Aberrant Stemness Transcription Signature Unveils a Mechanism of Chemotherapy Resistance through Blunting p53-Mediated Response in Acute Myeloid Leukemia

Stem cell transcriptional signatures are linked to poor outcomes in AML. Conventional chemotherapy (+/-HSCT) remains the only curative approach for AML, and survival largely reflects chemotherapy sensitivity. The mechanism by which a stem-cell-like transcriptional signature promotes chemotherapy resistance and relapse has not been determined. We identified a direct link between stemness and p53-mediated apoptosis involving a GATA2-RASSF4-MDM2-p53 axis.

To interrogate transcriptional states that might mediate resistance in AML patients, we conducted single-cell RNA sequencing and single-cell ATAC sequencing on 10 pediatric AML samples. We found that expression of the stem/progenitor cell transcription factor GATA2 identified an immature cell population that partially overlaps with populations characterized by ERG and MECOM expression and chromatin accessibility.

Using a Gata2 conditional retroviral KMT2A-MLLT3 mouse model, we established a model to study functional consequences of distinct GATA2 expression levels. This analysis revealed that AML originating from a murine stem cell or immature progenitor cell population showed higher Gata2 expression and were more resistant to standard AML chemotherapy agents. While GATA2 was not strictly required for leukemogenesis, Gata2^high cells were more resistant to doxorubicin than Gata2^low cells. In vitro treatment of bulk leukemias with doxorubicin resulted in selection for Gata2^high cells. GATA2 expression was also increased in patient AML samples at relapse compared to initial diagnosis.

Genetic inactivation of Gata2 profoundly enhanced chemosensitivity in Gata2^high, but not Gata2^low leukemias. Deletion of Gata2 in Gata2 ^high cells increased activation of p53-mediated apoptosis in response to nutlin-3. Importantly, loss of Gata2 decreased MDM2 protein stability indicating that GATA2 may blunt the p53 response by enhancing MDM2 stabilization. ChIP-Seq and RNA-Seq in murine KMT2A-MLLT3 leukemias showed that neither p53 nor Mdm2 are transcriptionally regulated by GATA2. We next sought to identify a potential regulator of MDM2 stability. Integrated RNA-Seq and ChIP-Seq data nominated Rassf4, which was significantly upregulated upon Gata2 ablation, and GATA2 occupied its promoter. Similar to leukemia cells, decreasing Gata2 expression in normal hematopoietic stem cells increased Rassf4 expression, a pattern that is partially rescued by re-expressing Gata2. The RASSF family has been shown to promote apoptosis by modulating MDM2 protein stability. We functionally validated whether the relationship between high Gata2 expression and drug resistance is mediated by Rassf4 repression. Gata2^-/- MA9 cells transduced with Rassf4 sgRNAs gained a competitive growth advantage under the pressure of nutlin-3a treatment versus cells transduced with non-targeting sgRNA indicating the emergence and expansion of drug-resistant cells. Conversely, overexpressing Rassf4 in Gata2^high MA9 cells sensitized them to nutlin-3a and chemotherapy. Furthermore, GATA2 and RASSF4 expression inversely correlated in our AML patient sample scRNA-Seq dataset, as well as publicly available AML patient datasets, supporting the notion that a GATA2 mechanism to suppress RASSF4 applies to multiple AML subtypes. We evaluated whether MDM2 inhibitors, such as Idasanutlin, in combination with doxorubicin, could overcome the drug resistance seen in Gata2^high leukemias. Combination therapy prolonged survival in vivo.

In conclusion, we identified a novel role for GATA2 in blunting p53-mediated apoptosis via transcriptional repression of Rassf4, a regulator of MDM2 protein stability. Our study supports a model where the “volume control” of p53-mediated apoptosis by a stem cell transcription factor is an integral part of stemness, which is imparted on leukemic cells arising from a stem-cell-like cell-of-origin. Our findings provide a mechanistic explanation for the well-established, but thus far unexplained observation that the expression of HSC signatures are associated with poor outcomes in AML.