A myeloid transcriptional signature stratifies outcomes in newly diagnosed ETP-ALL

Abstract

Abstract Background T-cell acute lymphoblastic leukemia (T-ALL) accounts for ~25% of adult ALL and is classified by immunophenotypic (IP) markers into cortical, mature, and early subtypes—the latter including early T-cell precursor (ETP) and near-ETP. ETP-ALL, originally defined by gene expression profiling, is genetically heterogeneous and thought to arise from leukemia-initiating cells with both lymphoid and myeloid differentiation potential. It is associated with poor outcomes due to resistance and relapse. Though clinically practical, IP-based classification often poorly reflects disease biology. We aimed to identify prognostic expression signatures within T-ALL to better understand heterogeneity and refine treatment strategies. Methods We analyzed 59 T-ALL patients, all treated with ALL-directed regimens. RNA was collected from bone marrow samples before therapy and sequenced using Illumina platforms. Raw counts were analyzed with DESeq2; molecular profiles were evaluated by gene set enrichment analysis (GSEA). Patients were clustered by expression signatures, and associations with event-free (EFS) and overall survival (OS) were assessed. Thymic and mature (T/M) subtypes were analyzed together. Results Samples from 59 patients were collected between 2000-2018: 26 newly diagnosed (ND) and 33 relapsed/refractory (R/R). The median age was 36 years (range, 18–77). IP subtypes included ETP (n=18, 31%), near-ETP (n=11, 19%), thymic (n=27, 46%), and mature (n=3, 5%). Karyotypes included diploid (n=28, 47%), hyperdiploid (n=4, 7%), hypodiploid (n=2, 3%), complex (n=4, 7%), near-tetraploid (n=2, 3%), and other (n=19, 32%). Most ND cases (81%) received Hyper-CVAD based regimens; most R/R (79%) received other anthracycline-based or investigational regimens. Transplantation rates were similar in ND and R/R groups (19%, 11/59 overall). With a median follow-up of 84 months, median EFS in ND patients was 20 months (ETP, n=9), 36 (near-ETP, n=5), and 44 (T/M, n=12) (p=NS); median OS was not reached. In the R/R cohort, median EFS was 1.9 (ETP, n=9), 4.8 (near-ETP, n=6), and 10.8 months (T/M, n=18) (p=0.02). Median OS was similarly poor across R/R subtypes: 14.5 months (ETP/near-ETP) and 11.1 (T/M) (p=NS). Unsupervised clustering of the 50 most variable differentially expressed genes identified four expression-based clusters. By IP, Clusters A and B contained both ETP/near-ETP and T/M cases, Cluster C was enriched for T/M, and Cluster D was enriched for ETP/near-ETP. Notably, when comparing ETP/near-ETP cases between Clusters A and D, striking differences emerged in myeloid gene expression: Cluster A showed high expression of a 15-gene myeloid differentiation signature (including MPO, MMP8, DEFA3; designated “Myeloid-up”), while Cluster D showed marked repression of the same genes (designated “Myeloid-down”). Expression clusters were prognostic only among ND ETP and near-ETP cases, not among T/M or R/R subgroups. ND ETP/near-ETP Myeloid-up patients (n=8) had superior outcomes vs Myeloid-down patients (n=6) with median EFS 153.4 vs 8.8 months (p=0.09) and median OS not reached vs 16.2 months (p=0.007). Myeloid-up samples were more frequently diploid (5/6) than Myeloid-down (0/8). ORR did not differ between Myeloid-up and Myeloid-down patients (75% vs 66%, p=NS). MRD and somatic mutation data were available in a limited subset and thus not analyzed. Flow cytometry markers did not correlate with cluster assignment. To characterize transcriptional differences between ND ETP-ALL cases in Myeloid-down and Myeloid-up clusters, we analyzed gene expression profiles from three samples per group. A total of 654 genes were differentially expressed, of which 506 were significantly downregulated in Myeloid-down relative to Myeloid-up cluster. GSEA revealed that these repressed genes were enriched for transcriptional programs silenced in hematopoietic and leukemic stem cells (NES 1.5, padj<0.001 for both) and activated in mature myeloid cells (NES 1.4, padj<0.001), indicating a stem-like transcriptional phenotype and restricted myeloid differentiation in Myeloid-down cluster. Conclusions ETP-ALL is transcriptionally heterogeneous. In ND ETP and near-ETP cases, a Myeloid-up transcriptional signature was associated with improved outcomes, while a stemness-associated Myeloid-down profile predicted poor survival—despite similar IP. These findings may improve risk stratification and guide consolidation and therapeutic decisions in ETP-ALL.