Predicting Response of CD19 Chimeric Antigen Receptor (CAR) T-Cell Therapy for Acute Lymphoblastic Leukemia (ALL) and Aggressive Non-Hodgkin Lymphoma (NHL) in ACIT001/EXC002, a Phase Ib Clinical Trial

Chimeric antigen receptor (CAR) T-cell therapy targeting CD19 is next generation immunotherapy that has curative potential in relapsed ALL and aggressive NHL. Though effective, about half of the patients do not respond to the therapy. In addition, the CAR T production process is difficult, time consuming, expensive and these costs are subsequently passed down to health care systems. Therefore, predicting the clinical efficacy and resistance to the CAR T cell therapy before going through all CAR T cell production process to rule out the non-responders is needed to save both time and cost. Herein we describe a holistic, correlative approach to identifying responders and non-responders on the phase 1b portion of ACIT001/EXC002, a clinical trial of decentralized production of anti-CD19/41BB/CD3z in multiply relapsed/refractory NHL and acute lymphoblastic leukemia (ALL) patients. The outcomes were correlated with the clinical response. The median transduction efficiency was 27% (range: 7.35% to 54.57%). Among 10 patients enrolled in phase I clinical trial, 3 patients never responded (2 primary refractory and 1 unevaluable), and 1 showed partial response. All 10 products were tested for cytotoxicity ex vivo in increasing effector to target ratios against a CD19+ cell line (Raji) after 4 and 24 hours in culture. Except one (A02), all of the complete responders exhibited effective CD19 positive cells killing while 2 out of 3 non-responders exhibited poor response in ex vivo cytotoxic assay. Although one of the non-responders expressed high level of PD1, expression was not statistically different between responders and non-responders. Flow cytometry analysis of CAR T products demonstrated that naïve T cell subset was negligible in all samples. Central memory subset was low in all CAR T samples except CAR T from one patient (A06). In general, proportion of effector memory subset was higher than other T cell subsets in CAR T from all patients and highest proportion was observed in the one of the CAR T from non-response group (A03). Gene expression analysis using nCounter CAR T characterization panel from Nanostring revealed upregulation and downregulation of gene sets in CAR T from non-response group as compared to the CAR T from complete response. Glucose-6-phosphate isomerase (GPI), Aldolase (ALDOC), Myc were the top downregulated genes and interferon-gamma-inducible-protein 30 (IFI30), chemokine receptor CX3CR1, FCGR3A/B were the top upregulated gene in CAR T from non-response group. Advanced analysis showed a trend in higher pathway score for apoptosis and exhaustion pathway and a trend in lower score for activation, glycolysis and JAK-STAT pathway in CAR T from non-response as compared to the complete response group. Since there was a substantial cross patient heterogeneity in gene expression of CAR T, more samples in each group are needed to validate significance of differences in gene expression. Interestingly, expression of top genes in CAR T from one the non-response group (A07) was different than other group although it showed good response in ex vivo cytotoxicity. While the three individual assays by themselves would not be sufficient to predict for response, taken together, the multiple criteria from ex vivo cytotoxicity assay, T cell phenotypic analysis and gene expression analysis is capable of identifying response to this second generation CAR T product. This study will continue to analyze the CAR T from upcoming phase II clinical trial and perform robust corelative analysis to predict the clinical outcomes based on multiple characteristics of CAR T products while striving to identify statistical weight behind these assays to develop a more reliable prediction model through use of machine learning.