nc886 is induced by TGF-β and suppresses the microRNA pathway in ovarian cancer

Abstract

Transforming growth factor-β (TGF-β) signaling and microRNAs (miRNAs) are important gene regulatory components in cancer. Usually in advanced malignant stages, TGF-β signaling is elevated but global miRNA expression is suppressed. Such a gene expression signature is well illustrated in a fibrosis (or mesenchymal) subtype of ovarian cancer (OC) that is of poor prognosis. However, the interplay between the two pathways in the OC subtype has not yet been elucidated. nc886 is a recently identified non-coding RNA implicated in several malignancies. The high expression of nc886 is associated with poor prognosis in 285 OC patients. Herein, we find that in OC nc886 expression is induced by TGF-β and that nc886 binds to Dicer to inhibit miRNA maturation. By preventing the miRNA pathway, nc886 emulates TGF-β in gene expression patterns and potentiates cell adhesion, migration, invasion, and drug resistance. Here we report nc886 to be a molecular link between the TGF-β and miRNA pathways.

Fig. 1

nc886 is epigenetically silenced but induced by TGF-β. a A model depicting the opposite tendency of TGF-β and miRNA activities in OC subtypes. The gradient of red to green colors indicates high to low activity. b A diagram showing the nc886 genomic region in chromosome 5. A wide nc886 locus spanning its flanking genes is on the top and a 1000-nt region is shown magnified on the bottom. The arrows indicate transcriptional direction. All symbols (nc886 RNA, wavy line; CpG island, blue bar; EpiTYPER region, dark magenta bar) on the magnified view are drawn to exact scale, with their nt coordinates counted based on the 5′-end of nc886 being +1. Among CpG sites (vertical bars), the ones measured by EpiTYPER and pyrosequencing are designated as dark magenta and purple, respectively. c Northern hybridization after treatment of AzadC at 10 µM. d, e Northern hybridization (d) and qRT-PCR (e) of indicated genes. f A scatter plot of nc886 and TGFBI expression values from 25 OC patients from the Cheil hospital. g, h A heat map view of EpiTYPER data for all measurable CpG sites (g) and pyrosequencing of indicated CpG sites (h). AzadC and TGF-β were treated for 96 h. All other descriptions are the same as (b–e). i qRT-PCR of nc886. DNMT1-expressing plasmid (and vector control) was transfected at 48 h post treatment of TGF-β at 10 ng ml^-1. Cells were harvested at 24 h afterwards

Fig. 2

nc886 phenocopies TGF-β in promoting OC metastatic capability. a nc886 expression level quantified from northern hybridization. A representative image is Supplementary Fig. 5b. See text and Supplementary Fig. 5a for cell line designation. “SKOV3_TGF-β” indicates SKOV3_vector cells treated with TGF-β. b–d Cell attachment (b), migration (c), and invasion (d) assays. In each panel, representative images and quantification graphs are shown. In images, red thick bars indicate 1 mm (b) and red thin bars indicate 100 μm (c, d). The graphs display an average and the standard deviation from pentaplicates. e Cell migration assays upon combined treatment. The order of treatment was: TGF-β for 96 h, transfection with siPKR for 24 h, and nc886 kd for 18 h. Representative images, quantification graphs, and northern hybridization are shown. Red tick bars indicate 1 mm. f Percent cell viability (calculated from MTT values) is plotted against paclitaxel concentration. The concentrations for 50% cell death (IC₅₀ in µM) and % apoptotic cells (from Supplementary Fig. 9) are indicated on the right

Fig. 3

nc886 emulates TGF-β in controlling gene expression. a A Venn diagram depicting the number of significantly altered genes upon ectopic expression of nc886 (“nc886_exp”) and in TGF-β treatment (“TGF-β”). The fc values are in log2 scale. See figure captions for details. b A scatter plot comparing fc values of 5221 genes altered by nc886 (x-axis) and TGF-β (y-axis). c Northern hybridization of nc886, SNORD38B as a nuclear marker, and 5S rRNA as a loading control. OSE80PC cells were transfected with an nc886-targeting anti-oligo (“anti-nc886”) or with a non-targeting anti-oligo (“anti-control”) at 100 nM. Cells were harvested for nuclear and cytoplasmic fractionation at 48 h post transfection. d A scatter plot comparing nc886-kd and nc886 expression, as described in (b). e A heat map showing the unsupervised hierarchical clustering of 1024 genes (see figure caption for the selection criterion). Their expression levels, which are array values relative to the median value across samples, are displayed as green to red (see the color bar below the map for scale). Seven experiments, each of which was triplicated, are partitioned into two groups: nc886-low and nc886-high (respectively blue bar and red bar on the top) which are consistent with nc886 levels expected from experimental manipulation (nc886 kd or ectopic expression and TGF-β treatment). The blue and red color signature (for nc886-low and –high respectively) was used throughout all bar graphs and plots

Fig. 4

nc886 suppresses the miRNA pathway. a Rank distribution plots of MIRs (top) and TFTs (bottom) upon nc886 kd (left panel) and TGF-β treatment (right panel). Z-scores were sorted from the smallest to the largest values and plotted against an anonymous x-axis. See text for details. b Scatter plots of MIR or TFT Z-scores between nc886 kd and TGF-β treatment are shown on the left side. Top five candidate nc886-associated miRNAs were selected through our workflow (see Supplementary Fig. 16a and the text) and their data points are red highlighted. On the right, a heat map shows clustering of genes that were significantly altered in nc886 kd and TGF-β treatment. One cluster contains 397 candidate miRNA target genes (see Supplementary Data File 6 for a complete list), which were cross-compared to the 5 miRNAs (see Supplementary Table 5 and Supplementary Fig. 16b for details) to yield CNN3, PDCD6, and ZEB2 as direct targets. c Taqman miRNA qRT-PCR assays, with small nuclear RNA U6 (U6 snRNA) for normalization control

Fig. 5

nc886 is a pseudo-substrate of Dicer and inhibits miRNA maturation. a A schematic diagram for the identification of nc886-interacting proteins. Dicer and its known interactor proteins are shown in a STRING™ protein–protein interaction network (https://string-db.org/). The complete list of our mass spectrometry data is shown in Supplementary Data File 7. b Dicer domains and truncated mutants. c Western blot (left panel) and nc886-binding assays (right panel). Plasmids for WT or mutant Dicers (“pcDNA3.1-FLAG-Dicer-” series in Supplementary Table 4) or pcDNA3.1-FLAG vector were transfected into 293T cells. After FLAG-IP, a minor portion of beads was resuspended in SDS-PAGE gel loading buffer and boiled for 2 min to elute the bound proteins which were detected by FLAG western blot (with the anti-FLAG antibody diluted to 1:2000). The remaining major portion was used for binding assays in which % input RNA was calculated from 2^-ΔCt values. d In vitro processing assays with FLAG-Dicer (WT) which was purified by FLAG-IP. Maturation of indicated pre-miRNAs was visualized by northern hybridization. Indicated amounts of competitors (unlabeled nc886 or vtRNA1-1) were added to processing reactions. Processed mature miRNA bands as well as unprocessed pre-miRNA bands were quantified to calculate % of processing [=mature miRNA / (mature miRNA+pre-miRNA)] that is displayed on the bottom. Each bar in the graph is aligned with the corresponding band. e In vitro processing reactions of nc886 and pre-miR-200c with titrating amounts of FLAG-Dicer. Pre-miRNAs, mature miRNAs, and degradation products were quantified and plotted (bottom panels). f qRT-PCR measurement (left panel) of nc886 from SKOV3_nc886 cells transfected with pcDNA3.1-FLAG-Dicer (WT). Cells were harvested at 24 h post transfection. Western blot (right panel) detecting Dicer with anti-Dicer antibody. g qRT-PCR of nc886 (left panel) and western blot of Dicer (right panel) at 48 h upon transfection of siRNA against Dicer

Fig. 6

The function of nc886 in tumor phenotypes is attributed to its inhibition of Dicer. a–d Cell attachment assays upon Dicer kd for 24 h. Western blot (a), miRNA qRT-PCR (b), representative images (c), and quantification graphs (d) are shown. Red thick bars indicate 1 mm (c). e–g Cell migration assays upon ectopic expression of Dicer. The SKOV3_vector cells were treated with TGF-β for 96 h, transfected with pcDNA3.1-FLAG-Dicer (WT) for 24 h, and then assayed. Western blot with anti-FLAG antibody (e), representative images (f), and quantification graphs (g) are shown. Red thin bars indicate 100 μm (f)

Fig. 7

nc886 association with poor clinical outcome and a summary model. a A schematic overview of the parameters of the prediction model. BCCP Bayesian Compound Covariate Predictor, LOOCV leave-one-out cross-validation. The cutoff of Bayesian probability: nc886 high >0.7 or nc886 low <0.3. b Kaplan–Meier plots of overall survival (OS) and recurrence-free survival (RFS). A total of 285 patients were stratified into 2 groups, as predicted by the BCCP algorithm. P values were generated by the log-rank test. The + symbols in the panel indicate censored data. c A matrix table showing drug sensitivity of the two groups of OC patients. Of the 285 patients, 243 were included in this analysis by excluding patients with undetermined subtypes (n = 33) and those without available chemotherapy data (n = 9). P values were determined by χ² test. d ROC analyses for the discriminatory value of the nc886 signature in OC patients treated with chemotherapy. Bayesian probability of nc886 signature was used to identify patients who are resistant to chemotherapy. AUC area under the curve, CI confidential interval. e A summary model. Red and green letters denote pro- and anti-tumorigenic features respectively. Bold highlighted are the features proven to be regulated by nc886 in this study