Gene expression RNAseq dataset (Level 3) and clinical data for The Cancer Genome Atlas Colon and Rectal Cancer (TCGA-COADREAD) cohort (19) were downloaded from the UCSC Xena (https://xena.ucsc.edu). CRC gene expression microarray data used in this study were downloaded from the publicly available GEO databases (http://www.ncbi.nlm.nih.gov/geo/): GSE21815 (20), GSE44076 (21), GSE44861 (22), GSE41258 (23) and GSE33113 (24). The GEO datasets used in this study include 562 CRC tissues and 222 NST from respective same patient groups. The downloaded raw data of GEO databases were normalized at the transcript and gene level using the Robust Multichip Average method (25). Cluster analysis was performed using Cluster 3.0 to classify the samples into statistically similar groups, and the resulting heatmaps were visualized in TreeView 1.6 (www.eisenlab.org/eisen). The four CerS genes present in the TCGA COADREAD, GSE44076 and GSE44861 cohorts were LASS2, LASS4, LASS5 and LASS6. The present study meets the publication guidelines provided by TCGA.

A total of 59 patients (mean age, 64.83±9.48; age range, 38–83; 34 males and 25 females) diagnosed with CRC were included in the present study. CRC and NSTs were obtained from patients undergoing surgery in Keimyung University Dongsan Medical Center (Daegu, Korea) between April 2008 and January 2010. Enrolled patients with CRC were classified according to the AJCC Tumor-Node Metastasis (TNM) staging criteria (26). Tissue samples were immediately frozen in liquid nitrogen and stored at −196°C until RNA isolation. Tissue samples were provided by Keimyung Human Bio-Resource Bank (Daegu, Korea). Written informed consent was obtained from each study participant and the protocols were approved by the Institutional Review Board of Keimyung University Dongsan Medical Center (approval no. 2015-11-059-001).

Total cellular RNA was extracted from tissues using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA). RNA was quantified using NanoDrop 1000 (Thermo Fisher Scientific, Inc.). Each cDNA was synthesized from 2 µg total RNA using MMLV reverse transcriptase (Promega Corporation, Madison, WI, USA), according to the manufacturer's protocol. qPCR was performed on the LightCycler^® 480 Real-Time PCR system (Roche Diagnostics GmbH, Mannheim, Germany) using the specific primer pairs presented in Table I and SYBR-Green Premix (Toyobo Life Science, Osaka, Japan). The qPCR was performed using the following thermocycling conditions: 95°C for 10 min; followed by 45 cycles of 95°C for 10 sec, 60°C for 10 sec, and 72°C for 12 sec. Melting curve was analyzed to determine primer specificity. b-actin was used as a housekeeping gene for normalization, and a no-template sample was used as a negative control. qPCR data were analyzed using the 2^−∆∆Cq method (27). Each experiment was performed three times.

Statistical analysis was performed using SPSS 22.0 (IBM Corp., Armonk, NY, USA). The cell viability data were analyzed using one-way analysis of variance and the Student-Newman-Keuls post hoc test. Differences between the groups were analyzed statistically using Student's t-test or Mann Whitney U test. The co-expression of the mRNAs of various CerS genes in TCGA-COADREAD cohort were searched using cBioPortal (http://cbioportal.org) (28). The association between inter-individual mRNA expression levels of CerS genes in Korean patients with CRC was assessed using Pearson's correlation coefficient analysis for continuous variables. Clinicopathological associations with the mRNA expression levels of various CerS genes in Korean CRC were analyzed using the Linear by linear association, the Pearson's Chi-square test and the Fisher's exact test for categorical variables. The mean value was used as the cut-off value (low and high) for categorical variables. P<0.05 was considered to indicate a statistically significant difference.

Various human colorectal adenocarcinoma cell lines, HCT116, HT29, SW403 and SW480 cells, were plated onto 6-well plates at a density 7×10⁵ cells/well and cultured overnight. pcDNA3.1-empty vector was used for plasmid constructs, including HA-tagged form of CERS2 (HA-CERS2) and HA-tagged form of CERS6 (HA-CERS6) constructs. All plasmids, including pcDNA3.1-empty vector, HA-CERS2 and HA-CERS6 were provided by Professor Anthony H. Futerman (Weizmann Institute of Science, Rehovot, Israel). The CRC cells were transfected with pcDNA3.1-empty vector, 2 µg HA-CERS2 and HA-CERS6 plasmid in 6-well plates using Lipofectamine reagent (Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. At 24 h after plasmid transfection, the subsequent experiments were conducted.

The transient transfected CRC cells were collected and washed twice with cold PBS, and cell pellets were prepared by suspending in modified radioimmunoprecipitation assay buffer (50 mM Tris-HCl pH 7.4, 1% NP-40, 0.25% Na-deoxycholate, 150 mM NaCl, 1 mM Na₃VO₄ and 1 mM NaF) containing protease inhibitors (100 µM phenylmethylsulfonyl fluoride, 10 µg/ml leupeptin, 10 µg/ml pepstatin and 2 mM EDTA). The lysates were centrifuged at 10,000 × g for 10 min at 4°C, and the supernatant fractions were collected. The total protein concentration was measured using Micro BCA™ Protein assay kit (Thermo Fisher Scientific, Inc.), according to the manufacturers protocol. Cellular proteins (60 mg) were mixed with protein 5X sample buffer (Elpis Biotech., Inc., Daejeon, Korea) and heated at 95°C for 5 min. The proteins were separated by 10% SDS-PAGE and then electrotransferred to Immobilon-P membranes (EMD Millipore, Billerica, MA, USA). The membranes were then blocked at room temperature with 5% skimmed dried milk in PBS/0.1% Tween-20 for 1 h, and incubated overnight at 4°C with anti-HA (1:2,000; mouse monoclonal; cat. no. SAB1411737) and anti-β-actin (1:2,000; mouse monoclonal; cat. no. A5441; both Sigma-Aldrich; Merck KGaA, Darmstadt, Germany). The membranes were then washed six times with PBS/0.1% Tween-20 (30 min each) and incubated with the corresponding secondary antibodies (horseradish peroxidase-conjugated, horse antibodies to mouse IgG; 1:2,000; cat. no. 7076; Cell Signaling Technology, Inc.) for 1 h at room temperature. Following washing six times in PBS/0.1% Tween-20, the specific protein bands were detected using an enhanced chemiluminescence western blotting kit (EMD Millipore), according to the manufacturer's protocol.

To investigate whether the sphingolipid metabolism-related genes (29) are dysregulated in CRC tissues, the present study re-analyzed the raw data of six independent CRC cohorts. To begin with, the cancer gene expression RNAseq datasets of 380 CRC patients were taken from the TCGA-COADREAD cohort through UCSC Xena. Next, CRC gene expression microarray data were downloaded from the publicly available Gene Expression Omnibus databases. The CRC gene expression microarrays, GSE21815, GSE33113, GSE41258, GSE44076 and GSE44861, were analyzed for potential transcriptome changes. Hierarchical clustering revealed that various sphingolipid metabolism-related genes were dysregulated in carcinomatous tissues compared with NST of patients with CRC (Fig. 1). To identify the significance of altered mRNA expression levels between CRC and NST, Student's t-test or Mann Whitney U test were performed (P<0.05). As demonstrated in Fig. 2, hierarchical clustering revealed that various sphingolipid metabolism-related genes were significantly dysregulated in CRC tissues compared with NST from the same patient groups. The list of analyzed sphingolipid metabolism-related genes is presented in Table II. Sphingosine kinase 1 (SPHK1) and UDP-glucose glycoprotein glucosyltransferase 2 (UGGT2) were significantly upregulated in the CRC tissues of all cohorts, while 15-hydroxyprostaglandin dehydrogenase (HPGD), lysophosphatidic acid receptor 1 (LPAR1), N-acylethanolamine acid amidase (NAAA), sphingomyelin phosphodiesterase 1 (SMPD1) and sphingomyelin phosphodiesterase acid-like 3A (SMPDL3A) were significantly downregulated in the CRC tissues of all cohorts (Fig. 2 and Table II).

Next, the present study evaluated whether the mRNA expression levels of the four CerS genes, which are abundant in colorectal tissues (30), are dysregulated in human CRC specimens with respect to NST. As demonstrated in Fig. 3, among six cohorts, CERS2 mRNA levels were significantly increased in five independent cohorts, while CERS5 and CERS6 were significantly upregulated in four independent cohorts. The specific platforms of each cohort and their associated studies are listed in Table III.

To determine whether there is altered CERS2, CERS4, CERS5 and/or CERS6 mRNA expression in Korean patients with CRC, the expression levels of these four CerSs were measured using qPCR in 59 paired CRC and NST specimens from Korean patients. Following exclusion of unqualified results, the qPCR data were analyzed. The present study revealed that mRNA expression levels of all four CerS genes were significantly upregulated in CRC tissues compared with corresponding NSTs (CERS2, P<0.001; CERS4, P=0.006; CERS5, P<0.001; CERS6, P<0.001; Fig. 4; Table IV).

It has previously been observed that CERS6-overexpression reduces the proliferation of CRC cells and induces apoptosis, whereas CERS2-overexpression increases the proliferation of CRC cells (18). To confirm the effect of overexpressing CerSs in CRC cells, HCT116, HT29, SW403 and SW480 cells were transiently transfected with constructs to overexpress HA-CERS2 and HA-CERS6, respectively. After 48 and 72 h, the numbers of viable cells were counted using a hemocytometer. As demonstrated in Fig. 5, overexpression of CERS2 and CERS6 decreased the viability of this panel of CRC cell lines.

Combinational patterns of CerS gene expression, including CerS hetero-complexes and co-expression of CerS genes, serve important roles in sphingolipid metabolism (31,32). Therefore, associations between the mRNA levels of each CerS gene were identified in the TCGA-COADREAD cohort and in the Korean CRC cohort. Using cBioPortal to analyze the TCGA-COADREAD cohort, co-expression analysis revealed that CERS4 and CERS5 had high correlation coefficients (Pearson's correlation=0.36; Spearman's correlation=0.48; Fig. 6A). Next, these correlations were assessed using Pearson's correlation coefficient analysis in the 59 Korean patients with CRC. There were significant correlations between CERS2 and CERS4, and also between CERS5 and CERS6, with a Pearson's correlation coefficient value of 0.532 (P<0.001; Fig. 6B) and 0.439 (P=0.003; Fig. 6C), respectively. Furthermore, significant correlations between the mRNA expression levels of CERS2 and CERS4 (P=0.009) and of CERS5 and CERS6 (P<0.001) were identified using Fisher's exact test (Table V).

To determine the clinicopathological implications of dysregulated expression of specific CerS genes in CRC, the association between CerS gene mRNA level and clinicopathological characteristics, which are used to represent progression and aggressiveness, were evaluated. Prior to the statistical analysis, the 44 patients, whose clinical data were available, were classified according to each clinicopathological characteristic (Table V). The results obtained from the statistical analysis of the Korean cohort revealed that altered mRNA expression levels of CerS genes were not significantly associated with any clinical parameters, including sex, age, Tumor-Node-Metastasis stage, body mass index or carcinoembryonic antigen titer.