The present study was approved (approval no. 2014-0043) by the Institutional Review Board of Nagoya University (Nagoya, Japan) and conformed to the ethical guidelines of the World Medical Association Declaration of Helsinki (2013) Ethical Principles for Medical Research Involving Human Subjects. Written informed consent for use of clinical samples and data, as required by the Institutional Review Board, was obtained from all patients.

Surgically resected gastric tissues from four patients with liver metastasis were subjected to transcriptome analysis. Global expression profiling was conducted using the HiSeq platform (Illumina, Inc.) to compare the expression levels of 57,749 genes in primary gastric cancer tissues with those of the corresponding noncancerous adjacent gastric mucosa as previously described (5).

A total of 14 gastric cancer cell lines (AGS, GCIY, IM95, KATO III, MKN1, MKN7, MKN45, MKN74, NUGC2, NUGC3, NUGC4, N87, OCUM1 and SC-6-JCK) were obtained from the American Type Culture Collection (ATCC) or the Japanese Collection of Research Bioresources Cell Bank. Cells were cultured at 37°C in RPMI-1640 medium (FUJIFILM Wako Pure Chemical Corporation) supplemented with 10% fetal bovine serum (Corning, Inc.) in an atmosphere containing 5% CO₂. The non-tumorigenic epithelial cell line FHs74 (ATCC) was used as a control. Primary gastric cancer tissues and corresponding normal adjacent tissues were collected from 300 patients who underwent gastric resection for gastric cancer without neoadjuvant therapy at Nagoya University Hospital (Nagoya, Japan) between January 2001 and December 2020. Tissue samples were immediately flash-frozen in liquid nitrogen and stored at −80°C. Tissue comprising >80% tumor components (H&E staining) without grossly visible necrotic regions (~5 mm²) was extracted from each tumor sample. Corresponding normal adjacent gastric mucosa samples were obtained from the same patient and were collected >5 cm from the tumor edge.

Specimens were histologically classified according to the guidelines of the Union for International Cancer Control (UICC), 8th edition (6). To determine whether the expression of CRABP1 differed according to tumor histology, patients were categorized into the histological subtypes of their tumors as follows: differentiated (papillary, well differentiated, and moderately differentiated adenocarcinoma) and undifferentiated (poorly differentiated adenocarcinoma, signet ring cell, and mucinous carcinoma). Since 2006, adjuvant chemotherapy using S-1 (an oral fluorinated pyrimidine) has been administered to all patients with gastric cancer with UICC stages II-III, unless contraindicated by the condition of the patient (7,8).

CRABP1 mRNA levels in primary gastric cancer tissues and corresponding normal adjacent tissues from 300 patients with gastric cancer were evaluated using the reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Patients included 84 women and 216 men, ranging in age from 26-96 years (mean, 70 years). Patients included those with pathologically diagnosed undifferentiated (n=181) or differentiated gastric cancer (n=119). Patients were diagnosed with stage I (n=50), stage II (n=71), stage III (n=109), or stage IV (n=70) gastric cancer and 230 patients with stages I-III under- went R0 resection. Patients classified with UICC stage IV (n=56 of 70) were assigned this diagnosis due to positive peritoneal lavage cytology, localized peritoneal metastasis, or distant lymph node metastasis. Among patients with stage IV disease, 12 had synchronous liver metastasis and 2 had lung metastasis. These patients underwent gastrectomy to control bleeding or allow ingestion of food.

CRABP1 mRNA levels in cell lines and clinical samples (n=300) were analyzed using RT-qPCR with an ABI StepOnePlus Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.). Total RNA (10 µg per sample) was purified using RNeasy Plus Mini kit (cat. no. 74136; Qiagen GmbH) according to the manufacturer's protocol. Complementary DNAs were generated using the M-MLV Reverse Transcriptase (cat. no. 28025013; Thermo Fisher Scientific, Inc.), dNTPs Mix (cat. no. U1511; Promega Corporation), the Primer Random pd(N)6 (11034731001, Roche Diagnostics) and RNase inhibitor (cat. no. 3335399001; Roche Diagnostics) according to the manufacturer's protocol, and amplified using primers specific for CRABP1 (Table I). RT-qPCR was performed using the SYBR-Green PCR Core reagents kit (Applied Biosystems; Thermo Fisher Scientific, Inc.) and absolute quantification was performed using the standard curve method. The following thermocycling conditions were used for qPCR: one cycle at 95°C for 10 min, 40 cycles at 95°C for 5 sec, and 60°C for 60 sec. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA served as an internal standard, and the expression level of each sample was determined in triplicate and calculated as the value of CRABP1 mRNA divided by that of GAPDH mRNA (9).

To identify genes coordinately expressed with CRABP1 in gastric cancer cell lines, PCR array analysis was performed using the Human Epithelial to Mesenchymal Transition (EMT) RT2 Profiler PCR Array (Qiagen GmbH). This array profiles the expression of 84 key genes including those that encode transcription factors, ECM proteins as well as proteins involved in the EMT, cell differentiation, morphogenesis, growth, proliferation, migration, cytoskeleton and major signaling pathways (10).

A total of two siRNAs specific for CRABP1 were designed at online sites and were pooled to inhibit CRABP1 mRNA expression with the aim of obtaining stable knockdown as previously described (Table I) (11,12). siCRABP1-1 and siCRABP1-2 were designed by siDirect (http://sidirect2.rnai.jp/) and i-Score Designer (https://www.med.nagoya-u.ac.jp/neurogenetics/i_Score/i_score.html), respectively, and supplied from Hokkaido System Science Co., Ltd. MKN1, MKN45 and NUGC4 cells were added to the wells of a 24-well plate (5×10⁴ cells/ml) and transiently transfected at 37°C the next day with 30 nM or CRABP1 siRNA or a control siRNA (siControl with sequence as follows: 5′-GCA AAC AUC CCA GAG GUA U-3′) combined with LipoTrust EX Oligo (Hokkaido System Science Co., Ltd.); total RNAs were extracted 72 h later. To evaluate the effect of siRNAs on CRABP1 mRNA expression, RT-qPCR analysis was performed as previously described (11,12). In addition, the knockdown efficacy of siCRABP1-1 or siCRABP1-2 alone in MKN1, MKN45 and NUGC4 cells was evaluated.

Cell proliferation was evaluated using the Cell Counting Kit-8 (Dojindo Molecular Technologies, Inc.) as previously described (11). MKN1, MKN45 and NUGC4 cells (at a density of 1.5×10³, 1.5×10³ and 5×10³ cells per well, respectively) were seeded into 96-well plates in RPMI-1640 medium supplemented with 2% FBS. Cell invasion was determined using BioCoat Matrigel invasion chambers (BD Biosciences,) according to the manufacturer's protocol as previously described (13). MKN1 and MKN45 cells (2.5×10⁴ cells/well) were suspended in serum-free RPMI-1640 and seeded in the upper chamber. After an appropriate incubation time (24 and 72 h, respectively), cells present on the surface of the membrane were fixed, stained, and counted using a light microscope in eight randomly selected fields as previously described (13). Cell migration was evaluated using wound-healing assays as previously described (14). The width of the wound was measured at 100-µm intervals (20 measurements per well, ×400 magnification). The invasion and migration assays were performed in duplicate (n=2; two wells for each assay). For the invasion assay, 8 fields were randomly selected from each well and numbers of invasive cells were counted. Thus, statistical analysis was carried out using 16 values for the untransfected, siControl and siCRABP1 groups. For the migration assay, the width of the wound was measured at 20 points for each well, indicating that statistical analysis was carried out using 40 values for the untransfected, siControl and siCRABP1 groups.

Animal experiments were performed between October and December 2021 according to the ARRIVE guidelines (15) and were approved (approval no. M210414-001) by the Animal Research Committee of Nagoya University (Nagoya, Japan). A total of 10 six-week-old male NOD/SCID (weight, 24.7 g) and 2 BALBc nu/nu mice (weight, 20.4 g) were obtained from Japan SLC, Inc. and housed at least 1 week before experiments in temperature-controlled rooms at 20-22°C with free access to food and water supply and a light/dark cycle of 14/10 h. MKN1 and NUGC4 cells transfected with CRABP1 siRNA or untransfected were implanted into the abdominal cavity of six-week-old male mice (MKN1: n=5 each, NUGC4: n=1 each) to analyze the peritoneal dissemination of the xenografts. MKN1 and NUGC4 cells (4×10⁶) in 500 µl of phosphate-buffered saline were injected into NOD/SCID and BALBc nu/nu mice, respectively. After 4 weeks of observations, these mice were euthanized after exposure to 100% CO₂ for 5 min and were observed for 20 min after confirmation of respiration cease. The flow rate of CO₂ was 50% of the chamber volume per min. After confirming euthanasia, the formation of peritoneal metastasis was observed under direct viewing.

The optimal cut-off value (0.0000325) of CRABP1 mRNA levels in primary gastric cancer tissues was determined using receiver operating characteristic curve analysis for evaluating the significance of the association of their levels with metastasis or recurrence. Patients were stratified according to the cut-off value of CRABP1 mRNA levels in gastric cancer tissues as follows: high CRABP1 expression (>cut-off value) and low CRABP1 expression (≤cut-off value). Correlations between the patterns of CRABP1 mRNA expression and clinicopathological parameters were evaluated. Correlation analysis of CRABP1 mRNA expression and recurrence patterns after curative surgery was applied to 230 patients who underwent curative surgery (i.e., stages I-III). Thus, the analysis of recurrence pattern specifically focused on initial recurrence after curative surgery. Outcome analyses of the overall survival and disease-free survival (DFS) rates and multivariate analysis were applied to 230 patients who underwent curative surgery. To validate the present data, an integrated microarray dataset comprising tissues of 1065 patients [Berlin, Bethesda, and Melbourne datasets (http://kmplot.com/analysis/)] was analyzed as previously described (16).

The significance of differences of the relative mRNA levels (CRABP1/GAPDH) between the two groups were analyzed using the Mann-Whitney test. The significance of a correlation between two variables was assessed using the Spearman's rank correlation coefficient. The χ² test was used to analyze the associations between the expression levels of CRABP1 and clinicopathological parameters. DFS rates were calculated using the Kaplan-Meier method, and the differences in the slopes of the survival curves were analyzed using the log-rank test. Multivariable regression analysis was preformed to identify prognostic factors using the Cox proportional hazards model, and variables with P<0.05 were entered into the final model. All statistical analyses were performed using JMP 15 software (SAS Institute, Inc.). P<0.05 was considered to indicate a statistically significant difference.

Transcriptome analysis of gastric tissues compared with corresponding noncancerous adjacent gastric mucosa from four patients with metastatic gastric cancer was first performed. Transcriptome analysis identified 26 candidate genes that were: i) Overexpressed in gastric cancer compared with the corresponding normal tissues and ii) Expressed at comparable expression levels in primary gastric cancer and metastatic tissues (Table II). A literature review of the functions of the identified genes was conducted and CRABP1 was selected for subsequent analyses for the following reasons: i) Insufficient evidence was available on the oncological roles of CRABP1; ii) CRABP1 mediates the activity of retinoid, which is involved in cancer progression; and iii) nucleotide sequence of CRABP1 is available from the United States National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/).

The relative levels of CRABP1 mRNA and those of mRNAs encoding potential CRABP1-interacting proteins in gastric cancer cell lines are presented in Fig. 1B. There were large differences in the levels of CRABP1 mRNA and those of other genes among gastric cancer cell lines. CRABP1 mRNA levels positively correlated with those encoding IGFBP4, MAP1B, ZEB2, STEAP1, VIM and TIMP1 and negatively with TFPI2 (Fig. 1C).

To characterize CRABP1 in gastric cancer, the levels of CRABP1 mRNA in 12 gastric cancer cell lines were next compared with those of a nontumorigenic epithelial cell line. CRABP1 mRNA levels were >2-fold higher in MKN1, MKN7, N87, IM95, GCIY, MKN45, NUGC2 and OCUM1 cells compared with FHs74 cells (Fig. 1A). CRABP1 mRNA levels did not significantly differ according to the extent of differentiation of the gastric cancer cells. MKN1, MKN45 and NUGC4 cells were selected for subsequent analyses, since MKN1 and MKN45 cells expressed relatively high levels of CRABP1 mRNA, and these three cell lines were easy to use in functional analyses.

The efficiency of CRABP1 knockdown by transfection of siCRABP1-1 and siCRABP1-2 alone was evaluated in MKN1, NUGC4 and MKN45 cells (Fig. S1). These two siRNAs were pooled to constitute a CRABP1-specific siRNA. To evaluate the function of CRABP1 in gastric cancer cells, MKN1 and NUGC4 cells were transfected with a CRABP1-specific siRNA. It was first determined that the knockdown efficacy of the CRABP1 siRNA in MKN1, MKN45 and NUGC4 cells was sufficient for analysis (Figs. 2A and S2). The proliferation of siRNA-transfected MKN1, MKN45 and NUGC4 cells as well as the invasiveness and migration of MKN1 and MKN45 cells were then evaluated. The proliferation of MKN1, MKN45 and NUGC4 cells was decreased as a result of CRABP1 knockdown starting from 72 h after transfection compared with the siControl-transfected cells (Figs. 2B and S2). Furthermore, the invasiveness of MKN1 and MKN45 cells was reduced by inhibiting CRABP1 expression (Fig. 3). The migration of MKN1 and MKN45 cells was reduced by inhibiting CRABP1 expression (Fig. 4).

MKN1 and NUGC4 cells transfected with CRABP1 siRNA or untransfected were injected into mice to identify the function of CRABP1 in recurrence and metastasis of gastric cancer. Observations in the abdominal cavity of the mice were performed after euthanasia. In the MKN1 xenograft model, peritoneal dissemination was not observed in the siCRABP1 group (Fig. 5). Peritoneal metastasis in the NUGC4-model mice was disseminated to a smaller extent in the siCRABP1 group compared with the untransfected group (Fig. S3).

The DFS rate of the CRABP1-high group was significantly lower compared with that of the CRABP1-low group (5-year DFS rates; 59.6% and 77.8%, respectively; P=0.012) (Fig. 6A) and were consistent with those of the extra-validation cohort (Fig. 6B).

Next, gastric cancer recurrence patterns were analyzed according to CRABP1 mRNA levels of 230 patients who underwent R0 resection (stages I-III). Among them, 57 (24.7%) experienced postoperative recurrence at 65 initial recurrence sites. Analysis of recurrence patterns revealed that high expression of CRABP1 mRNA was significantly associated with peritoneal recurrence (P=0.016) (Fig. 6C), but not with the other two recurrence patterns.

The correlations between CRABP1 expression and clinicopathological characteristics of patients were next examined (Table III). High CRABP1 expression was significantly associated with lymph node metastasis. Univariate analysis of DFS demonstrated that carbohydrate antigen 19-9 (37 IU/ml), tumor size ≥50 mm, macroscopic type (Borrmann type 4/5), pT4, lymphatic involvement, vascular invasion, invasive growth, lymph node metastasis and high CRABP1 mRNA expression in gastric cancer tissues were significant prognostic factors for adverse outcomes (Table IV). Multivariable analysis identified high CRABP1 mRNA expression as an independent prognostic factor of poor outcome (hazard ratio 1.89; 95% confidence interval, 1.15-3.09; P=0.012).