Cell and Disease Type Specific Fingerprints in the Mammalian Replication Program

The time at which DNA replicates during S-phase (replication timing; RT) is a precisely orchestrated, yet large-scale epigenetic property that offers an unparalleled window into the structure and regulation of the genome. While the timing program has been studied in many contexts and its significance is now well-established, the mechanisms that establish it have proven elusive, and recent studies have shed light on temporal and spatial aspects to replication control that were previously unanticipated. In this work, I and other members of the Gilbert laboratory worked to characterize the replication program in mammalian development, and described its structure, developmental regulation, and potential applications to medicine. In genome-wide studies of replication timing in mice, we found that replication timing profiles are both remarkably stable and cell type-specific, and are composed of coordinately regulated units (replication domains) that span one to several megabases. Changes in replication time typically occurred in 400-800kb units, and encompassed roughly 20% of the genome upon differentiation of embryonic stem cells (ESCs) to neural precursor cells (NPCs). These changes remarkably aligned domain timing values to genomic GC content and LINE-1 retrotransposon density. Consistent with previous results at individual loci, early replication was significantly (but not perfectly) associated with active transcription and active histone marks, and switches to later and earlier replication were accompanied by chromatin movement toward and away from the nuclear periphery respectively. Since H3K9 dimethylation was the only repressive histone mark with a moderate relationship to late replication, we next studied the regulation of the replication program with a cell line harboring an inducible conditional knockout of histone methyltransferase G9a. However, by comparison to the typical amount of timing differences between replicates (roughly 2-4% of the genome), we found no regions exhibiting unusually large timing changes upon G9a knockout in ESCs, or after differentiation of G9aCKO cells into NPCs. Nevertheless, many late-replicating H3K9me2-marked genes were transcriptionally upregulated, providing evidence that partially uncouples expression and histone mark changes from replication timing and nuclear location. To determine the extent of conservation between mouse and human replication program, we profiled several human ESC lines, differentiated NPCs, and lymphoblasts. Nearly all of the major properties from mouse were consistent in human cells, including domains sizes, timing changes in 400-800kb units, and relationships to activating histone marks and transcription. We also demonstrated that the replication program is well-conserved in regions syntenic between human and mouse. Importantly, hESCs aligned most closely not to mESCs, but to mouse EpiSCs, a more advanced population of cells derived from the epiblast and with comparatively limited plasticity. In studying the relation to histone marks we observed a peak of active marks 100kb within the border of most early replicating domains, but most remarkably (and unexpectedly), we found a correlation between replication timing profiles and Hi-C chromatin interactions stronger than any other genomic property, despite the Hi-C data deriving from an abstract computational model. As the robust cell type specificity of replication profiles suggested their potential for use in cell typing and studies of disease, I created (in collaboration with Jinfeng Zhang) a computational method to define "replication fingerprints"--collections of genomic regions with unique patterns of replication in defined collections of samples. Using these regions, 67/67 (34 mouse and 31 human) datasets could be correctly classified among 11 mouse and 9 human tissue types using a simple nearest-neighbor approach, and these results were confirmed through cross-validation and independent PCR assays. As a biological application, we created a fingerprint to isolate regions with common timing changes between pluripotent and committed cells, which revealed a conserved switch to later replication in the major histone H1 cluster that may help to explain the chromatin compaction previously observed during differentiation. To apply what we have learned about the replication program to the study of human disease, we collaborated with Drs. Bill Chang and Brian Druker to profile cell lines and pediatric patients with acute lymphoblastic leukemia (ALL). In contrast with normal B and T lymphocytes, leukemic samples displayed a high level of heterogeneity in replication profiles that offered intriguing potential for epigenetic fingerprints. Therefore, we applied the fingerprinting method to define regions with unique replication timing in high-risk patients and various genetic subtypes. To confirm the identity of leukemic samples and ability to detect known genetic lesions, we identified translocations and copy number variants in cell lines and patient samples known from CGH or karyotype information. These studies have opened paths to study less well-characterized subtypes of leukemia such as AML, which we plan to explore in future work.