The long noncoding RNA SNHG1 regulates colorectal cancer cell growth through interactions with EZH2 and miR-154-5p

The human colorectal cancer cell lines (HCT-116, HCT-8, SW-480, SW-620, DLD-1 and HT-29) and the human normal colorectal epithelial cell (FHC) were obtained from American Type Culture Collection. The SW-480, SW-620, DLD-1, HT-29 and FHC cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum. HCT-116 and HCT-8 cells were cultured in RPMI-1640 with 10% fetal bovine serum. Cells were maintained in a humidified atmosphere of 5% CO₂ at 37 °C.

Full-length complementary cDNAs of human SP1 and SNHG1 were synthesized and cloned into the expression vector pcDNA3.1 (Invitrogen, China). The small hairpin RNA (shRNA) of SNHG1 was synthesized and cloned into the pLVX-shRNA1 vector (GeneCreat, China). SP1, CCND2, EZH2, KLF2 and CDKN2B siRNAs were synthesized by GenePharma (China). SNHG1 siRNAs were designed and synthesized by Ambion (USA). MicroRNA mimics and inhibitors were purchased from RiboBio (China). The plasmid vectors and siRNAs were transfected into colorectal cancer cells using Lipofectamine 2000 (Invitrogen, USA) according to the manufacturer’s instructions. All siRNA and shRNA sequences are listed in Additional file 1: Table S1.

For the SNHG1 promoter luciferase reporter assay, the core promoter of the SNHG1 gene (− 201 to + 150, relative to the transcription start site of the SNHG1 gene, contained both E1 and E2), promoter region only contained E1 and promoter region only contained E2 were synthesized and cloned into the pGL3-basic firefly luciferase reporter (GeneCreat, China). The pRL-TK vector was employed as a control. For microRNA target gene luciferase reporter assays, target sequences containing predicted microRNA binding sites were respectively synthesized and inserted into the pmirGLO luciferase vector (GeneCreat, China). Luciferase activity was measured with the Dual Luciferase Assay system (Promega, USA). Renilla luciferase activity was normalized to firefly luciferase activity.

ChIP assays were performed using the ChIP Assay Kit (Beyotime, China) according to the manual with slight modifications. Briefly, HCT-116 and HCT-8 cells were cross-linked with 1% formaldehyde solution for 10 min at room temperature and quenched with 125 mM glycine. DNA fragments ranging from 200 to 500 bp were yielded via sonication. Then the lysates were immunoprecipitated with anti-SP1, anti-EZH2, anti-H3K27me3 or normal rabbit IgG antibody. Immunoprecipitated DNAs were analyzed by qRT-PCR. Human α Satellite Repeat Primers and Human DHFR Intron 1 Primers were purchased from Cell Signaling Technology (USA). ChIP primers are listed in Additional file 2: Table S2. Antibodies used in ChIP are listed in Additional file 3: Table S3.

FISH assays were performed using Fluorescent In Situ Hybridization Kit (RiboBio, China) according to the protocol. Cy3-labeled SNHG1 antisense probe 1 and cy3-labeled SNHG1 sense probe were designed and synthesized by RiboBio (China). Cy3-labeled SNHG1 antisense probe 2 was designed and synthesized by GenePharma (China). Cells were first fixed in 4% formaldehyde for 15 min. Then the cells were permeabilized in PBS containing 0.5% Triton X-100 at 4 °C for 30 min and pre-hybridizated at 37 °C for 30 min in pre-hybridization solution. After that, probes were added in the hybridization solution and incubated with the cells at 37 °C overnight in the dark. The next day, the cells were counterstained with DAPI and imaged.

RNA pull-down assays were performed using a Magnetic RNA Protein Pull-Down Kit (Pierce, USA) according to the manual. Biotinylated SNHG1 RNA were synthesized by RiboBio (China). For each assay, 50 pmol biotinylated RNA were incubated with 50 μl prewashed streptavidin-agarose beads for one hour at 4 °C. Then, RNA-bound beads were incubated with lysates from CRC cells cytosolic/nuclear extracts and eluted proteins were detected by western blot. 3′ untranslated-region of androgen receptor (AR) RNA which contained UC-rich regions for HuR was employed as a positive control. Poly(A)₂₅ RNA did not contain HuR binding sites was employed as a negative control.

The EZ Magna RNA immunoprecipitation Kit (Millipore, USA) was used following the guidelines. Briefly, HCT-116 and HCT-8 cells were lysed in RIP lysis buffer. Magnetic beads were pre-incubated with antibodies for 30 min at room temperature and the cell lysates was immunoprecipitated with beads for 6 h at 4 °C. Then, RNA was purified and detected by qRT-PCR. Antibodies information are listed in Additional file 3: Table S3.

Five-week-old male BALB/c nude mice were maintained under specific pathogen-free conditions and manipulated according to protocols approved by the Animal Care Committee of Nanjing Medical College. Stably transfected HCT-116 cells (5 × 10⁶/0.2 ml PBS) were implanted into two sides of the same nude mouse in the armpit. Xenografts were examined every 3 days with digital calipers and tumor volumes were calculated using the following equation: volume = 1/2 (length × width²). Sixteen days later, the mice were sacrificed, and tumors volumes were measured. The samples were embedded in paraffin for hematoxylin and eosin (HE) staining and immunohistochemistry staining.

Gene set enrichment analysis (GSEA) was used to explore pathways and gene sets associated with SNHG1 in colorectal cancer. Gene expression profiles of 481 colorectal cancer samples were downloaded from TCGA dataset. According to the SNHG1 expression level order, the top 25 % and the bottom 25 % of samples were grouped as high SNHG1 group and low SNHG1 group respectively. GSEA v3.0 was used to determine whether the members of the gene sets from the MSigDB database are randomly distributed at the top or bottom of the ranking [14]. If most members of a gene set were positively related to the SNHG1, the set was termed associated with SNHG1.

All statistical analyses were performed using SPSS 18.0 (SPSS, USA) and GraphPad Prism 6 (GraphPad, USA) software. A chi-square test was used to analyze the different distribution of clinical variables. Differences in gene expression levels were analyzed using the student’s t-test. Univariate and Multivariate Cox proportional hazards regression models were used to analyze potential factors associated with prognosis. Overall survival was estimated with the Kaplan–Meier method and log-rank test was employed to evaluate difference. For in vitro and in vivo experiments, the t-test or analysis of variance (ANOVA) was used to evaluate the difference between different groups. All P-values were two sided, and P < 0.05 was statistically significant. All data are presented as the mean ± standard deviation (SD) from at least three independent replicates.

A complete description of the methods, including tissue samples and clinical data collection, RNA isolation and quantitative reverse transcription polymerase chain reaction, protein extraction and western blot, flow cytometry, cell growth and colony formation assays, 5-Ethynyl-20-deoxyuridine (EdU) incorporation assay, subcellular fractionation location and immunohistochemistry are available in Additional file 4: Supplementary materials and methods.

In this study, we first analyzed RNA sequencing data of colorectal cancer and para-cancerous tissues downloaded from TCGA [15]. As expected, many lncRNAs were identified to be aberrantly expressed in colorectal cancer (Fig. 1a). Among these differentially expressed lncRNAs, we focused on SNHG1 given its abundance is relatively high. We then analyzed two independent microarray datasets downloaded from GEO [16, 17], SNHG1 is also significantly upregulated in colorectal cancer (Fig. 1b). Similar results were also observed in 80 paired colorectal cancer tissues and adjacent tissues analyzed by qRT-PCR, as shown in Fig. 1c, SNHG1 was significantly up-regulated in 83.8% (67 of 80) of colorectal cancer tissues. To explore the correlation between SNHG1 expression and clinic-pathological features, we divided enrolled patients into two groups based on SNHG1 expression value. Chi-square test was performed to analyze clinical characters between the two groups. As shown in Table 1, SNHG1 level was correlated with tumor invasion depth (P = 0.035), distant metastasis (P = 0.024) and TNM stage (P = 0.006). Besides, the SNHG1 expression in different tumor stages of colorectal cancer in TCGA cohort was shown in (Additional file 5: Figure S1a). We also investigated the relationship between SNHG1 expression and colorectal cancer prognosis. Survival analysis using the Kaplan–Meier method revealed that a higher SNHG1 level was associated with reduced overall survival in CRC patient (P < 0.001) (Fig. 1e). The same result was identified in an independent GEO dataset [18] (P = 0.035) and in TCGA cohort (P = 0.099) (Fig. 1d and Additional file 5: Figure S1b). To further confirm the prognostic role of SNHG1 expression in colorectal cancer, the univariate and multivariate survival analysis were performed. As shown in Table 2, univariate analysis identified three prognostic factors: SNHG1 expression (P < 0.001), TNM stage (P < 0.001) and distant metastasis (P < 0.001). In addition, multivariate analysis of the three prognosis factors revealed that SNHG1 expression was an independent prognostic biomarker (Hazard ratio [HR] = 2.82; 95% confidential interval [CI] = 1.53–5.18; P = 0.001) for colorectal cancer. Besides, we analyzed SNHG1 expression in human colorectal cancer cell lines. As shown in Fig. 1f, SNHG1 expression was upregulated in all six colorectal cancer cell lines (HCT-29, DLD-1, SW-620, HCT-8, HCT-116 and SW-480) compared with the human colorectal epithelial cell FHC. In addition, per cell copy numbers of SNHG1 in these cell lines were detected (Additional file 5: Figure S1c). Moreover, through analyses of RNA sequencing data in TCGA, we found SNHG1 was upregulated in most cancers (Additional file 5: Figure S1d). These results indicate that SNHG1 upregulation may play a critical role in the development and progression of colorectal cancer.

To investigate potential regulators of SNHG1 overexpression in colorectal cancer, we used the JASPAR CORE database to search transcription factor binding sites in SNHG1 promoter [19]. Putative SP1 binding sites (GCCCCGCCCCC, − 66 bp to − 54 bp upstream of transcription start site) got the highest score. We next analyzed ChIP-Seq data of HCT-116 downloaded from the Encyclopedia of DNA Elements (ENCODE) database [20]. As shown in Fig. 2a, SP1 was highly enriched in the SNHG1 promoter region. Immunohistochemistry analysis revealed that SP1 was up-regulated in CRC (Additional file 6: Figure S2a). We then knocked down SP1 in HCT-116 and HCT-8 cells, SNHG1 expression was decreased. Moreover, SP1 overexpression promoted SNHG1 expression (Fig. 2b and Additional file 6: Figure S2b). In addition, we found SNHG1 expression was positively correlated with SP1 expression in colorectal cancer sequencing data from TCGA (Additional file 6: Figure S2c), and the positive correlation was also observed in our samples (Fig. 2c). Furthermore, ChIP assays indicated SP1 bound to the SNHG1 promoter region directly. In SP1 ChIP assays, α-Satellite and DHFR were employed as negative and positive control respectively (Fig. 2d). Besides, luciferase report assays revealed that SP1 bound to the E2 sites (− 66 bp to − 54 bp upstream of transcription start site), but not the E1 sites (− 145 bp to − 134 bp upstream of transcription start site) (Fig. 2e). Overall, above results indicate that SNHG1 overexpression in colorectal cancer is at least partly due to SP1 activation.

We designed two independent small interfering RNAs (siRNAs) to silence SNHG1 expression. As shown in Fig. 3a, SNHG1 expression was strongly reduced when examined 24 h after siRNA transfection in HCT-116 and HCT-8 cells. Next, CCK-8 assays demonstrated that SNHG1 knockdown inhibited cell growth significantly (Fig. 3b). Similarly, clone formation assays showed that clone forming ability of HCT-116 and HCT-8 cells decreased following SNHG1 knockdown (Fig. 3c). We further explored whether SNHG1 could affect colorectal cancer growth in vivo. HCT-116 cells stably transfected with sh-SNHG1#1, pCDNA-SNHG1 or empty vector were injected into male nude mice. Sixteen days after the injection, tumors from the sh-SNHG1#1 group were significantly smaller compared with the control group. Conversely, tumors of the pCDNA-SNHG1 group were significantly larger than those in the control group (Fig. 3d). We performed qPCR analyses to confirm SNHG1 expression in xenografted tumor tissues. As expected, tumors formed from sh-SNHG1#1 cells exhibited reduced SNHG1 expression, whereas tumors that from pCDNA-SNHG1 cells exhibited increased SNHG1 expression (Fig. 3e). Besides, tumor tissues collected from the sh-SNHG1#1 group exhibited lower Ki67-positive rates, whereas the pCDNA-SNHG1 group exhibited higher Ki67-positive rates compared with the control group (Fig. 3f). These findings indicate that SNHG1 can affect colorectal cancer cells growth in vitro and in vivo.

Gene Set Enrichment Analysis (GSEA) revealed significant relations between the expression of cell cyclin related genes, DNA repair related genes and SNHG1 in CRC (Fig. 4a). We then employed Ethynyl deoxy Uridine (EdU) dye assays to observe cell proliferation rates. As we hypothesized, SNHG1 knockdown could inhibit colorectal cancer cell proliferation (Fig. 4b). Consistent with the EdU assay results, flow cytometry cell cycle analysis revealed that SNHG1 knockdown decreased the proportion of cells in S phase (Fig. 4c). In addition, we examined cell cycle related proteins by western blot. The results revealed that Cyclin D1, Cyclin D2, CDK4 and CDK6 were all decreased in si-SNHG1-transfected colorectal cancer cells (Fig. 4d). Furthermore, flow cytometry cell apoptosis analysis demonstrated a significantly increasd proportion of apoptotic cells following SNHG1 knockdown (Fig. 4e). Similarly, the expression of apoptosis related proteins, including cleaved Caspase-3, cleaved PARP and Bax, significantly increased when SNHG1 was silenced (Fig. 4f). Taken together, these results indicate that SNHG1 promotes colorectal cancer cell growth by affecting cell cycle progression and apoptosis.

To investigate the potential mechanisms by which SNHG1 contributed to the malignant phenotypes of colorectal cancer cells, we first examined its subcellular localization given that the function of one lncRNA depended on its subcellular distribution [21]. Using fluorescence in situ hybridization (FISH) and subcellular fractionation, we observed that SNHG1 was expressed both in the nucleus and cytoplasm, and a larger proportion of SNHG1 was observed in the nucleus (Fig. 5a and Fig. 5b). Many cytoplasmic lncRNAs have been reported to act as competing endogenous RNAs (ceRNAs) by competitively binding microRNAs. Using StarBase v2.0 software, we found that a set of microRNAs were predicted to have a high probability of binding to SNHG1 [22]. We then used dual luciferase reporter assays to determine which miRNAs could directly interact with SNHG1 in HCT-116 cells. As shown in Fig. 5c, miR-154-5p overexpression significantly reduced the luciferase activity. To further identify whether miR-154-5p could bind to the predicted target sites in SNHG1, we constructed wild type and mutant type (putative binding sites for miR-154-5p were mutated) SNHG1 luciferase reporter vectors. As expected, co-transfection of the wild-type SNHG1 vector (Luc-SNHG1-wt) with miR-154-5p mimics, but not the mutant SNHG1 vector (Luc-SNHG1-mut), significantly reduced luciferase activities in HCT-116 cells (Fig. 5d). In addition, RNA immunoprecipitation (RIP) assay results revealed that SNHG1 and miR-154-5p were significantly enriched in AGO2-containing micro-ribonucleoprotein complexes, suggesting that the AGO2 protein bound to SNHG1 and miR-154-5p directly in colorectal cancer cells (Fig. 5e and Additional file 7: Figure S3a). RNA pull-down assays also confirmed that SNHG1 could directly bind with AGO2 in colorectal cancer cells (Fig. 5f and Additional file 7: Figure S3b). Besides, we found SNHG1 overexpression reduced miR-154-5p expression, and knockdown of SNHG1 significantly increased miR-154-5p expression (Additional file 7: Figure S3c and Figure S3d). Above all, miR-154-5p overexpression weakened the proliferation promotion effect of SNHG1 overexpression, and miR-154-5p down-regulation rescued the growth inhibition caused by SNHG1 knockdown (Fig. 5g, h, Additional file 7: Figure S3e and Additional file 7: Figure S3f). Finally, correlation analysis revealed that SNHG1 expression is negatively associated with miR-154-5p in 30 colorectal cancer tissues (P = 0.008; Fig. 5i). The effects of miR-154-5p mimics and inhibitors were validated by qRT-PCR (Additional file 7: Figure S3 h). Altogether, our results indicated that SNHG1 acts as molecular sponge for miR-154-5p, and the tumor-promoting effect of SNHG1 was partly dependent on sponging miR-154-5p.

Some studies demonstrated that miR-154-5p could act as a tumor suppressor gene through targeting CCND2, STAG2, E2F5, HMGA2 and TLR2 [23‐27]. We then investigated the targets of SNHG1 ceRNA. Using expression analyses, we found that when SNHG1 was silenced, CCND2 but not STAG2, E2F5, HMGA2 or TLR2 was downregulated in HCT-116 cells (Additional file 8: Figure S4a). Besides, inhibition of miR-154-5p in SNHG1-downregulated cells reversed the reduction in CCND2 and overexpression of miR-154-5p abolished CCND2 increase in SNHG1-upregulated cells (Fig. 6a, b, Additional file 8: Figure S4b and Additional file 8: Figure S4c). Western blot assays also showed that inhibition of miR-154-5p could rescue CCND2 protein level decrease induced by SNHG1 knockdown (Fig. 6c and Additional file 8: Figure S4d). Then, we constructed luciferase reporter vectors Luc-CCND2 containing the 3′-untranslated region (3’-UTR) of CCND2 (Additional file 8: Figure S4e). As shown in Fig. 6d, miR-154-5p overexpression reduced Luc-CCND2 luciferase activity significantly in HCT-116 cells, indicating that CCND2 was a direct target of miR-154-5p. Then, Luc-CCND2 plasmids were co-transfected with the SNHG1 plasmids and miR-154-5p mimics. MiR-154-5p induced a reduction in luciferase activity that was completely abolished by SNHG1 overexpression. We next examined whether SNHG1 regulates CCND2 expression by putative miR-154-5p binding sequences. We silenced endogenous SNHG1 and then transfected cells with the SNHG1-mut vector, which contains mutations at the putative miR-154-5p binding site, or the SNHG1 vector. Following that, we measured CCND2 expression by western blot and found that SNHG1 knockdown downregulated CCND2 expression. In rescue experiments, transfection with wide type SNHG1 plasmids reversed the decrease in CCND2 expression, whereas transfection with mutant SNHG1 plasmids did not (Fig. 6e and Additional file 8: Figure S4f). Moreover, correlation analysis revealed that SNHG1 expression was positively associated with CCND2 expression in 30 colorectal cancer tissues (P = 0.036; Fig. 6f). A similar result was observed in colorectal cancer sequencing data in TCGA database (Additional file 8: Figure S4j).

Importantly, CCND2 silencing impaired SNHG1 overexpression mediated increases in colorectal cancer cells growth and proliferation (Fig. 6g, h, Additional file 8: Figure S4 g and Additional file 8: Figure S4 h). IHC assays showed that CCND2 expression was significantly decreased in xenografts formed from SNHG1 knockdown cells (Fig. 6i). Moreover, miR-154-5p was observed to be down-regulated in colorectal cancer tissues (Additional file 7: Figure S3 g), CCND2 was observed to be up-regulated in colorectal cancer tissues (Fig. 6j and Additional file 8: Figure S4i). Collectively, these results suggest that SNHG1 released CCND2 by sequestering endogenous miR-154-5p in colorectal cancers cells.

We next explored SNHG1 functions in the nucleus. GSEA results revealed a significant correlation between SNHG1 and genes in the PRC2 related pathway (Fig. 7a). Previous studies have demonstrated that some lncRNAs in the nucleus could participate in the recruitment of PRC2 to its target genes and influence their expression. Thus, we predicted the interaction probabilities of SNHG1 and proteins in PRC2 by RNA-protein interaction prediction software RPISeq [28]. SNHG1 exhibited a high possibility of interacting with EZH2, SUZ12 and EED (as the RF or SVM score > 0.5; Additional file 9: Figure S5a). Besides, through analyzing colorectal cancer RNA sequencing data in TCGA, we found SNHG1 expression was positively correlated with EZH2, SUZ12 and EED (Fig. 7b). These bioinformatics analyses indicated that SNHG1 might be involved in PRC2 mediated epigenetic repression. To verify our hypothesis, we next performed RNA immunoprecipitation assays with EZH2, SUZ12 and EED antibodies. As we expected, RIP results revealed that SNHG1 bound with EZH2, SUZ12 and EED in colorectal cancer cells (Fig. 7c). In addition, RNA pull-down assays also confirmed that SNHG1 could directly bound with EZH2 in colorectal cancer cells (Fig. 7d). We then explored whether SNHG1 silencing affected PRC2 mediated epigenetic repression. Previous studies have reported that tumor suppressor genes KLF2, RUNX3, CDH1, CDKN1C, CDKN2B, P14 and P16 were targets of PRC2 [29‐31]. So, we detected their expression in SNHG1 knockdown HCT-116 cells by qRT-PCR. As shown in Fig. 7e, downregulation of SNHG1 significantly increased KLF2 and CDKN2B expression compared with the control cells. Hence, we selected KLF2 and CDKN2B for further study. Next, we found that upregulation of SNHG1 could reduce KLF2 and CDKN2B expression, whereas EZH2 knockdown abolished these effects (Fig. 7f). Consistent with PCR results, western blot analysis revealed that SNHG1 silencing increased KLF2 and CDKN2B protein levels, SNHG1 overexpression reduced KLF2 and CDKN2B protein levels, and the decrease could be impaired by EZH2 knockdown (Fig. 7g). We also found that SNHG1 downregulation did not affect EZH2 expression (Additional file 9: Figure S5b). To further address whether SNHG1 was involved in transcriptional repression through enrichment of H3K27me3 to the promoter regions of KLF2 and CDKN2B, we conducted ChIP assays. We found that SNHG1 knockdown decreased EZH2 binding ability to their promoters. Similarly, PRC2 enrichment induced H3K27me3 modifications also reduced in their promoter regions after SNHG1 knockdown (Fig. 7h and Fig. 7i). Finally, we observed a negative correlation between SNHG1 and KLF2 or CDKN2B expression in colorectal cancer tissues by analyzing RNA sequencing data from TCGA database (Additional file 9: Figure S5e and Figure S5f). These data indicate that SNHG1 contributes to the epigenetic repression of KLF2 and CDKB2B through binding to EZH2.

To investigate whether KLF2 and CDKN2B were involved in the SNHG1 induced promotion of colorectal cancer cell growth, we performed rescue assays. CCK-8 results revealed that silencing of KLF2 or CDKN2B (P15) partially abolished SNHG1 knockdown induced growth arrest of colorectal cancer cells (Fig. 8a and Additional file 10: Figure S6a). Similarly, EdU results revealed that silencing of KLF2 and CDKN2B could partially rescue SNHG1 knockdown induced reductions in cancer cell proliferative capacity (Fig. 8b and Additional file 10: Figure S6b). Furthermore, EZH2 silencing partially impaired SNHG1 overexpression mediated proliferative capacity increase of colorectal cancer cells (Fig. 8c and Additional file 10: Figure S6c). In addition, immunohistochemistry analyses revealed that EZH2 was upregulated in colorectal tumor tissues, whereas KLF2 and CDKN2B expression was reduced in cancer tissues (Fig. 8d). Similar results were observed in RNA sequencing data (Additional file 9: Figure S5c and Figure S5d). Moreover, IHC results demonstrated that KLF2 and CDKN2B were increased in xenograft tissues formed from SNHG1 knockdown cells (Fig. 8e). These data suggest that SNHG1 increases colorectal cancer cell growth partly depends on regulation of KLF2 and CDKN2B expression.