Studies that use the genome sequence or gene expression patterns to draw biological conclusions are ten-a-penny, especially in the field of cancer. For example, specific gene expression signatures and fusions in DNA have been shown to be highly predictive of leukemia. However, have you ever considered whether the shape – rather than the sequence – of the genome is also altered in cancer? A new proof-of-principle study in Genome Biology shows that this just might be the case.
Given that chromatin proteins are becoming ever more linked to cancer, Josee Dostie and colleagues from McGill University, Canada, wondered whether cancer genomes may suffer not only from defects in their DNA sequences but also from changes in their shape.
To address this question, machine learning, in which a computer program uses a curated dataset to discover which characteristics are linked to which outcomes, was combined with chromatin conformation data generated by the ‘5C’ technique. These data identify contacts between sequences located some distance from each other along the DNA molecule.
Looking specifically at the shape of the HOXA gene cluster in leukemia cells, Dostie and colleagues found that the shape of the genome, is highly predictive of which type of leukemia a cancer cell originates from – not only whether the cell contains an MLL-fusion protein, but also to which gene MLL is fused, thereby defining the leukemia subtype.
Incredibly, the shape of the genome was just as good as, if not better than, gene expression in making predictions.
The mechanistic relevance of the distinctive shape of HOX gene clusters in each type of leukemia is not entirely clear and awaits further study. One interesting theory holds that 3D organization compartmentalizes genomes, which concentrate genomic loci for the purpose of preventing or promoting gene expression.
As is common in science, we have also yet to disentangle cause-and-effect: does the misshapen genome precipitate cancer? Or do the changes to DNA sequence and/or gene expression induce twists in the genome?
Whatever the biological truth underlying Dostie and colleagues’ findings, their study will hopefully provide thought-provoking fodder for cancer researchers. If the strong predictability shown here extends to a wider range of cancers, then perhaps the field will be encouraged to think more widely about changes to the genome that occur during tumorigenesis, rather than limiting research to a hunt for mutations in the DNA sequence or signatures of gene expression.
For more on the general biology of the 3D genome, please see the BioMed Central blog.
Written by Naomi Attar (@naomiattar), Senior Editor for Genome Biology.