scRNA-seq and CyTOF analyses of PNR and PS. (A) UMAP projection showing cells captured by scRNA-seq from pretreatment patient BMA. Cell type is classified and annotated using SingleR and referenced transcriptomic data sets.⁴⁴ (B) UMAP of bioinformatically isolated B-ALL cells colored per the cluster (1-5; upper) and clinical response to CD19-CAR (PNR and PS; lower). (C) Pie chart depicting relative frequency of PNR or PS leukemic cells per cluster. (D) Heatmap of most significantly differentially expressed genes based on the cluster. Scale depicts normalized expression of each gene per cluster. (E) RNA velocity analysis of scRNA-seq data overlaid on UMAP colored by cluster. Arrows represent high-dimensional vectors predicting the future cell state of cells within the UMAP. (F) Partition-based graph abstraction (PAGA) summarizes the predicted cell trajectories between the clusters. Notably, Cluster 5 is the most primitive cluster, and cells in Cluster 2 are projected to progress to Clusters 1, 3, and 4. (G) t-distributed stochastic neighbor embedding (tSNE) plot of CyTOF analyses of PNR and PS leukemias. Cluster 16 was the only cluster occurring at a significantly different frequency. It is characterized as being CD19⁺, CD22⁺, CD10⁺, CD34⁺, CD33⁺, CD27⁺, and HLA-DR^low. (H) Plot of frequency (%) of cells being clustered into 16 per sample stratified by PNR and PS. Cluster 16 is differentially expressed in PNR leukemias (P = .01; two sample t test). (I) GSEA of cell count normalized scRNA-seq data revealed a significantly decreased expression of antigen presentation and processing pathway genes (NES, −1.59; P = .004). (J) UMAP showing per cell average expression of genes in the antigen presentation and processing pathway demonstrates that the observed decrease in the antigen presentation and processing pathway is independent of cell cluster based on single-cell transcriptomics.