Gene expression and DNA methylation data for gastric cancer patients were obtained from the GDC data portal (https://portal.gdc.cancer.gov/). Gene expression data for gastric cancer patients with survival information was downloaded from the GEO database (GSE26253) (21).

Paired gastric tumor and normal tissue samples (n=242) were collected from Chungnam National University Hospital (CNUH; Daejeon, Korea) with informed consent obtained from all patients and among them 171 samples have clinicopathological information. The present study was approved by the Internal Review Board of CNUH.

Gastric cancer cell lines (SNU-001, SNU-005, SNU-216, SNU-016, SNU-484, SNU-520, SNU-601, SNU-620, SNU-638, SNU-668, SNU-719, AGS, KATOIII, MKN1, MKN45 and MKN74) were obtained from the Korean Cell Line Bank (http://cellbank.snu.ac.kr/main/index.html) and were maintained in complete RPMI-1640 and DMEM medium (Welgene, Inc., Gyeongsan-si, Korea) at 37°C in a humidified 5% CO₂ incubator. Complete media were supplemented with 10% fetal bovine serum (FBS; Welgene, Inc.) and 1% antibiotic-antimycotic solution (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA). The HOXC10 full cDNA clone was amplified by RT-PCR (primer sequences are listed in Table SI) and inserted into the pCDH-CMV-MCS-EF1-Puro (CD510B-1) lentiviral vector. An empty vector was used as a control. For viral particle generation, cloned DNAs and lentiviral packaging Mix (Sigma-Aldrich; Merck KGaA, Darmstadt, Germany) were transfected into 293T cells using X-tremeGENE™ HP DNA Transfection reagent (Roche Diagnostics, Basel, Switzerland) in Opti-MEM media (Gibco; Thermo Fisher Scientific, Inc.). After 2 h of incubation, the media was changed to complete media. After 48 h of incubation, virus particles were collected from the supernatant. For viral infection, 2×10⁵ gastric cancer cells were added to each well of a 6-well plate and incubated overnight at 37°C. Viral supernatant was added to plated cells, and after incubation for overnight at 37°C, the media was changed to complete media. After 24 h, infected cells were selected using puromycin treatment. HOXC10 transfection was confirmed by RT-PCR and western blot analysis. The siRNA in this study was purchased from Bioneer Corp. (Daejeon, Korea). The antisense and sense sequences of the siRNA are the following. siCST1: CUAUAACGCAGACCUCAAU and AUUGAGGUCUGCGUUAUAG. A total of 100 nM siRNA was transfected with the Lipofectamine RNAiMAX reagent (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. One million cells were treated with 10 µM 5-aza-dC (Sigma-Aldrich; Merck KGaA) in a 100-mm dish for three days with media changes every 24 h following by harvesting.

Total RNA was extracted using the RNeasy Plus Mini kit (Qiagen, Venlo, The Netherlands) according to the manufacturer's instructions. Reverse transcription was performed using 1 µg total RNA as the template and iScript cDNA Synthesis kit (Bio-Rad Laboratories, Hercules, CA, USA). RT-PCR assays were performed using a PCR Master Mix (MGmed, Inc., Seoul, Korea) for 5 min at 25°C, for 30 min at 42°C, for 5 min at 85°C. Quantitative real-time PCR (qPCR) reactions were performed in triplicate on a CFX96™ Real-Time PCR Detection System (Bio-Rad Laboratories) using SsoAdvanced™ Universal Inhibitor-Tolerant SYBR-Green Supermix (Bio-Rad Laboratories) for 3 min at 95°C, and 40 cycles for 20 sec at 95°C, for 20 sec at 60°C, 20 sec 72°C. The relative expression of cDNA was normalized to β-actin, quantified using the 2^−ΔΔCq cycle threshold method (22), and at least three independent biological replicates were included for each reaction. Primers used for PCR were designed either manually or using the Primer3 program (http://bioinfo.ut.ee/primer3-0.4.0/). All primer sequences are listed in Table SI. Cells were lysed using 2X Laemmli sample buffer [4% SDS, 20% glycerol, 120 mM Tris-Cl (pH 6.8)] with a protease inhibitor cocktail (Roche Diagnostics). Supernatants were collected and quantified using bovine serum albumin (BSA) (Thermo Fisher Scientific, Inc.) by Bradford assay. A total of 30 µg proteins were loaded onto 12% polyacrylamide gels, electrophoresed and transferred to a polyvinylidene difluoride (PVDF) membrane (Roche Diagnostics) and probed with the HOXC10 antibody (dilution 1:1,000; rabbit polyclonal; cat. no. ab153904; Abcam, Cambridge, MA, USA) and β-actin antibody (dilution 1:1,000; mouse monoclonal; cat. no. 061M4808; Sigma-Aldrich) at 4°C overnight. HRP-conjugated anti-rabbit and anti-mouse IgG antibodies (dilution 1:5,000; cat. nos. sc-2357 and sc-516102; Santa Cruz Biotechnology, Inc., Dallas, TX, USA) were used as the secondary antibody at room temperature for 2 h. Bands were detected with ECL reagent (Amersham Biosciences, Buckinghamshire, UK) and scanned with a Fujifilm LAS-4000 Imaging System and analyzed using MultiGauge Software (Fujifilm, Brookvale, Australia).

Suspensions of 1.0×10³ cells were seeded into 96-well plates. After 24, 48, 72 or 96 h of incubation at 37°C, EZ-cytox Cell Viability Assay kit mixture (DoGen, Seoul, Korea) was added. After 1 h of incubation, the fluorescence intensity ratio from 650 to 450 nm was measured. Suspended AGS (1×10³) cells were added to each well of a 6-well plate and incubated at 37°C for seven days. Colony forming cells were stained using 0.5% crystal violet staining solution (Sigma-Aldrich; Merck KGaA) and counted. Transwell chambers (Corning Inc., Corning, NY, USA) were coated with fibronectin (Sigma-Aldrich; Merck KGaA). Cells were suspended in serum-free media and seeded into the upper chamber at a density of 1×10⁵ cells (MKN74) and 1.5×10⁴ cells (AGS) per well, and serum-containing media was placed into the lower chamber. After incubation at 37°C for 18–20 h, cells penetrating the pores were stained using 0.5% crystal violet staining solution (Sigma-Aldrich; Merck KGaA) and observed under a light microscope. Suspended MKN74 cells (5×10⁴) were plated into each well of Culture-Insert 2 Well in µ-Dish 35 mm (ibidi Gmbh, Martinsried, Germany). After 24 h of incubation at 37°C, cells were gently removed from the Culture-Insert using sterile tweezers. The used dish was filled with cell-free medium and observed after 24, 48 and 72 h under a light microscope.

Genomic DNA was bisulfite modified using the EZ DNA methylation kit (Zymo Research, Irvine, CA, USA) and PCR amplified. PCR products were cloned into a pGEM-T Easy vector (Promega, Madison, WI, USA) and a few clones were randomly selected for sequencing. Bisulfite-modified DNA was amplified using a primer set designed to amplify a site between +832 and +1176 relative to the transcription start site using MethPrimer (http://www.urogene.org/cgi-bin/methprimer/meth-primer.cgi). Primer sequences were 5′-GGGTAAAGTGAGTTTTTTTGAGATTTTTAA-3′ (forward) and 5′-CTTCCCTAATCCCTAACCCAAAATC-3′ (reverse). Methylation levels are indicated as the percentage of CpG methylated sites among all CpG sites.

The RNA sequencing library was prepared using the TruSeq RNA Sample Prep kit (Illumina, San Diego, CA, USA; http://www.illumina.com), and sequencing was performed using the Illumina HiSeq 2000 platform to generate 100-bp paired-end reads. Sequenced reads were mapped to the human genome (hg19) using STAR (v.2.5.1), and gene expression levels were quantified with the count module in STAR (23). The TMM (trimmed mean of M-values)-normalized CPM (counts per million) value of each gene was set to a baseline of 1 and log₂-transformed for further analysis (i.e., clustering, heat map drawing and correlation analysis). NGS data were deposited in the NCBI Gene Expression Omnibus (GEO) under the accession no. GSE119196. Raw sequence tags were deposited in the NCBI Short Read Archive (SRA) under the accession no. SRP159087.

Chromatin immunoprecipitation (ChIP) was performed using Dynabeads Protein A and G (Thermo Fisher Scientific, Inc.). For ChIP-PCR/qPCR, 2.0×10⁷ cells were cross-linked with 1% formaldehyde for 10 min at 25°C. Cells were then resuspended and lysed for 10 min at 4°C with SDS lysis buffer. The lysate was aliquoted into a milliTUBE 1 ml AFA Fiber (Covaris, Inc., Woburn, MA, USA) and sonicated using a M220 Focused-ultrasonicator (Covaris, Inc.). Fragment sizes were ~200–500 bp. Samples were diluted with low-salt RIPA buffer and pre-cleared with 50 µl of Dynabeads Protein A and G for 1 h at 4°C. Primary antibodies were added to pre-cleared supernatants, and the mixtures were incubated overnight at 4°C. Antibodies used for the ChIP assay include HOXC10 (dilution 1:400; rabbit polyclonal; cat. no. ab153904; Abcam), phosphorylated RNA polymerase II (dilution 1:400; mouse monoclonal; cat. no. ab5408; Abcam), and IgG antibodies (dilution 1:1,000; cat. nos. sc-2357 and sc-516102; Santa Cruz Biotechnology, Inc.). Next, 50 µl Dynabeads Protein A and G were added to the samples, and the mixtures were incubated for 2 h at 4°C. The beads were subsequently washed with wash buffer (low-salt RIPA, high-salt RIPA, LiCl and TE). Precipitated chromatin was eluted in 200 µl elution buffer (0.1 M NaHCO3 and 1% SDS). Reverse cross-linking was performed overnight at 65°C, and chromatin was then treated with RNase A and 5M NaCl for 30 min at 37°C and proteinase K overnight at 65°C. DNA was purified using a QIAquick PCR Purification kit (Qiagen). ChIP-PCR assays were performed using PCR Master Mix and electrophoresis 1% agarose gel. ChIP-qPCR reactions were subsequently performed.

All experiments were triplicated and data are indicated as mean ± standard deviation. Statistical analysis was performed using Student's t-test. A P-value <0.05 was considered significant. The following parameters were obtained from the medical records of the 171 patients included in the study: Age, sex, histology, lymph node metastasis, tumor stage, and Helicobacter pylori infection status, and these parameters were compared using a Chi-square test. Statistical analysis of correlation between expression and methylation was performed using Pearson's correlation coefficient. All statistical analysis were performed using R statistical programming language (version 3.4.2; http://www.r-project.org/).

HOXC10 is upregulated in many types of cancer. To investigate whether HOXC10 is overexpressed in gastric cancer tissues, we used TCGA-STAD RNA-Sequencing expression dataset. HOXC10 expression was revealed to be higher in gastric cancer tissues than in normal tissues in TCGA dataset (Fig. 1A). The relationship between HOXC10 expression and recurrence-free survival was then investigated in gastric cancer patients from the GEO dataset (GSE26253), demonstrating that the HOXC10 high-expression group exhibited decreased recurrence-free survival compared to the HOXC10 low-expression group from the GEO dataset (Fig. 1B). Overexpression of HOXC10 was validated by performing RT-qPCR on paired normal and tumor tissue samples collected from CNUH (Fig. 1C). Clinicopathological characteristics with respect to HOXC10 expression from the CNUH dataset are summarized in Table I. A significant difference in HOXC10 expression was observed between diffuse and intestinal cancer types (P=0.039), but there was no significant difference in other parameters.

The functional effects of HOXC10 overexpression were first investigated in gastric cancer using in vitro assays. RT-PCR of HOXC10 among 16 gastric cancer cell lines illustrated that AGS and MKN74 cells expressed HOXC10 at low levels, while SNU-216, SNU-484, KATOIII and MKN45 cells expressed HOXC10 at high levels (Fig. 2A). Thus, AGS and MKN74 cells were selected for infection with lentiviruses ectopically expressing HOXC10. Overexpression of HOXC10 in AGS and MNK74 cells after lentivirus infection was confirmed by RT-qPCR and western blot analysis (Fig. 2B). HOXC10 overexpression increased proliferation (Fig. 2C) and colony formation (Fig. 2D) in AGS cells and increased migration in both AGS and MKN74 cells (Fig. 2E and F). Recent studies have demonstrated that overexpression of HOXC10 promotes cell proliferation and migration in gastric cancer GC-9811P, AGS and SGC7901 cells, and HOXC10 knockdown suppresses progression of gastric cancer in GC-9811P, SNU638 and SGC7901 cells (19,20,24). A mouse model injected with AGS cells containing upregulated HOXC10 exhibited increased tumor size (19). Collectively, it was concluded that HOXC10 overexpression promotes proliferation, colony formation and progression in gastric cancer.

To reveal the mechanism of HOXC10 overexpression in gastric cancer, publicly available TCGA STAD datasets were analyzed for DNA methylation and copy number alterations. For copy number alteration, the HOXC10 gene was altered in 1.69% of 295 cases (amplification, 1.02%; deep deletion, 0.68%) using cBioPortal (http://www.cbioportal.org/index.do). To determine whether methylation of HOXC10 CpG sites affected HOXC10 expression, we analyzed TCGA STAD 450K array dataset. It was observed that many CpG sites in the HOXC10 gene were hypomethylated in gastric cancer compared to normal tissues (Fig. 3A). We next attempted bisulfite sequencing to determine methylation levels of HOXC10 CpG sites. In fact, reduced methylation at 24 CpG sites of the first intronic region of HOXC10 (Chr12: 54,379,876-54,380,219; +832 to +1176) was confirmed in three gastric cancers compared to normal tissues (Fig. 3B). The region contains the promoter region (chr12:54,380,069-54,380,128; promoter ID: HOXC10_4) obtained from a Eukaryotic promoter database (https://epd.vital-it.ch/index.php). Specifically, CpG residues 3, 7, 8, 13 and 16 were significantly hypomethylated in tumor tissues (P<0.05, Student's t-test). By correlation analysis of DNA methylation (TCGASTAD 450K array) and gene expression (TCGA STAD RNA-seq) data, it was revealed that three CpG sites (indicated in Fig. 3B) exhibited significant negative correlations between methylation of HOXC10 CpG sites and HOXC10 expression (Pearson's correlation coefficient r=−0.594, −0.689 and −0.699) (Fig. 3C). The methylation levels and HOXC10 expression levels were examined in three paired gastric tumor and normal tissues (Fig. 3D). When two gastric cancer cell lines (AGS and SNU-620 with low HOXC10 expression) were treated with 5-aza-dC, an inhibitor of DNA methyltransferase, HOXC10 expression increased 45.24-fold (AGS) and 8.79-fold (SNU-620) (Fig. 3E). These results indicate that the methylation status of CpG sites in the HOXC10 first intronic region was important for the regulation of HOXC10 expression in gastric cancer.

It was demonstrated that HOXC10 was upregulated by hypomethylation of its CpG sites, and HOXC10 overexpression contributed to the progression of gastric cancer. Since HOXC10 is a transcription factor, the functional roles of HOXC10 overexpression were further explored in gastric cancer by identifying its target genes. Hence, RNA sequencing of ectopic HOXC10-overexpressing AGS cells was performed to identify target genes regulated by HOXC10. RNA-sequencing analysis confirmed overexpression of HOXC10 in AGS cells (3.21-fold increase compared to control vector-infected cells; Fig. 4A). Three hundred and sixty-seven (upregulated) and 366 (downregulated) genes were differentially expressed in response to ectopic HOXC10 expression [|log₂ (fold-change)| >1.5; Fig. 4B and Table SII]. Five of the highly upregulated genes (CST1, S100P, CKS2, FSCN1 and TRIB3) were validated in HOXC10-overexpressing cells compared to controls by RT-qPCR (Fig. 4C). The HOXC10 binding motif sequence predicted by the JASPAR database (http://jaspar.genereg.net/) was mapped to predicted promoter regions (TSS±2 kb) of RNA-seq fold-change >1.5 genes, and 185 genes were selected (P<0.001) (Fig. 4D and Table SIII). By combining two different data sets (RNA-seq fold-change >1.5 with binding motif of HOXC10 predicted by JASPAR database, and gene expression increased >1.5-fold in TCGA dataset), 19 genes were selected as potential target genes regulated by HOXC10 (Fig. 4E). For example, CST1 expression was significantly overexpressed in TCGA dataset (Fig. 4F). A recent study reported that CST1 is upregulated and promotes cell proliferation in gastric cancer (25).

Chromatin immunoprecipitation (ChIP) was then performed in cells with ectopic HOXC10 expression. First, a ChIP assay was conducted at the CST1 promoter region (Chr20:23,732,257-23,732,507) using the HOXC10 antibody. ChIP-PCR and ChIP-qPCR analysis demonstrated HOXC10 bound to CST1 promoter region (Fig. 5A). Additionally, to indicate whether transcriptional machinery operated by HOXC10 binding to the CST1 promoter region, a ChIP assay was designed at the transcription start site region of CST1 (Chr20:23,731,473-23,731,614) with phosphorylated RNA pol II antibody. Phosphorylated RNA pol II did bind to the CST1 transcription start site region (Fig. 5B). A ChIP assay was also performed on S100P (another predicted target) promoter regions and TSS regions (Fig. 5C and D). In addition, it was observed that gastric patients with high CST1 expression exhibited decreased recurrence-free survival compared to those with low CST1 expression (Fig. 6A). Combined analysis of CST1 and HOXC10 expression again confirmed that the CST1/HOXC10 high-expression group exhibited worse recurrence-free survival than the CST1/HOXC10 low-expression group (Fig. 6B). These results indicated that overexpression of CST1 mediated by HOXC10 contributes to worse prognosis in gastric cancer patients.

Expression of CST1 was upregulated in HOXC10-overexpressing AGS cells (Fig. 4C) and tumor tissues of the TCGA dataset. It was validated that HOXC10 regulated the expression of CST1 by using ChIP-PCR. Gastric cancer patients with both CST1 and HOXC10 overexpression exhibited the poorest survival (Fig. 6B). Thus, the roles of CST1 in gastric cancer progression were investigated using transient knockdown of CST1 in HOXC10-overexpressing cells. The silencing of CST1 was confirmed by RT-qPCR in AGS cells (Fig. 7A). Knockdown of CST1 inhibited cell proliferation (Fig. 7B) and colony formation (Fig. 7C) in AGS cells. CST1 expression was upregulated in HOXC10-overexpressing MKN74 cells and downregulated in CST1-silencing MKN74 cells (Fig. 7D). Knockdown of CST1 suppressed the migration ability of MKN74 cells (Fig. 7E). As another downstream target of HOXC10, S100P was also assessed, and it was revealed that knockdown of S100P suppressed cell proliferation and reduced colony size in AGS cells (Fig. S1). CST1 and S100P were selected for further validation based on previous studies revealing the role of those two genes in gastric carcinogenesis. For example, CST1 was revealed to increase cell proliferation (25) and S100P to regulate cell proliferation (26) and colony formation (27) in gastric cancer; these two genes exhibited oncogenic roles in other cancers as well (28–33). Altogether, CST1 regulated by HOXC10 contributes to gastric cancer development, and these two genes could be potential prognostic markers for gastric cancer.