Overcoming leukemia stem cells via targeting AKT-wnt/β-catenin by AI-powered mRNA-delivery of peptides

Abstract Leukemia stem cells (LSCs) are key drivers of chemoradiotherapy resistance, immune evasion, and frequent relapse of leukemia. Despite achieving initial remission, 30-50% of acute myeloid leukemia (AML) patients eventually relapse, even after allogeneic stem cell transplantation. Relapse samples show a 10- to 90-fold increase in LSC frequency, further implicating LSC's role in leukemia recurrence. The Wnt/β-catenin and PI3K/AKT pathways are frequently dysregulated in many cancers, and are crucial for LSCs' self-renewal and immune escape. β-catenin directly regulates the expression of multiple immune checkpoint genes. Additionally, AKT promote Wnt/β-catenin by phosphorylating and inhibiting GSK-3β to promote β-catenin accumulation, and directly phosphorylating β-catenin at Ser552 and Ser675 to enhance the nuclear transcriptional activity of β-catenin. Thus, hyperactivating the AKT-Wnt/β-catenin signaling plays a critical role in supporting LSCs' self-renewal and survival. The intricate relationship between the AKT-Wnt/β-catenin pathway and the mechanisms of immune escape and cancer recurrence underscores the urgent need for targeted interventions. We previously demonstrated that low-dose doxorubicin inhibits AKT-mediated Ser552 phosphorylation of β-catenin, and reduces LSCs in both T-cell acute lymphoblastic leukemia (T-ALL) mouse models and in a pilot clinical trial in refractory AML patients. However, due to topoisomerase-II binding, doxorubicin's cardiotoxicity limits its clinical utility. Over 20% of pediatric ALL survivors develop grade II cardiotoxicity post-chemoradiotreatment. Peripheral neuropathy occurs in approximately 38% of patients receiving neurotoxic chemotherapies, causing long-term sensory and motor deficits. Hence, there is an urgent need for safer LSC-targeting strategies. To address these challenges, we developed Artificial intelligence-powered, mRNA-delivery of Peptides (ARP), an integrated approach that combines AI-guided peptide designing with lipid-nanoparticle (LNP)-facilitated mRNA delivery for precise intracellular targeting of the AKT-β-catenin interaction. Leveraging AlphaFold-simulated docking in silico, and reporter assays in HEK293-AKT-β-catenin-TOPFlash-luciferase cells in vitro, we validated several highly potent peptide inhibitors, including β-catenin fragment 529-581. Next, we used RFdiffusion to generate 2,000 de novo binders targeting two hydrophobic hotspots within the AKT kinase-pocket. Following a streamlined de novo binder screening pipeline in silico, including competitive binding assay, immunogenicity and toxicity filters, we selected the top 35 candidates. Meanwhile, we created a library of 416 β-catenin fragments centered on the AKT phosphorylation site at Ser552. We then trained a machine-learning model by integrating (1) reporter assay data from these fragments in HEK293-AKT-β-catenin-TOPFlash-dGFP cells and (2) six structural and energetic features obtained from their AlphaFold3 docking conformations. Through this model, we ranked the last 35 candidates and predicted the top-ranked de novo peptide, Dnv, demonstrated its superior binding and stability in silico, potent inhibition in reporter cells, and enhanced stability relative to 529-581 in vitro. ARPs were then tested in SclCreER;Ptenfl/f;Ctnnb1Dex/+3(β-catenin) T-ALL mouse model. Linear or circular RNA encoding β-catenin fragments or Dnv was packaged in LNP and intravenously administered. ARP treatments were well tolerated and significantly reduced LSCs, partially restored hematopoietic stem/progenitor cells without obvious cardiotoxicity or hepatotoxicity. Competitive transplantation assays revealed that unlike control, ARP-treated bone marrow cells didn't drive leukemogenesis. ScRNA-seq of bone marrow cells confirmed a significant decrease in LSCs. Moreover, we found a pronounced shift of CD8⁺ T-cells that β-catenin fragments or Dnv treatment decreased stem-like CD8+ T memory cells, which are induced by elevated Wnt/β-catenin signaling. Effector and effector memory cells expanded, and their expressions of anti-tumor cytotoxic genes were elevated, accompanied by downregulation of exhaustion markers. Together, these results establish ARP as a versatile strategy to target intracellular AKT-Wnt/β-catenin signaling. By simultaneously disrupting LSCs' maintenance and reprogramming anti-tumor immunity, ARP offers a promising avenue for overcoming relapse and enhancing therapeutic responsiveness in leukemia.

Institutions

Publication Details

Journal
Blood
Published
2025-11-03
DOI
https://doi.org/10.1182/blood-2025-5078
Primary Topic
RNA Interference and Gene Delivery
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Overcoming leukemia stem cells via targeting AKT-wnt/β-catenin by AI-powered mRNA-delivery of peptides

Blood
RNA Interference and Gene Delivery
article

Overcoming leukemia stem cells via targeting AKT-wnt/β-catenin by AI-powered mRNA-delivery of peptides

article en

Abstract

Abstract Leukemia stem cells (LSCs) are key drivers of chemoradiotherapy resistance, immune evasion, and frequent relapse of leukemia. Despite achieving initial remission, 30-50% of acute myeloid leukemia (AML) patients eventually relapse, even after allogeneic stem cell transplantation. Relapse samples show a 10- to 90-fold increase in LSC frequency, further implicating LSC's role in leukemia recurrence. The Wnt/β-catenin and PI3K/AKT pathways are frequently dysregulated in many cancers, and are crucial for LSCs' self-renewal and immune escape. β-catenin directly regulates the expression of multiple immune checkpoint genes. Additionally, AKT promote Wnt/β-catenin by phosphorylating and inhibiting GSK-3β to promote β-catenin accumulation, and directly phosphorylating β-catenin at Ser552 and Ser675 to enhance the nuclear transcriptional activity of β-catenin. Thus, hyperactivating the AKT-Wnt/β-catenin signaling plays a critical role in supporting LSCs' self-renewal and survival. The intricate relationship between the AKT-Wnt/β-catenin pathway and the mechanisms of immune escape and cancer recurrence underscores the urgent need for targeted interventions. We previously demonstrated that low-dose doxorubicin inhibits AKT-mediated Ser552 phosphorylation of β-catenin, and reduces LSCs in both T-cell acute lymphoblastic leukemia (T-ALL) mouse models and in a pilot clinical trial in refractory AML patients. However, due to topoisomerase-II binding, doxorubicin's cardiotoxicity limits its clinical utility. Over 20% of pediatric ALL survivors develop grade II cardiotoxicity post-chemoradiotreatment. Peripheral neuropathy occurs in approximately 38% of patients receiving neurotoxic chemotherapies, causing long-term sensory and motor deficits. Hence, there is an urgent need for safer LSC-targeting strategies. To address these challenges, we developed Artificial intelligence-powered, mRNA-delivery of Peptides (ARP), an integrated approach that combines AI-guided peptide designing with lipid-nanoparticle (LNP)-facilitated mRNA delivery for precise intracellular targeting of the AKT-β-catenin interaction. Leveraging AlphaFold-simulated docking in silico, and reporter assays in HEK293-AKT-β-catenin-TOPFlash-luciferase cells in vitro, we validated several highly potent peptide inhibitors, including β-catenin fragment 529-581. Next, we used RFdiffusion to generate 2,000 de novo binders targeting two hydrophobic hotspots within the AKT kinase-pocket. Following a streamlined de novo binder screening pipeline in silico, including competitive binding assay, immunogenicity and toxicity filters, we selected the top 35 candidates. Meanwhile, we created a library of 416 β-catenin fragments centered on the AKT phosphorylation site at Ser552. We then trained a machine-learning model by integrating (1) reporter assay data from these fragments in HEK293-AKT-β-catenin-TOPFlash-dGFP cells and (2) six structural and energetic features obtained from their AlphaFold3 docking conformations. Through this model, we ranked the last 35 candidates and predicted the top-ranked de novo peptide, Dnv, demonstrated its superior binding and stability in silico, potent inhibition in reporter cells, and enhanced stability relative to 529-581 in vitro. ARPs were then tested in SclCreER;Ptenfl/f;Ctnnb1Dex/+3(β-catenin) T-ALL mouse model. Linear or circular RNA encoding β-catenin fragments or Dnv was packaged in LNP and intravenously administered. ARP treatments were well tolerated and significantly reduced LSCs, partially restored hematopoietic stem/progenitor cells without obvious cardiotoxicity or hepatotoxicity. Competitive transplantation assays revealed that unlike control, ARP-treated bone marrow cells didn't drive leukemogenesis. ScRNA-seq of bone marrow cells confirmed a significant decrease in LSCs. Moreover, we found a pronounced shift of CD8⁺ T-cells that β-catenin fragments or Dnv treatment decreased stem-like CD8+ T memory cells, which are induced by elevated Wnt/β-catenin signaling. Effector and effector memory cells expanded, and their expressions of anti-tumor cytotoxic genes were elevated, accompanied by downregulation of exhaustion markers. Together, these results establish ARP as a versatile strategy to target intracellular AKT-Wnt/β-catenin signaling. By simultaneously disrupting LSCs' maintenance and reprogramming anti-tumor immunity, ARP offers a promising avenue for overcoming relapse and enhancing therapeutic responsiveness in leukemia.

BloodVol. 146(Supplement 1)
Stowers Institute for Medical Research (US), University of Kansas (US), University of Kansas Medical Center (US), University of Missouri (US)
Good health and well-being
Openalex Percentile: Top 42%
RNA Interference and Gene Delivery
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.