Figure 1.
Molecular profiling of lncRNAs in 407 ovarian adenocarcinomas.
A, Relative abundances of gene categories in the GENCODE 11 annotation (unique loci). B, Polyadenylation status of lncRNAs, determined by polyA+ vs. total RNA-seq from a mixture of 16 tissues. C, LncRNA expression profiling using polyA+ RNA-seq across 407 tumors. In total 25.7 billion uniquely mapped read pairs, encompassing >3 terabases, were counted in GENCODE genes. The table shows per-tumor sequencing depth, based on all GENCODE-mapped reads or lncRNA subsets. D, Distributions of lncRNA and coding gene expression levels (maximum RPKM in all tumors). RPKM, reads per kilobase per million reads. E, Histograms of correlations between DNA copy-number and RNA level (based on 407 tumors with dual data). Left panel: lncRNAs overall (left panel, n = 10,066), showing lower correlations compared to coding genes. Right panel: improved correlations when considering genes expressed at RPKM > 3 (top 19% lncRNAs, n = 1920) that were also amplified or deleted in >15 samples (right panel, n = 125).
Figure 2.
LncRNAs associate with expression subtypes.
A, 455 lncRNAs showed increased or reduced expression in one of four previously defined expression subtypes (200 random tumors, left). These lncRNAs maintained their subtype-selective expression patterns in 200 independent tumors, and subtype could there be predicted based on their expression at 77% accuracy (right). B, The same analysis based on an intergenic lncRNA subset (n = 152, 73% accuracy, left). Nearest up- and downstream protein-coding neighbors of these lncRNAs (278 unique genes) lacked strong subtype-specific expression patterns and their combined signature was less informative of subtype (51% accuracy, right).
Figure 3.
Recurrent lncRNA amplification in serous ovarian carcinoma.
A, LncRNAs and coding genes in narrow regions of recurrent amplification or deletion identified by GISTIC (q < 0.05) in 486 HGS-OvCa tumors. Gene counts for 35 tight focal peaks with a maximum of 5 overlapping coding genes are shown. Known cancer genes and unambiguous coding targets are indicated. *, regions investigated in more detail. B, Detailed view of the ACBD6-XPR1 intergenic region (AXI region, dotted line) in a ~1 Mb genomic context. The AXI focal peak is centered on RP11-522D2.1/OVAL, an uncharacterized lncRNA on chr1q25. Red shading shows copy-number profiles for individual tumors. C, Tumors ordered by OVAL expression. OVAL RNA (y-axis) was mostly low or undetectable, but was induced in focally amplified cases (as defined in Methods). D, Neither ACBD6 nor XPR1 were notably induced by AXI region focal amplification. OVAL RNA was low also in broadly amplified cases. ND, not detected. E, Average RNA-seq read density in the AXI region (dotted line) for tumors with marked AXI focal amplification (n = 10) compared to remaining tumors (normalized read counts per 1000 nt segment). F, GSEA analysis showed than experimentally determined P53 regulated genes are induced in OVAL focally amplified tumors.
Figure 4.
A, OVAL locus on chromosome 1. GenBank mRNAs, conserved transcription factor binding sites (TFBS) predicted by the UCSC brower, and other features are indicated. B, Normal tissue expression profile of OVAL (RNA-seq). Heart expression was confirmed by reverse transcription PCR. PC3, human prostate cancer cell line; Heart, heart auricle; A7, human melanoma cell line C, OVAL and its coding neighbors have disparate expression profiles. D, Subcellular RNA-seq from 7 pooled cell lines (Gm12878, HelaS3, HepG2, Huvec, H1hesc, Nhek and K562) shows predominant cytoplasmic localization.
Figure 5.
The OVAL locus is focally amplified in serous endometrial tumors.
A, Low frequency focal amplification of the OVAL locus in endometrial cancer, but not 16 other TCGA cancers (see Figure S6 in File S1). B, 56% of focally amplified cases were of the serous subtype, compared to 21% overall (P = 0.025, Fisher’s exact test). C, OVAL RNA was strongly induced in a subset of tumors, and this coincided with focal amplification of the AXI region. ND, not detected. D, Average RNA-seq read density in the AXI region for tumors with marked focal amplification (n = 4) compared to remaining tumors (normalized read counts per 1000 nt segment). E, Similar to ovarian cancer, GSEA analysis revealed induction of P53 targets in OVAL amplified tumors.