Targeting the Creatine Kinase Pathway in EVI1-Positive Acute Myeloid Leukemia

Abnormal expression of the transcription factor EVI1 through chromosome 3q26 rearrangements has been implicated in the development of one of the most therapeutically challenging high-risk subtypes of acute myeloid leukemia (AML). Here we integrated genomic and metabolic screening of hematopoietic stem cells to reveal that EVI1 overexpression altered cellular metabolism. A pooled shRNA screen targeting metabolic enzymes identified the ATP-buffering, mitochondrial creatine kinase CKMT1 as a druggable dependency in EVI1-positive AML.

Of 18 screened AML cell lines harboring various genetic alterations, only the four EVI1-expressing lines exhibited markedly elevated CKMT1 protein expression and activity. Treatment of this cell line panel with either CKMT1-targeting shRNAs or cyclocreatine, an analog of the CKMT1 substrate creatine and inhibitor of the creatine biosynthesis pathway, showed that elevated CKMT1 protein expression correlated with sensitivity to CKMT1 pathway inhibition. Consistent with these data, flow cytometry analysis of a panel of 68 unselected primary AML patient specimens revealed that the four leukemias with the highest levels of EVI1 expression also had elevated CKMT1 protein levels and enhanced sensitivity to cyclocreatine treatment.

We next established that enforced EVI1 expression increased CKMT1 protein and mRNA levels and that three independent shRNA molecules targeting EVI1 drastically reduced CKMT1 expression in two EVI1-positive AML cell lines. A luciferase-based reporter system established that RUNX1 represses CKMT1 expression through direct binding to its promoter. ChIP-qPCR approaches were then applied to dissect the sequential events involved in EVI1-induced CKMT1 upregulation and the possible role of RUNX1 as an intermediate. In both primary AML samples and cell lines, we determined that EVI1 represses RUNX1 expression by directly binding to its promoter. This, in turn, eliminates repressive RUNX1 binding at the CKMT1 promoter and thereby promotes CKMT1 expression. Based on these data, we explored the relationship between EVI1 and RUNX1 expression with CKMT1 mRNA levels in two AML transcriptional datasets (GSE14468 and GSE10358). We divided these cohorts into four subgroups with high versus low expression of EVI1 and RUNX1. Consistent with our mechanistic analysis, primary AML samples within the EVI1^high/RUNX1^low subgroup were significantly more likely to express high levels of CKMT1 than AML samples in the other three subgroups.

CKMT1 promotes the metabolism of arginine to creatinine. To determine the effect of CKMT1 suppression on this pathway, we measured the metabolic flux of stable-isotope labeled L-arginine ¹³C₆ through creatine synthesis using mass spectrometry. CKMT1-directed shRNAs or cyclocreatine selectively decreased intracellular phospho-creatine and blocked production of ATP by mitochondria. Salvage of the creatine pathway by exogenous phospho-creatine restored normal mitochondrial function, prevented the loss of viability of human EVI1-positive AML cells induced by cyclocreatine or CKMT1-directed shRNAs, and maintained the serial replating activity of Evi1-transformed bone marrow cells.

Primary human EVI1-positive AML is frequently associated with somatic NRAS mutations. Thus, to investigate whether EVI1 over-expression sensitizes primary AMLs to CKMT1 inhibition in vivo, we transplanted primary Nras^G12D mutant AMLs with and without elevated Evi1 expression into congenic recipient mice. In this system, Ckmt1 knockdown did not significantly alter the outgrowth of control Nras mutant AML cells compared to a shControl (63% versus 71%). By contrast, Nras^G12D AML cells characterized by elevated Evi1 expression were profoundly depleted by Ckmt1 suppression to 2% versus 58% in shControl recipients. Consistent with these results, pharmacologic or genetic inhibition of the CKMT1-dependent pathway blocked disease progression and prolonged the survival of mice injected with human EVI1-positive cells but not with EVI1-negative cells, without noticeable cytotoxic effect on normal murine cells.

In conclusion, we have integrated "omic" approaches to identify CKMT1 as a druggable liability in EVI-positive AML. This study supports a potential therapeutic avenue for targeting the creatine kinase pathway in EVI1-positive AML, which remains one of the worst outcome subtypes of AML.