Molecular characterization of unresponsiveness to BiTE CD19-CD3 therapy in adult acute lymphoblastic leukemia

Acute lymphoblastic leukemia is a heterogeneous disease characterized by the sequential acquisition of genetic aberrations driving the leukemic clone’s onset and maintenance. The introduction of monoclonal antibodies has both increased over all survival rates and reduced the need of intensive and prolonged chemotherapy in relapsed/refractory (R/R) ALL. Blinatumomab is a BiTE (T-cell engaging bi-specific) antibody that redirects CD3-expressing T-cells to CD19-expressing leukemic cells. To identify predictive biomarkers of response/no-response and mechanisms underlying unresponsiveness to Blinatumomab, 26 B-ALL adult patients both responder and non-responder have been molecularly characterized by gene expression profiling. A bioinformatic analysis was performed employing the linear mixed model (LMM) in order to consider all the possible bias interfering with the results. The LMM output allows to classify training set patients (R or NR); the following LOPO (Leave One (Patient) Out) cross validation have been performed to avoid artifacts, determined by dataset structure. 649 genes pass the LOPO filter and the genes that contribute to less than the 1% to the total variance in the first PCA component have been discarded in order to obtain a small set of significant genes (MS4A1, CSRP2, MY05C, SEMA6A, CD200, CDR1, NEGR1, SCN3A, MME, DNTT, MIR1206). Moreover, the LMM capability of classify patients as responder or non-responder has been confirmed through its output blind application in validation set at baseline; only 1/8 patients is misclassified and additional data are needed to clarify if causes of only one patient misclassification are patient-related or structure dataset-related. Thus, the gene signature composed by 11 genes is capable of classify as responder or non-responder to Blinatumomab treatment adult B-ALL patients at baseline and easy to use for routine diagnostics. However, the dataset increase and deep molecular characterization (e.g. single-cell sequencing) are required to improve statistical significance and define strict associations between genomic characteristics and phenotypic features.