Large-scale identification of long-range epigenetic alterations in breast cancer

4140

Transcription can be affected by genetic or epigenetic alterations in cancers. These events take place at a single locus or over an entire chromosomal domain. Indeed single genes as well as chromosomal arms can be either deleted or gained in cancer leading to the down or up-regulation of the expression of the corresponding genes. Regarding epigenetic alterations, they are known to affect the expression of isolated genes such as CDKN2A or RB1, but some long-range effects have also been recently characterized in colon, bladder and breast cancers. Where genetic long-range alterations can easily be systematically identified by Comparative Genomic Hybridization methods (CGH), long-range epigenetic abnormalities remained poorly characterized. Therefore we performed an original large-scale analysis of such alterations by comparing transcriptome and genomic data for a same tumor set. This approach was experimentally validated on a set of 57 bladder tumors with the study of a region on 3p22 showing abnormal histone methylation leading to a loss of expression, but no DNA methylation (Nat Genet, 2006 dec). We have now applied it to a set of 128 invasive ductal breast carcinomas.

The bioinformatic analysis, involved in our study, uses a transcriptome correlation map TCM (Cancer Res 65: 1376) which calculates for each gene a transcriptome correlation score (TC score) between its expression profile and the one of its neighbors. A high score indicates that the expression of a gene is strongly correlated to the expression of its left and right neighbors. If several neighboring genes show a significant score, they form a “region of correlation”. We have identified 260 regions of co-expression for the breast cancer data. For each of these regions, we recalculate the TC scores using a subset of samples, the ones presenting no genetic alteration for the concerned region according to array CGH data. If the expressions of the genes are not anymore correlated, the original correlation is accounted for by the genetic alterations of the locus, and the regions are then called “DNA copy number-dependent”. On the other hand, if the expressions of the genes are still correlated, the original phenomenon is due to non-genetic alterations such as epigenetic ones for example, and the region is referred to as “DNA copy number-independent”. For the breast cancer data, we found 100 DNA copy number-independent regions, where 60% or more of the genes are conserved after the recalculation.

One of the DNA copy number-independent regions, the HOXA cluster, was recently published (Cancer Res 66: 10664) as affected by common abnormal epigenetic alterations: histone H3 and H4 hypoacetylation and DNA methylation leading to a loss of expression. This shows the biological validity of our systematic approach for breast cancer.

Altogether, the 100 regions form an exhaustive list of the potential long-range epigenetic alterations occurring in invasive ductal breast cancer.