Add this to the (increasingly long) list of papers whose basic research plans would once have gotten a net dropped over your head. It’s looking at variations in the BRCA1 gene, the one that is famously associated with breast cancer risk. There is no doubt at all that there are mutations in this gene that raise such risk – but at the same time, there are a lot of mutations out there. And it’s well known that all genes carry mutations if you go around carefully sequencing large numbers of individuals – it’s just that most of these mutations appear to be silent. So which are the silent mutations in BRCA1 and which are the alarming ones? There are plenty of single-nucleotide variations seen in it, and unfortunately, most of them are (1) rare by themselves and (2) of unknown risk.
BRCA1 (the gene, italicized) codes for BRCA1, a protein (un-italicized) that’s crucial in homology-directed DNA repair (the most common route of which is homologous recombination). This new paper uses a cell line that’s particularly susceptible to impairment of that pathway as a readout for how the mutated genes/proteins would function. Using the CRISPR system, everyone’s gene-editing friend, they targeted 13 of the gene’s 22 coding exons for study. These are the ones involved in its RING and BRCT domains, and some of those are already known to be hot spots, particularly exons 11-13. All the known single-nucleotide variations that are associated with increased cancer risk are in these regions. These exons were taken through saturation genome editing (SGE), which is just what it says: you make every change in every nucleotide base, and see what happens. They took it one exon at a time, and assayed the cells for how badly their growth (or survival) was hit in each variant.
That’s the aspect that (not so many years ago) would have caused that net to come down on you, because that represents a lot of mutations (3,893 of them, about 96% of all the possible ones in the targeted regions). Overall, about 73% of the variations were functional, 21% were nonfunctional, and 6% were intermediate. Looking more closely, the intermediate ones were enriched in missense mutations. Mutations that disrupted the known splice sites in the sequence gave proteins that were about 90% nonfunctional, as you’d imagine, while mutations deep in the introns or in the 5′ untranslated regions were almost never nonfunctional.
The correlation of what they found with the clinical picture for such mutations (where we know it) is very good indeed. There are 169 single-nucleotide variations in these regions that have been annotated in clinical practice as “deleterious”, and only 2 of these scored in this work as “functional”. And there are 22 variants that have been determined as clinically benign – of those, 20 scored as functional, one as intermediate, and one as nonfunctional. So the correlation is not perfect, but this assay is clearly on to something. (In fact, on closer inspection, in three cases where the two data sets completely disagree there’s reason to think that the clinical database is mistaken). And given the rate at which new variants are piling up (as more patients get sequenced), this sort of thing is probably the only way to keep even with the data. Modern sequencing methods will swamp us otherwise!
The next step, clearly, is to extend this approach to more of the BRCA1 sequence, and to try applying it to other targets. The first one should be pretty straightforward, but the second will depend on careful choice of the cell assay. Extending this to other cancer targets that are strongly affected through defects homologous DNA repair would be the most likely thing to try, you’d think, since the current cell line worked so well.