Abstract
Targeting NPM1, the most common abnormality in acute myeloid leukemia (AML), is still a significant challenge. Numerous efforts have focused on disrupting oligomerization, blocking nucleocytoplasmic shuttling, or promoting selective degradation of the mutant protein. Despite promising results, the clinical translation of these findings has been limited, and no NPM1-targeted therapies have progressed to advanced stages of clinical development. Therefore, identifying compounds that target NPM1-mutated AML cells, especially in the context of co-occurring high-risk genetic events, remains a translational priority.
We demonstrated that the RNA-binding protein PA2G4 (Proliferation-Associated 2G4) links MECOM (EVI1) to Myc signaling in 3q26 AML, highlighting PA2G4 as a potentially druggable driver in this disease. Given the central role of the Myc pathway in the pathogenesis of high-risk AML, we investigated the contribution of PA2G4 beyond 3q26 AML. We performed PA2G4 pulldown followed by nLC-MS/MS mass spectrometry in AML cell lines and demonstrated that 167 proteins were significantly enriched (P < 0.05) in the PA2G4 pulldown compared to the IgG-pulldown control. Only 40 of these proteins were part of chromatin-bound complexes associated with MECOM. Among the others, we isolated NPM1. Low-throughput validation studies conducted in models carrying NPM1 mutations (NPM1c), such as IMSM2 and OCI/AML3, confirmed PA2G4 binding to NPM1. Consistently, knockdown of PA2G4 using shRNA resulted in decreased levels of the PA2G4-NPM1 complex, as assessed by co-immunoprecipitation.
Next, we asked whether the disruption of PA2G4 interferes with NPM1c AML proliferation. To do this, we used WS6, a diarylurea-based compound reported to functionally inhibit PA2G4. However, to demonstrate that the effects of WS6 were due to PA2G4 binding, we functionalized the compound by tagging it with biotin at the terminal piperazine ring. Immunoprecipitation experiments, using an anti-biotin antibody, confirmed that WS6 interacts with PA2G4, supporting our validation efforts.
Cell lines and primary samples harboring NPM1 mutations exhibited significant sensitivity to WS6 treatment, with an IC50 ranging from 0.008 to 1.8 μM. Compared to cytarabine, azacitidine, or venetoclax, WS6 proved to be more or equally potent in these models and induced apoptosis in all NPM1c models. Following WS6 treatment, NPM1c proteins diminish more than their wild-type (WT) counterparts, supporting the hypothesis that NPM1c proteins are more sensitive to PA2G4 inhibition.
To understand the mechanism driving efficacy in NPM1c cells, we tested WS6 in IMSM2 cells after 24 hours of treatment. Quantification of transcript abundance revealed that 1079 genes were significantly upregulated and 1264 downregulated (adj.P < 0.05). Gene expression signature analysis identified a significant enrichment of genes involved in ribosome biogenesis (adj.P = 7.08x10-14) and in E2F Targets (adj.P = 4.5x10-44), indicating nucleolar stress induction following treatment. To evaluate this context, we examined the relocalization of NPM1 WT and NPM1c following WS6 treatment using immunofluorescence analysis and specific antibodies for mutants. The AML cells exhibited nucleolar disaggregation, resulting in the redistribution of NPM1 WT from the nucleolus to the nucleoplasm, whereas NPM1c accumulated near the nuclear membrane. As anticipated by the hypothesized mechanism, this effect led to an interaction between NPM1 and HDM2, resulting in the stabilization of p53, as shown by western blotting.
To validate the anti-leukemic effects of WS6, we established a patient-derived leukemia xenograft (PDLX) model transplanting NPM1-mutated AML cells in NOD/SCID mice. Mice were treated intraperitoneally with WS6 25 mg/kg per day for 15 days. Flow cytometry analysis at the end of treatment showed a significant decrease in human CD45⁺ cells in the bone marrow (P < 0.05). This result was further supported by immunohistochemistry, which showed a reduction in Ki-67, a marker of cell proliferation, in the bone marrow biopsies (P < 0.0001). Additionally, we observed a marked decrease in NPM1c transcript levels (fold change relative to control = 0.00086), as measured by quantitative real-time PCR.
In conclusion, the identification of PA2G4 as a direct interactor of NPM1 uncovers a novel vulnerability in NPM1c-mutated AML, particularly in cases where NPM1c appears to lose its association with a favorable prognosis.
