Measuring chromosomal interactions in living cells

3C based high-throughput sequencing methods such as Hi-C, 5C and 4C have substantially contributed to our current understanding of genome folding. These techniques have been instrumental in demonstrating that mammalian chromosomes possess a rich hierarchy of structural layers at the heart of which topologically associating domains (TADs) stand out as preferential functional units in the genome. TADs have been suggested to establish the correct interaction patterns between regulatory sequences, supported by genetic studies where the deletion of boundary elements resulted in ectopic gene expression in the neighboring domain. Within TADs, looping interactions occur between regulatory sequences and convergent binding sites of the architectural protein CTCF, the latter as a consequence of loop extrusion by cohesin that is blocked by CTCF bound to DNA in a defined orientation. The dominant role of CTCF in loop formation is further highlighted by induced depletion experiments and targeted deletions and inversion of CTCF sites manifesting in loss of these interactions.
Despite these fundamental discoveries and their implications for transcriptional control by cis-regulatory sequences, 3C and derivatives are based on formaldehyde crosslinking and ligation, which have been often criticized as a source of important experimental bias. This has actually raised the question if structures detected by 3C methods do really exist in living cells. Based on discrepancies between 5C and DNA-FISH data, it was suggested that 3C based methods might not always capture spatial proximity or molecular-scale interactions, but rather detect DNA fragments which are hundreds of nanometers apart through crosslinking of macromolecular protein complexes between them. At the same time, it was debated whether capturing of ligation products might be variable depending on sequence context, therefore over- or underrepresenting some interactions detected in 3C based methods. Even though several other methods including native 4C/Hi-C, GAM and SPRITE have also detected chromatin compartmentalization, TADs and looping interactions, they still involve substantial biochemical manipulation of cells, notably either crosslinking or ligation. Importantly, many mechanistic models of chromosome folding rely on 3C based data, making the assumption that crosslinking frequency is proportional to absolute contact frequency. However, a formal proof of this is still missing.
In order to measure chromosomal contacts directly in living cells, without using chemical fixation nor ligation, I developed an alternative approach based on the DamID technique that exploits detection of ectopic adenine methylation by the bacterial methyltransferase Dam. In the original version of DamID, Dam is fused to a DNA binding protein of interest resulting in adenine methylation within GATC motifs in the neighborhood of the DNA binding sites. The methylation-sensitive restriction enzyme DpnI is then used to detect methylated GATCs followed by high throughput sequencing of the restriction sites. After mapping the reads and normalizing for non-specific methylation by freely diffusing Dam, the binding sites of the protein of interest can be detected genome wide.
I established a new modified version of this technique called DamC, where Dam is recruited in an inducible way to ectopically inserted Tet operators through fusion to the reverse tetracycline receptor. The detection of methylated DNA by high-throughput sequencing then allows to identify chromosomal contacts at high genomic resolution across hundreds of kilobases around viewpoints. Importantly, modeling of this process provides a theoretical framework showing that the experimental output of DamC is indeed proportional to chromosomal contact probabilities.
DamC provides the first crosslinking- and ligation-free validation of key structural features of mammalian chromosomes identified by 3C methods. It confirms the existence of TADs and CTCF loops as well as the scaling of contact probabilities measured in 4C and Hi-C, which supports the validity of physical models of chromosome folding based on 3C-based data. Finally, it demonstrates that ectopic insertion of CTCF sites can lead to the formation of new loops with endogenous CTCF-bound sequences. This shows that chromosome structure can be engineered by inserting short ectopic sequences that rewire interactions within TADs, opening interesting avenues for modifying gene expression by altering chromosomal interactions rather than regulatory DNA sequences directly.