Major chromosomal breakpoint intervals in breast cancer co-localize with differentially methylated regions

Abstract

Solid tumors exhibit chromosomal rearrangements resulting in gain or loss of multiple chromosomal loci (copy number variation, or CNV), and translocations that occasionally result in the creation of novel chimeric genes. In the case of breast cancer, although most individual tumors each have unique CNV landscape, the breakpoints, as measured over large datasets, appear to be non-randomly distributed in the genome. Breakpoints show a significant regional concentration at genomic loci spanning perhaps several megabases. The proximal cause of these breakpoint concentrations is a subject of speculation, but is, as yet, largely unknown. To shed light on this issue, we have performed a bio-statistical analysis on our previously published data for a set of 119 breast tumors and normal controls (Wiedswang et al., 2003), where each sample has both high-resolution CNV and methylation data. The method examined the distribution of closeness of breakpoint regions with differentially methylated regions (DMR), coupled with additional genomic parameters, such as repeat elements and designated "fragile sites" in the reference genome. Through this analysis, we have identified a set of 93 regional loci called breakpoint enriched DMR (BEDMRs) characterized by altered DNA methylation in cancer compared to normal cells that are associated with frequent breakpoint concentrations within a distance of 1 Mb. BEDMR loci are further associated with local hypomethylation (66%), concentrations of the Alu SINE repeats within 3 Mb (35% of the cases), and tend to occur near a number of cancer related genes such as the protocadherins, AKT1, DUB3, GAB2. Furthermore, BEDMRs seem to deregulate members of the histone gene family and chromatin remodeling factors, e.g., JMJD1B, which might affect the chromatin structure and disrupt coordinate signaling and repair. From this analysis we propose that preference for chromosomal breakpoints is related to genome structure coupled with alterations in DNA methylation and hence, chromatin structure, associated with tumorigenesis.

Keywords: Alu repeat element; DNA methylation; breast cancer; copy number variation; genome instability; multi-modal analysis.

Figure 1

Analysis method in order to find DMRs associated with BERs. ROMA genome-wide copy number profiles from breast tumors were combined to partition the genome into variable intervals of stable copy number state in which we estimate DNA methylation levels using MOMA measurements from tumor and Normal samples. A Hotelling’s T²-test is performed to identify significant DMRs. On the other track, the locations of BERs are obtained from the ROMA profiles and the list of both significant DMRs and BERs are further evaluated for statistical association (Figure 3).

Figure 2

Significant DMRs tend to co-localize with breakpoints enriched regions. The copy number profile of all 108 breast tumors is shown on the top track (CNV). The middle track (DMR) shows the amplitude of the DNA methylation level change compared to normal across genome. Hypo-methylated regions are assigned a negative score, defined as log10(p) while hyper-methylated regions take a score equal to −log(p). Significant DMRs are marked by peaks with a score greater than ±2. The bottom track (BER) shows the locations of breakpoint enriched regions. Breakpoint enriched DMRs (BEDMR), i.e., DMRs occurring in the vicinity of a BER are marked by vertical yellow lines and black arrows.

Figure 3

Differentially methylated regions co-localize with breakpoints enriched regions. (A) DMRs tend to be more proximal to BER than expected (B): the most significant distance of the association between DMRs and BERs occurs at a distance of 1 Mb (shown in yellow).

Figure 4

Localization of BEDMRs in the genome. BEDMRs tend to occur in genomic contexts. For example (A) 5q31.3 (PCDHA,B,G cluster) (B) 7p11.2 and 7q11.23 (EGFR, HIP1) (C) 11q14.1 (GAB2). (D) 16p13.3, 16p11.2, 16q24.2 (TSC2, FUS, P53TG3, CBFA2T3). These regions contain important cancer related genes and can be both deleted and hyper-methylated (A) or amplified and demethylated (C).

Figure 5

Breakpoint enriched DMR tend to co-localize with Alu enriched regions. The statistical evaluation shows that Alu enriched regions and BEDMRs co-localize within a distance of 3 MB (A,B).

Figure 6

Alu enrichment at BEDMRs. Locations of significant DMRs (top track) near breakpoint enriched regions (middle track) are compared with local Alu repeat enrichment (bottom track). About 33 of the 93 identified BEDMRs overlap with an Alu enriched region. We show here four interesting regions on 1p21.1 (A), 5q31.2-3 (B), 8q24.3 (C), and 16p13.3 (D).