Single-cell multiomics dissection of basal and antigen-specific activation states of CD19-targeted CAR T cells

Abstract

Background: Autologous T cells engineered to express a chimeric antigen receptor (CAR) specific for CD19 molecule have transformed the therapeutic landscape in patients with highly refractory leukemia and lymphoma, and the use of donor-generated allogeneic CAR T is paving the way for further breakthroughs in the treatment of cancer. However, it remains unknown how the intrinsic heterogeneities of these engineered cells mediate therapeutic efficacy and whether allogeneic products match the effectiveness of autologous therapies.

Methods: Using single-cell mRNA sequencing in conjunction with CITE-seq, we performed multiomics characterization of CAR T cells generated from healthy donor and patients with acute lymphoblastic leukemia. CAR T cells used in this study were manufactured at the University of Pennsylvania through lentiviral transduction with a CD19-4-1BB-CD3ζ construct. Besides the baseline condition, we engineered NIH-3T3 cells with human CD19 or mesothelin expression to conduct ex vivo antigen-specific or non-antigen stimulation of CAR T cells through 6-hour coculture at a 1:1 ratio.

Results: We delineated the global cellular and molecular CAR T landscape and identified that transcriptional CAR tonic signaling was regulated by a mixture of early activation, exhaustion signatures, and cytotoxic activities. On CD19 stimulation, we illuminated the disparities of CAR T cells derived from different origins and found that donor CAR T had more pronounced activation level in correlation with the upregulation of major histocompatibility complex class II genes compared with patient CAR T cells. This finding was independently validated in additional datasets from literature. Furthermore, GM-CSF(CSF2) expression was found to be associated with functional gene productions, but it induced little impact on the CAR T activation.

Conclusions: Through integrated multiomics profiling and unbiased canonical pathway analyses, our results unveil heterogeneities in the transcriptional, phenotypic, functional, and metabolic profiles of donor and patient CAR T cells, providing mechanistic basis for ameliorating clinical outcomes and developing next-generation 'off- the-shelf' allogeneic products.

Keywords: chimeric antigen; immunotherapy; receptors; translational medical research.

Figure 1

A single-cell multiomics landscape of CD19-targeted CAR T cells generated from healthy donor and ALL patients. (A) Schematic of experimental design. (B) UMAP plot of pooled 23,349 single CAR T cells collected from all conditions of three samples. (C) Heatmap of differentially expressed genes of each identified cluster. Highlighted genes were used to annotate each cluster. (D) The distribution of surface protein ADT-CD4 and ADT-CD8 expression revealed by CITE-seq. Color bar represents normalized expression level. (E) Fraction of cells from each sample in each cluster. $x$ axis indicates sample resources and stimulation condition. ALL, acute lymphoblastic leukemia; CAR, chimeric antigen receptor; CART, without any stimulation; CD19, cocultured with 3T3-CD19; CR, patient with complete remission; HD, healthy donor; MESO, cocultured with 3T3-MESO (mesothelin); NR, non-responder.

Figure 2

Transcriptional and pathway signatures of basal CAR T cells indicate tonic signaling is regulated by a mixture of early activation, exhaustion, and cytotoxic activities. (A) UMAP clustering of basal single-cell profiles. (B) Canonical pathways identified in each cluster and their comparisons. $z$ score reflects the predicted activation level ($z\S amp;lt;0$, inhibited; $z\S amp;gt;0,$ activated; $z\ge 2$ or $z\le -2$ can be considered significant). (C) Fraction of cells from each sample in each cluster, separately calculated for CAR+ and CAR− cells. (D) The distribution of early activation related genes identified in cluster 7. (E) Expression level comparison of evaluated genes in (D) between CAR+ and CAR− cells. (F) The distribution of exhaustion-related genes identified in cluster 5. (G) Expression level comparison of evaluated genes (F) across all the single cells between CAR+ and CAR− cells. Color bar represents normalized expression level. Bar plots show mean±SEM. The p values were calculated with two-tailed Mann-Whitney test. **P<0.01, ***p<0.0001. ns, not significant. CAR, chimeric antigen receptor; UMAP, uniform manifold approximation and projection.

Figure 3

Phenotypic, functional, and metabolic states of basal CAR T cells. (A) Composition of T cell differentiation states of each sample determined by ADT-CD62L, ADT-CCR7, ADT-CD45RA, and ADT-CD45RO expression.(B) The average expression heatmap of the four ADT markers in each T cell state. (C) Comparison of early memory related ADT-CD62L and ADT-CCR7 expression between different samples. The p values were calculated with two-tailed Mann-Whitney test. (D) Volcano plot of DEGs upregulated in HD versus CR sample. (E) Corresponding signaling pathways regulated by DEGs identified in (D). (F) Volcano plot of DEGs upregulated in HD versus NR sample. (G) Corresponding signaling pathways regulated by DEGs identified in (F). $z$ score reflects the predicted activation level (${\displaystyle z0}$, inhibited; ${\displaystyle z>0,}$ activated; $z\ge 2$ or $z\le -2$ can be considered significant). (H–K) Comparison of oxidative phosphorylation (H), glycolysis (I), fatty acid oxidation (J) and apoptosis (K) correlative gene sets between HD, CR, and NR group. Literature-based gene sets were listed at left side. Scatter plot with bar shows mean±SEM. Symbols represent average expression value of each gene across all single cells. The p values were calculated with Wilcoxon signed-rank test. *P<0.05, **p<0.01, ***p<0.0001. ADT, antibody-derived tag; CAR, chimeric antigen receptor; CR, complete remission; DEGs, differentially expressed genes; HD, healthy donor; NR, non-responder; ns, not significant; T_N, naïve T cell; T_SCM, stem cell-like memory T cell; T_CM, central memory T cell; T_EM, effector memory T cell; T_EF, effector T cell.

Figure 4

CD19-specific stimulation of CAR T cells unveils their inherent characteristics and identifies signatures specific to HD, CR, or NR group. (A) UMAP clustering of 9744 CD19-3T3 stimulated CAR T cells identified 10 transcriptional clusters. (B) Distribution of CAR gene expression. (C) Single-cell expression violin plot of key immunologically relevant cytokine genes separated by CAR+ and CAR− as well as resource groups. The values listed at left represent normalized expression level of each cytokine. (D) UMAP reclustering of only activated CAR+ cells. (E) Volcano plot of DEGs when comparing activated HD with CR sample. (F) Signaling pathways regulated by DEGs identified in (E). $z$ score reflects the predicted activation level ($z\S amp;lt;0$, inhibited; $z\S amp;gt;0,$ activated; $z\ge 2$ or $z\le -2$ can be considered significant). (G) Comparison of activation related surface protein expressions between different groups. (H) The proportion of each T cell differentiation state in each group. CAR, chimeric antigen receptor; CR, complete remission; HD, healthy donor; NR, non-responder; T_N, naïve T cell; T_SCM, stem cell-like memory T cell; T_CM, central memory T cell; T_EM, effector memory T cell; T_EF, effector T cell; UMAP, uniform manifold approximation and projection.

Figure 5

CSF2 expression is correlative with functional genes production but induces no impact on the CAR T activation level. (A) Comparison of CSF2 expression in HD, CR, and NR group. (B) Comparison of CSF2 expression in CD4+ and CD8+ cells. Box middle line, median; whiskers, minimum and maximum. (C) CSF2-expressed single cells were manually classified into CSF2_HIGH and CSF2_LOW subgroup. (D) Violin comparison of functional genes expression between CSF2_HIGH and CSF2_LOW group. (E) The correlation between CSF2 expression and representative functional genes. (F) Violin comparison of activation surface markers between CSF2_HIGH and CSF2_LOW group. The $P$ values were calculated with two-tailed Mann-Whitney test. ****P<0.0001. CR, complete remission; HD, healthy donor; IL, interleukin; NR, non-responder.

Figure 6

Integrated analyses identify the upregulation of MHC II genes among functional heterogeneities in two more literature-derived activated donor samples. (A) UMAP clustering of two more literature-derived activated 4-1BB donor samples. The top five DEGs of specific clusters were displayed. (B) Proportion of cells from each sample in each cluster. HD-L1 and HD-L2 represent literature-derived samples. (C) Signaling pathways regulated by DEGs generated through comparing each donor sample with CR sample. $z$ score reflects the predicted activation level ($z\S amp;lt;0$, inhibited; $z\S amp;gt;0,$ activated; $z\ge 2$ or $z\le -2$ can be considered significant). (D) Dot plot of functional genes expression in CR and three donor samples. (E) Expression level comparisons of a full set of MHC II defining genes between three donor samples and CR. Each data point represents the average expression level of all the single cells of each gene. CR, complete remission; DEGs, differentially expressed genes; HD, healthy donor; MHC, major histocompatibility complex; NR, non-responder; UMAP, uniform manifold approximation and projection.

Figure 7

Comparison with CD28 CAR T cells identifies the enhanced memory signature in basal 4-1BB CAR T cells. (A) UMAP clustering of donor 4-1BB and CD28 CAR T cells at basal condition identified the separate cluster distribution of the two CAR T samples. (B) Upregulated or downregulated DEGs of directly comparing 4-1BB CAR T with CD28 CAR T sample. (C) The expression of memory T cell related genes in identified clusters. The numbers at lower left indicate the average expression level of all the single cells of specific gene in each sample. (D) The distribution of surface protein ADT-CCR7 and ADT-CD62L expression revealed by CITE-seq. CAR, chimeric antigen receptor; DEGs, differentially expressed genes; UMAP, uniform manifold approximation and projection.