FHL3 Contributes to EMT and Chemotherapy Resistance Through Up-Regulation of Slug and Activation of TGFβ/Smad-Independent Pathways in Gastric Cancer

Background Gastric cancer presents high risk of metastasis and chemotherapy resistance. Hence, it is important to understand the mechanisms of gastric cancer distant metastasis and chemotherapeutic resistance. Our previous study has revealed Four and a Half LIM Domains 3 (FHL3) plays as a binding partner of Glycogen Synthase Kinase 3 Beta (GSK3β), promoted tumor metastasis in pancreatic cancer. However, the role of FHL3 in gastric cancer still remains unclear. Methods TCGA database and clinical samples are used for exploring the role of FHL3 in disease progression and prognosis. Oxaliplatin (OHP) resistance cell lines were established to study the role of FHL3 in chemotherapy resistance. The experiments about cell proliferation, apoptosis, and metastasis were performed to measure the chemotherapy effects of sh-FHL3 on gastric cancer cell lines and in vivo. That FHL3 changed the EMT phenotype was verified by western blot. Finally, we explored the mechanism of FHL3-mediated EMT and chemotherapy resistance. Results mRNA and protein level of FHL3 were significantly up-regulated in gastric cancer tissues when compared with adjacent tissue. FHL3 higher expression is always accompanied with higher TNM stage and worse overall survival. FHL3 over-expressed could lead to OHP resistance. Knockdown of FHL3 slightly inhibited the cell growth, while it obviously sensitized the chemotherapy in vivo and in vitro. In addition, down-regulation of FHL3 increased the mesenchymal markers, such as Slug, Snail, Twist Family BHLH Transcription Factor 1 (Twist1), and Vimentin, while it decreased the epithelial marker E-cadherin. Cell and animal experiments also proved that down-regulation of FHL3 can decrease cancer cell metastasis. For mechanism study, FHL3 knockdown down-regulated the expression level of Mitogen-Activated Protein Kinase (MAPK)/Extracellular Regulated Protein Kinase (ERK) pathway and Transforming Growth Factor-β (TGFβ)/Phosphatidylinositol 3-Kinase (PI3K)/protein kinase B(Akt)/GSK3β-(Ring Finger Protein 146) RNF146/ubiquitin pathway. FHL3 competitively bonded the ubiquitin complex (Slug/GSK3β/RNF146) with Slug and inhibited ubiquitination of Slug. Mesenchymal phenotype cells hold higher level of Multidrug Resistance Gene1 (MDR1), and the FHL3 knockdown reverts the MDR1 in this type cell. Conclusion FHL3 high expression contributed to EMT and chemotherapy resistance via MAPK, and PI3K pathways were activated. FHL3 competitively bonded the ubiquitin complex with Slug, resulting in the up-regulation of Slug and leading to metastasis of gastric cancer.


INTRODUCTION
Gastric cancer (GC) is the fifth most common malignancy and accounts for the third leading cause of cancer death (1). Although prediction about the prognosis of GC is mainly according to the lymph node and distant metastasis status, increasing lines of evidence show that conventional staging criteria fail to sufficiently distinguish the prognostic differences of GC (2). Due to the high risk of lymph node metastasis and distant metastasis, about 65% cases of GC appear to have tumor recurrence and disease progression after surgical treatment, which leads to <35% rate of 5-year survival of GC (3). Many patients with clinical indications (advanced TNM stage) have to receive chemotherapy after resection surgery. As for the first-line chemotherapy drug for GC, 5-FU and oxaliplatin (OHP) usually appear tolerant and resistant during treatment; chemotherapy resistance is the main barrier that leads to unsatisfactory clinical outcomes (4). Thus, new prognostic biomarkers for GC are urgently needed to improve the early diagnosis and chemotherapy effectiveness of gastric cancer. Meantime, the mechanistic understanding of tumor metastasis and chemotherapy resistance should be improved to develop novel targets for advanced gastric cancer.
Most signaling pathways interfere with EMT process via the regulation of E-cadherin through EMT-TFs, such as Smadindependent TGFb pathway and Smad-dependent TGFb pathway (9)(10)(11). For the former one, TGFb can initiate the Wnt/b-catenin pathway, PI3K/Akt pathway, and MAPK/ERK pathways to up-regulate the level of EMT-TFs (11). For the later one, TGF-b activates Smad-complex/EMT-TFs to up-regulate the level of EMT-TFs (9).
Interestingly, recent studies implied that the roles of EMT beyond the traditional views in which it was the pivotal role in initiation of tumor metastasis and invasion and developing lines of evidence pointed out that the EMT-associated components (such as Snail, zeb, and Twist1) are closely related to drug resistance (12,13). And, the mechanism of EMT-associated chemotherapy is complex. Cancer stem cell (CSC) is a feature in mesenchymal phenotype tumor cells, which is closely related to chemotherapy resistance (14)(15)(16). In addition, some regulating factors directly related to EMT can promote chemotherapy resistance. EMT-TFs including Twist, Snail, Slug, and Forkhead Box C2 (FOXC2), regulate chemoresistance by increasing the expression of ATPbinding cassette (ABC) transporters in breast cancer, which is important for chemoresistance (17,18). Slug can inhibit the activity of caspase-9, leading to chemoresistance of tumor (19). However, single inhibition of EMT-TFs has few effects in reverting chemoresistance. The main reason of this phenomenon is other pathways which participate in the regulation of EMT process. According to previous studies, some EMT-associated pathways are also important in chemoresistance mechanism, such as PI3K/Akt pathway, MAPK pathway, and hypoxia pathway, through upregulation of ABC transporters and decreasing apoptosis (20)(21)(22)(23)(24)(25). Thus, the mechanism of EMT-associated chemotherapy resistance is complicated and unclear; more work is needed to figure out the mechanism of EMT-associated chemotherapy resistance.
LIM-only protein Four-and-a-Half LIM domain (FHLs), including FHL1, FHL2, and FHL3, are characterized by evolutionarily conserved LIM domains and one conserved LIM superfamily domain (26). FHLs are reported as the transcriptional factors, which participate in lots of signaling pathways. Simply, FHL1-2 are reported in the regulation of TGFb/Smadindependent pathway, such as PI3K/Akt, Wnt/b-catenin, and MAPK/ERK pathways. FHLs participate in the EMT process and chemo-radio-therapy resistance in pancreatic cancer, breast cancer, and osteosarcoma (7,26,27). Besides, FHL1-3 are considered as inhibitors of cell cycle checkpoints Cell Division Cycle 25 (CDC25) in pancreatic cancer and HeLa cell line and that they always lead to radioresistance (28). In addition, some studies show that FHLs can interact with Estrogen Receptor-aPolypeptide (ER-a) to make tumor progression in breast cancer (29,30). Another study shows FHL2 can directly interact with epithelial phenotype marker ZO-1 to promote tumor invasion in breast cancer (31). In our previous study, FHL3 also acts as a regulator in ubiquitin degradation process of EMT-TFs through Akt/GSK3b/ ubiquitin pathway in pancreatic cancer (7). However, some previous studies suggest that FHL1-3 perform as a tumor repressor in lung cancer, liver cancer, and breast cancer. As of now, no research article has revealed the role of FHL3 in gastric cancer, especially the regulatory mechanism of FHL3 in chemoresistance and metastasis.
In fact, our previous study has proved that FHL3 is an important role in the regulation of EMT. In this study, we explored the potential relationship between FHL3 and EMT/ chemotherapy resistance in gastric cancer. Simply, we explored the role of FHL3 in disease progression and overall prognosis by investigating TCGA database and clinical GC samples. Oxaliplatin (OHP) resistance cell lines were established to study the role of FHL3 in chemotherapy resistance. The experiments about cell proliferation, apoptosis, and metastasis were performed to measure the chemotherapy effects of sh-FHL3 on gastric cancer cell lines and in vivo. That FHL3 changed the EMT phenotype was verified by western blot. The mechanism of FHL3-mediated EMT and chemotherapy resistance was clarified.

Gastric Cancer Sample Preparation
This study was approved by The First Affiliated Hospital of Anhui Medical University Review Board and the ethics committees of Anhui Medical University. Patients gave their informed consent to use gastric cancer samples and slices in this study. 120 matched paraffin-embedded tumor tissue sections and 16 paired fresh frozen tissues were collected. All patients underwent total or partial gastrectomy at the First Affiliated Hospital of Anhui Medical University from 2013 to 2016. All patients with gastric cancer were confirmed by at least two pathologists. Follow-up time was estimated from the date of surgical treatment to that of an event (i.e., patient death or tumor recurrence) or withdrawal.

Bioinformatic Analysis
Gene profiles of GC and non-tumor adjacent tissues based on microarray were downloaded from the Gene Expression Omnibus database (GEO; https://www.ncbi.nlm.nih.gov/geo/). TCGA and GTEx RNA sequencing FPKM data of GC, non-tumor adjacent tissues, and normal stomach tissues were downloaded from the UCSC Xena database (https://xenabrowser.net/hub/). The proteomic data of GC and non-tumor adjacent tissues were downloaded from PRIDE Archive under the accession number PXD011821. KEGG pathway analysis was performed using R clusterProfiler package. Gene Set Enrichment Analysis (GSEA) was performed by the JAVA program using gene set collection (c2.cp.v7.1.symbols.gmt) from the MsigDB.

Cell Culture and OHP-Resistance Cell Lines
Gastric cancer cell lines (SGC-7901, HGC, AGS, and N87) and normal gastric epithelial cells (GES-1) were obtained from the cell bank of the Chinese Academy of Science. All cell lines were cultured in RPMI-1640 medium (Gibco, USA). All culture media were supplemented with 10% fetal calf serum and 100 units/ml penicillin and streptomycin. Gastric cancer cell lines (HGC and SGC) are firstly detected by the IC50 of OHP. Secondly, the cells are screened in this dose-treatment of OHP for at least three generations to establish the OHP-resistance cell lines. Then, the OHP IC50 of those cells is re-detected, and the cells are screened in the this IC50 OHP again. The screen cycle is performed at least six times.

Western Blot Analysis
Total protein extraction: Cells were harvested with a cytology brush, lysed with RIPA lysis buffer (Sigma-Aldrich, USA) supplemented with a phosphorylase and protease inhibitor mixture (Thermo Fisher Scientific, USA), and quantified by a BCA assay.

MTT Assay
The cells in the logarithmic phase were plated onto 96-well plates at a density of 5,000 cells per well in 200 µl of culture medium and incubated for 24, 48, and 96 h at 37°C with 5% CO 2 . A volume of 20 µl MTT solution (5 mg/ml; Solarbio Science & Technology, Beijing, China) was added into each well and incubated for another 4 h. The MTT solution was then removed, and 100 µl dimethyl sulfoxide (Sigma) was added to each well. The relative optical density (OD) was measured at 570 nm (Spectra Max, USA), and the experiment was repeated three times.

Immunofluorescence
Briefly, 2.5 × 10 4 cells were seeded in 24-well plates for 24 h, fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100 and blocked with 5% BSA (Sigma-Aldrich) for 1 h at 37°C. The samples were incubated with a primary antibody (FHL3, 1:200, Proteintech) overnight at 4°C. Subsequently, the cells were washed with PBS and incubated with secondary antibodies for 1 h at room temperature before being washed again. Finally, the nuclei were stained with 15 ml DAPI (Sigma-Aldrich, USA) before detection with a fluorescence microscope (Carl Zeiss, Germany).

Colony Forming Efficiency Assay
Firstly, cells were seeded in six-well plates at a density of 1,500 cells per well and incubated in 37°C for 10 days. Then cells were washed with PBS and fixed with 1 ml 4% formaldehyde solution. Then 1 ml crystal violet staining solution was added and washed with PBS for three times after 30 min. The newly formed colony units were counted by summing the number of different fields (100× microscape, five fields).

Live and Death Staining
Cells (10 × 10 4 /per well) were seeded in a six-well plate, and then the cells were transfected with three FHL3 siRNA for 48 h. After that, AO and PI solutions (1% AO and PI, 5 ul AO and 5 ul were added into dyeing diluent buffer) were added into plates and maintained for 0.5 h before observing by an inverted fluorescent microscope. AO could stain the live cells that show the green fluorescence; PI could stain the dead cells that show the red fluorescence. Fluorescence intensity was quantified by ImageJ.

Wound-Healing Assay
Cells (40 × 10 4 /per well) were cultured in a six-well plate and grown to 90% confluence in 2 ml of culture medium. A 200 µl plastic tip was used to create an artificial wound. Images were taken at 0, 24, and 48 h after scratching. The cell mobility = (0 h width-the indicated time points width)/0 h width × 100%.

Migration Assay
Transwell (Corning Life Sciences, Bedford, MA, USA) was used to assess GC cell migration. For migration assays, 1 × 10 5 cells were added to 200 µl serum-free DMEM in the upper chamber, and the lower chamber was filled with 600 µl culture medium. After incubation at 37°C in an atmosphere containing 5% CO 2 for 24 h, the non-migrated cells were carefully removed with a wet cotton swab. Finally, the cells were stained with Giemsa (Sigma, USA) for 10 min followed by imaging and counting under an inverted microscope (100× magnification).

Breeding Conditions and Method of Euthanization
Male nude mice and severe combined immunodeficiency (SCID) mice were bred in the SPF condition, 26-28°C, 40% humidity, 10 h illumination. When the in vivo experiments were performed, the nude mice were killed by cervical dislocation.

Subcutaneous Tumor Model
All animal procedures were performed in accordance with the Guidelines for Care and Use of Laboratory Animals of Anhui Medical University and approved by the Animal Ethics Committee of Anhui Medical University.
Male nude mice (4 to 6 weeks old) obtained from the SLAC (Shanghai, China) were randomly divided into two groups (three nude mice per group). A total of 1 × 10 6 cells (FHL3-NC and FHL3-SH1 cells) in 100 µl PBS were injected subcutaneously. 4 weeks later, the mother tumors are harvested to make the same volume transplanted tumors. Then the FHL3-NC-derived and FHL3-SH1-derived tumors are transplanted into mice of different groups (five nude mice per group). The tumor volume is investigated every day. All mice were sacrificed, and the tumors are harvested to determine the tumor volume (MaA MiA 2 /2; MaA = Major axis, MiA = Minor axis), followed by processing into sections for HE staining and Ki67 staining.

Orthotopic Transplantation Model
Simply, the transplanted tumors are transplanted into the subserosa of the stomach in SCID mice. 4 weeks later, mice were executed to investigate the volume and weight of tumors.

Lung Metastasis Model
Simply, 1 × 10 6 cells are injected into the tail vein of SCID mice after 4 weeks, and the lungs are harvested for frozen sections to investigate the metastatic tumors by FITC-tag.

Immunoprecipitation and Recombination Plasmid System
Primers of Slug, FHL3, and RNF146 are inserted into plasmid pcDNA 3.1(−) (Addgene). Briefly, cDNA templates are synthesized through PrimeScript RT Reagent Kit (TaKaRa, China); CDS of genes are amplified with PrimeSTAR ® GXL DNA Polymerase (TaKaRa, China); products are purified by SanPrep Column DNA Gel Extraction Kit (Sangon Biotech, China); the purified products and plasmids are treated with restriction endonuclease (Xho1, EcoR5, and Xba1 come from NEB, USA) respectively; recombination of plasmids are performed through homologous recombination with Hieff CloneTM Plus One Step Cloning Kit (Yeasen Biotech, China). Cells were transplanted into six-well plates for 24 h followed by transfection of RP for different times with Hieff TransTM Liposomal Transfection Reagent (Yeasen Biotech, China) to find out the best transfection efficiency, according to the manufacturer's instructions.
A total of 1 × 10 7 cells were harvested with a cytology brush and lysed with RIPA lysis buffer (Yeasen Biotech, 20118ES60) to isolate the protein supernatant, followed by adding magnetic beads (Anti-Myc, Anti-HA Bimake and Anti-Flag) with continuous slight mixing at 4°C for 24 h. Then, the magnetic beads were isolated with a magnet (Bimake), followed by washing with TBS. Finally, the products were boiled before being dissolved in 5× SDS (Yeasen) for 5-10 min for Western blot assays.

Statistics
All independent cell experiments were repeated three times. All experimental data are presented as the mean ± SD. Statistical Package for the Social Sciences version 21.0 (SPSS Inc., USA) was used for the statistical analyses. ANOVA, paired t-test, Chisquare () test, and a non-parametric test (Mann-Whitney U) were used for statistical analysis in different situations. Statistical significance was defined as P <0.05 (*P < 0.05; **P < 0.01; ***P < 0.001). All histograms and curves were constructed with GraphPad Prism 6 software (GraphPad Software, La Jolla, CA, USA).

FHL3 Plays a Disease Progression Role and Prognosis Prediction Biomarker in GC
To investigate the clinical relevance of FHL3 in GC, we systematically analyze multiple publicly available gene expression datasets [The Cancer Genome Atlas/Genotype Tissue Expression (TCGA/GTEx), GSE13861, GSE13911, GSE19826, GSE29998, GSE54129, and GSE63089], which contain >1,000 gastric cancer patients. We noticed that FHL3 mRNA expression is remarkably up-regulated in GC ( Figures 1A, B, D-G, P < 0.001). In line with this conclusion, FHL3 protein level was also significantly elevated in GC by proteogenomic analysis of Beijing dataset which contained 58 paired gastric tumor samples and non-tumor adjacent tissues (PXD011821, Figure 1H, P < 0.001). Furthermore, we explored the role of FHL3 in the progression of GC. We performed Kaplan-Meier analyses in both TCGA and KM-plotter cohorts, and the survival curves showed that patients with higher expression level of FHL3 have shorter overall survival time than those with lower expression levels in TCGA (HR = 1.40, 95% CI = 1.15-1.51, P = 0.039, Figure 1I) and KM-plotter cohorts (HR = 1.55, 95% CI = 1.31-1.84, P < 0.001, Figure 1J).
To confirm this observation, we examined the FHL3 level in clinical gastric tumor samples. According to the IHC staining result, up-regulated FHL3 expression was positively associated with the tumor TNM stage (Figure 2A). Consistently, the expression of FHL3 was dramatically up-regulated in GC tissues when compared with adjacent tissues in both 120 matched paraffin-embedded tumor tissue sections and 16 paired fresh frozen tissue (Figures 2B, D, E, P < 0.05). In addition, according to the FHL3 expression and the follow-up time of 120 GC patients, Kaplan-Meier analysis indicated that higher expression of FHL3 leads to worse prognosis in GC patients (P < 0.001, Figure 2C). In 16 paired GC tissues, higher expression level of FHL3 is always accompanied with worse TNM stage ( Figure 2F, P = 0.0484).

FHL3 Knockdown Reduces OHP Resistance In Vitro
In order to investigate the relationship between the expression of FHL3 and chemotherapy in GC, we first screened out the FHL3high-expression GC cell lines. We detected the FHL3 expression level in a normal gastric cell line (GES-1) and several gastric cancer cell lines (SGC, HGC, AGS and N87). As the results show, FHL3 is highly expressed in all GC cell lines in contrast with

FHL3 Knockdown Reduces OHP Resistance and Metastasis in Subcutaneous/Orthotopic Stomach Tumor Bearing-Model and Lung Metastasis Model
In in vivo experiment, we found the growth speed of subcutaneous tumors is slower in the single FHL3-SH1 group and OHP therapy compared with the FHL3-SH1 group ( Figures 5A, B, P < 0.05). Combination treatment (FHL3-SH1 and OHP) significantly inhibits the tumor growth when compared with the FHL3-SH1 + NS and FHL3-NC + OHP groups (P < 0.01, Figures 5A, B). We detected the tumor volume and weight. As Figures 5C-D show, combination treatment expressed better tumor growth inhibition (about one-fold) when compared with single OHP treatment or  single FHL3-knockdown treatment in HGC subcutaneous tumor model (P < 0.001). As shown in Figure 5E, IHC staining was performed, and the results showed that tumors with combination treatment have a weaker intensity of Ki-67. Those results suggested FHL3-knockdown could enhance the tumor growth inhibition of OHP therapy and sensitize the chemotherapy. Then, we validated the role of FHL3 in metastasis of gastric cancer cells in vivo. As Figure 6A show, the orthotopic tumor was performed, and tail intravenous injection of gastric cancer cells was done. 4 weeks later, the orthotopic stomach tumors and lung metastasis nodes were harvested for tumor detection. In this section, we found orthotopically transplanted tumors FHL3-SH1 treated with OHP were more than 50% smaller, when compared with FHL3-NC treated with OHP (P < 0.001, Figures 6B, E, F). According to the white nodules on the surface of the lungs ( Figure 6C), we could easily found the metastasis in situ in mice of the FHL3-NC + OHP group. We detected slices from these lung samples and performed immunofluorescence (Figures 6D,  G), the green area can be identified as metastatic tumor from the blood cycle in the lung (which may come from tail vein injection or orthotopic stomach in situ). According to our observation, lung metastasis occurred in 1/5 SCID mouse in the FHL3-SH1 + OHP group (20%), while it occurs in 4/5 SCID mice in FHL3-NC + OHP (80%). Those results indicated FHL3 knockdown can improve the OHP efficacy and decrease the lung metastasis more than 60%.

RNF146-Mediated Degradation Participates in FHL3-Induced EMT Process
In our previous study, we have found that FHL3 participates in Akt/GSK3b pathway-mediated ubiquitination degradation of EMT-TFs, and the E3 ligase RNF146 was firstly reported to interact with FHL3 in pancreatic cancer. Firstly, we treated the NC and sh-FHL3 GC cells with MG132 (an ubiquitin inhibitor) (5 uM, 3 h), and we found that when FHL3 was knocked down, the MG132 can reverse the degradation of Slug, and the protein level of Slug was up-regulated ( Figures 8A, B). Then, we explored the role of RNF146 in FHL3-mediated stabilization of EMT-TFs in GC. As Figure 8C  FHL3 and Slug ( Figure 8C). Those results implied that RNF146 participates in the ubiquitination degradation of Slug; FHL3 may interfere with EMT process by competitively bonding the ubiquitin complex with Slug.

FHL3 Induces Chemoresistance via MDR1
According to previous studies, TGFb/Smad-independent pathways containing MAPK/ERK and PI3K/Akt participated in chemotherapy resistance, and those pathways have been pointed out to be probably associated with OHP resistance in GC. However, more details are needed to figure out the mechanism of FHL3-mediated chemotherapy resistance. So, we detected the level of Hypoxia-Inducible  Figure 8G). The IHC showed that advanced GC samples with chemotherapy resistance have higher level of FHL3 and MDR1 when compared with early GC ( Figure 8H).

DISCUSSION
Although the morbidity and mortality of GC have declined over the past decade, we still face many problems and challenges in the screening and treatment of GC. TNM stage, encompassing the depth of invasion (T), lymph node metastasis (N), and distant metastasis (M) stages were regarded as the most significant prognostic factors of GC. Due to tumor heterogeneity, even GC patients with the same TNM stage may have different prognoses in survival time after complete surgical resection, indicating that prognosis cannot be accurate if we determined it based on the current staging system. Even receiving complete resection or targeted therapy, many advanced GC patients still die of local recurrence and/or distant metastasis, among which tumor metastasis and chemotherapy failure are the severe problems for all clinical doctors. Although various genes and pathways have been investigated in GC, the mechanisms of metastasis and chemoresistance are still unclear.
Recently, LIM domain-only protein family plays pivotal roles in tumor progression, including radiotherapy resistance and metastasis (7,28). Some studies showed that FHLs play as a tumor repressor in breast cancer, liver cancer, and lung cancer (32)(33)(34), while other studies pointed out that FHLs promote paclitaxel resistance and radiotherapy resistance in liver cancer and HeLa cell respectively (28,35), enhance tumor cell growth in liver cancer, glioma, and breast cancer (36,37), and lead to metastasis in breast cancer and pancreatic cancer (7,30). So far, the role of FHLs in gastric cancer is still unclear. In our study, we found the expression level of FHL3 was obviously up-regulated in GC both in mRNA and protein by analysis in TCGA, GTEx, and Beijing dataset (P < 0.05, Figures 1A-H). Meantime, the same results were obtained from 16 fresh frozen tumor tissues and 120 paraffin-embedded sections (P < 0.01). Then, 120 samples showed higher level of FHL3, leading to lower differentiation (P = 0.009, Table 1), metastasis trend (P = 0.002, Table 1) and worse stage of TNM (P = 0.039, Table 1) in GC. Besides, FHL3 was negatively associated with the prognosis in GC through Kaplan-Meier analysis in TCGA, KM-plotter cohorts and 120 GC samples (P < 0.05, Figures 1I, J, 2C). Furthermore, the univariate analysis showed higher level of FHL3 can accompany higher risk of GC progression ( Table 2, HR = 2.06, P = 0.005). In other words, those results suggested that FHL3 was a potential predictor of disease progression and prognostic factor in GC.
In the following experiments, the tumor growth was decreased~50% in HGC and~20% in SGC by the FHL3 knockdown ( Figures 3G, I, J). In subcutaneous tumor model, our study found that FHL3 knockdown reduced the tumor growth~25% of tumor volume and~25% of tumor weight ( Figures 5C, D). Besides, FHL3 knockdown enhanced the efficacy of OHP both in HGC and SGC ( Figures 3G, I, J), which was similar to the results in subcutaneous tumor model ( Figures 5A-D). In bioinformatic analysis, we found the level of FHL3 was positively related to the level of Slug, Snail, Twist1, Ncadherin and Vimentin, while it was negatively related to the level of E-cadherin ( Figures 4A1-6). FHL3 knockdown obviously decreased the expression level of Vimentin, Snail, and Slug and increased the level of E-cadherin and ZO-1 both in HGC and SGC ( Figures 4B1-7). Meantime, migration ability was reduced by FHL3 knockdown (Figures 4C-E2).
The mechanisms of how FHL3 regulates tumor chemotherapy resistance and metastasis were unclear in GC. To our knowledge, EMT was mainly responsible for tumor metastasis, and the lower level of E-cadherin was considered as the feature in the EMT process (6,9). E-box is a pivotal DNA reading frame in CDH1 sequence through which some transcriptional factors can directly bind to CDH1 to regulate the expression level of E-cadherin, such as Snail1/2, Twist1/2, Zeb1, and FOXC2 (9). Besides, other cellular junction proteins also play roles in tumor metastasis, such as N-cadherin and ZO-1. The regulation of EMT holds some recognized pathways, among which the TGFb-mediated Smad-dependent and Smadindependent ways were the major roles (6,9). For TGFb/Smaddependent way, Smad complex directly regulates the expression level of EMT-TFs to promote the EMT process (6,9). Previous studies have shown that FHLs promoted phosphorylation of Smad 2/3 and directly interacted with them, leading to nuclear accumulation of Smad complex in liver cancer (32). For TGFb/ Smad-independent way, PI3K/Akt, MAPK, Nuclear Factor Kappa-B (NF-kB), Hedgehog, and Wnt/b-catenin pathways were upstream regulators of EMT-TFs (38)(39)(40)(41). Some studies pointed out that FHLs can increase the activation and transcription of Akt to promote tumor growth and progression in glioma, breast cancer, and ovarian cancer (30,42,43). Other published papers showed that FHLs interfere with the MAPK/ERK pathway, resulting in radiotherapy resistance in pancreatic cancer (44). It was reported that FHL knockdown contributed to tumorigenesis prevention in osteosarcoma by down-regulation of Wnt/b-catenin pathways (27). FHLs were also associated with hepatocarcinogenesis by activating NF-kB pathway (45). Therefore, FHL3 was likely to regulate gastric cancer metastasis through TGFb/Smad-independent pathway. Encouragingly, our study found MAPK, PI3K/Akt, and TGFb pathways were close to FHL3 by KEEG and GO analysis ( Figures 5A, B). And in the following experiments, FHL3 knockdown obviously down-regulated the phosphorylation level of MAPK pathway downstream molecules ERK1/2, P38, and JNK in HCG (Figures 5C1-4). Besides, FHL3 knockdown obviously down-regulated the phosphorylation level of PI3K, Akt, and GSK3b in HCG (Figures 5D1-4). However, FHL3knockdown-induced down-regulation of TGFb has few effects on the phosphorylation level of Smad4. Collectively, FHL3induced EMT was associated with the activation of MAPK/ ERK|JNK|P38 and PI3K/Akt/GSK3b pathways. Furthermore, based on our previous study that FHL3 regulated the Akt/ GSK3b/ubiquitin-Snail1|Twist1 pathway to stabilize the EMT-TFs to promote EMT process in pancreatic cancer, we explored the role of FHL3 in the regulation of ubiquitin-mediated EMT. Here, we hypothesized that E3 ligase RNF146 can form a complex with FHL3 and Slug via GSK3b, and the higher level of FHL3 may induce up-regulation of Slug via inhibiting the degradation (Figures 8A, B).
The mechanism of chemotherapy resistance is complicated. ABC transporters, especially for MDR1, had a central role in chemo-drug efflux to make chemotherapy resistance in gastric cancer (46,47). MDR1 can be directly regulated by the NF-kB pathway. Hypoxia also participated in the regulation of MDR1; up-regulation of HIF-a can lead to chemotherapy in gastric cancer (23). Previous studies showed that inhibition of apoptosis was also important for chemotherapy, and the apoptosisinhibition-mediated chemotherapy was regulated by MAPK pathway and PI3K/Akt pathway (20)(21)(22). Besides, EMT was also considered to be responsible for chemotherapy resistance in pancreatic cancer and breast cancer (12,13). Nevertheless, the characteristic changes of some signaling pathways in mesenchymal phenotype may be related to the chemotherapy resistance. Up-regulation of EMT-TFs, such as Snail and Twist1, simultaneously increased the level of ABC transporters to endow chemotherapy resistance during its up-regulation of the EMT process (17)(18)(19). In fact, our study found that down-regulation of FHL3 promoted the mesenchymal-epithelial transition (MET), during which it may reduce chemotherapy resistance in HGC. Furthermore, our study found MDR1 was down-regulated by FHL3 knockdown (Figures 8D, E). In conclusion, those TGFb/ Smad-independent pathways were the regulators in FHL3mediated chemotherapy resistance.
Collectively, as the scheme of our hypothesis showed ( Figure  8I), FHL3 could competitively bond the complex (GSK3b/ RNF146) with Slug. Slug could induce the EMT process and promote cancer cell metastasis. In addition, FHL3 could induce drug resistance by activating the MAPK and PI3K pathway which may lead to the MDR1 overexpression.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the First Affiliated Hospital of Anhui Medical University Review Board and the ethics committees of Anhui Medical University. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by the First Affiliated Hospital of Anhui Medical University Review Board and the ethics committees of Anhui Medical University. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.