LGR4 maintains HGSOC cell epithelial phenotype and stem-like traits.

OBJECTIVE
High-grade serous ovarian cancer (HGSOC) is lethal mainly due to extensive metastasis. Cancer cell stem-like properties are responsible for HGSOC metastasis. LGR4, a G-protein-coupled receptor, is involved in the maintenance of stem cell self-renewal and activity in some human organs.


METHODS
TCGA and CCLE databases were interrogated for gene mRNA in ovarian cancer tissues and cell lines. Gain and loss of functions of LGR4, ELF3, FZD5 and WNT7B were performed to identify their roles in ovarian cancer cell epithelial phenotype and stem-like properties. In vivo experiments were performed to observe the effect of LGR4 on ovarian cancer cell growth and peritoneal seeding. The binding of ELF3 to LGR4 gene promoter was investigated by dual-luciferase reporter assays and ChIP.


RESULTS
LGR4 was shown to be overexpressed in HGSOCs and maintain the epithelial phenotype of HGSOC cells. LGR4 knockdown suppressed POU5F1, SOX2, PROM1 (CD133) and ALDH1A2 expression. Furthermore, LGR4 knockdown reduced CD133+ and ALDH+ subpopulations and impaired tumorisphere formation. To the contrary, LGR4 overexpression enhanced POU5F1 and SOX2 expression and tumorisphere formation capacity. LGR4 knockdown inhibited HGSOC cell growth and peritoneal seeding in xenograft models. Mechanistically, LGR4 and ELF3, an epithelium-specific transcription factor, formed a reciprocal regulatory loop, which was positively modulated by WNT7B/FZD5 ligand-receptor pair. Consistently, knockdown of ELF3, WNT7B, and FZD5, respectively, disrupted HGSOC cell epithelial phenotype and stem-like properties.


CONCLUSION
Together, these data demonstrate that WNT7B/FZD5-LGR4/ELF3 axis maintains HGSOC cell epithelial phenotype and stem-like traits; targeting this axis may prevent HGSOC metastasis.

LGR4, also known as GPR48, a G-protein-coupled receptor, plays an oncogenic role in human cancers.
Therefore, LGR4 is potentially associated with cancer stem cells. In the present study, LGR4 was identi ed to maintain HGSOC cell epithelial phenotype and stem-like traits.
LGR4 knockdown suppressed HGSOC cell growth in vitro and in vivo. Furthermore, LGR4 knockdown interfered with HGSOC cell peritoneal seeding.

Methods
Patient specimens: Clinical specimens were collected from Liaoning Cancer Hospital & Institute with the informed consent of the patients. 29 HGSOCs, 5 LGSOCs and 8 normal ovary tissues were used. The normal ovary tissues were from the other side ovary of the patients. The use of these specimens for research purposes was approved by Institutional Research Ethics Committee of China Medical University.
Immunohistochemistry: Immunohistochemical staining was performed on 4-μm sections of para nembedded tissues. Xylene and gradient alcohol were used to depara nize and hydrate, respectively. 3% H 2 O 2 was used to eliminate endogenous peroxidase activity. Sections were then incubated with citrate buffer to repair antigen, and blocked by BSA. Primary antibody was added overnight at 4℃. After incubation with corresponding second antibody, sections were stained with DAB. Subsequently, sections were re-stained with hematoxylin and dehydrated with gradient alcohol and xylene. For human specimens, protein expression was scored according to H-scoring method. The scoring formula was H-score=ΣPi*i, where "I" represents the intensity of staining (0-3), and "Pi" stands for the percentage of stained tumor cells (0%-100%). For xenograft tumors, the positive staining tumor cells in 5 randomly selected elds were counted. Primary antibodies were used as follows: LGR4 (Abcam, UK); ELF3 and Cleaved caspase-3 (Cell Signaling Technology, USA); Ki67 (Invitrogen, USA).
In silico analysis: The Cancer Genome Atlas (TCGA) database was interrogated for LGR4 mRNA expression in ovarian cancer tissues. Cancer Cell Line Encyclopedia (CCLE) database was interrogated for gene mRNA expression in a panel of ovarian cancer cell lines. Correlation between two genes was analyzed using Pearson statistics. Heat maps were generated by GraphPad Prism 8.0.
Cell culture and transfection: Human ovarian cancer cell lines OVCAR3 and CAOV3, and HEK293T cells were grown in DMEM medium with 10% fetal bovine serum and 1% penicillin/streptomycin at 37℃ and 5% CO 2 in a humidi ed incubator. 2.5×10 5 cells were transiently transfected with shRNA or overexpression plasmids using Lipofectamine 3000 in Opti-MEM medium according to the product manual. 5×10 4 cells were transfected with shRNA lentiviruses to stably knockdown gene expression, and the transfected cell were further selected using 2μg/ml puromycin.
Western blot: For the total protein extraction, cells were lysed in RIPA buffer containing 10% protease inhibitor on ice for 40 minutes. Protein fragments were centrifuged with 12,000g at 4℃ for 30minutes. Protein concentration was determined using a BCA protein assay kit. Protein lysate was operated by SDS-PAGE and transferred to polyvinylidene di uoride membranes (PVDF). Then 5% BSA was used for blocking. Membranes were incubation with various primary antibody at 4℃ overnight and with respective secondary antibodies. The primary antibodies are as follows: LGR4 (Abcam, UK); ELF3, FZD5, WNT7B, Ecadherin, Vimentin and GAPDH (Cell Signaling Technology, USA). ECL detection system and imaging system were used to analyze the signals.
Immuno uorescence: Cells were seeded in 6-well plates at 1×10 5 /well. After incubation for 48 hours, the cells were washed with cold PBS, xed with 4% formaldehyde, permeabilized with 0.5% Triton X-100, and then blocked with 5% donkey serum and 0.3% Triton X-100 in 1X PBS. Then the cells were incubated overnight at 4 °C with antibodies for E-cadherin (1:100, Cell Signaling Technology, USA) or Vimentin (1:200, Cell Signaling Technology, USA), followed by incubation with Alexa Fluor 488-conjugated antimouse IgG for 2 hours at room temperature in the dark. Subsequently, the nuclei were staining by DAPI for 5 minutes. Immuno uorescence results were observed under a Laser scanning confocal focus microscope.
Flowcytometry: ALDEFLUOR assay kit (StemCell Technologies, USA) was used to detect ALDH activity. In brief, cells were suspended in the ALDEFLUOR assay buffer containing an ALDH substrate and incubated for 45 minutes at 37°C. 1×10 6 cells were used to detect the uorescence signals by Flowcytometry (BD Biosciences, USA). CD133 positive cells were detected by Flowcytometry according to the standard protocol. 1×10 6 cells were harvested and washed with 1×PBS solution for three times, followed by incubation with 1μg uorescein isothiocyanate (FITC)-labeled antibody for CD133 (Cell Signaling Technology, USA) for 30 minutes. Cells were then washed twice and resuspended in 100μl uorescenceactivated cell sorting (FACS) buffer.
Cell viability: Cell viability was determined by CCK8 assay. Cells were seeded in 96-well plates at 2×10 3 /well. After incubation for 48 hours, CCK8 agent (Dojindo Molecular Technologies, Japan) was added. The absorbance at 450 nm was measured by a microplate reader (Bio-Rad Laboratories, USA) at the indicated time.
Colony formation: 1×10 3 cells were seeded in 3.5cm plates. After incubation for 14 days, cells were xed with paraformaldehyde for 15 minutes at room temperature and then dyed by crystal violet (Solarbio, China) for 30 minutes at 37°C. Subsequently, colonies were counted and photographed.
In vivo animal study: All animal experiments were in strict accordance with the institutional guidelines and approved by the Animal Ethics Committee of China Medical University. BALB/c nude mice (5-6 weeks of age, female, 18-20g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd.
(Beijing, China). 1×10 6 cells were subcutaneously injected into bilateral anks of the mice (n=5 per group). Tumor length and width were measured with a vernier caliper every 3 days. Tumor volume was calculated by the formula: V=1/2×length×width 2 . For the peritoneal metastasis assay, 5×10 6 cells were intra-peritoneally (i.p.) injected into BALB/c nude mice (n=5 per group). 30 days after injection, visible metastases in abdominal cavity were counted.
Chromatin Immunoprecipitation (ChIP): ChIP assays were performed using a ChIP assay kit (Beyotime, China) according to the manufacturer's instructions. Brie y, Cells in 10cm plates were xed for 10 minutes at 37 °C with 1% formaldehyde. 1.1ml glycine solution was subsequently added into the plates. To shear the chromatin, cells were treated with 1mM PMSF in SDS Lysis Buffer for 10 minutes at 4°C, followed by cell sonication for 15 minutes 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 1μg ELF3 (Cell Signaling Technology, USA) antibody at 4°C overnight. Then protein A+G agarose was added to precipitate the target protein recognized by the ELF3 antibody for 1 hour 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. The primers used were as follows: ELF3-LGR4promoter forward primers: GATCACGCCACTGGACAC, reverse primers: CTAAATAGGCTTTCATCGT; GAPDH-chip forward primers: TACTAGCGGTTTTACGGGCG, reverse primers: TCGAACAGGAGGAG CAGAGAGCGA.
Co-Immunoprecipitation (Co-IP): Cells were harvested and lysed in RIPA buffer containing 10% protease inhibitor on ice for 40 minutes. Protein fragments were centrifuged with 12,000 g at 4℃ for 30minutes. 500μg protein extract was incubated overnight at 4 °C with 1μg anti-FZD5 (Santa Cruz, USA) or anti-IgG antibody (Santa Cruz, USA) on a rotator. 20μl protein A/G-agarose beads (Santa Cruz, USA) were added for crosslink. The samples were then centrifuged and washed three times with PBS. The beads were harvested and resuspended in 20μl of gel loading buffer. Western blot was subsequently performed.
Statistical analysis: The data are expressed as mean±SD. GraphPad prism 8 was used to analyze the data. Differences were analyzed by Student's t test or one-way ANOVA. P value less than 0.05 was considered statistically signi cant.

Results
LGR4 is overexpressed in HGSOCs and maintains the epithelial phenotype of HGSOC cells.
LGR4 contributes to HGSOC cell stem-like traits. OVCAR3 and CAOV3 cells have been demonstrated with high and low tumorigenicity, respectively (14).
LGR4 promotes HGSOC cell growth and peritoneal seeding. The role of LGR4 in HGSOC cell growth was investigated in vitro and in vivo. Both CCK8 and colony formation tests showed that LGR4 knockdown interfered with the growth of in vitro cultured OVCAR3 cells (Fig. 3A, 3B). Xenograft models using nude mice further demonstrated that LGR4 knockdown delayed the appearance and hampered the growth of OVCAR3 tumors (Fig. 3C). Moreover, xenograft tumors with LGR4 knockdown demonstrated reduced Ki67 expression, whereas increased cleaved caspase-3 expression (Fig. 3D). As epithelial stemness facilitates tumor cells to colonize in secondary sites, a peritoneal seeding model was adopted to evaluate the role of LGR4 in HGSOC metastasis. OVCAR3 cells with LGR4 knockdown formed less and smaller metastatic colonies in abdominal cavity compared with those with control knockdown (Fig. 3E).
LGR4 and ELF3 form a reciprocal regulatory loop. To inquire whether a link exists between LGR4 and ELF3, an epithelium-speci c transcription factor, ELF3 expression was detected in HGSOCs by IHC. Hscore statistics revealed a similar expression pattern of ELF3 to that of LGR4 (Fig. 4A, 4B). Interrogation of CCLE database further uncovered a positive correlation between LGR4 and ELF3 (Fig. 4C).

Discussion
Based on the nding that LGR4 is preferentially expressed in HGSOCs, two HGSOC cell lines were selected for subsequent gain and loss of function study. On the one hand, LGR4 maintains the epithelial phenotype.
LGR4 regulates E-cadherin positively, whereas Vimentin and ZEB2 negatively. This is further supported by CCLE database interrogation showing that LGR4 is correlated with a series of epithelial factors including not only CDH1 (E-cadherin), but also EPCAM, ELF3, CLDN2, CLDN3, KRT8 and KRT19. On the other hand, LGR4 maintains cancer cell stem-like properties.
It was reported that breast cancer stem cells (BCSCs) exist in distinct mesenchymal and epithelial states (18,19). Mesenchymal BCSCs are characterized by epithelial-mesenchymal transition (EMT), CD24-CD44+, quiescence and being at the tumor invasive front; while epithelial BCSCs are ALDH1+, proliferative, located more centrally and characterized by mesenchymal-epithelial transition (MET).
Similarly, EOC stemness exhibits phenotypic and functional heterogeneity (20). Mesenchymal EOC stem cells express CD44, CD117 and TGF-β1 (21,22), while epithelial EOC stem cells are characterized by MET and prone to colonize at metastatic sites (23). Our study found that LGR4induces epithelial phenotype and stem-like properties which promotes EOC cell growth and peritoneal seeding.
ELF3 plays complicated roles in human cancers. ELF3 functions as a tumor suppressor in ampullary carcinoma (24,25), but as an oncogene in lung adenocarcinoma (26,27). There are even contradictory observations about the role of ELF3 in prostate cancer (28,29). Yeung et al. reported that ELF3 is a negative regulator of EMT (30). This is consistent with our nding. However, our study revealed that ELF3 is implicated in EOC stem-like properties, potentially inducing EOC cell growth and metastasis through regulating LGR4. The association of LGR4-ELF3 with epithelial phenotype may be related to the suppression of ZEB2. ELF3 negatively regulates ZEB1/2 expression in breast cancer cells (31).
WNT7A was shown to promote EOC cell growth, invasion and migration via FZD5 in a β-catenindependent way (32,33). Similarly, WNT7B/FZD5 pair was found to activate β-catenin pathway in our study. However, WNT7B/FZD5 pair modulates LGR4 and ELF3 expression and maintains EOC cell stemlike properties independent on β-catenin. Therefore, WNT7B/FZD5 pair seems to be involved in EOC via both canonical and non-canonical Wnt pathways.
In summary, our study for the rst time elucidated the role of LGR4 in HGSOC.
LGR4 maintains HGSOC cell epithelial phenotype and stem-like traits, and promotes HGSOC cell growth and peritoneal seeding. Mechanistically, LGR4 and ELF3 form a reciprocal regulatory loop, which is modulated by WNT7B/FZD5 pair via non-canonical Wnt pathway. Furthermore, ELF3, WNT7B and FZD5 are also implicated in HGSOC cell epithelial phenotype and stem-like traits. Targeting these molecules may potentially suppress HGSOC growth and metastasis.

Consent for publication
Not applicable.

Availability of data and materials
Not applicable.

Competing interests
The authors declare that they have no competing interests.  LGR4 is overexpressed in HGSOCs and maintains the epithelial phenotype of HGSOC cells. (A) LGR4 expression was detected in HGSOCs (n=29), LGSOCs (n=5) and normal ovaries (n=8) by Immunohistochemistry (IHC). (B) LGR4 expression was scored in LGSOCs (n=5) and HGSOCs (n=29). (C) LGR4 expression was scored in serous EOCs without metastasis (n=10) and those with peritoneal metastasis (n=24). (D) TCGA database was interrogated for LGR4 expression in Grade 1+2 (n=43) and Grade 3+4 (n=321) ovarian cancers. (E) A heat map was generated from CCLE database. (F) LGR4 expression was detected in OVCAR3 cells transfected with Control shRNA or different LGR4 shRNAs by   LGR4 promotes HGSOC cell growth and peritoneal metastasis. LGR4 shRNA were inoculated into nude mice (n=5 in each group). Tumor volume was calculated according to 1/2×length×width2. (D) Xenograft tumors were con rmed by HE staining. LGR4, Ki67 and cleaved caspase-3 (C-casp-3) expression was detected in xenograft tumors by IHC. (E) OVCAR3 cells transfected with Control shRNA or LGR4 shRNA were seeded into peritoneal cavity of nude mice (n=5 in Figure 4 LGR4 and ELF3 form a reciprocal regulatory loop. (A) LGR4 and ELF3 expression was detected in the same panel of serous EOCs (n=34) by IHC. (B) LGR4 and ELF3 expression was scored, and correlation between LGR4 and ELF3 was analyzed by Pearson statistics. (C) CCLE database was interrogated for LGR4 and ELF3 expression. Correlation of LGR4 with ELF3 was analyzed by Pearson statistics.   LGR4-ELF3 is modulated by WNT7B/FZD5 pair. (A) CCLE database was interrogated for FZD5, WNT7B, LGR4 and ELF3 expression. Correlation between two genes was analyzed by Pearson statistics. (B) Co-IP test was performed in OVCAR3 cells transfected with Control shRNA or FZD5 shRNA. (C) FZD5, ELF3, and LGR4 expression was detected in OVCAR3 cells transfected with Control shRNA or FZD5 shRNA by