Mapping T-ALL to T Cell Differentiation Trajectory Identifies an ETP-like Subgroup Arrested at an Early Stage of T Cell Development

Objectives: T cell acute lymphoblastic leukemia (T-ALL) is marked by the clonal proliferation of T cells arrested at distinct differentiation stages. The categorization of T-ALL primarily relies on immunophenotype, including early T-cell precursor (ETP)-ALL, near-ETP-ALL and non-ETP-ALL. However, recent studies have highlighted an incomplete alignment between the immunophenotype and the molecular characteristics of ETP-ALL, and indicated that ETP immunophenotype was not a prognostic factor for T-ALL. Therefore, molecular characteristics play a crucial role in the classification of T-ALL, and we aim to develop a score model to facilitate the accurate identification of ETP-ALL/near-ETP-ALL patients based on transcriptome.

Methods: Principal component analysis (PCA) was employed to delineate developmental trajectories of 24 T-ALL patients at distinct differentiation stages. Gene set variation analysis (GSVA) was performed to calculate differentiation scores of T-ALL patients. Subsequently, iterative clustering and guide gene selection (ICGS) clustering analysis was performed using AltAnalyze software on these 24 T-ALL patients. Lineage-correlated genes identified by ICGS were further subjected to key gene selection using the least absolute shrinkage and selection operator (LASSO) regression for model construction. Receiver operating characteristic (ROC) curves were employed for model validation.

Results: The developmental stages of 24 T-ALL patients (6 ETP-ALL, 5 near-ETP-ALL, 13 non-ETP-ALL) were mapped onto the differentiation trajectories of normal thymic development. ETP-ALL and near-ETP-ALL showed similar distribution patterns of hematopoietic stem cell (HSC), lymphoid-primed multipotent progenitors (LMPP), common lymphoid progenitor (CLP), suggesting that these patients arrested at an early thymic developmental stage. GSVA also indicated enrichment of HSC/LMPP features in ETP-ALL and near-ETP-ALL. Correspondingly, the expression of early lymphoid transcription factors such as LYL1, HHEX, and MEF2C showed a gradual decrease from ETP-ALL, near-ETP-ALL, to non-ETP-ALL. Then, we performed unsupervised clustering analysis of the 24 T-ALL patients using ICGS algorithm. ETP-ALL and near-ETP-ALL exhibited comparable transcriptome profiles and were both categorized within the ETP-like group, which exhibited characteristics of early T cell development. To identify ETP-like patients accurately, we utilized LASSO regression on the 406 lineage-specific genes identified by ICGS across the cohort of 24 T-ALL patients, leading to the establishment of an ETP-like score model. The efficacy of the ETP-like score model was assessed in four independent cohorts (TARGET-ALL-P2, GSE141140, GSE146901, OEP002748). ROC curves were generated, yielding AUC values of 0.87, 0.84, 0.925, and 0.8125 for the respective cohorts, which validates the accuracy of our ETP-like score model. Furthermore, we calculated the ETP-like score for 281 T-ALL patients enrolled in the AALL0434 (NCT00408005) clinical trial and observed that patients with high ETP-like score were associated with poor prognosis (P < 0.0001).

Conclusion: ETP-ALL and near-ETP-ALL display analogous transcriptomic characteristics and both exhibit a blockade at the initial stage of T cell development. Therefore, we define a new subgroup of T-ALL with HSC/LMPP features as the ETP-like group and develop an ETP-like score model based on their molecular characteristics. Our ETP-like score model provides a more objective method compared to immunophenotype for identifying ETP-like patients.

Wang:AbbVie: Membership on an entity's Board of Directors or advisory committees.