Although genetic basis of pediatric T-cell acute lymphoblastic leukemia (T-ALL) has been well characterized and we reported recurrent SPI1 fusion cases exhibiting uniformly dismal outcome, no comprehensive studies have yet explored the epigenetic profiles and their potential contribution to clinicopathological features of T-ALL.


A total of 79 cases with pediatric T-ALL samples from Tokyo Children's Cancer Study Group (TCCSG) and Japan Association of childhood Leukemia Study (JACLS) including 7 cases of SPI1 fusions were analyzed in this study. To describe epigenetic landscape of T-ALL, we performed methylation array analysis using Illumina HumanMethylation EPIC array, and conducted combined analysis of methylation data with our previous data of whole transcriptome sequencing (WTS) and targeted-capture sequencing for 158 ALL-related genes.


After normalization, 939 probes were selected to identify the most variable methylated probes, when a standard deviation of the beta-values across the samples was above 0.38. Unsupervised consensus clustering clearly identified 4 distinct sample clusters. Combined analyses with WTS data revealed that these 4 clusters were characterized by TAL1 fusions/high expression with intermediate methylation (M1 cluster; n=39), high TLX/HOX expression with intermediate methylation (M2 cluster; n=20), high TLX//HOX expression with high methylation (M3 cluster; n=11), and SPI1 fusions/high expression with low methylation (M4 cluster; n=9), respectively. Interestingly, these methylation clusters were well correlated with genetic features, expression status, and clinical outcomes. In M1 cluster, PTEN abnormalities and UPS7 mutations were especially enriched, and these aberrations were observed mutually exclusive. Furthermore, promoter regions in TAL1 and its target ALDH1A2, important for retinoic acid synthesis, were significantly hypomethylated and highly expressed in this cluster. In M2 cluster, genes related to chromatin remodeling and JAK-STAT were frequently mutated. PRR5L and DEPTOR, both related to AKT1 activation, were differentially methylated and expressed, suggesting an important role of PI3K-AKT1 pathway. M3 cluster was characterized by frequent mutations of epigenetic regulators related genes, such as PHF6, EZH2, and DNM2 . In contrast, mutations in JAK-STAT and PI3K-AKT1 pathways were less frequent. Since the methylation profile of M3 cluster was similar to normal T cells, abnormal DNA methylation status may be not crucial for the leukemogenesis in this group. Intriguingly, all SPI1 fusion cases were classified into M4 cluster, and remaining 2 cases without SPI1 fusions in this cluster showed very high SPI1 expression. Importantly, M4 cluster was characterized by high expression of RASGRP4, a family of Ras guanine nucleotide exchange factors, and low methylation of RASGRP4 promoter. In addition, FGFR1 expression was uniformly repressed and abnormalities of CDKN2A (89%) and KRAS/NRAS (44%) were frequently observed in this group. Furthermore, it should be noted that patients in M4 cluster showed significantly poor prognosis compared to other clusters; eight out of nine cases were died (Log-rank p =4.4×10-7). SPI1 fusion cases exhibited not only unique cytological and gene expression profiles but also distinctive methylation status.


Based on DNA methylation profiles, pediatric T-ALL is clustered into 4 distinct subtypes, which exhibited remarkable correlation with genetic signatures, expression features, and clinical outcomes. PI3K-AKT pathway may be essential in M1/M2 cluster, whereas genetic abnormalities of epigenetic regulators were characteristics in M2/M3 clusters. In addition, the involvement of JAK-STAT pathway would be unique in M2 cluster. Given that cases with M4 cluster showed a catastrophic prognosis and very high expression of RASGRP4 with high frequency of RAS mutation, our findings implicate that targeting therapy of RAS may improve the prognosis of the patients in this group. Although our cohort is limited, our results suggested that the biological phenotype of T-ALL is mediated by both genetic and epigenetic regulations, and explorations for aberrant DNA methylation along with genetic alterations might be helpful for a new therapeutic strategy for T-ALL.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.