Human DNA replication origins: reconciling disparate views – HUDROR

The goal of this project is to reach a rigorous, quantitative description of the human genome replication program. Understanding genome replication is essential as its perturbations are today recognized as a major threat to genome stability in stem cells, cancer, aging and diseases. However, the identification of human DNA replication origins has given rise to long-lasting controversies, due to variable results between techniques and laboratories. The aim of this project is to reconcile these disparate views and to establish a definitive picture of replication initiation in human cells, based on the generation of multiple high-quality datasets, integrative mathematical modelling and genetic dissection of the cis- and trans-acting determinants of replication initiation.
During previous ANR projects, we created novel methods to map replication genome-wide, at the single molecule (SM) level by high-throughput analysis in DNA nanochannels, and at the cell population level by sequencing newly replicated DNA (Repli-Seq) and Okazaki fragments (OK-Seq). We developed numerical and analytical models to exploit these huge datasets and we proposed a general, "cascade" model for replication, featuring efficient initiation at "master" origins (MaOris) in open chromatin followed by more dispersed, less efficient origin activation between MaOris. MaOris are broad (10-100 kb) but circumscribed initiation zones (IZs) that support a single initiation event per S phase. MaOris are non-transcribed, enhancer-rich segments often flanking topologically associating domains (TADs). MaOris only account for 10-30% of the total number of fired origins per S phase, suggesting "hidden", dispersed initiation. Correct modelling of all the data requires dispersed origins as a key ingredient. Positive evidence for dispersive initiation will necessitate high-throughput, SM analysis.
Origin mapping has relied on isolation of short bubble-containing fragments, short nascent single DNA strands (SNSs), or, more recently, nascent DNA synthesized in hydroxyurea-treated cells in the presence of 5-ethynyl-2'-deoxyuridine (EdU-seq-HU). Bubble-seq and EdU-seq-HU both reveal broad IZs, as OK-seq and SM data, but SNS-seq shows narrow peaks. Two hypotheses may explain this discrepancy. First, SNSs may be contaminated by non-origin DNA. Second, some SNSs may overreplicate and/or accumulate as aborted initiation intermediates, thus scoring high by SNS-seq, but low by bubble-seq or SM analysis due to template extrusion. Elongating SNS turn over faster and thus would score lower than aborted ones. To address these hypotheses, we will monitor initiation intermediates by multiple approaches. First, we are elaborating a new EdU-SNS isolation procedure, inspired from our success in complete purification of EdU-labelled Okazaki fragments for OK-seq. Crucially, SNS methods have never been tested in S. cerevisiae, where origins are known with precision. We will sequence yeast SNSs purified by several methods and assess their matching to known origins. The best method will then be applied to human cells. Second, we now perform automated, high-througput SM analysis of replicating DNA fibers in nanochannel arrays. This strategy will be applied to human cell-free extracts and intact cells to reveal human origins molecule by molecule, genome-wide, in an automated manner. These two independent approaches should confirm the location and reveal the internal structure of IZs, provide ampler evidence for dispersed initiation between IZs, and feed integrative mathematical models that will extract the kinetic parameters of the cascade model. A detailed understanding of DNA replication kinetics will allow us to decipher the influence of the chromatin 3D-structure on the replication program. Finally, we will use the gained information to dissect the cis- and trans-acting determinants of MaOris using genome editing by the CRISPR-Cas9 technology.