Transcriptome instability as a molecular pan-cancer characteristic of carcinomas
1 Department of Cancer Prevention, Institute for Cancer Research, The Norwegian Radium Hospital, Oslo University Hospital, P.O. Box 4953 Nydalen, Oslo NO-0424, Norway
2 Centre for Cancer Biomedicine, Faculty of Medicine, University of Oslo, P.O. Box 1078 Blindern, Oslo NO-0316, Norway
3 Department of Genetics, Portuguese Oncology Institute, Rua Dr. António Bernardino de Almeida, Porto 4200-072, Portugal
4 Cancer Genetics Group, Research Centre of the Portuguese Oncology Institute, Rua Dr. António Bernardino de Almeida, Porto 4200-072, Portugal
5 Institute of Biomedical Sciences Abel Salazar, University of Porto, Rua de Jorge Viterbo Ferreira n.° 228, Porto 4050-313, Portugal
BMC Genomics 2014, 15:672 doi:10.1186/1471-2164-15-672Published: 10 August 2014
We have previously proposed transcriptome instability as a genome-wide, pre-mRNA splicing-related characteristic of colorectal cancer. Here, we explore the hypothesis of transcriptome instability being a general characteristic of cancer.
Exon-level microarray expression data from ten cancer datasets were analyzed, including breast cancer, cervical cancer, colorectal cancer, gastric cancer, lung cancer, neuroblastoma, and prostate cancer (555 samples), as well as paired normal tissue samples from the colon, lung, prostate, and stomach (93 samples). Based on alternative splicing scores across the genomes, we calculated sample-wise relative amounts of aberrant exon skipping and inclusion. Strong and non-random (P < 0.001) correlations between these estimates and the expression levels of splicing factor genes (n = 280) were found in most cancer types analyzed (breast-, cervical-, colorectal-, lung- and prostate cancer). This suggests a biological explanation for the splicing variation. Surprisingly, these associations prevailed in pan-cancer analyses. This is in contrast to the tissue and cancer specific patterns observed in comparisons across healthy tissue samples from the colon, lung, prostate, and stomach, and between paired cancer-normal samples from the same four tissue types.
Based on exon-level expression profiling and computational analyses of alternative splicing, we propose transcriptome instability as a molecular pan-cancer characteristic. The affected cancers show strong and non-random associations between low expression levels of splicing factor genes, and high amounts of aberrant exon skipping and inclusion, and vice versa, on a genome-wide scale.