Long Noncoding RNA FER1L4 Promotes Papillary Thyroid Cancer Progression by Targeting miR-612/CDH4 Axis

Background: Long noncoding RNAs (lncRNAs) have emerged as crucial regulators in various cancers. However, the functional roles of most lncRNA in papillary thyroid cancer (PTC) are not detailly understood. This study aims to investigate the biological functions and the molecular mechanism of lncRNA FER1L4 in PTC. Methods: The expression of FER1L4 in PTC was determined via operating RT-PCR assays. Meanwhile, the clinical signicance of FER1L4 in PTC patients was described. The biological functions of FER1L4 on PTC cells were evaluated by gain and loss of function experiments. Moreover, animal experiments were performed to reveal the effect on tumor growth. Subcellular distribution of FER1L4 was determined by uorescence in situ hybridization and subcellular localization assays. Luciferase reporter assay and RNA immunoprecipitation assay were applied to dene the relationship between FER1L4, miR-612, and CDH4. Results: Upregulated expression of FER1L4 in PTC tissues was correlated with higher lymph node metastasis rate (p=0.020), extrathyroidal extension (p=0.013), and advanced TNM stage (p=0.013). In addition, knockdown of FER1L4 suppressed PTC cell proliferation, migration and invasion, whereas ectopic expression of FER1L4 inversely promoted these processes. Mechanistically, FER1L4 could competitively bind with miR-612 to prevent the degradation of its target gene Cadherin 4 (CDH4). This condition was further conrmed in the rescue assays. Conclusions: This study rstly demonstrates FER1L4 plays an oncogenic role in PTC via FER1L4-miR-612-CDH4 axis and may provide a new therapeutic and diagnostic target for PTC.

hepatocellular carcinoma, lncRNA AY could promote hepatocellular carcinoma metastasis via induction of chromatin modi cation for ITGAV transcription (9). In addition, it was reported upregulated expression of lncRNA BCRT1 in breast cancer remarkably accelerates tumor growth and metastasis by acting as a sponge for miR-1303 to attenuate its repressive effect on PTBP3 and by mediating hypoxia-induced malignant properties of breast cancer cells (10). However, researches on the underlying role of lncRNA in PTC carcinogenesis are still lacking yet.
LncRNA performs its complex biological functions through multiple ways, which is closely related with its subcellular localization (11). For lncRNAs largely localized in the cytoplasm, they mainly function by acting as decoys for miRNAs, in which lncRNA could control miRNA availability for its target gene, so named as competing endogenous RNAs (ceRNA) either (12). In case of thyroid cancer, lncRNA-GAS8-AS1, located in the cytoplasm of PTC cells, was found that could promote autophagy of PTC cells by sponging oncogenic miR-187-3p and miR-1343-3p and upregulating the expression of ATG5 and ATG7 respectively (13). In this study, lncRNA fer-1 like family member 4 (FER1L4) is identi ed as an oncogene which promotes PTC cell proliferation, migration and invasion via functioning as ceRNA for tumor suppressor, miR-612.
Cadherin 4, also named as retinal cadherin, is a classical cadherin from the cadherin superfamily consisting of epithelial-cadherin (E-cadherin, CDH1), neural-cadherin (N-cadherin, CDH2), placentalcadherin (P-cadherin, CDH3), and so on (14). Cadherins are transmembrane glycoproteins responsible for cell-cell adhesion, tissue pattering and carcinogenesis (15). Previous researches indicate that cadherins could bind to catenin proteins and then form the cadherin-catenin complex to activate the catenin signaling or not, and involve in the epithelial-mesenchymal transition (EMT) process (16). However, the function of CDH4 in PTC remains unknown and its role in above process, however, is still controversial (17). In the present study, CDH4 is considered as the downstream target of miR-612, which mediates the promoting role of FER1L4 in PTC cells.

Materials And Methods
Patients and tissue samples Eighty thyroid papillary cancer tissues and adjacent normal tissues were collected from 80 patients with PTC who received operation on the First A liated Hospital of Nanjing Medical University (NMU). None of the patients underwent any other treatment but surgery. All collected tissue samples were immediately snap frozen in liquid nitrogen and stored at −80 °C until required. Our study was approved by the Ethics Committee of the First A liated Hospital of NMU.

Cell lines
Four PTC cell lines (K-1, TPC-1, B-CPAP and IHH-4) and a normal thyroid follicular epithelium cell line (Nthy-ori3-1) were purchased from the American Type Culture Collection (ATCC). K-1, B-CPAP and Nthy-ori3-1 cells were cultured in RPMI1640 medium (Gibco, USA), while TPC-1 cells were cultured in DMEM with high glucose (Gibco, USA). Mixture (1:1) of RPMI 1640 and DMEM was used to culture the IHH-4 cell line. 1% antibiotics (100 U/ml penicillin and 100 mg/ml streptomycin) and 10% fetal bovine serum (Gibco, USA) were added to all of the above culture media. All cell lines were incubated in a humidi ed atmosphere at 37 °C containing 5% CO2.
RNA extraction and quantitative real-time PCR analysis Total RNA was extracted from tissues and cultured cell lines using TRIzol reagent (Invitrogen, USA). A PrimeScript RT reagent kit (Takara, Japan) was used to synthesize cDNA. MiRNAs were reverse transcribed after polyadenylation using Revert Aid First Strand cDNA Synthesis Kit (Thermo Scienti c, MA, USA). Quantitative real-time PCR (qRT-PCR) was performed with AceQ qPCR SYBR Green Master Mix (Vazyme, China). Results were calculated using the 2 -ΔΔCT method and normalized to the expression of GAPDH for mRNA or U6 for miRNA. Primers used in the study were listed in Additional le 1 Table S1.
Cell transfection PTC cells were transfected with 50nm siRNAs and plasmid vectors when they grew to 30-40% density using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's recommendations. In this study, siRNAs, miR-612 mimics, miR612 inhibitor and correspondent negative control were purchased from Genepharma. After 48h transfection, the cells were harvested for performing the following experiments. Plasmid vectors encoding FER1L4 and biologically active short hairpin RNAs (shRNA) targeting FER1L4 or CDH4 were generated either (Genepharma, Shanghai, China). Stable cell lines were obtained by using 2μg/ml puromycin (Sigma-Aldrich, St-Louis, Missouri, USA) for about three weeks. Nucleotide sequences mentioned above were listed in Additional le 2 Table S2.

Cell proliferation assay
In order to investigate the effect on cell proliferation of corresponding treatment to K-1 and TPC-1 cells, cell counting kit 8 (CCK8) assay, colony formation assay and 5-Ethynyl-2′-deoxyuridine (EdU) incorporation assay were applied. Detailed information has been described before (18).

Animal experiment
Four-week female BALB/c nude mice were purchased from the Animal Center of NMU and all experiments were approved by Committee on the Ethics of Animal Experiments of the Nanjing Medical University. For the tumorigenicity studies, a total of 20 mice were randomly assigned, stable transfected cells (1×10 6 cells/100μl of PBS) were subcutaneously injected into the ank of the nude mice. The tumor volume was measured every 4 days and calculated by the formula: volume = (length × width2)/2.

Immunohistochemical (IHC) analysis
All specimens were xed in 4% formalin and then embedded in para n. After blocking endogenic peroxides and proteins, these sections were incubated with primary antibodies speci c for Ki-67 (Abcam) or CDH4 (Abclonal) at 4°C overnight, lastly were counterstained with hematoxylin after incubating with the secondary antibodies at 37 °C for 1 h. Random images were obtained using a light microscope. (Olympus Corp. Tokyo, Japan)

Cell migration and invasion assays
Transwell chambers (Corning, USA) coated with or without Matrigel (BD Bioscience, USA) were used to evaluate the function of genes on cell migrative and invasive ability. Meanwhile, cell motility also examined by wound healing assay. Detailed information has been described before (18).

Flow cytometric analysis
Treated cells were collected for Flow cytometric analysis. According to the protocol of the reagent (MultiSciences, China), APC-Annexin V and Propidium Iodide (PI) were used to stain cells, and the rate of apoptosis was analyzed by a ow cytometry (FACScan, BD Biosciences). For cell cycle analysis, treated cells was stained by PI-staining solution, and then the percentages of cells in G 0 -G 1 , S, and G 2 -M phase were counted.

Western blot assay and antibodies
Western blot assay was performed following previous protocol. The primary antibody used were listed in Additional le 3 Table S3.

Subcellular fractionation and Fluorescence in situ hybridization (FISH)
The separation and puri cation of cytoplasmic and nuclear was implemented using the PARIS Kit (Life Technologies, USA) according to the manufacturer's instructions. For the FISH assay, The Cy3-labeled FER1L4 probe used in our study was synthesized by (RiBo Ltd, Guangzhou, China). Brie y, the prepared cells were incubated with speci c probes at 37 °C overnight after xation and permeabilization. Finally, the nuclei were stained by DAPI and observed using confocal microscope.

Dual-luciferase reporter assay
The sequences of FER1L4 and CDH4 3′-UTR containing wild-type or mutated miR-612 binding sites were synthesized and loaded into a pGL3 luciferase reporter vector (Promega, USA). TPC-1 cells (1×10 5 ) were cotransfected with miRNA mimics or control, and the luciferase reporter vectors. After 48h of incubation, the luciferase activities were measured using a Dual-Luciferase Reporter Assay System (Promega, Wisconsin, USA). Relative luciferase activity was normalized to Renilla luciferase.

RNA immunoprecipitation (RIP) assay
For detecting whether FER1L4 participates in the RNA-induced silencing complex mediated by miRNA, RIP assays were conducted using the Magna RIPTM RNA-binding protein immunoprecipitation kit (Millipore, USA). Brie y, the prepared cells were lysed and incubated with anti-Ago2 (Abcam, CA, MA, USA) or IgG antibody at 4 °C overnight. Then, cell lysates were incubated with the protein A magnetic beads for 4h.
The coprecipitated RNAs were collected for qRT-PCR analysis.

Statistics analysis
The data are described by the mean ± standard deviation (SD) in three independently experiments. For comparing statistical differences, Student's t-tests, Pearson Chi-square test, Wilcoxon test were performed as appropriate using SPSS v22.0 and GraphPad Prism 6. Spearman's correlation analysis was selected to analyze the correlations among FER1L4, miR-612 and CDH4. P value less than 0.05 is considered statistically signi cant.

FER1L4 is upregulated in PTC and associates with advanced stage
To identify genes that have pivotal roles in PTC tumorigenesis, expression signature of genes was screened based on TCGA database. In this study, we observed preferential upregulation of lncRNA FER1L4 in TC tissues (n=513) compared with normal tissues (n=58; Fig. 1A-B). Moreover, FER1L4 also showed concordance in a pattern of expression in PTC cell lines compared with an immortalized thyroid follicular epithelium cell line Nthy-ori-1 (Fig. 1C). Importantly, identical results were also detected via measuring the expression in 80 paired PTC tissues and adjacent normal tissues, in which 79 percent of patients exhibited higher expression of FER1L4 in PTC tissues (Fig. 1D). Furthermore, through investigating the clinical signi cance of FER1L4 in 80 PTC patients, it showed enhanced expression of FER1L4 correlated with higher lymph node metastasis rate (p=0.020), extrathyroidal extension (p=0.013), and advanced TNM stage (p=0.013) (Additional le 4: Table S4). In addition, analysis of TCGA cohort suggested that upregulation of FER1L4 predicted poor histological type, deeper tumor invasion depth (T stage), higher lymph node metastasis rate (N stage) and advanced TNM stage either (Fig. 1E-H). Therefore, these results supported an oncogenic role for FER1L4 in PTC.

FER1L4 promotes PTC cell growth and motility both in vivo and in vitro
To elucidate the potential function of FER1L4, TPC-1 and K-1 cells were transfected with FER1L4 targeting small interfering RNA (siRNA) or control siRNA ( Fig. 2A). Firstly, CCK8 assays demonstrated that knockdown of FER1L4 suppressed the cell growth of TPC-1 and K-1 cells (Fig. 2B). Meanwhile, overexpression of FER1L4 showed enhanced proliferative capacity of cells transfected with plasmid vectors encoding FER1L4 (Fig. 2C-D). Additionally, colony-formation assays indicated that silencing FER1L4 signi cantly reduced the colony numbers (Fig. 2E), while upregulation of FER1L4 drastically enhanced them (Fig. 2F). Moreover, EdU incorporation assays showed an analogous mode as above, in which EdU positive cell numbers were progressively reduced following depletion of FER1L4 (Fig. 2G), whereas overexpression of FER1L4 resulted in an increase of these cells (Fig. 2H). Importantly, there is successfully identi es that decreased apoptosis and cell-cycle promotion are two factors that could contribute to the outgrowth of cancer cell. Herein, we performed ow cytometric assays to examine the role of FER1L4 in these properties. Notably, knockdown of FER1L4 showed a growing number of apoptotic cells in TPC-1 and K-1 cells treated with FER1L4 targeting siRNA ( Fig. 3A; Additional le 5: Fig  S1A). Apart from increased cell apoptosis, FER1L4-knockdown PTC cells had a signi cantly higher percentage of cells in the S phase with detriment to the G0/1 phase cells than did cells expressed a negative control RNA, which indicated the induction of cell arrest at S phase after silencing FER1L4 (Fig.  3B). Remarkably, Western blot analysis showed depletion of FER1L4 was accompanied with reduced expression of apoptotic marker like Bcl2, along with enhanced expression of pro-apoptosis protein like Bax. Meanwhile, the expression level of S-phase checkpoint proteins such as CyclinA2 and CDK2 were also decreased in FER1L4 silenced PTC cells ( Fig. 3C; Additional le 5: Fig S1B). In parallel, in order to investigate the function of FER1L4 in vivo, TPC-1 cells transduced with lentivirus vectors carrying FER1L4 shRNA or scrambled sequence were subcutaneously injected into the ank of 4 weeks female nude mice. It showed that depletion of FER1L4 decreased the tumor growth of TPC-1 cells (Fig. 3D), and the tumor size and nal tumor weight in implanted tumors were signi cantly decreased in knockdown group (Fig.  3E-F). Furthermore, immunohistochemistry analysis of orthotopically implanted tumors showed diminished numbers of Ki67-positive cells by inhibition of FER1L4 (Fig. 3G). Thus, these data indicated that FER1L4 has a critically promoting role in PTC cell growth.
The molecular mechanisms behind the capacity for tumor cells to metastasize e ciently is an important topic. Therefore, it is necessary to evaluate whether FER1L4 involves in PTC metastasis. First, wound healing assays showed that FER1L4-knockdown led to attenuated migration of PTC cells (Fig. 3H; Additional le 5: Fig S1C). Transwell assays further suggested that loss of FER1L4 resulted in reduced numbers of cells getting through the chamber (Fig. 3I), whereas ectopic expression of FER1L4 showed elevated cell numbers (Fig. 3J). Consequently, our results suggested that FR1L4 essentially involves in the early metastasis of PTC cells. miR-612 is sponged by FER1L4 in PTC Long non-coding RNA plays many roles in cellular physiology which is partially dependent on its subcellular location. In this study, we found FER1L4 mainly located in the cytoplasm of PTC cell by conducting in situ hybridization assay and subcellular fractionation assay, which suggested that FER1L4 may regulate the expression of targets at the post-transcriptional level (Fig. 4A-B; Additional le 5: Fig  S1D). Then, combining gene expression pro ling from TCGA database, along with prediction tools such as regRNA2.0 (19), miRcode (20) and starbase (21), we selected ve potential miRNAs (miR-612, miR-140-3p, miR-92a-3p, miR-196b-5p and miR-784-3p) with binding sites in FER1L4 sequences ( Fig. 4C; Additional le 6: Fig S2A). However, only miR-612, miR-140-3p and miR-784-3p could bind with FER1L4 transcripts, which we tested through conducting dual luciferase reporter assays (Fig. 4D). Meanwhile, knockdown of FER1L4 also showed upregulated expression of miR-612, miR-140-3p and miR-784-3p in TPC-1 cells (Fig. 4E). Nevertheless, considering the highest binding a nity between FER1L4 and miR-612, we exclusively chosen miR-612 for further study. As illustrated in Figure 4F, the binding sequence of miR-612 to FER1L4 were mutated and then fused into the luciferase reported vector (MUT type). Notably, mutation of the miR-612 binding sites in FER1L4 3'UTR abolished the suppressive effects to FER1L4driven luciferase activity by miR-612 mimics in PTC cells (Fig. 4F). In addition, overexpression of FER1L4 caused reduction of miR-612 expression either (Fig. 4G). RNA-binding protein immunoprecipitation (RIP) assay further revealed that FER1L4 was successfully immunoprecipitated from cell extracts using Ago2 antibody (Fig. 4H). On the other hand, a signi cant inverse correlation between FER1L4 and miR-612 was acquired through analyzing of 20 PTC tissue samples (R 2 =0.38, P=0.0036) (Additional le 6: Fig S2B). Taken together, these studies identi ed that FRE1L4 could sequester miR-612 away to its targets, therefore involves in PTC tumorigenesis. miR-612 is su cient and necessary for FER1L4-induced effects Although above data con rmed miR-612 was one of the targets of FER1L4, but the function of miR-612 in PTC has never been determined yet. First, the expression of miR-612 was downregulated in K-1 and TPC-1 cells compared with normal cell line Nthy-ori-1 (Fig. 4I). Next, in order to investigate whether miR-612 modulates the biological behavior of PTC cells, the expression of miR-612 was successfully attenuated by miR-612 inhibitor and arti cially enhanced by miR-612 mimics (Fig. 4J). Overexpression of miR-612 inhibited cell growth (Fig. 5A-B) and colony formation ( Fig. 5C; Additional le 6: Fig S2C), whereas inhibition of miR-612 showed opposite effect. Flow cytometric assays also indicated reintroduction of miR-612 into PTC cells induced cell apoptosis (Fig. 5D) and cell cycle arrest in S phase (Fig. 5E). Moreover, transwell assays also indicated ectopic expression of miR-612 progressively inhibited PTC cell migration and invasion ( Fig. 5F; Additional le 6: Fig S2D). We next sought to explore whether FER1L4 promotes PTC cell growth and motility via regulating miR-612. PTC cells were co-transfected with siRNA against FER1L4 and miR-612 inhibitor, and the resulting cells were subjected to CCK8 assay and transwell assay. It revealed the inhibitory effect on PTC cell by knockdown of FER1L4 was nulli ed by miRNA-612 inhibitor (Fig. 5G-J; Additional le 6: Fig S2E-F). As a result, these data unveiled the mechanism of promotive effects on PTC cells mediated by FER1L4 is to sponge miR-612.

FER1L4 upregulates CDH4 expression via inhibition of miR-612
The downstream target of miR-612 was predicted by miRwalk (22), and CDH4 was identi ed with potential miR-612 binding sites. Intriguingly, the RNA level of CDH4 was not in uenced by silencing FER1L4 or overexpression of miR-612 (Fig. 6A-B). However, the protein level of CDH4 was signi cantly diminished after decreasing FER1L4 expression or enhancing the expression of miR-612 ( Fig. 6C; Additional le 7: Fig S3A-B). To this end, it may suggest miR-612 could negatively regulate CDH4 expression through translational repression. For subsequent investigation, dual luciferase reporter assays were performed to validate whether miR-612 directly binds to CDH4 in PTC cells. As presented in Figure. 6D, miR-612 seed sequence in CDH4 3'UTR were mutated (MUT-CDH4), and then the mutated 3'UTR sequence and the wild type 3′UTR sequence of CDH4 (WT-CDH4) were fused into a plasmid vector respectively. The luciferase activity was signi cantly decreased following co-transfection of miR-612 mimics and WT-CDH4 reporter vector, which is blocked by MUT-CDH4 (Fig. 6D). Of note, diminished expression of CDH4 protein in FER1L4 silenced PTC cells was rescued by co-transfection with miR-612 inhibitor ( Fig. 6E; Additional le 7: Fig S3C). Moreover, a negative correlation between FER1L4 and CDH4 (R 2 =0.56, P < 0.001) was found in 20 paired PTC tissues (Additional le 8: Fig S4A). Thus, on the basis of above observation, CDH4 was con rmed as a target of miR-612 and therefore was regulated by FER1L4.

FER1L4/miR-612/CDH4 axis promotes PTC cell proliferation and invasion
It is clear that the expression of CDH4 was upregulated in PTC tissues and cell lines (Additional le 8: Fig  S4B-C). To delineate further the functional signi cance of CDH4 in PTC cell proliferation and invasion, the expression of CDH4 was e ciently knocked down in K-1 and TPC-1 cells (Fig. 6F; Additional le 8: Fig  S4D). Firstly, loss of CDH4 repressed cell proliferation as presented by EdU incorporation (Fig. 6G; Additional le 8: Fig S4E) and CCK8 assays (Fig. 6H). Furthermore, silencing CDH4 displayed a signi cant induction in apoptosis ( Fig. 6I; Additional le 8: Fig S4F) accompanied by elevated Bax expression and reduced Bcl2 expression, as well as cell cycle arrest at S phase ( Fig. 6J; Additional le 8: Fig S4G) with reduced proteins level of CDK2 and cyclinA2 (Fig. 6K and 7A). Importantly, subcutaneous tumor formation experiments revealed that knockdown of CDH4 signi cantly inhibited the formation of subcutaneous tumors of PTC cells (Fig. 7B; Additional le 8: Fig S4I). In addition, IHC assay also showed upregulated expression of CDH4 in PTC tissue samples (Fig. 7C). As expected, wound healing assays uncovered a reduction of cell migrative rate in CDH4 depleted cells (Fig. 7D). Consistently, transwell assays displayed a sharp decline in cell numbers getting through the chamber by knockdown of CDH4 (Fig. 7E). Together, these experiments support an oncogenic role for CDH4 in promoting PTC cell proliferation and invasion. Subsequently, we performed several cell functional experiments in CDH4 depleted cells with simultaneous knockdown of miR-612. First, co-transfected with miR-612 inhibitor in PTC cells resulted in a reversal of CDH4 siRNA-induced repression of cell growth (Fig. 7F). Then, we further examined the effect on cell metastatic ability. As indicated by transwell assays, CDH4 siRNAmediated decrease in cell numbers were nulli ed by arti cially lowered expression of miR-612 (Fig. 7G-I; Additional le 8: Fig S4J). In conclusion, these results suggested there may exist an axis between FER1L4, miR-612 and CDH4 which involves in the tumorigenesis and aggressiveness of papillary thyroid cancer.

Discussion
Thyroid cancer development involves multiple genetic and epigenetic alterations such as point mutation of the BRAF and RAS genes which seem to be linked to speci c etiologic factors like exposure to ionizing radiation and chemical mutagenesis (23). In recent decades, technical advances in genome editing and high-throughput sequencing has promoted the deeply research of non-coding genomes (24). Notably, long noncoding RNA, the main components of non-coding genes, is considered to be essentially implicated in the process of tumor initiation or progression including thyroid cancer (25). Fer-1 like family member 4 is a lncRNA with 6.7 kb length, located in chromosome 20 q11.22. However, intriguingly, the literature documents context-dependent phenotype of FER1L4 in human cancers. It was previously reported that FER1L4 serves as tumor suppressor in colon (26) and gastric cancer (27), whereas promotes the proliferation and cell cycle of glioma cells (28). In a human pan-cancer analysis, it suggested FER1L4 may act as an oncogenic driver in human cancers (29). Thus, importantly, the current study unravels a new facet of FER1L4 which accelerates thyroid papillary cancer proliferation and metastasis. In agreement, upregulated expression of FER1L4 in PTC tissues was closely related to advanced clinicopathological features like lymph node metastasis, extrathyroidal extension, and advanced TNM stage. Accordingly, FER1L4 plays an oncogenic role in PTC which could be new biomarkers and treatment target of PTC.
In our study, the con rmation that FER1L4 is located in cytoplasm give us reason to hypothesize that FER1L4 may in uence the progression of thyroid papillary carcinoma through the mechanism model of ceRNA. Subsequently, miR-612 was validated to directly bind to FER1L4 in PTC cells via the dual luciferase reporter assay. MiR-612 has been proved to suppress tumor growth and metastasis in many tumors. For example, miR-612 negatively controlled the formation of invadopodia, matrix degradation and metastasis of hepatocellular carcinoma by HADHA-mediated lipid reprogramming (30). Nevertheless, the function of miR-612 in thyroid cancer has not been described before. Herein, in this study, miR-612 was also successfully identi ed as tumor suppressor which inhibits cell proliferation, migration and invasion of PTC. Meanwhile, functional experiments also suggested FER1L4-induced promotion of PTC cell is mediated by the negatively regulating to miR-612. Hence, our ndings uncovered that FER1L4 exerts oncogenic behavior partly via sponging miR-612 in PTC cells.
Cadherin 4 was proved to be the target of miR-612 in this study. However, knockdown of FER1L4 or overexpression of miR-612 only in uences the expression of CDH4 at the protein level, which indicates miR-612 may repress the translation of CDH4 and then attenuates its expression (31), yet the detailed mechanism of such regulation demand further exploration. Deregulation of CDH4 has been implicated in several human cancers. Dependent on the cell context, in most epithelial cancers like breast (32), colorectal and gastric cancer (33), CDH4 is epigenetically silenced by promoter hypermethylation and commonly suggested to act as a tumor suppressor. Nevertheless, some researches also support an oncogenic function of CDH4 in high-grade gliomas (17) and osteosarcoma (34). Although there are discrepancies to the function of CDH4 in human cancer, our study identi es and suggests CDH4 as an oncogenic driver in PTC that promotes PTC cell proliferation, migration, invasion and cell cycle progression. Consistent with its function in cell-cell adhesion, immunohistochemical analysis indicated that CDH4 shows crisp membrane localization in PTC tissues. In particular, Gene Set Enrichment Analysis (GSEA) of CDH4 using gene expression pro le from TCGA database also invalidates that CDH4 involves in process of adherens junction and extracellular matrix-receptor interaction (Additional le 9: Fig S5A-B) (35). On the other hand, FER1L4 was also characterized that is positively related to gene set of cell adhesion molecules (Additional le 9: Fig S5C). However, person correlation analysis suggested FER1L4 expression positively correlates with the expression of P-cadherin (R 2 = 0.2513, P < 0.001) but not the others (E-cadherin, N-cadherin and CDH4) (Additional le 9: Fig S5E-H), which was consistent with the RT-PCR results (Fig. 6A). Intriguingly, the expression of CDH4 also positively correlated with P-cadherin using TCGA database (R 2 = 0.1156, P < 0.001; Additional le 9: Fig S5K), indicating that FER1L4 may regulate the expression of P-cadherin via governing CDH4 or vice versa (36). However, it needs further experimental con rmation. Meanwhile, to elucidate the mechanism of CDH4-induced PTC cell proliferation is an important topic either. It was reported that CDH4 is necessary to the activation of ERK and p27 (17) and also the activation of c-Jun (34). GSEA analysis also suggested that CDH4 positively correlates with the wnt/β-catenin pathway, which may explain the promotion to PTC cells of CDH4 either (Additional le 9: Fig S5D). Collectively, different expression level and function of FER1L4 and CDH4 in human cancers may be caused by the heterogeneity of tumors (37), which suggesting the great potential of FER1L4 and CDH4 as diagnostic biomarkers. Additionally, the underlying mechanism of CDH4 in PTC carcinogenesis requires thoroughly exploration.

Conclusions
Taken together, we identi ed that FER1L4 promotes PTC development via working as a sponge for miR-612 to attenuate its repressive effect on CDH4. These results provide a better understanding of the role of lncRNA in thyroid cancer progression, as well as a potential therapeutic and diagnostic target.  TNM stage (H) from TCGA database. Error bars, mean ± SD. *, P < 0.05; **, P < 0.01; ***, P<0.001.