multicrispr: gRNA design for prime editing and parallel targeting of thousands of targets

Abstract

Targeting the coding genome to introduce nucleotide deletions/insertions via the CRISPR/Cas9 technology has become a standard procedure. It has quickly spawned a multitude of methods such as prime editing, APEX proximity labeling, or homology directed repair, for which supporting bioinformatics tools are, however, lagging behind. New CRISPR/Cas9 applications often require specific gRNA design functionality, and a generic tool is critically missing. Here, we introduce multicrispr, an R/bioconductor tool, intended to design individual gRNAs and complex gRNA libraries. The package is easy to use; detects, scores, and filters gRNAs on both efficiency and specificity; visualizes and aggregates results per target or CRISPR/Cas9 sequence; and finally returns both genomic ranges and sequences of gRNAs. To be generic, multicrispr defines and implements a genomic arithmetic framework as a basis for facile adaptation to techniques recently introduced such as prime editing or yet to arise. Its performance and design concepts such as target set-specific filtering render multicrispr a tool of choice when dealing with screening-like approaches.

Figure 1.. Schematic representation of CRISPR/Cas9 application and arithmetic.

(A, B) illustrate the basic mechanism of CRISPR/Cas9 and prime editing. Both systems target a genomic region based on complementarity to a 20-nucleotide spacer sequence (when followed by NGG on the opposite strand), and both involve cutting the PAM–strand spacer after position 17 (double or single strand). (B) The prime editor (B) additionally enables editing of the sequence following nucleotide 17 through reverse transcription of a template (light blue, provided as a gRNA component), a process which is initiated through pairing of the primer binding site (another gRNA component) with the primer (a portion of the spacer on the PAM–strand). (C) A graphical overview of existing CRISPR/Cas9 gRNA design tools as provided by Torres-Perez et al (2019) and their classification. (D, E, F, G, H, I) genomic arithmetic as needed for individual CRISPR/Cas9 applications as indicated. Black lines represent the target range, orange arrows indicate the spacer sequences, blue arrows are PAM sequences, orange crosses depict Cas9 cut sites, and large arrows mark the search region for spacer–PAM sequences.

Figure 2.. multicrispr workflow and validation.

(A) Selection of supported CRISPR applications and workflow of multicrispr. (B) Overlap of prime editing spacer output of multicrispr and spacers used for the sickle cell locus in the HBB gene, the Tay–Sachs locus in the HEXA gene, and the prion disease locus in the PRNP gene, as given by Anzalone et al (2019). Scatter plots indicate scores and #mismatches given for all spacers found by multicrispr for the respective loci. (C) Overlap of multicrispr spacers and spacers used to block Oct4 TFBS [−151, −137] upstream of the Nanog gene, as used in Shariati et al (2019). Scatter plots indicate scores and #mismatches given for all spacers found by multicrispr for the respective loci. (D) Overlap of spacers identified with multicrispr for all Brunello exons (Doench et al, 2016). Density plot indicates scores for spacers specific for multicrispr (blue) and overlapping Brunello (red). Bar plots indicate # mismatches for these spacer sets as well. (E) Runtime comparison of gRNA design tools: the x-axis depicts the increasing number of input sequences and total bases, respectively, whereas the y-axis shows the total time needed by individual tools to design respective gRNAs on a log₁₀ scale in seconds. Colors represent individual tools. Box plots represent repetitive processing of each input file (n = 10) to control for variability in computing performance.

Figure 3.. Prime editing spacers and use cases of multicrispr.

(A) Off-target benchmarking was performed using 10 prime editing spacers (colored solid lines) to target the four main prime editing loci of Anzalone et al (2019), (colored vertical bars); the cystic fibrosis locus in the CFTR gene (red), the sickle cell anemia locus in the HBB gene (green), the Tay–Sachs disease locus in the HEXA gene (blue), and the prion disease locus in the PRNP gene (purple). Nicking spacers are shown with black lines for completeness but were not used for off-target benchmarking. Genomic coordinates are shown on the y-axis, and additional offsets are shown on the x-axis. (B) The parallel targeting of 1,974 binding sites of the transcription factor SRF. Boxes show results for one particular binding site (chr13:119991554-69:+), indicating the genomic locus on y-axis and range width on x. multicrispr finds eight spacers for this binding site. Three of them are target-specific (nonspecific spacers are faded out). Two of them are predicted to have a good targeting efficiency (Doench2016 is mapped to line thickness). The resulting GRanges object is presented as a table (T, target [mis]match counts; G, genome [mis]match counts; off, off-target counts, number 0–2 indicates number of mismatches). (C) Prime editing the prion disease locus in the PRNP gene. Primer binding site and reverse transcription template, jointly referred to as 3′ extension, are shown with dotted lines.

Figure S1.. FlashFry runtime.

Comparison of runtime between FlashFry (red) and multicrispr (blue) on 2,700 input sequences.