates. Source data are provided as a Source Data file.

Fig. 6. Mutant RAS regulates PLK1 expression through the lysine methyltransferase KMT2A (MLL1).

A RNA-seq expression levels of enzymes potentially involved in monomethylation in pCMML (vs dCMML and healthy volunteers). B Representative fluorescence in situ hybridization (FISH) in two patients assessing for MLLT3-KMT2A fusions (n = 500 biological replicates). Arrows (right) indicate presence of fusion. C ChIP-PCR in NRAS mutant CMML patient-derived MNCs assessing KMT2A occupancy at the promoter of PLK1 with immunoprecipitation of KMT2A relative to isotype control (IgG). D, E ChIP-PCR assessing KMT2A occupancy at the PLK1 promoter in NRAS mutant (D) and KRAS mutant (E) CMML MNC after transfection with siNT or siNRAS/siKRAS. Western blots are representative validation of RAS knockdown from three experiments. F, G ChIP-PCR assessing KMT2A occupancy at the PLK1 promoter in NRAS wild type (F), and KRAS wild type (G) CMML MNC, after transfection with control vector or NRAS^G12D/KRAS^G12D. Western blots are representative validation of RAS expression. H, I ChIP-PCR assessing H3K4me1 at the promoter of PLK1 (H) and qPCR assessing levels of KMT2A and PLK1 (I) in NRAS mutant CMML MNC after transfection with siNT or siKMT2A. Western blots are representative validation of KMT2A knockdown. Western blots in panels D–I are representative of three experiments. Data in panels C–I are presented as mean ± SEM and indicated p-values by two-tailed Student’s t test. The indicated n represents the number of biologic replicates. Source data are provided as a Source Data file.

Fig. 7. Therapeutic efficacy of targeting PLK1 in pCMML.

A Daily cell counts of CMML MNC after transfection with either siNT or siPLK1. Representative western blot depicts validation of PLK1 knockdown from three experiments. Knockdown of PLK1 by qPCR was ≥90%. B Progenitor colony forming assay using CMML MNC with increasing doses of volasertib. Indicated p-value in panels A and B by two-tailed Student’s t test. C, D Histopathologic analysis with H&E staining and immunohistochemistry (IHC) of spleen (C) and BM (D) of murine patient-derived xenografts (PDX) after treatment with vehicle control or volasertib. Magnification is ×200 unless otherwise indicated. hCD45 is human CD45. E Representative flow cytometry of spleen, BM and PB of PDXs after treatment with vehicle control (left) or volasertib (right). The y-axis indicates hCD45 expression status while x-axis indicates murine CD45 (mCD45). Percentages indicate proportion of hCD45+ and mCD45- cells in the respective tissues. F Flow cytometry of proportion of hCD45+ cells in spleen, BM, and PB of PDX mice. Data are presented as mean ± SEM from nine mice. G, H Effect of vehicle versus volasertib treatment on PDX spleen size (G) and weight (H). Data in panel H are presented as mean ± SEM from seven mice treated with vehicle and eight treated with volasertib. Indicated p-value in panels F and H by two-tailed Student’s t test. Data in panels A and B are presented as mean ± SEM. The indicated n represents the number of biologic replicates. p-values are indicated with each comparison. Source data are provided as a Source Data file.