Selectivity of ORC binding sites and the relation to replication timing, fragile sites, and deletions in cancers

Abstract

The origin recognition complex (ORC) binds sites from which DNA replication is initiated. We address ORC binding selectivity in vivo by mapping ∼52,000 ORC2 binding sites throughout the human genome. The ORC binding profile is broader than those of sequence-specific transcription factors, suggesting that ORC is not bound or recruited to specific DNA sequences. Instead, ORC binds nonspecifically to open (DNase I-hypersensitive) regions containing active chromatin marks such as H3 acetylation and H3K4 methylation. ORC sites in early and late replicating regions have similar properties, but there are far more ORC sites in early replicating regions. This suggests that replication timing is due primarily to ORC density and stochastic firing of origins. Computational simulation of stochastic firing from identified ORC sites is in accord with replication timing data. Large genomic regions with a paucity of ORC sites are strongly associated with common fragile sites and recurrent deletions in cancers. We suggest that replication origins, replication timing, and replication-dependent chromosome breaks are determined primarily by the genomic distribution of activator proteins at enhancers and promoters. These activators recruit nucleosome-modifying complexes to create the appropriate chromatin structure that allows ORC binding and subsequent origin firing.

Keywords: DNA replication; ORC; chromatin; replication origins; replication timing.

Fig. 1.

ORC2 binding peaks are broader than those of typical transcription factors. (A) DNAs enriched in ORC2 ChIPs were analyzed by massive sequencing using the Illumina’s Solexa technology and DNA-tag enrichment at the LMNB2 replication origins visualized using IGV software. Input (blue), IgG (red), and ORC2 (black) are presented on similar y-axis scales. (B) Example genomic region showing binding peaks of ORC2 and transcription factors including E2F4 and JUN. Example ORC2 binding peaks are highlighted in gray boxes. (C) Distribution of peak sizes of ORC2 and example transcription factors. (D) Read distribution around summits of ORC2, GATA1, and JUN peaks.

Fig. 2.

Epigenetic features of ORC2 binding sites. (A) Distribution of ORC2, H3K27ac, DNase-seq, TBP, and Pol II reads around gene promoter regions. (B) Examples of ORC2 binding sites in the promoter of the CPSF1 gene and in an active enhancer region. (C) Examples of insulator and heterochromatin regions do not have ORC2 binding. (D) The DNase-seq and ChIP-seq read distribution around ORC2 binding sites, ranked by ORC2 binding levels.

Fig. 3.

ORC2 binds to active open chromatin regions. (A–D) Randomly selected 6,000 open chromatin regions with similar accessibility (A) but with or without ORC2 binding sites (B) were examined for H3K27ac (C) and H3K4me2 (D) levels in K562 cells. The Wilcoxon rank sum test P values comparing two groups of open chromatin regions are shown. (E–G) The AUC values measuring performances of logistic regression classifiers predicting ORC2 binding status, based on indicated training parameters. (H–K) Randomly selected 6,000 open chromatin regions with similar accessibility (H) but with or without ORC1 binding sites (I) were examined for H3K27ac (J) and H3K4me2 (K) levels in HeLa cells. The Wilcoxon rank sum test P values comparing two groups of open chromatin regions are shown.

Fig. 4.

Concordance among ORC2 binding sites, newly synthesized DNA domains measured by Repli-seq, and SNS sites. (A) An example genomic location showing ORC2 binding, replicating DNA domains in G1, S1, S2, S3, S4, and G2 phases and SNS sites. (B) Density of ORC2 binding sites in per 100 kb replicating DNA regions in G1, S1, S2, S3, S4, and G2 phases. (C) Distribution of distances between SNS sites and closest ORC2 binding sites. (D) Density of SNS sites in per 100 kb replicating DNA regions in G1, S1, S2, S3, S4, and G2 phases.

Fig. 5.

Lineage-specific early DNA replication timing is correlated with predicted ORC binding. (A) An example genomic region showing predicted lineage-specific ORC binding sites are correlated with differential early DNA replication. (B) Correlation between lineage-specific DNA replication and number of predicted ORC binding sites.

Fig. 6.

A simulated model for stochastic replication initiation. (A) Simulation based on the assumption that DNA replication initiates only at ∼52,000 ORC2 ChIP peaks. Shown is the correlation between replicating timing from simulation and that measured by Repli-seq choosing the value of 18 ORC sites firing per minute. (B and C) Simulations based on the assumption that DNA replication initiates at both ORC2 ChIP peaks and low-affinity, nontargeted ORC binding at nonpeak regions. Shown is the correlation between replicating timing from simulation and that measured by Repli-seq and with the indicated parameters for replication speed and ORC sites fired per minute.

Fig. 7.

Genomic regions lacking ORC2 binding sites are strongly linked to CFSs and recurrent deletions in cancer. (A) Snapshot of the genomic region of chromosome 2 encompassing fragile site FRA2T. The position of genes, the replication timing profile, as well as the position of ORC2 binding sites are reported. The location of the ORC2-poor regions, overlaying with a late replicating domain, is indicated. (B) Graph indicating the proportion of CFSs, aphidicolin-sensitive type, overlapping with ORC2-poor and Origin-poor regions of the genome. A detailed list is provided in SI Appendix, Table S3. (C) Graph indicating the proportion of recurrent deletions in cancer (43, 44) overlapping with ORC2-poor regions of the genome. A detailed list is provided in SI Appendix, Table S4.