DNA methylation profiling identifies two epigenetic subgroups with prognostic implications in pediatric acute promyelocytic leukemia: The JCCG study, JPLSG AML-P05

Abstract

Abstract Introduction: Acute promyelocytic leukemia (APL) is a unique subtype of acute myeloid leukemia (AML) characterized by the t(15;17)/PML::RARA fusion.Although all-trans retinoic acid and arsenic trioxide treatments have dramatically improved the prognosis of APL, relapses still occur. Moreover, the molecular biology underlying pediatric APL remains unknown. DNA methylation has been reported as a secondary change in adult APL, despite its involvement in the pathogenesis of the disease, and its prognostic relevance is unclear. In both adult and pediatric AML, DNA methylation patterns have been proposed as potential prognostic biomarkers. This study aimed to investigate the molecular biological basis of pediatric APL and to validate potential prognostic biomarkers. Methods: We retrospectively enrolled 38 pediatric patients with de novo APL who had been treated in the Japanese AML-P05 trial between April 2006 and March 2009. DNA samples from 34 of the patients were subjected to comprehensive DNA methylation analyses. Custom panel sequencing was performed with samples from 35 of the patients. The panel included 343 genes previously reported as drivers or targetable genes in hematological malignancies or solid tumors. Furthermore, we performed RNA sequencing for 34 patients with pediatric APL. Results: In the targeted deep sequencing analysis, we identified 41 mutations in 21 genes of 25 (71%) patients. In two patients, we found mutations of the tumor suppressor gene EP300, which encodes P300 histone acetyl transferase. This has not been previously reported in APL. We analyzed DNA methylation at 825,209 CpG sites in 34 of the pediatric APL patients. We performed unsupervised hierarchical clustering on the 1,000 CpG sites with the highest methylation variance. This profiling identified 48 characteristic CpG sites, which allowed us to classify the cohort into two clusters. Those in cluster 1 (n = 19) had sites characterized by hypo-methylation, while the sites of those in cluster 2 (n = 15) were characterized by hyper-methylation. We found significant differences between clusters 1 and 2 in the DNA methylation beta values of 979 CpGs (false discovery rate [FDR] <0.01, delta beta >0.2). All of these sites were highly methylated in cluster 2. A comparison of outcomes showed significantly worse prognoses in cluster 2 than cluster 1 patients (3-year overall survival 80% vs. 100%, p = 0.043; 3-year event-free survival 73.3% vs. 94.7%, p = 0.049). Four of the patients suffered relapses, all of whom were in cluster 2. Interestingly, the patients with EP300 mutations were both in cluster 2. The 979 highly methylated CpG sites were enriched in enhancer regions (1.99-fold enrichment over all probes, with p <2.2 x 10^(-16) by chi-squared test with Yates' continuity correction). This was indicative of epigenetic dysregulation at P300-regulated enhancers, at least in the two patients with EP300 mutations. The binding motif of EBF1 was substantially enriched in 806 of the 979 (82.3%) CpG sites. Based on the DNA methylation levels at these sites, those in cluster 2 were grouped with 64 other pediatric AML subtypes that we have previously reported. Cluster 1 was isolated. Our transcriptome analysis revealed significant differential expression of 1,432 genes between the two clusters (FDR <0.05). Extremely high MYC expression was observed in cluster 2, with five of the 14 exhibiting expression levels more than double the average of cluster 1. Two of these five patients died. There was also increased expression of ORC2, a MYC target, in cluster 2. Previous research has found worse survival outcomes in non-small cell lung cancer patients with an elevated c-Myc/ORC2 axis. Conclusions: We successfully classified pediatric APL patients into two risk groups based on the DNA methylation levels at certain CpG sites. Our findings suggest an association between APL prognosis and hyper-methylation at EBF1-binding sites. Further studies are needed to elucidate the relationship between EBF1 and c-Myc in APL.