SMAD7 expression in CAR-T cells improves persistence and safety for solid tumors

Abstract

Despite the tremendous progress of chimeric antigen receptor T (CAR-T) cell therapy in hematological malignancies, their application in solid tumors has been limited largely due to T-cell exhaustion in the tumor microenvironment (TME) and systemic toxicity caused by excessive cytokine release. As a key regulator of the immunosuppressive TME, TGF-β promotes cytokine synthesis via the NF-κB pathway. Here, we coexpressed SMAD7, a suppressor of TGF-β signaling, with a HER2-targeted CAR in engineered T cells. These novel CAR-T cells displayed high cytolytic efficacy and were resistant to TGF-β-triggered exhaustion, which enabled sustained tumoricidal capacity after continuous antigen exposure. Moreover, SMAD7 substantially reduced the production of inflammatory cytokines by antigen-primed CAR-T cells. Mechanistically, SMAD7 downregulated TGF-β receptor I and abrogated the interplay between the TGF-β and NF-κB pathways in CAR-T cells. As a result, these CAR-T cells persistently inhibited tumor growth and promoted the survival of tumor-challenged mice regardless of the hostile tumor microenvironment caused by a high concentration of TGF-β. SMAD7 coexpression also enhanced CAR-T-cell infiltration and persistent activation in patient-derived tumor organoids. Therefore, our study demonstrated the feasibility of SMAD7 coexpression as a novel approach to improve the efficacy and safety of CAR-T-cell therapy for solid tumors.

Introduction

Adoptive T-cell immunotherapy has thus far achieved promising therapeutic efficacy in treating hematologic maligancies [1–3]. In particular, chimeric antigen receptor T (CAR-T) cells that target anti-CD19 or B-cell maturation antigen (BCMA) have been approved for the treatment of relapsed and refractory B-cell malignancies [4, 5]. In contrast, CAR-T-cell therapy has yielded limited clinical responses in patients bearing solid tumors [6–8]. The hostile tumor microenvironment (TME), which is composed of various types of immunosuppressive cells and their secretory factors, such as cytokines, represents one of the main obstacles to the effective repression of solid tumors by CAR-T cells [9–11]. Moreover, the severe systemic toxicity caused predominantly by excessive inflammatory cytokine release remains to be overcome in CAR-T-cell therapy for cancer.

Transforming growth factor beta (TGF-β) is one of the pivotal cytokines that maintains the immunosuppressive TME and is secreted by multiple categories of cells, including cancer cells, stromal fibroblasts, macrophages, and regulatory T cells [12–15]. The TGF-β family of proteins consists of 3 members, of which TGF-β1 is most frequently involved in tumorigenesis [16, 17]. The signaling pathway is initiated by the engagement of TGF-β with its cognate heterodimeric receptor, TGF-β receptor I/II (TβRI/II), which recruits and phosphorylates SMAD2 and SMAD3. These so-called receptor-regulated SMADs (R-SMADs) subsequently form a trimeric complex with SMAD4 and accumulate in the nucleus to regulate the expression of target genes [18]. The hyperactivity of TGF-β signaling has been reported to promote metastasis, suppress the proliferation and activation of T cells, and restrain the tumoricidal capability of infiltrating lymphocytes [19–21]. Nonetheless, TGF-β signaling can be counteracted by a class of endogenous negative regulators, including the inhibitory SMAD protein (I-SMAD) SMAD7 [17, 22]. SMAD7 attenuates TGF-β signaling by forming a stable complex with TβRI, which consequently inhibits the phosphorylation of R-SMADs and induces the ubiquitylation and degradation of the receptor by recruiting two types of E3 ubiquitin ligases (SMURF1/2) [22–26]. In addition, SMAD7 impairs the transcription factor activity of the trimeric SMAD complex by dampening their association with SMAD-binding elements (SBEs) on target genes [27].

Accumulating efforts have been made to target the TGF-β pathway in an attempt to remodel the immune microenvironment and enhance the antitumor effect of immunotherapy. One approach is to edit the TGF-β receptor to render T cells unresponsive to TGF-β. For instance, knocking out endogenous TβRII using CRISPR/Cas9 was reported to promote the long-term efficacy of CAR-T cells against solid tumors [28]. Similarly, a preclinical study demonstrated that coexpression of a dominant-negative TGF-β receptor II (dnTβRII) endows tumor-specific T cells with stronger antitumor ability [29]. On this basis, Noh et al. incorporated an immunostimulatory interleukin (IL)-7 receptor (I7R) signaling endodomain into dnTβRII, which conferred CAR-T cells with significantly improved antitumor efficiency [14]. In addition to targeting the TGF-β receptor, disturbance of distal TGF-β signaling may be another potential way to block signal transduction. However, TGF-β also plays an anti-inflammatory role through SMAD7-mediated blockade of TGF-β-activated kinase 1 (TAK1) recruitment to the tumor necrosis factor (TNF) pathway [30]. We thus hypothesize that the introduction of SMAD7 into CAR-T cells might concomitantly improve the efficacy and safety of cancer therapy by remodeling the immunosuppressive TME and reducing the production of inflammatory cytokines, respectively. In the present study, we designed CARs that target human epidermal growth factor receptor 2 (HER2), which is overexpressed in various solid malignancies, and engineered T cells to coexpress these CARs and SMAD7. The effect of SMAD7 on the antitumor performance of CAR-T-cell therapy was evaluated, and the underlying mechanisms were explored by investigating the potential crosstalk between the TGF-β and NF-κB pathways.

Results

CAR-T cells coexpressing SMAD7 are cytotoxic and persistently proliferative in vitro

We designed typical second-generation CARs consisting of a single-chain variable fragment (scFv) against HER2 (P1h2) or CD19 (FMC63, as a control), a hinge and transmembrane region from CD8α, a costimulatory domain from CD28, and an intracellular domain from CD3ζ (Fig. 1A). By utilizing a T2A element, we generated cassettes that cotranslationally expressed CAR and SMAD7 (Fig. 1A, Fig. S1A). Recombinant lentiviruses harboring these cassettes were used to infect T cells prepared from human PBMCs. As a result, the cells displayed similar multiplicities of infection (MOIs) (Fig. S1B), and the engineered T cells expressed comparable levels of CAR (Fig. 1B). Next, HeLa and PC9 cells that ectopically expressed HER2 were generated and used as target cells to evaluate the effect of SMAD7 on the killing capability of CAR-T cells (Fig. S1C). We found that both groups of HER2-targeted CAR-T cells, but not those designed to recognize CD19, could efficiently induce the lysis of these target cells (Fig. 1C, D, Fig. S2A, B). When cocultured with target cells that produced modest levels of TGF-β1, such as PC9 cells transduced to express HER2 or the HER2-positive cell lines SKOV3 and SKBR3, CAR-T cells exhibited similar cytolytic activities regardless of SMAD7 coexpression (Fig. 1D, Fig. S1D, Fig. S2C, D). In contrast, SMAD7 significantly improved the cytotoxicity of CAR-T cells when cocultured with HER2-expressing HeLa cells, which secreted relatively high levels of TGF-β1 (Fig. 1D, Fig. S1D). We next performed long-term culture of CAR-T cells plus intermittent stimulation with malignant cells and found that the proliferation of CAR-T cells expressing SMAD7 was increased compared to that of CAR-T cells without SMAD7 when cocultured with HER2⁺ HeLa cells (Fig. 1E). Similarly, SMAD7 improved the survival and proliferation of CAR-T cells under stimulation with HER2⁺ PC9 cells and recombinant TGF-β1 (Fig. 1F). Thus, SMAD7 facilitates the persistent proliferation of CAR-T cells by counteracting the effect of neoplastic cell-derived TGF-β.

Fig. 1.

Establishment and validation of CAR-T-cell cytotoxic function in vitro. A Schematic diagram of CARs comprising a HER2 or CD19 scFv in the extracellular domain. B FCM assay for CAR expression on engineered primary human T cells. CAR-T cells were incubated with HER2⁺Fluc⁺ HeLa cells C or HER2⁺Fluc⁺ PC9 cells D at the indicated E/T ratio for 12 h, and tumor cell killing was assayed by measuring luciferase activity. NT, nontreated control. E CAR-T cells were cultured and intermittently stimulated with HER2⁺Fluc⁺ HeLa cells (arrows). Cells were then counted on the indicated days, and the numbers of cells were plotted. F CAR-T cells were cultured in medium with or without rhTGF-β1 (5 ng/mL) and intermittently stimulated with HER2⁺Fluc⁺ PC9 cells (arrows). Cells were then counted on the indicated days, and the numbers of cells were plotted. Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

Coexpression of SMAD7 ameliorates CAR-T-cell exhaustion and cytokine release

To better understand how SMAD7 affects the gene expression profiles of CAR-T cells, we performed RNA-seq after coculture with HER2⁺Fluc⁺ HeLa cells (Fig. S3A). We found that compared with conventional CAR-T cells, those coexpressing SMAD7 showed substantial downregulation of PDCD1, LAG3 and TIGIT, which are markers of T-cell exhaustion, and concurrent upregulation of CD27, Jun and KLF9, which suggests sustained T-cell activation and differentiation into a memory phenotype (Fig. 2A, Fig. S3B). We then incubated CAR-T cells with different concentrations of rhTGF-β1 and found that SMAD7-expressing CAR-T cells displayed lower ratios of subsets expressing the above exhaustion markers but a higher proportion of cells expressing TCF1, a biomarker for memory T cells (Fig. S3C). Moreover, flow cytometry assays validated that SMAD7-expressing CAR-T cells were more refractory to exhaustion than normal CAR-T cells after coculture with target cells (HER2⁺Fluc⁺ HeLa cells) (Fig. 2B). TGF-β promotes the differentiation of naive T cells to regulatory T (Treg) cells [31]. We next evaluated whether SMAD7 can prevent the differentiation of CAR-T cells toward the Treg subset. Indeed, SMAD7-expressing CAR-T cells displayed a lower proportion of CD4⁺FOXP3⁺ cells than cells without SMAD7 after coculture with malignant cells (Fig. 2C). Moreover, these CAR-T cells were more resistant to activation-induced apoptosis following incubation with HER2⁺Fluc⁺ HeLa cells (Fig. 2D). The same results were observed in conventional mesothelin-targeting CAR-T cells (M28z) and those coexpressing SMAD7 (S7M28z) (Fig. S4). KEGG pathway enrichment analysis of the RNA-seq data showed that the differentially expressed genes (DEGs) between CAR-T cells with or without SMAD7 expression were enriched in various pathways, including T-cell differentiation, proximal T-cell receptor signaling and cytokine-producing pathways (Fig. S3D). GSEA of the RNA-seq data also revealed a significant enrichment of downregulated genes (S7H28z vs. H28z) in cytokine-related pathways (Fig. 2E). Consistently, multiplexed cytokine analysis showed that coexpression of SMAD7 in CAR-T cells could significantly reduce the secretion of cytokines (Fig. 2F, G, Table S3). These results indicated that CAR-T cells expressing SMAD7 are characterized by mitigated exhaustion and decreased cytokine production upon antigen exposure.

Fig. 2.

Coexpression of SMAD7 prevents the exhaustion of CAR-T cells and reduces the release of inflammatory cytokines. A H28z and S7H28z CAR-T cells were stimulated with HER2⁺Fluc⁺ HeLa cells for 16 h and were subjected to RNA sequencing. Heatmap including the differentially expressed genes (DEGs) related to T-cell exhaustion and a memory phenotype. B FCM assays for expression of the indicated markers on CAR-T cells with (sti) or without (unsti) HER2⁺Fluc⁺ HeLa cells for 24 h. Representative FCM plots of CAR-T cells subjected to C expression of CD4/FOXP3 and D expression of Annexin V/7-AAD. E GSEA of the DEGs identified in A. Multiplex cytokine profiling of supernatants isolated from CAR-T cells cocultured with F HER2⁺ HeLa cells and G SKOV3 cells for 24 h at a ratio of 2:1 (E:T). Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

SMAD7 mediates crosstalk between the TGF-β and NF-κB signaling pathways in CAR-T cells

TGF-β signaling orchestrates target gene expression through the canonical SMAD2/3-dependent pathway and noncanonical pathways, such as TAK1-mediated activation of the NF-κB pathway, which is abrogated by SMAD7 (Fig. 3A) [32, 33]. We then investigated the mechanism by which SMAD7 inhibits TGF-β signaling and regulates cytokine secretion in CAR-T cells. Unlike conventional CAR-T cells, which showed a significant upregulation of TβRI when exposed to increasing concentrations of rhTGF-β1, those coexpressing SMAD7 exhibited a decline in TβRI levels in response to rhTGF-β1 treatment (Fig. 3B). While rhTGF-β1 dramatically improved the phosphorylation of SMAD2/3 in conventional CAR-T cells, this effect was not observed in cells coexpressing SMAD7 (Fig. 3C), suggesting that SMAD7 disrupted the positive reinforcement of TGF-β signaling in CAR-T cells. We next exploited the effect of SMAD7 on the NF-κB pathway in CAR-T cells and found that SMAD7 significantly reduced the phosphorylation of the NF-κB pathway components TAK1, IκB and p65 when unstimulated or stimulated with HER2-conjugated beads (Fig. 3D). Notably, SMAD7 induced the upregulation of background IκB levels before antigen stimulation, which is consistent with a previous report that SMAD7 could directly induce the expression of IκB and consequently suppress NF-κB activation [34]. We further analyzed the subcellular location of p65 using confocal microscopy and observed that SMAD7 expression in CAR-T cells impaired antigen crosslink-triggered nuclear translocation of p65 in CAR-T cells (Fig. 3E, Fig. S5A). In accordance with these results, RNA-seq and qPCR assays indicated that the production of cytokines was significantly decreased in SMAD7-expressing CAR-T cells compared with control CAR-T cells after stimulation with HER2⁺ tumor cells (Fig. 3F, Fig. S5B). These data suggest that SMAD7 impedes cytokine synthesis in CAR-T cells.

Fig. 3.

SMAD7 mediates cross-talk between the TGF-β signaling pathway and the NF-κB signaling pathway. A Schematic showing the involvement of SMAD7 in the TGF-β and NF-κB signaling pathways. CAR-T cells were incubated with the indicated doses of rhTGF-β1, and the expression of TβRI was determined via FCM, plotted B, permeabilized and subjected to FCM assay of phosphorylated SMAD2/3 C. CAR-T cells were stimulated with (sti) or without (unsti) HER2+ beads for 15 min, followed by Western blot analyses D or immunofluorescence staining E for the indicated proteins. Hoechst (blue), p65 (green), Scale bars, 5 μm. F CAR-T cells were stimulated with (sti) or without (unsti) HER2+ tumor cells, followed by qPCR validation of the differential expression of the indicated cytokines as revealed by RNA-seq (see also Fig. 2A). Data are presented as the mean ± SD, and the S7H28z CAR-T treatment that differed significantly from H28z treatment is denoted by asterisks above the box. Data were analyzed by one-way ANOVA with multiple comparison of LSD text or independent-samples T test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

SMAD7 contributes to the maintenance of CAR-T cell cytolytic activity during continuous antigen exposure

We next stimulated CAR-T cells with target cells for continuous antigen exposure (CAE) in the presence of different concentrations of rhTGF-β1 and examined the cytolytic capability of these CAR-T cells (Fig. 4A, Fig. S6A). The cytotoxicity of H28z CAR-T cells to HER2⁺Fluc⁺ PC9 cells was dramatically reduced after several rounds of CAE, which was exacerbated by increasing concentrations of rhTGF-β1 (Fig. 4B). However, CAR-T cells expressing SMAD7 exhibited a remarkably retarded decrease in cytolytic capacity during CAE, which was not significantly affected by rhTGF-β1 (Fig. 4C, D, Fig. S6B). CAR-T cells subjected to CAE were then cocultured with HER2⁺Fluc⁺ HeLa cells that produce relatively high levels of TGF-β1. As a result, we observed that S7H28z CAR-T cells showed higher antitumor ability and lower expression of exhaustion markers (including PD-1, TIM-3 and LAG-3) than H28z CAR-T cells (Fig. 4E, S6C). In addition, we found that the ratios of the effector subsets were significantly higher in SMAD7-expressing CAR-T cells after 7 rounds of CAE and subsequent coculture with HER2⁺Fluc⁺ HeLa cells (Fig. 4F, Fig. S6D). Together, these results suggest that SMAD7 contributes to the sustained cytolytic activity of CAR-T cells during CAE, probably by suppressing TGF-β signaling and thereby minimizing exhaustion and extending cytolytic activity in T cells.

Fig. 4.

SMAD7-expressing CAR-T cells remain highly toxic against target cells during CAE. A Scheme of continuous antigen exposure (CAE) of CAR-T cells with HER2⁺Fluc⁺ PC9 cells in the presence of 0, 5, and 10 ng/mL rhTGF-β1. B, C CAR-T cells were subjected to CAE via coculture with HER2⁺Fluc⁺ PC9 cells in the presence of rhTGF-β1, and the cytolytic activity of cells was examined through measurement of the luciferase activity in target cells. D Representative images of HER2⁺Fluc⁺ PC9 cells after 7 rounds of coculture with CAR-T cells in different concentrations of rhTGF-β1. CAR-T cells endured CAE cocultured with HER2⁺Fluc⁺ HeLa cells at an E/T ratio of 2:1 for 24 h. Cells were then subjected to FCM assay for the indicated T-cell exhaustion markers E or for the proportions of the indicated subsets F. The phenotypes for T-cell subsets were CD45RA⁺CD62L⁺ (T_N), CD45RA^-CD62L⁺ (T_CM), CD45RA^-CD62L^- (T_EM) and CD45RA⁺CD62L^- (T_EFF). Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

CAR-T cells expressing SMAD7 display high antitumor efficacy and low systemic toxicity in NSG mice

We assessed the in vivo antitumor performance of CAR-T cells using a xenograft model. NSG mice were inoculated subcutaneously with 5 × 10⁵ HER2⁺Fluc⁺ HeLa cells and treated with CAR-T cells 7 days after inoculation (Fig. 5A). A Bioluminescence (BLI) assays revealed that both H28z and S7H28z CAR-T cells could efficiently suppress tumor development (Fig. 5B, C). Notably, we observed that the mice treated with H28z CAR-T cells frequently died despite the elimination of tumors, which was in line with substantially decreased body weight and high levels of cytokines in the serum (Fig. 5D–F). In contrast, mice treated with S7H28z CAR-T cells survived with undetectable tumors and remarkably alleviated body weight loss (Fig. 5B–E). In addition, mice treated with SMAD7-expressing CAR-T cells had much lower levels of cytokines in the serum than those receiving treatment with conventional CAR-T cells (Fig. 5F). Therefore, the expression of SMAD7 improves the safety of CAR-T-cell therapy without affecting the efficacy of tumor eradication.

Fig. 5.

CAR-T cells expressing SMAD7 provide superior protection in tumor-challenged mice. A Schematic diagram of inoculation of NSG mice and treatment with infused CAR-T cells. Mice in A were subjected to bioluminescent imaging on the indicated days postadministration of CAR-T cells B and the total flux of tumors in treated mice was plotted C. The body weight and the survival of mice in each group were recorded, and Kaplan‒Meier survival curves were plotted (n = 5 per group) (D, E). F Measurement of cytokine levels in the serum of mice 10 days after injection with CAR-T cells. Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test, and survival curves were analyzed by log-rank (Mantel‒Cox) test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

SMAD7 expression endows CAR-T cells with persistent antitumor efficacy by preventing TGF-β-triggered exhaustion

To simulate the microenvironment of solid tumors in the context of CAR-T-cell therapy, we inoculated NSG mice with HER2-overexpressing malignant cells as described above and treated mice with a relatively low number of CAR-T cells (Fig. 6A). A bioluminescence (BLI) assay revealed that the administration of S7H28z CAR-T cells, but not H28z CAR-T cells, completely eliminated the tumor within 5 weeks posttreatment (Fig. 6B, Fig. S7A). While mice in all groups displayed comparable body weight, survival and serum hTGF-β levels, probably as a result of the inoculated tumor cells (Fig. S7B, C, Fig. 6C), we found a pronounced higher proportion of human T cells (hCD3⁺) in the peripheral blood and tumors of mice receiving treatment with S7H28z CAR-T cells (Fig. 6D, E, Fig. S7D), suggesting that these novel CAR-T cells are resistant to exhaustion induced by cytokines in the microenvironment such as TGF-β. We then isolated human T cells from treated mice and analyzed the phenotypes of these cells. Although cells prepared from mice treated with S7H28z CAR-T cells consisted of higher percentages of naive and effector T cells and a lower ratio of effector memory subset (T_EM, CD45RA^-CD62L^-) on Day 14 posttreatment, this was almost reversed 28 days after administration of CAR-T cells, as exemplified by a higher proportion of the T_EM subset (Fig. 6F). Collectively, our data indicated that the expression of SMAD7 confers CAR-T cells with persistent antitumor efficacy, presumably by circumventing TGF-β-triggered exhaustion in the TME.

Fig. 6.

SMAD7-expressing CAR-T cells display persistent antitumor efficacy in vivo despite a high TGF-β level in the TME. A Schematic diagram of inoculation of NSG mice and treatment with infused CAR-T cells. B Mice in A were subjected to bioluminescent imaging on the indicated days postadministration of CAR-T cells (n = 5 per group). Peripheral blood was collected from the mice described in A on Day 0 before treatment, and the concentration of hTGF-β1 in the serum of the mice was measured via ELISA C. CD3⁺ T cells were isolated and subjected to FCM assay D. E The tumors were separated to perform immunohistochemical staining for hCD3 Day 14 of treatment (n = 3 per group). Scale bars, 500 μm (4×), 50 μm (10×). F Mice were inoculated and treated as described in A, and hCD3⁺ T cells were isolated from the peripheral blood of mice on Days 14 and 28 posttreatment and subjected to FCM assay. Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test. Asterisks indicate a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

Coexpression of SMAD7 improves the antitumor performance of CAR-T cells in PDOs

PDOs represent a three-dimensional model that closely recapitulates the architecture, functions, and genetic signature of clinical tumor tissues [35, 36]. Moreover, PDOs allow the rapid evaluation of treatment efficacy [37, 38]. To further verify the antitumor efficacy of SMAD7-expressing CAR-T cells, we incorporated these CAR-T cells in ovarian cancer PDOs (Fig. 7A) [39]. We collected HER2⁺ ovary cancer specimens for the generation of PDOs (Fig. S8A) and found that most of these patients had high concentrations of TGF-β1 in the serum or tumor tissues (Fig. S8B, C). We also verified TGF-β1 accumulation and immune cell infiltration (T cells, B cells and macrophages) in PDOs after 5-7 days of culture (Fig. 7B, C, Fig. S8D). CAR-T cells were then added to the PDOs and cocultured for 48 h. As a result, we found a significantly higher density of cells undergoing apoptosis in PDOs cocultured with SMAD7-coexpressing CAR-T cells, which was in accordance with a more abundant PDO infiltration of these cells than the conventional CAR-T cells (Fig. 7D, Fig. S8E). Then, CAR-T cells were isolated from the PDOs for FCM analysis. We found that SMAD7 coexpression significantly increased the proliferation and decreased the apoptosis of isolated CAR-T cells and downregulated the expression of exhaustion markers in CAR-T cells (Fig. 7E–G). A previous study reported that a memory phenotype is essential for the persistent antitumor capacity of T cells [40]. Accordingly, we found a pronounced higher proportion of the T_CM (CD45RA^-CD62L⁺) subset in S7H28z CAR-T cells than in H28z CAR-T cells after coculture with PDOs (Fig. 7H). Consistent with the in vitro cytokine release assays (Fig. 2F, G), the production of cytokines such as IL-6 and IL-8 was significantly reduced in S7H28z CAR-T cells compared with H28z CAR-T cells cocultured with PDOs (Fig. 7I). These data suggest that SMAD7 coexpression in CAR-T cells promotes tumor regression while reducing cytokine production in models that mimic human solid malignancies.

Fig. 7.

SMAD7 promotes CAR-T cell antitumor efficacy and reduces cytokine release in PDOs. A Schematic diagram of the establishment of ovary cancer PDO models and treatment with CAR-T cells. B ELISA determination of TGF-β1 secretion by PDOs. C FCM assays for identifying the clusters of various immune cells in PDOs. D Representative images showing the location of CAR-T cells (CytoTell Red 650-labeled) in OC PDOs. Cells were also stained for cleaved caspase 3/7 (green). Scale bars, 20 μm. FCM assays for CAR-T cells isolated from PDOs. Staining for T cells (CytoTell Red 650, E), apoptotic cells F biomarkers of exhausted cells G and cells with varied phenotypes H was performed. I Multiplex cytokine profiling of supernatants prepared from PDOs cocultured with CAR-T cells for 48 h. Data were analyzed by one-way ANOVA with multiple comparison of LSD test or independent-samples T test. Asterisk indicates a statistically significant difference. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001

Discussion

T-cell exhaustion is a key limitation of CAR-T-cell therapy for solid tumors [41–43]. The complex tumor microenvironment (TME) represents a strong barrier to the successful treatment of solid tumors by CAR-T cells. TGF-β, as one of the predominant cytokines in the TME, has been proven to induce systemic immune suppression and impede immune surveillance against tumorigenesis. However, the effect of targeting the TGF-β signaling pathway on CAR-T-cell therapy for solid tumors still lacks in-depth study [44]. In the current study, we constructed CAR-T cells that coexpressed SMAD7, a negative regulator of TGF-β signaling, and found that these CAR-T cells displayed persistent antitumor efficacy during continuous antigen exposure largely due to improved resistance to exhaustion induced by the immunosuppressive TME. In addition, a previous study demonstrated that exhausted T cells express a series of inhibitory receptors and experience phenotypic alterations [45, 46]. Our present study proved that these inhibitory receptors were significantly downregulated by the introduction of SMAD7 into CAR-T cells during continuous exposure to antigens. Furthermore, we found that T-cell populations with a memory phenotype were increased in mice receiving treatment with CAR-T cells coexpressing SMAD7, suggesting that SMAD7 could improve the persistence of CAR-T-cell therapy. Moreover, a recent study reported that overexpression of SMAD7 with a 4-1BB costimulatory domain increased the antitumor function of CAR-T cells and exhibited higher levels of cytokine secretion [47]. We expressed SMAD7 in CAR-T cells employing the CD28 costimulatory domain, which resulted in a remarkable decline in cytokine release upon priming by neoplastic cells and thereby minimized the systemic toxicity of CAR-T-cell therapy. These findings suggest that the TGF-β pathway might be an ideal target for optimizing CAR-T-cell therapy as its repression simultaneously prevented T-cell exhaustion and excessive cytokine release.

TGF-β signaling is known to orchestrate the survival, activation and exhaustion of immune cells. The limited proliferation of CAR-T cells is one of the major obstacles to their clinical application in solid tumors [48, 49]. Several strategies to modulate the TGF-β pathway in immunotherapy have been developed, including the inhibition of TGF-β synthesis or blockade of interactions in intracellular signal transduction [50, 51]. Here, we showed that expressing SMAD7 in CAR-T cells is sufficient to impede TGF-β signaling, as revealed by the dramatically reduced phosphorylation of SMAD2/3. In addition, we also observed a significant downregulation of TβRI, corresponding to the reported role of SMAD7 in ubiquitination and proteasomal degradation of the TGF-β receptor [52].

Cytokine release syndrome, a potentially fatal adverse effect associated with CAR-T-cell therapy, is another major obstruction to the clinical application of CAR-T-cell therapy [53]. CRS is a systemic disease characterized by an abnormal elevation of cytokine levels from infused CAR-T cells [36, 54]. Apart from the canonical TGF-β/R-SMADs signaling pathway, TGF-β has been reported to activate noncanonical pathways such as the TNF receptor–associated factors (TRAFs)-TAK1 axis, which fine-tunes TNF-α-triggered activation of the NF-κB pathway and inflammatory cytokine synthesis [55]. These results are consistent with previous findings that show that SMAD7 not only induces IκB expression and inhibits its phosphorylation and degradation [34, 56] but also binds to the adaptors TAB2/TAB3 to block recruitment of TAK1 to the TNF receptor signalosome [30, 57]. Overall, we determined that SMAD7 in CAR-T cells impairs cytokine production by inhibiting the NF-κB pathway. Although the multifaceted roles TGF-β plays in carcinogenesis (e.g., blocking cell cycle progression and counteracting the inflammatory response) have prevented the systemic targeting of this essential pathway in cancer therapy, we demonstrated that specific expression of SMAD7 in CAR-T cells can prevent T-cell exhaustion and cytokine release by suppressing TGF-β signaling. However, further investigations are needed to determine whether inhibition of this pivotal pathway affects other properties of CAR-T cells, such as their infiltration into solid tumors and homing to secondary lymphoid organs [58].

The engagement of T-cell receptors, or chimeric receptors in the case of CAR-T cells, by specific antigens provides the initial and principal signals for T-cell activation. Antigen/receptor crosslinking induces the activation of a class of proximal kinases, including Lck and ZAP70, which subsequently connects to multiple canonical intracellular pathways involved in the release of both cytotoxic granules and inflammatory cytokines. However, the dysfunction of infused CAR-T cells, such as T-cell exhaustion, almost inevitably occurs after a primary antigen-primed response, which is induced by tonic signals or immunosuppressive factors in the TME. In addition, the phenotypic shift of CAR-T cells, particularly their differentiation into memory subsets, also dictates the outcome of CAR-T-cell therapy for solid tumors. While we demonstrated that the ablation of TGF-β signaling by SMAD7 mitigated CAR-T-cell exhaustion and facilitated memory T-cell differentiation, additional investigation is needed to determine how SMAD7 interacts with the above signaling events that control T-cell activation, exhaustion and differentiation. Nonetheless, the intensity of TGF-β signaling, especially the level of TβRI, was reported to act as a crucial criterion to determine T-cell quiescence and activation [26]. TGF-β signaling also plays essential roles in Treg cell differentiation through the regulation of FOXP3 expression [31]. Therefore, detailed studies including those utilizing single-cell RNA sequencing are needed to decipher the alterations in T-cell subpopulations and the remodeling of the TME upon expression of SMAD7 in CAR-T cells. Collectively, we established that artificial expression of SMAD7 in CAR-T cells prevented exhaustion upon continuous antigen exposure and reduced cytokine release through the concomitant inhibition of canonical TGF-β signaling and crosstalk with the NF-κB pathway. Thus, our study demonstrated the feasibility of targeting TGF-β signaling as a novel approach to improving the efficacy and safety of CAR-T-cell therapy for solid tumors.

Materials and methods

Cell lines and culture

The cells used in this study were all purchased from American Type Culture Collection (ATCC) and cultured in a humidified incubator at 37 °C with 5% CO₂. Human embryonic kidney 293 T cells were maintained in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS) (Lonza, Hopkinton, MA, America) and 1% L-glutamine and were used for packaging of the recombinant lentiviruses. Human cervical cancer HeLa cells and human lung cancer PC9 cells were transduced with lentivirus to express human HER2 antigen and firefly luciferase (Fluc). HeLa cells and human breast cancer MCF7 cells were cultured in DMEM (Gibco), PC9 cells were cultured in RPMI 1640 (Gibco), and human breast cancer SKBR3 cells and human ovary cancer SKOV3 cells were cultured in McCoy’5 A (Gibco) supplemented with 10% FBS.

Generation of lentiviral vectors

A mammalian expression plasmid (pCMV3) containing SMAD7 was purchased from Sino Biological (Beijing, China). Then, SMAD7 with EcoRI (GAATTC) and XbaI (TCTAGA) restriction endonuclease sites was generated from pCMV3-SMAD7 using polymerase chain reaction (PCR), and the primers were designed using Primer Premier 6.0 software and synthesized by TSINGKE Biological Technology Corporation (Xi’an, China). The primer sequences were as follows: Forward: 5′-GGCGAATTCGCCACCATGTTCAGGACCAAACGATCTG-3′, Reverse: 5′-CTCTCTAGACCGGCTGTTGAAGATGACCTCTAG-3′. The cDNA of SMAD7 was subcloned and inserted into a modified pLVX-EF1α-IRES-Puro CAR-containing vector. We constructed a group of CARs using anti-HER2 scFv (P1h2), anti-CD19 scFv (FMC63) or anti-MSLN scFv (SS) linked to a CD8α stalk, a costimulatory domain (CD28), and a CD3ζ intracellular domain (H28z, 1928z, M28z). The coding sequence of SMAD7 was fused to a T2A element and the 5′-end of the open reading frame of the CAR, which eventually achieved coexpression of the 2 proteins after intraribosomal cleavage of the translated polypeptide at the T2A site.

Lentivirus production

HEK293T was used for lentivirus production with pMD2. G and psPAX2 packaging systems. Then, the lentiviral supernatant containing S71928z, H28z, S7H28z, M28z, or S7M28z recombinant lentiviruses was collected at 48 and 72 h after transfection and concentrated with 3 M NaCl and PEG 6000. The viral titer in transduction units per milliliter was determined by real-time quantitative PCR (qPCR).

Primary human T-cell isolation and CAR-T-cell generation

All donors provided informed written consent, and the use of human peripheral blood was approved (Approval ID: KY20214016-1) by the Ethics Committee of Fourth Military Medical University (Xi’an, China). Human peripheral blood mononuclear cells (PBMCs) were prepared from healthy donors using Human Lymphocyte Separation Medium (Cat # 7111011, Dakewei, China). Primary human T cells were isolated from PBMCs using the MojoSort^TM Human CD3 T-Cell Isolation Kit according to the manufacturer’s instructions (Cat # 480131, BioLegend). After stimulation with activation/expansion CD3/CD28 beads (Cat # MBS-C001, ACROBiosystem, China) for 24 h, T cells were infected with the aforementioned recombinant lentiviruses. T cells were resuspended in lentiviral supernatant and centrifuged for 1.5 h at 800 g. Cells were then cultured in X-VIVO15 medium (Cat # 204-448Q, Lonza, Switzerland) supplemented with 10% FBS, 1% penicillin & streptomycin sulfate and 100 IU/mL recombinant human IL-2 (Cat # 200-02-100, Peprotech, America).

Analysis of infection efficiency

Quantitative PCR was performed to evaluate the multiplicity of infection (MOI) of CARs. Briefly, 7 days after infection, CAR-T cells were collected for genomic DNA (gDNA) extraction (1 × 10⁶ cells per group) using a TIANamp Genomic DNA kit following the manufacturer’s protocol (Cat # DP304-02, TIANGEN, China). Then, the expression of the universal WPRE sequence was used to evaluate the efficiency of CAR-T cells. The primer sequences were designed using Primer Premier 6.0 software and synthesized by TSINGKE Biological Technology Corporation (Xi’an, China). The primer sequences are listed in Table S1.

Quantitative PCR (qPCR)

qPCR was used to measure the expression of cytokines. NT, S71928z, H28z, and S7H28z CAR-T cells were collected after coculture for 16 h with HER2⁺ target tumor cells at an effector to target (E/T) ratio of 5:1. Total RNA was extracted using RNAiso Plus (Cat # 9108, Takara, Japan), and cDNA was synthesized using PrimeScript™ RT Master Mix (Perfect Real Time) (Cat # RR036A, Takara, Japan). The primer sequences of these cytokines were designed and are listed in Table S1.

ELISA

ELISA was used to measure endogenous TGF-β1 secreted by tumor cell lines (including HeLa, PC9, SKBR3, SKOV3 and MCF7). Briefly, cells were starved in 0.5% FBS for 24 h (1 × 10⁶/well/cell line), and the supernatants of these cells were harvested for measurement. A Human TGF-beta 1 ELISA Kit (Cat # KE00002, Proteintech, China) was used to quantify the level of TGF-β1 in the supernatant of different tumor cell lines according to the manufacturer’s protocol.

Flow cytometry

Flow cytometry was performed using a CytoFLEX instrument (Beckman Colter, America). The expression of CARs was determined by incubation of cells with a biotinylated rhCD19 or rhHER2 protein (Cat # 11880-H08H-B, Cat # 10004-HCCH-B, Sino Biological), followed by incubation with APC-streptavidin (Cat # 405207, BioLegend). In addition, the antibodies used in flow cytometry were as follows: Alexa Fluor^® 488 anti-human CD340 Antibody (erbB2/HER2, Cat # 324410, BioLegend), Alexa Fluor^® 647 Human Mesothelin Antibody (Cat # FAB32652R, R&D systems), FITC anti-human CD3 Antibody (Cat # 300305, BioLegend), PE/Cyanine7 anti-human CD4 Antibody (Cat # 300510, BioLegend), PE anti-human CD279 Antibody (PD-1, Cat # 379210, BioLegend), APC/Cyanine7 anti-human CD366 Antibody (TIM-3, Cat # 345026, BioLegend), APC anti-human CD223 Antibody (LAG-3, Cat # 369212, BioLegend), PE anti-human CD45RA Antibody (Cat # 304108, BioLegend), and APC/Cyanine7 anti-human CD62L Antibody (Cat # 304814, BioLegend). Moreover, PE/Cyanine7 Annexin V (Cat # 640951, BioLegend), its binding buffer (Cat # 422201, BioLegend), and 7-AAD Viability Staining Solution (Cat # 420404, BioLegend) were used.

To detect TGF-β receptor I, rabbit anti-TβRI antibody (Cat # ab235178, Abcam) and Alexa Fluor^® 488 Donkey anti-rabbit IgG (Cat # 406416, BioLegend) were used. Fixation buffer (Cat # 420801, BioLegend) was utilized to fix cells used for measurement of phospho-SMAD2/3 and FOXP3 levels. To measure phospho-SMAD2/3 levels, BD PhosFlow^TM PE mouse anti-SMAD2 (pS465/pS467)/SMAD3 (pS423/pS425) (Cat # 562586, BD Biosciences, America) was diluted in a phospho-protein staining buffer (True-Phos^TM Perm Buffer, Cat # 425401, BioLegend). For FOXP3 or TCF1 staining, the PE anti-human FOXP3 antibody (Cat # 320207, BioLegend) or PE anti-TCF1 antibody (Cat # 655207, BioLegend) was used with True-Nuclear^TM Transcription Factor Buffer Set (Cat # 424401, BioLegend) according to the manufacturer’s instructions. All data analyses were performed using FlowJo V10 (BD biosciences, America).

Western blot analysis

Equal numbers of CAR-T cells in each group were collected via centrifugation. Cell precipitates were vortexed after the addition of SDS‒PAGE loading buffer, followed by electrophoresis, transfer of proteins to nitrocellulose (NC) membrane, and incubation of the membrane in a blocking solution containing 3% BSA for 30 min. The membrane was then incubated with a primary antibody at 4 °C overnight and washed three times with 1 × TBST, followed by incubation with a secondary antibody at room temperature for 60 min. Then, the protein bands were visualized using the Efficient Chemiluminescence Kit (Proandy, China) and imaged using a miniCHEM imaging system (SinSage Technology Co., Ltd, China). The antibodies used for blotting were as follows: SMAD7 polyclonal antibody (Cat # 25840-1-AP, Proteintech), phospho-TAK1 (Thr 187) polyclonal antibody (Cat # 28958-1-AP, Proteintech), TAK1 polyclonal antibody (Cat # 12330-2-AP, Proteintech), phospho-NF-κB p65 (Ser 536) rabbit mAb (Cat # 3003, Cell Signaling Technology), NF-κB p65 (D14E12) XP^® rabbit mAb (Cat # 8242, Cell Signaling Technology), phospho-IκBα (Ser32) rabbit mAb (Cat # 2859, Cell Signaling Technology), IκBα (L35A5) mouse mAb (Cat # 4814, Cell Signaling Technology), HRP goat anti-mouse IgG (Cat # 15014, Proteintech), HRP goat anti-rabbit IgG (Cat # 15015, Proteintech), and GAPDH monoclonal antibody (Cat # 60004-1-Ig, Proteintech).

Immunofluorescence

Immunofluorescence staining was used to determine the localization of the NF-κB p65 subunit in CAR-T cells. Briefly, H28z and S7H28z CAR-T cells were stimulated by beads coated with HER2 antigen for 15 min, and then the CAR-T cells were collected. Cells were fixed and permeabilized with a True-Nuclear^TM Transcription Factor Buffer Set (Cat # 424401 BioLegend) according to the manufacturer’s instructions. For p65 staining, NF-κB p65 (D14E12) XP^® Rabbit mAb (Cat # 8242, Cell Signaling Technology) and Alexa Fluor^® 488 Donkey anti-rabbit IgG Antibody (Cat # 406416, Biolegend) were used. Then, CAR-T cells were resuspended in PBS, and images of the cells were acquired with a NikonA1R microscope (Nikon, Japan).

T-cell cytotoxicity assessment in vitro

Bioluminescence imaging (BLI) was used to evaluate the cytotoxicity of CAR-T cells. Briefly, HER2⁺Fluc⁺ tumor cells were seeded in a 96-well plate with a clear bottom at a quantity of 2 × 10⁴ cells/well. Four to six hours later, CAR-T cells were added to the plate at E/T ratios from 2:1 to 1:10. After 12 h, residual tumor cells in the plate were measured by adding D-Luciferin and potassium salt (Cat # AC19L012, LIFE iLAB BIO, China), and bioluminescence was recorded by Xenogen IVIS Lumina II (Caliper Life Sciences, America). Specific lysis was calculated as follows:

$$\mathrm{Specific}\mathrm{Lysis}\%=\frac{\mathrm{spontaneous}\mathrm{lysis}-\mathrm{experimental}\mathrm{lysis}}{\mathrm{spontaneous}\mathrm{lysis}}\times 100\%$$

Proliferation assay

HER2⁺Fluc⁺ HeLa cells or HER2⁺Fluc⁺ PC9 cells (1 × 10⁵) were seeded in a 12-well plate in the presence of 5 ng/mL recombinant human TGF-β1 (rhTGF-β1, Cat # 100-21, PeproTech, America). Then, CAR-T cells or nontransfected T cells were counted, and 1 × 10⁶ cells were added to each well. These T cells were counted every 3 days and stimulated by HER2⁺ tumor cells every 6 days.

Continuous antigen exposure (CAE) assays

HER2⁺Fluc⁺ PC9 cells were seeded in a 12-well plate at 2 × 10⁵ cells/well on the day preceding the addition of CAR-T cells. An E/T ratio of 1:5 was used, and the assays were performed in the absence or presence of 5 or 10 ng/mL rhTGF-β1. Three to four days later, the cocultures were suspended, the supernatants were collected and centrifuged at 800 × g for 5 min, and the supernatants were transferred into new tubes. CAR-T cells were resuspended in media containing equal volumes of supernatants and fresh X-VIVO15. The resulting CAR-T-cell suspension was counted and transferred into new HER2⁺Fluc⁺ PC9-coated plates to reach an E/T ratio of 1:5 for continuous coculture. The concentration of rhTGF-β1 was maintained at 0, 5 or 10 ng/mL during the whole CAE assays. This process was repeated for 7 rounds.

RNA-seq

H28z and S7H28z CAR-T cells were sorted by specific antigen proteins with streptavidin beads and then incubated with HER2⁺Fluc⁺ HeLa cells at an E/T ratio of 5:1 for 16 h. RNA extraction, PCR amplification, cDNA library construction and bioinformatics analysis were carried out by BIOMARK Technologies (Beijing, China).

Cytokine production assay

A total of 2 × 10⁵ CAR-T cells were cocultured with 1 × 10⁵ HER2⁺Fluc⁺ HeLa cells in media without IL-2 or other cytokines in a 96-well plate. The supernatant of the coculture media was collected after incubation for 24 h. Then, the supernatant was analyzed by a LEGENDplex^TM HU Cytokine Release Syndrome Panel (13-plex) w/VbP (Cat # 740930, BioLegend) according to the manufacturer’s protocol.

Evaluation of in vivo tumor regression by CAR-T cells

Four- to six-week-old NOD/scid/IL2rg^-/^- (NSG) mice were purchased from SPF Biotechnology Co., Ltd. (China) and used to generate human xenograft tumor models. The animal experiments were approved by the Animal Experiment Administration Committee of the Fourth Military Medical University (Approval ID: 20210565). The care and use of animals were reviewed and approved by the ARRIVE guidelines and the National Institutes of Health guidelines for the care and use of laboratory animals (NIH Publications No. 8023, revised 1978). Mice were injected subcutaneously (s.c.) with 5 × 10⁵ HER2⁺Fluc⁺ HeLa cells. Seven days later, the mice were injected with D-luciferin for tumor detection and were randomly divided into groups according to the tumor luminescent signals. Subsequently, NT, S71928z, H28z and S7H28z CAR-T cells were injected intravenously through the tail vein (i.v., 2 × 10⁶/mouse). Mice were s.c. inoculated with 5 × 10⁵ HER2⁺Fluc⁺ HeLa cells and i.v. administered of 2 × 10⁵ CAR-T cells and tumor suppression was observed. Tumor bioluminescence was acquired and analyzed using the IVIS Imaging System, and photon emission from tumor cells was expressed in photons per second. T cells from the peripheral blood were stained with anti-human CD3, CD45RA, and CD62L antibodies. The cytokines of mice were collected from serum and evaluated by a LEGENDplex^TM MU Cytokine Release Syndrome Panel (13-plex) w/VbP (Cat # 741024, BioLegend) according to the manufacturer’s protocol.

Immunohistochemistry

Tumor tissues were fixed in 4% paraformaldehyde at room temperature, embedded in paraffin wax and sectioned into 4 μm pieces. Endogenous peroxidase was then blocked by hydrogen peroxide (3%), and antigen retrieval was performed with sodium citrate buffer (0.01 M, PH = 6) (Cat # G1202-250ML, Servicebio). Samples were blocked in 1% bovine serum albumin for 30 min at room temperature and then incubated with the primary antibody at 4 °C overnight and the secondary antibody at room temperature for 1 h. Section images were acquired by a fluorescence microscope and camera system (OLYMPUS VS200). The antibodies used in IHC were as follows: anti-CD3 rabbit pAb (Cat # GB11014-100, Servicebio), anti-ErbB2 rabbit pAb (Cat # GB113971-100, Servicebio), and immunohistochemistry-specific goat anti-rabbit secondary antibody (Cat # G1213-100UL, Servicebio).

Patient-derived tumor organoid (PDO) establishment and cytotoxicity assessment

The volunteer patients were informed and fulfilled the ethical criteria included in the Institutional Review Board-approved protocols, and the use of patient tissues was approved (Approval ID: KY20213037-1) by the Ethics Committee of Fourth Military Medical University (Xi’an, China). Tumor tissues were obtained from patients undergoing ovarian tumor resection at the Department of Gynecology and Obstetrics, Xijing Hospital of Fourth Military Medical University. The tissues were subjected to immunohistochemical staining for HER2. In addition, TGF-β1 levels in the serum or tumor tissues of ovarian cancer patients were measured by ELISA. The fresh tissues were placed on ice and washed 3 times using DPBS and then dissected into pieces for digestion using collagenase type IV (type I). The samples were then centrifuged to collect the digested cell sediment and treated with RBC lysis buffer (Roche, 11814389001) in 3 times the cell suspension volume. Cells were washed with ADMEM/F12 (containing 1× Glutamax, 10 mM HEPES, Gibco) at 1000 rpm and resuspended in cold Cultrex growth factor reduced BME type 2 (Trevigen, 3533-010-02). Cell suspensions were then added to 24-well plates (Greiner, No. 662824) until the cells adhered to the plate, and then the culture medium (Supplemental Materials Table S2) was supplemented. The PDOs were cultured in humidified incubators with 5% CO₂ at 37 °C, and the media were changed every 2 days. CAR-T cells were introduced to the PDOs after culture for 5–7 days. T cells were stained with CytoTell™ Red650 (AAT Bioquest, No. 22255) in 1 μL per million cells before coculture, and tumor cells were stained with CellEvent™ Caspase-3/7 Green (Thermo Fisher Scientific, No. C10423) in 1 μL per milliliter medium. The images were recorded by an EVOS™ M5000 Imaging System (Thermo Fisher Scientific, No. AMF5000).

Statistical analysis

SPSS 23 (IBM, Chicago, USA) was used for statistical analysis. The data are expressed as the mean ± SD and were visualized by GraphPad Prism 8 (GraphPad Software, America). Significance was calculated by one-way ANOVA followed by LSD multiple comparison tests or independent-samples T tests, and the symbols indicate statistical significance as follows: ^* P ≤ 0.05, ^** P ≤ 0.01, and ^*** P ≤ 0.001. In addition, the log-rank (Mantel‒Cox) test was used to analyze the survival curves in vivo.

Author contributions

BLZ, BY and AGY conceived the study; BY and SXL designed the project and wrote the manuscript with inputs from SXL and RZ; BLZ, LTJ and BY provided the funding support; SXL, RZ, JL and YYW performed the majority of the experiments and data analysis; YJH, HD, XJZ, YTZ, PJW and RTM performed some of the experiments and contributed reagents, materials, and analysis tools; LTJ, AGY and BY revised the manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-023-01120-y.