H3K9 Trimethylation Silences Fas Expression To Confer Colon Carcinoma Immune Escape and 5-Fluorouracil Chemoresistance

Fas is a member of the death receptor superfamily. Despite other “nonapoptotic” cellular responses emanating from its signaling (1), the major and best known function of Fas is apoptosis (2–8). Fas is expressed on the tumor cell surface, and its physiological ligand, FasL, is expressed on activated CTLs. CTLs are essential for eliminating virus-infected cells and malignant cells in the host. CTLs destroy target cells primarily through one of two effector mechanisms: 1) perforin-mediated granzyme delivery following exocytosis of cytotoxic granule, and 2) FasL-induced apoptosis. Germline and somatic mutations or deletions of FAS or FASL gene coding sequences in humans lead to autoimmune lymphoproliferative syndrome (9, 10). Autoimmune lymphoproliferative syndrome patients also exhibited increased risk of both hematopoietic and nonhematopoietic cancers (9, 11). Furthermore, FAS and FASL gene promoter polymorphisms are associated with decreased Fas expression level and increased risk of both hematopoietic malignancies and nonhematopoietic carcinoma development in humans (12–14). The Fas protein level is high in normal human colon tissues. In human primary colorectal carcinoma (CRC), however, the Fas protein level is generally lower as compared with normal colon tissues, and complete loss of Fas protein is often observed in human metastatic CRC (15, 16). Furthermore, Fas-mediated apoptosis exerted by the CTLs is an important contributor of tumor regression, and acquisition of resistance to Fas-mediated apoptosis is linked to recurrence and adverse prognosis in human CRC patients (17, 18). These observations thus strongly suggest that human CRC cells use silencing Fas expression as a key mechanism to escape from host immune surveillance. The regulation of Fas expression has been subject of extensive studies, and it is clear that Fas expression is regulated by both transcriptional and epigenetic mechanisms (19–21). However, the molecular mechanism underlying FAS silencing in metastatic CRC cells (15, 16) remains to be determined. Furthermore, although Fas is a death receptor that mediates the extrinsic apoptosis pathway, interestingly, it has been shown that Fas also mediates colon carcinoma cell sensitivity to 5-fluorouracil (5-FU) (22, 23). 5-FU is the standard therapy for human CRC patients. However, acquisition of resistance to 5-FU is often inevitable in human CRC patients (24). Therefore, a novel chemotherapeutic agent that can effectively overcome metastatic human CRC 5-FU resistance is urgently needed.

Covalent modifications of DNA and histones, the two core components of eukaryotic chromatin, are the two major mechanisms of epigenetic regulation of gene expression. The methylation of lysine residues in histones, particularly in the N-terminal tails of histones H3 and H4 of the chromatin, play a fundamental role in the regulation of gene expression through modulating chromatin structure. Histone methyltransferases (HMTases) catalyze the methylation of histones to modify chromatin structure, thereby influencing gene expression patterns during cellular differentiation and embryonic development. Recent studies have firmly established a fundamental role of aberrant HMTase activity and human diseases, particularly human cancers (25). Unlike genetic mutations of oncogenes and tumor suppressor genes, which are permanent alterations in the cancer genome, histone methylation is a reversible process, which has made HMTases attractive molecular targets for cancer therapy. Thus, elucidation of the molecular mechanisms underlying HMTase-mediated tumor suppressor gene expression regulation and the use of HMTase inhibitors to induce re-expression of epigenetically silenced tumor suppressor genes can potentially lead to suppression of cancer growth or sensitization of cancer cells to specific therapeutic agents (25–29). DNA methyltransferase (DNMT) and histone deacetylase (HDAC) inhibitors have been under extensive development for human cancer therapy for the last two decades (30); in contrast, identification and development of HMTase inhibitors as therapeutic agents are still in their infancy (31–33). Furthermore, the specific HMTase targets associated with cancer progression remain to be determined.

In an attempt to identify new anticancer drug leads from nature, we screened wild mushroom extracts and identified verticillin A as a potent anticancer agent in vitro (34) and in vivo (35). In the present study, we report the discovery of verticillin A as a selective HMTase inhibitor that selectively inhibits SUV39H1, SUV39H2, G9a, and GLP to inhibit H3K9me3 in metastatic human colon carcinoma cells. Furthermore, we performed genome-wide H3K9me3 profiling in combination with genome-wide gene expression analysis using verticillin A as an H3K9me3 inhibitor, and identified FAS as a direct target of H3K9me3 in metastatic human colon carcinoma cells. SUV39H1, SUV39H2, and G9a/GLP have redundant functions in H3K9 trimethylation and FAS silencing in the metastatic human colon carcinoma cells. Our data indicate that H3K9me3-mediated FAS transcription silencing is a major mechanism by which colon cancer cells evade host cancer immune surveillance, and verticillin A is a new selective HMTase inhibitor that inhibits H3K9me3 to restore Fas expression to suppress colon cancer cell immune escape and chemoresistance.

Human CRC cell lines LS411N and SW620 were obtained from the American Type Culture Collection (Manassas, VA). The American Type Culture Collection has characterized these cells by morphology, immunology, DNA fingerprint, and cytogenetics. LS411N-5FU-R and SW620-5FU-R cells were selected from LS411N and SW620 cells, respectively, by culturing cells with increasing doses of 5-FU. Murine colon 26 cells were provided by Dr. William E. Carson, III (Ohio State University, Columbus, OH).

Cells were incubated in lysis buffer (20 mM Tris-HCl [pH 8], 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40) containing protease inhibitors for 30 min. Cells were then centrifuged for 5 min, and the pellet was resuspended in 90 μl 0.1 M HCl and incubated on ice for 30 min. The supernatant was neutralized with 10 μl 1 M Tris-HCl (pH 9.0). The extract was then resolved in standard protein gels, and the blot was probed with Abs specific for H3K9me2 (Active Motif) and H3K9me3 (Abcam). The blot was reprobed with anti-H3 (Cell Signaling Technology), which was used as the normalization control.

Verticillin A was tested in 10-dose IC₅₀ mode with 3-fold serial dilution starting at 10 μM at Reaction Biology (Malvern, PA). S-(5′-adenosyl)-l-homocysteine (SAH) was used as a positive control in each assay and tested in a 10-dose IC₅₀ mode with 3-fold serial dilution starting at 100 μM. Reactions were carried out at 1 μM [³H]S-adenosyl-methionine as substrate.

Genomic DNA was purified using a DNeasy tissue kit (Qiagen) according to the manufacturer’s instructions. Sodium bisulfite treatment of genomic DNA was performed using a DNA modification kit (Zymo Research) according to the manufacturer’s instructions. Methylation-sensitive PCR were carried out according to standard PCR procedures. The PCR primers are listed in Supplemental Table I.

The quality of RNA was analyzed on a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) and assured by a RNA integrity number of ≥7. The human gene 2.0ST array (Affymetrix, Santa Clara, CA), which covers 30,654 coding transcripts, was used for the gene expression profiling. Total RNA samples were processed using the Ambion WT expression kit (Life Technologies, Calsbad, CA) and a GeneChip WT terminal labeling kit (Affymetrix). The synthesized sense strand cDNAs were then fragmented and biotin-labeled using a GeneChip WT terminal labeling kit. The labeled cDNAs were hybridized onto the arrays using Affymetrix GeneChip fluidics station 450 systems according to the manufacturer’s protocol. The expression data were imported into Partek GS version 6.6 using a standard import tool with GeneChip robust multiarray analysis normalization. The differential expressions were calculated using ANOVA of Partek package.

Chromatin immunoprecipitation (ChIP) assays were carried out using anti-H3K9me3 Ab (Abcam), H3K9ac (Cell Signaling Technology), or RNA polymerase II (RNA Pol II; Santa Cruz Biotechnology) according to the manufacturers’ protocols. Fas promoter DNA was detected by PCR using promoter DNA-specific primers as listed in Supplemental Table I.

Native chromatin was prepared from LS411N-5FU-R cells and digested with micrococcal nuclease to generate mononucleosomes using the SimpleChIP enzymatic chromatin IP kit (Cell Signaling) according to the manufacturer’s instructions. Chromatin fragments were then incubated with anti-H3K9me3 Ab (Abcam) and the chromatin/Ab complexes were isolated using protein A–agarose and salmon sperm DNA (Millipore) according to the manufacturer’s instructions. The ChIP DNA fragments were then prepared using the TruSeq ChIP sample preparation kit (Illumina) and sequenced using HighSeq 2500. Sequence reads in the FASTQ format were generated by the Illumina pipeline CASAVA 1.8.2 and were analyzed using MACS 1.4.2 as described (36). Peaks were annotated using in-house scripts, and genes that have a peak in a promoter were fed into Panther to perform functional analysis.

Normalized read depths in wiggle files generated by MACS were used to generate window enrichment of 200 bp with step of 100 bp using an in-house script. The University of California, Santa Cruz genome graphs plotted the overall enrichment. The maximum enrichment of windows from 4000 bp upstream and 1000 bp downstream of transcription initiation site in a gene was correlated to gene expression. For each bin of gene expression, the distribution of ChIP sequencing (ChIP-Seq) enrichment was drawn in a box plot using R.

LS411N-5FU-R and SW620-5FU-R cells were plated at 1 × 10⁵ cells/well in 24-well plates. Cells were then treated with 0, 20, or 50 nM verticillin A for 24 h. Cells were harvested and stained with human anti-DR4 (TRAIL-R1, H5101, Alexis Biochemicals) or anti-DR5 (TRAIL-R2, H5201, Alexis Biochemicals) and analyzed using flow cytometry. Fas cell surface expression was analyzed in verticillin-treated or untreated cells using human anti-Fas Ab (BioLegend).

Cell viability assays were performed using the MTT cell proliferation assay kit (American Type Culture Collection) according to the manufacturer’s instructions.

Cells were treated with decitabine (0.5 μM), vorinostat (0.5 μM), verticillin A (20 nM), or MegaFasL (250 ng/ml) alone, or one each of these three epigenetic inhibitors in combination with MegaFasL for 24 h. Both floating and adherent cells were then collected and incubated with propidium iodide (PI) and annexin V and analyzed by flow cytometry. The percentage of apoptosis was calculated by the formula: % apoptosis = % PI⁺ and annexin V⁺ cells − % PI⁺ and annexin V⁺ cells in the control group.

A set of four human Fas-specific small interfering RNAs (siRNAs) were obtained from Qiagen. SW620-5FU-R cells were transiently transfected with these set of four siRNAs and analyzed for Fas expression by RT-PCR. One Fas siRNA (catalog no. S100000637) was used to silence Fas expression for functional studies. SW620-5FU-R cells were seeded onto 24-well plates at 2.5 × 10⁵ cells/well with scramble (200 nM) or Fas (200 nM) siRNA for 24 h. Cells were then harvested and replated at 2 × 10⁴ cells/well. Cells were then treated with or without verticillin A (20 nM) and MegaFasL (250 ng/ml) for 24 h. Cells were collected and analyzed for cell death by staining with annexin V and PI. For sensitivity to 5-FU and verticillin A, cells were treated with or without verticillin A (20 nM) and 5-FU (1 μg/ml) for 72 h and analyzed for cell death by staining with annexin V and PI and flow cytometry analysis. G9a-, SUV39H1-, and SUV39H2-specific siRNAs were obtained from Santa Cruz Biotechnology (catalog nos. sc-43777, sc-38463, and sc-106822). SW620-5FU-R cells were transfected as described above. Cells were then analyzed for G9a, SUV39H1, SUV39H2, and Fas expression using real-time RT-PCR. Acidic extracts were also prepared from these cells and analyzed by Western blotting using anti-H3K9me3 Ab (Abcam).

Mouse (BALB/c) was continuously anesthetized with isoflurane (1–3% in oxygen). A small abdominal incision at the right side near the rear leg was made and the cecum identified with sterile gauze. Tumor cells (2 × 10⁵ cells in 20 μl saline) were injected into the cecal wall from the serosal side using a sterile tuberculin syringe. The abdomen was closed with wound clips.

SW620-5FU-R cells were injected into nude mice s.c. at a dose of 3.5 × 10⁵ cells/mouse. Five days later, tumor-bearing mice were randomly divided into four groups. The tumor-bearing mice were injected i.v. with saline (control, n = 5), 5-FU (25 mg/kg body weight, n = 5), verticillin A (1 mg/kg body weight, n = 5), or 5-FU plus verticillin A (n = 5) at days 5, 7, 9, 11, and 13. Tumor size was measured after each treatment. Mice were sacrificed at day 15 after tumor transplant. Sera were collected and analyzed for liver enzyme profiles at the Georgia Laboratory Animal Diagnostic Service (Athens, GA).

Statistical analysis was performed using ANOVA and Student t test. A p value <0.05 was taken as statistically significant.

In a 3-y effort to identify natural anticancer agents, we identified the natural small molecule verticillin A (Supplemental Fig. 1) as an agent with potent activity against human cancer cells but low toxicity to normal human cells (35). Investigating verticillin A mechanism of action in metastatic human colon carcinoma cells identified BNIP3 as a target of verticillin A. We previously observed that verticillin A dramatically increases BNIP3 expression level in human colon carcinoma cells and that silencing BNIP3 decreased verticillin A–induced apoptosis in human colon carcinoma cells, suggesting that verticillin A induces tumor cell apoptosis at least partially through upregulating BNIP3 expression (35). However, the BNIP3 promoter DNA is hypermethylated in human colon carcinoma tissues and in human colon carcinoma cell lines (35, 37). Our initial hypothesis was that verticillin A acts as a DNA methylation inhibitor because it activates transcription of DNA methylation–silenced BNIP3 and a panel of other DNA methylation–silenced genes in human colon carcinoma cells (35). DNA methylation is carried out by DNA methyltransferases DNMT1, DNMT3a, and DNMT3b. To test this hypothesis, we then examined the effects of verticillin A on DNMT enzymatic activity in vitro and observed that verticillin A exhibited no inhibitory effect on the enzymatic activity of DNMT1, DNMT3a, or DNMT3b (Fig. 1A). Furthermore, verticillin A treatment did not result in DNA demethylation of the BNIP3 promoter region in human colon carcinoma cells (Fig. 1B). Therefore, our data suggest that verticillin A activates DNA methylation–silenced BNIP3 transcription through a DNA demethylation–independent mechanism.

To identify the molecular target of verticillin A, we screened the effects of verticillin A on HMTase activity. We used a 10-dose IC₅₀ mode with 3-fold serial dilutions starting at 10 μM verticillin A to determine whether verticillin A inhibits human histone methyltransferase activity. Among the known human HMTases tested, verticillin A selectively inhibits SUV39H1, SUV39H2, G9a, GLP (Fig. 2A), and to a lesser degree MLL1 and NSD2 (Table I). However, verticillin A, at the highest tested dose (10 μM) exhibited no inhibitory activity toward the other 15 HMTases (Table I). Our data thus suggest that verticillin A is a selective inhibitor for SUV39H1, SUV39H2, G9a, GLP, MLL1, and NSD2.

SUV39H1, SUV39H2, G9a, and GLP all mediate methylation of histone H3 at lysine 9 (H3K9) (38, 39). SUV39H1 and SUV39H2 have redundant functions and mediate di- and trimethylation of H3K9 within heterochromatin regions (38), whereas G9a and GLP form a heteromeric complex to catalyze H3K9 dimethylation (H3K9me2) (39). We therefore reasoned that if verticillin A inhibits SUV39H1/2 and G9a/GLP, then the global H3K9 methylation level should be decreased. To test this notion, LS411N and SW620 cells were treated with verticillin A and analyzed for overall H3K9 methylation. Histone acidic extracts were prepared from the untreated and treated cells and analyzed by Western blotting with H3K9Me2- and H3K9me3-specific Abs. H3K9me2 and H3K9me3 levels were diminished by verticillin A in the tumor cells (Fig. 2B).

To investigate H3K9me3 distribution in human CRC cells in a genome-wide unbiased fashion, we generated native chromatin from LS411N-5FU-R cells containing primarily mononucleosomes by micrococcal muclease digestion and performed ChIP with H3K9me3-specific Ab. The ChIP DNA fragments were sequenced with the Hi-Seq 2500 system. A total of ∼36 million sequence reads of H3K9me3 were generated and mapped to human genome by bioinformatics tools (Fig. 3A). It is clear that H3K9me3 markers are distributed all over the genome in a relatively even fashion without significant H3K9me3-enriched clusters in certain chromosome regions (Fig. 3A). Approximately 8000 H3K9me3 peaks were identified. Because verticillin A exhibits potent inhibitory activity against H3K9me3, ChIP-Seq analysis of verticillin A–treated cells is technically challenging and was not performed. Nevertheless, we mapped H3K9me3 to the human genome in human colon carcinoma cells.

To determine the genome-wide gene expression profiles, we performed genome-wide gene expression analysis in LS411N-5FU-R cells using an Affymetrix human gene 2.0 ST array. The entire dataset is available from the Gene Expression Omnibus database (accession no. GSE51262). To identify genes whose expression levels are regulated by H3K9 methylation, we also treated LS411N-5FU-R cells with verticillin A and analyzed the genome-wide gene expression. Genes whose expression was significantly changed as compared with untreated cells were identified by a combination of ANOVA (p < 0.01), intensity changes (≤1.5-fold), and minimum expression intensity (≥1.5). A total of 820 genes were identified. These verticillin A target genes were then functionally grouped. As expected, we identified a large set of genes with known functions in apoptosis and cell cycle regulation (Fig. 3C). The entire dataset is available in the National Center for Biotechnology Information Gene Expression Omnibus database (accession no. GSE51262: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE51262).

Because H3K9me3 is a hallmark of transcriptionally repressive chromatin (25), we reasoned that there should exist an inverse correlation between H3K9me3 level and gene expression in the genome-wide level. To test this hypothesis, we analyzed the expressed genes by quantitating the levels of H3K9me3 in the gene promoter region (1 kb downstream to 4 kb upstream of the gene transcription initiation site) and mRNA level as determined by genome-wide gene expression analysis as described above. Surprisingly, although a general trend of inverse correlation was observed, no statistically significant correlation was observed between H3K9me3 level and gene expression in the global level (Fig. 3B).

The lack of inverse correlation between H3K9me3 and gene expression in the genome scale is not totally surprising because some of the suppressive effects of H3K9me3 may be secondary. Additionally, some of the genes may be completely silenced by H3K9me3 and thus are not included for analysis. However, this inverse correlation should be true at least for some genes. Therefore, we then compared the gene expression and H3K9me3 profiles to identify individual genes whose expression is upregulated by verticillin A and whose promoter region is enriched with H3K9me3. FAS is identified as such a gene. ChIP-Seq analysis identified two H3K9me3 peak clusters in the FAS promoter region (Fig. 4A), and we confirmed these two H3K9me3 locations by individual ChIP analysis of the human FAS promoter region using H3K9me3-specific Ab (Fig. 4B). Fas is undetectable in the metastatic human colon carcinoma cells and is upregulated by verticillin A as revealed by DNA microarray (Fig. 3C). Furthermore, analysis of cell surface Fas receptor indicated that verticillin A significantly increased Fas receptor level (Fig. 4C) and Fas mRNA level (Fig. 4D) in the metastatic human CRC cells.

Next, we sought to determine the relationship between Fas expression and its promoter H3K9me3 level in human colon carcinoma cells. SW480 and SW620 are a pair of matched primary and metastatic human colon carcinoma cells. SW480 was established from primary colon carcinoma, whereas SW620 was derived from lymph node metastases of the same patients. Both cell lines were established prior to chemotherapy (40). The rationale is that if H3K9me3 controls Fas expression in human colon cells, then metastatic SW620 cells should have a higher degree of H3K9me3 level than the primary colon carcinoma SW480 cells. Flow cytometry analysis indicates that Fas is highly expressed in SW480 cells, but is silenced in SW620 cells (Fig. 4E). RT-PCR analysis indicated that Fas mRNA level is also higher in SW620 cells in SW480 cells (Fig. 4F). ChIP analysis shows that H3K9me3 level is significantly higher in both regions of the FAS promoter in SW620 cells as compared with that in SW480 cells (Fig. 4G). Thus, we have determined that H3K9me3-mediated FAS transcription silencing is a molecular mechanism of loss of Fas expression in metastatic human colon carcinoma cells.

We next sought to determine whether verticillin A affects H3K9me3 level in the FAS promoter region. SW620 cells were treated with verticillin A for 2 d and analyzed for H3K9me3 level. It is clear that verticillin A treatment decreased H3K9me3 level in the FAS promoter region SW620 cells (Fig. 5A). It is known that histone acetylation mediates Fas expression in embryonic fibroblasts (20), and H3K9ac is a mark for transcriptionally active chromatin (41). We therefore also analyzed H3K9ac level in the FAS promoter region in human colon carcinoma cells after verticillin A treatment. Interestingly, note that H3K9ac level is higher in the FAS promoter region in the primary human colon carcinoma SW480 cells than in the metastatic SW620 cells, and verticillin A dramatically increased H3K9ac level in the FAS promoter region in SW480 cells (Fig. 5B). Verticillin A also increased H3K9ac level in the FAS promoter region in the metastatic SW620 cells, albeit at a low degree (Fig. 5B). Our initial attempts did not identify an increase in the RNA Pol II level in the FAS promoter region in the human colon carcinoma cells after verticillin A treatment (Fig. 5C).

To determine which of the HMTases mediates H3K9me3-dependent Fas silencing in metastatic human colon carcinoma cells, we silenced SUV39H1, SUV39H2, and G9a expression using gene-specific siRNAs. Because G9a and GLP form a complex, we only silenced G9a expression. The expression levels of all three genes are effectively silenced by gene-specific siRNA (Fig. 6A). Silencing each of these three genes decreased H3K9me3 level in the metastatic human colon carcinoma cells (Fig. 6B), suggesting that SUV39H1, SUV39H2, and G9a function in catalyzing H3K9 trimethylation in human colon carcinoma cells. However, silencing each of these three genes only resulted in minimal increase in Fas expression in the metastatic human colon carcinoma cells (Fig. 6C). We then hypothesized that SUV39H1, SUV39H2, and G9a may have redundant functions in H3K9 trimethylation in the FAS promoter region. To test this hypothesis, we simultaneously silenced SUV39H1, SUV39H2, and G9a and analyzed Fas expression level. Indeed, silencing these three genes leads to a dramatic increase of Fas expression in the metastatic human colon carcinoma cells (Fig. 6D). We thus conclude that SUV39H1, SUV39H2, and G9a/GLP catalyze H3K9 trimethylation in the FAS promoter region to silence Fas expression in the metastatic human colon carcinoma cells.

The Fas–FasL effector mechanism plays a key role in cancer immune surveillance by host T cells (18). Our above observations thus suggest that colon carcinoma cells might use H3K9me3-mediated Fas silencing to escape from host cancer immune surveillance. To test this hypothesis, we examined colon carcinoma development in FasL-deficient mice. We first examined FasL expression in tumor-infiltrating CTLs. Tumor-infiltrating CTLs were gated and analyzed for FasL protein level on the cell surface. More than 60% of the tumor-infiltrating CTLs are FasL⁺ (Fig. 6A); as a negative control, CTLs from tumor-free mice are essentially FasL⁻ (Fig. 7A). Next, mouse colon carcinoma colon 26 cells were surgically transplanted into the cecal wall of syngeneic mice and observed for tumor growth. Significant fast tumor growth was observed in the FasL-deficient Fas^gld mice as compared with the wild-type mice (Fig. 7B). Because FasL is primarily expressed on activated CTLs under physiological conditions (42), our observations thus indicate that FasL-mediated cytotoxicity of host CTLs is a key component of the host cancer immune surveillance system against colon carcinoma development under physiopathological conditions.

Resistance to Fas-mediated apoptosis is a hallmark of human cancer in general (43) and of metastatic human CRC in particular (15, 44). It is known that epigenetic mechanisms including DNA methylation and histone acetylation modulate Fas transcription (20, 21). Therefore, we sought to determine the relative functions of these epigenetic mechanisms in terms of sensitization of Fas-resistant metastatic colon carcinoma cells to FasL-induced apoptosis. Sublethal doses of verticillin A, decitabine, and vorinostat were tested for their ability to sensitize tumor cells to FasL-induced apoptosis in vitro. As expected, both LS411N-5FU-R and SW620-5FU-R cells were resistant to FasL (Fig. 8A, 8B), and both decitabine and vorinostat increased the sensitivity of the tumor cells to FasL-induced apoptosis, albeit to a small degree under the conditions tested (Fig. 8B). Verticillin A exhibited significantly greater efficacy than either decitabine or vorinostat in sensitizing the two metastatic human CRC cell lines to FasL-induced apoptosis (Fig. 8A, 8B).

To determine whether verticillin A sensitized human colon carcinoma cells to FasL-induced apoptosis through Fas upregulation, SW620-5FU-R cells were transfected with scramble and Fas-specific siRNA. As expected, Fas-specific siRNA diminished Fas expression in the tumor cells (Fig. 8C). Treatment of these transfected cells with verticillin A and FasL revealed that silencing Fas expression diminished SW620-5FU-R cell sensitivity to apoptosis induction by verticillin A and FasL (Fig. 8D). Therefore, verticillin A sensitizes human colon carcinoma cells to FasL-induced apoptosis at least partially through upregulation of Fas expression.

The TRAIL-induced apoptosis pathway shares similar signaling pathways with the FasL-induced apoptosis pathway (44–46), and it plays an important role in suppression of tumor metastasis (47). We next analyzed the effects of verticillin A on TRAIL-induced apoptosis in human colon carcinoma cells. Verticillin A treated did not alter DR4 expression but increased DR5 expression level in both SW620-5FU-R and LS411N-5FU-R cells (Fig. 9A). Consistent with increased DR5 expression level, verticillin A significantly increased SW620-5FUR and LS411N-5FU-R cell sensitivity to drozitumab (Fig. 9B, 9C), a DR5 agonist mAb that has been developed to treat human cancers (44, 48, 49).

5-FU is the standard therapy for human CRC. However, acquisition of resistance to 5-FU is often inevitable in human CRC patients (24). Although Fas is a death receptor that mediates the extrinsic apoptosis pathway, interestingly, it has been shown that Fas also mediates CRC cell sensitivity to 5-FU (22, 23). Furthermore, 5-FU–resistant human CRC cells express similar level of H3K9me2 and H3K9me3 (Supplemental Fig. 2A) and are as sensitive to verticillin A as the parent cell lines (Supplemental Fig. 2B). Based on these observations, we hypothesized that in addition to its ability to sensitize colon carcinoma cells to FasL-induced apoptosis in vitro and FasL-mediated immune surveillance in vivo, verticillin A should also be effective in overcoming human CRC cell resistance to 5-FU. To test this hypothesis, LS411N-5FU-R and SW620-5FU-R cells were cultured in the presence of 5-FU, verticillin A, or both 5-FU and verticillin A and analyzed for apoptosis. It is clear that the combined verticillin A and 5-FU treatment induced significantly more apoptotic cell death than did either agent alone (Fig. 10A, 10B). To determine whether this finding can be extended to an in vivo situation, SW620-5FU-R cells were injected into nude mice s.c. Five days later, tumor-bearing mice were randomly divided into four groups. The tumor-bearing mice were injected i.v. with saline, 5-FU, verticillin A, and 5-FU plus verticillin A every 2 d for 10 d. At the dose used, 5-FU and verticillin A both exhibited tumor growth inhibitory activity (Fig. 10C). However, combined verticillin A and 5-FU exhibited the greatest inhibitory activity in terms of tumor size (Fig. 10D) and weight (Fig. 10E), and almost eradicated the established 5-FU–resistant tumor xenograft (Fig. 10C–E).

To determine whether verticillin A sensitized human colon carcinoma cells to 5-FU–induced apoptosis through Fas upregulation, SW620-5FU-R cells were transfected with scramble and Fas-specific siRNA. As expected, Fas-specific siRNA diminished Fas expression in the tumor cells (Fig. 8C). However, treatment of these transfected cells with verticillin A and 5-FU revealed that silencing Fas expression does not change SW620-5FU-R cell sensitivity to apoptosis induction by verticillin A and 5-FU (Supplemental Fig. 3). Therefore, verticillin A sensitizes human colon carcinoma cells to 5-FU–induced apoptosis through a Fas-independent mechanism.

To determine the toxicity of verticillin A, we measured serum liver enzyme profiles in mice after 5-FU therapy. Liver enzymes are diagnostic markers for liver damage. We collected serum from the above four groups of mice and measured liver enzyme and protein levels in the serum. Among all the enzyme analyzed in all four groups, only aspartate transaminase level was modestly increased (Table II). Therefore, the preliminary assays indicate that verticillin A has minimal liver toxicity at the efficacious dose.

Deregulation of H3K9me3 has been linked to various human cancers (50–52), and multivariate survival analysis showed that H3K9me3 levels are associated with cancer recurrence and poor survival rate in human patients with gastric cancer (53). We performed genome-wide H3K9me3 ChIP-Seq analysis and mapped H3K9me3 locations to the genome in the metastatic human colon carcinoma cells. Although H3K9 methylation often occurs within heterochromatin regions (38), interestingly, we observed that H3K9me3 marks exist relatively evenly all over the entire genome without significant clusters in a particular chromatin region in the metastatic human colon carcinoma cells (Fig. 3A). Our observations may thus suggest that H3K9me3 alone may not be a signature for heterochromatins at least in metastatic human colon carcinoma cells.

H3K9me3 is generally regarded as a mark of transcriptionally repressive chromatin (25). Therefore, it is expected that H3K9me3 marks should be inversely correlated with transcriptionally gene silencing. It is therefore a surprising observation that no correlations exist between gene promoter H3K9me3 and mRNA levels in the genome scale. It is possible that some of H3K9me3-regulated genes may be completely silenced to become undetectable and thus excluded in the correlation analysis. However, correlation analysis of H3K9me3 with mRNA expression levels in verticillin A–treated cells also did not find a correlation between H3K9me3 and gene expression in the genome scale in the metastatic human CRC cells. Therefore, it is possible that H3K9me3 alone may be insufficient to induce a heterochromatin structure to silence gene expression in the genome scale. Instead, H3K9me3 may act in concert with DNA hypermethylation to generate a transcriptionally repressive chromatin structure (54).

Although no genome-wide correlation was observed between H3K9me3 and mRNA expression, examination of individual gene promoter H3K9me3 and mRNA expression levels identified FAS as a target gene of H3K9me3. Two H3K9me3 clusters were identified in the FAS promoter region by genome-wide ChIP-Seq analysis and validated by ChIP analysis of H3K9me3 in the FAS promoter region in metastatic human colon carcinoma cells. DNA microarray study identified Fas as one of the genes whose expression was significantly upregulated by verticillin A, and we validated that Fas receptor is also significantly increased by verticillin A treatment on the surface of metastatic human colon carcinoma cells. It is known that Fas expression is regulated by its promoter DNA hypermethylation in certain human CRC cells (55). It is also known that Fas transcription is mediated by histone acetylation in embryonic fibroblasts (20). Furthermore, Fas expression can be upregulated by inhibition of DNA methylation and histone acetylation (21). In the present study, we show that Fas is silenced by its promoter H3K9me3. Moreover, inhibition of H3K9 methylation with verticillin A exhibits significantly greater apoptosis sensitization effects than does inhibition of DNA methylation or histone acetylation. Our data thus suggest that H3K9me3 acts as a dominant epigenetic mechanism over DNA methylation and histone acetylation by which the metastatic human colon carcinoma cells silence Fas expression (15).

Consistent with the notion that H3K9me3 is a dominant epigenetic mechanism underlying FAS transcriptional silencing in metastatic human colon carcinoma cells, verticillin A treatment dramatically decreased H3K9me3 level in the FAS promoter region in the metastatic human colon carcinoma cells (Fig. 5A). Interesting, although verticillin A treatment dramatically increased H3K9ac level in the primary human colon carcinoma SW480 cells, verticillin A treatment only induced minimal increase of H3K9ac in the metastatic human colon carcinoma SW620 cells. These observations suggest that lack of H3K9ac and high level of H3K9me3 at the FAS promoter region might be the key mechanisms underlying FAS silencing in metastatic human colon carcinoma.

RNA Pol II association with the FAS core promoter to form transcription preinitiation complex is very inefficient, but it supports fast transcription rate in human cancer cells (56). Therefore, FAS transcription rate may be controlled by the level of promoter-bound RNA Pol II and the polymerization rate of RNA Pol II. In this study, our initial attempts failed to identify increased RNA Pol II binding to the FAS promoter in the verticillin A–treated human colon carcinoma cells (Fig. 5C) even though verticillin A decreased H3K9me3 level that may likely create a more transcriptionally active promoter structure (Fig. 5A). The molecular mechanism underlying the effects of H3K9me3 on RNA Pol II association with the FAS promoter and FAS transcription in the metastatic human colon carcinoma cells requires further elucidation.

Systemic analysis of large cohorts of clinical specimens revealed that presence of tumor-specific CTLs in the tumor microenvironment is positively correlated with the absence of early metastasis and increased survival of CRC patients (17, 57, 58). We determined that tumor-infiltrating CTLs are FasL⁺, and CRC cells grow significantly faster in FasL-deficient mice in vivo, demonstrating a critical role of FasL of CTLs in immune suppression of CRC development under pathological conditions. Therefore, the fact that Fas expression is often completely lost in metastatic human CRC (15) and our finding that metastatic human colon carcinoma cells use H3K9me3 to silence Fas expression suggest that metastatic human colon carcinoma cells use H3K9me3 as a mechanism to silence Fas to evade the host CTL-mediated immune surveillance to progress and metastasize. Antitumor immune response is often suppressed by immune suppressor cells. However, recent studies have shown that the immunosuppressive regulatory T cells only selectively suppress the perforin pathway without inhibiting T cell activation in vivo (59, 60). Because FasL is primarily expressed on activated CTLs, the Fas-mediated effector mechanism thus should be still functional in the immunosuppressive tumor microenvironment. Therefore, pharmacological means to restore Fas expression in colon carcinoma cells can be an effective approach to sensitize tumor cells to CTL FasL-mediated cytotoxicity, thereby suppressing colon cancer immune evasion to suppress metastasis even under the immunosuppressive conditions.

We discovered verticillin A, a natural compound, as a selective HMTase inhibitor that is effective in restoring Fas expression in metastatic human CRC cells. Verticillin A selectively inhibits SUV39H1, SUV39H2, G9a, and GLP enzymatic activity with an IC₅₀ of 0.48–1.27 μM in vitro. It is known that SUV39H1, SUV39H2, G9a, and GLP all catalyze H3K9 methylation (38, 39). SUV39H1 and SUV39H2 have redundant functions and mediate di- and trimethylation of H3K9 within heterochromatin regions (38), whereas G9a and GLP form a heteromeric complex to catalyze H3K9 dimethylation (39). However, we observed that G9a also catalyzes H3K9 trimethylation to mediate FAS silencing in the metastatic human colon carcinoma cells (Fig. 6). In fact, SUV39H1, SUV39H2, and G9a all catalyze H3K9 trimethylation to cooperatively silence Fas expression in the metastatic human colon carcinoma cells.

Although verticillin A is effective in suppression of metastatic human CRC cell growth at a relatively high dose (i.e., 100 nM), verticillin A might be more valuable as a sensitizer to overcome human cancer cell resistance to apoptosis, and we observed that a sublethal dose of verticillin A effectively overcame metastatic human colon carcinoma cell resistance to FasL-induced apoptosis, and this sensitization is apparently Fas-dependent (Fig. 8C, 8D). Epigenetic agents, such as DNMT and HDAC inhibitors, were initially used at their maximally tolerated doses in clinical trials in human cancer patients, and they were often associated with extensive toxicity and minimal efficacy (61). Recent studies have shown that both clinical efficacy and tolerance improve when lower doses of these agents are used, and current clinical development efforts for DNMT and HDAC inhibitors are primarily focused on combining these agents with cytotoxic agents (29). HMTase inhibitors are much less studied as cancer therapeutic agents. In this study, we observed that H3K9me3 may play a dominant role over DNA methylation and HDACs in conferring metastatic human CRC cells to FasL-induced apoptosis. We observed that verticillin A can effectively inhibit H3K9me3 to reactivate FAS transcription. We envision that a sublethal and thus subtoxic dose of verticillin A may be used as an adjunct agent to restore Fas expression in human CRC cells to sensitize the tumor cells to CTL FasL-mediated cytotoxicity to suppress immune escape and to increase the efficacy of CTL immunotherapy to suppress CRC progression and metastasis.

TRAIL is a cancer-selective agent and has been shown to play an important role in suppression of tumor progression (46, 47). However, human clinical trials of TRAIL or TRAIL receptor agonist mAb have failed to advance beyond phase II because of limited efficacy. This outcome is expected because resistance to apoptosis is a hallmark of human cancer (62). Recent findings that modulating the TRAIL receptor complex to enhance signaling to overcome cancer cell TRAIL resistance to suppress cancer progression has induced new hope for TRAIL-based cancer therapy (63). In the present study, we observed that in addition to sensitizing metastatic human colon carcinoma cells to Fas-mediated apoptosis, verticillin A increased DR5 expression to effectively overcome metastatic human colon carcinoma cells to DR5-induced apoptosis. Therefore, verticillin A has the potential to be developed as an adjunct agent in TRAIL-based cancer therapy.

Alternatively, 5-FU is the standard and most commonly used therapeutic drug for patients with advanced and metastatic CRC (24, 64). During the last three decades, increased understanding of its mechanism of action has led to the development of strategies to improve its efficacy, primarily through combination with other therapeutic drugs (65, 66). However, despite these advances, the increase in clinical response to 5-FU has been minimal and cancer cell resistance remains a major challenge in 5-FU therapy (64). Our findings that 5-FU–resistant metastatic human CRC cells express similar levels of H3K9me3 and are sensitive to verticillin A suggest that 5-FU therapy does not alter verticillin A targets in the cancer cells. Moreover, a sublethal dose of verticillin A effectively increased 5-FU–resistant metastatic human colon carcinoma cells to 5-FU–induced apoptosis (Fig. 9A). This sensitization effect is unlikely due to increased Fas expression (Supplemental Fig. 3) and is likely due to altered expression of apoptosis-regulatory genes in the apoptosis signaling pathway (35). Verticillin A exhibited minimal toxicity to the liver in tumor-bearing mice in vivo. This outcome is also not surprising because verticillin A selectively targets HMTases to inhibit H3K9 methylation, an epigenetic event that is primarily involved in carcinogenesis and cancer progression but not in normal cellular activity. Therefore, verticillin A is potentially an effective and yet low toxic epigenetic agent for targeting 5-FU resistance in human patients with metastatic CRC, and thus holds great promise for further development as an “epi-drug” in human cancer therapy.

We thank Drs. Eiko Kitamura and Chang-Sheng Chang at the Georgia Regents University Cancer Center Integrated Genomics Facility for excellent technical assistance in genome-wide gene expression and ChIP-Seq analysis and Dr. A. Kaur and T. El-Elimat at the University of North Carolina at Greensboro for assistance in the purification of verticillin A.

The authors have no financial conflicts of interest.