T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy that is classified according to surface marker expression. In order to reveal the cells of origin in pediatric T-ALL and to understand mechanisms of relapse we used ATAC-Seq (Assay for Transposase-Accessible Chromatin Sequencing) to compare chromatin accessibility landscapes of healthy T-cell precursors to those of T-ALL cells obtained at initial diagnosis (INI) and relapse (REL).We have FACS sorted 7 differentiation stages of normal T-cell precursors contained in the thymus of infants undergoing cardiac surgery (DN2, DN3, ISP, DPCD3-, DPCD3+, CD4+ and CD8+) and subjected these to ATAC-Seq. Unsupervised learning by principal component analysis (PCA) clustered sorted populations according to the maturation stage, demonstrating that regulatory chromatin signatures of thymocytes are highly stage-specific and re-shaped during T-cell differentiation.We next compared normal T-cell precursors at different stages of maturation to pediatric T-ALLs and found fundamental differences with 30% of open chromatin regions to be more and 28% being less accessible in T-ALL (DESeq2, padj<.05), respectively. We next identified transcription factor (TF) binding sites in open chromatin regions of normal T-cell precursors and T-ALL cells by HOMER analysis. We found 76% of the accessible TF binding sites in double negative (DN2 and DN3) normal precursors to be also accessible in most (20/24) leukemias comparing to 17% of the sites accessible in the more mature (CD4+ and CD8+) stages. These data indicate that T-ALLs originate from cells with an epigenetic profile of early thymic progenitors.We then subjected the ATAC-seq data of all matched leukemia samples obtained at initial disease and at relapse to PCA. INI and REL samples derived from the same patient always clustered in close proximity and were separated according to the T-ALL driving fusion genes. A global analysis of differential accessibility revealed only 0.26% of ATAC-regions to be less- or more-accessible at relapse when compared to the matched initial samples (DESeq2, padj<.05). These data indicate chromatin accessibility to be largely determined by the cell of origin and to generally remain stable during progression from initial diagnosis to relapse. We then considered the 2 types of relapse separately, which we have previously characterized on the basis of subclonal mutation profiles (Kunz et al. Haematologica, 2015). These relapse types are defined by either clonal evolution of cells derived from the predominant clone at primary disease (type 1) or emergence and evolution of a minor initial clone showing a molecular profile that differs from the major initial clone (type 2). We found that the proportion of differentially accessible ATAC-regions between INI and REL is significantly higher in type 2 (1.3%) than in type 1 (0.006%) (DESeq2, padj<.05).Moreover, we trained the deconvolution algorithm CIBERSORT to recognize particular T-cell differentiation stages using ATAC-profiles of the 7 FACS-sorted healthy T-cell populations. We used regulatory chromatin landscape of non-sorted (total) thymus to assess the accuracy of deconvolution. Comparison of predicted fractions in total thymus to FACS measurements revealed highly accurate identification of the maturation stages (r2 = 0.95). CIBERSORT analysis confirmed that the profiles were largely preserved between INI and REL of each sample pair. Notably, however, while in T-ALLs that later developed into a type 1 relapse only one type of early T-cell progenitor dominated the deconvolution profile, T-ALLs that developed into a type 2 relapse showed heterogeneous profiles with contributions of progenitors at different maturation stages.In sum, these epigenomic analyses revealed that the chromatin landscape of normal T-cell precursors evolves in the course of thymic maturation and that the early maturation stages are the likely origin of T-ALL cells. Remarkably, pediatric T-ALLs that later develop a type 2 relapse consist of subclones with a variable profile of chromatin accessibility that define different stages of maturation. These data indicate that T-ALLs with the propensity to develop a type 2 relapse differ from type 1 in that they originate from early precursors that carry the potential of further development into different stages of maturation before the leukemia becomes apparent with a highly subclonal pattern. DisclosuresMuckenthaler:Novartis: Research Funding. Bourquin:Amgen: Other: Travel Support. Kulozik:bluebird bio: Consultancy, Honoraria.