FZD5 prevents epithelial-mesenchymal transition in gastric cancer

Background Frizzled (FZD) proteins function as receptors for WNT ligands. Members in FZD family including FZD2, FZD4, FZD7, FZD8 and FZD10 have been demonstrated to mediate cancer cell epithelial-mesenchymal transition (EMT). Methods CCLE and TCGA databases were interrogated to reveal the association of FZD5 with EMT. EMT was analyzed by investigating the alterations in CDH1 (E-cadherin), VIM (Vimentin) and ZEB1 expression, cell migration and cell morphology. Transcriptional modulation was determined by ChIP in combination with Real-time PCR. Survival was analyzed by Kaplan–Meier method. Results In contrast to other FZDs, FZD5 was identified to prevent EMT in gastric cancer. FZD5 maintains epithelial-like phenotype and is negatively modulated by transcription factors SNAI2 and TEAD1. Epithelial-specific factor ELF3 is a downstream effecter, and protein kinase C (PKC) links FZD5 to ELF3. ELF3 represses ZEB1 expression, further guarding against EMT. Moreover, FZD5 signaling requires its co-receptor LRP5 and WNT7B is a putative ligand for FZD5. FZD5 and ELF3 are associated with longer survival, whereas SNAI2 and TEAD1 are associated with shorter survival. Conclusions Taken together, FZD5-ELF3 signaling blocks EMT, and plays a potential tumor-suppressing role in gastric cancer. Video Abstract


Background
Frizzled (FZD) proteins are G protein-coupled receptors (GPCRs). The extracellular N-terminus contains a cysteine-rich domain (CRD) through which FZDs bind WNT ligands. The intracellular C-terminus binds the PDZ domain of Dishevelled (Dvl) and interacts with G proteins. Up to now, 10 FZDs have been identified in human. These receptors mediate β-catenin pathway and various β-catenin-independent pathways depending on cellular context. WNT/β-catenin pathway is oncogenic and involved in almost every aspect of tumor development. In addition, β-catenin-independent WNT/PCP and WNT/Ca 2+ -PKC pathways are also implicated in tumor progression [1][2][3][4]. In recent years, some novel pathways such as WNT/Stat3 and WNT/ Yap-Taz have been successively identified [5][6][7][8], indicating that FZD signalings are far more complicated and still incompletely comprehended.
Epithelial-mesenchymal transition (EMT) in tumor cells endows them with enhanced motility thereby increasing their metastatic potentiality. This process is characterized by reduced expression of epithelialrelated factors and increased expression of mesenchymal-related factors. Mechanistically, EMT is driven by some transcription factors such as SNAI1/2 and ZEB1/2. Studies have shown that FZD signalings induce tumor cell EMT and metastasis. FZD4, FZD7 and FZD10 promote EMT through β-catenin pathway in prostate, liver and breast cancer [9][10][11]. FZD2 mediates WNT5-induced EMT dependent on Stat3 signaling in liver, lung, breast and colon cancer [5], and FZD8 mediates WNT11-induced EMT by interacting with TGF-β signaling in prostate cancer [12].
In gastric cancer, FZD4 and FZD7 similarly promote EMT and metastasis by activating β-catenin pathway [13,14]. Genetic deletion of FZD7 was shown to inhibit the growth of gastric adenomas in vivo [15]. FZD2 and FZD8, through β-catenin pathway but not β-cateninindependent pathways, induce gastric cancer cell proliferation, migration, invasion and metastasis [16,17]. Unexpectedly our study uncovered that FZD5 prevents EMT and is associated with longer survival in gastric cancer, suggesting FZD5 as a putative suppressor in this type of cancer.

In silico analysis
Cancer Cell Line Encyclopedia (CCLE) database was interrogated for gene expression in 37 human gastric cancer cell lines. Correlation between two genes was analyzed by Pearson statistics. The Cancer Genome Atlas (TCGA) database was interrogated for GO pathway enrichment analysis in 375 gastric cancer tissues and 32 normal gastric tissues. GSE62254 database was interrogated in Kaplan Meier plotter website (http://kmplo t.com/analy sis/) to assess the survival [18,19].

Cell culture and transfection
Human gastric cancer cell line MKN45 and HGC27 were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum in a humidified incubator at 37 °C with 5% CO 2 . 5 × 10 4 MKN45 cells were transfected with shRNA lentiviruses to stably knockdown FZD5 expression; 5 × 10 4 HGC27 cells were transfected with overexpression lentiviruses to stably overexpress FZD5, and the transfected cells were further selected using 2 μg/ ml puromycin. For transient transfection, 2.5 × 10 5 cells were transfected with shRNA or overexpression plasmids using Lipofectamine 3000 in Opti-MEM medium according to the product manual. Target sequences for shRNA were listed in Table 1.

Real-time PCR
RNAiso Plus (Takara) was used to extract total RNAs, which were then reversely transcribed into cDNA using PrimeScript ™ RT reagent Kit with gDNA Eraser (TaKaRa)according to the instructions. Real-time PCR was performed using TB Green ™ Premix Ex Taq ™ II (TaKaRa). Primers for genes and transcription factor binding sequences were listed in Table 2. GAPDH was used as endogenous control. Expression difference was analyzed using 2 −ΔΔCT method.

Phalloidin staining
Cells on the slides were fixed with 4% formaldehyde at room temperature for 10 min. Then the slides were

Cell migration
Transwell chambers were used to determine cell migration ability. 2 × 10 4 cells were added into the upper chamber. Culture medium with 10% FBS was added to the lower chamber. After 24 h, non-migrated cells were removed with a swab. The chambers were fixed with 4% paraformaldehyde for 15 min at room temperature, and incubated with 0.1% crystal violet for 30 min at 37℃. Cells were determined in 5 randomly selected microscope fields.

Chromatin immunoprecipitation (ChIP)
A ChIP assay kit (Beyotime, China) was used according to the manufacturer's instructions. Briefly, Cells in 10 cm plates were fixed with 1% formaldehyde for 10 min at 37 °C. To shear the chromatin, cells were treated with 1 mM PMSF in SDS Lysis Buffer for 10 min at 4 °C, followed by cell sonication for 15 min at 4 °C. After a portion of the cross-linked chromatin was removed as input for the subsequent test, the remaining cell lysis was incubated with 1 μg primary antibody at 4 °C overnight. Then protein A + G agarose was added to precipitate the target protein recognized by the primary antibody for 1 h at 4 °C. IgG antibody was used as a negative control. The beads were then washed off and DNA was collected for subsequent Real-time PCR assays. The enrichment was indicated as % of input.
Then the agarose beads were washed to clear non-specifically bound proteins. The target protein was detected by Western blot.

Human specimens
55 gastric cancer tissues were obtained from Shengjing Hospital China Medical University with the informed consent of the patients. Institutional Research Ethics Committee of China Medical University approved the use of these tissues for research purposes.

Statistical analysis
GraphPad Prism 7.0 was used for statistical analysis. Data were expressed as mean ± SD. Student's t test or one way ANOVA was used to compare data among groups. Correlation analysis was performed using Pearson statistics. Survival was analyzed by Kaplan-Meier method. P value less that 0.05 was considered as significant.

FZD5 maintains epithelial-like phenotype
Interrogation of CCLE database revealed that FZD5 is positively correlated with epithelial-related factors CDH1 (E-cadherin), OCLN (Occludin), EPCAM and ELF3, while negatively correlated with mesenchymal-related  (Fig. 1a). Pathway enrichment analysis of FZD5 using TCGA database showed that FZD5 is negatively associated with EMT in 375 gastric cancer tissues and 32 normal gastric tissues. (Fig. 1b). FZD5 knockdown in MKN45 cells downregulated CDH1 but upregulated VIM (Fig. 1c- knockdown in MKN45 cells induced a morphological alteration from epithelial-like to mesenchymal-like and an increase in cell migration (Fig. 1f-h). In contrast, FZD5 overexpression in HGC27 cells increased CDH1 but reduced VIM (Fig. 2a, b). Furthermore, FZD5 overexpression caused a morphological alteration from mesenchymal-like to epithelial-like and a decrease in cell migration (Fig. 2c, d). These results indicate that FZD5 maintains epithelial-like phenotype.

SNAI2 transcriptionally inhibits FZD5
As FZD5 was shown to be inversely correlated with EMT transcription factor SNAI2 in mRNA level (Fig. 1a), whether SNAI2 modulates FZD5 transcription was investigated. Two binding sites (BS1 and BS2) of SNAI2 in FZD5 promoter were identified (Fig. 2e). ChIP in combination with Real-time PCR verified the binding of SNAI2 to these two sites in HGC27 cells and SNAI2-overexpressed MKN45 cells (Fig. 2f ). SNAI2 overexpression in MKN45 cells reduced FZD5 expression in both mRNA and protein levels (Fig. 2g). Furthermore, SNAI2 knockdown in HGC27 cells induced FZD5 expression (Fig. 2h). These findings confirmed that SNAI2 negatively modulates FZD5.

TEAD1 coordinates with SNAI2 in FZD5 transcription inhibition
YAP/TEAD was shown to interact with SNAI2 in modulating gene transcription [21,22]. Therefore, the role of TEAD1 in FZD5 transcription was subsequently evaluated. Two binding sites of TEAD1 in FZD5 promoter were identified, and the BS2 of TEAD1 is near the binding sites of SNAI2 (Fig. 3a).

ChIP in combination with
Real-time PCR verified the binding of TEAD1 to these two sites in HGC27 cells and MKN45 cells (Fig. 3b). TEAD1 knockdown in HGC27 cells upregulated FZD5 expression in both mRNA and protein levels (Fig. 3c). CoIP confirmed the interaction between SNAI2 and TEAD1 in HGC27 cells (Fig. 3d). TEAD1 knockdown in MKN45 cells had no effect on FZD5 expression (Fig. 3e), probably due to the absence of SNAI2 (Fig. 2g). However, TEAD1 knockdown in MKN45 cells blocked SNAI2 overexpression-induced FZD5 inhibition (Fig. 3f ), demonstrating that TEAD1 coordinates with SNAI2 to modulate FZD5 transcription.

LRP5 is required for FZD5 signaling
LRP5 is a co-receptor of FZDs. Interestingly, interrogation of CCLE database revealed that LRP5 is positively correlated with FZD5, CDH1, EPCAM and ELF3, while negatively correlated with VIM, SNAI2, ZEB1 and ZEB2 (Fig. 5a). FZD5 knockdown in MKN45 cells reduced LRP5 expression, indicating LRP5 is modulated by FZD5 (Fig. 5b). The role of LRP5 in FZD5 signaling and EMT was subsequently investigated. LRP5 knockdown in MKN45 cells repressed ELF3 expression but promoted ZEB1 transcription and cell migration (Fig. 5c-e). LRP5 knockdown in MKN45 cells also induced a morphological alteration from epithelial-like to mesenchymal-like (Fig. 5f ). Furthermore, LRP5 knockdown in HGC27 cells blocked the effects of FZD5 overexpression on ELF3 expression, ZEB1 transcription and cell migration ( Fig. 5g-i). These findings demonstrate that FZD5 signaling in modulating EMT requires its co-receptor LRP5.

WNT7B is a putative ligand for FZD5
Just like FZD5 knockdown, WNT7B knockdown in MKN45 cells inhibited CDH1 expression but induced VIM expression (Fig. 6a, b). Moreover, WNT7B knockdown in MKN45 cells reduced ELF3 expression but increased ZEB1 transcription (Fig. 6c, d). Similarly, WNT7B knockdown in MKN45 cells induced a morphological alteration from epithelial-like to mesenchymallike and an increase in cell migration (Fig. 6e, f ).

FZD5 signaling is related to patient survival
FZD5, ELF3, TEAD1 and SNAI2 expression in 55 gastric cancer tissues was detected by immunohistochemistry. Expression scoring showed that FZD5 is positively correlated with ELF3 while negatively correlated with TEAD1 and SNAI2 (Fig. 7a). EMT potentially promotes tumor metastasis, leading to unfavorable clinical outcomes.
Accordingly, the association of these factors with survival was analyzed. High FZD5 and ELF3 protein expression is correlated with longer overall survival (OS); in contrast, high TEAD1 and SNAI2 protein expression is correlated with shorter survival (Fig. 7b). GSE62254 database was interrogated to assess the association of these factors with survival in mRNA levels [18,19]. Consistently, high FZD5 and ELF3 mRNA expression is related to longer survival, while high TEAD1 and SNAI2 mRNA expression is related to shorter survival (Fig. 7c, d). These findings suggest that FZD5-ELF3 signaling contributes to good clinical outcome in gastric cancer.

Discussion
Among 19 WNT ligands in human, WNT7A/B and WNT5A have been reported to bind to FZD5. WNT7A-FZD5 signaling promotes endometrial and ovarian cancer cell proliferation and growth through β-catenin pathway [27,28]. Similarly, WNT7B-FZD5 signaling promotes RNF43-mutant pancreatic cancer cell proliferation and growth [29]. Upon binding WNT5A, FZD5 mediates β-catenin-independent pathways. WNT5A-FZD5 signaling increases cell motility in melanoma and classical Hodgkin lymphoma [3,30]. Together, these findings demonstrate FZD5 as a promoter in some human tumors. However, our study revealed that FZD5 is a putative suppressor in gastric cancer. Interrogation of databases showed an association of FZD5 with epithelial-like phenotype. Loss and gain of function studies further confirmed that FZD5 prevents cancer cell EMT. Prognosis analysis indicated that high FZD5 expression is correlated with longer survival period. Furthermore, our study showed that WNT7B is a putative ligand for FZD5 in gastric cancer. WNT7A/B seems to specifically bind to FZD5 among 10 FZDs [31]. Actually, WNT7A functions as an EMT inhibitor in lung cancer [32]. Moreover, interrogation of CCLE database showed that FZD5 is also associated with epithelial-like phenotype in lung cancer (data not shown).
As shown in our study, FZD5 is transcriptional inhibited by TEAD1-SNAI2 complex. Activation of tumorsuppressing Hippo pathway results in phosphorylation and degradation of Yap and Taz. When Yap and Taz are not phosphorylated, they translocate into the nucleus to bind transcription factor TEAD. Yap/Taz-TEAD signaling induces gastric cancer cell EMT, stemness, drug resistance and metastasis [33][34][35][36][37]. Intriguingly, Yap/Taz Fig. 7 FZD5 signaling is related to patient survival. a FZD5, ELF3, TEAD1 and SNAI2 expression in 55 GC tissues was detected by Immunohistochemistry. Protein expression was scored by H-score system. Staining of a representative sample was shown. Scalebar: 20 μm. Correlation of FZD5 with ELF3, SNAI2 and TEAD1 was analyzed by Pearson statistics. b The association of FZD5, ELF3, SNAI2 and TEAD1 with OS in 55 GC tissues was analyzed by Kaplan-Meier method. c The association of FZD5, ELF3, SNAI2 and TEAD1 with OS was analyzed using GSE62254 database in Kaplan Meier plotter website. d The association of FZD5, ELF3, SNAI2 and TEAD1 with PPS was analyzed using GSE62254 database in Kaplan Meier plotter website is modulated by WNT5A [7,8,38,39], suggesting that FZD5 signaling may be suppressed by WNT5A in gastric cancer. In addition, TEAD1 and SNAI2 coordinate to modulate gene transcription in both physiological and pathological processes [21,22,40].
ELF3 is a universal epithelial-limited transcription factor. Similar to FZD5, ELF3 has dual roles in human tumor. ELF3 induces breast epithelial cell malignant transformation [41,42], and promotes prostate cancer progression by interacting with NFκB [43]. In gastric cancer, ELF3 maintains epithelial-like phenotype, prevents EMT and is associated with longer survival, acting as a putative tumor suppressor. ELF3 also inhibits EMT in ovarian and breast cancer cells [44,45]. Mechanistically, ELF3 represses activity and expression of EMT transcription factor ZEB1 [46,47].
Activation of β-catenin pathway requires not only FZDs, but also their co-receptors LRP5/6 which is not involved in non-canonical β-catenin-independent pathway. A recent study reported that FZDs-LRP5/6 interaction can initiate β-catenin pathway in the absence of WNTs [48]. Surprisingly, another study revealed that FZDs-LRP5/6 interaction can block activation of FZDmediated non-canonical pathway, further preventing tumor metastasis [49]. These findings, in combination with our observations, suggest that FZDs-LRP5/6 interaction may act as a tumor promoter by activating β-catenin pathway or a tumor suppressor by antagonizing β-catenin-independent pathway, depending on tumor or cell context.

Conclusions
In summary, our study for the first time uncovers that FZD5-ELF3 signaling prevents EMT and is associated with favorable prognosis in gastric cancer. Therefore, FZD5 and ELF3 function as putative tumor suppressors in this type of cancer. Antibodies against FZD5 including ipafricept and vantictumab have been shown effective on certain cancer types [50]. However, our study warns out that targeting FZD5 should be cautious and tumor-specific, since treatment with these antibodies may potentially promote metastasis in some other types of cancers including gastric cancer.