NGS-Based Stratification Refines the Risk Stratification in T-ALL and Identifies a Very High-Risk Subgroup of Patients

Purpose

We previously reported a significantly better outcome in adult and pediatric T-cell acute lymphoblastic leukemia (T-ALL) harboring NOTCH1 and/or FBXW7 (N/F) mutations without alterations of K-N-RAS and PTEN (R/P) genes. High-throughput next-generation sequencing strategies (NGS) allowed us to refine the prediction of outcome in T-ALL.

Patients and Methods

198 adult T-ALLs in first remission (CR1) from the GRAALL-2003/2005 protocol were included in the study as the construction cohort, and 242 pediatric T-ALLs from FRALLE2000 were used as a validation cohort. Targeted whole-exome sequencing of 63 T-ALL-related oncogenes was performed. Primary outcome was cumulative incidence of relapse (CIR). To account for the large number of candidates genes, selection was performed using a LASSO penalization in a Fine and Gray model predicting CIR (Fu Z. et al. Lifetime Data Anal., 2017). To construct the final risk-stratification score, we used non-parametric clustering of CIR curves through k-medians algorithm (Villanueva N. et al. Stat. Med., 2019).

Results

We confirm the prognostic classifier NFRP in the NGS era and evaluate the impact of 39 new gene alterations in the adult cohort. Alterations affecting TP53, DNMT3A, IDH1/2, IKZF1, PI3K pathway (PTEN, PIK3CA, and PIK3R1), EP300, and PHF6 were independent prognostic factors in adult T-ALL. This led us to propose the first NGS-based classifier in T-ALL defining low risk patient (LR) as those with N/F, PHF6, or EP300 mutations without N-K-RAS, PI3K pathway, TP53, DNMT3A, IDH1/2, and IKZF1 alterations (234 of 440 patients, 53%). In the adult cohort, the NGS-based classifier separates a high-risk group (HR) (n=90/198, 45%) with a 5-year CIR of 47% (95%CI:36%-57%) and a low-risk group (LR) (n=108/198, 55%) with a 5-year CIR of 21% (95%CI:14%-29%) (p<0.0001). Our NGS-classifier was validated in the pediatric cohort, with a 5-year CIR of 35% (95%CI:26%-44%) in HR group (n=116/242, 48%) and 5-year CIR of 17% (95%CI:11%-24%) in the LR group (n=126/242, 52%) (p=0.001) (Figure A).

Since the NGS-based classifier is highly prognostic independently of minimal residual disease (MRD) at end of induction (cutoff 10^-4) and white blood cells count (WBC) (cutoff 100 x 10⁹/L), we then developed and externally validated a global risk stratification model incorporating MRD1, WBC at diagnosis and the NGS-classifier. This model identifies 3 subgroups at CR1: a large favorable Risk (CR1-FAV) group (231/332, 70%) with CIR at 5 years estimated at 19% (95%CI:14%-24%) (Figure B), a subgroup of Adverse risk (CR1-ADV) patients (30/332, 9%) with a 5-year CIR of 68% (95%CI:46%-82%) and an Intermediate risk (CR1-INT) group (71/332, 21%) with a 5-year CIR of 37% (95%CI:26%-48%).

Conclusion

T-ALL NGS-based stratification combined with WBC and MRD evaluation sharpens the prognostic classification in T-ALL and identifies a new subgroup of adverse risk patients who should benefit from innovative therapeutic approaches.