Novel Gastric Cancer Stem Cell-Related Marker LINGO2 Is Associated with Cancer Cell Phenotype and Patient Outcome

The expression of leucine-rich repeat and immunoglobulin-like domain-containing nogo receptor-interacting protein 2 (LINGO2) has been reported in Parkinson’s disease; however, its role in other diseases is unknown. Gastric cancer is the second leading cause of cancer death. Cancer stem cells (CSC) are a subpopulation of cancer cells that contribute to the initiation and invasion of cancer. We identified LINGO2 as a CSC-associated protein in gastric cancers both in vitro and in patient-derived tissues. We studied the effect of LINGO2 on cell motility, stemness, tumorigenicity, and angiogenic capacity using cells sorted based on LINGO2 expression and LINGO2-silenced cells. Tissue microarray analysis showed that LINGO2 expression was significantly elevated in advanced gastric cancers. The overall survival of patients expressing high LINGO2 was significantly shorter than that of patients with low LINGO2. Cells expressing high LINGO2 showed elevated cell motility, angiogenic capacity, and tumorigenicity, while LINGO2 silencing reversed these properties. Silencing LINGO2 reduced kinase B (AKT)/extracellular signal-regulated kinase (ERK)/ERK kinase (MEK) phosphorylation and decreased epithelial-mesenchymal transition (EMT)-associated markers—N-Cadherin and Vimentin and stemness-associated markers— POU class 5 homeobox 1 (OCT4) and Indian hedgehog (IHH), and markedly decreased the CD44+ population. These indicate the involvement of LINGO2 in gastric cancer initiation and progression by altering cell motility, stemness, and tumorigenicity, suggesting LINGO2 as a putative target for gastric cancer treatment.


Introduction
Gastric cancer is the fourth most common malignant tumor worldwide and the second leading cause of cancer death [1]. Although gastric cancer mortality has declined recently due to early diagnosis by endoscopy, its prognosis is still poor, and has a 5-year survival rate of 20% in most parts of the world, except in Korea and Japan [2]. The 5-year survival rate of gastric cancer is about 70% in Korea and Japan; this may be due to the effectiveness of the mass screening programs [3,4].
Cancer stem cells (CSCs) are a subpopulation of cancer cells that have a high self-renewal capacity within the tumors. Typically, the CSCs constitute less than 5% of total tumor cells and are critical in cancer initiation, invasion, metastasis, and drug resistance [5][6][7]. Recent studies have shown that cancer cells undergoing epithelial-mesenchymal transition (EMT) share many properties with CSCs [8][9][10]. Epithelial cells are tightly associated with neighboring cells through E-cadherin-containing adherent junctions. EMT is the process by which epithelial cells undergo phenotypic changes such as loss of cell-cell adhesion, loss of cellular polarity and enhanced migration and invasion capacity to become mesenchymal-like cells [11]. Aberrant activation of EMT disturbs normal epithelial homeostasis, inducing pathologic alterations such as fibrosis, cancer cell invasion, and metastasis to distant secondary sites. Thus, EMT is thought to be the initial and most important step in the cancer metastasis cascade [12][13][14]. Gastric CSCs were first reported by Takaishi et al. in experiments showing that CD44-positive gastric cancer cells have stem cell-like properties and showed an increased resistance to traditional chemotherapies and radiotherapies [15]. Moreover, CD44 expression was significantly associated with the expression of EMT-related molecules E-cadherin, Vimentin, Snail-1, and ZEB-1 in gastric cancer patients [16].
Using microarray analysis, we identified leucine-rich repeat and immunoglobulin-like domain-containing nogo receptor-interacting protein 2 precursor (LINGO2) as a novel gastric CSC-related marker. LINGO2 gene is located on chromosome 9 at 9p21.2. This membrane protein is 606 amino acids long and weighs 68 kDa. LINGO2 is a human paralog of LINGO1, with 61% sequence identity [17]. LINGO1 and LINGO2 are associated with essential tremor and Parkinson's disease; however, the function of LINGO2 in carcinogenesis is unknown [18,19].
In our studies, we explored the role of LINGO2 as a regulator of cell motility and stemness in gastric cancer cells. LINGO2 expression was significantly increased in patients with advanced gastric cancer. Moreover, the overall survival of patients with strong LINGO2 expression was significantly shorter than that of patients with weak LINGO2 expression. Cells that highly expressed LINGO2 showed increased tumorigenicity and angiogenic capacity compared to cells with low LINGO2 expression. In addition, suppressing LINGO2 expression by shRNA decreased cell motility and down-regulated CSC-related markers. Our data suggest that LINGO2 is a new therapeutic target for gastric cancer which targets EMT as well as CSC.

LINGO2 is Differentially Expressed in Gastric Cancer Sphere Cells
The sphere-forming capacity of different human gastric cancer cell lines including AGS, N87, SNU1, SNU5, SNU16, SNU484, SNU601, SNU638, and SNU668 was determined. N87 and SNU484 reproducibly formed spheres in vitro whereas other cell lines did not form spheres and remained as aggregated cell clusters ( Figure 1A). The expression of stemness-related genes including jagged canonical notch ligand 2 (JAG2), notch receptor 3 (NOTCH3), hes family bHLH transcription factor 1 (HES1), Indian hedgehog (IHH), smoothened, frizzled class receptor (SMO), GLI family zinc finger 1 (GLI1), frizzled class receptor 1 (FZD7), β-Catenin, Nanog homeobox (NANOG), POU class 5 homeobox 1 (OCT4) and phosphatase and tensin homolog (PTEN) were elevated in spheres as determined by polymerase chain reaction (PCR) analysis and was further confirmed by microarray analysis (Supplementary Figure S1 and Supplementary Figure S2A). Among the differentially expressed genes, we focused on membranous as well as secretory proteins and observed that LINGO2 was up-regulated by 1.31-folds in N87 cells and 2.88-folds in SNU484 cells (Supplementary Figure S1 and  Supplementary Table S1). We confirmed the elevated mRNA and protein expression of LINGO2 by PCR and western blot analysis, respectively ( Figure 1B,C). Treatment with an anti-LINGO2 antibody decreased sphere numbers and sizes in both N87 and SNU484 cells (Supplementary Figure S2B). LINGO2 was constitutively expressed in all gastric cell lines with differences in basal expression levels ( Figure 1D).

Figure 1.
Expression of LINGO2 in gastric cancer cell-spheres (A) Gastric cancer cell lines including AGS, N87, SNU1, SNU5, SNU16, SNU484, SNU601, SNU638, and SNU668 were maintained with growth media and plated onto culture dishes or ultra-low attached plates in sphere media for sphere-formation assays and N87 and SNU 484 formed spheres. (B) LINGO2 mRNA expression and (C) protein expression was analyzed by RT-PCR and western blot, respectively, and LINGO2 mRNA as well as protein expression was elevated in gastric cancer sphere cells. (A, adherent cells; SP, sphere cells) (D) LINGO2 expression in all gastric cancer cell lines was analyzed by PCR analysis.

Increase in LINGO2 Expression Elevates Cancer Stem Cell Characteristics
We sorted SNU484 and N87 cells based on LINGO2 expression by flow cytometry (Figure 2A, Supplementary Figures S3A and S4A). The protein levels, motility, and tumorigenicity of SNU484 cells with high LINGO2 expression (LINGO2 high ) and lower LINGO2 expression (LINGO2 low ) were compared in vitro and in vivo. Higher expression of stemness-associated proteins including OCT4, PTEN, GLI1, and HEY1 was observed in LINGO2 high cells compared to LINGO2 low . SNU484 LINGO2 high cells showed an approximately 2-fold increase (208% ± 24.6, p < 0.1) in cell migration and 4-fold increase (467% ± 15.8, p < 0.001) in clonogenic ability compared to SNU484 LINGO2 low cells ( Figure 2B-D). N87 LINGO2 high cells also showed a similar increase in clonogenicity compared to the N87 LINGO2 low cells, in vitro (Supplementary Figure S3A).
To determine tumor-initiating ability, sorted SNU484 cells were suspended in Matrigel and injected subcutaneously to the hind flanks of NOD/SCID mice (n = 3 per group). Tumor formation was observed with 250 LINGO2 high cells while LINGO2 low cells required more than 1000 cells to form a tumor mass ( Figure 2E). Tumor mass formed from the same number of LINGO2 high and LINGO2 low cells differed in not only its size but also the overall color; LINGO2 high tumors were reddish whereas LINGO2 low tumors were nearly white. Similar results were observed when LINGO2 high and LINGO2 low cells were injected in BALB/c nude mouse (n = 1, Supplementary Figure S4B). We immuno-stained the mouse tissue slides for LINGO2, stemness marker CD44, angiogenesis marker phopho-vescular growth factor receptor 2 (p-VEGFR2), blood vessel marker CD34, mesenchymal marker N-Cadherin, and epithelial marker Occludin, followed by hematoxylin and eosin (H&E) staining ( Figure 2F). SNU484 LINGO2 high tumors with up-regulated LINGO2 displayed up-regulated CD44, CD34, p-VEGFR2, and N-Cadherin but down-regulated Occludin compared to LINGO2 low tumors, suggesting the potential involvement of LINGO2 in angiogenesis and EMT.

Silencing LINGO2 Reduces Cell Proliferation and Motility
To determine the functional role of LINGO2, we suppressed LINGO2 expression in gastric cancer cell line SNU484 using shRNA. Cells transfected with LINGO2 shRNA became more rounded and cells with tapered ends disappeared ( Figure 3A). LINGO2 silencing led to a decrease in SNU484 cell proliferation by 23.6% ± 9.1% (p < 0.001) and migration by 95.5% ± 1.1% (p < 0.001) ( Figure 3B,C). Wound-healing ability was assessed, and wounds started to heal in 24 h in control cells while the healing process required more than 30 h in LINGO2 shRNA-transfected cells. Figure 3D shows the representative healing state at 24 h after creating the scratch in the cell monolayer. We analyzed the effect of LINGO2 silencing on the expression of proteins ( Figure 3E). When LINGO2 expression was suppressed stemness-associated proteins including OCT4, IHH, and β-Catenin expression was markedly altered-a decrease in OCT4 and IHH and an increase in β-Catenin was observed. Kinase B (AKT)/extracellular signal-regulated kinase (ERK) signaling-related proteins which are usually elevated in cancer cells were also altered with LINGO2 suppression-an increase in phospho-Glycogen synthase kinase-3 (pGSK3β) and a decrease in total AKT, phospho-AKT (pAKT), phospho-ERK kinase (pMEK), total ERK, and phospho-ERK (pERK) was observed. EMT-associated proteins including N-cadherin and Occludin were also altered in LINGO2 shRNA-transfected cells-a decrease in mesenchymal marker N-Cadherin and an increase in epithelial marker Occludin was observed. A marked decrease in mesenchymal marker Vimentin was also observed by immunofluorescence staining. A pronounced change in cell morphology was observed in LINGO2 shRNA-transfected cells ( Figure 3F). Flow cytometry analyses revealed a decrease in stemness-associated cell surface marker CD44. The percentage of CD44-positive cells was 19.7 ± 0.85% and 0.8 ± 0.14% in mock and LINGO2 shRNA-transfected cells, respectively ( Figure 3G).

LINGO2 Alters Secretion of Matrix Metallopeptidases and Angiogenic Factors
We performed western blot analysis and gelatin zymography to assay the expression and activities of matrix metallopeptidases (MMPs). The activity of MMPs is critical for cancer cells to invade through extracellular matrices. Significant reduction in MMP-1 and -9 protein expression was observed in LINGO2-silenced cells whereas no MMP-2 was detected in either LINGO2-silenced or parent cells. Active MMP-9 levels were markedly reduced in LINGO2-silenced cells, whereas MMP-1 and MMP-2 expression levels remained unaltered on gelatin zymography ( Figure 4A,B). Tube formation assay using human umbilical vein endothelial cells (HUVEC) in vitro assay to model the reorganization stage of angiogenesis was performed to measures the ability of endothelial cells to form capillary-like structures. Cell culture supernatants from LINGO2-silenced cells and mock translated cells supplemented with or without vesicular epithelial growth factor (VEGF) was used to culture HUVECs. LINGO2 silencing decreased tube formation by 35.6% ± 9.1% (p < 0.001) ( Figure 4C,D).

Clinical Classification of Gastric Cancer Patients Based on LINGO2 Expression
We evaluated the LINGO2 expression in 103 samples human gastric cancer by immunohistochemistry (IHC). Patients were divided into LINGO2 weak and LINGO2 strong groups based on LINGO2 expression. Patients with LINGO2 IHC staining scores of 0 or >1 were classified as LINGO2 weak group and patients with scores of >2 or >3 were classified as LINGO2 strong group ( Figure 5A).  The result of comparative analysis between LINGO2 weak and LINGO2 strong group are presented in Table 1. The LINGO2 strong group presented with a higher rate of advanced T stages (T3 or 4: 67.9% vs. 34.7%, p = 0.003), N stage (N2 or 3: 50.0% vs. 22.7%, p = 0.005) and an overall stage (stage III: 50.0% vs. 17.3%, p = 0.001). The OS was significantly shorter in LINGO2 strong group (25.5 vs. 121.2 months, p = 0.012) ( Figure 5B). Moreover, the yearly survival rate after surgery was significantly lower in the LINGO2 strong group (p = 0.004, 0.036, 0.039, and 0.051 at 2, 3, 4, and 5 years after surgery, respectively).
Univariate analysis using Cox-regression identified that the LINGO2 strong group was significantly associated with OS (HR 1.939, p = 0.014), older age and advanced stage (all p < 0.001, Table 2). Subsequent multivariate analysis revealed that older age (p < 0.001) and advanced disease (p < 0.001) were significantly associated with OS. However, LINGO2 strong group was not associated with OS (HR 1.245, 95% CI 0.674-2.301, p = 0.484) after adjustment for the effect of other compounding factors by multivariate analysis using Cox-regression. Although LINGO2 gene was identified as a CSC marker from cancer cells and cancer sphere cells, it is possible that non-tumor cells also express LINGO2. We performed IHC and immunofluorescence staining on non-tumor gastric tissues, gastric polyp tissues and spasmolytic polypeptide-expressing metaplasia (SPEM) tissues along with gastric cancer tissues ( Figure 6A-D). The expression of LINGO2 was barely detected in normal gastric tissues and gastric polyp tissues while moderate expression was observed in SPEM tissues. To visualize the location where LINGO2 expresses, we stained serial sections of SPEM tissues and stained with LINGO2 and SPEM marker, TFF2 ( Figure 6E,F) [20][21][22][23][24]. Interestingly, cells expressing TFF2 also expressed LINGO2 which suggest that LINGO2 expresses in SPEM cells. To clarify the expression site, we performed immunofluorescence staining with both TFF2 and LINGO2 and co-expression was consistently observed in independent SPEM tissues ( Figure 6G,H).

Discussion
LINGO2 is a novel protein encoded by the LINGO gene family first reported by Carim-Todd et al. in 2003; they observed LINGO2 expression in the central nervous system during early developmental stages and in the limbic system and cerebral cortex in adult tissues [25]. Haines et al. analyzed the LINGO expression during mouse embryonic development, reporting that LINGO2 was expressed in neuronal tissues along with LINGO4, NGL1, SALM1, SALM5, and TRKB [26]. Vilarino-Guell et al. analyzed the expression of and mutations in LINGO1 and LINGO2, reporting an association between LINGO1 and LINGO2 mutation and the risk of essential tremors and the incidence of Parkinson's disease [18]. The association of LINGO2 with carcinogenesis has only been discussed with respect to gene location involved in different mutations. Klorin et al. reported that LINGO2 is one of the recurrent homozygous deleted genes in malignant mesothelioma cell line and revealed that the LINGO2 gene located in the proximal region of 9p21.2 was co-deleted with the 9p21.3 region of cyclin-dependent kinase inhibitor 2A (CDKN2A)-a well-known tumor suppressor [27]. Li et al. reported that genetic polymorphisms in the 9p21 region, which include the LINGO2 gene are associated with the risk of multiple cancers by single nucleotide polymorphism (SNP) analysis [28]. The potential role of the LINGO family in embryogenesis and in adult neuronal cells has been suspected; however, neither the mechanism of action of the LINGO family, including LINGO2, nor the potential molecular targets of any LINGO family member are yet known.
The cancer stem cell theory, demonstrated by Reya et al. in 2001, proposes that cancers are derived from a stem cell compartment in a multistep process involving the accumulation of mutations in a variety of tumor suppressors and oncogenes [29]. CD44 is a well-known gastric CSC marker which is also a CSC marker in various other cancers [15,30]. Other markers including CD24/CD44, CD54/CD44, EpCAM/CD44, ALDH, CD90, and CD133 are also known to be gastric CSC markers. Cells expressing CSCs markers form cell-spheres and show increased cell motility [31][32][33][34][35][36][37]. The ability of a cell to form a sphere in nonadherent culture conditions is reflective of its self-renewal capacity. Cell-spheres are enriched in CSC population, and thus, can be used as sources to isolate CSCs [38,39]. We observed sphere-formation in two gastric cancer cell lines N87 and SNU484. The spheres passaged multiple times in vitro without the loss of sphere-forming capability or self-renewal capacity. Differentially expressed genes in the cell-spheres were analyzed by cDNA microarray analysis. Our studies identified LINGO2 as a novel CSC-associated gene.
In this study, we found that LINGO2 is expressed in gastric cancer tissues and regulates cell motility, tumorigenesis, and angiogenesis. Silencing LINGO2 by shRNA decreased stemness-associated genes including OCT4 and IHH as well as the CD44-positive cell population. LINGO2 silencing altered cell motility decreased N-cad, and Vimentin expression, as well as cell migration and wound-healing capacity. Cells highly expressing LINGO2 showed contrasting phenotypes including an elevated expression of stemness-associated molecules. LINGO2-silenced cells and LINGO2 low cells showed reduced in N-Cadherin, Vimentin, and MMPs, as well as cell migration compared to that of control vector, transfected cells, or LINGO2 high cells, suggesting that LINGO2 plays a role in EMT. LINGO2 silencing inhibited AKT and MEK/ERK phosphorylation, suggesting its role in the MEK-ERK signaling cascade. LINGO2 high showed an increase in tumorigenicity compared to LINGO2 low and generated a higher number of colonies in soft agar and required fewer cells to form a tumor in vivo compared to LINGO2 low cells. We performed tumorigenesis assay in two mouse strains NOD/SCID and BALBc/nude. LINGO2 high tumor mass showed a reddish color compared to that of white mass formed by LINGO2 low cells in both mouse models. This color difference suggested the possible relation of LINGO2 with angiogenesis. The elevated expression of pVEGFR2 and CD34 in LINGO2 high tumor tissues and the inhibition of HUVEC tube formation by cell culture supernatant from LINGO2-silenced cell supernatant which further supports the involvement of LINGO2 in angiogenesis. However, the specific role of LINGO2 in angiogenesis, cell motility, and stemness, and the mechanism underlying the LINGO2 mediated increase in stemness needs to be further investigated.
We analyzed LINGO2 expression in patient-derived tissues by IHC. We observed different levels of LINGO2 expression in all gastric cancer patients. Interestingly, patients with high LINGO2 expression (27%) showed an advanced clinical stage and decreased OS compared to patients with low LINGO2 expression (73%). These results were well supported by the functional analysis that silencing of LINGO2 reduced cell proliferation and motility in vitro, and tumor mass formation and angiogenesis were exaggerated in LINGO2 high cells than LINGO2 low cells in vivo. A more interesting finding was that LINGO2 expression was barely detected in both normal gastric tissues and adenomas polyp tissues, but a moderate expression, though not as strong as in cancer tissues, was observed in SPEM region in non-tumor gastric tissues. It is known that SPEM evolves from transdifferentiated chief cells triggered by the loss of parietal cells in the fundic mucosa [40]. With continuous inflammatory stimulation, intestinal metaplasia can occur from pre-existing SPEM and progress to cancer [41,42]. We observed the localization by immunofluorescence staining and LINGO2 expression was observed where TFF2, a known SPEM marker, was expressed. Therefore, LINGO2 was highly suspected to be associated with the development and growth of gastric cancer as well as the progression of gastric cancer from early to late stage. This is the first study to demonstrate the role of LINGO2 and its expression in human gastric cancer tissues.

Microarray Analysis
Total RNA was extracted from adherent cells and cell-spheres from N87 and SNU484 gastric cancer cell lines using RNeasy Mini kits (Qiagen, Valencia, CA, USA). Microarray procedures were carried out according to the manufacture0072's protocols. Briefly, 6-µg aliquots of total RNA were used to prepare double-stranded cDNA. cDNAs were amplified by PCR and labelled with biotin using the IVT labelling kit (Affymetrix, Santa Clara, CA, USA). Labelled cRNA was fragmented and hybridized to an Affymetrix GeneChip Human Genome U133 plus 2.0 high-density oligonucleotide arrays (Affymetrix). Microarrays were then washed using a GeneChip Fluidics Station 450 (Affymetrix) and scanned using a GeneChip Array Scanner 3000 7G (Affymetrix). Expression data were generated using Affymetrix Expression Console software version 1.1 using MAS5 algorithm normalization. Expression intensity data in CEL file was normalized using the MAS5 algorithm to reduce noise.

Immunofluorescence Staining
Cultured cells were fixed in methanol and immunostained with mouse monoclonal anti-Vimentin (1:100, Santa Cruz Biotechnology) overnight at 4 • C. Stained cells were incubated with goat anti-mouse Cy3-conjugated secondary antibody (Jackson ImmunoResearch Inc., West Grove, PA, USA) for 30 min. Nuclei were labelled with 4 ,6-diamidino-2-phenylindole (DAPI). Stained cells were analyzed on an Olympus BX51 microscope and images captured using an Olympus DP71 camera (Olympus America Inc., Center Valley, PA, USA). Tissue slides were were deparaffinized in xylene and rehydrated in graded alcohol. Endogenous peroxidase activity was blocked with 0.3% (v/v) hydrogen peroxide in methanol. Antigen retrieval was performed by microwaving the slides in sodium citrate buffer (0.01 M, pH 6.0) for 5 min. To block nonspecific staining, sections were incubated with 10% (v/v) normal donkey serum for 1 h and incubated with mouse monoclonal anti-LINGO2 (1:50, LifeSpan BioSciences Inc.) and mouse monoclonal anti-TFF2 (1:25, Leica Biosystems Newcastle Ltd., Newcastle Upon Tyne, UK) overnight at 4 • C. The slides were visualized using Cy3-goat anti-mouse IgG 2a (LINGO2, 1:100, Jackson ImmunoResearch Inc.) and Alexa488-Donkey anti-mouse IgM (TFF2, 1:50, Jackson ImmunoResearch Inc.) in antibody diluent. and incubated for 30min at room temperature. Between each step, there were three washing steps of 5min each on a rocking platform in PBS. The slides were cover slipped using mounting medium for fluorescence with DAPI (Vecta shield H-1200; Vector Laboratories, Inc. Burlingame, CA, USA).

Cell Proliferation Assays
Cells were detached and plated in triplicate at a density of 1.0 × 10 4 cells/well in 24-well plates in complete medium. Cells were detached every two days for eight days and stained with Trypan blue (Sigma-Aldrich, St. Louis, MO, USA) and counted using a hematocytometer.

Migration and Invasion Assays
For the migration assay, cells were detached and suspended at 1.0 × 10 5 cells/mL in serum-free media and plated at a density of 1.0 × 10 4 cells/well in 24-well Transwell plates (Costar, Bethesda, MD, USA). For the invasion assay, the upper chamber was pre-coated with Matrigel (1:4 diluted with serum-free medium, BD Biosciences, San Jose, CA, USA), and cells were seeded at 1.0 × 10 4 cells/well. The bottom chamber was filled with the culture medium from NIH-3T3 fibroblasts. Cells were incubated for 24 h for migration assays and 72 h for invasion assays. After incubation, cells were fixed with 5% glutaraldehyde for 30 min and stained with 0.1% crystal violet. Cells were completely removed from the upper surface of the membrane with a moist cotton swab. Migrated and invaded cells were counted and photographed under a microscope at 100× magnification. All assays were performed in triplicate.

Wound-Healing Assays
A scratch was made across a cell culture dish at approximately 90-95% cell confluence using a yellow pipet tip. The cells were incubated at 37 • C and photographed at 40× magnification during the time course of healing.

Clonogeinc Assays
For colony formation, the culture plate was pre-coated with base agar 0.6%, and cells were resuspended with 500 cells/well in top agar 0.3% with complete medium in 24-well plates. The plates were incubated at 37 • C with 5% CO 2 until colonies were visible. The colonies were then stained with 0.01% crystal violet and counted under an inverted microscope.

Gel Zymography
Cells were incubated in serum-free medium for 48 h, and the culture medium was collected and concentrated using Amicon ultrafilter devices (Millipore, Tullagreen, Ireland). Samples were loaded with non-reducing loading dye and electrophoresed through 10% zymogram gels (Bio-rad, Hercules, CA, USA). SDS was removed from the gels by incubation in 2.5% Triton X-100 (Sigma), and the gels were incubated at 37 • C for 16 h in development buffer (50 mM Tris-HCl, pH7.8, 200 mM NaCl, 5 nM CaCl 2 , 0.02% Brij-35), followed by staining with Coomassie blue. MMP activity was visualized as clear bands.

Tumorigenesis Assays
Gastric cancer cells sorted as LINGO2 high and LINGO2 low by flow cytometry, were washed with serum-free HBSS (Gibco BRL, Grand Island, NY, USA) and suspended in serum-free RPMI (Gibco BRL) and Matrigel (BD Biosciences PharMingen) (1:1 volume). Cells were subcutaneously injected into both flanks of NOD/SCID mice (n = 3 per group, Charles River Laboratories, Yokohama, Japan).
The animal experiments were approved by the Committee for the Care and Use of Laboratory Animals of Yonsei University College of Medicine. Tumor volume was calculated as V (mm 3 ) = (A 2 × B)/2, where A is the diameter perpendicular to the largest dimension B. After 14 to 16 weeks, mice were sacrificed and tumor tissues fixed in 4% paraformaldehyde. For histological evaluation, tissue samples were embedded in paraffin and stained with hematoxylin and eosin (H & E, Sigma-Aldrich).

Tube Formation Assay
Matrigel (BD Biosciences) was diluted with EBM-2 medium and coated in 96-well plates at 37 • C for 1 h. Then, 4 × 10 3 human umbilical vein endothelial cells (HUVECs) were seeded in the EBM-2 medium on Matrigel. The tube formation ability of HUVECs was measured at 6 h with RPMI with or without VEGF, or concentrated cell culture media. After incubation, the number of tubes and nodes of the tubular structures was quantified.

Patients
We performed a retrospective analysis of surgical tissue samples of gastric cancer patients between January 2002 and December 2002. All patients included in the study received gastric cancer surgery with curative intention. Information regarding patient demographics and clinical data was obtained from the electronic medical records, including age at diagnosis, sex, tumor stages at diagnosis, and serum levels of CEA. Tumors were staged based on the staging classification of the 7th edition of the American Joint Committee on Cancer (AJCC). All procedures involving human participants were performed in accordance with the ethical standards of the institutional research committee and the 1964 Helsinki declaration and its later amendments or comparable ethical standards. As this was a retrospective analysis, Yonsei University Health System, Severance Hospital, Institutional Review Board approval was obtained and the need for informed consent was waived-off.

Statistical Analysis
Data were analyzed using χ 2 and Fisher's exact for categorical data and Student's t-test and Mann-Whitney U test for continuous variables. Multivariate analysis was performed to evaluate possible significant factors, considering the influence of confounding clinical variables. Hazard ratios (HRs), 95% confidence intervals (95% CIs), and P values of multivariate analysis were calculated with a Cox proportional hazard model for OS. OS was estimated and compared using the Kaplan-Meier analysis with a log-rank test. All statistical analyses were performed using IBM SPSS Statistics for Windows, version 23.0 (IBM Corp., Armonk, NY, USA). A value of p < 0.05 was considered statistically significant. Asterisks (*, **, and ***) indicate significances at p ≤ 0.05, p < 0.1, and p < 0.01, respectively.

Conclusions
Our study reports for the first time that LINGO2 expression is elevated in stemness-enriched cells in gastric cancer cell lines, based on cDNA microarray analysis, and evaluation of tissues derived from gastric cancer patients. Furthermore, novel functions of LINGO2 in regulating cell migration, stemness, MMPs, tumorigenic ability, and angiogenesis were demonstrated, as well as the first clinical implication of LINGO2 expression correlated with gastric cancer progression. We propose that therapeutic targeting of LINGO2 may contribute to target gastric CSC to overcome conventional cancer therapy.

Conflicts of Interest:
The authors declare no conflict of interest.

LINGO2
leucine-rich repeat and immunoglobulin-like domain-containing nogo receptor-interacting protein 2 CSC cancer stem cells EMT epithelial to mesenchymal transition TMA tissue microarray analysis SPEM spasmolytic polypeptide-expressing metaplasia