Identification of early replicating fragile sites that contribute to genome instability

Abstract

DNA double-strand breaks (DSBs) in B lymphocytes arise stochastically during replication or as a result of targeted DNA damage by activation-induced cytidine deaminase (AID). Here we identify recurrent, early replicating, and AID-independent DNA lesions, termed early replication fragile sites (ERFSs), by genome-wide localization of DNA repair proteins in B cells subjected to replication stress. ERFSs colocalize with highly expressed gene clusters and are enriched for repetitive elements and CpG dinucleotides. Although distinct from late-replicating common fragile sites (CFS), the stability of ERFSs and CFSs is similarly dependent on the replication-stress response kinase ATR. ERFSs break spontaneously during replication, but their fragility is increased by hydroxyurea, ATR inhibition, or deregulated c-Myc expression. Moreover, greater than 50% of recurrent amplifications/deletions in human diffuse large B cell lymphoma map to ERFSs. In summary, we have identified a source of spontaneous DNA lesions that drives instability at preferred genomic sites.

Figure 1. Mapping replication-induced DNA damage in murine B lymphocytes

(A) FACS analysis showing DNA content of freshly isolated and ex-vivo stimulated splenic murine B lymphocytes in the absence and presence of HU. (B) Experimental plan describing cell synchronization and isolation for samples used in ChIP-Seq and RNA-Seq experiments. (C) For each RPA-bound site in response to 10 mM HU (y axis), each column depicts the presence of RPA (left) and γ-H2AX (right) within a window centered on the RPA-bound sites. Color-map corresponds to binding intensities where “black” represents no binding. K-mean clustering algorithm was used to group the protein-bound sites. (D) RPA, SMC5 and BRCA1 co-occupy 2204 genomic regions in response to 10 mM HU. The plot in each column, from left to right, represents the pattern of RPA, SMC5 and BRCA1 genomic occupancy in response to HU centered on RPA-bound sites. K-mean clustering algorithm is used to group the protein-bound sites. (E) The Venn diagram shows the overlap of sites bound by RPA, SMC5, and BRCA1 in response to 10 mM HU. The total number of bound sites is indicated for each shared and unique area. (F) Relative frequency of ERFSs in classes of repetitive sequences is shown. Dashed line indicates the expected frequency based on the permutation model (*: enriched repetitive element classes, p < 1×10⁻³). (G) ERFSs are enriched in CpG islands. Total CpG island sequences in all the 2204 ERFSs as indicated by the crossed red point is compared to the permutation model as indicated by the gray points. Each gray point corresponds to the total CpG island sequences covered in an iteration of the permutation model. The box-plot depicts the quantiles of total CpG sequences based on the permutation model (p < 1×10⁻⁵). (H) ERFS genomic regions are transcriptionally active. The line-plot represents the average RNA tag count (loess-smoothed) in a genomic window around the center of the ERFSs. (I) ERFSs are enriched in transcriptionally active convergent and divergent gene pairs. Count of divergent/convergent gene pairs coinciding with ERFSs as indicated by the crossed red point is compared to the permutation model as indicated by the gray points. Each gray point corresponds to the total number of divergent/convergent gene pairs observed in an iteration of the permutation model. The box-plot depicts the quantiles of the total convergent/divergent transcript pair count based on the permutation model (p < 1×10⁻⁵). For definition of convergent/divergent gene pairs see methods. See also Figures S1, S3, S4.

Figure 2. ERFS “hotspots” associate with highly transcribed gene clusters

(A) Gene tracks represent, from the top, ERFS and ERFS hotspot demarcations, bindings of RPA, BRCA1, SMC5, γH2AX occupancy, and BrdU incorporation near the IKZF1 locus. The y-axis represents the total number of mapped reads per million of mapped reads (RPM) in 200 nucleotide windows (sliding-window smoothed). (B) Genome-wide map of 619 ERFS hotspots. Each hotspot is represented by a green dot on the ideograms. The top fifteen hotspots are color-coded in red. (C) Table of the top fifteen ERFS hotspots. ERFS hotspots are ordered based on a ranked statistics of RPA/SMC5/BRCA1 binding strength (see methods). The first column depicts a representative gene within the hotspot. A hotspot containing at least three genes is designated as a “gene-cluster.” A hotspot with a gene transcript value greater than 1 RPKM (reads per kilobase exon model per million mapped reads) is designated as transcribed. ERFS rearrangements in B cell cancers are listed in Table S2. ERFS is designated as “AID-target” according to (Chiarle et al., 2011; Klein et al., 2011). For complete definition of columns see methods. See also Tables S1 and S2.

Figure 3. ERFS break in response to HU

(A) Upper panel, diagram of FISH probes. Lower panel, representative DNA aberrations identified by FISH. Blue is DAPI-stained DNA, green represents the BAC probe (MHC, GIMAP, SWAP70, BACH2, IKZF1 or FOXP1) and red marks telomeric DNA. (B) HU induced aberrations were found at ERFSs but not at “cold sites” (CNTNAP4, SLITRK6) or CFSs (FRA16D, FRA3B). Quantitation of abnormalities from FISH analysis of untreated cells (blue bars) or cells treated with 10 mM HU (red bars). The percent aberrations specifically at the BAC probes relative to the total damage is plotted. (C) Abnormalities detected by FISH in untreated (blue bars) and 10 mM HU-treated (red bars) XRCC2^−/− cells. (D) Upper panel, diagram of FISH probes. Lower panel: representative metaphase showing a spontaneous break at the GIMAP locus in an XRCC2^−/− cell. (E) Quantitation of abnormalities detected by FISH in untreated (blue bars) and 0.2 μM aphidicolin-treated (red bars) WT cells. (F) Upper panel, diagram of FISH probes. Lower panel: representative metaphases showing aphidicolin-induced breaks at the FRA14A2 and FRA8E1 loci in WT cells. See also Figure S2 and Table S3.

Figure 4. ERFS break in response to ATR inhibition and high transcription

(A) Quantitation of aberrations observed by FISH in response to overnight exposure to 1 μM ATRi in WT (blue bars) and XRCC2^−/− cells (red bars). (B) Gene tracks represent, from the top, ERFS demarcation, and transcription measured by RNA-Seq in T and B cells at the region flanking SWAP70 locus. (C) Relative transcriptional activities of GIMAP and SWAP70 loci in B and T cells and their relation to the ERFS fragility. GIMAP and SWAP70 hotspots are shown in separate facets. The x-axis shows the cell lineage. The y-axis upward depicts the log10(RPKM) in B and T cells by dark and light reds, respectively; the y-axis downward depicts the quantitation of aberrations observed by FISH in response to overnight exposure to 1 μM ATRi in B and T cells in dark and light blue, respectively. (D) Relative SWAP70 mRNA abundance (measured across exon 4) normalized to β-actin in WT and SWAP70^−/− B cells. (E) Quantitation of aberrations in WT and SWAP70^−/− cells at the GIMAP and SWAP70 regions in response to 10 mM HU. See also Figure S5 and Table S3.

Figure 5. ERFS fragility is observed in response to oncogenic stress and in human cancer

(A) Western blot for phosphorylated p53 in c-myc and EV-infected XRCC2^−/− B cells. (B) Aberrations in c-myc-infected and EV-infected XRCC2^−/− B cells. (C) Aberrations in WT (blue bars) and AID^−/− B cells (red bars) treated with 1 μM ATRi. (D) Spontaneous chromosome breaks (blue bars) and translocations (red bars) at the IgH and BACH2 locus in IgκAID/53BP1^−/− B cells. (E) Normal chromosomes and a translocation of BACH2 ERFS (red) to the IgH locus (green) is shown. (F, G) ERFSs significantly overlap with MCRs detected in DLBCL. The Venn diagram shows the overlap of ERFSs with MCR found in DLBCL. The total number of regions is indicated for each shared and unique area and color-coded based on the region’s title. (G) Significance of correlation between the ERFSs and MCRs is evaluated relative to the permutation model and CFS. The percent increase in the overlap between the ERFSs and MCRs relative to the permutation model’s expectation (p< 1×10⁻⁴) and CFSs are shown in the left and right bar-graphs, respectively. (H) ERFSs are enriched for known cancer genes. The pie-chart shows the fraction of putative cancer genes (Bignell et al., 2010) associated with ERFSs (p<6×10⁻²⁰). See also Figure S6 and Tables S3 and S4.

Figure 6

Model for recurrent rearrangements in B cell lymphomas. AID is active in G1 (Petersen et al., 2001) and targets IgH and various oncogenes (eg. c-myc). Replication fork collapse at ERFSs in S phase occurs at preferential sites including various cancer-associated genes (eg. BCL2, BACH2). An AID-generated break might be passed from G1 to early S, where it meets an ERFS, which may eventually result in a translocation (left panel). Alternatively, an ERFS (bearing unresolved a replication intermediate of under-replicated DNA) might break in mitosis and then become permissive to translocate to an AID-induced DSB in the next G1 phase of the cell cycle (right).