DNA hypomethylation can promote cancer development through activation of genes with oncogenic potential. Here, the authors found that CpGs in the promoter‐proximal DNA of ONECUT2 were hypomethylated in intestinal metaplasia and gastric cancers, and that hypomethylation was associated with ONECUT2 upregulation. Functional analysis demonstrated that ONECUT2 has oncogenic potential and could activate ACSL5, which is also expressed in intestinal metaplasia, suggesting that ONECUT2 and ACSL5 may cooperate to promote intestinal differentiation or development of gastric cancer. Taken together, the findings suggest that ONECUT2 and its downstream target ACSL5 could be used to develop early detection biomarkers and prevent gastric carcinogenesis.
acyl‐CoA synthetase long‐chain family member 5
caudal type homeobox 2
complement factor B
Chungnam National University Hospital
differentially methylated promoter
Gene Expression across Normal and Tumor tissue
Gene Expression Omnibus
hepatocyte nuclear factor 6
National Center for Biotechnology Information
ONECUT1 through ONECUT3
polymerase chain reaction
quantitative reverse transcription PCR
reduced representation bisulfite sequencing
short hairpin RNA
small interfering RNA
The Cancer Genome Atlas
transcription start site
University of California, Santa Cruz
Gastric cancer (GC) is currently the fourth most common cancer worldwide and is the second leading cause of cancer mortality globally, with over 700,000 deaths each year.11 Thus, a pressing need exists for useful biomarkers that define the malignancy potential of primary gastric tumors, inform prognosis, and could assist in establishing new therapeutic and preventive strategies for this disease. Gastric carcinogenesis can manifest only after a gradual accumulation of various genetic and epigenetic alterations. Among epigenetic alterations, there has been great interest in the direct hypermethylation of tumor suppressor genes.22 It has been commonly asserted that DNA hypomethylation can also promote cancer development through activation of genes with oncogenic potential,33 although examples of oncogenes that truly rely on DNA hypomethylation for activation in tumors have not been published. However, recent advances in global methylation profiling suggest that aberrant promoter hypomethylation is a frequent event in hematologic malignancies, such as a chronic lymphocytic leukemia.44
To pursue hypomethylation targets associated with gastric carcinogenesis, we previously performed a reduced representation bisulfite sequencing (RRBS) analysis using DNA samples from gastric mucosa (GM), intestinal metaplasia (IM) and GC cells from one patient with intestinal‐type GC (IGC).55 Regional hypomethylation was found in the promoter‐proximal DNA of ONECUT2 (OC2; also called hepatocyte nuclear factor 6B or HNF6B) in IM and GC cells. The ONECUT family in mammals comprises three members: OC1 (also called HNF6A), OC2 and OC3 . The encoded proteins activate genes involved in controlling cell differentiation in liver and pancreas as well as patterning of trigeminal sensory neurons.66 Although a role for OC2 in cancer is not well defined, it was recently reported that OC2 is associated with the development of colorectal cancers.77 More recent work showed that OC2 is hyperactivated in a substantial subset of human prostate cancers and therefore is a potential drug target for the metastatic phase of aggressive prostate cancers.88 The biological role, if any, of OC2 in GC has not yet been examined.
To address this shortcoming in knowledge, we investigated the molecular mechanism responsible for regulation of OC2 expression in GCs and elucidated its role in gastric carcinogenesis. Our results reveal that regional CpGs in the promoter‐proximal DNA of OC2 are predominantly hypomethylated in primary GCs and that the extent of methylation correlates negatively with OC2 expression. Functional analysis revealed that OC2 has oncogenic potential in GC cells and activates expression of acyl‐CoA synthetase long‐chain family member 5 (ACSL5), which is strongly expressed in IM but not GM. Our results suggest that OC2 and ACSL5 are markers for IM in the stomach that may play important roles in intestinal differentiation or GC development and may be useful as targets for prevention of GC development or as therapeutics for GC.
Sixteen GC cell lines (Fig. 4a), including SNU001 (RRID: CVCL_0099), SNU005 (CVCL_0078), SNU016 (CVCL_0076), SNU216 (CVCL_3946), SNU484 (CVCL_0100), SNU520 (CVCL_5072), SNU601 (CVCL_0101), SNU620 (CVCL_5079), SNU638 (CVCL_0102), SNU668 (CVCL_5081), SNU719 (CVCL_5086), AGS (CVCL_0139), KATOIII (CVCL_0371), MKN01 (CVCL_1415), MKN45 (CVCL_0434), MKN74 (CVCL_2791), were obtained from the Korean Cell Line Bank (http://cellbank.snu.ac.kr/main/index.html) and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% antibiotic‐antimycotic solution (Invitrogen, Carlsbad, CA). All experiments were performed with mycoplasma‐free cells. And all GC cell lines have been authenticated using STR profiling within the last 3 years. The origin and characteristics of 16 gastric cancer cell lines were summarized in Supporting Information Table S1. One hundred sixty frozen tumors paired with adjacent nontumor tissues were provided by the Biobank of Chungnam National University Hospital (CNUH), a member of the Korea Biobank Network. All samples were obtained with informed consent, and their use was approved by the Internal Review Board at CNUH.
Genomic DNA from GC cells or tissues was modified by sodium bisulfite using the EZ DNA Methylation kit (Zymo Research, Orange, CA). Bisulfite‐modified DNA was amplified in a 20‐μl reaction with primers (Supporting Information Table S2) for bisulfite sequencing region 1 (BS‐R1) around the transcription start site (TSS) and BS‐R2 in the promoter‐proximal DNA or exon 1 of OC2 (Fig. 1d). PCR products were cloned using the pGEM‐T Easy vector (Promega, Madison, WI), and 10 clones were randomly chosen for sequencing. The methylation percentage for each sample was calculated as the number of methylated CpG dinucleotides among the total number of CpGs.
Total RNA was isolated from GC cells or tissues using the RNeasy kit (Qiagen, Valencia, CA), treated with DNase I (Promega), and reverse‐transcribed with Superscript II reverse transcriptase (Invitrogen) as previously described.55 RT‐PCR for OC2 was performed as follows: 94°C for 5 min, followed by 35 cycles of 94°C for 30 sec, 64°C for 30 sec and 72°C for 30 sec, with a final cycle of 72°C for 7 min. β‐Actin served as the PCR control. The PCR products were analyzed on 1.5% agarose gels stained with ethidium bromide. The primer sequences for RT‐PCR are listed in Supporting Information Table S2. Real‐time qRT‐PCR for OC2 was performed using a C1000 Thermal Cycler (Bio‐Rad, Hercules, CA). cDNA (100 ng) was amplified as noted above by 45 cycles with 2× SYBR Green Supermix (Bio‐Rad). β‐Actin was amplified as a control. The relative amount of target mRNA was quantified using comparative threshold cycle (Ct) methods.
Four CpG sites in BS‐R2 (Fig. 1d ) were selected for the quantification of the extent of methylation. Bisulfite‐modified DNA (100 ng) was amplified by PCR in a 20‐μl reaction using 2× Dye Mix polymerase (Enzynomics, Daejeon, Korea) to yield a 199‐bp product using the primer set shown in Supporting Information Table S2. PCR was performed using an initial melting step of 94°C for 5 min, followed by 40 cycles of 94°C for 30 sec, 59°C for 30 sec and 72°C for 30 sec, with a final incubation at 72°C for 7 min. Pyrosequencing was performed as described99 using the sequencing primer in Supporting Information Table S2 and the PSQ HS 96A System (Biotage AB, Kungsgatan, Sweden).
Immunohistochemical staining for OC2 was performed with human gastric tissues containing GM, IM and GC regions using the streptavidin‐biotin labeling method after microwave‐assisted antigen retrieval. Briefly, formalin‐fixed, paraffin‐embedded 4‐μm‐thick sections were dewaxed in xylene, rehydrated through a graded series of alcohol, and placed in an endogenous peroxidase block for 15 min. Sections were then washed in water before antigen retrieval, placed in a citrate buffer and microwaved for 10 min. An OC2‐specific monoclonal antibody (Thermo Fisher, Waltham, MA; diluted 1:300) was then applied, and antibody binding was detected using an avidin–biotin–peroxidase complex kit (Vector Laboratories, Burlingame, CA) using diaminobenzidine tetrahydrochloride solution (Biogenex, San Ramon, CA).
Chromatin immunoprecipitation (ChIP) assays were performed with GC cells for ChIP‐PCR analysis or with OC2‐expressing and control cells following a protocol from the Myers' laboratory (http://hudsonalpha.org/myers-lab/protocols) with modifications. Specifically, cells were fixed with 1% formaldehyde, lysed, and sonicated using a Covaris M220 (Covaris, Woburn, MA). For ChIP‐PCR analysis, the sonicated lysates for GC cells were used by dividing the same amount into three tubes and 10% input. Normal Rabbit IgG (2 μg; Sigma‐Aldrich), Anti‐trimethyl‐Histone H3 (Lys4) (2 μg; Sigma‐Aldrich) and Anti‐trimethyl‐Histone H3 (Lys27) (2 μg; Sigma‐Aldrich) was prebound to 20 μl Dynabeads coupled with protein A or protein G (Invitrogen). One microliter of each sample was used, and PCR was performed in three regions around the OC2 TSS (Fig. 1d ). The PCR products were analyzed on 1.5% agarose gels stained with ethidium bromide. The primer sequences for RT‐PCR are listed in Supporting Information Table S2.
For ChIP‐seq analysis, the sonicated lysate for OC2‐expressing and control cells were used by dividing the same amount into two tubes and 10% input. Anti‐OC2 (10 μg; Proteintech, Rosemont, IL) was prebound to 50 μl Dynabeads coupled with protein A or protein G (Invitrogen). For OC2‐expressing and control cells, genomic libraries with ~250‐ to 400‐bp fragment sizes were generated with the input and OC2 ‐immunoprecipitated fragments using the TruSeq ChIP Sample Prep kit (Illumina, San Diego, CA) and sequenced using the NextSeq_500 (Illumina), producing 76‐bp single reads. Sequencing reads from ChIP‐seq were aligned on human reference genome (hg19) using Bowtie2 software (version 2.2.2).1010 ChIP peak calling was performed using a Hypergeometric Optimization of Motif EnRichment (Homer, version 4.8.2),1111 and ChIP peak annotation was performed using the annotatePeaks module of Homer. ChIP peaks within the −1.0 to +1.0 kb region relative to TSSs were considered OC2 binding sites in the promoters of corresponding genes. Consensus motifs were analyzed within a ±100 bp region on either side of the peaks.
Cells were seeded at a density of 1 × 106 per 100‐mm dish, cultured for 24 or 72 hr, and treated with 1 μM 5‐aza‐2‐deoxycytidine (5‐aza‐dC; Sigma‐Aldrich, St. Louis, MO) to achieve demethylation and then 0.5 μM trichostatin A (Sigma‐Aldrich) to inhibit histone deacetylase. After 2–5 days, cells were washed with PBS, and total RNA was extracted. Real‐time qRT‐PCR for OC2 was performed as described above. Independent triplicate experiments were performed.
To establish OC2‐expressing cell lines, the OC2 coding sequence was amplified from cDNA prepared from SNU719 cells by PCR using the primers XbaI‐OC2‐F and NotI‐OC2 ‐R (Supporting Information Table S2) and then cloned into vector pCDH‐CMV‐MCS‐EF1‐Puro. For lentivirus construction, 293T cells were cotransfected with MISSION Lentiviral Packaging Mix plus empty vector or OC2‐expressing vector. OC2‐knockdown (OC2‐KD) cells were established using TRCN0000013444 and TRCN0000013446 (OC2_sh#3 and OC2_sh#4, respectively; Sigma‐Aldrich), targeting OC2 mRNA. For lentivirus construction, 293T cells were cotransfected with MISSION Lentiviral Packaging Mix and control or OC2 short hairpin RNA (shRNA) vectors. Cells for ectopic expression or KD experiments were seeded onto six‐well culture plates. After 24 hr, supernatants containing the lentivirus were collected from the 293T cells, filtered, and applied to the target cells for lentiviral transduction. After 72 hr, the medium was changed to RPMI medium containing 1 μg/ml puromycin (Sigma‐Aldrich). After 2 weeks of puromycin selection, the change in expression was confirmed by western blotting and qRT‐PCR.
SDS‐PAGE was conducted using a Bio‐Rad western blotting system and a 10% acrylamide gel. Proteins were transferred to a polyvinylidene fluoride western blotting membrane (Sigma‐Aldrich) and probed with anti‐OC2 (Abcam, Cambridge, UK) followed by mouse anti‐rabbit IgG conjugated with horseradish peroxidase (Santa Cruz Biotechnology, Dallas, TX). Immunopositive bands were visualized using the enhanced luminescence image analyzer LAS‐4000 (FUJI Film, Tokyo, Japan).
For cell proliferation assays, 2 × 103 cells were plated in a 96‐well plate, and the proliferation over a given time course was measured with the EZ‐Cytox Cell Viability Assay kit (ItsBio, Seoul, Korea) using a microplate reader (Molecular Devices, San Jose, CA) at 450 nm. For colony‐forming assays, 5 × 102 cells were plated in a six‐well plate. After 2 weeks, colonies were stained with crystal violet (0.5% crystal violet, 3.7% formaldehyde, 30% ethanol) and counted.
Cell migration and invasion were examined in transwell chambers (6.5‐mm diameter, 8‐μm pore size; Corning, Corning, NY). Cells (2 × 104) were added to the upper chamber, which contained 0.5 ml RPMI 1640 medium supplemented with 10–20% fetal bovine serum and 1% antibiotic–antimycotic solution (Invitrogen). For invasion assays, the transwell membrane was coated with 1 mg/ml Matrigel. Cells were incubated for 24 hr at 37°C, and nonmigrating or noninvading cells were removed with cotton swabs. Cells migrating to or invading the bottom of the membrane were stained with crystal violet and counted. For each well, three microscopic fields were chosen randomly and counted.
Parental or OC2‐KD MKN01 cells (5 × 106 cells per mouse) were subcutaneously injected into nude mice (4–6 weeks old). Mice were monitored twice a week for tumor development. Tumor size was measured using Vernier calipers, and tumor volume was calculated according to the formula: volume = width2 × length/2. Experiments followed the Laboratory Animals Act, Animal Protection Act and Institutional Animal Care and Use Committee (IACUC) guidelines and were approved according to institutional guidelines.
Total RNA from 1 × 107 control and OC2 ‐expressing cells was extracted using the RNeasy Mini kit (Qiagen). An RNA sequencing library was prepared using the TruSeq RNA Sample Prep kit (Illumina), and sequencing was performed based on Nextseq500 platform (Illumina) to generate 75‐bp paired‐end reads. Sequenced reads were mapped to a human reference genome (hg19) using the STAR alignment tool (version 2.5.1),1212 and gene expression was quantified with the count module in STAR. The edgeR1313 package (version 3.12.1) was used to select differentially expressed genes based on the RNA‐seq count data. The value for each gene (in counts per million) was normalized by trimmed mean of M values normalization, set to 1, and log2‐transformed for further analysis.
To repress ACSL5, we used ACSL5 small interfering RNAs (ACSL5‐ siRNAs 1,001,692 and 1,001,688, Bioneer, Daejeon, Korea). Corresponding scrambled sequences for each siRNA were also designed using the siRNA Wizard Software 3.11414 and synthesized by Bioneer Company. For transfection, 2 × 105 MKN01 cells were plated in six‐well plates and incubated overnight. MKN01 cells were transfected with siRNAs and scrambled RNAs (100 nM) using Lipofectamine RNAiMax (Invitrogen) in Opti‐MEM medium (Thermo Fisher). After 24 hr incubation, the transfection medium was changed to complete medium. After 48 hr, KD was confirmed by qRT‐PCR.
Gene expression and 450K HumanMethylation BeadChip data for 230 primary GCs,1515 including subtype EBV (24), CIN (111), GS (50) and MSI (45), and 450K HumanMethylation BeadChip data for two normal tissues were downloaded from The Cancer Genome Atlas (TCGA) portal (https://portal.gdc.cancer.gov/) and the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/). Because the methylation data from normal tissue is limited in TCGA database, we obtained additional data for 10 gastric mucosa (four from GSE50192 and six from GSE31848) from public databases1616, 1717 and used them produced by same platform without any selection process. CpG island (CGI) coordinates were obtained from the University of California, Santa Cruz (UCSC) browser. ChIP‐seq and RNA‐seq data for OC2‐expressing cells are available from NCBI GEO (http://www.ncbi.nlm.nih.gov/geo/) via accession number GSE113045.
Previously, we identified 174 hypomethylated differentially methylated promoters (DMPs) using methylome data from laser‐capture microdissected GM, IM, and GC cells from a single patient with IGC.55 These included 91 GC‐specific and 83 early‐onset DMPs. Early‐onset hypomethylated DMPs indicate methylation differences of >twofold in IM and GC cells compared to GM. Of the 83 early‐onset DMPs, 21 were noted for the high frequency of methylated reads compared to total reads in GM (Fig. 1a). Among these, OC2 was a strong candidate for study of the development of GC, as the frequency of methylated reads to total reads was 0.66 in GM but only 0.02 and 0.05 in IM and GC, respectively (Fig. 1a ). Again, UCSC Genome Browser (hg19) for RRBS data revealed that CpG methylation signatures (purple vertical lines in Fig. 1c) at promoter‐proximal DNA of OC2 was detected only in GM cells, but it was disappeared in IM or GC cells (Fig. 1c), indicating that promoter‐proximal DNA of OC2 was methylated in GM, but demethylated in IM or GC.
Bisulfite sequencing (see strategy in Fig. 1d ) revealed that BS‐R1 was methylation‐free in all paired nontumors and tumors (left panels in Fig. 1e ). On the other hand, in BS‐R2 there were significant differences in methylation between nontumors and tumors (right panels in Fig. 1e), with average methylation approximately threefold higher in nontumors (37.71%) compared to tumors (12.21%). RT‐PCR data revealed that OC2 was silenced in nontumors but expressed in all tumors tested (Fig. 1f), suggesting that OC2 expression in tumors may be associated with CpG hypomethylation at promoter‐proximal DNA rather than the promoter region around the TSS.
From the RNA‐seq and HumanMethylation 450K BeadChip public databases (see Public data section in Supporting Information), we acquired methylation data for 242 stomach tissue samples, including 12 nontumor samples and 230 tumor samples, which were divided into four molecular subtypes:1818 Epstein–Barr virus‐positive (EBV), microsatellite instability (MSI), genomically stable (GS) and chromosomal instability (CIN). When we examined the CpG methylation level by the molecular subtype, we found that the greater variability in methylation on Exon 1 (BS‐R2) in the tumor samples was due to variability in each subtype (Fig. 2a ), indicating that the methylation on BS‐R2 was heterogeneous depending on subtype of gastric tumors. For example, the methylation level at two CpG sites (cg02250594 and cg20324206) on BS‐R2 was greatly increased in EBV subtype compared to that in normal, while in GS and MSI subtypes it was not much different to that in normal. On the other hand, the methylation level at the same CpG sites was clearly decreased in CIN subtype, which have mainly intestinal‐type GC histology.1515 When we tested the correlation between OC2 expression and methylation at four CpG sites on BS‐R1 and BS‐R2 in pooled tumors, the correlation coefficients (r = −0.28 for cg02250594 and r = −0.29 for cg20324206) for BS‐R2 were much higher than those (r = −0.17 for cg15892115 and r = −0.19 for cg06275813) for BS‐R1 (Fig. 2b). As it was tested at the same CpG sites only in CIN subtype tumors, the correlation coefficients for BS‐R2 (r = −0.21 for cg02250594 and r = −0.26 for cg20324206) were much higher than those (r = −0.07 for cg15892115 and r = −0.12 for cg06275813) for BS‐R1 (Fig. 2c). This results support the observation that CpG hypomethylation at promoter‐proximal DNA of OC2 is strongly associated with upregulation of OC2 mRNA in gastric tumors, especially in CIN subtype.
We next quantified the extent of methylation at four CpG sites (CG#30–#33) including the above cg02250594 on BS‐R2 by pyrosequencing (Fig. 1d) for 160 paired clinical tissues from the CNUH cohort and looked for a correlation with their qRT‐PCR data. Pyrosequencing revealed that the mean percentage of methylation at these four sites was 28.8 ± 7.6% in nontumors and 18.1 ± 10.6% in tumors, indicating a significant decrease in methylation in tumor samples (Fig. 2c, p < 2.2e−16). qRT‐PCR revealed that the mean percentage of OC2 expression was 13.29 ± 5.49% in nontumors and 16.70 ± 3.60% in tumors, indicating significant upregulation of OC2 in tumors compared to nontumors (Fig. 2d, p = 2.848e−09). An increase in OC2 expression was detected in 64% (102/160) of tumors, for which an increase in expression was defined as a >twofold increase in the values for tumors compared to paired nontumors. Additionally, there was a negative correlation between methylation at the four CpG sites and OC2 expression in tumors (Fig. 2e, r = −0.23, p = 4e−03).
We performed immunohistochemistry with anti‐OC2 to compare the amount of OC2 in nontumor, IM and GC tissues. The data revealed strong staining for OC2 in the nucleus but weak staining in the cytoplasm of goblet cells of IMs (Figs. 3b, 3f and 3j ) and IGCs (Figs. 3c, 3g and 3k), but OC2 was not detectable in paired normal tissue (Figs. 3a, 3e and 3i ). In addition, there is no staining in diffuse‐type GCs (DGCs; Figs. 3n and 3p ) or paired normal tissue (Figs. 3m and 3o). These results indicated that OC2 expression was repressed in normal cells but induced in IM and IGC cells, suggesting that OC2 may be associated with early development of IGC.
We performed RT‐PCR and bisulfite sequencing with GC cell lines in the same manner as in Figure 2. We also performed Western blot analysis to explore whether OC2 mRNA expression level correlates with its protein expression or not. The RT‐PCR and Western blot results revealed that the mRNA expression and protein expression varied among the 16 GC cell lines (Fig. 4a). Then we found that OC2 mRNA expression and protein expression levels were inconsistent with each other in some GC cell lines, such as SNU001, SNU016, SNU620, SNU638, AGS and MKN74, as shown in Figure 4a. For convenience, the 16 GC cell lines were divided into two groups based on OC2 protein expression (expressed or silenced), eight of these GC cell lines, which represent each group, were selected at random (Fig. 4b ), and used to examine the correlation between expression and methylation at BS‐R1 or BS‐R2. Bisulfite sequencing revealed that all CpG sites at BS‐R1 in all groups were not methylated (left panels in Fig. 4b ), similar to clinical tissues (left panels in Fig. 1e). On the other hand, the extent of BS‐R2 methylation depended on the OC2 expression group (right panels in Fig. 4b). In the OC2‐expressed group, SNU016 and SNU216 were almost methylation free (both, 1.43%), while SNU620 and MKN74 had heavy methylation at 78.48 and 93.14%. The OC2‐silenced group. In the OC2‐silenced group, SNU001, SNU638 and AGS had methylation ranging 21.42–96.85%, while SNU484 was almost methylation free (1.84%). Thus, we observed negative correlation between OC2 expression and BS‐R2 methylation in SNU016, SNU216, SNU001, SNU638 and AGS, but unusual pattern in SNU620, MKN74 and SNU484.
To characterize histone marks associated with chromatin remodeling in GC cell lines with unusual pattern, we performed ChIP‐PCR at the three regions shown in Figure 1d. The SNU620 and MKN74, in which OC2 expressed but had heavy BS‐R2 methylation, showed only H3K4me3 signatures (an active mark) at the three regions Ch‐R1, Ch‐R2 and Ch‐R3 (upper panels in Fig. 4c). On the other hand, the SNU484, in which OC2 silenced but had almost BS‐R2 methylation‐free, showed bivalent marks of H3K4me3 and H3K27me3 at the three regions (lower panels in Fig. 4c), suggesting that the bivalent marks may be indicative of an association with OC2 silencing. To determine whether OC2 expression is controlled epigenetically, we treated cells with the DNA methylation inhibitor 5‐aza‐dC and/or the histone deacetylase inhibitor trichostatin A. After treatment with either drug, OC2 expression was restored compared to untreated cells (Fig. 4d), indicating that OC2 expression is regulated by epigenetic mechanisms, including DNA methylation and/or histone modification.
We next examined the effect of ectopic OC2 expression in MKN74 or AGS cells, in which OC2 was repressed or weakly expressed (Fig. 4a). Western blotting and qRT‐PCR confirmed that OC2 was stably expressed in OC2‐transfected MKN74 (OC2‐MKN74) and AGS (OC2 ‐AGS) cells (Fig. 5a). Colony‐forming assays revealed that OC2 significantly induced colony formation in each of OC2‐MKN74 and OC2 ‐AGS cells (Fig. 5b). OC2 also significantly induced cell migration in both OC2 ‐transfected lines (Fig. 5c), invasion in OC2 ‐MKN74 cells (Fig. 5d ) and cell proliferation in both lines (Fig. 5e).
We examined whether OC2 expression in SNU216 or MKN01 cells could be knocked down by treatment of with one or two different shRNAs (Fig. 4a). Western blotting and qRT‐PCR confirmed that OC2 expression was significantly decreased in OC2 ‐KD SNU216 and MKN01 cells after treatment with either of the two shRNAs (Fig. 5f). Both shRNAs significantly decreased colony formation and cell proliferation in both OC2 ‐KD cell lines compared to control shRNA (Fig. 5g). One shRNA (sh#4) also significantly decreased the migration and invasion capacities of both OC2 ‐KD lines, whereas the other shRNA (sh#3) decreased migration and invasion only in MKN01 cells (Figs. 5h and 5i). These results indicated that shRNAs targeting OC2 mRNA could modulate the oncogenic potential of OC2. In addition, we examined the loss of function using OC2‐siRNAs in ectopic OC2 expression cells (OC2‐AGS). The result showed that OC2 was efficiently and specifically reduced in OC2‐AGS cells by OC2 ‐specific siRNAs compared to cells treated with the scrambled siRNAs (Supporting Information Fig. S4a ) and that potential of colony‐forming (Supporting Information Fig. S4b) and migration (Supporting Information Fig. S4c) were significantly reduced in OC2‐AGS cells by OC2 knockdown. Finally, an in vivo study revealed that tumors in mice xenografted with OC2 ‐KD cells were significantly smaller than those of mice xenografted with empty‐vector cells (Fig. 5k), demonstrating that tumorigenicity could be suppressed by shRNA targeting OC2 mRNA.
To further investigate the functional role of OC2 as a transcription factor, we performed ChIP‐seq and RNA‐seq analysis of OC2‐AGS and OC2‐MKN74 cells. ChIP‐seq analysis using Homer revealed that the consensus motif with the highest enrichment of a single OC2‐binding site was HNF6 (5′‐ATC/TGAT‐3′) in both cell lines and controls. That is, 76.51% (9,336) and 38.08% (652) of total OC2 peaks in OC2 ‐AGS and control (Supporting Information Fig. S1a) were significantly matched to the HNF6 motif (p = 1e−7,601 and 1e−50, left panel in Fig. 6a). In OC2‐MKN74 and control, 73.45% (74,543) and 52.02% (3,023) of total OC2 peaks (Supporting Information Fig. S1b) were significantly matched to the HNF6 motif (p = 1e−54,622 and 1e−1,678, right panel in Fig. 6a). We then focused on OC2 peaks at the HNF6 motif in only the promoter region (±1 kb of TSS), revealing a significant increase in OC2 peaks throughout the course of transfection of cells with OC2: 224 peaks in OC2 ‐AGC cells (Supporting Information Fig. S1c and Table S3) and 1,353 peaks in OC2 ‐MKN74 cells (Supporting Information Fig. S1d ; Supplementary Table S4) were significantly increased ≥fourfold compared to each control (p < 0.0001). Finally, there were 198 OC2 peaks in common between the 224 OC2‐AGS peaks and the 1,353 OC2 ‐MKN74 peaks (yellow circle in Fig. 6b).
RNA‐seq analysis showed that 507 genes, of which only 6 had common OC2 peaks at each promoter, were upregulated ≥twofold in OC2 ‐AGS cells compared to control (red circle in Fig. 6b ; Supporting Information Table S5). For OC2‐MKN74 cells, 414 genes, of which only five had common OC2 peaks at each promoter, were upregulated ≥twofold compared to control (blue circle in Fig. 6b ; Supporting Information Table S6).
Finally, we identified two genes, namely, ACSL5 and complement factor B (CFB), which were upregulated and had OC2 peaks at the promoter in OC2‐AGS and OC2 ‐MKN74 cells (Fig. 6b ). Using the Gene Expression across Normal and Tumor tissue (GENT) database,1919 we confirmed a positive correlation between the expression of OC2 and ACSL5 (r = 0.29, left in Fig. 6c) and between OC2 and CFB (r = 0.18, right in Fig. 6c) in 311 gastric tumor tissues. Using the UCSC Genome Browser, we found significant increases of OC2 peaks at the promoter region of ACSL5 and induction of ACSL5 expression in OC2‐AGS or OC2 ‐MKN74 cells (Fig. 6d). The induction of ACSL5 expression was confirmed in both OC2‐AGS and OC2 ‐MKN74 cells, as assessed with qRT‐PCR (Fig. 6e). To determine whether OC2 is an upstream regulator of ACSL5, we examined ACSL5 expression in each of OC2‐KD SNU216 and OC2 ‐KD MKN01 cells (Fig. 5f). qRT‐PCR revealed significantly decreased ACSL5 expression in OC2‐KD SNU216 and OC2 ‐KD MKN01 cells compared to control cells (Fig. 6f), suggesting that ACSL5 expression is dependent on OC2 activity.
To test the oncogenic potential of ACSL5 in vitro irrespective of OC2‐KD, we transfected MKN01 cells with two individual siRNAs targeting RNA sequences of ACSL5 (in those cells, ACSL5 expression had been induced; see Supporting Information Fig. S2). qRT‐PCR confirmed that ACSL5 expression was significantly decreased in ACSL5 ‐KD MKN01 cells after treatment with either of the siRNAs (Fig. 6g). siRNA#3 significantly decreased the colony‐forming potential of ACSL5 ‐KD MKN01 cells compared to scrambled siRNAs (Fig. 6h ). In addition, each of the siRNAs caused a decrease in the migration (Fig. 6i) of ACSL5‐KD MKN01 cells compared to cells treated with the scrambled siRNAs. These results demonstrated that siRNA‐mediated repression of ACSL5 inhibits the oncogenicity of MKN01 cells.
In our study, we demonstrated that CpG sites in the promoter region of OC2 were unmethylated in gastric tumors as well as nontumor cells but that regional CpG sites in OC2 exon 1 were aberrantly hypomethylated in tumors and that OC2 was also aberrantly upregulated in tumors. Notably, DNA hypomethylation occurred in IM as well as IGC, suggesting that OC2 expression may be induced during the precancer IM stage. Furthermore, we observed anti‐OC2 staining in IM as well as IGC, but not in normal mucosa or DGC. OC2 promoted GC cell proliferation and tumor growth in vitro and in vivo. To the best of our knowledge, our results are the first to indicate that regional CpG hypomethylation at OC2 exon 1 contributes to tumorigenesis as an early event in IGC development and that inhibition of this hypomethylation constitutes a potential therapeutic target for IGC.
To describe the relationship between promoter‐proximal DNA methylation and transcriptional efficiency, a model has been proposed with four alternative DNA methylation states.2020 First, methylation of the promoter region itself leads to robust silencing of the gene, as is frequently observed for CpG island promoters in cancer cells. Second, promoter‐proximal methylation (~300 bp from the TSS) may have a dramatic effect on transcription initiation, as a result of altering chromatin structure at the promoter. Third, heavy DNA methylation within ~1 kb from the TSS may have a modest effect on transcription elongation efficiency, perhaps by altering the chromatin structure of the promoter‐proximal region. Fourth, in the presence of distal DNA methylation (1–2 kb from the TSS), transcription may not be affected. We found that the promoter region of OC2 was completely unmethylated in all nontumor tissues and tumors, regardless of OC2 transcription level, suggesting that the promoter region (BS‐R1, −117 to +41 bp from TSS) may not be critical for regulating OC2 transcription. Among the above model, the second methylation state appears to reflect a more probable model for how OC2 transcription is regulated because CpG hypomethylation at the promoter‐proximal DNA of OC2 (BS‐R2, +209 to +685 bp from TSS) was associated with OC2 transcription in tumors.
Similarly, the OC2 promoter region in GC cell lines may not be critical for regulating transcription because, again, the promoter region was methylation‐free in all GC cell lines. However, the relationship between promoter‐proximal DNA methylation and transcriptional efficiency in the cell lines tested is less clear. For example, SNU620 cells had heavy methylation at OC2 promoter‐proximal region and expressed OC2 strongly, whereas SNU216 and SNU484 cells were methylation‐free at the OC2 promoter‐proximal region and expressed OC2 weakly or not at all. Thus, our data cannot explain the relationship between OC2 promoter‐proximal DNA methylation and OC2 transcription in GC cells. In the promoter‐proximal regions, on the other hand, we observed an H3K4me3 mark in GC cell lines with strong OC2 expression, but there was a bivalent H3K4me3/H3K27me3 mark in GC cell lines with weak or no OC2 expression. This result corresponds well with previous reports that such bivalent marks lead to transcriptional silencing and are common in adult tumor cells.2020, 2121 It is not known how heavy methylation at the OC2 promoter‐proximal DNA is associated with the H3K4me3 mark in SNU620, or whether methylation‐free promoter‐proximal DNA is associated with the bivalent mark H3K4me3/H3K27me3 in SNU216 or SNU484 cells. Further investigation is required to elucidate the unusual relationship between OC2 promoter‐proximal DNA methylation and chromatin structure.
The ONECUT transcription factors, including OC2 , are evolutionarily conserved proteins that play a role in the development of the liver, pancreas and neuronal system.2222, 2323, 2424 Although a role for OC2 in cancer is not well defined, it was recently reported that OC2 was aberrantly upregulated in a variety of cancer types, including hepatocellular carcinoma, prostate cancer, colorectal cancer and ovarian cancer.77, 2525, 2626, 2727 Recently, it has been proposed that during the multistep development human tumors acquire six hallmarks of cancer, including sustaining proliferative signaling, evading growth suppressors, resisting cell death, enabling replicative immortality, inducing angiogenesis and activating invasion and metastasis.2828 The conceptual progress has added two emerging hallmarks of reprogramming of energy metabolism and evading immune destruction to the list.2929 As underlying these hallmarks, genome instability and inflammation has been suggested.2929 In recent years, it has been suggested that specific shifts in chromatin configuration could occur during the acquisition of certain hallmark capabilities.3030 In our study, we demonstrated that OC2 acquired aberrant gain of function in GC as a consequence of specific epigenetic alterations and promoted GC cell proliferation and tumor growth in vitro and in vivo, thus suggesting that OC2 plays an important role during the acquisition of some hallmark capabilities in a variety of human cancers including GC. In addition, OC2 mRNA is a urine biomarker for the detection of PC,2626 and a risk‐associated genetic variant in prostate cancer patients has been associated with OC2 activity,3131 and OC2 acts as a master regulator and survival factor in aggressive prostate cancer variants.88 Moreover, OC2 is also a target for the microRNA miR‐429 or miR‐9, downregulation of which is associated with OC2 activation, which may contribute to the development of hepatocellular carcinoma2525 or colorectal cancer.77
Gastric carcinogenesis proceeds through a series of precursor lesions in the GM called Correa's cascade, comprised of multi‐atrophic gastritis, IM, dysplasia and finally GC.3232 In this process, IM is a trans‐differentiation of the gastric epithelium to yield an intestinal type, mostly induced by Helicobacter pylori infection and expression of homeobox genes, including caudal type homeobox 2 (CDX2 ).3333 There is emerging epidemiological evidence that IM may be reversible with long‐term follow up. For instance, a recent study showed that H. pylori eradication may reverse IM at least in a subset of patients and that reversibility may be associated with a decrease in CDX2 mRNA in patients with dysplasia as well as GC.3434 However, the results of earlier studies on the effects of H. pylori eradication for improving IM have been inconsistent.3535, 3636, 3737, 3838 Another recent study reported that selumetinib, an inhibitor of mitogen‐activated protein kinase, might reverse IM in a mouse model based on tamoxifen injection and lead to re‐establishment of normal gastric lineages.3939 Interestingly, in the normal gastrointestinal tract, ACSL5 in the small intestine is predominantly found in the villus epithelium and not in the GM, whereas in the stomach high levels of ACSL5 are found in the IM, suggesting that ACSL5 may play an important role in the progression of IM4040 and may be a target for reversing IM in the stomach. If ACSL5 is a downstream target of OC2, as shown in our present study, one would expect that its activity can be inhibited by modulation of OC2. In a mouse model of metastatic prostate cancer, OC2 was shown to be inhibited by CSRM617, a newly identified small molecule.88 Additionally, we confirmed a positive correlation between CDX2, OC2 and ACSL5 expression in 311 gastric tumor tissues (Supporting Information Fig. S3), suggesting that three IM markers may be associated with intestinal differentiation or development of IGC.
Taken together, our data indicate that OC2 is a novel IM marker that is hypomethylated longitudinally in IGC and its precancer lesion, IM. Further, we propose that OC2 and its downstream target ACSL5 could be used to develop early detection biomarkers and to prevent gastric carcinogenesis. Our findings contribute to the Pre‐Cancer Genome Atlas (PCGA), a concerted initiative to characterize the molecular alterations in premalignant lesions.4141 Further studies are required to clarify the exact role of ACSL5 in gastric carcinogenesis and to evaluate whether small molecules such as CSRM617 can modulate OC2 and thus ACLS5.
This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (no. 2017R1E1A1A01074883) and a Korea Research Institute of Bioscience and Biotechnology (KRIBB) Research Initiative grant.