Research ArticleEPIGENETICS

Portraying breast cancers with long noncoding RNAs

See allHide authors and affiliations

Science Advances  02 Sep 2016:
Vol. 2, no. 9, e1600220
DOI: 10.1126/sciadv.1600220
  • Fig. 1 lncRNA gene expression profiling in breast tissues reveals 215 dysregulated lncRNAs.

    (A) Description of the human breast tissues analyzed in this study. (B) Unsupervised consensus clustering of the samples on the basis of lncRNA gene expression. Primary tumors (823) and normal samples (172) were used for hierarchical clustering on the basis of the 500 most variant lncRNAs (based on SD). (C) Pie chart showing, among the lncRNAs that are dysregulated in breast tumors versus normal tissues, the numbers of down- and up-regulated ones. (D) Box plot for expression levels of the top dysregulated lncRNAs in the discovery cohort (top) and their expression in the TCGA RNA-seq cohort (bottom). Notches are used to compare groups; if the notches of two boxes do not overlap, the medians differ significantly. The whiskers extend to the most extreme data point, which is no more than 1.5 times the interquartile range of the box. (E) Same as in (B) for the TCGA RNA-seq cohort composed of 1052 breast tumors and 113 normal samples.

  • Fig. 2 lncRNA gene expression profiling identifies two main breast tumor categories differing with regard to ER status.

    (A) Dendrogram of 823 primary tumors and 172 normal samples obtained by consensus hierarchical clustering of the samples on the basis of expression of the top 500 most variant lncRNAs. Clusters I and III, encompassing almost all tumors, are related to the ER status. (B) Box plot illustrating the expression levels of five lncRNA genes differentially expressed between ER+ and ER tissues. Notches are used to compare groups; if the notches of two boxes do not overlap, the medians differ significantly. The whiskers extend to the most extreme data point, which is no more than 1.5 times the interquartile range of the box. (C) Heat map illustrating the expression of the 38 lncRNA genes (rows) of the ER signature across the breast tumors (columns). The lncRNAs in bold represent lncRNAs dysregulated between breast cancer and normal samples. The color scale of the heat map indicates the relative expression of each lncRNA gene. Hierarchical clustering reveals three clusters of lncRNAs. For each cluster, the most significant functional enrichment term from the guilt-by-association analysis is shown.

  • Fig. 3 Identification of four lncRNA-related clusters correlating with the known molecular subtypes and enriched in specific functional terms.

    (A) Dendrogram of the 823 breast tumors obtained by consensus hierarchical clustering according to the levels of the 500 most variant dysregulated lncRNAs, revealing four groups of tumors (clusters I to IV). How these clusters relate to the mRNA-based breast cancer subtype (based on PAM50) is also shown. (B) Histogram illustrating the number of specific lncRNAs in each molecular subtype of breast cancer. (C) Heat map illustrating the pathways whose activation or inhibition correlates with levels of subtype-specific lncRNAs and mRNAs. To relate subtype-specific lncRNAs to gene sets, an enrichment metascore was computed for each gene set. The P value of the metascore was defined as the proportion of random metascores being at least as high (low) as the metascore of the positively (negatively) subtype-specific lncRNAs. Conventional GSEA analysis was used to analyze specific enrichment in mRNAs, comparing one subtype to the three others. The significance score was defined as the log of the P value, adjusted by the sign of the enrichment metascore. (D to G) Heat maps illustrating the enrichment scores of subtype-specific lncRNAs for representative gene sets in the (D) luminal A, (E) luminal B, (F) HER2+, and (G) basal-like subtypes. A positive (negative) score is associated with the activation (repression) of the gene set.

  • Fig. 4 lncRNAs in breast cancer have prognostic value.

    (A) Forest plot showing the log2 HR with the SD (blue boxes) and the 95% confidence interval (bars) of the relapse-free survival analysis (multivariate Cox analysis). A negative HR reveals that a high lncRNA level is associated with a good outcome, and conversely. For example, five lncRNAs significantly related to relapse are shown. *P < 0.05, **P < 0.01. (B) Multivariate analysis with all the classical markers used clinically. RP11-863K10.2 is used as an example (see table S9 for the complete analysis). CI, confidence interval. (C) Exemplative Kaplan-Meier curves for RP11-863K10.2 (P = 0.01, log-rank test). (D) Same as in (B) for CYTOR. (E) Same as in (C) for CYTOR (P = 0.000776, log-rank test).

  • Fig. 5 CYTOR controls cell proliferation, cell migration, and cytoskeleton organization.

    (A) Expression of the gene encoding CYTOR in the discovery data set (823 tumors and 172 normal breast samples) and the TCGA RNA-seq data set GSE62944 (971 tumors and 103 normal breast samples). (B) UCSC (University of California, Santa Cruz) genome browser view of chromosome locus 2p11.2, which contains CYTOR. The magnified view depicts CYTOR with its associated histone marks (H3K27 acetylation, H3K4 trimethylation, and H3K4 monomethylation, from GSE49651) and methylation marks in MDA-MB-231 cells. The unfilled lollipop represents unmethylated CG dinucleotides from the Infinium HumanMethylation450. (C) Proliferation curve of MDA-MB-231 cells with and without LNA gapmer–mediated knockdown of CYTOR. A real-time cell analyzer (RTCA) software representative trace of a triplicate experiment is shown. (D) Migration kinetics of MDA-MB-231 cells assessed by continuous monitoring for approximately 24 hours. FBS (10%) in the lower chamber was used as chemoattractant, except for the control curve (purple), which represents untransfected cells with serum-free medium in the lower chamber. An RTCA software representative trace of a triplicate experiment is shown. (E) Pie chart and heat map showing the distribution of differentially expressed genes after CYTOR knockdown in MDA-MB-231 cells. (F) Enriched pathways (left) from Ingenuity Pathway Analysis in CYTOR knockdown cells. Representative differentially expressed genes of enriched gene sets are shown on the right (see also fig. S6E.) (G) Staining of F-actin with Acti-stain 488 fluorescent phalloidin (green) and of DNA with 4′,6-diamidino-2-phenylindole (blue) reveals reorganization of the actin cytoskeleton in MDA-MB-231 cells transfected with an LNA gapmer against CYTOR (right), as compared to cells transfected with a control LNA gapmer (left). Images captured with a 40× objective are shown.

  • Table 1 Predicted functions of CYTOR.

    A selected gene set significantly associated with CYTOR by the guilt-by-association analysis (see also table S5).

    Gene setScore
    Epithelial-mesenchymal transition (HALLMARK_EMT)2.75
    Proliferation (BENPORATH_PROLIFERATION)2.44
    EGFR signaling (EGFR_UP.V1)2.33
    MAPK signaling (MAPK CASCADE.POS)2.25
    RAS (MAPK) signaling (RAS.POS)2.14
    mTORC signaling (HALLMARK_MTORC_SIGNAL.)2.12
    RAF (MAPK) signaling (RAF_UP.V1.POS)2.12
    mTOR signaling (MTOR_UP.N4.V1.POS)1.99
    Migration (CELL_MIGRATION.POS)1.92

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/2/9/e1600220/DC1

    fig. S1. Dysregulated lncRNAs in breast tumors.

    fig. S2. Validation of the ER-related lncRNA signature in three data sets.

    fig. S3. Validation of the subtype-specific lncRNA signature in three data sets.

    fig. S4. Validation of the set of 27 lncRNAs predictive of relapse and dysregulated in breast cancer.

    fig. S5. Methylation of the CYTOR gene in relation to its expression.

    fig. S6. Impact of CYTOR depletion on cell cycle and gene expression.

    table S1. Clinical annotation of the 823 primary tumors and the 172 normal tissue samples from the discovery cohort, including age, size, lymph node status, ER status, HER2 status, grade, PAM50-associated subtype, and relapse information.

    table S2. Reannotation of the Affymetrix Human Genome U133 Plus 2.0 array.

    table S3. Description of the 215 dysregulated lncRNAs in breast cancer.

    table S4. Contingency table of the lncRNA-related clusters that correlate with ER status.

    table S5. Significant enrichment scores from the guilt-by-association analysis.

    table S6. Contingency table of the lncRNA-related clusters that correlate with the known molecular subtypes.

    table S7. lncRNAs signatures of the known molecular subtypes of breast cancer.

    table S8. Survival analysis: Univariate results.

    table S9. Survival analysis: Multivariate results.

  • Supplementary Materials

    This PDF file includes:

    • fig. S1. Dysregulated lncRNAs in breast tumors.
    • fig. S2. Validation of the ER-related lncRNA signature in three data sets.
    • fig. S3. Validation of the subtype-specific lncRNA signature in three data sets.
    • fig. S4. Validation of the set of 27 lncRNAs predictive of relapse and dysregulated in breast cancer.
    • fig. S5. Methylation of the CYTOR gene in relation to its expression.
    • fig. S6. Impact of CYTOR depletion on cell cycle and gene expression.
    • table S4. Contingency table of the lncRNA-related clusters that correlate with ER status.
    • table S6. Contingency tables of the lncRNA-related clusters that correlate with the known molecular subtypes.
    • table S7. lncRNAs signatures of the known molecular subtypes of breast cancer.
    • Legends for tables S1 to S3, S5, S8, and S9

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • table S1 (Microsoft Excel format). Clinical annotation of the 823 primary tumors and the 172 normal tissue samples from the discovery cohort, including age, size, lymph node status, ER status, HER2 status, grade, PAM50-associated subtype, and relapse information.
    • table S2 (Microsoft Excel format). Reannotation of the Affymetrix Human Genome U133 Plus 2.0 array.
    • table S3 (Microsoft Excel format). Description of the 215 dysregulated lncRNAs in breast cancer.
    • table S5 (Microsoft Excel format). Significant enrichment scores from the guilt-by-association analysis.
    • table S8 (Microsoft Excel format). Survival analysis: Univariate results.
    • table S9 (Microsoft Excel format). Survival analysis: Multivariate results.

    Download Supplementary tables S2, S3, S5, S8, and S9

    Files in this Data Supplement: