A Pan-Histone Deacetylase Inhibitor Enhances the Antitumor Activity of B7-H3–Specific CAR T Cells in Solid Tumors Free

Chimeric antigen receptor (CAR) T-cell therapy has emerged as one of the most promising forms of immunotherapy (1). This strategy relies on the use of genetically engineered T lymphocytes that specifically recognize membrane-bound tumor antigens (TAs) and subsequently eliminate target cells in an HLA-independent manner (2). To date, CAR T cells have produced impressive clinical responses in patients with hematologic malignancies; these results led to the FDA approval of CD19 CAR T cells for chemotherapy-refractory/relapsed acute lymphoblastic leukemia and non-Hodgkin lymphoma (3). Unfortunately, the application of CAR T-cell therapy in patients with solid cancers has generated disappointing results to date. The lack of efficacy is likely caused by the immune escape mechanisms utilized by targeted cancer cells, the poor CAR T-cell trafficking to tumors, and the hostile tumor microenvironment (TME; ref. 4). The poor antitumor activity of CAR T-cell therapy against solid tumors has stimulated interest in the development of combinatorial immunotherapeutic strategies that will enhance its efficacy. Among these strategies, cancer cells have been treated with chemotherapy (5) and/or radiotherapy (6) to enhance their susceptibility to CAR T-cell activity. Histone deacetylases (HDACs) are enzymes that catalyze the removal of acetyl functional groups from the lysine residues of diverse protein targets in addition to histone proteins (7). They have been considered crucial targets in various diseases, particularly in cancer. HDAC inhibitors (HDACis) are small molecules that inhibit the activity of HDACs, thus increasing the expression of specific genes at the transcriptional level (8). In recent years, six HDACis have been approved by the FDA as antitumor drugs based on their abilities to (i) inhibit cell growth, (ii) induce apoptosis, and (iii) regulate DNA damage, reactive oxygen species production, and proteasome activity (9). Moreover, recent studies have suggested that systemic treatment with HDACis potentially enhances the host immune responses by counteracting some of the mechanisms normally utilized by malignant cells to evade the host’s immune system (10).

According to a recent case study by Fraietta and colleagues, the disruption of TET2, a master regulator of blood cell formation, increases the therapeutic efficacy and persistence of CD19 CAR T cells (11). This finding suggested an alternative strategy to enhance CAR T-cell therapy efficacy against solid tumors by increasing the potency of CAR T cells, without inhibiting cancer cells. Therefore, the TET2 modification represents a useful tool to improve CAR T-cell–based therapy and provided us with the rationale to inactivate it in our proposed CAR T-cell–based combinatorial strategy. However, whether the combinatorial treatment we have designed disrupts TET2 expression in CAR T cells is not known.

We selected the TA B7-H3, also known as CD276, as the target in our strategy. B7-H3 is an immune checkpoint molecule belonging to the B7 superfamily of immune checkpoint inhibitors. Its biological function currently remains unclear; however, its function as both costimulatory and/or coinhibitory molecule appears to be unique and depends on the immune response in which it is involved (12). Numerous studies have described B7-H3 overexpression in human malignancies, including melanoma (13), leukemia (14), breast cancer (15), prostate cancer (16), ovarian cancer (17), pancreatic cancer (18), and colorectal cancer (19). B7-H3 expression has been correlated with a poor prognosis and clinical outcomes in patients (20). Another attractive feature of B7-H3 is its expression on the tumor-associated vasculature and tumor-associated macrophages, but very limited distribution in normal tissues (20). Therefore, immunotargeting of B7-H3 will not only affect cancer cells expressing B7-H3, but also disrupt the structure of the TME and inhibit neoangiogenesis. These properties make B7-H3 an attractive candidate for antibody-based immunotherapy against solid tumors, and have prompted many groups to design novel immunotherapeutic strategies targeting B7-H3 expressed on the surface of malignant cells, such as antibody–drug conjugates, bispecific antibodies, and CAR T cells (20, 21).

The purpose of this study is to assess the effect of the HDACi vorinostat (SAHA) on (i) B7-H3 expression on several solid cancer type cells, (ii) CAR expression and activation of multiple intracellular signaling pathways in human B7-H3.CAR T cells, and (iii) antitumor activity of B7-H3.CAR T cells toward several types of solid cancers by performing in vitro coculture studies and in clinically relevant mouse models.

One hundred eighty-six triple-negative breast cancer (TNBC) samples, 298 head and neck squamous cell carcinoma (HNSCC) samples, 152 non–small cell lung carcinoma (NSCLC) samples, and 89 skin cutaneous melanoma (SKCM) samples collected at Sun Yat-sen Memorial Hospital (SYSMH), Sun Yat-sen University (SYSU, Guangzhou, China), between April 2007 and December 2013 were used for B7-H3 staining and analysis. In addition, tumor samples obtained from 47 patients with HNSCC and 39 patients with TNBC at SYSMH, SYSU, between May 2016 and November 2018 were used to establish patient-derived xenograft (PDX) models. Fifty and 70 mL of peripheral blood was obtained from healthy donors or patients HNSCC-6, HNSCC-19, HNSCC-23, TNBC-12, and TNBC-20 and patients from whom the TNBC-S3 and HNSCC-Y2 cell lines were established, to generate CAR T cells. The clinical features of these patients are provided in Supplementary Table S1.

A lentiviral supernatant was collected from 293T cells transfected with pLVX-EF1α-IRES-AcGFP-(CAR-B7H3 or CAR-CD19) plasmids plus gag/pol, env, and VSV-G as previously described (22) following a 24- and 48-hour incubation. Peripheral blood mononuclear cells were isolated from whole blood samples drawn from healthy donors or patients at SYSMH, using Ficoll density gradient separation (GE Healthcare) and stimulated for 72 hours with anti-CD3/CD28 T-cell activation Dynabeads (Life Technologies) at a 1:1 bead:cell ratio. T cells were then transduced with the lentivirus at a defined multiplicity of infection. After removing the supernatant, 5 × 10⁵ activated T cells were plated and centrifuged at 1,000 × g for 10 minutes. T cells were collected, expanded in TexMACS Medium (Miltenyi Biotec) supplemented with penicillin–streptomycin (100 U/mL; Life Technologies), and treated with IL2 (100 U/mL; Life Technologies) every 48 hours. Dynabeads were removed 9 or 10 days after isolation. Mock-transduced T cells were stimulated using the previously described rapid expansion protocol (23). On days 12 to 14, cells were collected for in vitro and in vivo experiments. Thawed T cells were cultured in complete T-cell medium supplemented with IL7 (10 ng/mL; PeproTech) and IL15 (5 ng/mL; PeproTech); the medium was replaced every 2 days. T cells were cultured in IL7/IL15-depleted medium for 2 days prior to use in in vivo studies. Each in vitro or ex vivo experiment was repeated with T cells from a different donor or patient; T cells isolated from different donors were never pooled.

The cytolytic activity of T cells was assessed using ⁵¹Cr assays as previously described (24). CAR T cells were incubated with the tumor targets at the indicated effector cell/target cell (E/T) ratios for 4 hours. The mean percentage of specific cell lysis in triplicate wells was calculated using the following formula: (test release - spontaneous release)/(maximal release - spontaneous release) × 100.

Cancer cells were seeded in 24-well plates at a density of 1 × 10⁵ cells/well. T cells were added to the cultures at different ratios (E/T ratios of 1:1; 5:1, or 10:1) without the addition of exogenous cytokines. The cells were analyzed on days 3 to 5 to measure residual cancer cells and T cells using FACS. T cells were identified by the expression of GFP and CD45, whereas cancer cells were gated on the positive expression of B7-H3.

Single-cell suspensions were generated from harvested PDX tumors using a previously described method to analyze tumor-infiltrating T cells (25); cultures were incubated overnight to remove live cancer cells via adhesion to plastic. Viable cells were separated via density gradient centrifugation and added to an enzyme-linked ImmunoSpot (ELISpot) plate. The number of IFNγ-producing T cells was determined with an IFNγ ELISpot kit (Dakewe, 2110001 and 2110005) according to the manufacturer’s protocol.

Chromatin immunoprecipitation (ChIP) assays were performed as previously described (26). Briefly, cells (5 × 10⁶) were washed with PBS and incubated with 1% formaldehyde for 10 minutes at room temperature. Cross-linking was stopped by incubation with 0.1 mol/L glycine for 5 minutes. Cells were then washed twice with PBS, incubated for 1 hour at 4°C in a lysis buffer, and sonicated into chromatin fragments with an average length of 500 to 800 bp, as assessed using agarose gel electrophoresis. Samples were precleared with Protein A agarose (Roche) for 1 hour at 4°C on a rocking platform. Then, 5 μg of specific antibodies was added, and the samples were rocked overnight at 4°C. The immunoprecipitated DNA was purified using a QIAquick PCR purification kit (QIAGEN) according to the manufacturer’s protocol. The final ChIP DNA was then used as a template for qPCR with the primers listed in Supplementary Table S2. ChIP-grade anti–ac-H3 (Abcam, ab47915) and anti-RNA polymerase II (Abcam, ab5131) antibodies were used in this study.

The luciferase assay was performed as previously described (27), with modifications. Briefly, pGL4-B7-H3-wild-type was obtained by cloning a 2,000-bp DNA fragment (-2,000 bases to the BRCA1 transcriptional start site) into the pGL4.20-Basic vector upstream of the luciferase reporter gene (Supplementary Table S2). The pGL4.20-derived reporter vectors were transfected into cells, and stable cell lines were obtained through puromycin selection for 2 weeks. The pRL-TK plasmid containing Renilla luciferase was cotransfected as a control. Luciferase activities were measured using the Dual Luciferase Reporter Assay Kit (Promega). The target effect is presented as the luciferase activity of the reporter vector with the target sequence relative to a reporter vector without the target sequence.

Flow cytometry was performed using antibodies specific for human B7-H3, CD3, CD4, CD8, CD19, CD45, CD45RA, CCR7, granzyme B, and perforin from BD Biosciences or Invitrogen. Isotype control antibodies were used according to the manufacturer’s instructions. B7-H3.CAR expression on human T cells was assessed using a recombinant human B7-H3 Fc chimeric protein (Bio-Techne, 1027-B3). Intracellular staining was performed utilizing cells pretreated with the Intracellular Fixation and Permeabilization Kit (eBioscience, 88-8824) according to the manufacturer’s instructions. Cells were subsequently acquired using multicolor flow cytometry (BD, FACSVerse) with BD FACSDiva software (BD Biosciences). For each sample, a minimum of 10,000 events were acquired, and data were analyzed using FlowJo v10 software.

Protein extracts were resolved through 8% SDS-PAGE, transferred to polyvinylidene difluoride membranes (Bio-Rad), sequentially probed with a human B7-H3–specific (Cell Signaling Technology, 14058, 1:1,000), ac-H3–specific (Abcam, ab4729, 1: 1,000), CD3-ζ–specific (Abcam, ab119827, 1: 1,000), or β-actin–specific (Proteintech) antibody followed by a peroxidase-conjugated secondary antibody (Proteintech), and visualized using chemiluminescence (GE).

B7-H3.CAR T cells were incubated at 37°C with or without 0.5 μmol/L SAHA for 5 days for mRNA profiling. Briefly, the RNA samples that passed quality control on a Nanodrop ND-1000 and Bioanalyzer 2100 were amplified and labeled using the Agilent Quick Amp Labeling Kit and hybridized to an Agilent whole-genome oligo microarray in Agilent SureHyb hybridization chambers. After hybridization and washing, the processed slides were scanned using an Agilent DNA microarray scanner according to the guidelines provided by Agilent Technologies. The .txt files were extracted from Agilent Feature Extraction software (version 10.5.1.1) and imported into Agilent GeneSpring GX software (version 11.0) for further analysis. The microarray datasets were normalized in GeneSpring GX using the Agilent FE one-color scenario.

For the PDX models, human TNBC tumor fragments were orthotopically implanted into the mammary fat pad of 8- to 10-week-old NOD-Prkdc^em26Cd52Il2rg^em26Cd22/Nju (NCG) mice, whereas HNSCC tumor fragments were implanted into the buccal submucosa of NCG mice. Approximately 30 to 60 days after tumor engraftment, if the volume of the P0 tumors was approximately 1,000 mm³ using the formula TV (mm³) = length × width² × 0.5, the P0 tumors were divided into 20 pieces and implanted into NCG mice to establish P1 tumors. At 28 days after the engraftment of the P1 tumors, CD19.CAR T cells or B7-H3.CAR T cells (1 × 10⁷ cells/mouse) were injected i.v. into mice with at least 15 tumors on ultrasound images or a physical examination; the mice in the combined treatment group also received 2.0 mg/kg SAHA (PBS vehicle), which was injected i.p. daily for 5 days. Fifty-eight days following P1 tumor inoculation, all mice were sacrificed, and tumors were collected for additional studies as indicated. For the metastatic models, 8- to 10-week-old female NCG mice were injected i.v. with the cell line Fluc-TNBC-S3 or Fluc-HNSCC-Y2 (1 × 10⁶ cells/mouse). Fifteen days following cancer cell inoculation, nontransduced T cells (NTs), CD19.CAR T cells, or B7-H3.CAR T cells (5 × 10⁶ cells/mouse) were injected i.v. (n = 6 mice per group). The mice in the combined treatment group also received 2.0 mg/kg SAHA (PBS vehicle), which was injected i.p. daily during the 5-day treatment cycles starting on days 15 and 45. Investigators were blinded to the groups in this study, and the tumor burden was monitored by bioluminescence imaging using an IVIS Lumina imaging system. Mice were euthanized when signs of discomfort were detected by the investigators or as recommended by veterinarians from a laboratory animal research unit who monitored the mice three times per week. Mice were sacrificed approximately 100 days following cancer cell inoculation, and tumor specimens, organs, and peripheral blood samples were collected from the mice for the indicated studies. To test the toxicity of SAHA, NCG mice (n = 6) received 2.0 mg/kg SAHA, which was injected i.p. daily during the 5-day treatment cycles starting on days 1 and 31. Mice were sacrificed on day 40, and tumors and organs were harvested for hematoxylin and eosin staining. Cell line–derived xenograft (CDX) models were generated using the SUM149, SCC-9, and M21 cell lines. SUM149 cells (1 × 10⁶ cells/mouse) were implanted into the mammary fat pad, SCC-9 cells (1 × 10⁶ cells/mouse) were implanted into the buccal submucosa, and M21 cells (2 × 10⁶ cells/mouse) were implanted s.c. into the flank of the NCG mice (n = 6 mice per group). The schedule of treatment was similar to that of the PDX models, and CAR T cells (1 × 10⁷ cells/mouse) were injected i.v. on day 7 with or without 5-day treatment with 2.0 mg/kg SAHA. Mice were sacrificed on day 37.

Mice were injected with D-luciferin (150 mg/kg, i.p. injection, 15 minutes before imaging), anesthetized (3% pentobarbital), and imaged with a Xenogen IVIS Lumina system (Caliper Life Sciences). Bioluminescent flux (photons/s/cm²/steradian) was calculated to monitor tumor migration.

Frozen biopsy samples of harvested PDX tumors were routinely cut into 4-μm-thick sections. Tissue sections were then mounted on glass slides and fixed with 4% paraformaldehyde for 3 minutes at room temperature. Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) staining was performed with the In situ Cell Death Detection Kit (Roche, 11684817910) according to the manufacturer’s instructions. Tissue sections were then washed twice with 0.01 mmol/L PBS, pH 7.4, incubated with a human B7-H3–specific mouse antibody (Santa Cruz Biotechnology, sc-376769, 1:200) overnight at 4°C, and subsequently stained with an Alexa Fluor 594–conjugated secondary antibody (Thermo Fisher Scientific, A-11012) for 1 hour at room temperature. DAPI was then used to counterstain nuclei, and images were acquired with an upright fluorescence microscope (PerkinElmer Vectra 2.1).

The patient study was conducted in accordance with the principles of the Declaration of Helsinki. All samples were collected from patients who had provided written informed consent, and all related procedures were performed with the approval of the Institutional Review Board (IRB) of SYSMH (approved protocol: 2016158). Normal human tissue samples were collected from Shanghai Outdo Biotech Co., Ltd. (XT16-012). The use of this material did not require IRB approval. For the mouse models, tumor samples, mouse organs, and peripheral blood samples were obtained under protocols approved by the Institutional Animal Care and Use Committee at Sun Yat-sen University.

All data are presented as mean ± SEM. Comparisons between two groups were performed with two-tailed Student t tests using the SPSS 18.0 package (SPSS). For multiple-group comparisons, one-way ANOVA or two-way ANOVA was used to determine statistically significant differences among samples. Kaplan–Meier survival curves were plotted, and a log-rank test was performed. Experimental sample numbers (n) are indicated in the figure legends, and statistical analysis methods are also described in individual figure legends. At least three independent experiments were performed for all cell culture experiments. Throughout the study, P values less than 0.05 were considered significant.

Data have been deposited in the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/gds) under the accession number: GSE13537.

First, we evaluated B7-H3 expression in solid cancer and normal tissue samples. Tumor samples obtained from 186 patients with TNBC, 298 patients with HNSCC, 152 patients with NSCLC, and 89 patients with SKCM at SYSMH, SYSU, between April 2007 and December 2013 were used for B7-H3 staining. As shown in Fig. 1A, Supplementary Fig. S1A, and Supplementary Table S3, the four types of tumors showed strong positive staining for B7-H3, with the antigen expressed by both cancer cells and surrounding stroma. We found that neoadjuvant chemotherapy (NCT) did not change B7-H3 expression in patients with HNSCC (Supplementary Fig. S1B and S1C). Furthermore, TNBC, HNSCC, and NSCLC cell lines showed no significant change in B7-H3 expression on the cell membrane following cisplatin treatment at half maximal inhibitory concentrations (Supplementary Fig. S1D). Moreover, normal human tissue microarrays showed that the brain, lungs, and heart were B7-H3–negative, whereas the bladder, skin, and rectum exhibited weak cytoplasmic staining in the epithelium or stroma. The stomach, colon, liver, pancreas, testis, and prostate showed positive staining, but the staining intensity was much lower than that in solid tumors (Fig. 1B and C; Supplementary Table S3). Our data are consistent with previous studies showing that B7-H3 is broadly expressed across many solid cancer types, but exhibits limited expression in normal tissues (20). Interestingly, The Cancer Genome Atlas (TCGA) RNA-sequencing data showed ubiquitously expression of the B7-H3 transcript at a higher level in a wide spectrum of human solid cancer types than in normal tissues (Supplementary Fig. S2A). A univariable Cox regression analysis indicated that patients with TNBC, HNSCC, and NSCLC with low B7-H3 expression levels had a longer overall survival (OS; Fig. 1D; Supplementary Tables S4–S9). Furthermore, a multivariable Cox regression analysis revealed that low B7-H3 expression is an independent prognostic marker for a prolonged OS in patients with TNBC, HNSCC, and NSCLC (Supplementary Tables S4–S9). Consistent with these findings, elevated B7-H3 expression correlated with relatively shorter OS for patients with several malignancies, such as glioblastoma multiforme, breast-invasive carcinoma, HNSCC, lung adenocarcinoma, bladder urothelial carcinoma, and SKCM in the TCGA database (Supplementary Fig. S2B). Thus, B7-H3 is not only a prognostic biomarker for solid cancers, but also an attractive target for cancer immunotherapy.

We generated a B7-H3.CAR using a single-chain variable fragment (scFv) derived from the humanized 8H9 mAb (28) and included CD28 and 4-1BB endodomains (Supplementary Fig. S3A). Third-generation CD19.CAR T cells were also used as controls in this study (Supplementary Fig. S3A). The transduction efficiency of activated T cells both from healthy donors and patients was generally greater than 60% (Supplementary Fig. S3B). An analysis of the expansion kinetics did not reveal significant differences between the two types of CAR T cells (Supplementary Fig. S3C). B7-H3.CAR T cells and CD19.CAR T cells were composed of central memory T cell, effector memory T cell, and T-memory stem cell populations (Supplementary Fig. S3D).

Because B7-H3 is broadly expressed on many solid cancer types, we investigated the antitumor activity of B7-H3.CAR T cells with two TNBC cell lines (SUM149 and TNBC-S3), two HNSCC cell lines (HNSCC-Y2 and SCC-9), one NSCLC cell line (A549), and one SKCM cell line (M21; Supplementary Fig. S3E). TNBC-S3 and HNSCC-Y2 are primary cell lines. We also utilized the B7-H3–knockout cell line SUM159-B7-H3 KO and the CD19-overexpressing cell line K562-CD19 as negative and positive controls, respectively, in the coculture system (Supplementary Fig. S3F). Then, we assessed the cytolytic activity of B7-H3.CAR T cells and found that B7-H3.CAR T cells could efficiently lyse cancer cells (Supplementary Fig. S4). Furthermore, the killing of cancer cells over 3 days significantly increased as the T-cell (effector cell) to tumor cell (target cell) ratio (E/T) in the coculture system increased (Supplementary Fig. S5A and S5B). The specificity of the cytolytic activity of B7-H3.CAR T cells was confirmed by the release of cytokines, including GMCSF, IFNγ, IL2, and TNFα, into the culture supernatant (Supplementary Fig. S5C).

Antigen heterogeneity in solid tumors is one of the major barriers to the therapeutic efficacy of CAR T cells (29–31). In fact, antigen-rich regions facilitate the recruitment of CAR T cells and interactions between CAR T cells and target cells (32, 33). HDACis have been shown to positively upregulate transcriptional activation (34). Therefore, we tested whether HDACis can increase B7-H3 expression on solid cancer cells. Considering the dose-limiting toxicities and other adverse events related to HDACis observed in clinical trials (35), we screened a number of HDACis, including entinostat, mocetinostat, SAHA, tubastatin A, and tucidinostat for their ability to upregulate B7-H3 expression on various types of cancer cells after treatment with a low dose. SAHA was the most effective HDACi at upregulating B7-H3 expression, as 0.5 μmol/L SAHA induced an approximately twofold increase in B7-H3 expression level but only produced negligible changes in cell viability (Fig. 2A; Supplementary Fig. S6A). The effect was both dose- and time-dependent (Supplementary Fig. S6B), but prolongation of the SAHA incubation time with cells did not cause a significant increase in B7-H3 expression after 5 days. The increase in B7-H3 expression on the cell membrane could last for at least 7 days following SAHA withdrawal. By day 14, the B7-H3 expression level returned to the original value (Fig. 2B). In addition, increased levels of the B7-H3 transcript and protein were also detected in SAHA-treated cells and were consistent with its upregulation on the cell surface (Fig. 2C and D). We performed ChIP-qPCR and luciferase reporter assays to further confirm that SAHA increases B7-H3 transcription and found that SAHA induced B7-H3 transcription in multiple types of cancer cells as indicated (Fig. 2E and F). We evaluated B7-H3 expression on various types of normal human cells treated with SAHA, including primary tongue muscle cells, primary stomach epithelial cells, primary duodenal cells, primary hepatocytes, primary pancreatic cells, and primary ovarian cells to prevent “on-target” toxicity by B7-H3.CAR T cells in normal tissues. SAHA did not increase B7-H3 expression on the cell surface of any human normal cells; B7-H3 was at most barely detectable on these types of normal cells (Supplementary Fig. S7).

We then tested whether SAHA enhances B7-H3.CAR expression on transgenic T cells. First, SAHA reduced the viability of human T cells; in this case, the effect was dose- and time-dependent, but treatment of human T cells with 0.5 μmol/L SAHA for 5 days did not produce significant damage (Fig. 3A). SAHA (0.5 μmol/L) increased the surface expression of B7-H3.CAR on T cells, as evidenced by a twofold increase in the level of the B7-H3 protein (Fig. 3B). However, the expression of GFP was not increased by the SAHA treatment (Fig. 3B). Despite the ability of SAHA to induce epigenetic changes, only a negligible increase in the CAR mRNA level was detected using qRT-PCR (Fig. 3C). Similarly, although the level of acetylated histone 3 (ac-H3) was increased, the levels of the wild-type and chimeric CD3-ζ chain were not (Fig. 3D), consistent with the evaluation of GFP expression (Fig. 3B). Together, these results argue in favor of the possibility that the SAHA-induced B7-H3.CAR upregulation reflected a posttranslational mechanism.

To assess the functional relevance of the B7-H3 and B7-H3.CAR upregulation induced by SAHA on cancer cells and transduced T cells, respectively, the antitumor activity of SAHA-treated B7-H3.CAR T cells against cancer cells or SAHA-treated cancer cells was tested in vitro. Therefore, we first performed a standard cytolytic assay and found that pretreated B7-H3.CAR T cells showed the greatest tumor-lytic potential against pretreated cancer cells, whereas pretreated CAR T cells or pretreated cancer cells exhibited equivalent extent of lysis (Fig. 4A). Then, B7-H3.CAR T cells were cocultured with cancer cells at an E/T ratio of 1:1 for up to 3 days. As shown in Fig. 4B and Supplementary Fig. S8, there was a significant increase in the elimination of target cells in the cocultures containing either pretreated B7-H3.CAR T cells or pretreated cancer cells compared with cocultures containing only untreated cells. Moreover, the highest antitumor activity of B7-H3.CAR T cells was observed in the coculture with SAHA-pretreated CAR T cells and -pretreated cancer cells. Consistently, there was a significant increase in the release of cytokines GMCSF, IFNγ, IL2, and TNFα into the culture supernatant (Fig. 4C). Together, these results indicate that SAHA sensitizes human solid cancer cells to B7-H3.CAR T-cell attack by upregulating B7-H3 expression and CAR expression.

To the best of our knowledge, the effect of epigenetic therapy on CAR T cells has not been systematically studied. Therefore, we performed mRNA profiling to detect transcriptomic changes in B7-H3.CAR T cells under SAHA treatment. As shown in Fig. 5A, the SAHA treatment markedly changed the transcriptome, with 6,172 significantly upregulated genes and 6,225 significantly downregulated genes. Then, we applied gene set enrichment analysis (GSEA) to identify the changed phenotype of CAR T cells. Because we found upregulation of CAR expression at the posttranslational level, we focused on the protein transport and immune-related signaling pathways. As expected, the GSEA revealed the upregulation of protein transport–related genes, whereas immunosuppression- and unfolded protein response (UPR)–related genes were downregulated (Fig. 5B). The corresponding associated genes included the upregulated genes GPC1-4 and ZG16 and downregulated genes CTLA-4 and TET2 (Fig. 5C; Supplementary Fig. S9). A detailed description of the gene sets, including upregulated Golgi lumen-related genes and downregulated UPR-related genes, PERK-regulated genes, and T-cell receptor signaling pathway genes, is provided in Fig. 5D.

Interestingly, TET2, a methylcytosine dioxygenase, was recently shown to be a potent regulator of CD19.CAR T-cell fate (11). Therefore, we investigated the potential role of TET2 in our model. Knockdown of TET2 expression in B7-H3.CAR T cells increased cytokine release (GMCSF, IFNγ, IL2, and TNFα) and the cytotoxic activity of B7-H3.CAR T cells in the coculture system (Fig. 5E–H). Consistently, TET2 knockdown in B7-H3.CAR T cells increased perforin and granzyme B expression (Fig. 5I). Thus, SAHA increases B7-H3.CAR expression on the T-cell surface possibly through the upregulation of the protein transport signaling pathway. In addition, SAHA influences critical components of the cytotoxic machinery in B7-H3.CAR T cells.

SAHA markedly enhanced the cytotoxic activity of B7-H3.CAR T cells in vitro by upregulating B7-H3 expression on solid cancer cells and B7-H3.CAR expression on the surface of human T cells (Fig. 5J). Then, we tested the efficacy of our combinatorial therapy in vivo. First, we found that a low dose of SAHA with two cycles of treatment had no significant toxicity in NCG mice (Supplementary Fig. S10A and S10B). The CDX models using SUM149, SCC-9, and M21 cell lines indicated that SAHA and B7-H3.CAR T-cell treatment significantly inhibited tumor growth (Supplementary Fig. S10C–S10E). Furthermore, PDX tumor models provide a faithful recapitulation of the histology, genetics, and transcriptome of a donor patient-derived tumor (36, 37). Based on our plans to translate the results obtained from animal models to a clinical setting, we established PDXs as a first step. We first established passage 0 (P0) orthotopic TNBC-PDXs and performed buccal engraftment of HNSCC-PDXs in NCG mice. When the P0 tumor volume was approximately 1,000 mm³, each tumor was divided into 20 parts and implanted into 20 mice to establish P1 models. The P0 success rates of the TNBC-PDXs and HNSCC-PDXs were 20.51% (8/39) and 23.40% (11/47), respectively (Fig. 6A). Then, we chose the mice with successful tumor growth for treatment with a single autologous B7-H3.CAR T-cell infusion and 5 consecutive days of SAHA treatment on day 28 if the success ratio of P1 was greater than 15, because five mice were required for each group (Fig. 6A). The tumor burdens of all mice were assessed using ultrasound measurements or physical examinations (Fig. 6A). Compared with previous studies (34), the administration of SAHA at the indicated low dose did not cause potential systemic side effects. B7-H3 expression was confirmed to be retained in vivo upon engraftment in P1 mice (Fig. 6B). Then, two TNBC-PDX models and three HNSCC-PDX models were selected for treatment. Compared with B7-H3.CAR T-cell transfer alone, the combinatorial therapy significantly controlled tumor growth by day 30 after the CAR T-cell infusion, the time point at which the experiment was terminated, because the mice treated with CD19.CAR T cells showed a high tumor burden (Fig. 6C and D; Supplementary Fig. S11A and S11B). Consistently, the combinatorial therapy–treated group also showed the highest percentage of apoptotic cancer cells, B7-H3 expression level, and tumor infiltrating T-cell number (Fig. 6E–H). Moreover, more infiltrating IFNγ-producing T cells were also identified in the SAHA and B7-H3.CAR T-cell–treated group than in the other groups, and this group also exhibited the lowest expression of the immunosuppressive molecules TET2 and CTL-4 (Fig. 6I and J).

We established metastatic TNBC and HNSCC models by infusing TNBC-S3 and HNSCC-Y2 cells into NCG mice via the tail vein to further investigate the therapeutic efficacy of B7-H3.CAR T cells on metastases. As shown in Fig. 6K, B7-H3.CAR T cells alone or in combination with SAHA controlled TNBC and HNSCC cell growth up to day 100 after tumor challenge, when the experiment was terminated (Fig. 6K and L). In contrast, CD19.CAR T cells and untransduced T cells did not exert a detectable effect on TNBC-S3 and HNSCC-Y2 cell growth. Notably, the mice treated with B7-H3.CAR T cells in combination with SAHA experienced significantly longer survival than the B7-H3.CAR T-cell–treated group (Fig. 6K and L). Furthermore, B7-H3.CAR T cells persisted in peripheral blood, bone marrow, and spleen until the day mice were sacrificed (Fig. 6M). Interestingly, no lesions were detected in tissue sections from the mice treated with the combination of B7-H3.CAR T cells and SAHA. The expression of mouse B7-H3 in normal tissues from the SAHA-treated group was similar to that in the normal tissues from the untreated mice (Supplementary Fig. S11C and S11D). Based on these results, SAHA enhances the antitumor activity of B7-H3.CAR T cells in vivo possibly by upregulating B7-H3 expression on cancer cells and increasing the cytolytic activity and recruitment of B7-H3.CAR T cells in the TME. The low dose of SAHA used is unlikely to cause organ toxicity or upregulate B7-H3 expression in normal tissues.

In this study, we have shown that epigenetic therapy enhances the antitumor activity of CAR T cells with solid tumors. We selected B7-H3 as the target and generated third-generation B7-H3.CAR T cells. We found that a low dose of SAHA upregulated B7-H3 expression on many types of solid cancer cells, including TNBC, HNSCC, NSCLC, and SKCM cells. Interestingly, SAHA could also increase B7-H3.CAR expression on the surface of transduced human T cells and downregulate the expression of immunosuppressive molecules, such as TET2 and CTLA-4. We established orthotopic PDX and metastatic xenograft models of TNBC and HNSCC to further validate the clinical relevance of this combinatorial therapy in vivo and found that low-dose SAHA significantly enhanced the therapeutic efficacy of B7-H3.CAR T cells.

The combination of various HDACis and types of immunotherapy shows synergistic effects that occur in different ways (38). PD-1 checkpoint blockade produces a limited response rate in patients with lung cancer but yields promising antitumor effects when combined with romidepsin, an FDA-approved HDACi. The underlying mechanism includes increased T-cell recruitment mediated by HDACi-induced T-cell chemokine expression by both cancer cells and T cells, increased cancer cell immunogenicity mediated by the HDACi, and upregulated IFNγ expression in T cells induced by the anti–PD-1 antibody (39). Another study identified a mechanism that depended on tumor cell apoptosis mediated by SAHA/panobinostat, which stimulated the uptake of dead cancer cells by antigen-presenting cells (APCs), and immune-activating bispecific antibodies designed to promote the function of APCs and enhance the proliferation and survival of cytotoxic T cells (40). Other rationales for designing combinatorial therapies include targeting mouse granulocytic MDSCs that cause resistance to PD-1 and CTLA-4 blockade (41) and inhibit apoptosis in CD4+ T cells within a tumor to enhance antitumor immune responses and suppress tumor growth (42). At present, several clinical trials which utilize combinatorial therapies are ongoing. Particularly, SAHA is one of the most used HDACis and is currently being evaluated in the clinical setting in combination with the immune checkpoint inhibitor pembrolizumab for the treatment of advanced renal and urothelial cell carcinoma (NCT02619253), NSCLC (NCT02638090), newly diagnosed glioblastoma (NCT03426891), and recurrent HNSCC (NCT02538510). However, dose-limiting toxicities and other adverse events related to HDACis have been reported in clinical trials (35), and the trial NCT00976183 was terminated because of toxicities. These results prompted us to use a low dose of SAHA in our experiments with the expectation that this approach will facilitate the clinical translation of the results obtained in in vitro experiments and in animal models.

The potential of CAR T-cell–based immunotherapy as a treatment option for cancer is indicated by some leukemia patients’ impressive responses to adoptive cell transfer of CAR T cells. However, the results of pilot clinical trials in solid cancers have not been very positive thus far. Several obstacles, including the limitations of the available TAs used as targets, inefficient trafficking of CAR T cells to tumor sites, hostile solid TME, and potential “on-target/off-tumor” toxicities, remain to be overcome for the successful application of CAR T cells in solid cancers (43).

It also remains challenging to select appropriate target antigens for the therapeutic development of safe and effective CAR T-cell–based therapies for solid cancers. Toxicities upon an infusion of CAR T cells have been observed when targeting molecules such as HER2 or CAIX, as these molecules are expressed in some normal tissues (37, 44). We and other groups (37, 45) have used different mAbs to stain normal human tissues/cell lines and several malignant tissues/tumor cell lines. Consistently, all the data have shown B7-H3 overexpression in many cancer types, but limited expression in normal tissues. The experiments in immunocompetent mice further illustrate the lack of on-target/off-tumor toxicity of B7-H3.CAR T cells; these results may reflect the lower B7-H3 density in normal tissues than in malignant tumors (37). On the other hand, the therapeutic activity of CAR T cells can be limited due to the rapid tumor escape that occurs when a targeted antigen shows heterogeneous expression or a low density within a tumor (32, 33, 37, 46), as was recently reported in patients with glioblastoma treated with EGFRvIII-specific CAR T cells (37, 46). Therefore, targeting multiple antigens or increasing the antigen density can potentially overcome this obstacle, but toxicities related to the recognition of antigens expressed by normal tissues remain a major concern. Unlike most conventional CARs that use a scFv derived from a mouse mAb, our B7-H3.CAR was derived from the humanized mAb 8H9 (28), and thus it would not be expected to induce human anti-mouse antibody or T-cell responses, which may impair therapeutic efficacy in humans (47). The B7-H3.CAR T cells in our study displayed significant antitumor activity against many solid cancer cells and xenografts. Importantly, we found that a low dose of SAHA enhanced B7-H3 transcriptional activity and the B7-H3 density on cancer cells. The cytotoxic activity of B7-H3.CAR T cells was significantly increased by SAHA in vitro, and the xenograft models also showed that our combinatorial strategy increased the recruitment of B7-H3.CAR T cells to tumor sites with superior therapeutic efficacy. Notably, previous studies showed that the signaling pathways of cancer cells or stromal cells, such as apoptosis, metabolism, and antigen presentation, may be regulated by SAHA (40, 48). Therefore, we cannot exclude that these changes contribute to the synergistic antitumor activity of our combinatorial strategy. It is also noteworthy that the low-dose SAHA treatment did not produce any detectable change in human B7-H3 expression on normal human cells or mouse B7-H3 expression in vital mouse organs, suggesting that the combination treatment would not induce “on-target/off-tumor” toxicities in the clinical setting.

To our surprise, SAHA at a low dose increased the membrane expression of B7-H3.CAR on transduced T cells. However, no SAHA-induced change in the transcriptional level and total cellular protein level of the CAR was detected. These findings are consistent with the possibility that SAHA may play a role in the posttranscriptional regulation of B7-H3.CAR expression. Based on mRNA profiling, GSEA results showed that UPR and immunosuppressive signaling pathways are downregulated, whereas the protein transport pathway is upregulated in SAHA-treated–transduced T cells. Although the process involved in the modification and transport of cellular proteins to the cell surface is complicated (49), we could partially explain the increase in B7-H3.CAR expression by upregulation of the Golgi lumen pathway, although other potential signals could not be excluded and require further investigation. Pretreatment with low-dose SAHA could significantly enhance the elimination of cancer cells by B7-H3.CAR T cells. We also found that CTLA-4 and TET2 levels were both decreased by SAHA in B7-H3.CAR T cells. TET2 encodes a methylcytosine dioxygenase and has been recently reported to be a master regulator of T-cell fate (11). Knocking down TET2 expression in B7-H3.CAR T cells increased granzyme B and perforin levels as well as their cytotoxic activity; these results are in agreement the information in the literature (11). Generally, approximately equal numbers of genes are up- and downregulated after HDACi treatment (50); the results of our study are relatively consistent with these findings (Fig. 5A). The mechanisms by which HDACis repress the expression of specific genes have been investigated to a limited extent. However, HDACis mediate the transcriptional repression of specific genes in a STAT5-or Sp1/Sp3-dependent manner (50, 51). Additional studies are needed to elucidate the mechanism underlying CTLA-4 and TET2 repression by SAHA.

To translate our preclinical findings to a clinical setting, we utilized PDX models to assess the therapeutic efficacy of the combination of SAHA and B7-H3.CAR T cells. Barriers that remain to be overcome to achieving effective CAR T-cell therapy include the antigenic heterogeneity of solid tumors and the immunosuppressive microenvironment that limit T-cell infiltration of tumors. Therefore, we utilized orthotopic TNBC-PDX and HNSCC-PDX models and found that the combination of SAHA and autologous B7-H3.CAR T-cell infusion improves tumor control; furthermore, metastatic models showed that the described combinatorial therapy prolonged the survival of mice. We also detected no organ lesions in the metastatic models after the mice received two B7-H3.CAR T-cell infusions and two cycles of SAHA treatment at a low dose. Even in these models, no side effects caused by B7-H3.CAR T cells could be detected in the present study; our data show the clinical relevance of the combination of low-dose SAHA with B7-H3.CAR T cells. Several clinical trials are targeting B7-H3, utilizing mAb (NCT02628535, NCT02982941, NCT02923180, NCT01502917, NCT03275402, NCT01099644, NCT01391143, and NCT03729596) and CAR T-cell approaches for recurrent or refractory glioblastoma (NCT04077866). Although we cannot exclude potential toxicities, B7-H3.CAR T cells derived from the mAb 376.96 (37) control PDAC tumor growth in immunocompetent mice, and the mAb m276 (37, 52) or a 131 I-conjugated anti-B7-H3 antibody (37) cause no toxicities in xenograft mouse models or phase I clinical trials. Therefore, we have already started to register a clinical trial for a combination treatment with low-dose SAHA and B7-H3.CAR T cells derived from the humanized mAb 8H9 in solid cancers. Finally, our and those of other studies (4, 37, 53) show that CAR T cells cannot completely eradicate tumors, because numerous obstacles (31) in solid tumors impair the trafficking, penetration, and survival of CAR T cells. Therefore, additional T-cell engineering and combination with chemotherapy, irradiation, small molecules, or other biological agents should be explored (31, 37, 53).

In conclusion, we generated B7-H3.CAR T cells from the humanized mAb 8H9, screened HDACis, and showed the SAHA-mediated upregulation of B7-H3 expression on solid cancer cells and of B7-H3.CAR T-cell expression on human T cells. The SAHA and B7-H3.CAR-T cell combination displayed increased antitumor activity with several solid cancer types. The information we have presented provides a convincing rationale to translate the combinatorial strategy we have developed to a clinical setting.

No disclosures were reported.

X. Lei: Data curation, software, formal analysis, visualization, writing–original draft. Z. Ou: Data curation, software, formal analysis, visualization, methodology, writing–original draft. Z. Yang: Resources, data curation, investigation. J. Zhong: Data curation, investigation. Y. Zhu: Resources, data curation. J. Tian: Software, visualization. J. Wu: Resources. H. Deng: Resources. X. Lin: Software, validation. Y. Peng: Methodology. B. Li: Software, investigation, visualization. L. He: Resources, investigation. Z. Tu: Validation, investigation. W. Chen: Methodology. Q. Li: Resources, data curation, software. N. Liu: Data curation, software, visualization. H. Zhang: Software, visualization, methodology. Z. Wang: Investigation, methodology. Z. Fang: Resources, funding acquisition, investigation. T. Yamada: Resources, data curation, software, funding acquisition, investigation. X. Lv: Data curation, software, validation, investigation. T. Tian: Validation, investigation, methodology. G. Pan: Data curation, methodology. F. Wu: Data curation. L. Xiao: Data curation, methodology. L. Zhang: Methodology. T. Cai: Formal analysis, methodology. X. Wang: Resources, formal analysis. B.A. Tannous: Resources. J. Li: Conceptualization, supervision, writing–original draft, project administration, writing–review and editing. F. Kontos: Conceptualization, resources, supervision, project administration, writing–review and editing. S. Ferrone: Conceptualization, supervision, validation, writing–original draft, writing–review and editing. S. Fan: Conceptualization, data curation, supervision, validation, writing–original draft, writing–review and editing.

The authors are grateful to Hongwei Du and Yingjin Lin for critically reading the article. They acknowledge Sha Fu’s help to analyze the tissue slides.

This study was supported by grants from the National Natural Science Foundation of China (grant Nos. 81472521, 81402251, 81502350, 81672676, and 81772890), Guangdong Science and Technology Development Fund (grant Nos. 2015A030313181, 2016A030313352, and 2017A030311011), Science and Technology Program of Guangzhou (grant Nos. 201607010108 and 201803010060), Fundamental Research Funds for the Central Universities (grant Nos. 16ykpy10 and 19ykzd20), National Clinical Key Specialty Construction Project for Department of Oral and Maxillofacial Surgery, The Key Laboratory of Malignant Tumor Gene Regulation and Target Therapy of Guangdong Higher Education Institutes, Sun Yat-sen University (grant No. KLB09001), and the Key Laboratory of Malignant Tumor Molecular Mechanism and Translational Medicine of Guangzhou Bureau of Science and Information Technology (grant No. (2013)163). S. Fan was supported by NIH grants R01DE028172, R03CA216114, R03CA223886, R03CA231766, and R03CA239193 and by DOD grant W81XWH-20-1-0315. This research was also supported by a grant from Guangdong Science and Technology Development Fund (2021A1515012355).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.