Low FAT4 expression is associated with a poor prognosis in gastric cancer patients

In this study, we investigated the role of Fat atypical cadherin 4 (FAT4) in gastric cancer (GC) progression. Immunohistochemical analysis showed lower FAT4 expression in tumor tissues from GC patients than in normal gastric epithelium. Lower FAT4 expression was associated with poor prognosis, tumor size and invasion, and lymph node and distant metastases. Multivariate analysis showed that TNM stage, lymph node and distant metastases, Lauren classification, and FAT4 expression were independent prognostic factors in GC. Methylation-specific PCR analysis showed increased FAT4 promoter methylation in GC tumor tissues and cell lines. Higher FAT4 promoter methylation was associated with low FAT4 expression and a poor prognosis. BGC-823 cells showed increased FAT4 expression upon treatment with 5-azacytidine, demethylating agent. FAT4 knockdown in BGC-823 cells led to increased cell proliferation, migration and invasiveness. Moreover, xenografts of BGC-823 cells with FAT4 knockdown showed enhanced tumor growth and metastasis in nude mice. These findings demonstrate that low FAT4 expression is associated with a poor prognosis in GC patients.


INTRODUCTION
Despite recent developments in the surgical techniques and improved efficacy of anticancer drugs, mortality rates of gastric cancer (GC) patients remains very high. The overall 5-year survival rate of GC is about 40% because of late diagnosis (stage III and IV), which is associated with lymph node and distant metastasis [1]. Therefore, new prognostic and therapeutic biomarkers are necessary to improve the survival of GC patients.
The Fat gene family was originally identified in Drosophila as a member of the cadherin super-family with tumor suppressor functions [2,3]. It regulates cell proliferation and planar cell polarity during Drosophila development by the Hippo signaling pathway [4,5]. They encode a type 1 trans-membrane protein with 34 cadherin repeats, 4 epidermal growth factor (EGF)-like repeats, a transmembrane domain and a cytoplasmic domain that is distinct from the classical cadherin proteins [6,7]. In humans, four members of the Fat family have been identified, namely, FAT1, FAT2, FAT3 and FAT4, which are structurally similar to the Drosophila Fat protein [3]. In mammals, FAT4 is the true structural ortholog of the Drosophila FAT, whereas FAT1, FAT2 and FAT3 are closely related to the Drosophila Fat-like 8 . FAT4 mediates key developmental functions such as planar cell polarity and regulates the Hippo signaling pathway, which controls the size of the organs [3,4]. Whole-exome sequencing revealed that TP53, PIK3CA and ARID1A were critical cancer driving genes in GC and cell adhesion genes were frequently mutated [8]. Genomic mutations (5%) and deletions (4%) in FAT4 were observed in GC, whereas FAT4 silencing decreased cellular adhesion and increased migration and invasion of GC cells [8]. FAT4 suppresses growth and invasion in breast [9] and GC cancers [10][11][12] by activating the Yes-associated protein (YAP) and/or WNT signaling and silencing the epigenetic modifier proteins. However, the definitive role of FAT4 in GC distant metastases remains unknown. Therefore, in this study, we investigated the role of FAT4 in GC progression and prognosis. Table 1 shows the association between FAT4 expression and clinicopathological features based on IHC analysis of 449 clinical GC tissue microarray (TMA) samples. FAT4 staining was exclusively cytoplasmic in normal gastric epithelial cells and few cases of tumor cells ( Figure 1). Low FAT4 expression was observed in GC tissues than in adjacent noncancerous tissues ( Figure 1A-1F). FAT4 expression was detected in 263 (58.6%) out of 449 GC tumor samples analyzed. FAT4 expression reduced from 75.9% to 32.4% in the patients with lymphatic metastasis with increasing N grades (N0 to N4 grades). FAT4 expression decreased from 89.3% to 20.3% with increasing TNM grades (TNM Grade I-IV, P<0.05). FAT4 expression was lower in GC patients with lymph node metastasis (48.4%, or 137/283) than patients without lymph node metastasis (75.9%, or 126/166, P<0.05). FAT4 expression was lower in patients with distant metastasis (20.3%, or 13/64) than patients without distant metastasis (64.9%, or 250/385; P<0.05). Lower FAT4 expression correlated with tumor size, tumor invasion depth (T Grade) and vascular invasion (P<0.05, Table 1). FAT4 expression was not associated with age, sex, location, differentiation, Lauren classification and histological type of GC (P>0.05, Table 1).
The IHC and qRT-PCR data showed that FAT4 expression was higher at the early stages of the GC progression (N0 or T1+T2 grade), but gradually decreased during lymph node and distant metastatic events (from group1 to group3, Figure 1L). Low FAT4 expression during cancer metastasis was associated with decreased survival ( Figure 1K).

FAT4 promoter methylation is associated with poor prognosis in GC
Next, we analyzed FAT4 methylation status in 173 paired GC tissue samples to determine if methylation was associated with low FAT4 expression and GC progression. We identified a CpG island region of approximately 4.5 kb in the human FAT4 gene promoter region ( Figure 2A). We analyzed the methylation status of the FAT4 promoter region in paired GC and adjacent noncancerous tissues by methylation specific PCR (MSP) analysis. We found that 31.21% (54/173) of GC tissues were methylated (Table 3 and Figure 2B). FAT4 was methylated in AGS, MKN-28 and MKN-45 GC cancer cell lines and unmethylated in the normal GES-1 gastric epithelial cells ( Figure 2C). We performed pyrosequencing on 36 GC tissue samples and GC cell lines to verify the results from the MSP method. After re-extraction and quantification, we obtained 32 GC tissues and 6 cell lines for analysis. Four cases (No.7, 23, 29 and 36) had low quality DNA and hence could not be analyzed. We sequenced 6 CpG sites and determined the mean methylation percentage of each site relative to the corresponding matched noncancerous tissue. We classified the specimens based on mean methylation percentages into low (0-5%) and high (> 5%) methylated groups www.impactjournals.com/oncotarget   Figure 3A). These data demonstrated that FAT4 promoter was methylated in advanced GC cases.

Association between FAT4 promoter methylation status and GC clinicopathological factors
We treated GC tumor cell lines AGS, MKN-28 and MKN-45 with 1μM of 5-aza-2'-deoxycytidine (5-Aza-dC) for 48h and demonstrated increased expression of FAT4 mRNA (Supplementary Figure 4). This indicated that methylation of the FAT4 promoter silenced the expression of FAT4.

Association between FAT4 promoter methylation status and GC prognosis
We analyzed the association between FAT4 methylation status and prognosis of GC patients. In the cohort of 173 patients, the overall survival time was 33.69±1.32 months and the overall survival rate was 12  We classified the patient samples (n=179) based on lymph node and distant metastases status into three groups namely, group1 (N0, M0, n=50), group 2 (N>1, M0, n=64) and group 3 (N>1, M1, n=59). There is no difference of FAT4 expression between cancer tissue and adjacent noncancerous tissues in group1 and group2 (P>0.05), but in group3 the FAT4 expression in cancer tissue is significant lower than adjacent noncancerous tissues (P<0.05). FAT4 expression was significant lower in cancer tissues from group 2 and group3 patients than group 1 patients (P<0.05). www.impactjournals.com/oncotarget

FAT4 levels determine in vitro growth and invasiveness of GC cell lines
We knocked down FAT4 expression in the BGC-823 cell line that exhibits relatively higher FAT4 expression than other GC cell lines ( Figure 2C

FAT4 knockout GC cells promote tumorigenesis and lung metastasis in nude mice xenograft model
Next, we tested the effects of FAT4 knockdown on GC tumorigenesis and invasiveness in a nude mice xenograft model by transplanting BGC-823-FAT4-KO and BGC-823-KO-NC cells.
We analyzed the metastatic status in tail vein injection models after 5 weeks by PET-CT scanning. CT scans indicated the presence of lung nodules in the BGC-823-FAT4-KO group (red arrow), which was absent or ill-    Figure 6A). PET evaluation of the two groups of mice indicated high 18 F-FDG activity in the brain, bladder, heart and muscle tissues, but not in the lungs. Despite this observation, the BGC-823-FAT4-KO group mice showed higher 18 F-FDG activity and multiple tumors in the lungs than the BGC-823-KO-NC group mice ( Figure 6A; blue arrows). These findings are consistent with the presence of the macroscopic lung metastatic cancer nodules in the lung tissues ( Figures 6A  and 7A). Gross anatomical analysis of lung metastasis nodules showed higher and larger number of nodules in the BGC-823-FAT4-KO group than the BGC-823-KO-NC group (117.5±16.65 vs. 15.83±5.31, P<0.05; Figure 6B). Furthermore, tumors of the BGC-823-FAT4-KO group showed weak FAT4 staining and higher Ki-67 index than the BGC-823-KO-NC group ( Figures 5C-5D and 7B-7C). These findings were evident in both subcutaneous implant tumors and lung metastatic cancer nodules. Overall, these results indicate that loss of FAT4 expression promotes GC progression and metastasis.

DISCUSSION
Many developmental genes that are involved in cell proliferation and apoptosis regulation are deregulated in human diseases including cancers. The Fat gene family is a member of the cadherin super family, which controls cell proliferation and planar cell polarity during Drosophila development [13,14]. In mammals including humans, FAT family members regulate Hippo signaling pathway [9,12,13]. Whole-exome sequencing revealed mutations (5%) and deletions (4%) in FAT4 of GC cases [8]. FAT4 knockdown increased the migratory and invasive ability of breast cancer cells [9]. The Hippo pathway in mammals involves a kinase cascade that involves phosphorylation and activation of large tumor suppressor 1/2 (LATS1/2) by Ste20-like kinases 1/2 (MST1/2). This kinase cascade inhibits the activities of two transcriptional co-activators, Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ) [15]. When YAP and TAZ are activated,  they translocate into the nucleus and induce expression of a wide range of genes that regulate tissue homeostasis and organ size control [16]. Dysregulation of the Hippo pathway is associated with cancer development [17]. Mao Y et al showed that the planar cell polarity complex was composed of transmembrane cadherins, Fat and Dachsous, which regulated the Hippo pathway [18]. In Drosophila, Dachsous binds to Fat, which activates the Hippo pathway by inhibiting the interaction between Zyxin and warts (Wts) (17). In vertebrates, the homolog of Fat is FAT4.
In the present study, loss of FAT4 expression was associated with poor prognosis, tumor size, invasion depth, vascular invasion, lymph node and distant metastases. FAT4 mRNA levels were lower in GC tumor samples than adjacent noncancerous tissues. We observed gradual decrease in FAT4 expression as tumor progressed from a non-metastatic state to tumor associated with lymph node and distant metastasis. FAT4 expression was similar between cancer and adjacent noncancerous tissues in patients without lymph node metastasis. But, FAT4 expression was lower in cancer tissues than adjacent normal tissues in patients with distant metastasis. Our data is in accordance with previous studies that demonstrate the tumor suppressive role of FAT4 in various cancers [9,19].
In addition to point mutations, epigenetic modifications such as gene methylation inactivate tumor suppressor genes in tumors are very common [20][21][22]. Methylation of FAT4 has been reported previously in breast cancer [9]. In the present study, FAT4 promoter was methylated in 31.2% of GC patients. Methylation of the FAT4 promoter correlated with FAT4 expression levels. Patients with lymph node and/or distant metastases showed higher methylation at the FAT4 promoter than patients without metastasis. Therefore, high methylation at the FAT4 gene promoter was associated with poor prognosis for GC patients. The FAT4 promoter region Data represent mean ± SD from 3 independent experiments. Scale bar, 100 μm. * denotes P < 0.05 compared to control. www.impactjournals.com/oncotarget was also methylated to various degrees in different GC cell lines. S-azacytidine, a demethylating agent increased FAT4 expression in GC cell lines. This indicated that the decrease in FAT4 expression was partly due to promoter methylation as previously reported in breast cancer [9] and few stage I lung adenocarcinoma patients [23].
These data suggested that decrease in FAT4 levels was critical for GC progression. The reversal of FAT4 expression alone could not completely block the tumorigenic properties of the GC cells because of acquired genetic changes during their transformation. We demonstrated the critical role of FAT4 in GC progression by showing that FAT4 knockout BGC-823 cells showed increased tumor growth and progression, both in vitro and in vivo.
In conclusion, our study demonstrates that decreased FAT4 expression is associated with GC progression resulting in poor prognosis. Moreover, FAT4 promoter methylation decreases FAT4 expression. Thus, our study demonstrates that FAT4 is a potential prognostic marker and therapeutic target for GC patients.

GC patient samples
We performed tissue microarray and immunohistochemical analyses on FFPE samples obtained from 449 GC patients enrolled from January 1998 to January 2004 at the Zhejiang Provincial People's Hospital. The patients were followed up for over 5 years until December 2009. The survival time was estimated from the date of surgery to the end of the follow-up period or the date of death due to carcinoma recurrence. The patients did not receive radiotherapy and/or chemotherapy prior to surgery. We obtained written informed consent from all patients prior to analysis. The age of the GC patients ranged from 17 to 80 (median age: 59.3 y). The patients were classified according to the World Health Organization pathological classification (2010) of tumors. The clinicopathological characteristics of the GC patients are summarized in Table 1. The core tumor area (tumor occupying >50%) was determined by the pathologist based on H&E staining of the 449 FFPE GC tissue wax blocks. Subsequently, 35 (approximately 2 mm diameter) individual paraffin embedded GC blocks were arranged in a recipient paraffin microarray block (tissue array blocks) using a trephine. Finally, 13 tissue array blocks were generated from the 449 GC cases. Each block contained more than three internal controls consisting of normal gastric mucosa.
FAT4 loss of function was investigated in 173 frozen GC tissue samples. Methylation analysis was conducted on RNA and DNA samples isolated from the 449 samples. This cohort included 50 cases without lymph node and distant metastases, 59 cases with distant metastases (40 with liver metastasis and 19 with ovary metastasis) and 64 cases with lymph node metastasis alone. The clinicopathological characteristics of the GC patients are summarized in Table 3.

Immunohistochemical staining and evaluation
The GC patient tumor tissue microarray (TMA) and the whole sections of distant metastases tissues were used to analyze FAT4 by IHC. The sections were deparaffinized, rehydrated, and subjected to antigen retrieval combined with signal detection as described previously [24]. The sections were incubated with primary FAT4 antibody (ab198905, 1:100 dilution) overnight at 4°C. They were developed with 3, 3-diaminobenzidine (DAB) and counterstained with hematoxylin. Fat4 staining intensity was determined under a light microscope on a scale of 0-3. The proportion of positively stained cells (0, <5%; 1, 5-25%; 2, 26-50%; 3, 51-75%; 4, 76-100%) were independently estimated by two pathologists in the absence of clinical information as described previously [24,25]. The intensity and proportion scores were subsequently multiplied to obtain a composite score. A score of 0 to 3 was considered negative, whereas scores of 4 to 12 were considered positive.

RNA and DNA extraction
The E.Z.N.A. ® DNA/RNA Isolation Kit (R6731-01, Omega Bio-tek, GA, USA) was used to simultaneously isolate genomic DNA and total RNA from 90 GC tissues and cell line extracts. Briefly, the tissue samples were initially lysed and homogenized in a denaturing buffer and the supernatant was applied to a HiBind DNA spin column that binds the DNA. The RNA pellet was dissolved and purified by a HiBind RNA spin column. The purity and concentration of the samples were measured by a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA).

Quantitative real-time PCR (qRT-PCR)
FAT4 mRNA levels were analyzed by qRT-PCR. Total RNA from tissue samples and cultured cells was reverse transcribed to cDNA with PrimeScript™ RT Master Mix (Perfect Real Time) kit (RR036A, Takara) according to manufacturer's instructions. Real-time PCR was carried out by the SYBR Green master mix kit in a CFX96 thermal cycling instrument (Bio-Rad). GAPDH was used as endogenous control. The primer sequences were as follows: FAT4-F: 5'-TATCACAAAACGCCCTTGCT-3', FAT4-R: 5'-TGGATTGTCATTGATATCCTG-3'. GAPDH-F: 5'-TGAAGGTCGGAGTCAACGG-3' and GAPDH-R: 5'-CTGGAAGATGGTGATGGGATT-3'. The PCR cycling conditions were as follows: 95°C for 4 min followed by 40 cycles of 95°C for 10 s, 60°C for 30 s and 72°C for 30 s. The relative expression was determined by the 2 -∆∆Ct method with the Ct values obtained from the melting curves.

Methylation specific PCR and analysis
The methylation status of FAT4 gene was determined by methylation specific PCR (MSP) using bisulfite modified genomic DNA as the template. Bisulfite modified genomic DNA was prepared with the EZ DNA Methylation-Gold™ Kit (Zymo Research, D5006, CA, USA) according to the manufacturer's protocol. MSP was carried out in a 50 μl reaction mixture containing 5 μl of bisulfite treated DNA. The MSP assay conditions were as follows: 95°C for 4 min followed by 40 cycles of 95°C for 10 s, 62°C for 30 s and 72°C for 30 s. The reverse primer sequence for both methylated and unmethylated templates was 5'-CCTATATCTAAAATATATAAAAAATC-3'. The forward primers for methylated and unmethylated templates were 5'-GTTTTAGCGGTTATTGTCGGC-3' and 5'-GTTTTAGTGGTTATTGTTGGT-3', respectively. After completion of the PCR, the samples were analyzed on a 1.5% agarose gel and the methylated or unmethylated PCR products (~280 bp) were detected by the Bio-Rad ChemiDoc™ MP System.

Methylation pyrosequencing
We used pyrosequencing analysis to quantify the methylation levels in 36 pairs of GC and adjacent noncancerous tissues and 7 GC cell lines to verify the MSP results. For bisulfite conversion of the target sequences, EpitTect Bisulfite Kit (Cat No.: 59104, QIAGEN, Germany) was used according to the manufacturer's manual. Pyrosequencing was performed using a PyroMark Gold Q96 Reagents Kit (Cat No.: 972804, QIAGEN, Germany) in a PyroMark Q96 ID System (QIAGEN, Germany) according to the manufacturer's instructions. CpG site methylation was quantified by Pyro Q-CpG 1.0.9 software (QIAGEN, Germany).
In brief, double stranded PCR products were denatured with 0.1M sodium hydroxide (NaOH) followed by PBS washes. It was then annealed to the sequencing primer (5'-CTACAACCCCCCTCCA-3'). The pyrosequencing reaction was started at the 3'-end of the sequencing primer by adding A, T, C, and G nucleotides into each sample well, one at a time. As the bases were incorporated into the growing DNA strand based on complimentary base pairing, each incorporated base resulted in light production as a result of activation of an enzymatic cascade. The light intensity was graphed in a program and analyzed by the Pyromark CpG software to determine average methylation at all CpG sites. FAT4 knock-out by the CRISPR/Cas9 system To knockout FAT4, sgRNA sequence, 5'-GGGCAGGCTCTCGCGGTCGA-3' that targeted exon1 of FAT4 was designed with the Zhang Lab web application (http://crispr.mit.edu/). The FAT4 specific sgRNA was cloned into the pL6-U6-Bbsl_ Bbsl_gRNA scaffold plasmid (Addgene, Cambridge, USA), linearized by digestion with Bbsl (Thermo Scientific, ER1011, USA) for 30 min at 37°C and purified by gel electrophoresis. Subsequently, pair of oligonucleotide sequences was designed to target the FAT4 exon 1, namely, FAT4-KO-sgRNA-F: ACCGGGGCAGGCTCTCGCGGTCGA and FAT4-KO-sgRNA-R: AAACTCGACCGCGAGAGCCTGCCC. The two oligonucleotides were annealed, phosphorylated and ligated to the linearized pL6 vector. BGC-823 GC cells were transfected in a 6-well plate with 250 ng of pL6-U6-Bbsl_Bbsl_gRNA scaffold and/or an empty pL6-U6-Bbsl_Bbsl_gRNA scaffold plasmid and 500 ng of pXC-FLAG-hCas9 plasmid using polyethylenimine (PEI, Polyplus transfection ® SA, 115-010, USA). After 48 h, the cells were selected for 3 weeks with 1 μg/ ml puromycin. Then, single colonies were tested for authentic clones by PCR amplification of the CRISPR binding site using the following primers: forward: 5'-CCCGTTGCACACTCTATCAGTATCT-3' and reverse: 5'-TGTCTTGCACAGTCACGTTTACCT-3'. PCR products were analyzed by Sanger sequencing.

Real time cellular analysis (RTCA) of cellular growth, migration and invasion
The xCELLigence system (ACEA Biosciences Inc., Hangzhou, China) was used in order to monitor the cellular growth. The system consisted of microtiter plates (E-plates) with integrated gold microarrays at the bottom of the wells for continuous real time and label-free measurements of cellular status by electrical impedance. www.impactjournals.com/oncotarget Cell index (CI) was determined based on relative changes in impedance at different time points (T) in comparison to baseline (T 0 ). In standard E-plates, the CI values are proportional to the number of cells attached. The kinetic profiles reflect adhesion and spreading in approximately 6 h following cell seeding. In the present study, BGC-823-FAT4-KO, BGC-823-shFAT4 and the corresponding control cells were seeded in E-plates at a density of 5000 cells/well in 10% FBS supplemented RPMI-1640 medium. The E-plates were transferred to the RTCA-DP instrument for automated real-time monitoring at standard incubator conditions with quadruplet readouts of CI every 30 min for the following 3 days.
The rates of migration and invasion were further monitored by the xCELLigence-system in cell invasion or migration (CIM) plates as previously described [26]. CIM plates have microelectronic sensors located on the underside of a microporous membrane insert. The migrating cells adhere to the sensors resulting in higher impedance and CI. For the migration assays, BGC-823-FAT4-KO, BGC-823-shFAT4 and the corresponding control cells were seeded in the upper chamber of the CIM-plates in serum-free RPMI 1640 medium at a density of 40000 and 200000 cells per well for migration and invasion assays, respectively. The bottom chambers were filled with 30% FBS supplemented RPMI 1640 medium to promote migration across membranes by generating a serum gradient. For the invasion assays, the upper chambers were loaded with 30 μl of 1:20 diluted matrigel to generate a 3D biomatrix film. Following cell seeding, CIM-plates were transferred to the RTCA instrument for continuous readouts during the next 70 h. CI was registered solely by cells that migrated through the 8 μm porous membranes. The data were analyzed by the RTCA software 2.0.

MTT, wound healing and transwell assays
The RTCA data of cell growth, wound healing, migration and invasion were confirmed by the MTT, wound healing and the Transwell assays as previously described [25].

Nude mice xenografts
Four to five week old female BALB/c athymic nude mice were purchased from Slac Laboratory Animal Co. Ltd. (Shanghai, China). They were housed in pathogenfree conditions according to our institutional guidelines. The study protocol was approved by the Animal Care committee of our institution.
The mice (n=5) were simultaneously injected with BGC-823-FAT4-KO and BGC-823-KO-NC cells (2 X 10 6 cells/ 100 μl, in the left and right flanks, respectively. Tumor size was measured every 7 days by a slide caliper and tumor volume was calculated as (length×width 2 )/2.
To evaluate the effect of FAT4 on GC metastasis, 1×10 6 BGC-823-FAT4-KO and BGC-823-KO-NC cells were injected into the lateral tail vein of the mice. Metastatic lung nodules were scanned by whole-body micro-PET/CT and quantified by H&E staining using a dissecting microscope at each endpoint. The mice were sacrificed after 30 days and subcutaneous tumor or metastatic lung tissues were rapidly collected. A portion of the tissue was stored in liquid nitrogen, whereas the remaining was fixed in 10% neutral buffered formalin and embedded in paraffin. The tissues were cut into 4 μm thick sections and stained with H&E. Some sections were subjected to IHC staining with antibodies against FAT4(NBP1-78381F, rabbit polyclonal anti-FAT4 antibody, 1: 150, Novus Biologicals, Littleton, CO, USA) and Ki-67 (#GA62661, Clone MIB-1, Ready to use, DAKO, USA). The Ki-67 proliferation index was determined by a senior pathologist by estimating the percentage of positive nuclei in 1,000 cells in hot spots.

Whole-body 18 F-FDG PET/CT
Nude mice with GC xenografts were injected with approximately 0.2 mCi of 18 F-FDG for 1 h under continuous anesthesia. The mice were placed at the centre of the Siemens Inveon combined microPET-CT scanner (Siemens Preclinical Solution USA, Inc., Knoxville, TN, USA) in a prone position. MicroCT scans were conducted under the following conditions: X-ray tube voltage: 80 kV; current intensity: 500 μA; exposure time: 150 ms and 120 rotation steps. PET static acquisition was carried out for 10 min and the images were reconstructed using the OSEM (ordered set expectation maximization) algorithm for 3D PET reconstruction. The images were analyzed by the Inveon Research Workplace 4.1 (Siemens, Erlangen, Germany). The standardized uptake (SUV g/ml) was determined as SUV = [(RTA/cm 3 )/RID] × BW (RTA is the measured www.impactjournals.com/oncotarget radiotracer tissue activity in mCi; RID is the radiotracer injected dose in mCi and BW is the mouse body weight in grams).

Statistical analysis
All statistical analyses were performed with the SPSS 13.0 statistical software. Differences between groups were estimated by 2-tailed paired Student's t test or independent sample t-test. The relationship between FAT4 expression and clinicopathological characteristics were analyzed by the Chi-square test. Kaplan-Meier survival curves were analyzed by the logrank test. The significance of various survival related variables were assessed by Cox regression model in the multivariate analysis. P<0.05 was considered statistically significant.