The β-catenin/TCF-4-LINC01278-miR-1258-Smad2/3 axis promotes hepatocellular carcinoma metastasis

Hepatocellular carcinoma (HCC) metastasis is largely responsible for HCC-associated recurrence and mortality. We aimed to identify metastasis-related long non-coding RNAs (lncRNAs) to understand the molecular mechanism of HCC metastasis. We first identified that miR-1258 was downregulated in HCC tissues both in The Cancer Genome Atlas (TCGA) and Sun Yat-sen University Cancer Center (SYSUCC) dataset. MiR-1258 expression negatively correlated with recurrence-free survival and overall survival of HCC patients. MiR-1258 overexpression inhibited migration and invasion of HCC cells both in vitro and in vivo, whereas miR-1258 downregulation promoted cell metastasis. Luciferase assays verified direct binding of miR-1258 to Smad2 and Smad3, thereby attenuating TGF-β/Smad signaling. We further established that lncRNA LINC01278 was a negative regulator of miR-1258. In vivo and in vitro assays demonstrated that LINC01278-mediated HCC metastasis was dependent on miR-1258 expression. Furthermore, miR-1258 downregulation in turn increased LINC01278 expression. We also observed that TCF-4 could bind to the LINC01278 promoter site. In addition, LINC01278 downregulation decreased migration and invasion of HCC cells induced by β-catenin and TGF-β1 both in vitro and in vivo. We uncovered a novel mechanism for β-catenin/TCF-4-LINC01278-miR-1258-Smad2/3 feedback loop activation in HCC metastasis, and the study indicated that LINC01278 could serve as a therapeutic target for HCC metastasis.


Introduction
Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide, with at least 250,000 new cases every year, nearly half of them in China [1]. Although the treatment of HCC, especially early stage HCC, has greatly improved the survival of HCC patients, one half of patients with early stage HCC will develop recurrence after surgery [2][3][4]. The formation of metastases is largely responsible for HCC-specific recurrence and mortality [3,5]. Therefore, it is important to understand the molecular mechanism in HCC metastasis in order to identify novel effective therapeutic targets.
Non-coding RNAs (ncRNAs) participate in various pathophysiological activities, especially in tumor development and progression [6]. Although the functions and mechanisms of microRNAs (miRNA) have been widely elaborated, the role of long non-coding RNAs (lncRNAs) in human cancers remains largely unknown. Genes can be positively or negatively regulated by lncRNAs through a diversity of mechanisms, and identification of cancerrelated lncRNAs remains challenging [7]. Competitive endogenous RNA (ceRNA) hypothesis is one of the many hypotheses to explain the mechanism of lncRNAs [8]. This hypothesis proposed that lncRNAs competitively bind with miRNAs and harbor miRNA-response elements, thereby relieving the inhibitory effects of miRNA on the target mRNAs [8]. The "lncRNA-miRNA-mRNA" network has been validated in many human cancers [9,10].
The function of these networks in HCC metastasis remains poorly understood. The activation of TGF-β/Smad and Wnt/β-catenin signaling plays an important role in metastasis of early stage HCC [11,12]. It has been reported that phospho-Smad2/3 could stabilize β-catenin in the cytoplasm and facilitate translocation of β-catenin into the nucleus [13,14]. However, very few studies documented that lncRNAs mediated the cross talk between the TGF-β1/Smad pathway and the Wnt/β-catenin signaling pathway [15].

MiR-1258 expression negatively correlates with survival of HCC patients
We identified 29 differentially expressed miRNAs between 49 normal tissues and 369 HCC tissues in the TCGA dataset using the significant analysis of microarray (SAM) algorithm, including 25 upregulated miRNAs and 4 downregulated miRNAs (Table 1). MiRNA-1258 was the top downregulated miRNA (fold change = 0.32, P < 0.001).
Our ROC analysis showed that the optimal cutoff of miR-1258 expression in TCGA dataset was 1.226 (Fig. S1) and was used categorized the study patients into low versus high miR-1258 expression group. The clinicopathologic characteristics of 117 patients from the SYSUCC dataset stratified by low versus high expression of miR-1258 and LINC01278 are summarized in Table S1. We next sought to validate the differentially expressed miRNAs in 20 paired early stage HCC and adjacent normal tissues from the SYSUCC dataset. We found that HCC tissues had significantly lower levels of miR-1258, miR-1248, miR-369, and miR-7704 than paired adjacent normal tissues ( Fig. 1a and S2).
We next analyzed correlation between miR-1258 expression and recurrence-free survival (RFS) and overall survival (OS) of HCC patients in the TCGA dataset. Kaplan-Meier analyses showed that HCC patients with high miR-1258 expression had significantly better median RFS and OS than those with low miR-1258 expression (P < 0.001) (Fig. 1b). Significant difference in RFS (P = 0.002) and OS (P < 0.001) was also observed between the low and high miR-1258 expression group in the SYSUCC dataset (Fig. 1c).

MiR-1258 inhibits in vitro and in vivo migration of HCC cells
We further investigated the effects of miR-1258 on the migration of HCC cells in vitro and in vivo. miR-1258 expression was significantly lower in four HCC cell lines  (Fig. 1d). We infected SMMC-7721 and Hep3B cells with lentiviruses overexpressing miR-1258 or antisense miR-1258 (Fig. S3). The wound scratch assays and transwell invasion assays demonstrated that ectopically expressed miR-1258 significantly suppressed, whereas miR-1258 downregulation marked enhanced the migration and invasion of HCC cells (Fig. 1e, f). Furthermore, mice bearing SMMC-7721 xenografts overexpressing miR-1258 had a significantly lower number of metastatic lung nodules than the control mice. Meanwhile, mice bearing Hep3B xenografts expressing antisense miR- 1258 had a significantly higher number of metastatic lung nodules than the control mice ( Fig. 1g and S4).

MiR-1258 is regulated by LINC01278
LncRNAs have been increasingly shown to correlate with miRNA expression regulation. Therefore, the bioinformatic tool LncBase v.2 was used to identify candidate lncRNAs capable of regulating miR-1258 (Table S2). Pearson correlation analysis of the TCGA dataset demonstrated a significant negative correlation between miR-1258 expression and LINC01278 among the top ten predicted lncRNAs ( Fig. 3a), which was also observed in 117 early stage HCC samples from the SYSUCC dataset (Fig. 3b). Moreover, LINC01278 significantly reduced while shRNAs against LINC01278 markedly elevated miR-1278 levels ( Fig. 3c). Subcellular fractionation analysis revealed that LINC01278 was mainly cytoplasmic both in HCC cells and HCC tissues (Fig. 3d, e). Our dual-luciferase gene reporter assays further showed that co-transfection of LINC01278-wt and miR-1258 significantly decreased luciferase activities while cotransfection of LINC01278-mut and miR-1258 had no effect on luciferase activities compared with the control group (Fig. 3f).
RNA immunoprecipitations were undertaken to understand the mechanism whereby LINC01278 regulates miR-1258. In the control group, the amount of LINC01278 and miR-1258 that immunoprecipitated with Ago2 was higher than the IgG group. In the anti-miR-1258 group, the amount of LINC01278 and miR-1258 that immunoprecipitated with Ago2 was significantly decreased compared with the control group (Fig. 3g). The results suggested that LINC01278 regulated miR-1258 expression by mediating with RNAinduced silencing complex (RISC). Notably, the amount of LINC01278 was higher than the IgG group in the anti-miR-1258 group, implying that miR-1258 may have an inhibitory feedback in LINC01278 expression (Fig. 3g).
We further investigated the effect of LINC01278 on the expression of Smad2 and Smad3, which are the targets of miR-1258. We observed that LINC01278 overexpression significantly increased while shRNAs against LINC01278 markedly depressed the protein and mRNA levels of Smad2 and Smad3 in HCC cells (Fig. 4e, f). MiR-1258 upregulation in HCC cells overexpressing LINC01278 markedly reduced while miR-1258 downregulation significantly increased the protein and mRNA transcript levels of Smad2 and Smad3 (Fig. 4e, f). In HCC tissues metastasized to the lungs, LINC01278 overexpression increased the protein levels of phospho-Smad2, Smad2, phospho-Smad3, Smad3, PAI-1, and vimentin and reduced the levels of E-cadherin (Fig. 4g). Meanwhile, shRNAs against LINC1278 decreased the protein levels of phospho-Smad2, Smad2, phospho-Smad3, Smad3, PAI-1, and vimentin and increased the levels of E-cadherin. Furthermore, in the HCC tissues metastasized to the lungs, LINC01278 overexpression significantly increased while shRNAs against LINC1278 markedly reduced the number of metastatic lung nodules ( Fig. 4h and S6). (LINC01278-mut) was used for luciferase reporter assay. The relative luciferase activities were detected in 293 T cells transfected by LINC01278-WT and LINC01278-mut. g RNA-IP was used to identify the inhibition of miR-1258 by LINC01278. The expression levels of LINC01278 and miR-1258 were detected using qRT-PCR. LINC01278, ectopic LINC01278 expression in HCC cells. shLINC01278, HCC cells were transfected by shRNA target LINC01278. *P < 0.05.

LINC01278 is a mediator between the Wnt/β-catenin and TGF-β/Smad signaling pathways
The RNA immunoprecipitation results prompted us to speculate that miR-1258 may have an inhibitory feedback on LINC01278 expression. Our RT-PCR assays showed that ectopically expressed miR-1258 significantly depressed LINC01278 expression (Fig. 5a). Analysis of the promoter sequence of LINC01278 using the online PROMO algorithm identified a putative TCF-4/LEF-1 binding site (Fig. 5b). Luciferase reporter assays further revealed that HCC cells co-transfected with LINC01278-wt and TCF-4 showed significantly increased luciferase activities relative to the control group, whereas HCC cells co-transfected with TCF-4 and LINC01278 with the binding sequence mutated showed no significant change in luciferase activities versus the control group (Fig. 5c). The mRNA levels of LINC01278 were significantly decreased by suppressing the expression of β-catenin, TCF-4, and Smad2/3 (Fig. S7). Chromatin immunoprecipitation (ChIP) assays further showed that TCF-4 bound to the LINC01278 promoter. Moreover, TCF-4 binding to LINC01278 promoter increased upon β-catenin overexpression (Fig. 5d). In addition, β-catenin overexpression caused a significant increase in LINC01278 expression in HCC cells, which, however, was aborted by co-transfection with siTCF-4 (Fig. 5e). Furthermore, TGF-β1 (0.5 ng/ml) upregulated while co-transfection with siTCF-4 diminished LINC01278 expression (Fig. 5f). The expression of signaling molecules of β-catenin/TCF-4-LINC01278-miR-1258-Smad2/3 feedback loop was listed in Fig. S7.
We further evaluate whether LINC01278 could be as a therapeutic target for metastatic HCC. Wound scratch assays and transwell assays showed that ectopically expressed β-catenin promoted the migration and invasion of HCC cells, which was aborted by shRNAs against LINC01278 (Fig. 5g, h). In addition, TGF-β1 (0.5 μg/week) significantly increased the number of metastatic lung nodules while TGF-β1 plus shLINC01278 reduce the number of metastatic lung nodules (Fig. 5i, and S8).
To identify HCC metastasis-related lncRNAs, we first explored the TCGA dataset to identify differentially expressed miRNAs. Investigation of E-HCC's metastasis is important in providing useful insight into treatment strategy, and we selected an E-HCC database from SYSUCC for validation. We found that miR-1258 was one of the most significantly decreased miRNAs both in the TCGA dataset and the SYSUCC dataset. Furthermore, miR-1258 was associated with HCC prognosis in both cohorts. In vitro assays demonstrated that miR-1258 expression negatively correlated with metastasis of HCC cells. MiR-1258 regulates cell cycle progression and inhibits cell proliferation, and suppresses tumor metastasis in many cancer types [16][17][18]. In this study, we screened the targets of miR-1258 by online tools. We noticed that both Smad2 and Smad3 were the putative targets of miR-1258. Luciferase reporter assays demonstrated miR-1258 binding to Smad2-3′-UTR and Smad3-3′-UTR, thereby inhibiting Samd2/3 expression. In the classic Smad pathway, TGF-β signaling activates Smad2 and Smad3 through a tetramer complex of type I and type II receptors and binds to Smad4 [19]. The Smad complex translocate to the nucleus to mediate inhibition or activation of target genes, thereby promoting epithelial-mesenchymal transition [20,21]. MiR-1258 specifically targets Smad2 and Smad3, block the TGF-β/Smad signaling pathway, and inhibit metastasis of HCC cells.
The "ceRNA" hypothesis speculates that lncRNA could regulate mRNA expression by sponging miRNAs [22]. We used lncRNA predictive online tools to screen putative lncRNAs which could affect miR-1258 function. We identified LINC01278, located on chromosome Xq11.1, as a novel lncRNA that could directly target miR-1258. LINC01278 was downregulated in papillary thyroid carcinoma compared with normal tissues and acted as a tumor suppressor in papillary thyroid carcinom progression [23]. However, in this study, we found that the expression of LINC01278 was significantly increased in HCC tissues compared with adjacent normal tissues in the TCGA dataset and the SYSUCC dataset. Moreover, high LINC01278 shLINC01278, LINC01278+miR-1258, and shLINC01278+anti-miRNA-1258. g The protein levels of P-Smad2, Smad2, P-Smad3, Smad3, PAI-1, E-cadherin, and vimentin in metastasis tumor tissues. LINC01278, ectopic LINC01278 expression in HCC cells. h The lung metastasis of HCC cells transfected by LINC01278 and shLINC01278. shLINC01278, HCC cells were transfected by shRNA target LINC01278. miR-1258, HCC cells were transfected by miR-1258 expression vector. Anti-miR-1258, HCC cells were transfected by miR-1258 antisense plasmid. *P < 0.05. expression was associated with worse prognosis in HCC. Both in vitro and in vivo assays demonstrated that LINC01278 promoted HCC cell metastasis dependent on miR-1258 expression. The results suggest that LINC01278 may play different roles in different tumors. Previous studies demonstrated that RISC, which is the key molecular for siRNA or miRNA-mediated gene silencing, participates in the "ceRNA" network, and Ago2 is the core component of RISC [24]. We conducted RNA immunoprecipitations to investigate the underlying mechanism whereby LINC01278 targets miR-1258. We found that the levels of LINC01278 and miR-1258 immunoprecipitated with Ago2 increased. MiR-1258 upregulation significantly increased the levels of LINC01278 immunoprecipitated with Ago2. The results suggested that LINC01278 could regulate miR-1258 by RISC complex, and other miRNAs could also bind to LINC01278.
We noticed that the levels of miR-1258 could reciprocally regulate the expression of LINC01278. We first speculated that the promoter of LINC01278 may have a Smad responsive element (SRE) [25]. We used the online PROMO algorithm to analyze putative factor binding sites. However, no SRE was present in the LINC01278 promoter. According to the online tool, transcription factors TCF-4 and LEF-1 bind to the promoter of LINC01278 which was confirmed by luciferase reporter assays and ChIP assays. Our quantitative RT-PCR analysis further showed that both β-catenin and TGF-β1 induced LINC01278 expression. LINC01278 downregulation could inhibit the HCC metastasis induced by β-catenin and TGF-β1. Many studies have shown that phospho-Smad2/3 complex could stabilize β-catenin in the cytoplasm, and promote translocation of β-catenin into the nucleus [13,26]. In this study, we found that the "β-catenin/TCF-4-LINC01278-miR-1258-Smad2/ 3" axis could in turn induce the expression of LINC01278, forming a positive feedback loop. It should be noticed that the amount of TGF-β1 secreted by hepatic stellate cells was usually elevated under long-term chronic inflammation, such as alcohol or hepatitis virus [27]. The TCGA network also showed that Wnt signaling is largely altered in HCC [28]. Therefore, LINC01278, which is an important crossmediator between TGF-β/Smad signaling and Wnt signaling, could serve as a new therapeutic target to inhibit HCC metastasis.
In conclusion, our results showed that the "LINC01278-miR-1258-Samd2/3" axis promotes HCC metastasis. The axis forms a positive feedback loop and increases the stability of TGF-β signaling and Wnt/β-catenin signaling, thereby promoting HCC metastasis. LINC01278 could serve as a new therapeutic target for HCC metastasis.

Cells
Four HCC cell lines (SMMC-7721, Huh-7, HepG2, and Hep3B), and immortalized human liver cells (THLE-3) (China Center for Type Culture Collection, Wuhan, China; American Type Culture Collection, Manassas, VA, USA) were grown in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, USA) with 10% newborn calf serum at 37°C in a humidified atmosphere containing 5% CO 2 . All cell lines were authenticated based on short tandem repeat fingerprinting.

Acquisition of tissue specimens
Twenty paired HCC tissue and adjacent normal tissue specimens archived at Sun Yat-sen University Cancer Center (SYSUCC) between June 1, 2017 and July 30, 2018 were obtained. In addition, 117 early stage HCC (defined as diameters < 5 cm) samples surgically resected at SYSUCC between January 1, 2008 and December 30, 2013 were acquired. Major exclusion criteria were preoperative ablation, radiotherapy, transarterial chemoembolization, or liver transplant. All eligible patients had definitive diagnosis and follow-up data. Informed content was obtained from all subjects. The study protocol was approved by Institutional Review Board of SYSUCC.

Wound scratch assays and transwell assay
For wound scratch assays, cells were cultured into a confluent monolayer. A P200 pipette tip was used to introduce a scratch, and the fraction of cell coverage was measured at 48 h. For invasion assays, 1 × 10 4 cells were seeded into the upper Matrigel invasion chamber (BD Biosciences, NJ, USA) and FBS was added into the lower chamber. The invading cells were counted on the lower side of the chambers at 24 h in six randomly chosen microscopic fields per filter at ×200 magnification using an IX71 inverted microscope (Olympus Corp.). The assays were performed at least three times independently in triplicate.

Cell fractionation assays
Cell fractionation assays were performed using the PARIS Kit (Life Technologies, Carlsbad, CA, USA) as instructed by the manufacturer. Actin was used as the positive marker for the cytoplasmic fraction and small nuclear RNA U2 was used as a marker for the nucleus.

Quantitative real-time PCR
Total cellular RNA was extracted using TRIzol reagents (Invitrogen) and cDNA synthesis was done using the Pri-meScript RT reagent Kit (Promega, Madison, WI, USA) as instructed by the manufacturer. Real-time PCR was done on an ABI 7900HT Fast Real-time PCR system (Applied Biosystems, Foster City, CA, USA). MiR-1258 expression was examined using Bulge-Loop TM miRNA kit (RuiBo, Guangzhou, China) with U6 as a control. All the primers were purchased from Gene Copoeia Co. (Guangzhou, China). The sequence of primers is shown in Table S3.

RNA immunoprecipitation assays
RNA immunoprecipitation assays were performed using EZ-Magna RIP RNA-binding protein immunoprecipitation kit (Millipore) following the manufacturer's instructions. The cellular lysates were incubated with magnetic beads conjugated with human anti-Argonaute2 (Ago2) antibody (Millipore). Mouse IgG (Millipore) served as the negative control. The immunoprecipitated RNA was isolated using proteinase K buffer and the target RNAs were analyzed by quantitative RT-PCR as detailed elsewhere [33].

ChIP assays
The nuclear extract was obtained using the Nuclear Extraction Kit (Active Motif, California, USA) as instructed by the manufacturer. Then, immunoprecipitation of the nuclear lysates was performed using a commercially available ChIP assay kit (Abcam) according to the manufacturer's protocol. Rabbit anti-TCF-4 antibody (1:50, Abcam) and nonspecific IgG (1:200, Sigma) were used for immunoprecipitation. The immunoprecipitated promoter fragment containing the TCF response element was amplified by PCR and visualized by agarose gel electrophoresis.

Luciferase reporter assays
We cloned the 3′UTR of Smad2 and Smad3 (Gene Copoeia) to the downstream of the luciferase gene in pGL3-basic vector (Promega), and then transfected the miR-1258mimic, miR-1258-mut, anti-miR-1258, and miR-control HCC cells with the modified vectors. The LINC01278 promoter and LINC01278-binding-nut promoter in which the TCF-4-binding site was mutated (Gene Copoeia) were cloned into pGL3-Basic vector (Promega), and the TCF-4 overexpression and control HCC cells were transfected with the modified vectors. The Renilla luciferase reporter pRL-TK was used as a control. LINC01278 gene and LINC01278-mut sequence (Gene, Copoeia) were cloned into pmirGLO vector (Promega). After 48 h, firefly and Renilla luciferase activities were measured using Dual-Luciferase Reporter Kit (Promega).

HCC xenograft assays
The animal study was approved by the Animal Care and Use Committee of SYSUCC and carried out in strict accordance with the established institutional guidelines and the USA NIH guidelines on the use of experimental animals. Totally 5 × 10 6 HCC cells were subcutaneously inoculated into the flanks of 4-week-male BALB/C-nu/nu athymic nude mice. After the subcutaneous tumors reached 1 mm 3 in volume, they were implanted in the left hepatic lobe of nude mice. On day 50 post hepatic implantation, mice were sacrificed and lung metastatic nodules were enumerated by consecutive tissue sections as described [34].

Statistical analysis
The miRNA expression profiles of HCC patients in the TCGA dataset were examined using the SAM algorithm and miRNA with a fold change ≥2 was considered differentially expressed. ROC analysis was used to determine optimal cutoff value of miR-1258 and LINC01278. Statistical analysis was performed using SPSS 17.0 software (SPSS, Chicago, IL, USA). RFS was calculated from the date of surgery to pathological confirmation of tumor recurrence of surgical. OS was calculated from the date of surgery to the date of death of any cause. Survival was censored at the date of the last follow-up visit. Survival curves were calculated by Kaplan-Meier method and compared by the log-rank test. Pearson correlation analysis was used to assess correlation between two variables. Chisquare test was used to assess correlation between miR-1258 and LINC01278 levels and clinicopathologic features. Data were presented as mean ± SD and assessed by Student's t test. P values < 0.05 were considered statistically significant.