Demethylation of the SFRP4 Promoter Drives Gastric Cancer Progression via the Wnt Pathway

Although the incidence and mortality rate have steadily decreased, gastric cancer remains the sixth most common malignant neoplasm and fourth leading cause of cancer-related death worldwide (1). More importantly, due to the mild and atypical symptoms at the early stage, over 80% of patients are clinically diagnosed at an advanced stage, which generally indicated a dismal prognosis (2). Thus, understanding key molecules in the carcinogenesis and progression of gastric cancer and targeting the underlying mechanisms may provide a novel targeted therapy for gastric cancer.

Secreted frizzled-related protein 4 (SFRP4) belongs to the family of secreted frizzled-related proteins (SFRP; ref. 3). SFRPs share homology with frizzled receptors and are believed to impede Wnt–Fz interactions, thereby blocking Wnt activity in cells (4). The Wnt pathway is involved in cell proliferation, differentiation, apoptosis, migration, and stem cell self-renewal (5). However, aberrant Wnt signaling has been identified as an important contributor to tumor development (6, 7). Considering the inhibitory effect on Wnt signaling, SFRPs tend to play a role as tumor suppressors in previous research (8, 9). Nevertheless, unlike other SFRP family members, recent studies have shown conflicting evidence for the expression level and function of SFRP4 in different cancers (10). SFRP4 expression was indicated to be either silenced or reduced in the majority of tumors (11–13). However, in colorectal cancer, SFRP4 was reported to be overexpressed and have less frequent promoter hypermethylation in tumors than normal mucosa, suggesting that SFRP4 may serve as an oncogene (14–16). In gastric cancer, the prognostic value of SFRP4 and its precise role in modulating tumor progression remain unknown and need to be further explored.

In this study, we aim to investigate the expression and oncogenic role of SFRP4 in gastric cancer and explore the effects of SFRP4 on Wnt signaling pathway.

Raw data of GSE27342 was downloaded from the Gene Expression Omnibus (GEO) dataset: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE27342. R software (version 3.6.3) was used to analyze the data. The data was processed with rma algorithm. Low-expressed probeset was filtered, and differential expression analysis was performed using limma (trend algorithm). P value was adjusted by Benjamini–Hochberg procedure. Hugo gene symbol is used in gene annotation.

The study enrolled 240 patients with gastric cancer who had received a radical (R0) resection with a D2 lymphadenectomy from Zhongshan Hospital, Fudan University between 2004 and 2008. None of them received any preoperative therapy. We retrospectively collected the demographic characteristics and the clinicopathologic parameters of each patient, including age, gender, tumor size, location, tumor differentiation, Lauren classification, vessel invasion, and tumor–node–metastasis (TNM) stage. Another independent group of 50 paired frozen gastric cancer and matched normal mucosa tissues was also collected from Zhongshan Hospital in the year 2018. Normal mucosa tissues were identified as tissues that were more than 60 mm away from the primary lesions. The study was approved by the ethics committee of Zhongshan Hospital and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Informed consent was obtained from each patient.

Tissue microarrays applied by formalin-fixed paraffin embedded (FFPE) surgical specimens were constructed as previously described (17). IHC staining was performed according to the protocol supplied by the Thermo Scientific. Primary antibody (Anti-SFRP4 antibody: 1:150 dilution, ab, Abcam) were used for IHC staining. Evaluation of SFRP4 staining depended on both the area and the intensity of immunopositive cells. The area was categorized as follows: 0, <5%; 1, 5%–25%; 2, 26%–50%; 3, 51%–75%; 4, >75%. The intensity was categorized as follows: 0, negative; 1, weak; 2, moderate; 3, strong. The score was calculated by multiplying staining area and staining intensity, yielding a series of results ranging from 0 to 12. According to the area under the receiver operating characteristic (ROC) curves (AUC), the value on the curve closest to the point (0, 1) which maximized both sensitivity and specificity for the survival was the cut-off score (18). As shown in Supplementary Table S1, the score higher than 4 were considered as high expression. The stained sections were assessed by two gastroenterology pathologists who were blind to patients' clinical data.

The human gastric cancer cell lines MGC80–3 and AGS were purchased from Cell Bank of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). All cell lines obtained from the cell bank were tested for authentication using short tandem repeat fingerprinting and passaged for fewer than 6 months. Cells were cultured in DMEM or RPMI1640 supplemented with 10% FBS (catalog no. 16000–044; Gibco) at 37°C in a humidified atmosphere containing 5% CO₂. Mycoplasma was tested and was negative in all cells using the qPCR method (Thermo Fisher Scientific). The primary antibodies including GSK-3β (#12456), β-catenin (#8480), LEF-1 (#2230), c-Myc (#5605), c-Jun (#9165), CD44 (#3570), and MMP-7 (#3801) were purchased from Cell Signaling Technology; SFRP4(ab154167) was purchased from Abcam (Abcam); Wnt2 (abs137544), Wnt3a (abs123674), Wnt4 (abs138560), Wnt5a (abs137051) were purchased from Absin; Wnt5a(NBP2–75714) antibody for Flow cytometry was purchased from Novus Biologicals. The recombinant human SFRP1 (1384-SF), SFRP4 (1827-SF) and Wnt5a(645-WN) protein were purchased from R&D Systems.

The siRNA (siSFRP4–1: AAGTCCCGCTCATTACAAATT; siSFRP4–2: GUCCCGCUCAUUACAAAUUTT; siSFRP4–3: AAGUCCCGCUCAUUACAAAUU), and negative control siRNA (UUCUCCGAACGUGUCACGUTT) were constructed by Biotend Research. The shSFRP4 plasmid, SFRP4 overexpression plasmid, and negative control plasmid were constructed by YouBio. Transfections were performed with Lipofectamine 3000 (Life Technologies) according to the manufacturer's instructions.

For the immunoprecipitation (IP) assay, cells were solubilized with IP buffer (50 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 0.1% NP-40, 15 mmol/L MgCl₂, and 5 mmol/L EDTA), and incubated with specific antibody immobilized onto Protein G-Sepharose for 2 hours at 4°C with gentle rotation. Beads were then washed four times with IP lysis buffer, boiled, and centrifuged (19). For Western blotting (WB) assay, proteins extracted from samples were separated by SDS–PAGE and transferred onto polyvinylidene difluoride membranes. Then membranes were incubated with specific antibodies and followed by secondary antibody.

Total RNA of samples was isolated by TRIzol Reagent (15596026, Thermo Fisher). The cDNA was generated by the Takara RNA PCR Kit. Real-time PCR was performed using SYBR Premix Ex Taq (Takara) agent with StepOne Plus system (Life Technologies). The primers were listed in Supplementary Table S2.

Cell proliferation was measured by using the Cell Counting Kit-8 (CCK8, Dojindo). One-hundred microliters of cell suspension was dispensed in 96-well plates (Corning) at a density of 5,000 cells/well. After the cells were cultured for 48 hours, 10 μL of CCK8 reagent mixed with 90 μL DMEM without FBS were added to each well and incubated for indicated times. The absorbance of optical density (OD) at 450 nm was read in an 800-TS microplate reader (Bio-Tek Instruments).

Cell-cycle alterations and the occurrence of apoptosis were assessed using flow cytometry. Cells were collected by trypsin and washed twice with PBS. For cell-cycle assay, the collected cells were stained with DNA staining solution and propidium iodide (PI) for 30 minutes by using Cell Cycle Staining Kit (Lianke Bio). Cellular apoptosis was determined with FITC Annexin V and PI using FITC Annexin V Apoptosis Detection Kit I (BD Biosciences).

To detect the Wnt5a binding to the receptors, suspended MGC80–3 cells in all groups were stained with primary Wnt5a antibody followed by Alexa Fluor 488–labeled secondary antibody. Stained cells were assessed by flow cytometer and data were analyzed by FlowJo software (TreeStar).

Cells were seeded on 12-well culture plates at a density of 300 cells/well in triplicate and cultured for 7 consecutive days at 37°C in an atmosphere with 5% CO₂. The naturally formed colonies (>50 cells per colony) were fixed with acetic acid–methanol and stained with 1% crystal violet (Sigma-Aldrich), followed by manual counting under an inverted microscope (Nikon).

Genomic DNA of tumor tissues from 5 patients with gastric cancer were isolated with QIAamp DNA Mini Kit (QIAGEN), and bisulfite modification of the genomic DNA was carried out using an EpiTect Bisulfite Kit (QIAGEN) according to the manufacturer's instructions (20). Quantitative methylation analysis of the promoter of SFRP1, SFRP2, SFRP4, and SFRP5 coding gene was performed by the Sequenom MassARRAY platform (Ouyi) according to the manufacturer's protocol. The primer of SFRP1, SFRP2, SFRP4, and SFRP5 were listed in Supplementary Table S2. The PCR conditions for all primer sets were as follows: hot start of 94°C for 4 minutes, 94°C for 20 seconds, 56°C for 30 seconds, and 72°C for 1 minute for 45 cycles, and a final extension at 72 1C for 3 minutes.

Study protocols involving mice were approved by the Animal Ethics Committee of Zhongshan Hospital and were performed according to the Guidelines for Animal Health and Use. All sections of this report adhere to the ARRIVE Guidelines. Immunocompromised nude mice were purchased from Shanghai SLAC Laboratory Animal Company (Shanghai, China) and were housed in individually ventilated cages with free access to water and food in a specific pathogen-free room. 1 × 10⁷ MGC80–3 cells stably transfected with the indicated plasmids, were suspended in 100 μL PBS and then injected subcutaneously into the flanks of 6-week–old male nude mice. In a rescue experiment, starting 1 week later, IWR-1 was dissolved in DMSO and was administered i.p. for the SFRP4-OE + IWR-1 group (5 mg/kg, twice per week, for 3 weeks). Tumor size was monitored using Vernier caliper every week. At 4 weeks postinjection, the nude mice were sacrificed by cervical dislocation after carbon dioxide inhalation, and tumors were collected and weighted.

Pearson test or Fisher exact test was used for categorical variables. Student t test was used for continuous variables, and Wilcoxon test was used for nonparametric variables. One way ANOVA test was used for analysis of three or more groups. Kaplan–Meier analysis was used to determine the survival rate, and the log-rank test was used to compare two survival rate curves. Cox's proportional hazards regression mode was performed for multivariate analysis of independent prognostic factors. ROC curve was used to judge the predictive accuracy of the prognostic models. Delong. Delong. Clarke–Pearson test was used to compare each two ROC curves. All analysis was conducted by SPSS program (Version 22). All statistical analyses were two-sided, and P < 0.05 was considered statistically significant.

Tissue samples and clinicopathologic data were obtained from Zhongshan Hospital, Fudan University. Written informed consent from each patient was achieved. The research was approved by the Research Ethics Committee of Fudan University and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

The results shown here are in whole or part based upon data generated by the TCGA Research Network (https://www.cancer.gov/tcga), GEO dataset (GSE27342), and Oncomine database (www.oncomine.com). All the data supporting the findings of this study are available from the corresponding authors upon reasonable request.

To examine the differential gene expression between different TNM stages of gastric cancer and investigate the main tumor-interrelated genes, genome analysis of 80 gastric cancer tissue samples was performed using online data from GEO dataset (GSE27342). The genome profiles of different TNM stages of gastric cancer tissues were shown in Supplementary Fig. S1. Then through data normalization, quantitation accuracy evaluation, and differentially expressed protein selection, the 20 most upregulated genes in advanced gastric cancer (TNM stage III–IV) were identified (Supplementary Table S3). Among them, the Wnt signaling pathway regulator SFRP4 ranked fourth, and was further studied.

Next, we performed searches and a meta-analysis of SFRP4 expression in microarray datasets from the Oncomine database (www.oncomine.com) and found that SFRP4 was most highly expressed in colorectal cancer, gastric cancer, brain and central nervous system cancer, and breast cancer (Fig. 1A). In gastric cancer, SFRP4 was confirmed to be upregulated in the Stanford, GEO and The Cancer Genome Atlas (TCGA) datasets (Fig. 1B). In addition, we also compared the expression of other SFRP family members in gastric cancer and found that SFRP1 and SFRP5 were expressed at low levels in the most of cancer types, including gastric cancer, and SFRP2 was expressed at slightly higher level in gastric cancer (Supplementary Fig. S2).

To verify the microarray analysis results, qPCR and immunoblotting were performed on 50 pairs of gastric cancer specimens and adjacent normal mucosa tissues. As shown in Fig. 1C and E, both the mRNA and protein level of SFRP4 were significantly increased in cancer compared with normal tissues. Next we performed IHC staining of SFRP4 on a series of TMAs including 240 gastric cancer samples. The result demonstrated an apparently increased intensity of SFRP4 staining in gastric cancer tumors compared with peritumor tissues (Fig. 2A and B).

According to the cut-off value calculated by the ROC curve, 137 (57.1%) and 103 (42.9%) cases were identified as low and high expression. The associations between SFRP4 expression and clinicopathologic characteristics in patients with gastric cancer were showed in Table 1. Among all variables, tumor size (P = 0.001), late T- stage (P < 0.001), lymph node metastasis (P < 0.001), and advanced TNM stage (P < 0.001) were significantly associated with high SFRP4 expression. The percentage of high SFRP4 expression also displayed an increase with the progression of T stage, N stage, as well as TNM stage (Fig. 2C–E; and Supplementary Fig. S3).

The correlation between SFRP4 expression and survival of patients was then evaluated. High expression of SFRP4 in tumor tissues indicated an unfavorable prognosis for patients with gastric cancer (P < 0.001; Fig. 2F), both in early-stage and advanced stage groups (Fig. 2G and H; Supplementary Fig. S4). The prognostic significance of SFRP4 expression in gastric cancer was confirmed by TCGA and GEO dataset and online survival analysis software (http://www.kmplot.com/analysis/index.php?p=service&cancer=gastric). The results also demonstrated that high expression of SFRP4 was significantly associated with poor overall survival (OS) and disease-free survival (DFS) in patients with gastric cancer (Supplementary Fig. S5).

Multivariate analysis also indicated that intratumoral expression of SFRP4 was an independent predictive factor for overall survival of patients with gastric cancer (Supplementary Fig. S6A). In addition, combining SFRP4 expression with TNM stage displayed a superior prognostic validity compared to SFRP4 expression or TNM stage alone (Supplementary Fig. S6B).

Because high SFRP4 expression suggested positive association with tumor progression in clinical samples, we further explored the biological function of SFRP4 in gastric cancer cells. On the basis of Western blotting result, the level of SFRP4 in majority of gastric cancer cells was higher than that in the normal gastric cell line GES-1 (Fig. 3A). MGC80–3 and AGS cells with high SFRP4 expression were selected for in vitro assays. Our result showed that after depletion of SFRP4 expression in MGC80–3 and AGS cells (Fig. 3B), cell viability was remarkably suppressed by CCK-8 assay (Fig. 3C). Colony formation assays also indicated that the number of clones was significantly reduced in SFPR4-silenced gastric cancer cells (Fig. 3D). Subsequent flow cytometry analysis revealed that knockdown of SFRP4 could induce apoptosis and G₁ phase arrest in gastric cancer cells (Fig. 3E and F).

As Wnt signaling plays a pivotal role in tumor progression, we first performed a meta-analysis of the expression of Wnt pathway-related genes in 12 gastric cancer microarray datasets from the Oncomine database and found that SFRP4 was the most upregulated gene (Fig. 4A). Then, we detected the expression of Wnt and its downstream genes in SFRP4 knockdown gastric cancer cells, and the result showed a significant reduction in both the mRNA and protein levels of Wnt pathway components compared with cell cultures without siRNA treatment (Fig. 4B–D). In addition, the nuclear localization of β-catenin was also decreased in SFRP4-compromised gastric cancer cells, indicating that SFRP4 could modulate the translocation of β-catenin from the cytosol to the nuclear (Fig. 4E). Together, these data suggested that SFRP4 overexpression might promote the activation of Wnt pathway in gastric cancer.

Next, we examined the expression of Wnt ligands in different gastric cancer cells and found that Wnt5A was highly expressed in AGS and MGC80–3 cells (Fig. 5A). Then the flow cytometry result showed that recombinant SFRP4 could remarkably enhance the binding ability of Wnt5A to its receptors (Fig. 5B). The IP assay also indicated that the interaction of SFRP4 with β-catenin was increased in the presence of Wnt5A (Fig. 5C). Taken together, these results indicated that SFRP4 might play a both extracellular and intracellular function in activating Wnt pathway signaling, which was consistent to previous report (21).

To further explore the biological function of SFRP4 in gastric cancer, we examined the expression of Wnt pathway markers after treating cells with recombinant SFRP1 and SFRP4 proteins. The result showed that SFRP1could remarkably inhibit the activation of the Wnt pathway in MGC80–3 cells; however, the suppressed activation mediated by SFRP1 could be rescued by recombinant SFRP4 treatment (Fig. 5D). In addition, SFRP4 competitively antagonized the inhibitory effect of SFRP1 on Wnt pathway activation in a dose-dependent manner (Fig. 5E). Next, we tested the cell proliferation under the treatment of recombinant SFRP1 and SFRP4. Our result showed that reintroduction of recombinant SFRP4 in gastric cancer cells could efficiently rescue SFRP1-reduced cell viability (Fig. 5F) and colony formation (Supplementary Fig. S7), indicating that SFRP4 might play an important role in the antagonistic effect on biology function of SFRP1.

After depletion of SFRP4 in MGC80–3 and AGS cells with siRNA, reduced cell viability and colony formation were observed. In contrast, SFRP4 overexpression in gastric cancer cells could significantly increase cell viability and colony formation in vitro (Fig. 6A and B). To address whether SFRP4 promotes tumor progression by regulating Wnt pathway in gastric cancer, we introduced Wnt signaling inhibitor IWR-1 to treat SFRP4-overexpressing gastric cancer cells. The results showed that IWR-1 could remarkably reduce the increased cell viability and colony formation mediated by SFRP4 overexpression in MGC80–3 and AGS cells (Fig. 6A and B). Furthermore, IWR-1 significantly suppressed tumor growth mediated by SFRP4 overexpression in a nude mouse model (Fig. 6C–E).

SFRPs were frequently epigenetically silenced by promoter hypermethylation in many cancer types (4); thus, their antagonistic role in Wnt pathway activation was diminished. However, unlike other SFRPs, the expression of SFRP4 was upregulated in gastric cancer. Some studies revealed that SFRP4 had less frequent promoter hypermethylation in colorectal carcinoma than other members of the SFRP family (16). To investigate the methylation status of SFRPs in gastric cancer, we first analyzed the methylation data from the GEO and TCGA datasets. The results showed that the methylation of most SFRPs in tumors were higher than that in normal tissues; however, the methylation of SFRP4 showed lower level in tumors (Supplementary Fig. S8A–S8C), especially in early-stage tumors (Supplementary Fig. S8D). Unlike other family members, SFRP4 also displayed a decreased methylation compared to SFRP1, SFRP2, and SFRP5 in gastric cancer specimens (Supplementary Fig. S8E–S8G). Interestingly, the methylation of SFRPs in tumors did not show significant association with tumor stages (Supplementary Fig. S8H). The lower methylation level of SFRP4 was also observed in different gastric cancer cells (Supplementary Fig. S8I). Then, we studied the methylation status of SFRPs in 5 gastric cancer specimens by quantitative methylation analysis based on Sequenom MassARRAY platform. Consistent with the online results, the methylation level of SFRP4 was significantly lower than that of other SFRP family members in gastric cancer (Supplementary Fig. S8J).

Despite the considerable efforts in early screening and improved treatment of gastric cancer, the mortality of this fatal malignancy remains high (22). The Wnt signaling pathway has been widely described in cancer and has been recognized as an important contributor to tumor development (23). SFRPs are a family of glycoproteins that have been recognized as antagonists of Wnt ligands and are commonly silenced in cancers (24–26). However, we first identified that SFRP4, a family member of SFRPs, was significantly overexpressed in gastric cancer, especially in advanced gastric cancer, by genomes analysis of 80 gastric cancer tissue samples from the GEO dataset (GSE27342). We also found that high expression of SFRP4 could promote cell viability and proliferation, inhibit cell apoptosis, and regulate the cell cycle. Our results suggested that SFRP4 may act as an oncogene and serve as a potential biomarker in gastric cancer.

SFRP4 has showed disparate expression patterns in previous studies (12, 27–29). The expression of SFRPs has been reported to be regulated by promoter methylation (4, 30). In our study, we found that the methylation status of SFRP4 was significantly lower than that of other SFRPs members, which was consistent with the results in colorectal cancer (14). The decreased methylation of SFRP4 may explain the possible mechanism for its overexpression in gastric cancer. Our study also demonstrated that unlike other SFRPs members that played antagonistic roles in Wnt signaling, SFRP4 could promote the activation of Wnt pathway. The different biological function of SFRP4 may be attributed to the fact that SFRP4 is the largest member of the SFRPs and demonstrated the least structural homology when compared to other family members (31). SFRP4 comprises a netrin-like domain (NLD) and a cysteine-rich domain (CRD), which had been shown to increase intracellular calcium levels, leading to the activation of the Wnt/Ca2+ signaling pathway (32, 33). The new function of SFRP4 in Wnt pathway activation (34) was also proposed by Suzuki (14) and He (35). In addition, as SFRPs were secreted proteins, we cultured gastric cancer cells with recombinant SFRP4 and SFRP1, and then examined the expression of Wnt pathway markers. The result showed that the suppressed activation of the Wnt pathway mediated by SFRP1 could be rescued by recombinant SFRP4 in a dose-dependent manner. We speculated that SFRP4 and SFRP1 could competitively bind to Wnt ligands. In gastric cancer, SFRP1 was downregulated; however, SFRP4 was upregulated, and as a result, SFRP4 had a better binding capacity than SFRP1.

In conclusion, our study demonstrated that SFRP4 expression was markedly increased at both the mRNA and protein levels and was associated with tumor progression and poor patient survival. In vitro and in vivo studies showed that SFRP4 could promote tumor progression by activating Wnt pathway signaling. In addition, SFRP4 was upregulated by the decrease of methylation, and might act a competitive inhibitor of other SFRPs, resulting in activation of the Wnt signaling pathway. All these results indicated that SFRP4 had significant clinical relevance with gastric cancer, and could serve as a novel molecular target for gastric cancer therapy.

No disclosures were reported.

H. Li: Data curation, investigation, visualization, methodology, writing–original draft, project administration. J. Zhao: Investigation, visualization, methodology, writing–original draft, project administration. J. Sun: Investigation, visualization, methodology, writing–original draft, project administration. C. Tian: Visualization, methodology, writing–original draft. Q. Jiang: Software, methodology, project administration. C. Ding: Data curation. Q. Gan: Data curation, software. P. Shu: Data curation. X. Wang: Conceptualization, supervision, funding acquisition. J. Qin: Writing–review and editing. Y. Sun: Conceptualization, supervision, funding acquisition.

We thank the pathologists in Zhongshan Hospital for their work of tissue microarrays construction, IHC staining, and evaluation of tissue samples in our study. This work was supported by grants from the National Natural Science Foundation of China (grant nos. 31670806, 81872425, and 81972228).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.