The methods of DNA extraction and next-generation sequencing/data processing are provided in the online supplemental methods.

Chidamide was purchased from ApexBio. The human sarcoma cell lines were purchased from ATCC. MCA205 is a 3-methylcholanthrene-induced fibrosarcoma syngeneic to C57BL/6 mice.17 HT1080, SK-LMS-1, T778, RD and SW872 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum. MCA205 cells were cultured in RPMI 1640 supplemented with 10% fetal bovine serum.

Written informed consent for the collection and publication of their medical information was obtained from each patient at their first visit to our center.

C57BL/6 mice were inoculated subcutaneously in the right flank with 2×10⁵ MCA205 tumor cells on day 0. Mice were then randomized into the following treatment groups: IgG control, chidamide alone, anti-PD-1 antibody alone and combination. Treatment with the drugs was started on day 5 after inoculation. The tumor volume was evaluated and calculated with the formula: (width²×length)/2. Chidamide was dissolved with 0.2% carboxymethyl cellulose (CMC) and 0.1% Tween 80 and administered once daily for 15 days (12.5 mg/kg/day, orally). The groups treated with out chidamide were administered using 0.2% CMC and 0.1% Tween 80 as controls. The anti-PD-1 antibody was injected through intraperitoneal injection (10 mg/kg, BioXCell) on days 6, 12, and 18 after tumor inoculation. Mice were sacrificed when the tumor’s diameter exceeded 20 mm.

Tumors were sufficiently chopped using scalpels and then digested with 0.2 mg/mL collagenase intravenous and 0.1 mg/mL DNase I at 37°C for 1 hour, and then passed through a 70 µm strainer to determine the presence of the infiltrating T cell population. Mouse-specific antibodies including anti-CD3, anti-CD4, anti-CD8, anti-CD11b, and anti-Ly6G antibodies were purchased from BD Pharmingen. Anti-IFN-γ and PD-1 antibodies were purchased from eBioscience. The human-specific antibody against PD-L1 was purchased from BD Pharmingen. Major histocompatibility complex (MHC) Class I (H-2Kb) antibody, MHC Class II (I-A/I-E) antibody, human MHC Class I/ HLA-ABC antibody (W6/32) and human MHC Class II/ HLA-DR antibody (LN3) were purchased from eBioscience. Intracellular staining for IFN-γ was performed after stimulation with PMA at 37℃ for 4–6 hours as described in the protocol of the Fixation and Permeabilization Buffer Kit (BD Bioscience). For detection of intracellular IFN-γ, brefeldin A was used to block secretion of cytokines during the last few hours of the stimulation. Cells were then subjected to flow cytometry.

Total cell lysates were electrophoretically separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane. Then the membrane was blocked with 5% skim milk for 1 hour at room temperature. Next, the membrane was incubated with primary antibody at 4°C overnight. Antibodies against acetylated H3K27, total H3, PD-L1, STAT1, and β-actin were purchased from Cell Signaling. Antibodies against HDAC1, HDAC2 and HDAC3 were purchased from Abcam.

Total RNA was extracted using TRIzol reagent (Invitrogen) according to a standard protocol, and the concentration of RNA was determined. Then, cDNAs were synthesized with a reverse transcriptase kit (Invitrogen). Expression was assessed using a qPCR analysis with SYBR Green on a Bio-Rad CFX96 PCR instrument. The resulting data were assessed, and relative mRNA expression was calculated with the formula 2[−(delta delta Ct)]. Primers are listed in online supplemental table S2.

HDAC1, HDAC2 and HDAC3 amplification were analyzed by using q-PCR with the method described by Ma and Chung.19 The level of amplification of HDAC genes were calculated as described by Lee et al.20 Positive amplification was considered when the copy number was greater than two times that of the reference gene. Primers are listed in online supplemental table S3.

ChIP assays were performed as described by Zeng et al.21 Briefly, 5×10⁶ HT-1080 cells were treated with or without chidamide for 24 hours. A total of 5 µg primary antibodies against acetylated-histone H3K27 from Cell Signaling and IgG control from R&D Systems were used for each immunoprecipitation. Protein A beads were incubated with the reactions for 2 hours at 4℃. Then, the beads were washed and eluted, and a PCR Purification Kit (Qiagen) was used to purify DNA. DNA was subjected to a q-PCR analysis using Bio-Rad CFX96 PCR instrument. PD-L1 promoter primers are listed in online supplemental table S4.

HT-1080 cells were transfected with the STAT1-targeted siRNA or siRNA NC overnight and then cotransfected with the pGL3-basic vector or the PD-L1 promoter luciferase reporter gene plasmid and the pRL-TK plasmid for 48 hours. Chidamide or dimethylsulfoxide (DMSO) was added for another 24 hours. Cell lysates were harvested for the dual-luciferase assay, which was performed according to the manufacturer’s instructions (Promega, Madison, Wisconsin, USA). Data were normalized to Renilla luciferase activity.

Cells were treated with chidamide or the DMSO control for 48 hours. Total RNA was isolated using TRIzol reagent. Library construction was performed with the generated 100 bp paired-end reads. Then the libraries were sequenced using the Illumina HiSeq 2500 platform by Annoroad Gene Technology (Beijing, China).

Sarcoma cells were seeded in 96-well plates and incubated overnight. Then, cells were treated with the control or various concentrations of chidamide for 0–5 days. Add 20 µL of 0.5 mg/mL 3-(4,5-Dimethylthiazol-2-yl)−2,5- diphenyltetrazolium-bromide (MTT) solusion into each well and incubate the plates at 37°C for 4 hours. Then, the media were carefully discarded, and the cells in each well were lysed with 200 µL of DMSO. The absorbance was measured at optical density (OD)=490 nm within 1 hour.

Colony formation assays were performed as described by Yuan et al.22 Briefly, cells were plated in six-well plates and cultured with DMEM in the presence or absence of chidamide for 14 days. Then the colonies were fixed with methanol and stained with 0.5% crystal violet. Three independent wells were established for each chidamide treatment concentration.

Sarcoma cells were transfected with siRNAs using Lipofectamine RNAiMAX Reagent (Invitrogen, Waltham, Massachusetts, USA) according to the manufacturer’s instructions.

Overal survival (OS) and disease-free survival (DFS) were analyzed using Kaplan-Meier and log rank tests. The significance of difference in expression in different groups was determined by an unpaired, two-tailed t-test or one-way analysis of variance. Statistical analyzes were performed using GraphPad Prism V.6.0 software and SPSS V.19.0. Statistical significance was defined as p<0.05.

We recruited 11 Chinese patients with pathologically confirmed liposarcoma and performed WES of the tumor–blood sample pairs from these patients. In this cohort, we detected 328 (mean 29.82) somatic non-silent mutations in 306 genes. TP53, which was previously reported to be the most recurrently mutated gene in sarcomas,23 was recurrently mutated in these patients (online supplemental figure S1).

Then, we identified somatic copy number alterations (SCNAs) and detected significant large segment copy number gains, including gains at chromosomes 6q24.3, 12p13.31 and 12q14.1 (online supplemental figure S2). The gains at chromosome 12q13~15 were previously reported as highly recurrent focal amplifications in all subtypes of sarcoma, and in our patients, the SCNA peak at 12q14.1 was the most significant amplification. Because gene amplification is a common basis for resistance to anticancer drugs, we analyzed SCNAs at the gene level and tried to identify obvious amplification patterns in drug-target genes (figure 1A). As expected, we detected CDK4 and MDM2 amplification in all samples and 10 of 11 samples, respectively. The expression of these two genes has been reported in well and dedifferentiated liposarcoma.24 CDK4 inhibitors, such as palbociclib, are FDA approved for breast cancer therapies,25 and MDM2 inhibitors, including nutlin-3, also display exciting prospects.26 Interestingly, we found that the HDAC gene family was also extensively amplified in 73% of the samples (HDAC1 in 2/11 patients, HDAC2 in 4/11, HDAC3 in 1/11, HDAC4 in 1/11, HDAC5 in 1/11, HDAC7 in 3/11, HDAC9 in 6/11 and HDAC10 in 2/11 patients; figure 1B). These genes were also idendified as frequently amplified in The Cancer Genome Atlas (TCGA) liposarcoma cohort (figure 1C). Furthermore, the HDAC gene family was extensively amplified in 76.65% (197/257) of all sarcoma samples with different subtypes in TCGA cohort and were particularly amplified in fibrosarcoma (22/24, 91.67%), undifferentiated sarcoma (34/34, 100%) and leiomyosarcoma samples (76/101, 75.25%) (online supplemental figure S3). Based on this finding, HDAC inhibitors may be potentially effective drugs for sarcoma treatment.

We analyzed the TCGA public database of 263 STS samples (cBioPortal.org) to identify the potential role of HDACs in sarcoma and observed an association between higher expression of class I HDACs, including HDAC1, HDAC2 and HDAC3 with shorter overall survival (p<0.001) (figure 2A). We also performed real-time PCR (RT-PCR) with cDNAs from human STS samples and analyzed the clinical significance of class I HDAC expression including HDAC1, HDAC2 and HDAC3, in 49 patients with different subtypes of STS treated at our cancer center. Patients with higher HDAC1 and HDAC2 mRNA expression exhibited a significantly lower overall survival rate (p<0.05). A trend, although not significant, toward shorter overall survival was observed for patients with higher expression of the HDAC3 mRNA (figure 2B). The characteristics of these 49 patients with sarcoma are shown in online supplemental table S1. In particular, patients with liposarcoma and fibrosarcoma patients accounted for 40.8% of all the patients evaluated (10 patients with liposarcoma and 10 patients with fibrosarcoma).

Then we used siRNAs to knockdown class I HDACs, respectively (figure 2C). As shown in figure 2D, siRNA-mediated knockdown of class I HDACs (HDAC1, HDAC2, and HDAC3) significantly promoted cell apoptosis. At 48 hours after siRNA transfection, knockdown of HDAC1 resulted in the highest apoptosis rate including early and late apoptosis stages. Moreover, knockdown of class I HDAC genes was responsible for the upregulation of PD-L1 in sarcoma cells (figure 2E–F).

Thus, class I HDAC expression is associated with a poor prognosis, and the inhibiton of class I HDAC expression induces apoptosis and increases PD-L1 expression in sarcoma.

Because class I HDAC expression indicated a poor prognosis and chidamide is a potential HDAC inhibitor targeting class I HDACs, we detected the effects of chidamide on sarcoma cells and used the upregulation of acetylated histone H3K27 as a marker of class I type HDAC inhibition. HT-1080, SK-LMS-1 and T778 sarcoma cells were treated with chidamide at the indicated doses. As shown in online supplemental figure S4A, chidamide increased the acetylation of H3K27. Since class I HDACs affect cell proliferation,27 we also detected the expression of HDAC1, HDAC2, and HDAC3 after treatment with chidamide and founded inhibited expression of HDAC1, HDAC2, and HDAC3 in the three sarcoma cell lines (online supplemental figure S4B). MTT assays and colony formation assays were performed in cells treated with various concentration of chidamide to examine the effect of chidamide on the proliferation of sarcoma cell lines (online supplemental figure S4C-F). Chidamide suppressed the proliferation of sarcoma cells in a dose-dependent manner.

HDAC inhibitors regulate the antitumor immune response in certain malignancies.28–30 We observed the upregulation of PD-L1 expression following the knockdown of class I HDACs. Thus, we hypothesized that chidamide would also upregulate PD-L1 expression. To verify this, we used RNA-seq to analyze transcriptional genes expression alteration on chidamide treatment in HT-1080 cell. A hallmark pathway analysis of these upregulated genes revealed 34 significantly affected pathways (online supplemental table S5). Among these pathways we focused on the immune-related pathways. Specifically, 17 genes involved in the allograft rejection pathway were transcriptionally modulated, including HLA-DMA, HLA-DMB, and HLA-DOA, which play important roles in antigen processing and presentation. Importantly, 14 genes involved in the interferon gamma response were upregulated including HLA-B, HLA-C, PD-L1, IL2RB and TNFAIP6, suggesting the reprogramming of the immune microenvironment of sarcoma cells on chidamide inhibition (figure 3A).

PD-L1 expression was evaluated in five sarcoma cell lines with different pathological types using flow cytometry and real-time PCR to confirm the results of this analysis (figure 3B and C). PD-L1 expression was consistently increased in a dose-dependent manner in sarcoma cell lines. In addition to PD-L1, MHC class I and MHC II genes were upregulated, according to the flow cytometry data, indicating that chidamide might enhance antigen presentation to promote antitumor immune response (online supplemental figure S5). Additional HDAC inhibitors including belinostat, panobinostat and vorinostat also showed potency in upregulating MHC class I and class II expression in sarcoma cell lines (online supplemental figure S6). However, PD-L2 expression was not increased by chidamide (figure 3C, (online supplemental figure S6). Next, we detect whether chidamide may exert similar effects on mouse sarcoma tumors. Chidamide also induced MHC class I and class II genes expression and PD-L1 expression in MCA205 cells (figure 3D–3F). Mice bearing xenograft tumors composed of MCA205 cells were treated with chidamide for 7 days (n=8/per group). Then tumors were harvested and subjected to RNA extraction. pd-l1 mRNA expression and MHC class I and II genes including H2-Aa, H2-Ab, and H2-L mRNA expression were analyzed by q-PCR (online supplemental figure S7). We also harvested the tumor and analyzed tumor-infiltrating lymphocytes (TILs). Both CD8⁺ and CD4⁺ TILs upregulated PD-1 after HDAC inhibition (figure 3G–3H). However, there was no significant changes in CD8⁺ PD-1⁺, and CD4⁺ PD-1⁺ cells were observed in spleen (data not shown). Based on these data, the chidamide treatment enhanced the antigen presentation process and upregulated PD-1/PD-L1 signaling in murine models.

We used an immunocompetent murine model of sarcoma cells to determine the therapeutic efficacy of combining the chidamide and an anti-PD-1 antibody. MCA205 cells were subcutaneously injected into C57BL/6 mice. Approximately 4 days later, mice were randomized to treatment with chidamide (12.5 mg/kg), an anti-PD-1 antibody (10 mg/kg), the combination of chidamide and anti-PD-1 antibody or IgG control (n=8 mice /group) as shown in figure 4A. The chidamide group and anti-PD-1 therapy group both exhibited decreased tumor growth compared with the control group (p<0.01). However, the combination group showed a generally significantly (p<0.001) decreased tumor burden compared with that of the IgG control group (figure 4B). Chidamide combined with anti-PD-1 therapy group significantly improved survival (p<0.05) compared with the control group (figure 4C). The control group survived for 18 days, and anti-PD-1 and chidamide-treated groups both survived for 20 days. The combination therapy group survived for 22.5 days. Only the combination group survived significantly longer than the control group (p<0.05).

The percentage of TILs in the four groups was assessed using flow cytometry. The CD3⁺CD8⁺ and CD3⁺CD4⁺ ratios were dramatically increased in the group receiving the combination therapy (figure 4D). The CD3⁺CD8⁺ ratio was also significantly increased in the chidamide group and anti-PD-1 group compared with the control group. However, the percentages of CD4 (CD3⁺CD4⁺) and CD8 (CD3⁺CD8⁺) cells in the spleens of the chidamide or anti-PD-1 groups did not display a significant difference compared with the untreated groups (data not shown). We cultured TILs for 4–6 hours, and then examined IFN-γ production in CD4⁺ and CD8⁺ TILs. The combination treatment significantly increased the production of IFN-γ of TILs compared with monotherapy or the control (figure 4E). Additionaly, the frequency of tumor CD8⁺ and CD4⁺ cells that produce IFN-γ ex vivo in response to MCA205 cells in an ELISPOT assay was significantly increased after exposure to the combination therapy (online supplemental figure S8). Chidamide combined with anti-PD-1 therapy was associated with a significant decrease in the population of MDSCs (CD11b+Ly6G+) compared with the control group (p<0.01). The ratio of MDSCs in the anti-PD-1 therapy group also showed a significant decrease compared with the control group (p<0.05) (figure 4F). Thus, the combination of chidamide with an anti-PD-1 antibody retards tumor growth and enhance the antitumor response by promoting and optimizing the tumor microenvironment.

Based on our preclinical data, we investigated the efficacy and toxicity of chidamide combined with toripalimab (an anti-PD-1 antibody) in patients with advanced STS treated at our institute. These patients were all diagnosed with metastatic disease and at least two lines of therapies had failed.

At the time of cut-off, a cohort of seven patients was enrolled to receive chidamide (30 mg/day, twice a week), combined with toripalimab 240 mg, q3wks, and underwent a confirmed response evaluation (figure 5A). The follow-up time for these patients was 40 weeks. Three patients achieved a partial response (PR), two pateints had stable disease, and two patients progressed (figure 5B–C). Currently, all the patients are still alive. All the included patients experienced at least one treatment-related adverse event (AE). The majority of these AEs were mild and tolerable, with no patient discontinuing treatment due to AEs (figure 5D). Then, we performed a q-PCR analysis to detect HDAC1, HDAC2 and HDAC3 amplification in patients #1, #3 and #6, who achieved PR (figure 5E). In patient #1, the copy numbers of HDAC2 and HDAC3 displayed a low-level gain, and HDAC1 was amplified, while in patient #3, the copy numbers of HDAC1 displayed a low-level gain. In patient #6, HDAC2 was amplified. However, in patient #2, #4, #5 and #7, the amplification was not detected (online supplemental figure S9). Thus, HDAC class I gene amplification with STS patients who achieved a PR, indicating that chidamide combined with an anti-PD-1 antibody shows a promise as treatment for patients with frequent HDAC amplification in the future.

We then performed chromatin immunoprecipitation (ChIP) assays to determine the potential chromatin state of the PD-L1 promoter and examined the active histone mark H3K27Ac in sarcoma cells following drug treatment with chidamide for 24 hours. We identified several potential binding sites along the first 2 kbp of the promoter region. Peak acetylation was observed at approximately at −743 to −597 bp upstream of the first exon of the PD-L1 gene (figure 6A).

We wanted to explore the regulatory effect of chidamide on the PD-L1 promotor and hypothesized that chidamide could regulated PD-L1 transcription by affecting the transcription factors that bind to the promoter. Forty-six genes were coexpressed with PD-L1, and genes with correlation coeffcients greater than 0.5 were filtered from the cBioPortal website. On the PROMO website, 75 transcription factors (TF) were predicted to bind to the −2 kb promoter of PD-L1 (online supplemental table S6). Then, a Venn diagram analysis was constructed and showed that STAT1 was the only potential TF (figure 6B). A scatter gram showed that the mRNA expression of PD-L1 and STAT1 was strongly correlated between (SARC, TCGA, Provisional; figure 6C). Western blot analysis also revealed increases in PD-L1 and STAT1 levels following treatment with chidamide (figure 6D). We performed the dual-luciferase reporter assays to explore the regulatory effect of STAT1 on the PD-L1 promoter. As shown in figure 6E, the promoter exhibited significantly enhanced luciferase activity on treatment with chidamide, and this increase was impared by STAT1 silencing. Moreover, knockdown of STAT1 by transfecting siRNAs into chidamide-treated sarcoma cells decreased the surface protein and mRNA expression of PD-L1 (figure 6F,G). Interestingly, knockdown of STAT1 in non-treated HT1080 cells did not significantly change the expression of PD-L1. Base on these data, STAT1 was involved in the chidamide-induced increase in PD-L1 expression.