Multiomic signatures of durable response in a Phase I/II trial of tandem aCD22-aCD19 CAR-T therapy for relapsed B-lineage acute lymphoblastic leukaemia

Abstract Background: CAR-T therapeutic failures are multifactorial and include a lack of cell proliferation post-infusion, reduced effector differentiation, and inadequate persistence to provide longer-term immune-surveillance to prevent B-ALL relapse. We conducted a phase I/II trial of the dual aCD22-aCD19 CAR-T therapy (MB-CART2219.1) for adults and children with relapsed/refractory B-lineage acute lymphoblastic leukemia across three tertiary insitutions. We treated 11 R/R B-ALL patients on clinical trial and a further 3 patients per protocol, with a 91% MRD-negative complete remission rate. The 12-month overall survival (OS) and leukaemia-free survival (LFS) probability was 82% +/- 12% and 64% +/- 15%, respectively, without transplant consolidation, improving upon LFS of 50% with aCD19 CAR-T therapies. Median overall survival and LFS was not reached at 24 months. Aim: We aimed to define functional correlates of durable response through integrated multi-omic analysis of MB-CART2219.1 products and post-infusion profiles. Methods: Selected CAR-T cell products and post-infusion PBMCs were analyzed using 27-colour spectral flow cytometry on a Cytek Aurora (Cytek Biosciences, Fremont, CA) and single-cell RNA sequencing with the Chromium Controller (10X Genomics, Pleasanton, CA). We compared immune signatures and gene pathways enriched in CAR+ T cells between responders and non-responders. Responses were evaluated up to 12 months post-infusion for inclusion in the multi-omic analysis. Statistical Analysis Flow cytometry data from 11 patients (7 responders, 4 non-responders), including pre-infusion CAR-T products and early and late post-infusion blood samples, were analysed using OMIQ (OMIQ.ai) for dimensionality reduction (t-SNE, UMAP), clustering, and heatmap visualization. Manual gating was performed using FlowJo v10 (FlowJo LLC, Ashland, OR). Single-cell RNA-seq data from 9 patients (6 responders, 3 non-responders) were analysed to characterize rare CAR⁺ T cell subsets pre- and post-infusion. Reads were aligned to a custom reference (GRCh38 + CAR transgene) using Cell Ranger v7.0.1 (10X Genomics). Cells with <1,000 genes, high UMI counts, or >20% mitochondrial reads were excluded. Expression data were normalized and integrated using Seurat v4.3, with batch correction via canonical correlation analysis and PCA. UMAP was used for dimensionality reduction, and clusters were annotated using curated T cell subset and functional gene signatures (e.g., T_SCM, T_CM, T_EM, T_EFF; activation, exhaustion, cytotoxicity, proliferation). DEGs between responders and non-responders were identified using Wilcoxon rank-sum tests with Bonferroni or FDR correction. Pathway enrichment was performed with Enrichr. TCR clonotypes were called with Cell Ranger V(D)J, mapped to expression clusters, and categorized by size (singleton, small, expanded). Clonality distributions were compared across outcome groups. Concordance between scRNA-seq and flow cytometry was assessed for matched markers across timepoints. Results: Clinical response was associated favourably with CAR-T product effector memory signatures (CD62L-, TIM3-, low Treg) and a cytotoxic transcriptome (GZMB, IFNG, TNFSF10). Post-infusion, CD8+ CAR+ activation (41BB+), TEM and TCM CAR⁺ T cell expansion, and the acquisition of proliferative memory (CD127) at day 28, were all associated with durable response. Post-infusion trajectories of CAR-T cells in responders showed large clusters of cytotoxic or proliferative CD8⁺ clonotypes emerging from Day 10 through Day 28. In contrast, non-responders had CAR-T products skewed toward naïve (TSCM, CD62L+, CD127⁺) and T regulatory (Treg) phenotypes. These patients exhibited early upregulation of exhaustion markers on CAR⁺ T cells (PD1, TIM3, TIGIT, LAG3), and on CAR⁻ immune subsets (TIGIT⁺, LAG3⁺, CD38⁺). Notably, there were also differences within the non-CAR immune cells at late timepoints after infusion. Tregs were elevated in non-responders, while responders showed enrichment of CD39⁺ NKT cells, suggesting a role of bystander immune activation in long-term response. Conclusion: Tandem aCD22-aCD19 CAR-T (CART2219.1) as a stand-alone infusion in relapsed/refractory B-ALL is promising, with a median overall survival and LFS not reached at 24 months. Multiomic analysis of product and post-infusion signatures highlight effector functionality at infusion and memory transition post-infusion as key determinants of long-term CAR-T efficacy.