Chimeric antigen receptor-induced antigen loss protects CD5.CART cells from fratricide without compromising on-target cytotoxicity

Abstract

Chimeric antigen receptor T cells (CART) targeting lymphocyte antigens can induce T cell fratricide and require additional engineering to mitigate self-damage. We demonstrate that the expression of a chimeric antigen receptor (CAR) targeting CD5, a prominent pan-T cell antigen, induces rapid internalization and complete loss of the CD5 protein on T cells, protecting them from self-targeting. Notably, exposure of healthy and malignant T cells to CD5.CART cells induces similar internalization of CD5 on target cells, transiently shielding them from cytotoxicity. However, this protection is short-lived, as sustained activity of CD5.CART cells in patients with T cell malignancies results in full ablation of CD5⁺ T cells while sparing healthy T cells naturally lacking CD5. These results indicate that continuous downmodulation of the target antigen in CD5.CART cells produces effective fratricide resistance without undermining their on-target cytotoxicity.

Keywords: CAR; CD5; T cell fratricide; T cell malignancies; chimeric antigen receptor; clinical trials; fratricide; off-tumor toxicity.

Graphical abstract

Figure 1

Fratricide-resistant CD5 CART cells internalize and lose detectable CD5 (A) Schematic of the CD5.CAR and ΔCD5.CAR. (B) Representative histograms and MFI of surface CD5 expression in control or CD5 knocked-out (CD5KO) primary human T cells non-transduced (NT) or gammaretrovirally transduced with CD19.CAR, CD5.CAR, and ΔCD5.CAR constructs. Representative data from 6 individual healthy donors from 2 experiments. p values calculated using one-way ANOVA with Dunnett’s correction. Data are represented as mean ± SD. (C) CD5 protein expression in control or CD5KO primary human T cells expressing CD19.CAR, CD5.CAR, and ΔCD5.CAR. Representative data from 2 donors. Data are represented as mean ± SD. (D) Schema of fluorescent mScarlet-tagged CD5, fluorescent mEmerald-tagged ΔCD5.CAR, or bi-cistronic ΔCD5.CAR with CD5.mScarlet (top) and the timeline of live-imaging microscopy experiment (bottom). (E) Maximal-projection images from live-imaging fluorescent microscopy of T cells serially transduced with CD5.mScarlet and ΔCD5.CAR-mEmerald. mScarlet localization (top). mEmerald localization (middle). Differential interference contrast (DIC) overlay with mScarlet and mEmerald (bottom). (F) Z-slice images of mScarlet localization in T cells when CD5.mScarlet is expressed in the absence (top) or presence of ΔCD5.CAR (bottom). (G) Schema of dual-tagged myc/FLAG CD5 co-expressed with ΔCD19.CAR or ΔCD5.CAR lacking both CD28 cytoplasmic domain. (H) Representative histogram of CD5 expression on the surface of primary human T cells either control non-transduced (NT) or transduced with dual-tagged CD5 and either ΔCD19.CAR (Ctrl, gray) or ΔCD5.CAR (purple). Representative data from 6 donors. (I) Representative flow plot of myc and FLAG expression on non-transduced (left) or transduced human T cells expressing dual-tagged CD5 with either ΔCD19.CAR (center) or ΔCD5.CAR (right, purple) Representative data from 6 donors. p values calculated using paired Student’s t test. Data are represented as mean ± SD. (J) myc and FLAG tag expression in ΔCD19.CART cells (Ctrl) or ΔCD5.CART cells expressing dual-tagged CD5. Representative data from 3 donors. (K) CD5 mRNA expression of in primary human T cells either non-transduced (NT) or transduced with CD19.CAR or CD5.CAR (n = 3 donors). Data are represented as mean ± SD.

Figure 2

Target T cells downmodulate CD5 protein upon exposure to CD5.CART cells (A) Representative time-course histograms of surface CD5 expression on healthy T cells cultured alone (left), with CD5.CART cells (middle) or with ΔCD5.CART cells (right) in 24-well plate at a 1:1 ratio under static (top) or dynamic (bottom) conditions. Representative data of 3 donors. Data are represented as mean ± SD. (B) Kinetics of CD5 downregulation on T cell targets cocultured with ΔCD5.CART cells over 24 h. (n = 3 donors). Data are represented as mean ± SD. (C) Maximal-projection images from live-imaging fluorescent microscopy of CD5.mScarlet-expressing T cells cultured with ΔCD5.CAR-mEmerald-expressing T cells. mScarlet localization (Top). mEmerald localization (middle). DIC overlay with mScarlet and mEmerald (bottom). T = 0 m indicates time of T cell contact with CAR T cell. (D) Representative histogram of surface CD5 levels on T cells expressing dual-tagged CD5 cultured either alone (gray) or with ΔCD5.CART cells (purple) (n = 3). p values calculated using paired Student’s t test. Representative data of 3 donors. Data are represented as mean ± SD. (E) Representative flow plot of myc and FLAG expression on control cells (left), or T cells expressing dual-tagged CD5 cultured alone (center) or with ΔCD5.CART cells (right). Percentage of total myc/FLAG expression obtained by normalizing myc/FLAG MFI on target T cells cultured with CART cells to myc/FLAG MFI on target T cells cultured alone, respectively (n = 3). p values calculated using paired Student’s t test. Representative data of 3 donors. Data are represented as mean ± SD. (F) CD5 protein expression on T cells cultured with CD19.CART cells (Ctrl) or with ΔCD5.CART cells. Representative western blot of 3 donors. (G) CD5 mRNA transcript expression of ATCs cultured with either donor-matched CD19.CAR T cells or ΔCD5.CAR T cells. Data normalized to the average of housekeeping genes (18S rRNA, B-actin, and beta-2-microglobulin). Data are represented as mean ± SD. (H) Schematic of in vivo mouse experiment evaluating CD5.CAR trans removal of CD5 on primary autologous ATCs. n = 3 mice. (I) CD5 expression on autologous donor ATCs injected alone (top) or together with ΔCD5.CAR T cells. Representative data of 3 mice. (J) CD5 mRNA transcript expression of autologous donor ATCs injected alone or together with ΔCD5.CAR T cells. CD5 mRNA level in ΔCD5.CAR T cells is shown as a gray bar (n = 3 mice). Data are represented as mean ± SD. (K) CD5 expression of T cells cocultured with ΔCD5.CART cells and then sorted to remove CART cells from culture. Time indicates hours post-sort. Each line represents 2 individual donors. Data are represented as mean ± SD. (L) Schematic of mixed coculture experiment with CD5⁺ or CD5KO (Ctrl targets) T cell targets pre-cultured with either CD19.CART cells (Ctrl effectors) or ΔCD5.CART cells for 20 h, and then subsequently cultured with cytotoxic CD5.CART cells for either 6 or 20 h. (M) CD5 expression on T cell targets prior to coculturing with cytotoxic CD5.CART cells. p values calculated using one-way ANOVA with Sidak's correction (n = 8 donors from 3 experiments). Data are represented as mean ± SD. (N) Percentage of killing of primary autologous activated human T cells pre-cultured with either CD19.CAR.mEmerald (control) or ΔCD5.CAR.mEmerald-expressing T cells for 20 h before adding cytotoxic CD5.CART cells. CD5 CRISPR knocked-out (CD5KO) primary autologous activated T cells pre-cultured with CD19.CAR-mEmerald-expressing T cells were used as a control. p values calculated using one-way ANOVA with Dunnett’s correction (n = 8 donors from 3 experiments). Data are represented as mean ± SD. (B, D, G, M) Percentage of total CD5 expression obtained by normalizing CD5 MFI on target T cells cultured with CART cells to CD5 MFI on control target T cells cultured alone at each time point. (D, E, M, N) Each dot denotes an individual donor.

Figure 3

T cells resistant to CD5.CART cell elimination are enriched for CD8⁺ effector T cells (A) Percentage of residual targets following coculture of CD5.CART cells with resting T cells, CCRF-CEM, or Jurkat tumor cells for 48 h at an E:T ratio of 1:3. Resting T cells were CD3⁺ MACS isolated from fresh peripheral blood mononuclear cells. p values calculated using one-way ANOVA with Dunnett’s correction. n = 3 donors. Data are represented as mean ± SD. (B) Freshly isolated T cells were cocultured for 36 h with autologous CD19.CART cells or CD5.CART cells. Residual target cells were sort-purified and subjected to CITE-Seq. ssRNA-seq clustering by Seurat is shown. (C) Jaccard Similarity Index of Seurat UMAP of autologous T cells cocultured with either CD19.CART cells (left) or CD5.CART cells (center). Table denoting enriched and depleted clusters (right). (D) CD4, CD8, CD45RA, and CD62L surface expression of T cells (CITE-Seq) cultured with either CD19.CART cells or CD5.CART cells overlaid on Seurat UMAP. Gradient denotes log UMI counts of surface transcripts from CITE-seq. (E) CD5 protein surface expression (orange, top) and CD5 transcript levels (green, bottom) in residual T cells after coculture with either CD19.CART cells or CD5.CART cells overlaid on Seurat UMAP. Gradient denotes log UMI counts of surface and gene transcripts from CITE-Seq and RNA-seq, respectively. (F) Heatmap denoting fold change of genes in enriched clusters 3, 4, and 5, or depleted cluster 0 and 2 vs. all other cells; clustering by Seurat. (G) Differential abundance neighborhood groups. Color denotes significantly upregulated/downregulated genes with fold change differences above in CD5 CART treatment (red) or below compared to CD19 CART treatment (blue) with half log threshold using Milo. (H) Differentially expressed genes between CD19.CART treatment (left) and CD5.CART treatment (right) using Milo. Genes upregulated under CD5.CART treatment on the right.

Figure 4

Long-term resistance of normal T cells to CD5.CART cells is mediated by the loss of CD5 gene expression (A) Gating strategy (top). CD5 expression of T cells in patient #5 treated with CD5.CART cells that produced an abbreviated expansion (bottom). (B) CD3⁺ T cells in peripheral blood of patients with T cell lymphoma (TCL) or T cell acute lymphoblastic leukemia (T-ALL) treated with CD5.CART cells (NCT03081910). (C) CD5 expression of T cells in patient #7 (top), #8 (middle), and #9 (bottom) with durable CD5.CART cell persistence and achieved complete response. (D) CD5 mRNA expression of healthy T cells isolated from a lymph node biopsy at week 5 in patient #7 (left) and week 4 in patient #8 (right) normalized to respective donors’ T cells pre-lymphodepletion. (E) T cell phenotype of healthy T cells based on CCR7 and CD45RA expression of patient #7, #8, and #9. Naive (CCR7⁺/CD45RA⁺), central memory (CCR7⁺/CD45RA⁻), effector memory (CCR7⁻/CD45RA⁻), TEMRA (CCR7⁻/CD45RA⁺), and CD4/CD8 T cell distributions in healthy circulating T cells in patient #7, #8, and #9.