Comparative optimization of combinatorial CRISPR screens

Abstract

Combinatorial CRISPR technologies have emerged as a transformative approach to systematically probe genetic interactions and dependencies of redundant gene pairs. However, the performance of different functional genomic tools for multiplexing sgRNAs vary widely. Here, we generate and benchmark ten distinct pooled combinatorial CRISPR libraries targeting paralog pairs to optimize digenic knockout screens. Libraries composed of dual Streptococcus pyogenes Cas9 (spCas9), orthogonal spCas9 and Staphylococcus aureus (saCas9), and enhanced Cas12a from Acidaminococcus were evaluated. We demonstrate a combination of alternative tracrRNA sequences from spCas9 consistently show superior effect size and positional balance between the sgRNAs as a robust combinatorial approach to profile genetic interactions of multiple genes.

Fig. 1. Comparison of combinatorial CRISPR libraries.

a Three main combinatorial CRISPR systems evaluated: eight distinct combinations of alternative spCas9 tracrRNAs sequences (orange boxes represent libraries generated), enCas12a, and orthologous spCas9-saCas9 system. b Schematic representation of the CRISPR screen pipeline. Ten distinct libraries are screened in IPC298. spCas9-saCas9, enCas12a, VCR1-WCR3, and WCR2-WCR3 libraries are screened in PK1 and MELJUSO. c Schematic of sgRNA constant regions with indicated alterations compared to original spCas9 tracrRNA. Dark blue represents canonical sequence; light blue represents regions where insertions are introduced; orange represents base changes.

Fig. 2. Performance of combinatorial systems at the single-gene knockout level.

a–c The single-gene receiver operating characteristic (ROC)—area under the curve (AUC) curves derived from pan-essential and nonessential genes across distinct libraries in (a) IPC298, (b) MELJUSO, and (c) PK1. AUC values are indicated in parentheses. d Separation between gene scores for common essential genes and nonessential genes computed using null-normalized median difference (NNMD). e Correlation between the average LFC of right versus left sgRNAs coupled with sgAAVS1 cutting control in IPC298. Pearson’s correlation coefficient is indicated (r value) in black. Percentage of pan-essential genes with LFC less than −1 for sgRNAs in both right and left positions are indicated in red. Pan-essential genes are annotated by red dots. Source data are provided in the Source Data file.

Fig. 3. Positional effects and recombination rate on gene targeting.

a Evaluation of tracrRNA effect calculated by the difference in LFC between sgRNAs utilizing the VCR1 and WCR3 tracrRNAs from the VCR1-WCR3 and WCR3-VCR1 screens in IPC298. Analyses separated by promoters expressing the sgRNA. b Evaluation of the promoter effect by correlation between the LFC from the U6 and H1 promoter. For ten genes, same crRNA sequences are used on both left and right position to remove crRNA variability. Analyses separated by tracrRNA used. c Schematic illustration evaluating the recombination rate by next-generation sequencing (NGS) reads. crRNA and tracrRNA in both positions were sequenced by NextSeq paired-end 150 bp. Reads were either mapped to crRNA only or crRNA+tracrRNA (sgRNA). d Recombination rate calculated by percent of reads mapped between sgRNA to crRNA in pDNA for WCR2-WCR3 (green) and VCR1-WCR3 (blue). e LFC calculated by reads mapped to crRNA or sgRNA. The percentage of pan-essential genes with LFC less than −3 are indicated in red. Source data are provided in the Source Data File.

Fig. 4. Benchmarking of dual knockouts by pan-essential paralog pair.

a–c The ROC-AUC curves derived from pan-essential and nonessential paralog pairs across distinct libraries in (a) IPC298, (b) MELJUSO, and (c) PK1. AUC values are indicated in parentheses. d Correlation between the average LFC of dual knockouts with inverted positions of sgRNAs in IPC298. Pearson’s correlation coefficient is indicated (r value) in black. Percent of pan-essential paralog pairs with LFC less than −1 for both sgRNAs in positions are indicated in red. Pan-essential paralog pairs are annotated by red dots. Source data are provided in the Source Data File.

Fig. 5. Positive and negative regulators of the MAPK pathway exhibit synergistic dependencies in NRAS-mutant melanoma.

a Manhattan plot of FDRs corresponding to GEMINI synergy scores and color-coded by the LFC of dependencies for paralog pair knockouts in IPC298. The dotted line represents the FDR of 1e-3. b Scatter plot of GEMINI synergy score and LFC of dual knockout in IPC298. Curated paralog pairs of the MAPK pathway are annotated by red dots. Pan-essential paralog pairs or genes are annotated by blue or yellow dots, respectively. c LFC of single or combinatorial gene knockouts of paralog pairs associated with MAPK signaling pathway in IPC298 (n = 3 biological replicates; 6 independent sgRNAs per gene and 18 independent sgRNA combinations per gene pair). The centerline, lower hinge, and upper hinge correspond to the 50th, 25th, and 75th percentiles, respectively. The upper and lower whiskers extend from the upper and lower hinges to the largest and smallest values no further than 1.5 * IQR (interquartile range). Source data are provided in the Source Data File.