Efficient gene knockout and genetic interactions: the IN4MER CRISPR/Cas12a multiplex knockout platform

Abstract

Genetic interactions mediate the emergence of phenotype from genotype, but initial technologies for combinatorial genetic perturbation in mammalian cells suffer from inefficiency and are challenging to scale. Recent focus on paralog synthetic lethality in cancer cells offers an opportunity to evaluate different approaches and improve on the state of the art. Here we report a meta-analysis of CRISPR genetic interactions screens, identifying a candidate set of background-independent paralog synthetic lethals, and find that the Cas12a platform provides superior sensitivity and assay replicability. We demonstrate that Cas12a can independently target up to four genes from a single guide array, and we build on this knowledge by constructing a genome-scale library that expresses arrays of four guides per clone, a platform we call 'in4mer'. Our genome-scale human library, with only 49k clones, is substantially smaller than a typical CRISPR/Cas9 monogenic library while also targeting more than four thousand paralog pairs, triples, and quads. Proof of concept screens in four cell lines demonstrate discrimination of core and context-dependent essential genes similar to that of state-of-the-art CRISPR/Cas9 libraries, as well as detection of synthetic lethal and masking/buffering genetic interactions between paralogs of various family sizes, a capability not offered by any extant library. Importantly, the in4mer platform offers a fivefold reduction in the number of clones required to assay genetic interactions, dramatically improving the cost and effort required for these studies.

Figure 1.. Comparative analysis of synthetic lethality screens.

A) Different multiplex CRISPR perturbation methods applied to assay paralog synthetic lethality. B) Tested paralog pairs in each study. Upset plot shows the intersection of pairs across different studies. C) Quantifying synthetic lethality between paralog pairs. Single mutant fitness (SMF) is the mean log fold change of gRNAs that target an individual gene. Expected double mutant fitness (DMF) is calculated as the sum of SMF of gene 1 and gene 2. Delta Log Fold Change (dLFC) is the difference between observed and expected fold change and is used as a measure of genetic interaction. D) dLFC vs. Cohen’s D in one data set, A549 screen in Dede . E) Comparison of union of hits across all cell lines in each study. F) Jaccard coefficient comparing hits across all pairs of cell lines within each study. G) The “paralog score” is the weighted sum of hits minus the weighted sum of misses; i.e. where the gene pair is assayed but not a hit. Weights are the median of the platform-level Jaccard coefficients from (F). H). Histogram of paralog scores of 388 hits across all 5 studies. I) Histogram of paralog scores across 26 hits in >1 study. J) Thirteen candidate “paralog gold standards” with paralog score > 0.25 and hit in more than one study.

Figure 2.. Multiplexing beyond 2 guides with Cas12a.

(A) 7mer arrays were constructed with all combinations of either an essential or nonessential guide at each position (2^7=128 species), using the same DR sequences at each position, in three independent sets with unique gRNA sequences targeting the same genes at each position (n=384 total). (B) Guide sets were evenly represented in the combined pool before and after packaging and transduction (C) 7mer guide array representation is consistent across replicates and variation is consistent with high quality screens. (D) Fold change of guide arrays vs. number of essential guides on the array (n=384 arrays). (E) Fold change vs. position of essential guide on array, for all arrays encoding one essential guide (six nonessentials). (F) Fold change of guide arrays encoding one essential per array, forward vs. reverse orientation. Essential guides expressed at positions 6 and 7 deviate from the diagonal, indicating position-specific loss of editing. (G) A linear regression model that learns single gene knockout effects can be used to predict combinatorial target phenotype, an accurate null model for genetic interactions.

Figure 3.. In4mer platform for whole-genome screening.

(A) Inzolia human whole-genome library targets single genes and paralog pairs, triples, and quads with arrays of 4 Cas12a gRNA. Each gene or gene family is targeted by two arrays encoding the same gRNA in different order. Commercially synthesized oligo pools are cloned into the one component pRDA_550 lentiviral vector; schematic created in Biorender. (B-F) Screening in K562 CML cells and A459 lung cancer cells. (B) Read counts from the plasmid and experimental timepoints after lentiviral transduction. (C) Correlation of sample read counts. Endpoint replicates are highly correlated. (D) Fold change distributions of arrays targeting reference essential (red) and nonessential genes (blue) in four cell lines. (E) Precision/recall analysis from ranked mean fold change of arrays targeting each gene, calculated against reference essential and nonessential genes.

Figure 4.. Paralog synthetic lethality with Inzolia.

(A) Fold change vs. dLFC for >4,000 paralog families in Meljuso cells. (B) dLFC vs. Paralog Score from meta analysis of published paralog screens. Of 12 paralogs with score > 0.25 (red), 9 show dLFC < −1 in Meljuso cells. C) Fold change vs. paralog score in Meljuso cells. Most scored paralogs are essential, regardless of synthetic lethality. D) Selected synthetic lethals in Meljuso cells showing single and double knockout fitness phenotype. Bar chart, mean fold change. Points indicate fold change of single array of gRNA (mean of 2 replicates). E) Fraction of synthetic lethal paralogs by amino acid sequence similarity in Meljuso cells. F) Pathway activation by BCR-ABL1 fusion in K562 cells. Red, essential gene in in4mer screen; blue, nonessential; orange, synthetic lethal paralog pair. G) Fraction of dead cells, normalized to controls, for single, double, and triple knockouts of RAS genes in K562. KRAS-NRAS joint knockout shows increased cell death. ARCN1, control essential gene. ADH7, control nonessential gene. H) Single, double, and triple knockout phenotype of RTK/MAP kinase pathway genes in all four cell lines. White, target not in library.

Figure 5.. Synthetic chemogenetic interactions.

Meljuso cells were cultured in the presence of MEK inhibitor selumetinib and screened for chemogenetic interactions (A) DrugZ scores of single gene knockouts. Selected genes in the MAPK and Hippo pathways highlighted. (B) Selected GSEA results for gene sets conferring sensitivity (ERK signaling) or resistance (Hippo signaling) to MEKi. (C) Comparing DrugZ scores of paralogs (x-axis) vs. the most extreme Z score of the single gene knockout (y-axis) shows that most pairwise perturbagens yield similar phenotype as singletons. Outliers in red (synergistic) or blue (suppressor). (D) Synergistic and suppressor paralog knockouts from (C). Asterisk indicates functional buffering in the Hippo pathway, masking phenotype in monogenic knockout screens.

Figure 6.. Library size comparison.

(A) Representative Cas9 and Cas12a whole-genome libraries. Inzolia library targets 19k protein coding genes and additionally includes 9,822 guide arrays targeting paralog doubles, triples, and quads. (B) Five recent publications screening for genetic interactions between paralogs. Bar plot shows number of reagents per paralog pair tested, including single and double knockouts.