Paper Details

ABSTRACT

Background: Relapsed T-cell acute lymphoblastic leukemia (T-ALL) has a poor prognosis, with survival <30%. Identifying novel therapies to prevent and salvage relapse is critical. Target selection and prioritization is complicated by somatic tumor biologic heterogeneity and a paucity of targets aberrantly expressed or dysregulated in T-ALL blasts and not healthy cells. Much of T-ALL profiling has been RNA-based, which can be misleading for target evaluation due to disconnect between RNA-protein levels. Defining RNA–protein concordance is essential to determine which targets can be inferred from transcriptomic data and which require protein-level assessment for biomarker-driven precision medicine.

Methods: We previously performed comprehensive genomic profiling of 1,309 pediatric and young adult T-ALL cases treated on the Children’s Oncology Group AALL0434 international phase 3 clinical trial, defining 15 major molecular subtypes with distinct genomic profiles (PMID: 39143224). To understand RNA-protein relationships, which targetable pathways are dysregulated in different genomic subtypes, and identify novel targets for immunotherapy, we performed mass spectrometry (MS) proteomics on samples from 114 genomically characterized AALL0434 T-ALL patients that included global proteomics, glycosylation, phosphorylation, and ubiquination. Global proteomics data was generated using data-independent acquisition (DIA) and tandem-mass tag (TMT) MS methods. We also generated DIA MS data from pediatric healthy thymi (n=6) and bone marrow (n=3).

Results: We quantified ~10,000 proteins by DIA and ~11,000 by TMT, of which 8,279 and 10,491, respectively, were detected in at least 50% of samples. Median gene-wise RNA–protein correlation was relatively low (DIA, ρ=0.16; TMT, ρ=0.20), demonstrating RNA expression is not uniformly predictive of protein abundance in T-ALL. Nevertheless, 687 DIA-quantified and 1,023 TMT-quantified proteins were at least moderately correlated (ρ>=0.5). Membrane associated proteins represented 26% and 23% of these proteins in DIA and TMT, respectively. Clinically actionable targets with strong RNA–protein concordance (TMT and DIA ρ≥0.5; P<0.001) included LCK, BCL2, CD7, CD38, CD5, and PLK1 supporting the use of RNA-based assays as surrogates for protein abundance for these targets. A small number of proteins (n=16) including GTF2H3, previously linked to T-ALL, had negative RNA-protein correlation (TMT and DIA); mechanistic studies are underway to understand the negative correlation.

RNA expression did not consistently capture genomic alteration-associated differences in RUNX1, PTEN, NOTCH1, and CCND3. Blasts with NOTCH1 and CCND3 activating alterations had significantly higher protein abundance than unaltered cases (P<0.001), despite no corresponding differences in RNA, highlighting the importance of proteomic evaluation. Proteomic profiling identified maturation-associated targetable dependencies. BCL2 protein abundance, a biomarker of venetoclax sensitivity, was enriched in developmentally immature T-ALL (ETP-like and BCL11B) compared to more mature double-positive T-ALL (TLX1, NKX2-1, and TAL1 DP and αβ-like). In contrast, LCK protein abundance, a biomarker of dasatinib sensitivity, was enriched in the double-positive T-ALL group compared to immature T-ALL (Wilcoxon BH-adjusted P<0.001 for both).

We used ImmunoTar (PMID: 39932005) to identify novel potential targets for immunotherapy that were abundant on T-ALL blasts but had low to no protein abundance in healthy hematopoietic cells. These targets include LILRB2 (CD85d), a member of the leukocyte immunoglobulin-like receptor (LIR), and CD96 (TACTILE), a type I transmembrane glycoprotein belonging to the immunoglobulin superfamily. Flow cytometry validation is ongoing.