C57BL/6 mice were obtained from Cavens Laboratory Animal (Changzhou, China). CD8 ovalbumin T-cell receptor transgenic (Tg) OT-I mice were kindly provided by Professor Hai Qi (Tsinghua University). Hdac9-KO mice were generated in a C57BL/6 background using the clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 technique, as previously described.22 The single guide RNA sequence was as follows: 5′-CGGACGACAGGATCCACCAC-3′. Mouse sequences were analyzed using Sequencher V.5.4 (GeneCodes, Ann Arbor, Michigan, USA). Primers used for genomic screening were as follows: forward 5′-GCTTACATGCGCTATTCAGGG3′, reverse 5′CTCCCATGGGTCGGTCTTATG3′. All mice were maintained under specific pathogen-free conditions, and experiments were performed at 6–8 weeks of age. Lewis lung carcinoma (LLC) cells and B16 melanoma cells derived from C57BL/6 mice were obtained from the American Type Culture Collection and maintained in Dulbecco’s modified Eagle’s medium (GIBCO, Nanjing, China) supplemented with 10% fetal bovine serum (FBS; Biolnd, Nanjing, China). LLC cells were infected with pMiT-ovalbumin retrovirus, packaged in 293 T cells, in the presence of 10 µg/mL polybrene after 24 hours of pretreatment with 250 ng tunicamycin (Sigma-Aldrich, Shanghai, China). Flow cytometry was performed to confirm ovalbumin expression in LLC cells.

Recombinant mouse Fms-like tyrosine kinase 3 (Flt3) ligand was purchased from Peprotech (Suzhou, China). Fluorescein-conjugated monoclonal antibodies against CD45 (30-F11), CD11c (N418), CD11b (M1/70), adhesion G protein-coupled receptor E1 (F4/80, BM8), Gr-1, programmed cell death ligand 1 (PD-L1, 10 F .9G2), CD3 (17A2), CD4 (GK1.5), CD8 (53 – 6.7), IFNγ (XMG1.2), and FOXP3 (MF-14); isotype control monoclonal antibodies, and carboxyfluorescein diacetate succinimidyl ester (CFSE) were purchased from Biolegend (Beijing, China); leukocyte activation cocktail (LAC) was purchased from BD Pharmingen (Shanghai, China). Microbead-conjugated monoclonal antibodies against CD8 or CD11c were purchased from Miltenyi Biotec (Shanghai, China). The iTAg Tetramer/PE-H-2Kb OVA (SIINFEKL) detection kit was purchased from MBL (Shanghai, China). Ovalbumin peptide SIINFEKL (OVA257-264), collagenase type IV, hyaluronidase, and deoxyribonuclease were purchased from Sigma-Aldrich (Shanghai, China). Dextran sulfate sodium (DSS; molecular mass 36,000–50,000 Da) was purchased from MP Biomedicals (Beijing, China). Monoclonal antibodies against human CD8 (EP1150Y) and HDAC9 (EPR5223) were purchased from Abcam (Shanghai, China).

The femurs and tibiae of wild-type and Hdac9^-/- C57BL/6 mice were harvested and the bone marrow was collected by Roswell Park Memorial Institute (RPMI) 1640 flushing. Red cell counts were removed by lysis. Bone marrow cells (1×10⁶/mL) were cultured in RPMI 1640 medium (GIBCO, Nanjing, China) supplemented with 10% FBS, 50 µM β-mercaptoethanol (GIBCO, Nanjing, China), and 100 ng/mL Flt3-ligand as described previously.12 On day 7, dendritic proliferating clusters were collected and purified using anti-CD11c microbeads (Miltenyi Biotec, Shanghai, China). The purity of DCs was confirmed to be >90% by flow cytometry.

LLC (2×10⁵), LLC-OVA (3×10⁵), or B16 (1×10⁵) cells suspended in 100 µL of saline solution were subcutaneously inoculated into the right flank of each C57BL/6 or Hdac9^-/- mouse. Tumors were measured with a caliper twice weekly, and the volume was calculated as tumor size (mm³) = ab²/2, where a is the length of the longest axis, and the b is the length at a right angle from a. Mice with tumors <300 mm³ were considered to be surviving. CD11c-DTR mice were injected with diphtheria toxin (DT; Sigma-Aldrich, Shanghai, China; 10 ng/g body weight in saline solution), and the next day, bone marrow–derived dendritic cells (BMDCs) from wild-type or Hdac9^-/- C57BL/6 mice were coinjected with tumor cells as described above. To maintain low levels of endogenous DCs, mice were injected with low-dose DT (4 ng/g body weight) every 3 days. Tumor-bearing mice were sacrificed before the tumor diameter reached 2 cm. Murine tumor protocols complied with all relevant laws and institutional guidelines and were approved by the Animal Care and Use Committee of Nanjing Medical University.

Surface and intracellular molecule staining was performed as previously described.11 Tumors and draining lymph nodes (DLNs) were collected from mice and minced into pieces smaller than 1 mm³, followed by digestion with collagenase type IV, hyaluronidase, and deoxyribonuclease for 30 min at 37°C on a rotating platform. Samples were then filtered through a 70 µm cell strainer and washed twice with staining buffer (phosphate-buffered saline containing 2% fetal calf serum and 1 mM ethylenediaminetetraacetic acid). Cells were resuspended in staining buffer, blocked with an Fc-blocking monoclonal antibody for 15 min on ice, and stained with fluorescently labeled antibodies against CD45, CD11c, CD11b, F4/80, Gr-1, PD-L1, CD3, CD4, or CD8 on ice for 30 min. For IFNγ and FOXP3 staining, cells were fixed and permeabilized. OT-I-specific T cells were stained using iTAg Tetramer/PE-H-2Kb OVA (SIINFEKL). After a washing step, flow cytometry was performed on a BD FACSCanto II (San Jose, California, USA). Flow cytometry data were analyzed using FlowJo software V.8 (Tree Star, New Jersey, USA).

OT-I cells were isolated from the spleens and lymph nodes of OT-I mice and CD8⁺ T cells were purified using Miltenyi magnetic-activated cell sorting system. To functionally identify tumor-associated DCs, single cell suspensions were isolated from mouse tumor tissue by Percoll density gradient centrifugation followed by CD11c⁺ cell sorting. Tumor-infiltrating DCs were pulsed with 1 µg/mL SIINFEKL peptide (OVA257-264) and cocultured with naïve CFSE-labeled OT-I CD8⁺ T cells (2.5 µM CFSE for 10 min at 37°C) at a 1:5 ratio (DCs: 0.2×10⁶, T cells: 1×10⁶) in 24-well tissue culture plates for 3–5 days. On the day of detection, cells were restimulated with LAC for 4–6 hours, stained for intracellular IFNγ, and analyzed by flow cytometry.

Single cell suspensions of DLN cells from LLC-OVA tumor-bearing mice were obtained and restimulated in vitro with 1 µg/mL SIINFEKL peptide. After 72 hours of culture, cells were collected and analyzed for intracellular IFNγ expression by flow cytometry.

Paraffin-embedded tumor tissues from 17 patients with lung cancer were obtained with informed consent, using a protocol that complied with relevant ethical regulations and was approved by Changzhou No. 2 People’s Hospital review board (No. [2018]KY306-01). The microtome was used to prepare 5 µm sections. For antigen retrieval, heat-induced epitope retrieval was conducted in citrate buffer at pH 6 using a boiler. For detection, a horseradish peroxidase–conjugated secondary antibody was used along with the chromogen diaminobenzidine (DAB; Roche, Shanghai, China). The specific binding of an antibody to its antigen resulted in brown staining. Tissue sections were counterstained with hematoxylin and dehydrated with a graded alcohol series and xylene. Fiji software (a distribution of ImageJ) was used to quantify the mean HDAC9 intensity and the number of CD8⁺ cells.

DSS colitis was induced by adding 2.5% DSS to the mice’s drinking water for 1 week, followed by normal drinking water for the remainder of the experiment. Body weights were monitored daily over the course of the experiment. At the end of the experimental period, mice were sacrificed, and their colon lengths were measured. Then, a 1 cm segment of the distal colon was immediately embedded in optimal cutting temperature compound (SAKURA, Nanjing, China) and frozen. After sectioning into 3 µm sections, colon segments were stained with H&E to assess tissue integrity.

Sort-purified BMDCs (5×10⁶) were collected in duplicate from wild-type and Hdac9^-/- mice. Total RNA was extracted with TRIzol and genomic DNA was removed using DNase I (TaKaRa, Beijing, China). RNA quality was determined and high-quality samples (optical density (OD) 260/280=1.8–2.2, OD 260/230≥2.0, RNA integrity number ≥6.5, 28S:18S ≥1.0, >10 µg) were used to construct RNA-seq transcriptome libraries with the TruSeq RNA Sample Preparation Kit (Illumina; San Diego, California, USA), using 1 µg of total RNA. To identify differentially expressed genes (DEGs) between samples from wild-type and Hdac9^-/- mice, the expression levels of each transcript were calculated using the fragments per kilobase of exon per million mapped reads (FRKM) method. The R statistical package edgeR was used for differential expression analysis. Significant DEGs were defined by a logarithmic fold change >2 with a false discovery rate <0.05. To examine DEG functions, gene ontology (GO) functional enrichment analysis was performed in the Database for Annotation, Visualization and Integrated Discovery. DEGs were considered significantly enriched in GO terms and metabolic pathways when their Bonferroni-corrected p values were <0.05.

The unpaired Student’s t-test or one-way analysis of variance were used for comparisons between two groups. P values ≤0.05 were considered significant. For survival curves, the log-rank (Mantel-Cox) test was used. To determine correlations between HDAC9 levels and CD8⁺ cells, immunohistochemistry experiments were performed with blinded staining and blinded analysis, and Pearson’s correlation test was used. Statistical analysis was performed in GraphPad Prism V.7.0 software, and results are the mean±SE of the mean (SEM). No statistical methods were used to predetermine sample size.

To analyze the specific role of HDAC9 in tumor progression, we generated stable Hdac9 KO mice using CRISPR/Cas9 technology. Deletion of a fragment of Hdac9 exon 5 was confirmed by DNA sequencing (online supplementary figure S1a). Western blot analysis showed complete loss of HDAC9 expression in Hdac9 ^-/- mice (online supplementary figure S1b).

We subcutaneously inoculated LLC cells into wild-type and Hdac9 ^-/- mice. Compared with wild-type mice, Hdac9 ^-/- mice displayed faster tumor growth and shortened survival (figure 1A–C). Similar results were observed using B16 melanoma cells (online supplementary figure S2).

To examine HDAC9-associated immune effects in vivo, we next quantified the tumor-infiltrating immune cells (CD45⁺ cells) in these mice. The frequencies of CD3⁺CD8⁺ T cells and CD11c⁺CD8⁺ DCs in the TME and DLNs were significantly decreased in Hdac9^-/- mice compared with wild-type mice, while CD3⁺CD4⁺ T cells are not remarkably altered in the HDAC9-deficient mice (figure 2A–C). CD11c⁺PD-L1⁺ DCs were markedly increased in the TME of HDAC9-deficient mice, but not in the DLNs. There were no significant differences in the frequencies of total CD11c⁺ DCs, CD11c⁺CD11b⁺ DCs, CD11b⁺F4/80⁺ macrophages, CD11b⁺Gr-1⁺ MDSCs, T cells, or Treg cells in the TME (online supplementary figure S3a-c). In addition, loss of HDAC9 did not affect the composition of DC subpopulations in naïve mice, nor did it affect the numbers of other immune cells (online supplementary figure S3d). The absence of HDAC9 did not significantly influence the expression of MHCII, CD80, CD40, CD86, CCR7 on the DCs within the tumor (online supplementary figure S3e).

Next, we examined whether the cytokine profiles in the tumor milieu were altered between wild-type and Hdac9^-/- mice. Tumors were excised, and the total RNA was extracted for qRT-PCR. As shown in figure 2D, the mRNA levels of cytokines and molecules known to support antitumor immunity, such as IFNγ, GZMB, TBX21, IL12A, and IL12B, were not significantly changed. Interestingly, the production of molecules that suppress the immune response, such as GATA3, NR4A1, FOXP3, TGFβ1, and IL10, was significantly increased, while CD8A mRNA was significantly decreased within the TME of Hdac9^-/- mice. These data suggest that HDAC9 is capable of reprogramming immune effects at the tumor site.

To investigate whether HDAC9 deficiency reduced DC cross-presentation of tumor antigens, resulting in decreased CD8⁺ T cell priming, we used the LLC-OVA tumor model (figure 3A) and analyzed the abundance of H2Kb-SIINFEKL complexes on DCs from tumor-bearing wild-type and Hdac9 ^-/- mice. As shown in figure 3B, H2Kb-SIINFEKL complex levels were significantly lower in tumor-infiltrating Hdac9 ^-/- DCs than in wild-type DCs, suggesting that DCs from Hdac9 ^-/- mice possess diminished antigen-presentation activity relative to wild-type DCs. To verify the ability of the DCs to promote CD8⁺ T cell priming, CFSE-labeled OT-I cells were cocultured with sorted DCs from the TMEs of LLC-OVA tumor-bearing mice. As expected, Hdac9 ^-/- DCs drastically decreased CD8⁺ T cell proliferation and differentiation into effector cells, as measured by IFNγ production (figure 3C). These results suggest that loss of HDAC9 decreases the cross-priming capacity of DCs.

To determine whether antigen-specific CD8⁺ T cells were differentiating in vivo, we assessed the frequency of LLC-OVA tumor-infiltrating CD8⁺ T cells expressing the SIINFEKL-major histocompatibility complex-I tetramer in wild-type and Hdac9^-/- mice. Compared with wild-type mice, Hdac9^-/- mice displayed a substantial decrease in CD8⁺ T cells activated against the OVA antigen in vivo (figure 3D). Next, lymphocytes from DLNs were restimulated with OVA-derived SIINFEKL peptide in vitro for 3 days. There were significantly less CD8⁺IFNγ⁺ T cells in the Hdac9^-/- cells than in the wild-type cells (figure 3E), indicating that HDAC9 loss decreases T cell priming against tumor antigens.

To investigate whether the observed immune suppression relies on the specific loss of HDAC9 from DCs, we generated BMDCs from wild-type mice and Hdac9^-/- mice using Flt3 ligand. After sorting, BMDCs were subcutaneously coinjected with LLC tumor cells into CD11c-DTR transgenic mice that had been depleted of host DCs expressing CD11c-DTR by administration of DT. Endogenous CD11c⁺ DC depletion was verified by flow cytometric analysis 24 hours after injection of 200 ng DT or control vehicle (online supplementary figure S4). Notably, although mice injected with Hdac9^-/- DCs displayed less tumor growth than mice injected with LLC cells alone, wild-type DCs were much more efficient at preventing tumor progression (figure 3f). These results demonstrate that DCs participate in the impaired antitumor immunity of Hdac9^-/- mice.

DC activation is associated with colitis,23 so we next investigated whether HDAC9 is involved in DC-mediated colitis. We challenged wild-type and Hdac9^-/- mice with DSS, a widely used chemical irritant that induces intestinal inflammation with the clinical, immunological, and histological features of human inflammatory bowel disease. The protocol is shown in figure 4A. Body weight loss was significantly delayed in Hdac9^-/- mice compared with wild-type mice (figure 4B). Reduced colon length often serves as a surrogate macroscopic indication of colonic injury, and we observed significantly longer colons in Hdac9^-/- mice than in wild-type mice following DSS treatment (figure 4C). Furthermore, histological analysis of the distal colon demonstrated severe disrupted crypt architecture and abundant inflammatory cell infiltration in DSS-treated wild-type mice, while these phenotypes were alleviated in Hdac9^-/- mice, with less crypt loss and decreased immune cell infiltration (figure 4D). These findings demonstrate that HDAC9 deficiency reduces inflammation, inhibiting the onset and clinical signs of DSS-induced colitis.

To systematically assess the role of HDAC9 in the regulation of DC immune functions, we next performed RNA-seq to identify DEGs in Hdac9^-/- BMDCs compared with wild-type BMDCs. We identified 72 significant DEGs in Hdac9^-/- DCs, including 59 upregulated genes and 13 downregulated genes (figure 5A).

To identify HDAC9-associated functional pathways, we performed GO term enrichment analysis. As shown in figure 5B, with regard to biological processes, HDAC9-associated transcripts were enriched in pathways associated with immune system processes (particularly the innate response), drug responses, neuronal differentiation, and circadian rhythms. Further analysis revealed that the translation of a group of transcripts responsible for inflammatory responses and antigen recognition was changed in Hdac9^-/- DCs compared with in wild-type DCs (figure 5C). Conversely, the translation of costimulatory or inhibitory molecules and cytokines was not significantly altered in Hdac9^-/- DCs (data not shown). These data suggest that HDAC9 extensively regulates DC immune functions. Specifically, the transcription of GATA3, which is generally identified as T cell-specific, was dramatically upregulated in DCs with HDAC9 deficiency, as also shown in figure 2C. This indicates that HDAC9 may participate in the regulation of DC differentiation from common myeloid and lymphoid progenitor cells.

To examine the effects of HDAC9 expression on the human TME, we collected tissue sections from patients with non-small cell lung cancer and performed immunohistochemical staining for CD8 and HDAC9. Consistent with the observations in mouse models, patients with low HDAC9 expression in the tumor stroma tended to have a lower number of CD8⁺ cells than patients with high HDAC9 expression (online supplementary figure S5, figure 6). The mean intensity of HDAC9 in the stroma was positively correlated with the number of CD8⁺ cells, indicating that low HDAC9 expression might control the amount of T cells in the non-inflamed TME.