Epigenetically silenced linc00261 contributes to the metastasis of hepatocellular carcinoma via inducing the deficiency of FOXA2 transcription

Abstract

Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide. In recent decades, long non-coding RNAs (lncRNAs) have attracted increasing attention and have been reported to play important roles in human cancers, making them ideal candidates for precise disease assessment and treatment. Our previous study found that the loss of linc00261 was significantly correlated with the malignant biological behaviors of HCC, particularly MVI, and serves as an excellent independent prognostic factor for recurrence-free survival. In this study, our in-depth research demonstrated that linc00261 inhibits epithelial-mesenchymal transition (EMT) in liver cancer cells, thereby suppressing migration, invasion, and the formation of lung metastatic lesions. Moreover, linc00261 and its neighbor gene FOXA2 were positively correlated in HCC, the gain- and loss-of-function analyses indicated that linc00261 transcriptionally promotes the expression of FOXA2. Additionally, bioinformatic analysis and rescue assays confirmed that linc00261 partially suppresses migration, invasion, and EMT by upregulating FOXA2 expression. Molecular mechanism studies showed that linc00261 transcriptionally upregulates FOXA2 in cis by recruiting SMAD3. Finally, we identified EZH2 is responsible for linc00261 transcription repression via modulating trimethylation of H3K27 at Lys27 (H3K27Me3), both EZH2 and H3K27Me3 were negatively correlated with linc00261 expression in HCC. In conclusion, these findings demonstrated a crucial role of linc00261 in HCC metastasis, and that EZH2/linc00261/FOXA2 axis might reveal potential prognostic factors and be applied as therapeutic targets for HCC metastasis.

Introduction

Primary liver cancer is the sixth most frequent cancer and fourth leading cause of cancer-related death worldwide; 75-85% of these cases are hepatocellular carcinoma (HCC) [1]. Although various therapeutic options are available, such as surgical resection, liver transplantation, ablation, transarterial chemoembolization, radiotherapy, and systemic treatments involving multidisciplinary approaches, tumor recurrence and metastasis remain the main factors resulting in poor prognosis [2]. A better understanding of the risk factors and molecular mechanisms underlying HCC progression is necessary to advance diagnostic and therapeutic inventions.

Long non-coding RNAs (lncRNAs) post-transcriptionally modulate gene expression by affecting mRNA stability, functioning as miRNA sponges, and regulating translation in the cytoplasm; however, they are more able to activate or repress gene transcription in cis or trans by interacting with various chromatin-modifying complexes in the nucleus [3-5]. Our previous study discovered a long intergenic noncoding RNA (lincRNA), linc00261, which is significantly associated with the presence of microvascular invasion (MVI) and dysregulated in HBV-related HCC; the loss of linc00261 in HCC tissues was correlated with worse post-operative recurrence-free survival (RFS) in HBV-related HCC and enhanced cellular motility and invasion in HCC cells [6], which possibly results from the inhibition of Notch signaling through the suppression of Notch-1 and Hes-1 expression [7]. However, the precise effect and underlying mechanism of linc00261 on HCC progression, as well as the reason of linc00261 downregulation in HCC, is not well-understood.

Accumulating studies have suggested that the loss of differentiation-associated transcription factors (such as GATA3, ELF5, and NKX2-1) and embryonic cell fate regulating transcription factors (such as SOX2, MYC, KLF4, FOXA2, and OCT4) induces stem cell-like properties, which are related to metastasis initiation [8]. Interestingly, these development-associated transcription factors are always preferentially surrounded by lincRNAs in vertebrates [9], whereas only approximately 3% of all human lincRNAs are positively correlated with their neighboring genes [10]. Therefore, it is rare that development-associated lincRNAs influence cancer metastasis by interacting with neighboring genes encoding development-associated transcription factors. Encoding one of the pioneer transcriptional factors in liver specification, FOXA2 was found to be located upstream of linc00261 and induced by linc00261 during endoderm differentiation [11,12]. In lung adenocarcinoma, FOXA2 also inversely induce linc00261 expression, their expression are together triggered by DNMT1 through CpG island methylation of the entire lnc00261-FOXA2 locus [13]. Moreover, FOXA2 and its targets are central modulators of the sexual dimorphism of HCC, indicating that FOXA2 is tightly coupled with HCC initiation or progression [14]. Besides, FOXA2 was recently reported to be involved in regulating AFP production both in AFP-producing gastric adenocarcinomas and HCC cells [15]. Some studies have also demonstrated crucial suppressive roles for FOXA2 in HCC metastasis [16,17]. Taken these together, linc00261/FOXA2 axis is supposed to regulate HCC progression, yet it has not been systemically explored.

EZH2 (Enhancer of zeste 2) is the key subunit of PRC2 (ploycomb repressive complex 2), which catalyzes trimethylation of H3K27 at Lys27 (H3K27Me3) via its SET domain, and then triggers the recruitment of PRC1 complex to the promoter region of target genes, ultimately results in the compression of chromosomes and silencing of target genes [18]. It also causes transcriptional inhibition through a variety of histone independent mechanisms. For example, EZH2 directly recruits DNMT1 and DNMT3A/B to the promoter region of the target gene, causes methylation of the promoter CpG island, thus interferes the binding of transcriptional factors to the promoter and regulating the transcription of target genes [19]. EZH2 is commonly overexpressed in a number of caners, and correlated with tumor aggressiveness and poor prognosis, including HCC [2,3]. It induces gene silencing of many tumor suppressors, such as miRNAs (miR-101, miR-622, miR-34a) [20,21] and lncRNAs (SPRY4-IT1, lncRNA-SVUGP2) [22,23]. However, little is known about the effect of EZH2 on the epigenetic regulation of linc00261 in HCC.

In this study, we further discovered that the loss of linc00261, partially suppressed by EZH2 mediated H3K27Me3, might be a key risk factor for the generation of metastasis-initiating cells. Subsequent exploration of the potential mechanism underlying linc00261-associated HCC metastasis showed that it transcriptionally upregulates the expression of its neighbor gene, a known tumor suppressor, FOXA2, by interacting with the transcriptional factor SMAD3, thereby restraining HCC metastasis.

Materials and methods

Patients and tissue specimens

Patients with hepatitis B virus-related HCC who underwent radical resection between November 2010 and November 2016 in Nanfang Hospital, Southern Medical University were enrolled in this study. HCC and adjacent noncancerous tissues (exceeding the outer edge of tumor at least 2 cm) were obtained immediately after resection. The specimens were stored in liquid nitrogen until subjected for real-time quantitative PCR (RT-qPCR) analysis or were paraformaldehyde-fixed and paraffin-embedded. All patients included in this study satisfied the following inclusion criteria: (a) treatment-naïve before surgery; (b) diagnosed pathologically and serologically as having hepatitis B virus-related HCC, and without hepatitis C virus infection; (c) the complete clinical parameters and follow-up data were well-documented. The Ethical Committee of Nanfang Hospital, Southern Medical University approved the usage of tissues from patients with HCC for retrospectively analysis at January 15^th, 2018 (NFEC-2018-004). As required by the Declaration of Helsinki (6th revision, 2008), written consent was signed by each patient.

For cohort 1 (n=100), fresh HCC and adjacent noncancerous tissues were used for linc00261 relative expression determination by RT-qPCR. For cohort 2 (n=79), paraformaldehyde-fixed and paraffin-embedded HCC and adjacent noncancerous tissues were subjected to immunohistochemical staining. Overlapping cases (n=44) between cohorts 1 and 2 were evaluated by both assays.

Patient follow-up

After discharge from the hospital, the patients were further monitored for survival analysis via regular outpatient follow-up. Relapse was diagnosed based on increased post-operative serum alpha-fetoprotein levels and the presence of space occupying lesion in or out of the liver, using at least one imaging examination (ultrasonic examination, computed tomography scan, or magnetic resonance imaging). Follow-up began on the date of surgery and terminated on August 31, 2017.

Hematoxylin-eosin (HE) staining and immunohistochemical staining

Paraffin-processed sections with 3 μm thickness were prepared and mounted onto polylysine-coated slides, and then dewaxed in xylene and rehydrated in gradient alcohol. For HE staining, the slides were subsequently stained with hematoxylin and eosin according to standard protocols. For immunohistochemical staining, the antigens were firstly retrieved in citric acid buffer (PH6.0) by microwave antigen retrieval. After that, they were immersed in 3% H₂O₂, blocked with 5% BSA solution, and incubated with primary rabbit polyclonal antibodies (Supplementary Table 5) overnight at 4°C. The following day, the sections were incubated with horseradish peroxidase-conjugated goat-anti-rabbit secondary antibody (ZSGB-BIO, Beijing, China), and developed with peroxidase substrate diaminobenzidine (DAB; ZSGB-BIO). The pathological histology and expression of target proteins were observed under an upright microscope (Olympus, Tokyo, Japan). The expressions of target proteins were evaluated semi-quantitatively by multiplying the scores of staining intensity and positive rate as we previously reported [24].

Immunofluorescence staining

Cells at logarithmic phase were planked on coverslips. After adherence, cells were fixed in 4% paraformaldehyde for 20 min, permeabilized with 0.3% Triton X-100 diluted with PBS buffer for a specified interval set (without this step for membrane proteins, 10 min for cytosolic proteins and 30 min for nuclear proteins), and blocked with 3% BSA in PBS buffer for 1 hour at room temperature. Then, cells were incubated with primary antibodies (Supplementary Table 5) overnight at 4°C, and with Alexa fluor 594-conjugated goat-anti-rabbit secondary antibody (Proteintech) for 1 hour at room temperature in the dark. In the end, nuclei were stained using 4’,6-diamidino-2-phenylindole (DAPI; Solarbio, Beijing, China) and images were captured under an inverted fluorescence microscope (Olympus).

FISH

The expression and localization of linc00261 in tissues and cell lines were determined as the instructions obtained from Exonbio (Guangzhou, China) who designed and synthesized linc00261-specific probes (targeting 3409-4497nt of linc00261 and labeled with digoxin). Briefly, the paraffinized sections were dewaxed with xylene and rehydrated in gradient alcohols, digested with 0.25% pepsin solution for 30 min at 37°C. For cell lines, the coverslips were incubated with 4% paraformaldehyde and permeabilized with 0.3% Triton X-100 for 20 min. Then, the sections/cell slides were prehybridized with hybridization solution (Exonbio) for 2 hours at 55°C followed with refixation with 4% paraformaldehyde. In the meanwhile, the probes diluted with hybridization solution (1:100) were denatured for 5 min at 85 ± 2°C and kept at 37°C for use in a temperature-controlled water bath. Next, the sections/cell slides were incubated with denatured probes overnight at 37°C, blocked with 3% BSA, and incubated with rhodamine-conjugated IgG fraction monoclonal mouse anti-digoxin (Exonbio) for 1 hour at 37°C in the dark. Nuclei were stained using DAPI (Solarbio) and images were captured under an inverted fluorescence.

RNA pull-down assay and mass spectrometry (MS) analysis

RNA pull-down assay was conducted using Pierce TM Magnetic RNA-Protein pull-down kit (20164; Thermo Scientific, MA, USA). Briefly, linc00261 was firstly transcribed in vitro and 3’-end labeled with a single desthiobiotinylated cytidine bisphosphate using T4 RNA ligase. Then, the labeled RNA was captured by magnetic beads and incubated with MHCCLM3 cell lysates. The same amount of 3’untranslated-region of androgen receptor RNA was used as the negative control, the sequence was as follow: 5’-CUGGGCUUUUUUUUUCUCUUUCUCUCCUUUCUUUUUCUUCUUCCCUCCCUA-3’. After washing and elution, the RNA-binding proteins was separated by 10% SDS-PAGE and visualized using silver staining. Ultimately, the proteins were identified with MS and western blotting analyses. For evaluation of the MS results, samples from 3 replications were merged and subjected for MS detection. Each peptide identified has a score, pep_expect, If the pep_expect score < 0.05, it was considered that this peptide matches with its target protein. The ultimate results of linc00261 potentially binding proteins refers to Supplementary File 1. For verifivation of the binding between linc00261 and SMAD3 (Abcam, Cambridge, England), the proteins retrieved in the RNA pull-down assay were subjected to western blotting for 3 times.

RNA immunoprecipitation (RIP)

MHCCLM3 cells were subjected to RIP assay according to the protocol of using RNA Binding Protein Immunoprecipitation Kit (BersinBio, Guangzhou, China). Briefly, cells were lysed in polysome lysis buffer containing protease inhibitor and RNase inhibitor. After removal of DNA, the lysates were divided into three groups (IP, IgG and Input groups), and the former two were respectively immunoprecipitated with equal amount of anti-SMAD3 rabbit antibody and control IgG at 4°C overnight, and incubated with prepared protein A/G beads at 4°C for 1 hour. Then, RNA was extracted from the protein-RNA complexes and applied to the subsequent RT-qPCR. U1 snRNP and GAPDH were used as control, and the primers were the same as above. The experiment was repeated for 3 times.

Chromatin immunoprecipitation (CHIP)

To evaluate the binding of SMAD3 to the promoter region of FOXA2, and EZH2, H3K27Me3 to the promoter region of linc00261, CHIP assay using SimpleChIP® Enzymatic Chromatin IP Kit (Agarose Beads; Cell signaling technology, MA, USA; CST9002) was performed with MHCCLM3 and HepG2 cell lines, respectively, according to the manufacturer’s instruction for 3 times. Control rabbit IgG and antibodies against SMAD3 (Abcam), EZH2 (Cell signaling Technology) and H3K27Me3 (Cell signaling Technology) were presented in Supplementary Table 5. The purified DNA was further used for RT-qPCR with a subset of primers targeting FOXA2 and linc00261 promoters (Supplementary Table 6).

The other materials and methods, such as cell lines and culture conditions, cell proliferation assay, plate colony-forming assay, Transwell migration and invasion assays, RNA extraction and RT-qPCR, Western blotting, subcellular fractionation and quantification of RNAs, establishment of transient knockdown models, lentivirus construction and transfection, and in situ tumor model, refers to the Supplementary Materials and Methods.

Statistical analysis

All results were presented as the mean ± standard error of mean and analyzed using SPSS Statistics 20.0 software (SPSS, Inc., Chicago, IL, USA). The enumeration data were analyzed by χ²-test. The measurement data obtained from two groups were compared by Student’s t test or Mann-Whitney U test, and analysis of variance followed with least significant difference-t test or Dunnett’s T3 test were used to compare multiple groups. Pearson correlation was used to analyze the correlation coefficient between two groups. Survival data were assessed by the Kaplan-Meier method and log-rank test was used to evaluate the difference. Univariate and multivariate analyses were based on the Cox proportional hazards regression model. P < 0.05 was considered as statistically significant.

Results

Downregulation of linc00261 is associated with HCC progression as an independent risk factor

To study the effect of linc00261, we consulted online databases to examine its expression in various adult normal tissues and cancerous tissues. First, the NCBI gene database revealed that linc00261 was annotated as specifically-expressed in endoderm-derived adult tissues/organs, with the liver showing the highest level compared to that in other organs or tissues (Supplementary Figure 1A). Then, we determined that patients with lower linc00261 levels in cancerous tissues had poorer overall survival in colon adenocarcinoma, HCC, lung adenocarcinoma, pancreatic adenocarcinoma, and uterine corpus endometrial carcinoma (Supplementary Figure 1B, 1C) as shown in the sequencing data of Oncolnc (http://www.oncolnc.org/) [25]. Our previous preliminary study suggested that linc00261 can serve as a prognostic biomarker to predict postoperative outcomes in patients with HCC [6]. To further explore its expression and clinical significance, we measured its expression in another cohort with HCC. The results showed that linc00261 was downregulated in 32.0% of patients (Figure 1A). Further, Chi-square and Mann-Whitney U tests revealed that patients with tumor sizes ≥ 5 cm or with MVI had lower linc00261 expression (Supplementary Table 1 and Figure 1B-E) and that patients with lower relative linc00261 expression had a shorter RFS than those with higher expression (median ± standard deviation, 7.89 ± 2.59 vs. 31.34 ± 13.40 months; Figure 1F). Moreover, univariate and multivariate Cox regression analyses indicated that tumor number (95% confidence interval [CI], 1.350-5.436; P=0.005), MVI (95% CI, 1.065-4.595; P=0.033), differentiation (95% CI, 0.306-0.824; P=0.006), and relative linc00261 expression (95% CI, 0.294-0.978; P=0.042) were independent prognostic factors of RFS in patients with HCC (Supplementary Table 2).

Figure 1.

Linc00261 expression in patients with HCC and the relationship with clinicopathological parameters and survival. (A) Linc00261 was down-regulated in 32.0% (32/100) patients with HCC determined by RT-qPCR; (B-E) Linc00261 expression was compared in patients with different pre-operative AFP levels (B), tumor size (C), MVI status (D) and TNM staging (E) determined by RT-qPCR and Mann-Whitney U test; (F) Recurrence-free survival was analyzed by Kaplan-Meier method in HCC patients with different relative linc00261 expression (n=91); (G) Linc00261 expression comparison between normal liver cells and MVI cells evaluated by FISH and t-test (n=3); (H). Representative images of linc00261 staining by FISH in noncancerous liver tissues. In H(a) the red circle displays the MVI cells in the portal vein, and the green box indicates the normal liver cells. N.S. not significant; *P < 0.05.

Microvascular invasion (MVI), usually defined as the presence of microscopic tumor emboli within the portal or hepatic veins of the surrounding liver tissues close to the HCC lesions, has been regarded as the early stage of HCC intra- and extra-hepatic metastasis, which markedly increases the risk of post-operative recurrence and death [26-28]. We further screened HE-stained slides in patients with MVI; specifically, three cases with MVI within the noncancerous liver tissues were selected and subjected to further fluorescence in situ hybridization (FISH) analysis, which demonstrated that linc00261 was downregulated in cells undergoing MVI compared to expression in normal liver cells (Figure 1G, 1H).

Linc00261 suppresses migration, invasion, and EMT in HCC cell lines

The previously mentioned results indicated that the downregulation of linc00261 in liver tumor cells was significantly associated with MVI, thereby contributing to metastatic recurrence. To further understand the potential underlying mechanism, gene ontology (GO) analysis were conducted using DAVID (https://david.ncifcrf.gov/tools.jsp) [29] with linc00261 negatively co-expressed mRNAs in HCC obtained from cBioPortal [30,31]. Interestingly, the analyses of biological process (BP), cellular component (CC), and molecular function (MF) showed that linc00261 was significantly associated with cell-cell adhesion (BP, CC, and MF), extracellular matrix disassembly (BP), microtubule-based movement (BP), and microtubule motor activity (MF); further, the cell-cell adhesion-associated co-expressed mRNA counts ranked among the highest levels, indicating that linc00261 is closely related to cell motility, EMT, and even cancer metastasis (Supplementary Figure 2A).

Based on RT-qPCR detection in a series of cell lines (Supplementary Figure 2B), we constructed linc00261-knockdown models using MHCCLM3 and SNU-449 cell lines (Supplementary Figure 2C), as well as models stably overexpressing linc00261 based on HepG2 and SMMC-7721 cell lines (Supplementary Figure 2D). Neither linc00261 knockdown or overexpression influence the proliferation, as revealed by CCK-8 and plate clone formation assays (Supplementary Figure 2E-G); however, migration and invasion were significantly promoted after linc00261 knockdown (Figure 2A), and obviously suppressed after overexpressing linc00261 (Figure 2B). We also determined the expression of EMT-associated proteins in linc00261-overexpressing and knockdown cells. The knockdown of linc00261 promoted ZEB1 and vimentin expression and suppressed E-cadherin expression (Figure 2C, 2D), whereas its overexpression resulted in the opposite effects on ZEB1, E-cadherin and Vimentin expressions (Figure 2E-G). Additionally, ectopic expression of linc00261 in SMMC-7721 cells significantly inhibited the formation of lung metastatic lesions in male Biocytogen-NOD-PrkdcscidIL2rgtm1/Bcgen (B-NDG) mice based in-situ tumor model, although few liver orthotopic xenograft tumors were observed (Figure 3A-F). Mice implanted with SMMC-7721-vector cells showed earlier death compared to that in the linc00261-overexpressing groups (Figure 3G).

Figure 2.

Linc00261 overexpression suppressed migration, invasion and EMT process in HCC cell lines. (A) Representative images and statistical analyses of migration and invasion assays after linc00261 knockdown in MHCCLM3 and SNU-449 cells; (B) Representative images and statistical analyses of migration and invasion assays after linc00261 overexpression in HepG2; (C-F) The protein expressions of epithelial (E-cadherin) and mesenchymal associated markers (N-cadherin, β-catenin and Vimentin)/transcription factor (ZEB1) determined by western blotting after linc00261 knockdown in MHCCLM3 and SNU-449 (C) or linc00261 overexpression in HepG2 (E), as well as the comparisons of gray values of these proteins using ANOVA (D and F); (G) Representative images and statistical analyses (t-test) of ZEB1, E-cadherin and Vimentin expression evaluated by immunofluorescence after linc00261 overexpression in HepG2 (n=3). N.S. not significant; *P < 0.05; **P < 0.01; ***P < 0.001. The original western blotting images refers to Supplementary Figure 5.

Figure 3.

Forced expression of linc00261 in SMMC-7721 significantly suppresses lung metastasis and results in early death in B-NDG mice. Nine and 7 B-NDG mice were separately implanted with SMMC-7721-vector and -Lv-linc00261 cells in the livers and maintained in SPF-grade condition. (A) The representative microscope photos of the livers of B-NDG mice revealed by HE staining (A, left) and the gross specimens (A, right); (B) Comparison of the hepatic lesions growth rates between two groups using chi-square test; (C) The representative microscope photos of the lungs of B-NDG mice revealed by HE staining (C, left) and the gross specimens (C, right); (D and E) Comparisons of the lung metastatic lesions growth rates (D; chi-square test) and the tumor numbers (E; t-test) between two groups; (F) The representative images of in vivo orthotopic (in the livers) and metastatic (in the lungs) tumors of B-NDG mice; (G) Overall survival comparison of B-NDG mice after SMMC-7721 cells implantation using Kaplan-Meier method. B-NDG, Biocytogen-NOD-PrkdcscidIL2rgtm1/Bcgen; HL, hepatic lesion; LM, lung metastasis; N.S. not significant; *P < 0.05; ***P < 0.001.

Linc00261 suppresses migration, invasion, and EMT by upregulating its neighboring gene, FOXA2

It is well-known that lincRNA transcriptionally regulates neighbor gene expression in cis by interacting with transcription factors [5,32]. FOXA2 is located upstream of linc00261 in the genome. Thus, we examined FOXA2 expression in another HCC cohort (cohort 2; n=79), which showed that FOXA2 protein was primarily localized in the nucleus and was significantly decreased in HCC tissues compared to that in adjacent noncancerous tissues (Supplementary Figure 3A, 3B). Further, FOXA2 protein levels were negatively associated with MVI (P=0.045), portal vein tumor thrombus (P=0.015), and TNM stage (P=0.008) (Supplementary Table 3); in addition, patients with lower FOXA2 protein levels in HCC tissues had a shorter RFS compared to those with higher levels (Supplementary Figure 3C). Univariate and multivariate Cox regression analyses indicated that portal vein tumor thrombus (95% CI, 1.330-7.432; P=0.009), differentiation (95% CI, 0.091-0.586; P=0.002), and FOXA2 protein expression (95% CI, 0.152-0.706; P=0.004) were independent prognostic factors for RFS (Supplementary Table 4). Furthermore, the knockdown of FOXA2 promoted the migration and invasion capabilities of HCC cells (Supplementary Figure 4A-C), suppressed expression of the epithelial marker E-cadherin, and increased levels of the mesenchymal markers ZEB1, β-catenin, and vimentin (Supplementary Figure 4D). These results confirmed that FOXA2 plays a suppressive role in HCC metastasis.

To explore whether linc00261 suppresses HCC metastasis by modulating its neighboring gene FOXA2, correlation analysis between linc00261 and FOXA2 mRNA based on RNA sequencing data using GEPIA (http://gepia.cancer-pku.cn/detail.php?gene=) [33] was performed, indicating that they were positively correlated in both normal liver (P=0.00039, r=0.48; Figure 4A, left) and HCC tissues (P=7.6E-32, r=0.56; Figure 4A, right); data from a series of cell lines and tissues from patients with HCC (n=44, intersection of the previous two cohorts) also demonstrated a strong correlation between linc00261 and FOXA2 mRNA (Figure 4B, 4C) and protein (Figure 4D). Next, linc00261 localization was evaluated by RT-qPCR following the separation of nuclear and cytoplasmic RNA fractions (Figure 4E), as well as by FISH (Figure 4F), which demonstrated that linc00261 was present in both the nucleus and cytoplasm, but primarily in the nucleus. Further, knockdown of linc00261 significantly attenuated FOXA2 expression at the mRNA level (Figure 4G), whereas the ectopic expression of linc00261 upregulated FOXA2 expression at the mRNA level (Figure 4H); and the FOXA2 protein expression altered in accordance with mRNA in both linc00261 knockdown and overexpression experiments (Figure 4I-K). Moreover, linc00261- and FOXA2-co-expressed mRNAs in patients with HCC were separately obtained from cBioPortal (http://www.cbioportal.org/) [30,31]. Genes co-expressed with linc00261 significantly overlapped with those co-expressed with FOXA2 (Figure 5A), and the common negatively co-expressed mRNAs were significantly enriched in the following terms as revealed by GO analysis: cadherin binding involved in cell-cell adhesion, extracellular matrix disassembly and collagen catabolic process (BP), cell-cell adherens junction (CC), and cadherin binding involved cell-cell adhesion (MF) (Figure 5B) [29]; this was in agreement with our previous findings (Supplementary Figure 2A). Ultimately, rescue experiments showed that knockdown of FOXA2 in Lv-linc00261-HepG2 cells could abrogate the inhibitory effect of linc00261 on migration, invasion, and EMT (Figure 5C, 5D).

Figure 4.

Correlation between linc00261 and its neighbor gene FOXA2 in HCC. (A) Correlation between linc00261 and FOXA2 mRNA expression in normal liver tissues (left; Pearson r=0.48, P=0.00039) and HCC tissues (right; Pearson r=0.56, P=7.6 10E-32) obtained from GEPIA; (B) Correlation between linc00261 and FOXA2 mRNA expression in a subset of HCC cell lines evaluated by RT-qPCR (n=11, Pearson r=0.92, P=0.000); (C) Correlation between relative linc00261 and FOXA2 mRNA expression in patients with HCC evaluated by RT-qPCR (Pearson r=0.6111, n=44, P=0.000); (D) Correlation between relative linc00261 expression (detected by RT-qPCR) and FOXA2 protein expression (detected by immunohistochemical staining) in patients with HCC (Pearson r=0.353, n=44, P=0.019); (E) Linc00261 allocation in nucleus and cytoplasm of MHCCLM3 and SNU-449 cells detected by nucleus/cytoplasm separation followed with RT-qPCR, U1 snRNP and GAPDH were used as controls; (F) Linc00261 staining in MHCCLM3, SNU-449 and HepG2 cell lines by FISH; (G, H) FOXA2 mRNA expression after linc00261 knockdown (MHCCLM3 and SNU-449 cell lines) or after linc00261 overexpression (HepG2) examined by RT-qPCR; (I-K) FOXA2 protein expression after linc00261 knockdown (i; MHCCLM3 and SNU-449) or overexpression (J and K; HepG2) examined by western blotting (I and J) and immunofluorescence staining (K), as appropriate; the gray values of the protein bands were evaluated by image J. N.S. not significant; *P < 0.05; **P < 0.01; ***P < 0.001. The original western blotting images refers to Supplementary Figure 5.

Figure 5.

Bioinformatic analysis of linc00261-FOXA2 co-expressed mRNAs and the effects of linc00261/FOXA2 axis on migration, invasion and EMT in HCC. A. Pearson correlation analysis of the correlation coefficient (r) of linc00261 and FOXA2 co-expressed mRNAs obtained from cBioPortal (Pearson r’=0.6762, P < 0.0001, n=20090); B. Linc00261-FOXA2 common negatively (r ≤ -0.25; n=1187) co-expressed mRNAs were subjected for further GO analysis conducted by DAVID; C. Rescue assays demonstrated that the suppressive roles in migration and invasion are reversed by FOXA2 knockdown using siRNAs in linc00261-overexpressed HepG2 evaluated by ANOVA followed LSD-t test; D. Rescue assay indicated that the effect of linc00261 overexpression on EMT associated proteins is reversed by FOXA2 knockdown using siRNAs in HepG2. GO, Gene Ontology; ***P < 0.001. The original western blotting images refers to Supplementary Figure 6.

Linc00261 interacts with SMAD3 to activate the transcription of FOXA2

Considering the positive regulatory effect of linc00261 on FOXA2 expression at both the mRNA and protein levels, we predicted the involvement of chromatin-modifying complexes controlling FOXA2 transcription. To understand the potential mechanism, we performed an RNA pull-down assay and mass spectrometric analysis in MHCCLM3 cells, which suggested that 532 proteins were specifically bound to linc00261 (Figure 6A and Supplementary File 1). By constructing a Venn diagram comprising the 396 transcriptional factors predicted to modulate FOXA2 transcription using JASPAR (Supplementary File 2) [34], SMAD3, FOXO3 and CUX1 were anticipated to be recruited by linc00261 to the promoter region of FOXA2 to regulate its transcription (Figure 6B). We noticed that SMAD3, and a SMAD2/3/4 complex were predicted targeting to FOXA2 promoter region. Interestingly, a previous study discovered that DEANR1 (also termed linc00261) facilitate endoderm differentiation by upregulating the expression of endoderm factor FOXA2 via inducing SMAD2/3 recruitment to the FOXA2 promoter. Therefore, we focused on SMAD3 for further research. Proteins retrieved in the RNA pull-down assay were subjected to western blotting for linc00261-protein enrichment evaluation, which verified that SMAD3 does bind to linc00261 (Figure 6C). RNA immunoprecipitation (RIP) followed by RT-qPCR also demonstrated that SMAD3 was specifically enriched with linc00261 in MHCCLM3 cells (Figure 6D); Chromatin immunoprecipitation (CHIP) showed that SMAD3 could bind to the FOXA2 promoter in MHCCLM3 cells, and knockdown of linc00261 significantly attenuated the binding of SMAD3 to FOXA2 promotor region (Figure 6E). Moreover, SMAD3, primarily located in the nucleus, was significantly downregulated in HCC tissues compared to levels in adjacent noncancerous tissues, as revealed by immunohistochemical staining of 40 randomly selected cases from cohort 2 (Figure 6F, 6G), which might also attenuate linc00261-induced FOXA2 transcription. SMAD3 expression in HCC tissues was also positively correlated with FOXA2 protein expression (Figure 6H; Pearson r=0.3627, n=40, P=0.021). Finally, the knockdown of SMAD3 in HCC cell lines significantly reduced FOXA2 expression (Figure 6I). These results suggest that linc00261 recruits SMAD3 to activate the transcription of FOXA2.

Figure 6.

Linc00261 regulates FOXA2 transcription by recruiting SMAD3 to the promoter. A. SDS-PAGE electrophoresis and silver staining of linc00261 and negative control RNA binding proteins obtained from RNA pull-down assay, and the Venn diagram showed 532 linc00261 specifically binding proteins; B. Venn diagram representing the potential 3 transcription factors (SMAD3, FOXO3, and CUX1) that not only specifically binding with linc00261, but also predicted to regulate the transcription of FOXA2 using JASPAR; C. Proteins retrieved in the RNA pull-down assay were subjected to western blotting for linc00261-protein (SMAD3) enrichment evaluation in MHCCLM3 cells; D. SMAD3 bound directly to linc00261 revealed by RIP assay in MHCCLM3 cells, U1 snRNP and GAPDH were used as internal control; E. CHIP demonstrated binding of SMAD3 to the promoter region of FOXA2 in MHCCLM3 cells, which was dependent on linc00261 expression; the binding positions of the promoter region are indicated in the upper diagram and Supplementary Table 6; F. Representative images of various SMAD3 protein levels in HCC and adjacent normal tissues; G. Comparison of SMAD3 protein expression in HCC and adjacent normal tissues (n=40). H. Correlation between SMAD3 and FOXA2 protein expressions in patients with HCC detected by immunohistochemical staining (Pearson r=0.3627, n=40, P=0.021); I. FOXA2 expression in SMAD3 knockdown MHCCLM3 and SNU-449 cells revealed by western blotting. N.S., not significant; *P < 0.05. The original western blotting images refers to Supplementary Figure 6.

Linc00261 was repressed by EZH2 induced H3K27Me3

Accumulating studies revealed that histone methylation works cooperated with DNA methylation in silencing lncRNAs expression, which is partially induced by EZH2, a crucial co-regulator of H3K27Me3 and CpG island methylation associated-lncRNA epigenetic silencing. A recent study discovered DNMT1-mediated CpG island methylation within linc00261-FOXA2 locus is responsible for linc00261 downregulation in lung adenocarcinoma cell line [13]. We speculated that EZH2 may be participated in H3K27Me3-associated linc00261 transcription repression in HCC. Treatment with GSK126 at a concentration of 10 μM for 48 h, a selective inhibitor of EZH2 methyltransferase activity [35], significantly increased linc00261 expression in liver cancer cells (Figure 7A); Furthermore, western blotting revealed that H3K27Me3 expression was decreased after treatment with GSK126 (Figure 7B). Additionally, EZH2 was predicted to binding with linc00261 promoter region at 5 sites, and to study whether EZH2 and H3K27Me3 directly bind to the promoter region of linc00261, 5 pairs of primer were designed across the promoter (Figure 7C). ChIP-qPCR assay with anti-H3K27Me3 or anti-EZH2 antibodies showed that both EZH2 and H3K27Me3 enriched the promoter compared to IgG groups, which is exactly right within the CpG islands; besides, treatment with GSK126 for 48 h at a concentration of 10 μM obviously decreased the enrichments of EZH2 and H3K27Me3 (Figure 7D). These results indicated that linc00261 expression is partially transcriptionally controlled by EZH2-mediated H3K27Me3 level in HCC.

Figure 7.

EZH2 represses linc00261 expression via modifying H3K27 trimethylation in liver cancer cells. A. RT-qPCR analysis of linc00261 expression in HepG2 or SMMC-7721 cells after treatment with GSK126 for 48 h at a concentration of 10 Μm; B. Treatment with GSK126 for 48 and 72 h at a concentration of 10 μM in HepG2 cell line, the H3K27Me3 expression was detected using western blotting; C. Distributions of the CpG islands (predicted by MethPrimer website) and primers for CHIP-qPCR assay in linc00261 promoter region; D. CHIP-qPCR analysis of EZH2 and H3K27me3 enrichment at the promoter of linc00261 in HepG2 cells after GSK126 treatment with GSK126 (10 μM) for 48 h; E and F. Representative images of EZH2 and H3K27Me3 staining using immunohistochemistry in human HCC and adjacent liver tissues, both of them had a higher expression in tumor tissues than that in the adjacent tissues; G. Pearson correlation analysis of linc00261 with EZH2 and H3K27Me3 proteins. *P < 0.05; **P < 0.01; ***P < 0.001. The original western blotting images refers to Supplementary Figure 6.

To determine the clinical relation of linc00261 with EZH2, H3K27Me3 expression in HCC, we detected EZH2 and H3K27me3 expression by immunohistochemical staining in tumors and adjacent normal specimens. Both EZH2 and H3K27Me3 were expressed at high levels in tumor tissues compared to non-tumor regions (Figure 7E, 7F). Moreover, the Pearson correlation analysis revealed that linc00261 expression was negatively correlated with EZH2 (n=35; r=-0.3779, P=0.0252) and H3K27Me3 (n=35; r=-0.3354, P=0.0489) expression in HCC tissues (Figure 7G).

Discussion

In this study, we discovered that lower expression of linc00261 in tumor tissues is associated with poorer prognosis for multiple cancers including HCC. Linc00261 and its positively correlated neighbor gene FOXA2, were negatively associated with malignant biological behaviors, particularly MVI, and served as excellent independent prognostic factors for RFS in patients with HCC. Linc00261 inhibits migration, invasion, and EMT partially by transcriptionally upregulating FOXA2 through an interaction with SMAD3, thereby suppressing HCC metastasis. Additionally, we identified EZH2 is responsible for linc00261 transcription repression via modulating H3K27Me3, thereby contributes to HCC progression.

To date, few lncRNAs have been successfully used as therapies or for predicting HCC metastasis. Linc00261 was found to be dysregulated in gastric cancer [36], lung cancer [13,37], endometrial carcinoma [38], choriocarcinoma [39], and HCC [6,7]. These studies proposed that linc00261 inhibits tumor proliferation by inducing G2-M cell cycle arrest [13] or promoting apoptosis and suppresses migration and invasion by promoting slug degradation [36] or inhibiting Notch signaling [7]. However, our present study found that the transient knockdown or reintroduction of linc00261 in HCC cells, rarely affected proliferation, and the GO enrichment results didn’t show cellular proliferation, apoptosis or cell cycle associated process, which is a more comprehensive lens to look at this contradiction, and further mechanistic research is required [7]. In addition, the loss of linc00261 in cells undergoing MVI was observed; combined with the effect of lin00261 on EMT and lung metastasis suppression, we concluded that the loss of linc00261 might be a key risk factor for the generation of metastasis-initiating cells in HCC.

FOXA2 is tightly coupled with HCC initiation and progression as reported by the present results and others [14,16,17]. Moreover, Combining the observations that linc00261 is specifically expressed in adult endoderm-derived tissues/organs and the liver, which exhibited the highest expression level compared to other tissues, the effect of linc00261 on HCC metastasis, and the interaction between linc00261 and FOXA2 in endoderm differentiation and HCC, we propose that the linc00261-FOXA2 axis, typically involved in endoderm differentiation, is hijacked during HCC metastasis. It is thus of greater clinical significance to determine the role of the linc00261-FOXA2 axis in HCC than in other cancers. This is despite the fact that a recent study reported that the loss of linc00261, regulated by FOXA2, promotes malignant phenotypes such as G2-M cell-cycle acceleration and the activation of DNA damage in lung adenocarcinoma, whereas whether FOXA2 functions as a tumor suppressor in lung adenocarcinoma is still under controversy [13]. In brief, it is interesting and valuable to note that two neighboring development-associated genes, namely linc00261 and FOXA2, are strongly and positively correlated, and coincidentally play a crucial suppressive role in MVI and subsequent HCC metastasis. Much more similar mechanisms during embryonic development might be hijacked in oncogenesis and cancer metastasis.

SMAD3 contains a DNA-binding domain (Mad homology 1), followed by a proline-rich region (also known as the linker region) and a Mad homology 2 domain. Phosphorylation sites in the linker region and C-terminus integrate various signals from multiple signaling pathways such as CDKs, ERK, JNK, MAPK, and TGF-β, as well as cross-talk with these pathways. This influences the formation of heteromeric complexes with other transcription factors and the binding of SMAD3 to target promoters, ultimately resulting in divergent functions, particularly with respect to the modulation of TGF-β signaling induced EMT and cancer progression [40]. To address the molecular mechanism underlying these effects, we focused on SMAD3, and found that SMAD3 binds the promoter of FOXA2 and promotes its transcription with the assistance of linc00261, thereby contributing to the suppression of HCC progression. Moreover, SMAD3 was found to be downregulated in HCC tissues compared to expression in adjacent normal tissues, which together with linc00261 downregulation contributed to the attenuation of FOXA2 transcription to some extent. However, the underlying mechanism through which linc00261 transcriptionally upregulates FOXA2 requires further in-depth analysis. For example, it is unclear whether linc00261 interacts with the promoter of FOXA2 by base pairing or if this linc00261/SMAD3 complex is sufficient for binding to the promotor and the initiation of transcription. Moreover, it remains to be determined if FOXO3, or any other factors such as SMAD3, function during this process.

Few studies demonstrated the mechanism of linc00261 downregulation in cancers. In lung adenocarcinoma, DNMT1 epigenetically silences linc00261 by promoting the CpG islands methylation that within linc00261-FOXA2 locus. Interestingly, EZH2 serves as a recruitment platform for DNMT1 and DNMT3A/3B to mediate DNA methylation. It is thus possible that EZH2 epigenetically represses linc00261 transcription by directly recruiting DNMT1 to the CpG islands. Nevertheless, we discovered for the first time that EZH2 mediated H3K27Me3 is responsible for the suppression of linc00261 transcription, and both EZH2 and H3K27Me3 bind to the CpG islands of linc00261 promoter region, indicating EZH2 as a co-regulator of two epigenetic repression systems in regulating linc00261 expression. The effect and mechanism of EZH2/linc00261 axis is essential for HCC progression and needs further investigation.

In conclusion, we demonstrated that the loss of linc00261, induced by EZH2-mediated H3K27Me3, contributes to HCC metastasis by regulating EMT, which is partially achieved via the deficiency of transcriptional upregulation of FOXA2 through an interaction between linc00261 and SMAD3. A more detailed understanding of the EZH2/linc00261 axis, linc00261/SMAD3/FOXA2 axis might reveal potential prognostic factors and therapeutic targets for HCC metastasis.

Disclosure of conflict of interest

None.