Session: 617. Acute Myeloid Leukemias: Biomarkers, Molecular Markers and Minimal Residual Disease in Diagnosis and Prognosis: Poster II
Hematology Disease Topics & Pathways:
Research, Acute Myeloid Malignancies, AML, Translational Research, genomics, Diseases, Myeloid Malignancies, Biological Processes
We initially performed unbiased hierarchical clustering on CpG sites that are differentially methylated between AML and normal bone marrow to identify distinct AML-restricted methylation (ARM) signatures. This analysis demonstrated that that these ARM clusters were enriched in specific cytomolecular subtypes of pediatric AML. Specifically, we identified methylation signatures that recapitulated their adult counterparts (e.g., IDH1/2, DNMT3A, WT1) and pediatric specific methylation signatures that clustered with RUNX1::RUNX1T1, CBFB::MYH11, KMT2A-rearrangement, NUP98::NSD1, NUP98::KDM5A, and CEBPA mutant AML. Next, we investigated the transcriptional and methylation signatures for the relapsed cases and compared them to diagnosis. We demonstrated that relapsed transcriptional and methylation signatures clustered within their cytomolecular subtype counterparts at diagnosis (i.e., AMLs at relapse more closely resemble AMLs at diagnosis with the same underlying gene fusion than other relapsed AMLs) (Fig. 1A). However, analyzing our data in the context of these cytomolecular subtypes, we found that leukemias uniformly acquired a more stem-like state at relapse.
We focused our biomarker discovery efforts on KMT2A-rearranged AMLs since (as a cohort) these patients have an intermediate risk of relapse and existing molecular biomarkers (e.g., KMT2A fusion partner and measurable residual disease detection) result in only modest improvements in relapse prediction. First, we found that CpG islands associated with CD34 and HOXA gene expression were frequently (but variably) hypermethylated, which was also associated with downregulation of CD34 and HOXA gene expression and favorable survival outcomes. Next, we performed non-negative matrix factorization on diagnosis and relapsed KMT2A-rearranged AMLs and identified a novel gene set associated with relapse and poor outcomes. These three distinct methylation-transcriptional findings (i.e., CD34, HOXA, and relapsed signatures) were subsequently transformed into a novel methylation-transcription biomarker score (MTBS) using machine learning algorithms (i.e., LASSO Cox regression). We used KMT2A-rearranged AML patients from AAML1031 as our training cohort (N = 129) to generate a MTBS that was able to distinguish KMT2A-rearranged cases with disparate outcomes. This algorithm was validated in an independent cohort from AAML0531 (N = 28) (Fig. 1B). Additional methylation-transcriptional biomarker discovery efforts for pediatric AML are underway.
Since pediatric AML is associated with a low mutational burden, understanding leukemia-specific transcriptional and epigenetic alterations is of utmost importance. Our study demonstrates the translational promise offered by deep functional genomic characterization of pediatric AML and we believe that further interrogation may provide additional novel insights into leukemia initiation, relapse prediction, and novel treatment approaches.
Disclosures: Farrar: Novartis: Research Funding.