Question 18 of 20 Advanced Interpretation
Why can a CRISPR screen call a gene essential when it is not?
Short answer
Cutting DNA is itself toxic, and more cuts mean more toxicity. In a region the cell line has amplified, every guide targeting it makes many double-strand breaks, arrests the cell cycle and reduces growth, whichever gene is targeted. The screen then scores whole amplified regions as essential.
Why
Aguirre and colleagues screened 33 cancer cell lines and found a strong relationship between gene copy number and loss of viability after editing. Guides against unexpressed genes and even intergenic sequence in amplified regions reduced proliferation, through a G2 cell cycle arrest, and the effect tracked the number of target sites rather than the gene hit. RNA interference screens in the same lines did not show this pattern, because they do not cut DNA.
The consequence is that a cancer line with an amplified region will appear to depend on everything in it. Meyers and colleagues built CERES to estimate and subtract this copy-number effect, which improved specificity. Practical handling has several parts.
Know the copy number of the line and flag guides in amplified regions. Use guides that are unique in the genome, since multi-target guides cut repeatedly. Apply a correction method.
Confirm hits with independent guides, with an orthogonal method such as RNA interference or a degradation tag, and with a rescue construct. The point generalizes. Any screen has technology-specific artifacts, and a hit list that has not been corrected for them is a list of candidates rather than a list of dependencies.
A strong answer covers
- Double-strand break toxicity scales with cut number
- Amplified regions score as essential regardless of gene
- Effect seen for unexpressed genes and intergenic sites
- Computational correction such as CERES
- Confirm with unique guides and orthogonal methods
Common mistakes
- Reading uncorrected screen scores in amplified cancer lines as true dependencies
- Using guides with many genomic matches
- Validating a hit only with more CRISPR guides and no orthogonal method
The interviewer might ask next
How would you confirm a dependency found in an amplified region of a cancer cell line?