CD4+ anti-TGF-β CAR T cells and CD8+ conventional CAR T cells exhibit synergistic antitumor effects

Abstract

Transforming growth factor (TGF)-β1 restricts the expansion, survival, and function of CD4⁺ T cells. Here, we demonstrate that CD4⁺ but not CD8⁺ anti-TGF-β CAR T cells (T28zT2 T cells) can suppress tumor growth partly through secreting Granzyme B and interferon (IFN)-γ. TGF-β1-treated CD4⁺ T28zT2 T cells persist well in peripheral blood and tumors, maintain their mitochondrial form and function, and do not cause in vivo toxicity. They also improve the expansion and persistence of untransduced CD8⁺ T cells in vivo. Tumor-infiltrating CD4⁺ T28zT2 T cells are enriched with TCF-1⁺IL7R⁺ memory-like T cells, express NKG2D, and downregulate T cell exhaustion markers, including PD-1 and LAG3. Importantly, a combination of CD4⁺ T28zT2 T cells and CD8⁺ anti-glypican-3 (GPC3) or anti-mesothelin (MSLN) CAR T cells exhibits augmented antitumor effects in xenografts. These findings suggest that rewiring TGF-β signaling with T28zT2 in CD4⁺ T cells is a promising strategy for eradicating solid tumors.

Keywords: CAR T; SMAD4; T cell exhaustion; TGFβ1; glypican-3; mesothelin; mitochondrial fission.

Graphical abstract

Figure 1

Anti-TGF-β CAR T cells reduced tumor growth and promoted T cell expansion in vivo (A) Anti-TGF-β CAR vector (T28zT2), anti-GPC3 CAR vector (G28zT2), and anti-CD19 CAR vector (1928zT2) based on an anti-TGF-β scFv (US20140127230A1), anti-GPC3 scFv (GC33), or anti-CD19 scFv (FMC63), respectively. All contained expression cassettes encoding a human CD8 leader signal peptide, CD28, CD3ζ, and TLR2 signaling domain along with eGFP using 2A self-cleaving peptide (2A). The eGFP expression was used to monitor CAR-transduced cells. (B) The percentage of Huh7 cells with 1928zT2, G28zT2, or T28zT2 T cell-induced lysis after 72 h; data are the mean percentage of tumor cell-specific lysis ± SEM values; n = 3 independent experiments; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p ≤ 0.0001. (C–F) A total of 4 $\times$ 10⁵ 1928zT2, T28zT2, or G28zT2 T cells were treated with PBS or TGF-β1 (10 ng/mL) for 15 min. These T cells were then cocultured with 1 $\times$ 10⁵ Huh7 cells for 72 h at a 4:1 effector (E):target (T) ratio in 12-well round bottom plates for 72 h. Supernatants were harvested and analyzed with a multiplex immunoassay to determine the concentrations of the indicated cytokines. (C) Representative crystal violet images of Huh7 cells with 1928zT2, T28zT2, or G28zT2 T cells-induced lysis after 72 h. Quantification of residual tumor cells (D) and summary of IFN-γ (E) and Granzyme B (F) released by CAR T cells (from 4 independent experiments); data are the mean $\pm$ SEM values; one-way ANOVA with Tukey’s multiple comparisons test; ∗∗p$\le$ 0.01, ∗∗∗p$\le$ 0.001, ∗∗∗∗p$\le$ 0.0001. (G) Eight-week-old male NSI mice were inoculated subcutaneously with 2 $\times$ 10⁶ Huh7 cells into the right flanks. A total of 5 $\times$ 10⁶ CAR T cells or PBS was injected peritumorally when the xenograft volume was ∼50 mm³ (day 0). The majority of mice exhibited severe graft-versus-host disease (GVHD) symptoms, halting animal experiments on day 27. Growth curves of Huh7 tumors in NSI mice post-infusion of T28zT2, G28zT2, and 1928zT2 T cells or PBS treatment (n = 8 mice/group); data are the mean $\pm$ SD values; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p$\le$ 0.0001. (H–J) The percentages of CD4⁺ (H), CD8⁺ (I), CAR⁺(GFP⁺) CD4⁺ (square, J), and CAR⁺CD8⁺ (triangle, J) T cells (n = 8 mice/group) of all nucleated cells in murine peripheral blood (PB) were determined by flow cytometry for the T28zT2, G28zT2, and 1928zT2 groups on day 27; data are the mean $\pm$ SD values; one-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p$\le$ 0.0001. (K–M) The percentages of CD4⁺ (K), CD8⁺ (L), and CAR⁺ (M) T cells in all nucleated cells from Huh7 tumors in the 1928zT2, T28zT2, and G28zT2 groups on day 27 determined by flow cytometry (n = 3 mice per group). Data are shown as the mean $\pm$ SD values; one-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗p$\le$ 0.001, ∗∗∗∗p$\le$ 0.0001. (N–P) Representative images of CD4⁺ (N, top) and CD8⁺ (N, bottom) T cells (brown) in Huh7 tumors from the T28zT2, G28zT2, and 1928zT2 groups on day 27. The frequencies of CD4⁺ (O) and CD8⁺ T cells (P) were calculated by ImageJ software (n = 4 mice per group). Scale bar, 20 μm. Data are shown as the mean $\pm$ SD values; one-way ANOVA with Tukey’s multiple comparisons test; ∗∗p$\le$ 0.01, ∗∗∗∗p$\le$ 0.0001. See also in Figures S1 and S2.

Figure 2

CD4⁺ anti-TGF-β CAR T cells exhibit memory-like T cell phenotypes in xenografts (A) 2D projection of the sample distribution (left) and subclusters (right) of purified splenic CAR⁺(GFP⁺) CD4⁺ T cells from the T28zT2 group (blue) and G28zT2 (red) group using t-SNE. T28zT2 clusters contain clusters 8–15, defined as IL7R⁺PD-1⁻LAG3⁻ T cell subsets; G28zT2 clusters contain clusters 1–6, defined as IL7R⁻PD-1⁺LAG3⁺ T cell subsets or IL7R⁻PD-1⁺LAG3⁻ T cell subsets. (B) PhenoGraph cluster distribution comparing CAR⁺CD4⁺ T cells between the T28zT2 group (blue) and G28zT2 (green) group. (C) Differences in gene expression between the T28zT2 and G28zT2 groups of individual purified splenic CAR⁺(GFP⁺) CD4⁺ T cells in the t-SNE projection, including CD4, IL7R (CD127), TCF-1, CXCR3, PD-1, and LAG3. (D) Volcano plot of DEGs showing upregulated (red) and downregulated (blue) DEGs and non-DEGs (gray) identified by RNA-seq in T28zT2 CAR⁺CD4⁺ T cells compared to G28zT2 cells. Adjustment for the false discovery rate (FDR) results in an adjusted p value called the q value. The y axis shows the significance value after −log₁₀ transformation of the FDR (−Log₁₀(FDR)). The x axis shows the fold difference threshold between the T28zT2 and G28zT2 groups (Log₂(T28zT2/G28zT2)). (E) GSEA of the Wnt signaling pathway (p = 0.016) in CAR⁺CD4⁺ T cells. From left to right, the genes in the rank-ordered list are enriched in the T28zT2 and G28zT2 groups. (F) Heatmap of DEGs involved in T cell differentiation, exhaustion, and activation in CAR⁺CD4⁺ cells identified in comparisons between the T28zT2 and G28zT2 groups. Cutoff: absolute log2 (fold change) ≥ 1; adjusted p value ≤ 0.05. See also in Figure S3.

Figure 3

CD4⁺ but not CD8⁺ T28zT2 T cells are effective for tumor growth inhibition (A) CD4⁺ and CD8⁺ T cell compartments were sorted, and CD8⁺ T28zT2 T cells and CD4⁺ T28zT2 cells were generated. We also mixed the CD4⁺ and CD8⁺ T28zT2 T cells at a ratio of 1:1 to obtain CD3⁺ T28zT2 T cells. We then infused a total of 5 $\times$ 10⁶ of the three types of T28zT2 T cells, CD3⁺ 1928zT2 T cells, or CD3⁺ M28zT2 T cells peritumorally into subcutaneous xenografts of NSI mice inoculated with 1 $\times$ 10⁶ AsPc-1 cells (day 0). Tumor volumes were monitored on the indicated days (n = 8 mice/T28zT2 CD3, T28zT2 CD4, and T28zT2 CD8 groups; n = 6 mice/1928zT2 CD3, M28zT2 CD3 and PBS groups); data are the mean $\pm$ SD values; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗p$\le$ 0.01, ∗∗∗∗p$\le$ 0.0001. (B–D) CRISPR-Cas9-RNP knockout of GZMB or IFNG expression in CD4⁺ T28zT2 T cells. A total of 4 $\times$ 10⁵ T28zT2 CD4 control single guide RNA (sgctrl), T28zT2 CD4 single guide RNA targeting granzyme B (sgGZMB), T28zT2 CD4 single guide RNA targeting interferon-gamma (sgIFNG), G28zT2 CD4 sgctrl, or 1928zT2 CD4 sgctrl T cells were cocultured with 1 $\times$ 10⁵ Huh7 cells for 72 h. Shown are the levels of IFN-γ (B) and Granzyme B (C) (from 4 independent experiments) detected by ELISA assay; (D) the relative viability of Huh7 cells (from 4 independent experiments); data are the mean $\pm$ SEM values; one-way ANOVA with Tukey’s multiple comparisons test; ∗p < 0.05, ∗∗∗p$\le$ 0.001, ∗∗∗∗p$\le$ 0.0001. (E) Curves showing variations in the volume of Huh7 tumors in NSI mice post-infusion of T28zT2 CD4 sgctrl, T28zT2 CD4 sgIFNG, T28zT2 CD4 sgGZMB, G28zT2 CD4 sgctrl, or 1928zT2 CD4 sgctrl T cells (n = 5 mice/group); data are the mean $\pm$ SD values; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p$\le$ 0.0001. See also in Figures S4 and S5.

Figure 4

A combination of CD4⁺ anti-TGF-β CAR T cells and CD8⁺ anti-MSLN CAR T cells exhibits augmented antitumor effects in AsPc-1 tumor models (A) Mixed CAR T cells consisted of CD4⁺ T28zT2 T cells and CD8⁺ M28zT2 T cells, designated T4M828zT2 T cells. Graphics were created with BioRender.com (agreement number VP27TZMIXE). (B) The percentage of AsPc-1-GL cells with 1928zT2, M28zT2, T28zT2 CD4, or T4M828zT2 T cell-induced lysis overnight; data are the mean percentage of tumor cell-specific lysis ± SEM values; n = 3 independent experiments; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p ≤ 0.0001. (C) A schematic diagram of the experimental design. 8-week-old male NSI mice were inoculated subcutaneously with 2 $\times$ 10⁶ AsPc-1 tumor cells into the right flank. A total of 5 $\times$ 10⁶ T4M828zT2 T, M28zT2 T, and 1928zT2 T cells were injected peritumorally (day 0). At each endpoint, the splenic CAR T cells were sorted by fluorescence-activated cell sorting (FACS) and cryopreserved. Graphics were created with BioRender.com (agreement number GX26UUJLAJ). (D) Tumor volumes were monitored on the indicated days (n = 11 mice/1928zT2 and M28zT2 group, n = 10 mice/T28zT2 CD4 group, n = 15 mice/T4M828zT2 group). The majority of mice in all groups exhibited serious symptoms of GVHD, halting animal experiments on day 35 post-injection. Data are shown for 2 independent experiments; displayed as the mean $\pm$ SEM values; a repeat measures ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p$\le$ 0.0001 (T4M28zT2 T cells vs. 1928zT2 T cells on day 35, T4M28zT2 T cells vs. M28zT2 T cells on day 35, and T4M28zT2 T cells vsT28zT2 CD4 T cells on day 35). (E) The percentages of CAR⁺(GFP⁺) CD4⁺ T cells in murine peripheral blood (PB) populations of mice from the T4M828zT2 and M28zT2 groups on day 35 were determined by flow cytometry. n = 6 mice/group; data are the mean $\pm$ SD values; one-way ANOVA with Tukey’s multiple comparisons test; ∗p < 0.05; ∗∗∗∗p $\le$ 0.0001. (F–H) The percentages of PD-1⁺LAG3⁺ expression among CAR⁺CD4⁺(F and G) and CAR⁺CD8⁺ (F and H) T cells in murine peripheral blood (PB) of mice from the T4M828zT2 and M28zT2 groups on day 35 were determined by flow cytometry. n = 6 mice/group; (G and H) data are the mean $\pm$ SD values; unpaired two-tailed t test; ∗∗p$\le$ 0.01, ∗∗∗p$\le$ 0.001. (I) 8-week-old male NSI mice were inoculated subcutaneously with 2 $\times$ 10⁶ AsPc-1 tumor cells into the right flank. A total of 5 $\times$ 10⁶ T4M828zT2 T, T28zT2 CD4, M28zT2 T, or 1928zT2 splenic T cells were injected peritumorally. Tumor volumes were monitored on the indicated days (n = 5 mice/group); data are shown as the mean $\pm$ SD values; ∗∗∗p$\le$ 0.001, ∗∗∗∗p$\le$ 0.0001. See also in Figure S6.

Figure 5

A combination of CD4⁺ anti-TGF-β CAR T cells and CD8⁺ anti-MSLN CAR T cells exhibits augmented antitumor effects in NSCLC PDX (A) NSCLC PDX tumors were diced into ∼30 mm³ pieces, and tissue were inoculated subcutaneously into the right flanks of 8-week-old male NSI mice. 5 $\times$ 10⁶ T4M828zT2, T28zT2 CD4, M28zT2, or 1928zT2 T cells were injected peritumorally (day 0). Tumor volumes were monitored on the indicated days (n = 6 mice/group); data are the mean $\pm$ SD values; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗p$\le$ 0.01, ∗∗∗∗p$\le$ 0.0001. (B–D) The mean fluorescence intensity (MFI) of PD-1 among tumor-infiltrating CAR⁺CD4⁺ (B, left) and CAR⁺CD8⁺ (B, right) T cells from the T4M828zT2 and M28zT2 groups on day 21 determined by flow cytometry (n = 4 mice/group). (C and D) Data are shown as the mean $\pm$ SD values; unpaired two-tailed t test; ∗p < 0.05, ∗∗p$\le$ 0.01. (E) A schematic diagram of the experimental design. Tumor tissue from 1928zT2, M28zT2, or T4M828zT2 groups was obtained from NSCLC PDX models at the endpoint (day 21). Tumor tissues were prepared into single-cell suspension. Tumor-infiltrating CAR⁺(GFP⁺) CD8⁺ T cells from the M28zT2 and T4M828zT2 groups and tumor-infiltrating CD8⁺ T cells from the 1928zT2 group were sorted by FACS. These tumor-infiltrating T cells were then stimulated with CD3/CD28 monoclonal antibodies (mAbs) and subjected to functional experiments. Finally, the cytotoxicity, cytokine production, and T cell expansion of tumor-infiltrating CD8⁺ T cells from the 1928zT2, M28zT2, and T4M828zT2 groups were evaluated. Graphics were created with BioRender.com (agreement number QW27PPO9DF). (F–I) Tumor-infiltrating CD8⁺ 1928zT2, CAR⁺CD8⁺ M28zT2, or CAR⁺CD8⁺ T4M828zT2 cells were incubated with AsPc-1 cells at a 2:1 effector (E):target (T) ratio for 72 h. (F) Representative images of 0.1% crystal violet staining of AsPc-1 cells cocultured with CD8⁺ 1928zT2, CAR⁺CD8⁺ M28zT2, or CAR⁺CD8⁺ T4M828zT2 tumor-infiltrating T cells ex vitro. (G) The relative viability of AsPc-1 cells with 1928zT2, M28zT2, or T4M828zT2 T-cell induced lysis after 72 h n = 4 mice/group. (H and I) Supernatants were harvested and analyzed with a multiplex immunoassay to determine the concentrations of the indicated cytokines. n = 4 mice/group. The concentrations of IFN-γ (H) and Granzyme B (I) were measured by ELISA assay; data are the mean ± SD values; one-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. (J) The expansion of tumor-infiltrating CD8⁺ 1928zT2, CAR⁺CD8⁺ M28zT2, and CAR⁺CD8⁺ T4M828zT2 cells was detected by flow cytometry on day 0, 3 and 7; data are the mean ± SD values; n = 4 mice/group; two-way ANOVA with Tukey’s multiple comparisons test; ∗∗∗∗p ≤ 0.0001. See also in Figure S6.

Figure 6

TGF-β1 suppressed OXPHOS activity in CD4⁺ human T cells (A–D) CD4⁺ and CD8⁺ T cells (1 $\times$ 10⁶) were activated with CD3/CD28 mAbs for 24 h, followed by PBS or TGF-β1 (10 ng/mL) treatment for 16 h. (A) Mitochondrial morphology of CD4⁺ and CD8⁺ T cells upon PBS or TGF-β1 (10 ng/mL) treatment, as determined by spinning disk confocal microscopy. Mitochondria are green (MitoTracker Green), and nuclei are blue (DAPI). Scale bar, 5 μm. (B) Relative lengths of mitochondria, as analyzed by ImageJ software, in CD4⁺ and CD8⁺ T cells (2 independent experiments). Each dot represents the mean relative length of the mitochondria in a sample. Data are shown as the mean $\pm$ SEM values; paired two-tailed t test; ∗∗∗∗p$\le$ 0.0001. (C) Immunoblot analysis of cellular protein extracts probed with antibodies against pSMAD2^S465/467 (top)/pSMAD3^S423/425 (bottom) (pSMAD2/3), SMAD2/3, OPA1, MFF, pDRP1^S616, DRP1, and β-actin. (D) The relative expression of pDRP1^S616, MFF, and OPA1 was analyzed by ImageJ software (3 independent experiments). Data are shown as the mean $\pm$ SEM values; paired two-tailed t test; ∗p < 0.05, ∗∗p$\le$ 0.01, ∗∗∗p$\le$ 0.001. (E) Immunoprecipitation (IP) of MFF in activated CD4⁺ T cells after treatment with PBS or TGF-β1 (10 ng/mL) for 2 h and subsequent immunoblot (IB) analysis of the indicated proteins. (F) Mitochondrial morphology of CAR⁺CD4⁺ T28zT2, CAR⁺CD4⁺ G28zT2, and untransduced CD4⁺ T cells upon treatment with PBS or TGF-β1 (10 ng/mL) as determined by spinning disk confocal microcopy. Mitochondria are red (MitoTracker Deep Red), CAR T cells are green (GFP), and nuclei are blue (Hoechst 33342). Scale bar, 5 μm. (G) Lengths of mitochondria, as analyzed by ImageJ software (2 independent experiment), in CAR⁺CD4⁺ T28zT2, CAR⁺CD4⁺ G28zT2, and untransduced CD4⁺ T cells. Each dot represents the mean relative length of the mitochondria in a sample. (H–M) OCR profile (H and K), ATP-coupled OCR (I and L), and SRC (J and M) of CAR⁺CD4⁺ T28zT2 T cells and CAR⁺CD4⁺ G28zT2 T cells (3 independent experiments). Data are shown as the mean $\pm$ SEM values; unpaired two-tailed t test; ∗p < 0.05, ∗∗p$\le$ 0.01. See also in Figure S6.