Characterization of SLC22A18 as a tumor suppressor and novel biomarker in colorectal cancer.

SLC22A18, solute carrier family 22, member 18, has been proposed to function as a tumor suppressor based on its chromosomal location at 11p15.5, mutations and aberrant splicing in several types of cancer and down-regulation in glioblastoma. In this study, we sought to demonstrate the significance of SLC22A18 as a tumor suppressor in colorectal cancer (CRC) and provide mechanistic bases for its function. We first showed that the expression of SLC22A18 is significantly down-regulated in tumor tissues using matched normal-tumor samples from CRC patients. This finding was also supported by publically accessible data from The Cancer Genome Atlas (TCGA). Functionally, SLC22A18 inhibits colony formation and induces of G2/M arrest consistent with being a tumor suppressor. Interestingly, suppression of KRAS by RNA interference promotes SLC22A18 expression, and expression of SLC22A18 in turn inhibits KRAS(G12D)-mediated anchorage independent growth of NIH3T3 cells indicating a mutual negative interaction. Finally, we evaluated diagnostic and prognostic values of SLC22A18 using clinical and gene expression data from TCGA which revealed a significantly worse long-term prognosis for patients with low level SLC22A18 expression. In sum, we established SLC22A18 as a tumor suppressor in colon epithelial cells and propose that SLC22A18 is potentially a marker of diagnostic and prognostic values.


INTRODUCTION
CRC is currently the one of the highest-ranked cancers in terms of both mortality and incidence rates [1]. In the United States alone, nearly 140,000 new cases and over 50,000 deaths associated with CRC are projected for 2014 [2]. Although such numbers signify a slight decline of colorectal cancer cases in United States, the incidence has been on the rise in other parts of the world including many populous Asian countries where colonoscopic screening for early detection is still limited [3,4].
Clearly, novel molecular markers would be beneficial both for better diagnosis and prognosis and for characterizing the molecular mechanism of CRC development. We have previously reported isolation of novel markers of CRC from meta-analysis of publicly available gene expression data [5]. Of the 34 genes we have isolated, we validated 9 of them to be significantly up-regulated in cancer tissues compared to normal surrounding colon tissues. At least some of these genes (ECT2, ETV4, DDX21, RAN, S100A111, RPS4X, HSPD1, CKS2 and C9orf140), three of which have been shown to be activated by c-MYC or KRAS, likely contribute to oncogenic processes [5]. www.impactjournals.com/oncotarget In contrast, the genes down-regulated in cancer tissues represent potential tumor suppressors. Here, we report characterization of SLC22A18, a member of a large family of cytoplasmic membrane associated transporters [6], as a potential tumor suppressor of CRC. Also known as IMPT1, BWR1A, TSSC5 and ORCTL2, SLC22A18 is located in 11p15.5, a chromosomal region which is frequently deleted in a variety of cancers [7] and encompasses several imprinted genes most of which including SLC22A18 show preferential expression of the maternal allele [8,9]. Taken together with that the gene silencing and the loss-of-heterozygosity (LOH) involve mostly the maternal allele [10][11][12], it has been proposed that at least one tumor suppressor gene is present in this chromosomal region. Missense mutations of SLC22A18 have been found in a variety of cancer types including breast cancer, lung cancer and rhabdomyosarcoma, and aberrant splicing of this gene has been reported in multiple Wilms' tumor cases [13,14]. Together, these reports suggest that SLC22A18 functions as a tumor suppressor in certain cell types. Consistently, Chu and coworkers showed that down-regulation of SLC22A18 was frequently associated with glioma development and that ectopic expression of SLC22A18 led to reduced proliferation of U251 glioma cells in vitro and in vivo [15]. Furthermore, low SLC22A18 expression has been reported to be correlated with poor prognosis for glioma patients and breast cancer patients [16,17].
Here, we provide for the first time evidences indicating that SLC22A18 functions as a tumor suppressor in CRC. The gene is down-regulated in CRC and can retard the growth of CRC cell lines by inducing G2/M arrest. Remarkably, expression of SLC22A18 is negatively regulated by KRAS, and SLC22A18 in turn appears to negatively regulate KRAS signaling. In addition, our preliminary analysis of public data from TCGA indicates that SLC22A18 may also be a potential prognostic marker of CRC.

Down-regulation of SLC22A18 in CRC tissues
We have previously reported isolation of 34 genes differentially expressed in CRC from public data analysis and confirmation of 9 among them as potential biomarkers of CRC based on quantitative RT-PCR results using matched normal and tumor tissues [5]. These had been selected based on significantly elevated expression in tumor tissues of 10 randomly selected patients (P-value < 0.01; Supplementary Figure 1). Several genes including SLC22A18 also showed significantly differential expression but with lower levels in tumor tissues (Supplementary Figure 1). SLC22A18, a member of membrane bound solute carrier gene super family, has been proposed to be a tumor suppressor based on its chromosomal location in 11p15.5 as well as frequent mutations found in various tumors and was thus selected for further analyses.
First, real-time RT-PCR assay was carried out with samples from 29 patients. Without exception, the expression was down-regulated in tumor tissues compared to matched normal tissues ( Figure 1A). This was also confirmed by conventional RT-PCR followed by gel electrophoresis using 5 representative patient samples ( Figure 1B). We were able to obtain RNA-seq data from 26 normal-tumor matched samples of colon adenocarcinoma (COAD) from TCGA and determine normalized expression values of SLC22A18. Consistent with RT-PCR results from our patient samples, virtually all tumor tissues showed lower expression of SLC22A18 than matched normal tissues ( Figure 1C). We obtained paraffin-embedded tumor tissues from an independent group of patients and examined SLC22A18 expression by immunohistochemical staining. Consistent with results from RT-PCR, tumor areas from all samples examined showed down-regulated expression of SLC22A18 ( Figure  1D and Supplementary Figure 2). We next investigated the mutation status of SLC22A18. First, using our patient genomic DNA samples from normal and tumor tissues, we PCR-amplified all coding portions of SLC22A18 and examined the nucleotide sequence by Sanger sequencing. We found no mutations aside from novel polymorphisms and no evidence for LOH (data not shown). All cases of heterozygosity found in tumor tissues were also seen in normal tissues and vice versa. Second, we examined exome sequences of 151 colon adenocarcinoma patients with matched normaltumor samples from TCGA database. Our analysis using GATK (version 2.5.2) program to call for heterozygous loci and VarScan2 (version 2.3.5) program to call for LOH revealed only two cases of LOH out of 151 patients (data not shown). These data together indicate that by and large mutations or LOH is not the cause for lowered expression of SLC22A18 in colorectal cancer cases.
We have expanded the analysis of TCGA RNAseq data to all cancer types with matched normal-tumor samples. Down-regulation in tumor tissues appears to be restricted to COAD as other types of cancer showed no significant down-regulation in tumor tissues (Supplementary Figure 3). Although rectal cancer showed down-regulation in 6 out of 6 tumor tissues, the statistical significance was not attained due to the limited number of patient samples. Still, if the COAD and rectal cancers are pooled together into CRC, the correlation should be even more highly significant. Of note, for lung adenocarcinoma, thyroid cancer and breast cancer, SLC22A18 level appears to be higher in tumor tissues (Supplementary Figure 3). www.impactjournals.com/oncotarget

Inhibition of colony formation by SLC22A18
We next examined if SLC22A18 has tumor suppressor activity using colony formation assay. Three CRC cell lines, HCT116, SW480, and HT29 were selected, and two expression constructs of the full length SLC22A18 with epitopes on the N-terminus or C-terminus were generated. Both constructs showed similar effects in inhibiting formation of colonies, and all three cell lines tested showed inhibited colony formation ( Figure 2A) upon ectopic expression of SLC22A18 although the inhibitory effect of SLC22A18 differed considerably among the three cell lines ( Figure 2B). HCT116 showed the most dramatic response in that the number of colonies was only about 20% of what was seen in the control transfection case. Subsequent experiments were mostly carried out with HCT116 cells.

G2/M cell cycle arrest induction by SLC22A18
Tumor suppressors often induce cell cycle arrest. We therefore examined the effect of ectopic SLC22A18 Figure 2: Examination of the tumor suppressor activity of SLC22A18 in colony formation assay. A. Ectopic expression of SLC22A18 suppresses growth of CRC cell lines. Two SLC22A18-expression vectors with V5 or Xpress epitopes respectively at the C-or N-termini have been prepared and transfected to three independent CRC cells, HCT116, SW480, and HT29. For the control, pcDNA3.1 plasmid with no insert has been used. Colonies after G418 selection and Coomassie staining are shown. Each transfection has been carried out three independent times. B. Bar graphs showing results from colony formation assay. For all three cell lines, transfection of SLC22A18 expression plasmids led to statistically significant decrease in colony numbers compared to control plasmid transfection. (*) represents P-value of < 0.05. Note the particularly strong effect in the case of HCT116 cells. expression on cell cycle progression in HCT116 cells using PI staining in combination with flow cytometric analyses. V5 epitope-tagged SLC22A18 was incorporated into a retroviral vector with IRES-GFP in tandem. Control virus expressed only GFP. Compared to cells not infected or infected with the control virus, SLC22A18 virusinfected cells showed a significant decrease in the fraction of cells in S phase and increase in the fraction in G2 phase consistent with induction of G2/M arrest ( Figure 3A, 3B). The sub-G1 fraction also showed a low level albeit statistically significant increase ( Figure 3B). The extent of the increase suggests that apoptosis which typically follows prolonged G2/M arrest was taking place. We also checked for changes in cell cycle markers after ectopic expression of SLC22A18. Again, consistent with G2/M arrest, we saw decreased levels of p-Cdc2, Cyclin B1 and Cdc25c and an increase in the level of p21 in cells transduced with SLC22A18 virus compared to control cells ( Figure 3C). Taken together with inhibition of colony formation, the G2/M arrest induced by SLC22A18 strongly suggests that SLC22A18 indeed functions as a tumor suppressor. The KRAS level is expressed relative to the level seen in transfection reagent only control. Knock-down of KRAS mRNA by two specific siRNAs is shown (top panel). NC represents non-targeting control siRNA transfection. Note that SLC22A18 level increased over 2 fold (bottom panel). The graphs show results as mean ± S.D. of three independent experiments. (*) represents P-value of < 0.05 compared to NC. B. Immunoblot analyses of siRNA transfection assay. HCT116 lysates were examined after transfection of siRNAs targeting KRAS mRNA. Treatment with the transfection reagent only and transfection of control siRNA were used as controls. Knock-down of KRAS and elevation of SLC22A18 proteins were seen. For the loading control, α-tubulin was used. www.impactjournals.com/oncotarget

Negative interaction between SLC22A18 and KRAS
Oncogenic mutations in KRAS gene is found in up to 50% of CRC cases [18]. Specifically, mutations of Gly12 or Gly13 lead to constitutive activation of KRAS signaling pathway. HCT116 contains the G13D mutation of KRAS which allows a chance to see if KRAS signaling affects expression of SLC22A18. First, two independent siRNAs targeting KRAS mRNA were shown to induce efficient down-regulation of KRAS by real-time RT-PCR ( Figure  4A). Their effects were also confirmed by immunoblotting ( Figure 4B). Importantly, SLC22A18 expression in HCT116 cells was shown to be up-regulated upon specific targeting of KRAS mRNA ( Figure 4A). Again, this was confirmed by immunoblotting for SLC22A18 ( Figure  4B). These observations are consistent with that KRAS signaling down-regulates SLC22A18 as part of its proproliferation activity. Anchorage independent growth is one of the hallmarks of cellular transformation. We used a wellestablished protocol for KRAS-induced anchorage independent growth of NIH3T3 cells on soft agar to test if SLC22A18 negatively affects KRAS signaling and consequently the transformation. Pseudotyped retroviruses expressing SLC22A18 or KRAS G12D were generated and used to infect NIH3T3 cells which were subsequently grown in soft agar. Cells infected with either control or SLC22A18 virus showed virtually no growth in soft agar as expected ( Figure 5A, 5B). In contrast, KRAS virus transduction led to a dramatic increase in the number of multicellular colonies consistent with induction of transformation. Interestingly, co-infection of SLC22A18 virus with KRAS virus led to a significant decrease in the number of visible multicellular colonies ( Figure 5A, 5B). Taken together with KRAS-induced down-regulation of SLC22A18, this result indicates that SLC22A18 and KRAS form a mutually inhibitory regulatory loop.

Inhibition of xenograft tumor growth by SLC22A18
We next examined the effect of SLC22A18 expression using a xenograft model. HCT116 cells infected either with control virus or with SLC22A18 virus were used to inoculate female BALB/c Slc-nu/nu mice. Tumor volumes were measured over the course of 5 weeks. A definitive and statistically significant difference was noted (Figure 6 and Supplementary Figure 4) between control group (n = 10) and SLC22A18 group (n = 15). Specifically, while all 10 control mice developed tumors, only 2 out of 15 from SLC22A18 did so, consistent with tumor suppressor activity of SLC22A18.

Evaluation of SLC22A18 as a prognostic marker of CRC
SLC22A18 has been previously reported as a potential prognostic marker in breast cancer and glioma [16,17]. Thus, we investigated the relationship between SLC22A18 expression level and patient survival using the TCGA COAD data. Patients were divided into three groups according to the expression levels of SLC22A18 gene, and the resulting Kaplan-Meier survival curves are shown in Figure 7. Even though the survival analysis of 430 patients gave statistically insignificant result (P-value = 0.78), we observed that three curves were well separated beyond three years. Restricting the patient cohort to 85 survivors beyond three years, we obtained the P-value = 0.012. Comparison between two groups (55 patients in sum) of high and low expressions of SLC22A18 yielded evidence for more significant difference (P-value = 0.0032). We have also carried out Cox Proportional Hazards regression analysis to determine the correlation between SLC22A18 expression levels and survival based on individually matched data without categorizing patients into groups. We used TCGA survival data from 259 COAD patients whose gene expression data were uniformly obtained via HiSeq sequencing platform. The result supports conclusions from the grouping-based analysis. Specifically, even though the cohort of 259 patients gave overall a statistically insignificant result (P-value = 0.1939), restricting the patient cohorts to 84 survivors beyond two years and to 58 survivors beyond 3 years led to P-values of 0.0312 and 0.0190 respectively. This suggests that the expression level of SLC22A18 has a predictive power for long-term survival in COAD patients. Kaplan-Meier survival analysis shows that after three years, the survival rate is significantly co-related with the SLC22A18 expression level (P-value estimated to be 0.012). Comparison between low and high groups led to a more significant correlation (P-value estimated to be 0.0032; see Results section). www.impactjournals.com/oncotarget DISCUSSION Multiple lines of evidence have suggested that SLC22A18 functions as a tumor suppressor. They include chromosomal location of the gene at 11p15.5 as well as missense mutations and aberrant splicing found in tumor tissues of various cancer types [7][8][9]. Furthermore, gene silencing and loss-of-heterozygosity of this paternally imprinted gene mostly occurs to the maternal allele [10][11][12]. Although low-level expression in glioma has been reported to be associated with poor prognosis [15], mechanistic studies for its tumor suppressor activity have been sparse. To the best of our knowledge, this study is first to report G2/M arrest induction by SLC22A18 and the mutually inhibitory activity with KRAS. Although inhibition of colony formation, induction of cell cycle arrest and inhibition of xenograft tumor growth were seen from overexpression of SLC22A18 which in itself could have toxic effects, epistatic relationship with KRAS supports that the observed tumor suppressor activities of SLC22A18 is endogenous to the gene.
Interestingly, the down-regulation of SLC22A18 in tumor tissue is not seen in other types of cancer but is restricted to colon and rectal cancers although the analyses based on matched normal-tumor tissues were available only for a limited number of cancer types. Of note, at least in the case of glioma, for which TCGA does not provide matched sample data, one report indicated that SLC22A18 functions as a tumor suppressor with lower expression in tumor tissues [15]. That SLC22A18 functions as a tumor suppressor is strongly supported by down-regulated expression in CRC, but it remains to be determined if the protein can inhibit cell cycle progression or KRAS signaling in other types of cancer cells, particularly in those where mutations of SLC22A18 have been found. We have also shown that SLC22A18 is a potential prognostic marker for CRC: lower expression appears to be associated with poor long-term survival. Clearly however, the prognostic value of SLC22A18 needs to be re-examined with a larger number of patients.
Among the critical remaining questions is how the organic ion transporting activity of SLC22A18 is connected with its tumor suppressor activity. The class of organic solutes which SLC22A18 transports is not known. In fact, the assignment of SLC22A18 as an organic ion transporter is based on the SLC22 subfamily membership but not on exact functional analysis. Still, it should be noted that SLC22A18 is expressed highly in organs with metabolite transport functions including intestine, kidney, placenta and liver [8]. Importantly, it has been proposed that solute carriers play key inter-and intracellular functions in sensing and signaling to maintain homeostasis in response to physiological changes [19]. Therefore, a better understanding of the exact function of SLC22A18 as a solute transporter is likely to be a step toward understanding its function as a tumor suppressor of CRC.

Patient samples and RNA isolation from tissues
Primary colorectal cancer and noncancerous colon samples used in RT-PCR study have been previously described [5]. The study was approved by the local institutional review board. The procedures for tissue processing and RNA isolation have been described [5].

Cell lines and culture conditions
Human colorectal carcinoma cells, HCT116, HT29 and SW480, were obtained from the American Type Culture Collection (ATCC, Manassas, VA). HCT116 and HT29 cells were cultured in McCoy's 5A, and SW480 cells were cultured in RPMI-1640 supplemented with 10% fetal bovine serum (Hyclone, Logan, UT). NIH3T3 fibroblast cells were also purchased from the ATCC and cultured in DMEM supplemented with 10% calf serum (Invitrogen, Carlsbad, CA).

Real-time RT-PCR and conventional RT-PCR analysis
Total RNA from cultured cells was extracted using the TRI Reagent ® (Ambion, Austin, TX). Single-stranded cDNA was synthesized from 2 μg of total RNA using ImProm-II ™ reverse transcriptase (Promega, Madison, WI). For the quantitative analysis of SLC22A18 and KRAS mRNA levels, cDNA generated from 10 ng of total RNA was subjected to PCR amplification using Kapa SYBR Fast qPCR kit (Kapa Biosystems, Boston, MA) on a CFX96 Real-time PCR detection system (Bio-Rad, Hercules, CA). ACTB and HPRT1, two internal control genes, were used as dual reference genes. Primer sequences are available online (Supplementary Table 1). Cycling conditions were as follows: pre-denaturation for 2 min at 95°C, a 2-step reaction (40 cycles) of 10 sec at 95ºC and 40 sec at 60ºC, and a dissociation peak analysis. The mRNA expression value of target genes was calculated with Bio-Rad CFX Manager Software. For conventional RT-PCR analysis, cDNA was amplified using Platinum Taq DNA polymerase (Invitrogen) and SLC22A18 primers. Cycling conditions were as follows: pre-denaturation for 2 min at 95°C, a 3-step reaction (28 cycles) of 15 sec at 94°C, 15 sec at 60°C and 30 sec at 72°C with a final extension of 10 min at 72°C. ACTB expression was used as an internal control, and PCR products were visualized by agarose gel electrophoresis.

Immunohistochemical staining
Immunohistochemistry (IHC) for SLC22A18 was performed on 4-µm frozen sections of clinically obtained colon cancer tissues. Tissue sections were deparaffinized with xylene, hydrated in serial dilutions of alcohol, and immersed in 3% H 2 O 2 . Following antigen retrieval in citrate buffer (pH 6.0), the tissue sections were incubated with protein blocking agent (ScyTek, Logan, UT) to avoid non-specific antibody binding for 30 minutes at room temperature and then incubated overnight with primary antibody against SLC22A18 (1:400; LSBio, Seattle, WA) at 4°C in a humidified chamber. After washing with PBS three times, the sections were incubated with a biotinylated secondary antibody and streptavidin conjugated to horseradish peroxidase (ScyTek) for 60 minutes at room temperature followed by a PBS wash. The liquid 3, 3'-diaminobenzidine (ScyTek) for chromogen was used for development and was followed by counterstaining with Meyer's hematoxylin.

Retrovirus vector construction
To generate retroviruses expressing SLC22A18 or KRAS G12D , the coding region of each gene was subcloned into the LZRS retroviral vector plasmid (pLZRS) along with IRES-GFP (internal ribosome entry site-green fluorescent protein) as previously described [20,21]. The control virus contained just the IRES-GFP in pLZRS. Production of high titer virus preparation in 293GPG cells was carried out as described with minor modifications [22]. Further details on construction of viral vectors and generation of pseudotyped viral particles will be provided upon request. Retroviruses encoding SLC22A18 or KRAS were infected with polybrene (4 µg/ml) into HCT116 or NIH3T3 cells. After 48 hours, the efficiency of infection was determined using fluorescence from GFP.

Flow cytometry
Two days after retroviral infection, the cells were trypsinized and rinsed with ice-cold PBS and resuspended at a concentration of 7x10 5 cells/ml in PBS. After fixation with 70% ethanol and washing with PBS, the cells were stained with 50 µg/ml propidium iodide (Sigma, St. Louis, MO) in PBS containing 0.1% Triton X-100 and 1 µg/ml RNase (Sigma) for 20 min. Flow cytometric analysis was carried out on a BD LSRFortessa cell analyzer (BD Biosciences, San Jose, CA). Typically, fluorescence distribution of a total of 10,000 nuclei was analyzed using BD FACSDiva™ software.

Transfection of siRNAs
Specific siRNAs for KRAS and the scrambled negative control (NC) siRNA were purchased from Qiagen (Germany). Target sequences are available on line (Supplementary Table 2). The siRNA duplexes were transfected to 1.5x10 5 cells in 35mm dishes using Lipofectamine RNAiMAX (Invitrogen) at the concentration of 50 nM for 48 hours according to the manufacturer's instructions.

Western blotting analysis
For immunoblotting, cells were lysed in cold lysis buffer (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 2 mM Na 2 EDTA, 1% NP-40, 0.1% sodium dodecyl sulfate, 0.5% Na-deoxycholate, and 10 mM NaF) supplemented with a mixture of protease inhibitors (Sigma) and phosphatase inhibitors (Sigma). After incubating on ice for 20 min, the supernatants were isolated by centrifugation at 13,000 rpm for 20 min. The concentration of total protein was determined using the BCA protein assay kit (Thermo Scientific Pierce, Rockford, IL). For detection of SLC22A18, cells were resuspended in Laemmli buffer, boiled for 10 min and centrifuged. Antibodies against p-cdc2 (T161), p21 Waf1/Cip1 (12D1), Cyclin B1 and cdc25C were purchased from Cell Signaling Technology (Beverly, MA). Monoclonal antibody against the V5 epitope was purchased from Invitrogen. Antibodies to KRAS, SLC22A18 and α-Tubulin were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), LSbio and AbFrontier (Seoul, Korea), respectively. Proteins were detected using peroxidase-conjugated anti-mouse-IgG or anti-rabbit-IgG antibodies in combination with enhanced chemiluminescence detection kit (Amersham-Pharmacia Biotech, Piscataway, NJ).

Colony formation assay
Two SLC22A18 expression plasmids, one with V5 epitope in the C-terminus and another with Xpress epitope in the N-terminus were constructed in pcDNA3.1 vector. Cells (1.5x10 5 ) were seeded in a 35mm dish and transfected with SLC22A18 plasmids or pcDNA3.1 plasmid (2 µg each) using Lipofectamine 2000 (Invitrogen) for HCT116 cells and HT29 and Lipofectamine LTX (Invitrogen) for SW480 cells following the manufacturer's protocols. Forty-eight hours after transfection, HCT116 and HT29 cells were placed under selection with G418 at 800 µg/ml (Invitrogen) respectively for 10 and 18 days. In the case of SW480, cells were treated with G418 at 800 µg/ml for 8 days initially and at 400 µg/ml for 6 days subsequently. Colonies were stained with 0.1% Coomassie Blue in 45% methanol and 10% acetic acids solution, and the colony numbers were determined using Gel Doc XR system (Bio-Rad) with Quantity One ® 1-D analysis software (Bio-Rad).

Soft agar assay
For soft agar assays, 5x10 3 (per 21mm dish) retrovirus-transduced NIH3T3 cells were suspended in 0.35% agar containing DMEM with 10% calf serum and overlaid onto a 0.9% agar solution. Colonies were stained after 21 days with MTT solution at the concentration of 1 mg/ml overnight. Colony numbers were determined using Gel Doc XR system with Quantity One ® 1-D analysis software. Details are available upon request.

Tumor xenograft experiments
One day after infection with control viruses or viruses encoding SLC22A18, HCT116 cells were trypsinized and counted. A total of 2 x 10 6 cells in 100 μL of phosphate-buffered saline were injected subcutaneously into the right flank of 7-week-old female BALB/c Slc-nu/ nu mice (Japan SLC, Japan). Tumor size was measured three times weekly with calipers for approximately 5 weeks, and the tumor volume was calculated using the equation: volume = (W 2 xL)/2 (W, the shortest dimension; L, the longest dimension). This study was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Ewha Womans University.

Statistical analysis
Paired t-test (one-or two-tailed) was used to evaluate the significance of results from expression analyses and colony formation, cell cycle and soft agar assays. For tumor xenograft experiments, unpaired t-test was used.

Analysis of the TCGA COAD data
Gene expression data from TCGA colon adenocarcinoma (COAD) consortium were downloaded at level 3 from the Broad GDAC Firehose website (released on April 16, 2014). RNA-Seq experiments of TCGA COAD had been performed in two different Illumina platforms of HiSeq and Genome Analyzer for 430 patients (n = 259 and 171, respectively), and SLC22A18 expression was analyzed separately from each platform data. For survival analysis, patients were divided into three groups according to SLC22A18 expression levels. The log-rank test was performed for Kaplan-Meier survival analysis of three groups using the Survival package version 2.37-7 (http://CRAN.R-project.org/package=survival).