Paper Details

ABSTRACT

Background: CD19-directed chimeric antigen receptor (CAR) T-cell treatment promotes long-term remission in about a third of relapsed or refractory large B-cell lymphoma patients. Emerging evidence suggests durable responses depend on activation of endogenous antitumor immunity in addition to CAR-T-mediated direct tumor cell killing. Our previous work implicates tumor MHC-I loss in resistance to CD19-directed CAR-T therapy, but the role of tumor antigen presentation in shaping immune remodeling in response to treatment is unknown. Single-cell interrogation of CAR-T responses in real time in immunocompetent models provides opportunity to fill this gap and generate new comprehension of the complex interactions occurring among CAR-T cells, tumor cells, and endogenous immune populations in the tumor microenvironment (TME).

Methods: A20 wild-type (WT) or B2m-knockout (B2m-KO) lymphomas were established in syngeneic BALB/c mice. Hosts were treated or untreated with 2 x 10⁶ strain-matched anti-CD19 CAR-T cells, and we harvested tumors 5 days later during peak CAR-T activity in TMEs. We subjected tumors from 16 animals (WT-Control, n = 5; WT-CAR-T, n = 5; B2m-KO-Control, n = 3; B2m-KO-CAR-T, n = 3) to 10X Genomics GEM-X Flex single-cell RNA sequencing. Following quality control, 328,229 cells were suitable for analysis using Seurat with Harmony integration, SingleR-based annotation, differential gene-expression analysis, Hallmark pathway enrichment, and immune-cell subclustering.

Results: Clustering resolved malignant B cells and major TME populations and revealed key differences in immune infiltration. CAR-T therapy increased T-cell abundance in WT, but this effect was absent in B2m-KO (p<0.00001). In contrast, monocyte and neutrophil populations were enriched in B2m-KO, both with and without CAR-T treatment (p<0.00001 for both relative to WT), suggesting inflamed myeloid-enriched TMEs known to associate with poor CAR-T outcomes in DLBCL patients. We therefore investigated how these TMEs impacted CAR-T penetration and activation plus secondary impacts on endogenous T-cell populations. We re-clustered 21,930 total T cells in the data and used mCherry expression to distinguish CAR from non-CAR cells in the treated samples. B2m-KO tumors were dramatically less infiltrated by CD8+ CAR-T cells (p<0.00001). Moreover, while endogenous CD8 (eCD8) T cells expanded compared to untreated in response to CAR-T in WT, they declined in B2m-KO. Examining eCD8 sub-clusters showed deficiency in B2m-KO of specific populations associated with CAR-T responses, including central-memory, effector-memory, cytotoxic, and proliferative (p<0.00001 for all). MHC-I deficiency also altered the endogenous CD4 compartment demonstrating a shift toward T regulatory cells and immunoregulatory TH17 cells (p<0.00001 for both). Next, we compared treated WT and B2m-KO T-cells by gene-expression and Hallmark pathway-enrichment analyses. WT tumors showed significantly (Log2FC>0.5, Padj<0.05) higher expression of cytotoxic effector genes, including Gzmb, Prf1, Gzma, Serpinb6b, Serpinb9, Nkg7,and Klrk1. Hallmark analysis showed enrichment of E2F targets, G2M checkpoint, MYC targets, oxidative phosphorylation, and mitotic-spindle programs in T cells from treated WT tumors (Benjamini–Hochberg FDR-adjusted P<0.05), while B2m-KO T cells were enriched for IL-2-STAT5, inflammatory-response, TNF-NF-kB, TGF-beta, IL-6-JAK-STAT3, interferon-alpha, and hypoxia-associated programs. Strikingly, these altered patterns were broadly similar also for CAR-T cells within the treated TMEs.  

Conclusions: Overall, tumor MHC-I licensed both the numerical expansion and cytotoxic differentiation of endogenous T cells after CAR-T therapy. In an absence of tumor MHC-I, endogenous T cells failed to fully acquire an effector program and instead demonstrated inflammatory and immunoregulatory features. An absence of tumor cell-surface MHC-I therefore adversely impacted CAR-T penetration to TMEs and secondary activation of endogenous T-cell populations, associated with increased myeloid infiltration and inflammatory gene expression. These data align with our previous results from patient samples and immunocompetent models demonstrating impaired CAR-T responses in tumors lacking MHC-I expression. Loss of MHC-I, common in LBCL pathogenesis, is an emerging tumor-intrinsic driver of impaired CAR-T efficacy in need of novel solutions in development of future cellular immunotherapies.