SUV39H1 Represses the Expression of Cytotoxic T-Lymphocyte Effector Genes to Promote Colon Tumor Immune Evasion

Colon cancer was the first human neoplasia that was found to be under immunosurveillance by the adaptive immune response (1–3). The type, density, and location of tumor-infiltrating immune cells within the tumor microenvironment correlate with tumor growth, progression, and recurrence, and predicts patient clinical outcomes, and therefore, human colon tumors are considered immunogenic (1–8). Given the presence of helper T cells (Th1) and CTLs in the tumor microenvironment, it might be expected that the host immune system would naturally eliminate these immunogenic colon tumors. However, colon cancer is the third most commonly diagnosed cancer and the second leading cause of cancer-related death in men and women combined, suggesting that this is not the case. Why this host defense mechanism fails to occur is not yet known. One possibility is that the microsatellite instable (MSI) colon tumors selectively upregulate multiple counter-inhibitory checkpoint molecules affording immune escape (9). Consistent with this notion, PD-1 blocking antibody immunotherapy yields high response rates in MSI colon cancer patients (10).

However, despite breakthroughs in immune checkpoint inhibitor (ICI) immunotherapy for many types of human cancers, colon cancer stands out as one of the few human cancers that does not respond to ICI immunotherapy (except for the small subset of MSI-positive patients, and these represent only ∼4% of all colon cancer cases; ref. 10). Why the majority of human immunogenic colon cancer cases do not respond to ICI immunotherapy is unknown (11).

ICI immunotherapy, such as anti–PD-1 and anti–PD-L1, block interactions between PD-L1 and PD-1 to reverse immune suppression to unleash CTLs that kill tumor cells in the tumor microenvironment. CTLs kill tumor cells through inducing tumor cell apoptosis, and for this to occur, CTLs must be activated and the target tumor cells must be sensitive to apoptosis induction. Resistance to apoptosis is a hallmark of human cancer (12), particularly in human colon cancer (13–15). If tumor cells are not sensitive to apoptosis induction, then regardless of how potent the CTLs are, the tumor cells will not be eliminated. Other PD-L1–independent mechanisms of immune suppression, such as myeloid-derived suppressive cells (MDSCs), may suppress ICI-activated CTLs in the tumor microenvironment (16) in addition to PD-L1–mediated immune suppression. The fact that CTL infiltration is observed in both microsatellite stable (MSS) and MSI human colon cancer suggests that a lack of CTL infiltration is unlikely to be the sole mechanism underlying human colon cancer nonresponse to ICI immunotherapy (1, 3, 4, 16, 17). Therefore, targeting the effector gene expression and function of tumor-infiltrating CTLs can potentially bypass immune suppression to directly augment their effector function to control colon carcinoma development. To test this hypothesis, we developed a novel small-molecule inhibitor (termed F5446) specific to histone methyltransferase SUV39H1 that decreased H3K9me3 at T-cell effector promoters to increase effector gene expression in tumor-infiltrating CTLs, thereby suppressing colon carcinoma growth in vivo.

BALB/c and C57BL/6 mice were obtained from the Charles River Frederick mouse facility. B6.129S7-Rag1^tm1Mom/J mice were obtained from the Jackson Laboratory. Six- to 8-week-old female mice were used. All mice were housed, maintained, and studied in accordance with approved protocols by Augusta University and the Ohio State University Institutional Animal Use and Care Committees. Mouse colon carcinoma CT26 cells were obtained from ATCC in 2013 and stored in liquid nitrogen as aliquots. Cells were used within 30 passages. ATCC characterized these cells by morphology, immunology, DNA fingerprint, and cytogenetics. Murine colon carcinoma MC38 cells were provided by Dr. Jeffrey Schlom (National Cancer Institute, Bethesda, MD) and characterized as described previously (18). Cells were cultured in RPMI1640 medium (Invitrogen) plus 10% HyClone FBS (catalog no. SH30396.03, GE Healthcare). Cells are tested for Mycoplasma every 2 months, and all cells used in this study were Mycoplasma-negative.

Human dataset of SUV39H1 expression in human primary tumor (n = 380) and normal solid tissue (n = 51) was extracted from TCGA Colon and Rectal Cancer (COADREAD) ploy A⁺ IlluminaHiSeq pancan normalized RNA-seq datasets using UCSC Xena Cancer Genomics Browser (University of California, Santa Cruz, CA). H3K9me3 ChIP-Seq dataset were extracted from the Gene Expression Omnibus (GEO) Database (accession# GSM1058783) and analyzed with IGV 2.4x program (Broad Institute, Cambridge, MA, and University of California, San Diego, CA).

Synthesis of F5446 was conducted by Leadgen Labs LLC using reagents purchased from Sigma-Aldrich and methods as described previously (19). ¹H NMR spectra were recorded at Brucker 400 MHz, and ¹³C NMR spectra were recorded at 100 MHz at room temperature using CDCl₃ as solvent and tetramethylsilane (TMS) as an internal standard. Chemical shifts were reported in ppm relative to TMS as the internal standard. The following abbreviations are used to describe spin multiplicity: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, brs = broad singlet, dd = doublet of doublets.

The LC/MS was recorded on an Agilent 1100 with LC/MSD SLMobile Phase: A, H₂O+0.1%HCOOH; B, MeCN+0.1HCOOH. 5: 6-(4-Chloro-benzenesulfonyl)-4,5-dioxo-3,4,5,6-tetrahydro-3,6-diaza-asindacene-1,7-dicarboxylic acid 1-benzyl ester 7-methyl ester: beige crystalline powder; m.p. 141-152°C; ¹H NMR (400 MHz, CDCl₃, δ) 11.05–11.02 (m, 1H), 8.56 (d, J = 0.8 Hz, 1H), 8.54 (d, J = 0.8 Hz, 1H), 7.82–7.81 (m, 2H), 7.53–7.51 (m, 2H), 7.45–7.30 (m, 5H), 5.37 (s, 2H), 4.00 (s, 3H); ¹³C NMR (100 MHz, CDCl₃, δ) 169.3, 169.2, 141.1, 136.7, 135.0, 133.9, 133.9, 133.9, 131.4, 131.2, 128.7, 128.3, 128.1, 128.0, 127.2, 125.9, 117.4, 114.8, 66.5, 53.1. LC/MS for C₂₆H₁₇ClN₂O₈S [M+H]⁺ 552.95, found 553.0.

Step A: Methyl 1-(4-chlorobenzenesulfonyl)-4-formyl-1H-pyrrole-2-carboxylate. Solution of pyrrole 1 (1.0 g, 6.5 mmol) in tetrahydrofuran (THF; 15 mL) was slowly added to a suspension of NaH (0.36 g, 60% in mineral oil, 9.0 mmol) in THF (15 mL). The reaction mixture was then stirred for 1 hour at room temperature, followed by slow addition of p-ClPhSO₂Cl (1.65 g, 7.8 mmol) in 15 mL of THF. The reaction mixture was then stirred at room temperature overnight followed by quenching with CH₃COOH.

After solvent was evaporated, the crude product was purified by column chromatography using hexane/EtOAc = 7/3. Product 2 was isolated with the yield of 1.7 g, 70%.

Step B: Methyl 4-[(1E)-3-(benzyloxy)-3-oxoprop-1-en-1-yl]-1-(4-chlorobenzenesulfonyl)-1H-pyrrole-2-carboxylate. (EtO)₂P(O)CH₂CO₂Bn (1.69 g, 5.9 mmol) was added dropwise at −30°C to a suspension of NaH (0.29 g 60% in mineral oil, 6.7 mmol) in THF (12 mL) and the mixture was then warmed up to 0°C. After the mixture was cooled to −30°C, the pyrrole 2 (1.7 g, 5.2 mmol) in THF (10 mL) was added at −10°C. After 5 minutes of stirring, the mixture was quenched with 10% aqueous NH₄Cl (7 mL), and the THF was evaporated under reduced pressure. The residue was extracted with CH₂Cl₂ (3 × 50 mL), dried (Na₂SO₄), and evaporated to give a tan solid (2.1 g). Recrystallization from acetone gave product 3: 1.56 g, 74%.

Step C: Methyl 4-{4-[(benzyloxy)carbonyl]-1H-pyrrol-3-yl}-1-(4-chlorobenzenesulfonyl)-1H-pyrrole-2-carboxylate. Freshly prepared LiHMDS [prepared from HMDS (0.88 g, 5.4 mmol) and 2.5 mol/L n-BuLi (1.56 mL, 3.9 mmol) in THF (8 mL)] was added at −78°C to a solution of TosMIC (0.64 g, 3.3 mmol) and product 3 (1.13 g, 2.5 mmol) in THF (25 mL). After stirring for 15 minutes at −70°C, the reaction mixture was quenched with 10% aqueous NH₄Cl (8 mL) and evaporated under reduced pressure to remove THF. The residue was then extracted with CH₂Cl₂ (4 × 20 mL), dried (Na₂SO₄), and filtered.

After evaporation, the residue was chromatographed over silica gel using 5% EtOAc/25% petroleum ether/70% CH₂Cl₂ to give product 4: yield 0.55 g, 45%.

Step D: 1-Benzyl 7-methyl 6-(4-chlorobenzenesulfonyl)-4,5-dioxo-3H,4H,5H,6H-pyrrolo[3,2-e]indole-1,7-dicarboxylate. Oxalyl chloride (0.11 mL, 1.3 mmol) was added dropwise to a solution of the bipyrrole 4 (0.52 g, 1.0 mmol) in CH₂Cl₂ (20 mL) at 0°C. More oxalyl chloride (0.06 mL) was slowly added in an hour, and the mixture was left at 0°C for 24 hours. The mixture was cooled to −78°C and SnCl₄ (0.118 mL, 1.0 mmol) was added dropwise. After 1.5 hours, the mixture was quenched by dropwise addition of water (4.0 mL). The CH₂Cl₂ layer was separated, and the aqueous phase was extracted with CH₂Cl₂ (3 × 10 mL). Combined organic phases were dried (Na₂SO₄) and evaporated under reduced pressure. The residue was then chromatographed on silica gel using CHCl₃/MeOH = 25/1.

Structures in the NCI Diversity Set V were supplied in 2D form (https://wiki.nci.nih.gov/display/NCIDTPdata/Compound+Sets). Three-dimensional representations were generated with computer program obprep from the Open Babel suite (20). Docking of these structures was done with computer program Autodock Vina 2. 0 (ref. 21; The Scripps Research Institute, San Diego, CA) to predict binding modes of a drug candidate against the protein target using a box centered on the ligand-binding site. The figure-of-merit in the scoring output of Autodock Vina is expressed as a binding affinity in kcal/mol, with a standard error of 2.85 kcal/mol. Suppressor of variegation 3–9, enhancer of zeste and trithorax (SET) domain structure of human SUV39H1 protein was used for molecular docking (22).

Recombinant SUV39H1 protein (250 nmol/L, catalog no. SUV39H1-406H; Reaction Biology Corp) was incubated with human histone 3 peptide (amino acids 1–21, 5 μmol/L), S-Adenosyl-L-[methyl-³H] methionine (1 μmol/L, catalog no. NET15500, PerkinElmer), and the test compounds (0–50 μmol/L) in reaction buffer [50 mmol/L Tris-HCl pH 8.5, 50 mmol/L NaCl, 5 mmol/L MgCl₂, 1 mmol/L dithiothreitol (Sigma-Aldrich), and 1 mmol/L phenylmethylsulfonyl fluoride (Sigma-Aldrich)] at 30°C for 60 minutes. The reaction was stopped by adding EDTA to 50 mmol/L. the reaction mixture (6 μL) from each reaction was spotted onto a 1.2 × 1.2 cm Whatman P81 ion exchange paper (catalog no. 3698-915, obtained from Reaction Biology Corp) disc. The filter discs were air-dried at room temperature for 30 minutes, washed three times in washing buffer (50 mmol/L Na₂CO₃ and NaHCO₃ pH 9.2) for 10 minutes each time, and air-dried again. The filters were transferred to 20-mL scintillation vials containing 1 mL water and 10 mL Ultima Gold scintillation cocktail (catalog no. 6013329, PerkinElmer) and counted in a LS6500 Multi-Purpose Scintillation Counter (Beckman). Enzymatic inhibition was calculated using Prism GraphPad 6.0. F5446 inhibitory activity was determined in a 10-dose EC₅₀ mode with 3-fold serial dilution starting at 10 μmol/L essentially as described above in Reaction Biology Corp.

C57BL/6 mice (7–8 weeks of age) were injected intraperitoneally with vehicle [10% Cremophor EL (catalog no. C5135, Sigma-Aldrich) in PBS] or F5446 at doses of 10 and 20 mg/kg body weight (n = 5 per group), respectively, every 2 days for a total of 7 doses. Mouse weights were measured every 2 days. Mice were euthanized approximately 24 hours after the last dose. Blood was collected from mice immediately after CO₂ euthanasia via cardiac puncture. Whole blood from each mouse was dispensed into a microvette 100 K3E tube (catalog no. 20.1278.100, Sarstedt) and a 1.5 mL microcentrifuge tube, which were then submitted to the Comparative Pathology and Mouse Phenotyping Shared Resource at The Ohio State University Comprehensive Cancer Center (Columbus, OH; https://cancer.osu.edu/research-and-education/shared-resources/comparative-pathology-and-mouse-phenotyping) for the determination of complete blood counts and serum biochemical panel, respectively. Complete blood counts were performed on a FORCYTE Autosampler 10 Hematology Analyzer (Oxford Science, Inc.), followed by manual confirmation of white blood cell differential counts. Serum biochemical panel was determined on a VetAce Chemistry Analyzer (Alfa Wassermann) using undiluted serum collected manually as supernatant after removal of blood clot by centrifugation.

MC38 tumor cells (1.5 × 10⁵ cells/mouse) were injected into C57BL/6 or Rag1 KO mice, and CT26 cells (2 × 10⁵ cells/mouse) were injected into BALB/c mice. MC38 tumor–bearing mice were treated 8 days later, and CT26 tumor-bearing mice were treated 10 days later, respectively, with vehicle (10% cremophor in PBS) and IgG (200 μg/mouse; Control), F5446 only (10 mg/kg body weight), anti–PD-1 only (200 μg/mouse, clone 29F.1A12; BioXcell), or F5446 plus anti–PD-1 every 2 days for a total of 7 times. To determine CD8⁺ CTL function, tumor-bearing mice were treated with vehicle and IgG, F5446, or F5446 plus a CD8-specific neutralization mAb (200 μg/mouse, clone BE0061; BioXcell) at same concentrations. Tumor size was measured in two dimensions with a digital micrometer caliper at the indicated time points. Tumor volume was calculated by the formula: (tumor length × tumor width²)/2. MC38 tumor–bearing mice were sacrificed 22 days after tumor injection, and CT26 tumor–bearing mice were sacrificed 24 days after tumor injection.

MC38 tumor cells (1.5 × 10⁵ cells/mouse) were injected to C57BL/6 mice subcutaneously (s.c.). Tumors were excised 22 days after tumor cell injection and digested in collagenase solution–collagenase (1 mg/mL; Sigma-Aldrich), hyaluronidase (0.1 mg/mL; Sigma-Aldrich), and DNaseI (30 U/mL; Roche Diagnostics)–in RPMI1640 medium plus 10% FBS at 37°C on a magnet stirrer for 30 minutes to obtain single cells. The digested tumor mixtures were centrifuged in a Sorvall Legend RT centrifuge (Kendro Laboratory Products) at 1,500 rpm for 5 minutes. The cell pellet was resuspended in red cell lysis buffer (0.15 mol/L NH₄Cl, 10 mmol/L KHCO₃, and 0.1 mmol/L EDTA pH7.4), incubated at room temperature for 5 minutes, centrifuged, and washed with PBS once. The cells were then resuspended in PBS and the MojoSort mouse CD45 nanobeads (catalog no. 480028, BioLegend) were added to the cell suspension. The cell and beads mixtures were incubated in 4°C with rotation for 30 minutes, and then placed on DYNAL MPC-S magnet stand (Invitrogen) for 2 minutes. The supernatant was collected and used as the CD45⁻ (>95% CD45⁻ cells) tumor cells. To isolate tumor-infiltrating CTLs, tumor mixtures were incubated with mouse CD8 Dynabeads (catalog no. 114.47D, clone lyt-2, Invitrogen Dynal AS) in PBS and incubated at 4°C with rotation for 30 minutes. The bead and cell mixtures were placed on a DYNAL MPC-S magnet stand (Invitrogen) for 2 minutes. The supernatant was removed and the beads were resuspended in 1 mL PBS, and placed on the magnet stand for 2 minutes. This washing step was repeated once. The bead-bound CD8⁺ cells were directly used for analysis.

Spleens were collected from C57BL/6 mice and broken into single cells in 10-mL PBS by using two sterile frosted microscope slides (catalog no. 12-552-3, Thermo Fisher Scientific). The cell suspension was passed through a 100-μm cell strainer (catalog no. 22363549, Thermo Fisher Scientific) and centrifuged. The cell pellet was resuspended in PBS and washed twice. The cell pellet was then resuspended in PBS, and CD3⁺ T cells were purified using the MojoSort mouse CD3 T cell isolation kit (catalog no. 480031, BioLegend) according to the manufacturer's instructions. The CD3⁺ T-cell purity was above 90%. For T-cell activation, purified T cells (1.5 × 10⁶ cells/well in 24-well plate) were cultured in RPMI1640 medium plus 10% FBS in tissue culture plates coated with anti-mouse CD3 (8 μg/mL, clone 145-2C11; BioXcell) and anti-mouse CD28 (10 μg/mL, clone 37.51; BioXcell) for 2 days.

Total RNA was isolated from CD3⁺ T cells, purified tumor-infiltrating CD8⁺ T cells, and fresh whole tumor tissues using the GeneJET RNA isolation kit according to the manufacturer's instructions (Thermo Fisher Scientific). RNA concentrations were measured using a NanoDrop Spectrophotometer (Thermo Fisher Scientific). RNA purity was determined on the basis of OD260/OD280 ratio, and only samples with ratio >1.90 were used. cDNA was synthesized from 500 ng total RNA in a 20 μL reaction volume using the ThermoScript RT-PCR kit (catalog no. 11146-016, Invitrogen) with the substitution of ThermoScript reverse transcriptase by M-MLV reverse transcriptase (catalog no. M170B, Promega). PCR was performed in triplicates with 1 μL cDNA mixture using gene-specific primers (Supplementary Table S1) in the StepOne Plus Real-Time PCR System (Applied Biosystems). Gene expression of Gzmb, Prf1, Faslg, and Ifng was determined as ΔΔC_t with β-actin as internal normalization control.

Western blotting analysis was performed as described previously (23). Briefly, resting and in vitro–activated spleen CD3⁺ T cells and tumor-infiltrating CD8⁺ T cells were lysed in total cell lysate buffer (20 mmol/L Hepes pH 7.4, 20 mmol/L NaCl, 10% glycerol, 1% Triton X100) plus protease inhibitor cocktail (catalog no. 535140, Millipore) on ice for 60 minutes. The cell lysate was centrifuged at 13,000 rpm in 4°C for 10 minutes in a Eppendorf 5415R microcentrifuge (Eppendorf). The supernatant was collected and measured for protein concentration using Protein Assay Dye Reagent (catalog no. 500-0006, Bio-Rad Laboratories) according to the manufacturer's instructions using BSA (catalog no. BP1605, Thermo Fisher Scientific) as concentration standard. Total lysates (20 μg per sample) were loaded to a 4%–20% mini-protean TGX precast gel (catalog no. 456-1096, Bio-Rad Laboratories), and the proteins were transferred to the Immun-Blot PVDF membrane (catalog no. 162-0177, Bio-Rad Laboratories) after electrophoresis using a Pierce power blot cassette. The membrane was incubated in 5% blotting-grade milk (Bio-Rad Laboratories) in PBS plus 1% Tween-20 (catalog no. P1379, Sigma-Aldrich) at room temperature for 1 hour. The membrane was then washed three times in PBS plus 1% Tween 20 and incubated with anti-mouse CD8 (BD Pharmingen, clone 53-6.7) at 1:1,000, anti-mouse CD11b (Abcam, clone M1/70) at 1:1,000, anti-mouse perforin (clone H315, Santa Cruz Biotechnology) at 1:500, and anti-mouse β-actin (clone: AC-74, Sigma-Aldrich) at 1:10,000, respectively, in PBS plus 1% Tween 20 and 5% blotting-grade milk (Bio-Rad Laboratories) at 4°C on a shaker overnight. The membrane was then washed three times with PBS plus 1% Tween 20 and incubated with anti-rabbit IgG at 1:3,000, anti-mouse IgG at 1:3,000, and anti-rat IgG at 1:3,000 dilution, respectively, for 45 minutes. The membrane was developed using the Western Lightning Plus-ECL kit (PerkinElmer) for 1 minute, exposed to X-ray film (catalog no. E3012, Denville Scientific), and processed in the SRX-101A medical file processor (Konica Minolta Medical and Graphic Inc.).

Resting and activated CD3⁺ T cells (approximately 1.5 × 10⁶ cells) were blocked with TruStain fcX (catalog no. 101320, BioLegend), stained with mAb specific for Zombie Violet (catalog no. 423108, BioLegend) and FasL (clone MFL3, BioLegend), fixed and permeabilized using the Cytofix/Cytoperm Plus Kit (BD Biosciences) according to the manufacturer's instructions, and stained with mAbs specific for granzyme B (clone GB11, BioLegend), perforin (clone eBioOMAK-D, eBiosciences), and IFNγ (clone XMG1.2, BioLegend). Aliquots of fresh tumor digestion mixtures were incubated with mAbs specific for Zombie Violet, CD45.2 (Clone 104, BioLegend), and CD8 (BioLegend, clone 53-6.7). Zombie⁻ live cells and CD45⁺ cells were gated and analyzed for CD8⁺ T cells. The stained cells were analyzed using LSR Fortessa cytometer and FACSDiva software (BD Biosciences). Data were analyzed with FlowJo V10 program (Flow Jo).

Deidentified human MSS (n = 9) and MSI (n = 8) colon carcinoma specimens were obtained from the Cooperative Human Tissue Network (CHTN) Southern Division (University of Alabama, Birmingham, AL). Human tonsil tissue was obtained from Georgia Cancer Center Biorepository (Augusta, GA) and used as a positive control for CD8⁺ T-cell staining. IHC staining was carried out by the Georgia Esoteric Molecular Labs, and the CD8 antibody used was obtained from Dako (clone C8/144B, Dako). The tissues were then probed with mouse-specific IHC polymer detection kit (catalog no. ab209101, Abcam) according to the manufacturer's instructions. The stained tissues were counterstained with hematoxylin (Richard-Allan Scientific). Tumor tissues stained without the primary antibody were used as negative controls. CD8⁺ T-cell infiltration was graded by a board-certified pathologist (N.M. Savage) as low, medium, and high, and was also quantified by double-blind counting of CD8⁺ T cells per view field for each stained specimen.

Purified CD3⁺ T cells and the bead-bound isolated tumor-infiltrating CD8⁺ T cells, as described above, were fixed with 1% formaldehyde (catalog no. BP531, Thermo Fisher Scientific) in RPMI1640 medium plus 10% FBS for 10 minutes. Chromatin immunoprecipitation (ChIP) analyses were performed using anti-H3K9me3 (catalog no. ab8898, Abcam) and the Simple ChIP Plus Enzymatic Chromatin IP Kit (catalog no. 9004, Cell Signaling Technology) according to the manufacturer's instructions. The H3K9me3-bound DNA of Gzmb, Prf1, Faslg, Ifng, and Nxf2 gene promoters was quantified by qPCR in triplicates using gene promoter–specific primers (Supplementary Table S1).

All statistical analysis was performed using SAS 9.4 (SAS Institute Inc.), and statistical significance was assessed using an alpha level of 0.05. A repeated-measures mixed model with fixed effects of treatment groups and day, as well as all two- and three-factor interactions between these effects, was used to examine changes in tumor size over time. A one- or two-factor ANOVA with fixed effects of F5446 and anti–PD-1 groups and the two-factor interaction of these effects was used to examine differences in mean tumor size and tumor weight. A Tukey–Kramer multiple comparison procedure was used to examine post hoc pairwise differences between the groups. Statistical analysis was also performed using two-tailed t test. EC₅₀ was calculated using GraphPad Prism 6.0 (GraphPad Software).

Human MSS and MSI colon carcinoma specimens were analyzed by IHC staining of CD8⁺ T cells. As expected, all eight MSI colon carcinoma specimens exhibited medium to high infiltration of CTLs. However, eight of the nine MSS colon carcinoma specimens also exhibited medium infiltration by CTLs (Fig. 1A). No statistically significant difference in CTL tumor infiltration was observed between MSI and MSS colon carcinoma specimens (Fig. 1B).

The above observation that no significant difference in CTL infiltration exists between MSS and MSI colon carcinomas suggests that a lack of CTL infiltration might not be the sole mechanism underlying colon cancer immune escape. It is, therefore, possible that the functional status of the tumor-infiltrating CTLs is also responsible for colon cancer immune escape. We screened the TCGA database for altered gene expression between human colon carcinoma tissues and normal colon tissues to identify genes with known function in regulating CTL functions. We found that the expression of SUV39H1, which encodes H3K9me3-specific histone methyltransferase, recently shown to regulate CTL effector expression (24), was significantly elevated in the tumor tissue (Fig. 2A).

Next, we sought to determine Suv39h1 expression in tumor-infiltrating CTLs in the MC38 colon carcinoma mouse model. CD45⁻ tumor cells and CD8⁺ CTLs were isolated from fresh MC38 tumor tissues and analyzed for Suv39h1 mRNA levels. As in the human colon carcinoma tissues, tumor-infiltrating CTLs, as well as tumor cells, had high expression of Suv39h1 in the MC38 tumor model (Fig. 2B).

SUV39H1 catalyzes H3K9me3 (25, 26). Analysis of H3K9me3 ChIP-Seq data in the GEO Database revealed that H3K9me3 is enriched in the promoter regions of GZMB, PRF1, FASLG, and IFNG in human T cells (Fig. 3A). To validate this finding in the mouse T-cell genome, we purified CD3⁺ T cells from C57BL/6 mouse spleens and performed ChIP analysis using an H3K9me3-specific antibody and PCR primers that covered the promoters approximately from −2,000 to +1,000 relative to the transcription start sites of these four effector genes (Fig. 3B). ChIP analysis of these four individual gene promoter regions indicated that H3K9me3 was enriched in the promoter regions of all 4 effector genes in resting mouse T cells (Fig. 3B). These observations indicate that the expression of T-cell effector genes might be silenced by an H3K9me3-mediated epigenetic mechanism in resting T cells.

The above observations that tumor-infiltrating CTLs expressed high SUV39H1 and that H3K9me3 was enriched in the effector gene promoters in quiescent T cells suggest that effector gene expression may be repressed by the SUV39H1–H3K9me3 axis in tumor-infiltrating CTLs. Therefore, targeting the SUV39H1–H3K9me3 axis may increase effector gene expression and thereby, suppress tumors without the need to activate CTLs in the tumor microenvironment. To develop a SUV39H1-specific inhibitor, the molecular docking program Autodock Vina (21) was used to screen the NCI small chemical virtual library based on the SET domain structure of human SUV39H1 protein. The top 43 hits were selected and tested for inhibition of human SUV39H1 enzyme activity in vitro (Fig. 4A). Compound A8 and H1 exhibited inhibitory activity against recombinant SUV39H1. These two compounds were further tested at Reaction Biology Corp. Compound A8 [8-O-Benzyl 2-O-methyl 3-(4-methylphenyl) sulfonyl-4,5-dioxo-6H-pyrrolo(3,2-e) indole-2,8-dicarboxylate, C₂₇H₂₀N₂O₈S] was then selected and structurally modified. A chemical synthesis procedure was developed for this A8 derivative, termed F5446 (Fig. 4B). F5446 has a molecular weight of 552.95 g/mol (Fig. 4C) and exhibited an EC₅₀ of 0.496 μmol/L in an in vitro enzymatic activity assay (Fig. 4D). The ¹H and ¹³C spectra (Supplementary Figs. S1 and S2) and HPLC/MS data of F5446 (Supplementary Fig. S3) confirmed the structure and composition.

To determine the in vivo tolerability of F5446, mice were dosed every two days with vehicle or 10 or 20 mg/kg body weight F5446 for a total of 7 doses. Observations of body weights during the 2-week dosing period indicated mice in the 20 mg/kg group lost weight on average (<10%). The 10 mg/kg body weight group maintained a stable body weight throughout the dosing period, whereas the vehicle-treated control group increased average body weight by approximately 10%. These data are summarized in Supplementary Fig. S4 and Supplementary Table S2. Complete blood counts indicated no significant differences between vehicle and the F5446 treatment groups (Supplementary Table S3). Serum chemistry data suggested no difference between vehicle and drug-treated groups with all markers, except amylase and lipase, which were reduced, although still within normal levels in the F5446 treatment groups (Supplementary Table S4).

H3K9me3 is a repressive chromatin marker and often silences gene expression (24, 25, 27). We, therefore, reasoned that inhibition of SUV39H1 by F5446 would decrease H3K9me3 at these effector gene promoters, thereby, unsilencing the gene expression. CD3⁺ T cells were cultured in anti-CD3/CD28-coated plates in the absence or presence of F5446 and analyzed for H3K9me3. Stimulation of T cells decreased H3K9me3 at the promoter regions of Gzmb and Prf1 (Fig. 5A). F5446 treatment further reduced H3K9me3 in the promoter regions of Gzmb, Prf1, Faslg, and Ifng in activated T cells (Fig. 5A). As a negative control, we analyzed the effect of F5446 on H3K9me3 in the Nxf2 promoter in T cells because the Nxf2 gene promoter has no H3K9me3 deposition in resting human T cells (Fig. 5B). No H3K9me3 deposition was detected in Nxf2 gene promoter region in resting and activated mouse T cells, and F5446 did not alter H3K9me3 in the Nxf2 gene promoter in activated T cells (Fig. 5B). To determine whether reduced H3K9me3 leads to increased gene expression, the expression of the four effector genes was analyzed. Consistent with decreased H3K9me3 in their promoter regions, F5446 treatment increased the expression of all four effector genes in a dose-dependent manner (Fig. 5C). F5446 significantly increased the percentage of granzyme B-, IFNγ-, and FasL-positive T cells (Fig. 5D). F5446 treatment also significantly increased the protein levels (MFI) of granzyme B, IFNγ, and FasL in the T cells (Fig. 5E). Finally, F5446 treatment increased perforin protein in activated T cells, albeit to a small degree (Fig. 5F). Taken together, we conclude that H3K9me3 directly represses Gzmb, Prf1, Faslg, and Ifng transcription in T cells.

IFNγ and lymphocytes are the two primary components of the host cancer immunosurveillance system (28). The perforin–granzyme and FasL–Fas cytotoxic pathways are the two primary effector mechanisms by which T lymphocytes induce apoptosis of tumor cells (29–31). Colon tumor-infiltrating CTLs expressed high SUV39H1 in the tumor microenvironment (Fig. 2B). We, therefore, hypothesized that targeting SUV39H1 should increase effector expression in the tumor-infiltrating CTLs to suppress tumor growth in an immune competent host. To test this hypothesis, we used the MC38 colon tumor mouse model. MC38 is immunogenic and responds to anti–PD-1 immunotherapy (32). We, therefore, included anti–PD-1 immunotherapy as a positive control. MC38 tumor–bearing mice were treated with IgG, F5446, anti-PD-1, or F5446 + anti–PD-l. Both F5446 and anti–PD-1 treatments suppressed the established MC38 tumor growth in a time-dependent manner (Fig. 6A and B). Both F5446 and anti–PD-1 treatments significantly reduced tumor growth as measured by tumor size and tumor weight (Fig. 6C). However, F5446 and anti–PD-1 did not show additive or synergistic effects (Fig. 6C). To determine whether inhibition of SUV39H1 increases T-cell effector gene expression in vivo, tumor tissues were analyzed for the expression of the four effector genes. qPCR analysis revealed that F5446 treatment significantly increased the expression of granzyme B, perforin, FasL, and IFNγ (Fig. 6D).

To validate the above finding, we then tested F5446 in the complimentary CT26 colon tumor mouse model. CT26 tumor–bearing mice were treated as the MC38 tumor–bearing mice. Similar to observations in the MC38 tumor model, both F5446 and anti–PD-1 treatments suppressed the established CT26 tumor growth in a time-dependent manner (Fig. 6E and F). Both F5446 and anti–PD-1 treatments significantly reduced tumor growth as measured by tumor size and tumor weight (Fig. 6G). However, F5446 and anti–PD-1 did not show additive or synergistic effects (Fig. 6G). Analysis of total tumor tissues revealed that F5446 treatment also significantly increased the expression of granzyme B, perforin, FasL, and IFNγ (Fig. 6H).

The above observations indicate that F5446 inhibits SUV39H1 to increase T-cell effector gene expression to suppress tumor growth. To validate this, MC38 tumor cells were injected to Rag1 KO mice, and the tumor-bearing mice were treated with F5446. The tumor suppression efficacy of F5446 disappeared in the Rag1 KO mice (Fig. 7A). A complimentary approach was then used to further determine the role of CTLs in F5446-mediated tumor suppression. MC38 tumor–bearing mice were treated with F5446 in the absence or presence of CD8 neutralization antibody. Again, the efficacy of F5446 in tumor suppression was abolished when CD8⁺ CTLs were neutralized in the tumor-bearing mice (Fig. 7B). F5446 did not alter the level of tumor-infiltrating CTLs in both MC38 and CT26 tumors (Fig. 7C).

We then isolated CD8⁺ CTLs from tumor tissues of the control and the F5446-treated groups. ChIP analysis of H3K9me3 in the promoter regions of Gzmb, Prf1, Faslg, and Ifng indicated that H3K9me3 is enriched in the promoter regions of these four T-cell effector genes (Fig. 7D). Treatment with F5446 decreased H3K9me3 in the promoter regions of all four effector genes in the tumor-infiltrating CTLs (Fig. 7E). As a negative control, F5446 did not alter H3K9me3 deposition at the Nxf2 promoter region in T cells (Fig. 7F). qPCR analysis determined that the expression of granzyme B, perforin, FasL, and IFNγ were significantly increased by F5446 treatment (Fig. 7G). As a quality control, the purified CD8⁺ CTLs were analyzed by Western blot analysis, and the purified cells confirmed to be CD8⁺ T cells (Supplementary Fig. S5). Taken together, these data indicate that F5446 inhibits SUV39H1 to increase the expression of granzyme B, perforin, FasL, and IFNγ in tumor-infiltrating CTLs, resulting in enhanced CTL lytic function to suppress tumor growth.

The H3K9me3-specific histone methyltransferase SUV39H1 plays an essential role in epigenetic regulation of gene silencing in T cells under physiological conditions (24, 33, 34), as it regulates constitutive heterochromatin (25, 27, 35). We observed significantly elevated SUV39H1 expression in human colon carcinoma, suggesting that the SUV39H1–H3K9me3 pathway may play a critical role in silencing gene expression in tumor-infiltrating T cells and tumor cells in the tumor microenvironment under pathological conditions (36–40). To target SUV39H1, we screened the NCI small-molecule virtual library and developed an SUV39H1-specific small-molecule inhibitor F5446. Chaeotocin and verticillin A are natural products that inhibit the enzymatic activity of several histone methyltransferases including SUV39H1 (37, 41). Although very potent as histone methyltransferase inhibitors, chaetocin and verticillin A have multiple targets. Because of their structural complexity, synthetic procedures have not been developed for either of them. We chemically synthesized F5446 and confirmed its composition using an array of analytic methods. F5446 is a small-molecule inhibitor of SUV39H1and exhibited potent and specific tumor suppressor activity in vivo. Our in vivo tolerability and efficacy studies demonstrated F5446 can be administered at an efficacious dose of 10 mg/kg body weight without severe toxicities in mice.

SUV39H1 expression in tumor-infiltrating CTLs was as high as in the tumor cells. It is known that SUV39H1-H3K9me3–mediated gene silencing is critical for CD4⁺ T-cell lineage differentiation under physiological conditions (33, 34). SUV39H1 represses cytokine gene expression through H3K9me3-dependent transcription silencing to control Th1/2 cell lineage commitment and Th17 cell differentiation (33, 34). It is also reported that SUV39H1-dependent H3K9me3 controls the expression of a set of stem cell–related genes in CD8⁺ T cells and that IFNγ⁺ and granzyme B⁺ cells decrease in SUV39H1-deficient mice compared with wild-type mice (24). In this study, we observed H3K9me3 enrichment in the promoter regions of genes encoding granzyme B, perforin, FasL, and IFNγ in resting human and mouse T cells, which is consistent with the inactivation of transcription of these four effector genes in resting T cells. However, the repressive epigenetic silencing marks are diminished to create a transcriptionally active chromatin conformation in activated T cells, thereby, allowing activation of these effectors under physiologic conditions (42). F5446 treatment did not alter the tumor-infiltrating CTL levels in both MC38 and CT26 colon tumor models. Further studies are needed to determine whether SUV39H1regulates survival and differentiation of subsets of CTLs in the tumor microenvironment (24). However, it is clear that SUV39H1 regulates expression of CTL effector genes in the tumor microenvironment. Our observations that tumor-infiltrating CTLs expressed high SUV39H1 indicate that these tumor-infiltrating CTLs were not functionally activated, or at least only partially activated in their lytic function. The tumor microenvironment consists of PD-L1⁺ tumor cells, MDSCs, immunosuppressive cytokines, chemokines, metabolic factors, and other immunosuppressive cells. These immunosuppressive cells and factors can directly suppress T-cell activation and function through PD-L1–independent mechanisms (11, 43–45), thereby reducing tumor-infiltrating CTL activity and preventing suppression of immunogenic colon cancers. Therefore, targeting tumor-infiltrating CTL effector gene expression is a potentially effective approach to enhance the antitumor immune response.

In summary, H3K9me3 was enriched in the promoter regions of T-cell effector genes for granzyme B, perforin, FasL, and IFNγ in resting T cells and tumor-infiltrating CTLs. IFNγ and lymphocytes are the two key components of the host cancer immunosurveillance system (28). T cells, particularly CD8⁺ CTLs, are the major lymphocytes that suppress tumor development (46). CD8⁺ CTLs use the perforin–granzyme and FasL–Fas cytotoxic pathways as the two primary effector mechanisms to induce tumor cell apoptosis (15, 29–31, 47–49). F5446 treatment decreased H3K9me3 at the promoter regions of all four effector genes and increased their expression in activated T cells in vitro and in vivo. Treatment of mice bearing colon tumors with F5446 significantly increased the expression of granzyme B, perforin, FasL, and IFNγ and suppressed established tumor growth in a CD8⁺ CTL-dependent manner to a degree comparable with the anti–PD-1 immunotherapy (32). However, F5446 and anti–PD-1 showed no additive or synergistic effects on suppression of colon tumor growth in vivo, suggesting that F5446 and anti–PD-1 may use similar mechanisms to inhibit tumor growth. Nevertheless, our data document that colon tumor-infiltrating CTLs are partially inactivated by a PD-L1–independent mechanism in the tumor microenvironment and that targeting the SUV39H1–H3K9me3 pathway is potentially an effective mechanism to augment effector gene expression in tumor-infiltrating CTLs to suppress tumor growth.

No potential conflicts of interest were disclosed.

Conception and design: C. Lu, I.K. Oh, A.H. Colby, M. Grinstaff, K. Liu

Development of methodology: C. Lu, D. Yang, J.D. Klement, I. Lebedyeva

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Lu, D. Yang, J.D. Klement, I.K. Oh, Z. Xie, S.K. Kulp, M.A. Phelps, T. Albers, I. Lebedyeva

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Lu, I.K. Oh, N.M. Savage, J.L. Waller, M. Grinstaff, Z. Xie, C.C. Coss, M.A. Phelps, T. Albers, K. Liu

Writing, review, and/or revision of the manuscript: C. Lu, N.M. Savage, J.L. Waller, A.H. Colby, M. Grinstaff, N.H. Oberlies, C. Pearce, M.A. Phelps, K. Liu

Study supervision: K. Liu

This work was supported by grants from NIH/NCI (CA182518 and CA133085, to K. Liu; CA016058, to M.A. Phelps; CA227433, to M. Grinstaff, N.H. Oberlies, and K. Liu; and CA221414, to C. Lu) and VA Merit Review (BX001962, to K. Liu).

We thank Dr. Kimya Jones at Georgia Esoteric Molecular Labs for IHC staining of tumor tissues and The Ohio State University Pharmacoanalytical and Comparative Pathology and Mouse Phenotyping Shared Resources for F5446 mouse tolerability studies.

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.