Multiplex enCas12a screens detect functional buffering among paralogs otherwise masked in monogenic Cas9 knockout screens

Abstract

Background Pooled library CRISPR/Cas9 knockout screening across hundreds of cell lines has identified genes whose disruption leads to fitness defects, a critical step in identifying candidate cancer targets. However, the number of essential genes detected from these monogenic knockout screens is low compared to the number of constitutively expressed genes in a cell. Results Through a systematic analysis of screen data in cancer cell lines generated by the Cancer Dependency Map, we observe that half of all constitutively expressed genes are never detected in any CRISPR screen and that these never-essentials are highly enriched for paralogs. We investigated functional buffering among approximately 400 candidate paralog pairs using CRISPR/enCas12a dual-gene knockout screening in three cell lines. We observe 24 synthetic lethal paralog pairs that have escaped detection by monogenic knockout screens at stringent thresholds. Nineteen of 24 (79%) synthetic lethal interactions are present in at least two out of three cell lines and 14 of 24 (58%) are present in all three cell lines tested, including alternate subunits of stable protein complexes as well as functionally redundant enzymes. Conclusions Together, these observations strongly suggest that functionally redundant paralogs represent a targetable set of genetic dependencies that are systematically under-represented among cell-essential genes in monogenic CRISPR-based loss of function screens.

Comprehensive prediction of robust synthetic lethality between paralog pairs in cancer cell lines

“…However, the predictions made by our classifier significantly outperform these features alone. Furthermore, we show that our classifier, trained with data from single gene perturbation screens, can accurately predict the results of two independent combinatorial CRISPR screens of paralog pairs (Thompson et al , in press) and (Dede et al , 2020). Finally, we generate synthetic lethality predictions, with local explanations based on feature credit, for ~36.6k paralog pairs.…”

“…To evaluate the classifier using an orthogonal approach, we next evaluated its ability to predict the results of dual-gene knockout screens of paralog pairs. We obtained data from two independent combinatorial CRISPR screening efforts - one using a Cas9 based approach (Thompson et al , in press) and one using a Cas12a approach (Dede et al , 2020) (see Methods). For these screens individual and paired guide RNAs are used to determine the expected and observed fitness consequences, respectively, of perturbing each gene pair.…”

“…(A) ROC and PR curves showing the performance of our predictions (black) and sequence identity (orange) in distinguishing paralog pairs that were found to be SL in 2-3 vs.0 cell lines in genetic interaction screens carried out by Thompson et al (B) Same as (A) but for paralog pairs screened by (Dede et al , 2020). (C) Left: Venn diagram showing the number of paralog pairs that were identified as SL in the indicated cell lines by Thompson et al Right: Boxplots showing the distribution of prediction scores for pairs essential in zero, one, two or three cell lines.…”

“…Indirect evidence for the contribution of paralog buffering to genetic robustness is provided by the finding that, in both model organisms and cancer cell lines, paralog genes are less likely to be essential than genes with no identifiable paralogs (De Kegel & Ryan, 2019; Gu et al , 2003; Wang et al , 2015; Blomen et al , 2015; Dandage & Landry, 2019). More direct evidence for functional buffering is provided by observations of synthetic lethality between paralog pairs (DeLuna et al , 2008; Dean et al , 2008; VanderSluis et al , 2010; Dede et al , 2020; De Kegel & Ryan, 2019). Synthetic lethality occurs when the perturbation of two genes individually is well tolerated but their combined perturbation is lethal.…”

“…ARID1A mutant cell lines were found to be sensitive to loss of its paralog ARID1B through analysis of genome-wide shRNA screens of 165 tumour cell lines (Helming et al , 2014)) or through the rational testing of individual paralog pairs ( ME3 was tested for synthetic lethality with ME2 because of an understanding of their role in metabolism (Dey et al , 2017)). Recently, combinatorial screening using CRISPR approaches has emerged as an alternative approach for determining synthetic lethality between paralog pairs, allowing researchers to quantify synthetic lethal relationships between hundreds of pairs in pooled screens (Gonatopoulos-Pournatzis et al , 2020; Dede et al , 2020; Thompson et al , in press). However, there are tens of thousands of paralog pairs in the human genome (Zerbino et al , 2018) and consequently there is a need for approaches to prioritize those pairs most likely to be synthetic lethal.…”