Prediction of Patients at Risk of CD19^Neg Relapse Following CD19-Directed CAR T Cell Therapy in B Cell Precursor Acute Lymphoblastic Leukemia

CD19-directed chimeric antigen receptor T cell (CAR-T) therapy has shown impressive results in children and adults with relapsed or refractory B-ALL or diffuse large B-cell lymphoma. However, 30 - 70% of initial responders will eventually relapse with CD19 antigen loss (CD19^Neg) (Maude SL, et al. N Engl J Med. 2018). To avoid CD19^Neg relapse, patients may undergo a hematopoietic stem cell transplant (HSCT). HSCT is an expensive and often morbid procedure that many physicians would prefer to avoid. The development of tools to accurately predict which patients are at risk for CD19^Neg relapse would guide treatment decisions regarding HSCT or alternative therapies. Since CD19^Neg relapses also occur in patients treated with other CD19-directed immunotherapies, like blinatumomab (Mejstríková E, et al. Blood Cancer J. 2017), a predictive model to detect patients at risk of CD19^Neg relapse would have broader therapeutic impact.

To address this problem, we performed CyTOF and RNA-seq analysis from paired patient samples collected before CD19-directed CAR-T administration and after CD19^Neg relapse. High dimensional phenotyping by CyTOF clustered patient samples based on their mechanism of CD19 expression loss (frameshift mutation versus expression of intracellular isoforms), even before CAR-T administration. In addition, we identified identical immunoglobulin heavy and light chain RNA sequences before CAR-T administration and after CD19^Neg relapse, suggesting that the clones destined to cause relapse are present at the time of CAR-T administration. Altogether, these results support our hypothesis that resistant tumor cells are present before CAR-T administration and could be discovered and interrogated for CD19^Neg relapse prediction.

To identify cell subpopulations responsible for driving CD19^Neg relapse, we used the B cell developmental classifier previously developed in our lab (Good Z, et al. Nat Med. 2018). We observed a significant increase in the Early-non-BI population (CD38^Pos CD24^Pos CD19^Neg CD20^Neg CD3^Neg CD16^Neg CD61^Neg cells) after CD19^Neg relapse, suggesting that CD19 loss is associated with the loss of other B cell features. Since our classifier relies on CD19 to classify cells, we compared the resulting classification of cells when CD19 was included or excluded in the classifier. This change had minimal impact in cell classification from healthy bone marrow controls. However, when applied to the samples collected before CAR-T administration, we found a subpopulation of CD19^Pos Pro-B cells that classified as Early-non-BI cells when CD19 was excluded from the classification. We hypothesize that these Pro-B "discordant" cells are those that lose CD19 expression to escape the immune pressure exerted by the CD19-directed CAR-T and mediate CD19^Neg relapse. Further, we found Pro-B "discordant" cells in 77% of independent cohort of 22 B-ALL samples collected at the time of diagnosis, suggesting these cells exist in de novo B-ALL. We likewise identified a CD19^Neg IgM^Pos Early-non-BI subpopulation in 4 healthy bone marrow and further studies are ongoing to characterize these cells. We continue to interrogate this candidate population as that responsible for CD19^Neg relapse after CAR-T cell therapy.

In addition, we performed differential expression analysis between paired samples collected before (CD19^Pos) and after (CD19^Neg) CAR-T therapy. Through the application of the developmental classifier, we identified that CD19 loss is associated with upregulation of key B cell transcription factors IKAROS, PAX5 and glucocorticoid receptor in the pre-pro-B to Pre-B stages. Moreover, after CD19 loss, there are also increases in levels of phosphorylated proteins pSYK, pSRC and pSTAT5, involved in IL7 receptor and pre-BCR signaling pathways, essential for B cell development. This suggests that CD19^Neg cells activate unique tumorigenic pathways that may provide novel therapeutic opportunities. Exploration and validation of these therapeutic targets could significantly improve clinical outcome and care of patients with CD19^Neg B-ALL.

In conclusion, these results support the feasibility to predict patients at risk for CD19^Neg relapse together with the mechanism behind it. Future studies will be conducted to confirm unique tumorigenic pathways in CD19^Neg B cells and determine their therapeutic potential.