AUTHORS
Yara Raquel Rodriguez, PhD, Gabriela Alexe, PhD2, Audrey Taillon, PhD3, Mariya Atanasova4, Paige Stocker5, Allen Basanthakumar6, Olasunkanmi Olaoye, PhD7, Silvi Salhotra, PhD8, Kieren Marini, PhD9, Lucy Merickel10, Constanze Schneider, PhD11, Oivin Guicher12, Kimberly Stegmaier, MD13
AFFILIATIONS
1 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
2 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
3 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
4 Broad Institute, Cambridge, Massachusetts, United States
5 Broad Institute, Cambridge, Massachusetts, United States
6 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
7 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
8 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
9 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
10 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
11 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
12 Broad Institute, Cambridge, Massachusetts, United States
13 Broad Institute, Cambridge, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
ABSTRACT
Acute myeloid leukemia (AML) treatment has advanced substantially over the past decade. Patients with TP53-mutant (TP53-Mut) AML, however, continue to have dismal outcomes, with a median overall survival of only seven months. Using the Cancer Dependency Map (DepMap), we identified a TP53-Mut AML vulnerability network composed of ADAR1, PRKRA, IPPK, and USP18. These genes suppress endogenous dsRNA sensing by Protein Kinase R (PKR), encoded by the gene EIF2AK2, preventing PKR from triggering cell death in response to elevated self-derived dsRNAs. Knockout of these genes induced PKR activation and cell death in TP53-mut AML.
We determined that TP53-Mut AML cells harbor higher basal dsRNA levels than TP53-WT AML and rely on viral-mimicry conditioning to restrain PKR activation, preventing PKR from triggering protein translation inhibition and cell death. ADAR1 catalyzes A-to-I editing of dsRNAs, altering their conformation to enable evasion from cytoplasmic dsRNA sensors. Both ADAR1 expression and editing activity are elevated in TP53-Mut versus TP53-WT AML cells and TCGA patient AML samples. Accordingly, ADAR1 knockout activates innate immune signaling and triggers cell death to a greater extent in TP53-Mut than TP53-WT cells. In vivo, knockout of ADAR1 reduced the leukemic burden in both bone marrow (54.4% vs. 24.2%, p<0.001) and spleen (24.6% vs. 11.5%, p<0.001), as measured by human CD45+ cell abundance in a TP53-Mut AML xenograft model.
To define the immunogenic dsRNA repertoire underlying this dependency, we used dsRNA structural predictions coupled with RIP-sequencing of RNAs bound to the dsRNA sensors MDA5 and PKR in ADAR1-knockout AML cells. For MDA5, whose putative dsRNA target sites are heavily A-to-I edited, we integrated this list with mapped A-to-I edited sites. Because PKR targets do not require A-to-I editing, we did not impose this additional filter. All told, we nominated over 5300 candidate immunogenic dsRNA regions. These dsRNA arms were predominantly hosted in the 3'UTRs of protein-coding genes, were enriched in Alu elements, and mapped to 639 genes that are enriched by Gene Set Enrichment Analysis (GSEA) for interferon and other immune signatures, dsRNA sensing pathways, and cellular stress.
Notably, the PKR (EIF2AK2) 3'UTR was itself among these immunogenic regions. Overexpressing the EIF2AK2 3'UTR upstream of a BFP reporter in an AML cell line with ADAR1 knockout increased IFNβ and ISG15 expression relative to BFP-only controls. Moreover, overexpression of the EIF2AK2 3'UTR in combination with ADAR1 knockout increased phosphorylation of both PKR and eIF2α, suggesting that this 3’UTR also self-regulates PKR kinase activity.
To directly exploit the vulnerability of TP53-Mut AML cells to the hyperactivation of PKR, we performed a high-throughput screen of 6,778 FDA-approved and investigational compounds and identified erdafitinib (FDA-approved) and rogaratinib as non-canonical PKR activators. Both compounds are reported FGFR inhibitors. We found, however, that they activate PKR by promoting its dimerization independent of FGFR inhibition. When combined with IFNβ treatment, both compounds strongly trigger activation of the integrated stress response (ISR), evidenced by increased expression of canonical ISR genes (DDIT3, ASNS, CHAC1, and PPP1R15A) by RNA-sequencing, increased phosphorylation of eIF2α, and increased ATF4 protein levels. The activation of the ISR response led to inhibition of protein translation and subsequent cell death. This effect was ablated by EIF2AK2 (PKR) knockout, confirming that PKR mediates the phenotype. In a TP53-Mut AML orthotopic xenograft, erdafitinib combined with pegylated IFNβ produced the greatest reduction in the percentage of human CD45+ leukemic burden in the bone marrow (vehicle = 54.9%, erdafitinib = 41.3%, IFNβ = 20.7%, combination = 10.4%). The combination significantly reduced leukemic burden compared with erdafitinib alone (p<0.001) and pegylated IFNβ alone (p<0.05).
Together, these findings reveal PKR hyperactivation lethality in TP53-Mut AML and the mechanism of action behind this liability. Importantly, these data nominate the combination of erdafitinib and IFNβ as a promising strategy for further study in this aggressive, treatment-resistant disease.
DISCLOSURES
Yara Raquel Rodriguez, PhD: No relevant conflicts of interest to declare.