Cancer cells exploit an orphan RNA to drive metastatic progression

Abstract

Here we performed a systematic search to identify breast-cancer-specific small noncoding RNAs, which we have collectively termed orphan noncoding RNAs (oncRNAs). We subsequently discovered that one of these oncRNAs, which originates from the 3' end of TERC, acts as a regulator of gene expression and is a robust promoter of breast cancer metastasis. This oncRNA, which we have named T3p, exerts its prometastatic effects by acting as an inhibitor of RISC complex activity and increasing the expression of the prometastatic genes NUPR1 and PANX2. Furthermore, we have shown that oncRNAs are present in cancer-cell-derived extracellular vesicles, raising the possibility that these circulating oncRNAs may also have a role in non-cell autonomous disease pathogenesis. Additionally, these circulating oncRNAs present a novel avenue for cancer fingerprinting using liquid biopsies.

Conflict of interest statement

Competing interests: The authors declare no competing interests.

Figures

Fig 1:. Discovery, annotation, and validation of cancer-specific orphan non-coding RNAs in breast cancer.

a, Heatmap representing the relative abundance of 437 small non-coding RNAs that are significantly expressed in breast cancer lines but not in normal HMECs. HMEC samples were prepared in biological triplicate while all other cell lines were prepared in duplicate. Cell lines color coded by sub-type: HMEC (purple), triple negative breast cancer (TNBC; red), HER2 positive (green), and luminal (yellow). b, Of the 437 small RNAs identified in (a), 201 were significantly expressed in breast tumor biopsy small RNA gene expression profiles collected as part of The Cancer Genome Atlas (TCGA-BRCA), and these 201 were also largely absent from the adjacent normal tissue collected from the ~200 individuals in this dataset. c, These 201 cancer-specific small RNAs were classified as oncRNAs and were independently validated in a third dataset comprised of small RNA profiles from four normal epithelial samples and 10 patient-derived breast cancer xenograft models (PDX models).

Fig 2:. The oncRNA T3p is associated with breast cancer progression.

a, A volcano plot comparing the expression of oncRNAs in poorly metastatic breast cancer cells relative to their highly metastatic derivatives. T3p is highlighted in red. n = 4 biologically independent experimental comparisons. P values and log fold change were calculated using the DE-seq2 package (two-sided, no correction for multiple testing). b, T3p (in grey) maps to the 3’ end (CR7 domain) of TERC, the RNA component of telomerase. c, Normalized (per million reads mapped) coverage plots for reads mapped to TERC in smRNA-seq data from highly metastatic MDA-LM2 cells and poorly metastatic MDA-MB-231 parental cells in our previously published data (identical y-axes), and validation of this T3p upregulation by qRT-PCR (n = 6, comprised of two biologically independent sets of biologically independent triplicates, using a two-tailed Mann-Whitney test). Shown are mean ± s.e.m. d, Expression of T3p (count-per-million; cpm) in breast tumor biopsies and their matched normal tissue in the TCGA-BRCA dataset. n = 96 biologically independent paired samples. A paired two-way Wilcoxon test was used to calculate the associated P-value. e, Violin plots of T3p expression across the TCGA-BRCA dataset. n = 923 biologically independent samples. A two-way Mann-Whitney test was used to calculate the P value. Violin plots show the distribution along the minima and maxima for each cohort (median = 0 for normal and 0.34 for BRCA). f, Survival analysis in the TCGA-BRCA dataset for patients stratified into tertiles based on the expression of T3p in their tumors. All tertiles are shown. P value calculated from a two-sided log-rank test. Hazard-ratio (HR) equals 0.5 for the low vs. high tertiles. g, Expression of T3p across normal, stage I, and stage II and III samples in the TCGA-BRCA dataset shown as violin and boxplots (quartiles). Normal: n = 104 (median = 0), Stage I: n = 169 (median = 0.42), Stage II-III: n = 799 (median = 0.46). The violin plots show the distribution along the minima and maxima. The boxplots show the quartiles. The whiskers indicate quartile ± IQR (inter-quartile range). Outliers are also shown as points. P value calculated using a two-way Mann-Whitney test.

Fig 3:. T3p regulates gene expression and drives metastatic progression.

a, A volcano plot of gene expression changes following transfection of anti-T3p LNA relative to scrambled LNA in highly metastatic MDA-LM2 cells. n = 2 biologically independent experiments. P and log fold change values were calculated using the Lumi package (significance threshold was set at two-tailed adjusted P = 0.01). b, Scatter plot comparing gene expression changes induced by anti-T3p LNA in MDA-LM2 cells versus T3p mimetic in MDA-MB-231 parental cells. Reported is the associated Pearson correlation (R = −0.234, P ~ 0). c, Venn diagrams showing overlap of genes upregulated in LNA-transfected and downregulated in T3p mimetic-transfected MDA-LM2 cells, as well as overlap of genes downregulated in LNA-transfected and upregulated in T3p mimetic-transfected MDA-LM2 cells. P-values from hypergeometric distribution. d, Bioluminescence imaging plot of lung colonization by MDA-LM2 cells transfected with anti-T3p LNA (LNA-T3p) or scrambled LNA (LNA-Scr); n = 4 (LNA-Scr) or n = 5 (LNA-T3p) biologically independent animals per cohort; mean ± s.e.m. are reported for each day; two-way ANOVA. Area under the curve is presented as mean ± s.d.; P was calculated using a one-tailed Mann-Whitney test. e, Bioluminescence imaging plot of lung colonization by MDA-LM2 cells stably expressing T3p-TuD or a control hairpin; n = 5 biologically independent animals per cohort; mean ± s.e.m. are reported for each day; two-way ANOVA. Area under the curve is presented as mean ± s.d.; P was calculated using a one-tailed Mann-Whitney test.

Fig 4:. T3p biogenesis and function.

a, A network analytical approach combined with a computational gene prioritization step (see Methods) was used to identify two candidate proteins that may be involved in T3p biogenesis: TARBP2 and DROSHA. RNA-seq experiments were performed in biological replicates. b, c Read densities from previously published CLIP datasets showing direct interactions between TARBP2 (b) and DROSHA (c) proteins with the 3’ end of TERC RNA,. d, Normalized T3p levels in smRNA-seq data from MDA-LM2 cells transfected with each of the indicated siRNAs. n = 2 biologically independent samples. Normalized expression was calculated using the DE-Seq2 package (center = mean). e, Genome browser view of AGO2 CLIP read density at the T3p locus. f, A heatmap depicting the activity of miRNAs predicted to be targeted by T3p in vivo. The miRNA seed sequences were used to evaluate gene expression changes in their targets induced by an anti-T3p LNA. For each listed miRNA, we observed a significant enrichment of the miRNA target genes among those downregulated in T3p silenced cells. Also shown are the mutual information values (MI, measured in bits) and the associated z-score for each association. Yellow boxes with red borders indicate significant enrichment and blue boxes with dark blue borders indicate significant depletion. n = 2 biologically independent experimental comparisons. g, AGO2 CLASH-qPCR was performed on candidate and control miRNAs (hsa-mir-940 and hsa-mir-107) to test for in vivo interactions with T3p. Assays performed in biological duplicate.

Fig 5:. T3p-mediated inhibition of miR-10b and miR-378c results in overexpression of metastasis promoters NUPR1 and PANX2.

a, Heatmaps showing significant upregulation of miR-10b and miR-378c targets upon transfection of MDA-MB-231 cells with T3p mimetic. Significant upregulation of these targets was also observed in highly metastatic MDA-LM2 cells compared to poorly metastatic MDA-MB-231 cells. Mutual information values (MI, measured in bits) and the associated z-scores are also shown. n = 2 biologically independent experimental comparisons. b, qRT-PCR for NUPR1 and PANX2 in MDA-LM2 and the parental MDA-MB-231; n = 3 biologically independent experiments. c, qRT-PCR for NUPR1 and PANX2 in control, miR-378c, and miR-10b knockdown cells; n = 3 biologically independent experiments. d, NUPR1 and PANX2 expression levels in cells expressing a Tough Decoy against T3p (T3p-TuD) relative to control cells; n = 3 biologically independent experiments. A one-tailed Mann-Whitney test was used to calculate P for (b-d). e, Bioluminescence imaging plot of lung colonization by MDA-LM2 cells stably expressing CRISPRi guide RNAs against NUPR1, PANX2, or a control guide; n = 4 biologically independent animals per cohort. Statistical significance of knockdown versus control cohorts was measured using two-way ANOVA. The area under the curve was also calculated for each mouse; mean ± s.d. shown (change in normalized lung photon flux times days elapsed); P was calculated using a one-tailed Mann-Whitney test. f, NUPR1 and PANX2 expression levels were measured using qRT-PCR across 96 clinical samples composed of normal tissue and tissue from the indicated breast cancer stages. P was calculated using a one-tailed Mann-Whitney test (without adjustment). g, A schematic of model for T3p-mediated control of target genes in highly metastatic cells.

Fig 6:. Systematic profiling of oncRNAs in the extracellular compartment.

a, Small RNA sequencing of RNA collected from conditioned media (CM) from breast cancer cell lines and normal HMECs. CM from each cell line was prepared in biologically independent duplicate, which were combined prior to count-per-million calculations. Heatmap shows the detection of oncRNAs in the extracellular compartment. T3p is indicated with an arrow. b, The detection of oncRNAs in serum samples collected from breast cancer patients with stage II and III disease. As a point of reference, we have also included data from 35 healthy individuals from two independent studies (green: 11 samples from the exoRNA atlas; blue: 24 samples from). T3p is indicated with an arrow.