Quantification and correlative analysis of subclonal and non-clonal imperfect dsDNA break repair
Introduction
In many types of cancer, including breast, ovarian, prostate, and pancreatic, cancer cells often have defective dsDNA break repair due to some dysfunction in the homologous recombination repair (HRR) pathways. HRR is the cell’s most reliable method for repairing double-stranded DNA breaks, and when it fails, such as due to mutations in the BRCA1 and/or BRCA2 genes, DNA repair is redirected to more error-prone mechanisms. These mechanisms may introduce errors that are not just simple substitutions, but exhibit complex characteristics, distinct from replication errors, with deletions being the most common genomic scar. Deletion-containing mutational signatures have been identified previously in cancer tissues, for instance, in cancer cells with both copies of the HRR-associated genes BRCA1 and BRCA2 deactivated. However, for a deletion to be included in this mutational signature, the same deletion had to be clonally amplified, meaning it needed to be observed independently multiple times in sequencing reads. Long before a deletion is observed a certain number of times in sequencing results, the defective HRR may generate many more deletions that occur only once or twice in all the cells in an organism or an organ. As these somatic deletions are randomly and sparsely distributed in the genomic DNA, there are currently no efficient methods to identify them before a very small number of them get amplified by cancer growth.
Methods
We developed a computational method that identifies and quantifies these rare deletional events by applying hierarchical filtering to sequencing data acquired with standard methods. This method does not require any modifications to standard sequencing library preparation protocols. We tested our approach by analyzing hundreds of whole genome sequencing datasets from the International Cancer Genome Consortium (ICGC).
Results
After separating the signatures of non-clonal and sub-clonal deletions from the signatures of experimental systematic effects, the analysis revealed patterns of correlations which stratified patients into distinct groups and provided insight into the somatic evolution of cancers. One intriguing result is the presence of a non-clonal deletion signal in patients who do not carry mutated versions of BRCA1/2. While the presence of clonal deletions is a hallmark of BRCA1/2 inactivation, the non-clonal deletion signal is found only in a subgroup of these patients.
Conclusion
Our method provides a reliable approach for analyzing non-clonal deletions as rare as those happening in a single sequencing read, offering a strategy for tracking features of cancer evolution. It may also help to identify new therapeutic targets and new patient populations that could benefit from therapies targeting NHEJ and related DNA repair pathways.