Identifying Genetic and Epigenetic Changes Underlying Acute Myeloid Leukemia Progression

Acute myeloid leukemia (AML) is characterized by impaired differentiation and aberrant proliferation of hematopoietic stem and progenitor cells (HSPCs), preventing normal bone marrow hematopoiesis. Significant progress has been made in improving the overall survival rate in AML; however, preventing and managing relapsed cases pose substantial challenges in AML treatment regimens. Therefore, we aim to evaluate the prognostic importance of somatic mutations in de novo AML, predict the likelihood of relapse in diagnostic AML patients, and characterize genetic and epigenetic changes throughout AML progression.

Herein, we conducted a systematic review and meta-analysis focusing on de novo AML to assess the prognostic impact of somatic mutations on overall survival (OS) and relapse-free survival (RFS). Random effects model and Cochrane standard statistical test were employed for statistical analyses and heterogeneity among primary studies, respectively. We found that CEBPA and NPM1 mutations have favorable outcome on OS and RFS, whereas FLT3-ITD and DNMT3A mutations were associated with adverse outcomes on OS and RFS, underscoring their detrimental effects in AML patients. Additionally, subgroup analysis of cKIT mutations revealed favorable outcomes on RFS in children and adverse outcomes on RFS in adults.

Next, we have characterized the open chromatin landscape of AML patients at diagnosis using ATAC-seq. We found that leveraging the presence or absence of active regulatory regions in a LightGBM model could moderately predict relapse; however, it did not explain the molecular mechanisms associated with relapse. Therefore, we assessed whether the transcription factor binding site (TFBS) repertoire of open chromatin regions could predict the outcome. The model achieved an AUC of ~0.80 (70–73% accuracy). Additionally, the lowest percentage of predicted relapse peaks among relapse samples was 55%, whereas the highest percentage of predicted relapse peaks among remission samples was 40%. Therefore, relapse samples could be predicted with 100% accuracy based on the proportion of predicted relapse peaks. Upon interrogating the features for each correctly predicted region, we identified multiple networks of TFBSs that may be significant predictors of pediatric AML relapse.

Finally, we characterized genetic and epigenetic changes in AML progression using multi-omics approaches to elucidate the underlying mechanisms of relapse. Differential interaction analysis showed significant 3D chromatin landscape reorganization between relapse and diagnosis samples. Comparing global open chromatin profiles revealed that relapse samples had significantly fewer accessible chromatin regions than diagnosis samples. In addition, we discovered that relapse related upregulation was achieved either by forming new active enhancer contacts or by losing interactions with poised enhancers/potential silencers.

In conclusion, our investigation examined the landscape of AML progression and relapse. Through systematic review and meta-analysis, machine learning, and multi-omics characterization, we have identified prognostic and predictive markers and molecular mechanisms underlying AML relapse. These findings provide insights for refining treatment strategies and advancing personalized therapeutic approaches in hematological oncology.