Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Mar;23(3):301-313.
doi: 10.1038/nm.4283. Epub 2017 Feb 13.

The Creatine Kinase Pathway Is a Metabolic Vulnerability in EVI1-positive Acute Myeloid Leukemia

Affiliations
Free PMC article

The Creatine Kinase Pathway Is a Metabolic Vulnerability in EVI1-positive Acute Myeloid Leukemia

Nina Fenouille et al. Nat Med. .
Free PMC article

Abstract

Expression of the MECOM (also known as EVI1) proto-oncogene is deregulated by chromosomal translocations in some cases of acute myeloid leukemia (AML) and is associated with poor clinical outcome. Here, through transcriptomic and metabolomic profiling of hematopoietic cells, we reveal that EVI1 overexpression alters cellular metabolism. A screen using pooled short hairpin RNAs (shRNAs) identified the ATP-buffering, mitochondrial creatine kinase CKMT1 as necessary for survival of EVI1-expressing cells in subjects with EVI1-positive AML. EVI1 promotes CKMT1 expression by repressing the myeloid differentiation regulator RUNX1. Suppression of arginine-creatine metabolism by CKMT1-directed shRNAs or by the small molecule cyclocreatine selectively decreased the viability, promoted the cell cycle arrest and apoptosis of human EVI1-positive cell lines, and prolonged survival in both orthotopic xenograft models and mouse models of primary AML. CKMT1 inhibition altered mitochondrial respiration and ATP production, an effect that was abrogated by phosphocreatine-mediated reactivation of the arginine-creatine pathway. Targeting CKMT1 is thus a promising therapeutic strategy for this EVI1-driven AML subtype that is highly resistant to current treatment regimens.

Conflict of interest statement

COMPETING FINANCIAL INTERESTS

The authors declare no competing financial interests.

Figures

Figure 1
Figure 1. EVI1 overexpression imparts new metabolic dependencies on AML cells
(A) Quantitative comparison by GSEA of the c2 collection of curated gene sets available from MSigDB v.5.0 for murine LinLow Evi1-overexpressing versus control bone marrow samples (GSE 34729). Data are presented as a volcano plot of −log10 (FDR) versus the normalized enrichment score for each evaluated gene set. Triangles indicate sets related to proliferation (green), cell differentiation (blue), and metabolism (red), and gray dots indicate all other c2 gene sets. (B) Metabolic network showing gene sets altered in the Evi1 overexpression signature generated from GSE 34729. Black, dark grey, light grey, and white colors indicate an enrichment odds ratio (OR) above 3, between 2 and 3, between 1 and 2, and below 1, respectively. (C) Pathway analysis integrating enrichment and pathway topology analyses (MetaboAnalyst) of the steady-state metabolite profile from control versus Evi1-overexpressing LinLow murine bone marrow cells. Top 15 out of a total of 45 metabolic pathways are represented based on a FDR ≤ 0.1 and p value ≤ 0.05. “Hits” represents the number of metabolites that scored in the steady-state profile and “Total” represents the number of metabolites present in the given metabolic pathway. Metabolic pathways with a - log10 (FDR) ≥ 1.5 are depicted in red, 1 ≤ − log10 (FDR) < 1.5 are in orange, and − log10 (FDR) < 1 in beige. (D) Scatterplot of hits significantly depleted (in blue) or enriched (in red) from an shRNA library targeting 67 metabolism-related genes in the human TF-1 (top) and UCSD-AML1 (bottom) cell lines. Hits which did not score are depicted in black. Data are presented as volcano plots of −log10 (FDR) versus the RIGER enrichment score for each gene. (E) Heatmap of the metabolism genes differentially expressed by genomic profiling in GSE 34729 upon Evi1 overexpression in LinLow bone marrow cells (left panel), and heatmap of genes either depleted (in blue) or enriched (in red) in the shRNA screen in TF-1 and UCSD-AML1 cells (right panel). Shown in blue text are genes both upregulated upon Evi1 overexpression and depleted in the shRNA screen. CT, control. Each column for each condition represents a technical replicate (n=3 per condition). (F) Growth of TF-1 cells infected with hairpins directed against either CKMT1, SDHA, or CS (left panel). Immunoblot confirming shRNA target knockdown (right panel). shCT, control shRNA. Error bars represent mean ± SD. * and # p value ≤ 0.05 was calculated on the latest time point using a nonparametric Kruskall-Wallis test and Dunn’s multiple comparisons test. Data are representative of two independent experiments.
Figure 2
Figure 2. EVI1-positive cells highly express and are dependent on CKMT1
(A) Immunoblot for EVI1, CKMT1, and HSP60 (loading control) from a panel of human AML cell lines with high or low EVI1 expression. (B-C) Creatine kinase activity (B) and distribution of IC50 (C) for human EVI1 high versus low cell lines treated with cyclocreatine (Ccr). p value calculated using nonparametric Mann-Whitney test. Error bars represent mean ± SD of 8 low EVI1-expressing and four high EVI1-expressing human cell lines. (D) Eleven human AML cell lines were infected with two CKMT1-directed miR30-shRNAs. Growth after treatment with 0.5 μg/ml doxycycline is normalized to the control shRNA and shown relative to day 0 (time of seeding), with error bars representing the mean ± SD of seven technical replicates. Each experiment in A–D was performed at least two independent times. (E) Z Score normalization of EVI1 expression across a panel of CD13/33-gated bone marrow cells from 68 patient AML samples. Red bars indicate the four patients with the highest EVI1 expression level. (F) Immunoblot for EVI1, CKMT1, and GAPDH (loading control) from bone marrow samples sub-selected from panel (E). (G) Distribution of IC50 for EVI1 high versus low patient samples in response to treatment with cyclocreatine (Ccr). Red squares represent the four patient samples from panel (F) highlighted in red. p value calculated using nonparametric Mann-Whitney test. Error bars represent mean ± SD of 10 low EVI1-expressing and four high EVI1-expressing human primary samples.
Figure 3
Figure 3. EVI1-mediated RUNX1 downregulation promotes CKMT1 expression
(A–B) Immunoblot for Evi1, Ckmt1 and Gapdh (A) and qRT-PCR for Ckmt1 expression (B) from LinLow, c-Kit+ murine bone marrow cells infected with an empty or Evi1-encoding construct. Error bars represent the mean ± SD of four technical replicates. * p value ≤ 0.05 in comparison with empty control vector calculated using a Mann-Whitney test. (C) qRT-PCR for CKMT1 expression level in human TF-1 and UCSD-AML1 cells infected with shControl (shCT) or three EVI1-targeting shRNAs. Error bars represent mean ± SD of four technical replicates. * and # p value ≤ 0.05 in comparison with shCT calculated using a Mann-Whitney test. (D) Luciferase reporter assay from human 293E cells co-expressing a vector encoding EVI1, and wild-type or several truncated forms of the human CKMT1 promoter flanking a luciferase cassette. (E) Luciferase reporter assay from human 293E cells co-expressing a vector encoding EVI1 and a luciferase-flanked human CKMT1 promoter either wild-type or deleted for the RUNX1-binding site. (F) Luciferase reporter assay from human 293E cells co-expressing vectors encoding EVI1 and RUNX1. (G) Immunoblot for EVI1, RUNX1, CKMT1, and HSP60 (loading control) from human TF-1 cells infected with the indicated combinations of EVI1- and RUNX1-directed shRNAs. (H) Representation of EVI1 binding peaks on RUNX1 gene tracks available from two ChIP-sequencing experiments performed by Bard-Chapeau et al. and Glass et al.. (I-L) LinLow, c-Kit+ murine bone marrow cells overexpressing a vector encoding Evi1 (I–J), or human UCSD-AML1 cells infected with shControl (shCT) and three EVI1-directed shRNAs (K), or human UCSD-AML1 cells infected with combinations of EVI1- and RUNX1-directed shRNAs (L) following ChIP with the indicated ChIP antibodies (ChIP Ab) and qPCR on the indicated promoter regions. Results are represented as fold enrichment over input. (D–L) Error bars represent mean ± SD of three technical replicates.* p value ≤ 0.05 calculated using a Welch’s t-test. Each experiment in A–L was performed at least two independent times. (M) TF-1-Luc cells engineered to overexpress CSF2 for efficient engraftment were infected with a combination of two EVI1- and one RUNX1-directed shRNAs before tail vein injection. Bioluminescence was quantified weekly as a measure of disease burden. Data are represented as mean ± SEM of 5 mice per cohort. * p value ≤ 0.05 was calculated on the latest time point using nonparametric Mann-Whitney test. (N) Bar graph showing the number of primary patient AML samples with either EVI1high/RUNX1low, EVI1high/RUNX1high, EVI1low/RUNX1low, or EVI1low/RUNX1high expression and displaying CKMT1 high versus low expression level (absolute cut-off of a z-score ≥ 0.58 is high expression and a z-score ≤ − 0.58 is low expression) in two cohorts: GSE14468 and GSE10358. p value calculated using a Fisher exact test.
Figure 4
Figure 4. Blockade of the arginine-creatine pathway upon CKMT1 inhibition impairs both mitochondrial respiration and ATP production in EVI1-positive AML cells
(A) Schematic of arginine metabolism and enzymes involved in this process. (B–C) Heatmaps of metabolite products of arginine metabolism that are differentially regulated in human UCSD-AML1 cells infected with either control (shCT) or CKMT1-directed shRNAs (B) or in human TF-1 cells treated with 3 mM cyclocreatine (Ccr) for 12 hours in combination with 1 mM phospho-creatine (P-Cr) (C) before a 30-min pulse labeling with L-Arginine 13C6. Metabolites scoring with a p value ≤ 0.05 are shown as log2 fold change normalized to the average control condition. Each column for each condition represents a technical replicate (n=3 per condition). (D–E) Growth inhibition of human TF-1 and UCSD-AML1 cells treated with increasing concentrations of cyclocreatine (Ccr) (D) or infected with control (shCT) or CKMT1-directed shRNAs (E) in combination with 1 mM phospho-creatine (P-Cr). Error bars represent mean ± SD of five technical replicates.* p value ≤ 0.05 in comparison with shControl calculated using a Mann-Whitney test. (F) Colony-forming assay for 5 rounds of serial replating from control or Evi1-overexpressing murine LinLow, c-Kit+ bone marrow cells treated with 3 mM cyclocreatine (Ccr) in combination with 1 mM phospho-creatine (P-Cr). # p value ≤ 0.05 calculated using Welch’s t-test in comparison with the control condition. Error bars represent mean ± SEM of three technical replicates. (G) Heatmap of the top upregulated or downregulated metabolites (p value ≤ 0.05 and absolute log2 fold change ≥ 1.5 in at least one of the two CKMT1-directed shRNAs condition versus average shControl condition) identified by steady-state metabolism profiling in human TF-1 and UCSD-AML1 cells infected with either control (shCT) or two CKMT1-directed shRNAs. Each column for each condition represents a technical replicate (n=3 per condition in two individual human EVI1-positive cell lines). (H) Pathway analysis integrating enrichment and pathway topology analyses (MetaboAnalyst) of the list of top metabolites from panel (G). Top enriched metabolic pathways are represented based on a FDR ≤ 0.25 and p value ≤ 0.05. “Hits” represents the number of metabolites that scored in the steady-state profile and “Total” represents the number of metabolites present in the given metabolic pathway. Metabolic pathways with a − log10 (FDR) ≥ 1.5 are depicted in red, 1 ≤ − log10 (FDR) < 1.5 are in orange, and − log10 (FDR) < 1 in beige. (I) ADP / ATP ratio in a panel of 12 human AML cell lines treated with 5 mM cyclocreatine (Ccr) for 12 hours. * p value ≤ 0.05 in comparison with the control condition calculated using a Mann-Whitney test. Error bars represent mean ± SD of four technical replicates. (J) Intracellular ATP level in the indicated human EVI1-positive AML cell lines treated with 5 mM cyclocreatine (Ccr) in combination with 1 mM phospho-creatine (P-Cr) for 12 hours. * and # p value ≤ 0.05 calculated using a Mann-Whitney test in comparison with the control condition and Ccr-treated condition respectively. Error bars represent mean ± SD of four technical replicates. (K) Mitochondrial Oxygen Consumption Rate (OCR) of human UCSD-AML1 and TF-1 cells treated with 5 mM cyclocreatine (Ccr) in combination with 1 mM phospho-creatine (P-Cr) for 12 hours prior to measurement under basal conditions, following sequential injections of 1 μM ATP synthase inhibitor oligomycin, 1 μM of the uncoupling agent FCCP, and 1 μM of the complex I and III inhibitors rotenone and antimycin A, respectively. * p value ≤ 0.05 calculated using a Mann-Whitney test in comparison with the control condition. Error bars represent mean ± SD of 4 technical replicates. Each experiment was performed at least two independent times except in G and H in which two different EVI1-positive human cell lines transduced with two different CKMT1-directed shRNAs were profiled using metabolomics.
Figure 5
Figure 5. Inhibition of the creatine kinase pathway alters the viability of EVI-1-positive AML cells via cell cycle blockade and apoptosis induction
(A) Heatmap of the top common up- and down-regulated genes in human TF-1, UT-7, and UCSD-AML1 cell lines treated with cyclocreatine for 24 hours. p value ≤ 0.05, FDR ≤ 0.05, and absolute fold change for log2 (FPKM) scores ≥ 1.5. Each column for each condition represents an independent biological replicate (n=3 per condition in three individual human EVI1-positive cell lines). (B–C) A cyclocreatine signature identified by RNAseq was interrogated by GSEA across the MsigDB and DMAP database for vehicle versus cyclocreatine-treated AML cells. (B) Top enriched upregulated (in red) and downregulated (in blue) gene sets upon cyclocreatine treatment are depicted in a functional network. (C) Quantitative comparison of these gene sets of interest (in blue or red) versus all other available gene sets (in gray). Data are presented as a volcano plot of -log10 (p value) versus Normalized Enrichment Score (NES) for each evaluated gene set. (D–E) May-Grünwald-Giemsa staining of indicated human AML cell lines treated with 8 mM cyclocreatine for 9 days (D), or infected with a control (shCT) or a CKMT1-directed shRNA (E). (F) Cell size measurement of indicated human AML cell lines treated with 8 mM cyclocreatine for 9 days (top panel) or infected with CKMT1-directed shRNAs (bottom panel). * p value ≤ 0.05 calculated using Welch’s t-test in comparison with the control condition. Error bars represent mean ± SD of three technical replicates. (G) Cell cycle analysis at the indicated time points of human EVI1-positive cell lines treated with cyclocreatine. (H) FACS plots showing Annexin V (AV) / propidium iodide (PI) profile of the indicated human AML cell lines treated with 8 mM cyclocreatine (Ccr) for 5 days. (I) FACS analysis of the expression of CD117 cell surface marker in the indicated human AML cell lines treated with 8 mM cyclocreatine for 9 days (bottom panel) or infected with CKMT1-directed shRNAs (top panel). A representative FACS plot from each cell line is shown. (J) Colony formation assay for 2 to 3 rounds of serial replating from indicated human AML cell lines treated with either 3 mM or 10 mM cyclocreatine (Ccr). Results represent the average of triplicate assays. * p value ≤ 0.05 calculated using a Welch’s t-test in comparison to the control condition. Error bars represent mean ± SEM. Each experiment in D–J was performed at least two independent times.
Figure 6
Figure 6. CKMT1 knockdown preferentially impairs development of EVI1-positive human and murine myeloid leukemias, without affecting normal progenitor cell viability
(A) Western immunoblot for Evi1/EVI1, Ckmt1/CKMT1, and Gapdh/GAPDH (loading control) in murine bone marrow cells from wild-type (WT), N-RasG12D, or N-RasG12D + Evi1 mouse. Human UCSD-AML1 cell line is used as positive control for EVI1 and CKMT1 expression. (B) Immunoblot indicating the level of Ckmt1 knockdown in murine RFP-sorted bone marrow cells seven days post-doxycycline induction. (C) Kaplan-Meier curves showing overall survival of mice (n = 5 for each group except for N-RasG12D + Evi1 shCT + dox group, for which n = 6) transplanted with cells expressing each combination of indicated constructs. Statistical significance determined by log-rank (Mantel-Cox) test. * p value ≤ 0.05 by comparison with shControl (shCT) within the N-RasG12D or the N-RasG12D + Evi1 group. (D) Spleen weight and white blood cell (WBC) count of four mice per group when shControl mice became moribund. P value calculated using a Mann-Whitney test. Error bars represent mean ± SEM. (E) Proportion of Mac-1+ / Gr-1+ cells in spleens from four mice per group. p value calculated using a Mann-Whitney test. Error bars represent mean ± SEM. (F) Proportion of CMP (CD16/32/CD34+), GMP (CD16/32+/CD34+), and MEP (CD16/32/CD34) cell populations on gated LinLow/Sca-1/c-Kit+ myeloid progenitors when shControl became moribund. RFP and Venus expression were evaluated on every leukemic cell population. One representative moribund mouse from each group is shown. (G–H) Low U937-Luc (G) and high TF-1 CSF2+-Luc (H) EVI1-expressing human cell lines were either infected with two doxycycline-inducible CKMT1-directed miR30-shRNAs (top panel) before tail vein injection, or treated with 1g/kg/day cyclocreatine (Ccr) 10 days post-injection (bottom panel). TF-1-Luc cells were engineered to overexpress CSF2 for efficient engraftment in mice. Bioluminescence was quantified weekly as a measure of disease burden. Arrows indicates beginning of doxycycline (Dox) or cyclocreatine (Ccr) treatments. Data are represented as mean ± SEM of 5 mice per cohort. p value was calculated on the latest time point using nonparametric Mann-Whitney test. Kaplan-Meier curves showing overall survival of each group of mice is shown in the right panel. Statistical significance determined by log-rank (Mantel-Cox) test. n.s, not significant. (I) Low U937-Luc (top panel) and high TF-1 CSF2+-Luc (bottom panel) EVI1-expressing human cell lines were treated with 1g/kg/day cyclocreatine (Ccr) in combination with phospho-creatine (P-Cr) at the indicated time point. Bioluminescence was quantified weekly as a measure of disease burden. Data are represented as mean ± SEM of seven mice per cohort. p value was calculated on the latest time point using a nonparametric Kruskall-Wallis test and Dunn’s multiple comparisons test. Kaplan-Meier curves showing overall survival (right panel). Statistical significance determined by log-rank (Mantel-Cox) test. * and # p value ≤ 0.05 by comparison with vehicle- and Ccr-treated groups respectively. n.s, not significant.

Similar articles

See all similar articles

Cited by 12 articles

See all "Cited by" articles

MeSH terms

Feedback