DNA is packaged in the cell as chromatin, which folds into organized domains. Mapping of chromatin contacts in single cells sheds light on the dynamic evolution of these domains between cell divisions. See Article p.61
The division of a cell's replicated DNA into two daughter cells is arguably the most fundamental of all cellular processes. This mitotic division is followed by a period called interphase, in which chromosomes expand to allow gene expression, cell growth and DNA replication, before compacting again to lead into another round of mitosis. Methods for analysing the spatial organization of chromosomes, especially a technique called Hi-C, have transformed our understanding of 3D chromosome structure during interphase1,2,3. On page 61, Nagano et al.4 present a study of cell-to-cell variability in chromosome structure in dividing cells, revealing a continuous process of folding and reorganization throughout the mitotic cell cycle.
In Hi-C, physical contacts between different genomic regions are captured by isolating interacting fragments of chromatin (the DNA–RNA–protein complex in which chromosomes are packaged). Sequencing of these fragments produces a map of interactions across the genome that can be used to investigate the 3D folding of chromosomes. Most previous studies have used millions of cells mixed together. By contrast, Nagano et al. generated single-cell Hi-C maps for 1,992 individual mouse embryonic stem cells at all stages of the cell cycle. They then designed a computational strategy to assign each cell to a specific phase of the cycle using the Hi-C data.
Although cells in the same phase might have been expected to display very similar chromatin structures, individual cells could be ordered along a gradient of chromatin folding. Using this information, Nagano and colleagues built a cyclical model of chromosome changes, and validated this using an established technique for separating cells on the basis of their cell-cycle phase. The authors conclude that chromosome conformation is not stable at any stage. Rather, the chromosome-wide expansion that occurs after mitosis is followed by a period of further chromatin unfolding in some smaller regions, which lasts until DNA replication begins. At this point, chromatin begins to refold, and later re-compacts for the next mitosis. Mouse embryonic stem cells have an unusually fast cell cycle, which is characterized by a very short G1 (the phase after mitosis but before DNA replication), so it will be interesting to see whether this finding holds true for cells that have more-typical cycles.
Previous Hi-C studies in bulk cell populations have identified three key features of interphase chromatin folding: loops, topologically associating domains (TADs) and compartments5. Loops are created by an interaction between two small genomic regions typically separated by between 100 and 750 kilobases6, and the connecting regions are often associated with the DNA-binding protein CTCF. TADs are medium-sized genomic regions (100 kilobases to 2 megabases; ref. 7) that interact weakly with neighbouring regions, but strongly within themselves. Compartments are groups of TADs that either contain actively expressed genes or are mostly inactive. The relationships between these features are not entirely clear. For example, loops are often observed between the sequences at either end of a TAD, and CTCF binding is enriched at compartment boundaries. None of the features have so far been observed during mitosis1,3, and their behaviour during the rest of the cell cycle has not previously been explored in detail.
Nagano et al. report that chromatin loops are relatively stable throughout the cell cycle, whereas the borders of TADs weaken after G1, leading to an increase in contacts that cross the TAD boundary. Compartmentalization shows the opposite trend, increasing throughout the cell cycle and peaking just before mitosis (Fig. 1). These differences imply that the three levels of organization might have distinct origins and roles. This idea is supported by the fact that experimental depletion of different chromatin-associated proteins can have differential effects on each feature8,9,10,11. In addition, a recent Hi-C analysis12 of the genomes derived from sperm and eggs immediately after the cells fuse during fertilization showed that maternal chromatin has loops and TADs, but lacks compartments, implying that these features arise independently.
Single-cell Hi-C studies are of great value. But even if cells can be arranged in pseudotime, as achieved by Nagano and colleagues, they can provide only static snapshots of chromosome conformation. Unlocking the temporal dynamics of chromatin folding will require the analysis of live cells over time using microscopy.
An open question that could be resolved using live imaging is whether specific TADs inferred by Hi-C maps really exist in individual cells. TADs were first identified in Hi-C maps generated by averaging the contacts found in millions of cells. As such, they could arise from weak preferences summed across many cells, with real contacts in individual cells frequently crossing TAD boundaries. Single-cell Hi-C (which does measure contacts in individual cells) has not yet given a definitive answer: the current study provides evidence that TADs do indeed exist in individual cells, whereas other recent analyses12,13 have reached the opposite conclusion.
Single-cell Hi-C data sets yield the most information if computational modelling is used to infer full models of 3D chromosome structure for whole, individual cells. Nagano et al. are only the second group to achieve this, and do so for cells exiting mitosis13. A major technical hurdle has prevented the wider adoption of this potentially powerful computational modelling — cells normally contain maternal and paternal copies of each chromosome, and these have similar DNA sequences but are folded differently. Because most of the contacts identified by single-cell Hi-C cannot be unambiguously attributed to either parental genome, only a few contacts are left available for modelling approaches. The two studies that inferred full models avoided this issue by using specific mouse embryonic stem cells containing only one copy of each chromosome, and by examining only cells in which the DNA had not yet replicated.
There is much research into whether changes in chromatin folding can explain why thousands of variants in regions of the genome that do not encode proteins are linked to human disease. Using single-cell Hi-C in combination with 3D chromatin modelling could be a powerful way to unpick the variation between individual healthy and diseased cells. It remains to be seen whether improvements in sequencing technology or in the efficiency of single-cell Hi-C will eventually allow 3D modelling of human cells that have two copies of each chromosome. This advance would be invaluable in the study of medically relevant patient samples.