Sp1-mediated up-regulation of lnc00152 promotes invasion and metastasis of retinoblastoma cells via the miR-30d/SOX9/ZEB2 pathway

Abstract

Purpose

Previously, we found that long non-coding RNA (lncRNA) MEG3 may act as a tumour suppressor in retinoblastoma. Overall, however, little is known about the role of lncRNAs in retinoblastoma. Here, we aimed to determine the expression and clinical significance of lnc00152 in retinoblastoma.

Methods

Lnc00152 and its downstream targets were selected using GEO datasets. The level of lnc00152 in primary patient samples was determined using RT-qPCR. Odds ratios of invasion and metastasis were calculated using logistic regression analysis. Recurrence-free survival was assessed using Cox regression analysis. Scratch wound healing, transwell and tumorigenesis assays were used to determine migration and invasion abilities of retinoblastoma cells in vitro and in vivo. Levels of EMT-related proteins were measured using Western blotting. Binding sites between lnc00152 and its targets were validated using dual-luciferase reporter and RNA pull-down assays. Lnc00152 activating transcription factors were determined using ChIP assays.

Results

We found that Lnc00152 was significantly up-regulated in retinoblastoma tumour tissues, and was a risk factor for tumour invasion, metastasis and recurrence. Lnc00152 overexpressing retinoblastoma cells exhibited a tendency to transform into mesenchymal cells, with significantly increased migration and invasion capacities, significantly decreased E-cadherin expression levels, and significantly increased N-cadherin, SOX9 and ZEB2 expression levels. In addition, we found that lnc00152, which was activated by Sp1, could inhibit miR-30d as an endogenous miRNA ‘sponge’, thereby regulating the expression of SOX9 and ZEB2.

Conclusions

Our data indicate that Lnc00152 may be associated with retinoblastoma invasion, metastasis and prognosis. In addition, we conclude that Lnc00152, which can be activated by Sp1, can induce EMT via the miR-30d/SOX9/ZEB2 pathway and, by doing so, promote the invasion and metastasis of retinoblastoma cells.

Electronic supplementary material

The online version of this article (10.1007/s13402-020-00522-8) contains supplementary material, which is available to authorized users.

Introduction

Retinoblastoma (RB) is the most common intraocular malignancy in infants and young children. The global incidence of retinoblastoma is approximately 1:12000 to 1:18000, with ~ 9000 new cases per year [1]. There are ~ 1100 new patients each year in China, 84% of which are advanced cases. The mortality rate of retinoblastoma is 3–5% in developed regions such as Europe and the United States, and ~ 37% in China, which is significantly higher [2, 3]. It is considered particularly important to study the molecular mechanisms underlying retinoblastoma invasion and metastasis in order to improve the treatment outcome of patients with metastatic disease.

Recent studies have indicated that long non-coding RNAs (lncRNAs) can be abnormally expressed in various cancers, such as cervical, colorectal and prostate cancer [4–9]. LncRNAs have also been found to be related to tumour invasion, metastasis and prognosis, and to serve as tumour biomarkers in clinical practice [10–12]. Furthermore, molecular data indicate that lncRNAs can regulate tumour epithelial-mesenchymal transition (EMT) processes via various cellular signalling pathways [13]. Our previous studies showed that lncRNA MEG3 may serve as a tumour suppressor in retinoblastoma and that the methylation status of the MEG3 promoter can be used as a biomarker for retinoblastoma [14, 15]. Overall, however, still little is known about the role of lncRNAs in retinoblastoma.

Here, the Gene Expression Omnibus (GEO) database was used to identify the most differentially expressed lncRNA (lnc00152) and its downstream targets (miR-30d, SOX9, ZEB2) in retinoblastoma. The transcription factor that could directly bind to the promoter region of lnc00152 was identified using bioinformatics tools. Next, we examined the expression level of lnc00152 and its clinical significance in retinoblastoma. The effect of lnc00152 on the migration and invasion of retinoblastoma cells was tested using in vitro and in vivo (animal) experiments, and concomitant changes in the expression of EMT-related proteins were determined. The molecular mechanism by which lnc00152 can promote retinoblastoma EMT was investigated via regulatory networks of competing endogenous RNAs (ceRNAs).

Materials and methods

Patient sample collection

Tissue specimens of 63 retinoblastoma cases from our previous study and 7 new cases were collected [14]. The method of specimen collection was the same as in our previous study. The experimental procedures were approved by the Ethics Committee of Shenzhen People’s Hospital, China. All patients in the experimental group provided signed informed consent. The patient information is summarized in Table 1.

Table 1.

Logistic regression analysis for clinicopathological characteristics

Characteristics	Patients (n)	Lnc00152 (mean ± SEM)	OR (95% CI)	P-value
Gender
Male	40	3.18 ± 0.29	0.968 (0.761–1.231)	0.791
Female	30	3.05 ± 0.40
Age (years)
≤ 2.5	40	3.39 ± 0.34	0.849 (0.662–1.089)	0.198
> 2.5	30	2.77 ± 0.32
Laterality
Unilateral	60	3.20 ± 0.26	0.864 (0.602–1.239)	0.864
Bilateral	10	2.66 ± 0.64
Differentiation
Well and Moderately	44	2.78 ± 0.26	1.271 (0.985–1.639)	0.065
Poorly and Undifferentiated	26	3.70 ± 0.45
IIRC stage
Early states (A,B,C,D)	47	2.77 ± 0.24	1.321 (1.015–1.721)	0.039
Advanced stages (E)	23	3.84 ± 0.50
Optic nerve invasion
Negative	44	2.65 ± 0.24	1.403 (1.074–1.833)	0.013
Positive	26	3.92 ± 0.47
Nodal or distant metastasis
Negative	62	2.91 ± 0.23	1.640 (1.086–2.476)	0.019
Positive	8	4.80 ± 0.79

OR (odds ratio)

Cell lines and culture conditions

The Y79 and Weri-Rb1 retinoblastoma cell lines were purchased from the American Type Culture Collection (ATCC; Manassas, USA). Both cell lines were cultured in RPMI-1640 medium containing 10% foetal bovine serum, in an incubator containing 5% CO₂ at 37 °C. The medium was changed at intervals of 2–3 days.

Transient transfection of oligonucleotides

PcDNA-Sp1, pcDNA-NC, si-Sp1, si-NC, miR-30d mimics, miR-30d inhibitor, NC-mimics and NC-inhibitors were designed and synthesised by GenePharma (Shanghai, China) and TaKaRa (Otsu, Japan), respectively. The sequences are listed in Supplementary File 1: Table S1. The oligonucleotides were thoroughly mixed with the transfection reagent Lipofectamine 2000, according to the instructions of the manufacturer (Invitrogen, ca., USA) and Opti-MEM medium (Invitrogen), placed at 25 °C for 5 min, added to the cell culture and placed in an incubator. The medium was replaced with complete culture medium 12 h later after which cell culture was continued.

Establishment of stable transfected cell lines

A lentiviral vector mediating lnc00152 overexpression and its control vector, as well as a lentiviral vector mediating expression of a shRNA targeting lnc00152 and its control vector, were all purchased from GenePharma (Shanghai, China). The lentiviral vectors were applied to four experimental groups termed lnc00152 group, Vector group, lnc00152 shRNA group, and NC shRNA group. All vectors carried a green fluorescent protein (GFP) gene. Cell transfections were performed according to an established lentiviral protocol: Y79 and Weri-Rb1 cells were seeded into 12-well plates at 0.5 × 10⁵ cells/well and allowed to grow to 40–50% confluence. Next, 1 × 10⁷ U/ml lentivirus and complete medium containing 5 µg/ml polybrene were added for incubation. The solutions were replaced with complete fresh medium 24 h later. Viral infection was evaluated by fluorescence microscopy 72 h later. Culture medium containing 2 µg/ml puromycin was used for selection and changed every 2 days. Stable cell lines were obtained after 2 weeks. The lnc00152 shRNA sequences are shown in Supplementary File 1: Table S1.

RNA extraction and RT-qPCR

Total RNA of retinoblastoma tissues, para-tumour normal retinal tissues, and retinoblastoma cell lines was extracted using TRIzol reagent (TaKaRa). The extracted RNA was reverse transcribed into cDNA using a M-MLV Reverse Transcription Kit (Thermo Scientific, Belmont, MA, USA). PCR was carried out according to the instructions of the SYBR Premix Ex Taq GC Kit (Thermo Fisher Scientific), and RT-qPCR analysis was performed using a 7900HT Fast Real-Time PCR detection platform (Thermo Fisher Scientific). The primers used were synthesised by Invitrogen and the sequences are shown in Supplementary File 2: Table S2. The results were normalised to the expression levels of GAPDH or U6, and the expression of lnc00152 relative to GAPDH and miR-30d relative to U6 was calculated and normalised using the 2^−ΔΔCt method.

Scratch wound healing assay

Y79 and Weri-Rb1 cells were seeded into 6-well plates at 5 × 10⁴ cells/well, and three replicate wells were assigned to each group of cells. Cells were incubated for 24 h in 5% CO₂ at 37 °C after which a sterilised 10 µl pipette tip was used to draw a cell-free trace (‘wound’) of approximately 1 mm width at the bottom of the wells. The plates were rinsed 3 times with PBS to remove floating cells and fresh serum-free medium was added to continue the culture. Migration of the cells at the edge of the scratches was observed using an inverted phase-contrast microscope at different time points (0 and 48 h), and the distance of the migrated cells into the wound was measured to calculate the scratch healing rate: scratch healing rate (%) = (scratch distance at 0 h - scratch distance at 48 h) / scratch distance at 0 h × 100%.

Transwell invasion assay

A layer of Matrigel (diluted at 1:5 with serum-free medium, 50 µl/well, incubated at 37 °C for 30 min to form a gel) was applied to Corning chamber filters, after which 400 µl serum-free medium containing 3 × 10⁴ cells was added to the upper chamber, and 600 µl culture medium containing 10% serum was added to the lower chamber. After 24 h of routine culture, the Matrigel and cells in the upper chamber were wiped off. Cells that had crossed the Matrigel were fixed with paraformaldehyde, stained with crystal violet for 5 min and photographed under a microscope at 200 × magnification. Five microscopic fields were randomly selected for cell counting, and the mean values were calculated.

In vivo tumorigenesis assay

Forty female NOD/SCID mice, 4–5 weeks old, were purchased from the Shanghai Laboratory Animal Centre, China. After 1 week of acclimatisation under specific pathogen-free (SPF) conditions, the mice were inoculated with approximately 1 × 10⁶ retinoblastoma cells suspended in 150 µl PBS via the tail vein. The mice were divided into 4 groups according to the cell type inoculated, i.e., lnc00152 group (Y79 cells), Vector group (Y79 cells), lnc00152 shRNA group (Weri-Rb1 cells) and NC shRNA group (Weri-Rb1 cells), with 10 mice in each group. Five weeks after the tail vein injection, the mice were sacrificed, and tumours on the lung surface were visually observed and counted. The lung tissues were conventionally fixed with 10% paraformaldehyde, dehydrated and paraffin-embedded, after which 3 µm thick sections were prepared, stained according to routine methods and evaluated under an optical microscope. The experimental procedures were approved by the Ethics Committee of Shenzhen People’s Hospital.

Western blotting

Cells in logarithmic growth phase were used for protein extraction using RIPA (Thermo Scientific), after which protein quantification was performed using a BCA protein quantification kit (Tiangen, Beijing, China). Next, 50 µg cellular protein was added to each well of a 10% SDS-PAGE gel. After electrophoresis, the proteins were transferred onto PVDF membranes, blocked using 5% skim milk at room temperature for 2 h, and washed with TBST 3 times. Primary antibodies directed against E-cadherin, N-cadherin, ZEB2, SOX9 and GAPDH (1:1000, Cell Signaling Technology, Massachusetts, USA) were added to the PVDF membranes for overnight incubation at 4 °C, after which they were washed 3 times with TBST. Subsequently, a secondary antibody (HRP-labelled goat anti-rabbit IgG, 1:2000, Santa Cruz Biotechnology, Dallas, TX, USA) was added to the membranes for 2 h at room temperature, after which they were developed using an ECL solution (Thermo Scientific). Images were captured and recorded using a GBOX gel imaging system and its accompanying software.

Immunofluorescence assay

After washing Y79 and Weri-Rb1 cells with PBS, they were fixed with 4% paraformaldehyde at room temperature for 20 min, treated with 0.5% Triton X-100 for 20 min, washed with PBS, and blocked with 1% BSA for 30 min at room temperature. Rabbit anti-SOX9 and anti-ZEB2 (1:500, Santa Cruz Biotechnology, Dallas, TX, USA) were used as a primary antibodies for overnight incubation at 4 °C. Next, after PBS washing, the cells were incubated for 2 h in the dark with TRITC-labelled anti-mouse IgG (1:200; Sigma-Aldrich, St. Louis, MO, USA) as a secondary antibody. Subsequently, DAPI was added at room temperature for 5 min and washed away with PBS. Finally, Gel Mount Aqueous Mounting Medium (G0918, Sigma-Aldrich) was applied, and the cells were observed under an inverted fluorescent microscope, photographed and recorded.

Dual luciferase reporter assay

Luciferase plasmids (Promega, Madison, WI, USA) containing wild-type (ZEB2-WT, SOX9-WT) or mutant (ZEB2-MUT, SOX9-MUT) 3’-UTR sequences were constructed, based on the relationship between miR-30d and ZEB2 or SOX9 3’-UTRs. Lipofectamine 2000 was used for transient co-transfection of the plasmids with miR-30d mimics or mimics-NC, respectively. Forty-eight hours after transfection, luciferase activity was measured according to the instructions of the Dual Luciferase Assay kit (Promega). A Renilla luciferase reporter plasmid (pRL-TK) was used as reference, and each experiment was repeated 3 times.

RNA immunoprecipitation (RIP) assay

RIP assays were performed using a Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, Massachusetts, USA). Experimental procedures were carried out according to the kit instructions: cells were collected and lysed in RIP lysis buffer, after which the cell lysates were mixed with RIP buffer containing magnetic beads conjugated with an anti-Argonaute 2 (Ago2) antibody (Millipore) or normal mouse IgG (Millipore) as a negative control for co-immunoprecipitation. Proteinase K was used to separate co-precipitated RNAs and, finally, the enrichment of lnc00152 and miR-30d in the immunoprecipitated RNAs was analysed by RT-qPCR.

RNA pull down assay

Cells were transfected with biotinylated miR-30 WT/MUT and biotinylated NC probes. After 48 h, the cells were dissolved in wash/binding buffer and subsequently incubated with streptavidin-coupled magnetic beads (M-280 Dynabeads; Thermo Fisher Scientific) at 4 °C for 3 h to pull down the biotin-labelled RNA complexes. Next, the beads were washed and the complexes were purified with TRIzol (Takara), after which the abundance of lnc00152 was analysed by RT-qPCR.

Chromatin immunoprecipitation (ChIP) assay

ChIP assays were performed using a Magna ChIP™ A/G Chromatin Immunoprecipitation Kit (Millipore) following the manufacturer’s protocol. Briefly, nuclear fractions were sonicated to reduce the DNA length to 200–500 bp, after which protein/DNA complexes were immunoprecipitated using an anti-Sp1 antibody (Cell Signaling Technology) and normalized to an IgG antibody (Cell Signaling Technology). Precipitated DNA was analysed by qPCR, and data were calculated as percentage relative to the input DNA using the equation 2 [Input Ct- Target Ct] × 0.1 × 100. The primer sequences are shown in Supplementary File 2: Table S2.

Statistical methods

Statistical analyses were performed using SPSS 20.0 software (Chicago, IL, USA). All experiments were repeated three times, and data were expressed as mean ± standard error. Statistical significance analyses were performed using Student’s t test, χ2 test, Fisher’s exact test, or Wilcoxon test, as appropriate. To investigate correlations, Pearson correlation analysis was used. Logistic regression analysis was used for odds ratio (OR), and Cox regression analysis was used for Hazard Ratio (HR). The Kaplan-Meier method was used to calculate recurrence-free survival rates (log-rank test for comparison). Differences were considered statistically significant when p < 0.05.

Results

Identification of differentially expressed lncRNAs and their downstream targets using GEO datasets

GEO dataset GSE125903 contains lncRNA sequencing data of retinoblastoma tumour (n = 6) and normal retina (n = 2) tissues. Through data analysis, we selected lncRNAs that showed significant expression differences between these tissues (Fig. 1a). These lncRNAs were listed as candidates. To associate these candidates with EMT-related proteins GEO dataset GSE59983, which contains RNA sequencing data of 76 retinoblastoma tumours, was used. We found that lnc00152, a novel lncRNA, showed a significantly low expression in retinoblastoma tumour tissues compared to normal retina tissues. In addition, we found that lnc00152 was positively correlated with SOX9 and ZEB2 protein levels in the retinoblastoma tumour tissues (Fig. 1b). Therefore, we chose lnc00152 for further analysis. We specifically searched for concomitantly down-regulated miRNAs in the GEO GSE7072 dataset, which contains miRNA sequencing data from three retinoblastoma tumour samples and three healthy retina samples (Fig. 1c). Subsequently, we assesses whether candidate miRNAs could simultaneously bind to lnc00152 and the 3’UTRs of the SOX9 and ZEB2 mRNA sequences. We found that miR-30d-5p was down-regulated in the retinoblastoma tumour tissues and could indeed bind to lnc00152 and the 3’UTR of the SOX9 and ZEB2 mRNA sequences (Figs. 1c, 6c and 7d). Based on these data, we conclude that miR-30d-5p may serve as a bridging molecule between lnc00152 and SOX9/ZEB2. A flow-chart of the analysis is shown in Fig. 1d.

Fig. 1.

Analysis of differentially expressed lncRNAs using GEO datasts. (a) Expression patterns of lncRNAs analysed using GSE125903, which includes six retinoblastoma samples and two healthy retina samples. (b) Correlation between differentially expressed lncRNAs and EMT proteins analysed using GSE125903, which includes 76 retinoblastoma samples. (c) miRNA profiles analysed using GSE7072, which includes three retinoblastoma samples and three healthy retina samples. (d) Flowchart of the analysis

Fig. 6.

Lnc00152 down-regulates miR-30d expression by competitive binding. (a) Expression of miR-30d in retinoblastoma and normal paraneoplastic tissues. (b) Correlation between expression levels of lnc00152 and miR-30d in retinoblastoma tissues using Pearson correlation analysis. (c) RT-qPCR analysis of the expression level of miR-30d in the lnc00152, Vector, lnc00152 shRNA and NC shRNA groups. (d) Sequence alignment of binding sites between miR-30d and lnc00152 wild type (WT) or mutant type (MUT). Dual-luciferase reporter assays (e) and RNA pull down assays (f) showing that lnc00152 interacts directly with miR-30d via a putative binding site. (g) RIP assays performed after treatment of cell lysates with anti-Ago2 or anti-IgG (negative control) antibodies. **p < 0.01, *p < 0.05

Fig. 7.

Targeted regulation of SOX9 and ZEB2 proteins by miR-30d. After transfection of miR-30d mimics and miR-30d inhibitor into Y79 and Weri-Rb1 cells, RT-qPCR was used to determine expression levels of miR-30d (a), SOX9 and ZEB2 (b). (c) Western blotting was used to determine the expression of SOX9 and ZEB2 proteins after overexpression or inhibition of miR-30d. (d) Prediction of binding sites between miR-30d and SOX9 or ZEB2 3’-UTRs. Dual-luciferase reporter assays showing that after co-transfection of SOX9-WT/MUT plasmid and miR-30d mimics (e), or co-transfection of ZEB2-WT/MUT plasmid and miR-30d mimics (f), miR-30d mimics inhibited the luciferase activity of the WT plasmid, but did not affect the luciferase activity of the MUT plasmid. *p < 0.05

Expression of lnc00152 in retinoblastoma tissues correlates with prognosis

The expression levels of lnc00152 in retinoblastoma tissues (n = 70) and its matched paraneoplastic normal retinal tissues (n = 70) were determined by RT-qPCR. We found that the expression of lnc00152 in retinoblastoma tissues was significantly higher than that in the matched normal tissues (p < 0.01, Fig. 2a). Next, we divided the patients into an optic nerve invasion-positive group (n = 26) and a -negative group (n = 44), as well as a nodal or distant metastasis-positive group (n = 08) and a -negative group (n = 62), and found that the expression levels of lnc00152 in the invasion- and metastasis-positive groups were significantly higher than those in the negative groups (p < 0.05, Fig. 2b, c). Subsequent logistic regression analysis revealed that lnc00152 was a risk factor for optic nerve invasion with an OR of 1.403 and, in addition, was a risk factor for nodal or distant metastasis with an OR of 1.640 (Table 1). No significant correlation was found between the lnc00152 level and any other of the parameters tested, including age, gender, laterality, differentiation and IIRC stages (Table 1). A subsequent receiver operator characteristic curve (ROC) analysis revealed that when lnc00152 was used as a diagnostic marker for optic nerve invasion, the AUC (the area under the ROC curve) was 0.651. When the cut-off value was set at 4.8, the sensitivity was 50.0% and the specificity was 93.2% (Fig. 2d). When lnc00152 was used as a diagnostic marker for nodal or distant metastasis, the AUC was 0.738, and when the cut-off value was set at 3.5, the sensitivity was 77.8% and the specificity was 63.9% (Fig. 2e).

Fig. 2.

Clinical significance of lnc00152 in retinoblastoma. RT-qPCR analysis showing that (a) the expression level of lnc00152 in retinoblastoma tissues (n = 70) was significantly higher than that in paraneoplastic normal retinal tissues (n = 70). (b) The expression level of lnc00152 in the optic nerve invasion-positive group (n = 26) was significantly higher than that in the negative group (n = 44). (c) The expression level of lnc00152 in the nodal or distant metastasis-positive group (n = 8) was significantly higher than that in the negative group (n = 62). (d), (e) ROC curve analyses of the diagnostic performance of lnc00152. (f) Kaplan-Meier analysis showing that the recurrence-free survival of the High- lnc00152 group was significantly lower than that of the Low- lnc00152 group. **p < 0.01, *p < 0.05

Subsequently, we divided the 70 patients into a lnc00152-Low group (n = 35), and a lnc00152-High group (n = 35), according to the median expression level of lnc00152, and assessed the recurrence-free survival of the two groups by Kaplan-Meier analysis. We found that the recurrence-free survival of the lnc00152-High group was significantly lower than that of the lnc00152-Low group (p = 0.0356, Fig. 2f). In addition, Cox regression was carried out to assess whether lnc00152 could serve as a risk factor for recurrence. The univariate analysis revealed that a high lnc00152 expression was significantly associated with a poor recurrence-free survival (Table 2), and as determined by the multivariate analyses, lnc00152 expression was also found to serve as an independent prognostic factor for recurrence-free survival of retinoblastoma patients (Table 2).

Table 2.

Univariate and multivariate Cox regression analyses for recurrence-free survival

Risk factors	Univariate analysis		Multivariate analysis
	HR (95% CI)	P-value	HR (95% CI)	P-value
Lnc00152	1.481 (1.187∼1.847)	< 0.001	1.480 (1.173∼1.868)	0.001
IIRC stage (A/B/C/D, E)	21.196 (6.179∼72.709)	< 0.001	11.936 (2.798∼50.925)	0.001
Nodal or distant metastasis (Negative, Positive)	13.852 (5.491∼34.943)	< 0.001	3.836 (1.347∼10.924)	0.012
Optic nerve invasion (Negative, Positive)	15.324 (4.486∼52.340)	< 0.001	2.095 (0.499∼8.790)	0.312
Differentiation(Well/Moderately, Poorly/Undifferentiated)	2.088 (0.887∼4.919)	0.092
Laterality (Unilateral, Bilateral)	1.017 (0.209∼3.503)	0.95
Age	1.355 (0.993∼1.847)	0.055
Gender (Male, Female)	1.038 (0.437∼2.464)	0.933

HR (hazard ratio)

Establishment of stable retinoblastoma cell lines using lentivirus-mediated transfections

Retinoblastoma cell lines stably and highly expressing lnc00152, and retinoblastoma cell lines stably expressing lnc00152 at low levels were generated using lentiviral transfections. Using RT-qPCR, we found that the lnc00152 transfected cells exhibited significantly elevated lnc00152 expression levels compared to the control (Vector) cells (p < 0.05, Fig. 3a, S1A). In contrast, we found that the lnc00152 shRNA cells exhibited a significantly decreased lnc00152 expression compared to the control (NC shRNA) cells (p < 0.05, Fig. 3a, S1A). Using inverted phase-contrast fluorescence microscopy we simultaneously found that the lnc00152, Vector, lnc00152 shRNA, and NC shRNA cells all showed obvious green fluorescence, and that the transfection efficiency was over 95% (Fig. 3b, lower). These results indicate that the exogenous recombinant vectors were successfully expressed in Y79 and Weri-Rb1 cells.

Fig. 3.

Migration and invasion abilities of retinoblastoma cells detected in vitro. RT-qPCR analysis showing that (a) the expression level of lnc00152 in Y79 cells stably transfected with lnc00152 was significantly increased, and that the expression level of lnc00152 in Weir-Rb1 cells stably transfected with lnc00152 shRNA was significantly decreased. (b, upper panels) After overexpression of lnc00152, the cell morphologies changed from epithelial to mesenchymal. After inhibition of lnc00152 expression, the number of tight junctions increased significantly. (b, lower panels) Expression of green fluorescence visualised by inverted phase-contrast fluorescence microscopy. (c) Scratch wound healing test showing that down-regulation of lnc00152 inhibited the migration of Weri-Rb1 cells. Up-regulation of lnc00152 promoted migration of Y79 cells. (d) Transwell assay showing that down-regulation of lnc00152 inhibited Weri-Rb1 cell invasion and that up-regulation of lnc00152 promoted Y79 cell invasion. *p < 0.05

Effects of lnc00152 on morphology of retinoblastoma cells

Using light microscopy, we found that most of the cells in the Vector group and the NC shRNA group were polygonal, with filopodia, and that the cells were loosely connected, although there were still some tight junctions. After lnc00152 shRNA transfection, the cells were polygonal, tightly arranged, with significantly decreased numbers of filopodia, and increased numbers of tight junctions, and the cell morphology tended towards that of epithelial cells. Cells in the lnc00152 transfected group showed an increase in the numbers of slender filopodia, a lack of intercellular connections, looser intercellular arrangements, and significantly increased numbers of free cells. The cell morphology tended towards that of mesenchymal cells (Fig. 3b, upper).

Effects of lnc00152 on migration and invasion abilities of retinoblastoma cells

Using a scratch wound healing assay, we found that the wound healing rate of Weri-Rb1 cells in the lnc00152 shRNA group at 48 h was significantly lower than that of the control group (p < 0.05, Fig. 3c). The wound healing rate of Y79 cells in the lnc00152 group at 48 h was found to be significantly higher than that in the control group (p < 0.05, Fig. 3c). These results indicate that lnc00152 can promote the migration of retinoblastoma cells.

Using a Transwell assay, we found that the number of invasive cells that had crossed the Matrigel in the lnc00152 shRNA group was significantly lower than that in the the control group (p < 0.05, Fig. 3c, S1B), whereas the number of invasive cells in the lnc00152 group was significantly higher than that in the blank control group (p < 0.05, Fig. 3c, S1B). The above results indicate that lnc00152 can promote retinoblastoma cell invasion.

lnc00152 affects the metastatic ability of retinoblastoma cells in vivo

Mice were divided into four groups (lnc00152, Vector, lnc00152 shRNA or NC shRNA groups) and were injected with retinoblastoma cells via the tail vein. The mice were sacrificed on day 21. All the mice were found to have different numbers of white and round tumours on the lung surface. The lnc00152 group (Y79 cells) had a significantly higher number of tumours compared to the control group, while the lnc00152 shRNA group (Weri-Rb1 cells) had a significantly lower number of tumours compared to the control group (p < 0.05, Fig. 4c). After microscopic examination, we found that the lnc00152 group showed tightly arranged, rounded, or oval cancer cells, with large and deeply staining nuclei, and that the normal lung tissue was severely damaged. In contrast, only a small number of cancer cells were seen in the lnc00152 shRNA group, and there was still some normal lung tissue (p < 0.05, Fig. 4a, b).

Fig. 4.

Metastatic tumours of the lung after injection of retinoblastoma cells into the tail vein of mice. (a, b) Pathological examination of lung tumours; scale bars, 5 mm (a) and 200 µm (b). (c) Comparison of the number of lung tumours in each group, *p < 0.05

Effects of lnc00152 on expression levels of EMT-related proteins

Western blotting was used to assess the expression levels of EMT-related proteins in the lnc00152, Vector, lnc00152 shRNA, and NC shRNA groups. We found that after lnc00152 up-regulation, the expression of E-cadherin in the retinoblastoma cell lines was significantly decreased compared to that in the control group, while the expression of the mesenchymal cell markers N-cadherin, ZEB-2 and SOX9 was significantly increased (Fig. 5a, S2A). Conversely, after lnc00152 down-regulation, the expression level of E-cadherin was found to be significantly increased in the retinoblastoma cell lines, while the expression levels of N-cadherin, ZEB-2 and SOX9 were significantly decreased (Fig. 5a, S2A). Cellular immunofluorescence assays confirmed that the fluorescent staining of the ZEB2 and SOX9 proteins in Y79 cells of the lnc00152 group was significantly increased compared to the control group, and that the fluorescent staining of the ZEB2 and SOX9 proteins in Weri-Rb1 cells of the lnc00152 shRNA group was significantly decreased compared to the control group (Fig. 5b).

Fig. 5.

Effect of lnc00152 on expression levels of EMT-related proteins. (a) Western blotting showing that up-regulation of lnc00152 significantly decreased the expression of E-cadherin, and increased the expression of N-cadherin, SOX9 and ZEB2. Down-regulation of lnc00152 resulted in a significant increase in the expression of E-cadherin, and a significant decrease in the expression levels of N-cadherin, SOX9 and ZEB2. (b) Immunofluorescence assays showing changes in the expression levels of SOX9 and ZEB2 in each group of cells

Lnc00152 down-regulates miR-30d expression through competitive binding

Based on the above results (Sect. 3.1), we first set out to examine the expression of miR-30d in primary retinoblastoma tissues and normal paraneoplastic tissues. Using RT-qPCR we found that miR-30d was lowly expressed in retinoblastoma tissues and highly expressed in normal paraneoplastic tissues (p < 0.01, Fig. 6a). Pearson correlation analysis revealed a negative correlation between lnc00152 and miR-30d expression in retinoblastoma tissues (r = -0.608, p < 0.01, Fig. 6b). Subsequently, we set out to determine the expression of miR-30d in the lnc00152, Vector, lnc00152 shRNA and NC shRNA groups using RT-qPCR. We found that after lnc00152 expression up-regulation, compared to the control group, the expression level of miR-30d in the retinoblastoma cell lines was significantly decreased (p < 0.05, Fig. 6c, S2B). After lnc00152 expression down-regulation in the retinoblastoma cell lines, the expression level of miR-30d was found to be significantly increased (p < 0.05, Fig. 6c, S2B). Using bioinformatics analysis, we found that a potential binding site was present involving lnc00152 and miR-30d (Fig. 6c). Based on this notion, we constructed lnc00152 wild-type (WT) and mutant (MUT) reporter vectors. Subsequent dual luciferase reporter assays revealed that, compared to miR-NC and no miRNA control groups, miR-30d could significantly downregulate the luciferase activity in the lnc00152 WT group (p < 0.05, Fig. 6e), whereas no significant effect was observed in the lnc00152 MUT group. A subsequent RNA pull down assay indicated that lnc00152 WT was indeed more enriched in the miR-30d pull down fraction than lnc00152 MUT (p < 0.05, Fig. 6f). In addition, we used an Ago2 antibody in RIP assays to confirm potential endogenous interactions between lnc00152 and miR-30d. We found that the Ago2 antibody significantly precipitated both lnc00152 and miR-30d compared to the control IgG antibody (p < 0.05, Fig. 6g).

MiR-30d down-regulates the expression of SOX9 and ZEB2

Retinoblastoma cell lines were transiently transfected with miR-30d mimics and miR-30d inhibitor. Using RT-qPCR we found that miR-30d was overexpressed in miR-30d mimics transfected cells, while miR-30d was down-regulated in miR-30d inhibitor transfected cells (p < 0.05, Fig. 7a, S2C). Subsequent RT-qPCR and Western blotting revealed that the expression of SOX9 and ZEB2 in the cells over-expressing miR-30d was significantly lower than that in the negative control cells. However, in the cells in which miR-30d was down-regulated, the expression of SOX9 and ZEB2 was significantly higher than that in the negative control cells (p < 0.05, Fig. 7b, c, S2D).

Using bioinformatics analysis, we found that miR-30d has putative binding sites in the 3’-UTR regions of SOX9 and ZEB2, and the sequence is shown in Fig. 7d. Based on these binding sites, we constructed SOX9 WT and MUT reporter vectors, as well as ZEB2 WT and MUT reporter vectors. Subsequent dual luciferase reporter assays revealed that the luciferase activity was significantly decreased after co-transfection of miR-30d mimics with SOX9-WT or ZEB2-WT compared to that in the control group, respectively. When SOX9-MUT or ZEB2-MUT was co-transfected with miR-30d mimics, no significant differences were observed compared to the control group (Fig. 7f, e).

Lnc00152 regulates EMT via the miR-30d/SOX9/ZEB2 pathway in retinoblastoma cells

First, we mutated the binding sites of miR-30d in lnc00152. Using RT-qPCR and Western blotting, we found that lnc00152-WT, but not lnc00152-MUT, could down-regulate miR-30d and up-regulate SOX9 and ZEB2 (p < 0.05, Fig. 8a, b). The scratch wound healing test also indicated that lnc00152-WT could increase the migration efficiency of Y79 cells, whereas lnc00152-MUT could not (p < 0.05, Fig. 8c). Next, we set out to reverse the regulation of miR-30d by lnc00152 through transient transfection of miR-30d mimics in the lnc00152 group of Y79 cells, and transient transfection of miR-30d inhibitor in the lnc00152 shRNA group of Weri-Rb1 cells. Using RT-qPCR, we found that the miR-30d level was the same in the Vector + NC mimics group and the lnc00152 + miR-30d mimics group, and the same in the NC shRNA + NC inhibitor group and the shRNA lnc00152 + miR-30d inhibitor group (p < 0.05, Fig. 8d). Subsequently, we detected the expression of SOX9 and ZEB2 in each group using Western blotting. We found that the expression levels of SOX9 and ZEB2 were significantly decreased in the lnc00152 + miR-30d mimics group compared to the lnc00152 + NC mimics group, while they were significantly increased in the shRNA lnc00152 + miR-30d inhibitor group compared to the shRNA lnc00152 + NC inhibitor group (Fig. 8e). Transwell assays also showed that the invasion ability of the retinoblastoma cells was dramatically inhibited in the lnc00152 + miR-30d mimics group compared to the lnc00152 + NC mimics group, or was significantly activated in the shRNA lnc00152 + miR-30d inhibitor group compared to the shRNA lnc00152 + NC inhibitor group (p < 0.05, Fig. 8f). Together, these results suggest that lnc00152 can modulate EMT in retinoblastoma cells by regulating miR-30d (Fig. 9f).

Fig. 8.

Lnc00152 regulates expression of SOX9 and ZEB2 via miR-30d. RT-qPCR and Western blotting showing that lnc00152-WT can down-regulate miR-30d (a) and up-regulate SOX9/ZEB2 (b), but that lnc00152-MUT cannot. (c) Scratch wound healing test showing that lnc00152-WT can promote cell migration, but that lnc00152-MUT cannot. (d) RT-qPCR analysis showing that miR-30d mimics and miR-30d inhibitor reverse the regulation of miR-30d by lnc00152. (e) Western blotting showing that the regulation of SOX9 and ZEB2 expression by lnc00152 was significantly attenuated after reversal of the regulation of miR-30d by lnc00152. (f) Transwell assay showing that regulation of the invasive ability by lnc00152 was significantly attenuated after reversal of the regulation of miR-30d by lnc00152. *p < 0.05

Fig. 9.

Transcription factor Sp1 regulates expression of Lnc00152. (a) Bioinformatics analysis showing Sp1 binding sites within the lnc00152 promoter. (b) Western blotting showing that pcDNA-Sp1 can up-regulate Sp1 expression, and that si-Sp1 can down-regulate Sp1 expression in Weir-Rb1 and Y79 cells. (c) RT-qPCR analysis showing that Sp1 over-expression can up-regulate lnc00152 expression, and that Sp1 knock-down can down-regulate lnc00152 expression. (d) Sp1 is up-regulated in Weir-Rb1 and Y79 cells compared to hTERT RPE-1 cells (retinal epithelial cell line). (e) Lnc00152 is up-regulated in Weir-Rb1 and Y79 cells compared to hTERT RPE-1 cells. ChIP assays (f) and dual-luciferase reporter assays (g) showing that Sp1 can bind to specific sites within the lnc00152 promoter. (h) A proposed model illustrating the function and mechanism of lnc00152 in retinoblastoma. *p < 0.05

Lnc00152 is regulated by transcription factor Sp1 in retinoblastoma cells

First, we searched for transcription factors that could bind to the promoter of lnc00152 using bioinformatics tools. We found that there were two specific sites for Specificity protein 1 (Sp1) in the lnc00152 promoter (Fig. 9a). Next, we found using RT-qPCR that lnc00152 was up-regulated when Sp1 was overexpressed and, as expected, that lnc00152 was down-regulated when Sp1 was knocked down (p < 0.05, Fig. 9b, c). Moreover, we found that the Sp1 expression level was higher in Weri-Rb1 and Y79 cells than in hTERT RPE-1 cells (retinal epithelial cell line) (Fig. 9d) and that also the lnc00152 expression level was higher in Weri-Rb1 and Y79 cells than in hTERT RPE-1 cells (p < 0.05, Fig. 9e). An additional ChIP assay showed that Sp1 was more enriched in the lnc00152 promoter of Weri-Rb1 and Y79 cells than in that of hTERT RPE-1 cells (p < 0.05, Fig. 9f). Finally, dual-luciferase reporter assays revealed that Sp1 knockdown significantly reduced the luciferase activity of the wild type lnc00152 promoter, but that the activity of a reporter containing mutated Sp1 sites was not affected (p < 0.05, Fig. 9g). These findings indicate that upregulation of lnc00152 is mediated by Sp1 in retinoblastoma cells (Fig. 9h).

Discussion

EMT is an important biological process for the migration and invasion of epithelial-derived malignant cells. Elucidating the molecular mechanisms underlying EMT in malignant tumour cells may facilitate the design of diagnostic methods for, and targeted therapies of, metastatic tumours [16, 17]. LncRNAs have been found to be widely involved in all aspects of EMT. LncRNAs HOTAIR and LncRNA XIST have, for example, been shown to promote cell invasion and migration by affecting EMT processes in retinoblastoma [18, 19]. Here, we aimed to assess aberrant expression of lncRNAs in retinoblastoma, and to explore whether they participate in EMT and, by doing so, in promoting retinoblastoma invasion and metastasis. To this end, we analysed RNA-seq data of the GSE125903 and GSE59983 datasets and found that a novel lncRNA, lnc00152, was not only highly expressed, but also showed a positive correlation with SOX9 and ZEB2 expression, proteins known to related to EMT [20–22], in retinoblastoma. Previously, lnc00152 has been found to be highly expressed in colorectal cancer and that high levels of lnc00152 were associated with advanced pathological stages and poor overall survival rates in patients with colorectal cancer [23]. So far, a role of lnc00152 in retinoblastoma has not been reported. We first confirmed that lnc00152 was overexpressed in 70 primary retinoblastoma tissues and, more importantly, that lnc00152 was a risk factor for invasion and metastasis in retinoblastoma patients. These results are similar to those reported for colorectal cancer [23]. The results of our primary retinoblastoma tissue experiments and those of the RNA-seq data deposited in GEO were mutually confirmatory.

Lnc00152 has been reported to promote colorectal cancer cell viability and invasion by up-regulating N‑cadherin and down-regulating E‑cadherin [23]. Here, we set out to assess whether lnc00152 can also induce EMT in retinoblastoma cells. Microscopically, we found that lnc00152 can promote transformation of the morphology of retinoblastoma cells from that of epithelial to mesenchymal cells. These changes lead to a decreased adhesion of tumour cells and an increased migration motility. Further, we examined the role of lnc00152 in EMT in retinoblastoma cells by performing a number of in vitro experiments. We found that lnc00152 can significantly increase the scratch wound healing (migration) and Matrigel traversing (invasion) abilities of retinoblastoma cells, supporting the notion that lnc00152 may promote invasion and metastasis of retinoblastoma cells by promoting EMT. To further verify whether lnc00152 is involved in EMT, we examined the effect of lnc00152 on the expression of EMT-related proteins in retinoblastoma cells by Western blotting. We found that lnc00152 significantly decreased the expression of the epithelial marker E-cadherin in retinoblastoma cells, and increased the expression of the mesenchymal markers N-cadherin, SOX9 and ZEB2. These results again indicate that lnc00152 can induce EMT in retinoblastoma cells, thereby promoting tumour invasion and metastasis.

In 2011 Poliseno et al. [24] proposed the competing endogenous RNA (ceRNA) hypothesis: endogenous RNAs such as mRNAs, circRNAs and lncRNAs contain miRNA-binding sites, and through these binding sites endogenous RNAs can competitively bind the same miRNA, thereby reducing inhibition of the target mRNA and increasing the expression level of the corresponding target gene. The ceRNA concept has changed the RNA interaction mode from “miRNAs→RNAs” to “RNAs→miRNAs→RNAs”. A large number of experimental studies has demonstrated that complex networks can be formed among miRNAs, mRNAs, lncRNAs and circRNAs. These networks coordinate the expression levels of different RNAs, and once the homeostatic balance of the ceRNA network is broken, it can lead to disease, including cancer [25, 26]. In cancer, lncRNAs may play important roles as ceRNAs. Cheng et al. [18] found, for example, that lncRNA XIST can induce EMT by regulating miR-101 in retinoblastoma. Yang et al. [19] found that lncRNA HOTAIR regulates EMT in retinoblastoma via the miR-613/c-Met regulatory axis. More importantly, it has been reported that lnc00152 can function as a ceRNA sponging miRNA‑206 in colorectal cancer [23]. A further clarification of this means of gene expression regulation may provide new perspectives for studying the mechanisms underlying tumorigenesis, and for the development of novel anti-tumour therapies. Here, we pursued this perspective and attempted to elucidate the molecular mechanisms by which lnc00152 promotes EMT in retinoblastoma. First, candidate miRNAs, which were down-regulated in retinoblastoma tissues, were identified using RNA-seq data of the GSE125903 dataset. Next, we found that miR-30d, which was one of the candidate miRNAs, harboured potential binding sites for lnc00152 and the 3’UTRs of SOX9 and EB2 using bioinformatics analysis. So we speculated that miR-30d may act as a bridge molecule linking lnc00152 to SOX9/ZEB2. Subsequently, we not only found that miR-30d was downregulated in retinoblastoma tissues, but also that its expression was negatively correlated with that of lnc00152, suggesting that there may indeed be interactions between lnc00152 and miR-30d. We also found that lnc00152 could inhibit the expression of miR-30d in retinoblastoma cells. We also confirmed the binding sites between the lnc00152 and miR-30d. To this end, we first mutated the binding sites of lnc00152 and, using dual-luciferase reporter assays, found that lnc00152-WT could directly bind miR-30d, whereas lnc00152-MUT did not. Next, we mutated the binding sites of miR-30d and, using RNA pull down assays, found that miR-30d-WT could bind lnc00152, but that miR-30d-MUT could not. In addition, through a RIP assay we found that both lnc00152 and miR-30d could be enriched using an anti-Ago2 antibody, which is the key protein for ceRNA function. In summary, we conclude that lnc00152 likely regulates the expression of miR-30d as a ceRNA.

We further explored the downstream targets of miR-30d and found that both the 3’-UTRs of SOX9 and ZEB2 harbour miR-30d binding sites. Subsequently, we found that miR-30d could significantly inhibit the expression of SOX9 and ZEB2 in retinoblastoma cells. Additional dual-luciferase reporter assays also showed that SOX9 and ZEB2 are likely to be downstream targets of miR-30d. Next, we explored whether lnc00152 regulates SOX9 and ZEB2 by binding miR-30d. We found that lnc00152-MUT could not down-regulate miR-30d and also could not up-regulate SOX9/ZEB2, indicating that miR-30d is a key factor for lnc00152 to regulate SOX9/ZEB2 expression. Next, we set out to assess whether lnc00152 can promote EMT through the miR-30d/SOX9/ZEB2 axis. We found that lnc00152-MUT could not promote the migration ability of retinoblastoma cells. We also reversed the regulation of miR-30d by lnc00152 in retinoblastoma cells and found that, by doing so, the regulation of SOX9 and ZEB2 by lnc00152 was significantly attenuated. A transwell invasion assay also revealed that miR-30d partially abolished the promotion of cell invasion induced by lnc00152. These results strongly suggest that lnc00152 is likely to regulate SOX9 and ZEB2 via miR-30d, which ultimately leads to EMT in retinoblastoma and promotes its invasion and metastasis.

We also explored the mechanism underlying abnormal lnc00152 expression in retinoblastoma. Bioinformatics analysis suggested that there are two binding sites for Sp1 in the promoter region of lnc00152. Sp1 is a transcription factor that can bind directly to gene promoter regions and activate transcription in retinoblastoma [27]. We performed dual-luciferase reporter and ChIP assays to determine whether the Sp1 binding sites were responsive to Sp1-mediated transcriptional activation of lnc00152. We found that Sp1 was more enriched in the lnc00152 promoter in retinoblastoma cells than that in retinal epithelial cells, resulting in activation of lnc00152 expression in retinoblastoma cells. Mutation of the Sp1 sites in the lnc00152 promoter led to inhibition of lnc00152 expression, meaning that Sp1 binding is essential for lnc00152 expression activation.

In addition, we preliminary explored the clinical value of lnc00152. We found that the specificity of lnc00152 in diagnosing optic nerve invasion is good (92.3%), and that the sensitivity of lnc00152 in diagnosing nodal or distant metastasis is also good (75.0%). As such, lnc00152 may hold promise for the diagnosis and prediction of invasion and metastasis in retinoblastoma. We also found that lnc00152 appears to be a high risk factor for recurrence of retinoblastoma, so evaluation of the expression level of lnc00152 may provide a basis for postoperative treatment and follow-up strategies for patients at high risk of tumour recurrence.

In conclusion, we found that lnc00152 is overexpressed in retinoblastoma, and is closely related to tumour invasion, metastasis, and prognosis. Our experimental data indicate that lnc00152 is activated by Sp1 and likely to induce EMT in retinoblastoma cells via the miR-30d/SOX9/ZEB2 pathway, thereby providing new directions to explore therapeutic regimens and prognostic markers for metastatic retinoblastoma.

Compliance with ethical etandards

All procedures performed in studies involving human participants were in accordance with the standards of the ethic committee of the Shenzhen People’s Hospital. All patients signed inform consent forms.

Conflict of interest

All authors declare that they have no conflict of interest to declare.