Utilizing directed evolution to interrogate and optimize CRISPR/Cas guide RNA scaffolds

Abstract

CRISPR-based editing has revolutionized genome engineering despite the observation that many DNA sequences remain challenging to target. Unproductive interactions formed between the single guide RNA's (sgRNA) Cas9-binding scaffold domain and DNA-binding antisense domain are often responsible for such limited editing resolution. To bypass this limitation, we develop a functional SELEX (systematic evolution of ligands by exponential enrichment) approach, termed BLADE (binding and ligand activated directed evolution), to identify numerous, diverse sgRNA variants that bind Streptococcus pyogenes Cas9 and support DNA cleavage. These variants demonstrate surprising malleability in sgRNA sequence. We also observe that particular variants partner more effectively with specific DNA-binding antisense domains, yielding combinations with enhanced editing efficiencies at various target sites. Using molecular evolution, CRISPR-based systems could be created to efficiently edit even challenging DNA sequences making the genome more tractable to engineering. This selection approach will be valuable for generating sgRNAs with a range of useful activities.

Keywords: CRISPR; Cas9; DNA editing; SELEX; aptamer; guide RNA; molecular evolution.

Figure 1.. Synthesis of biased sgRNA library template.

A library based on the WT SpCas9 sgRNA sequence was synthesized with 5’ and 3’ constant regions (the Targeting Region and Loop 3, respectively, shown in black) and a 60 nucleotide partially randomized region (shown in gold). The random region was synthesized using phosphoramidite mixes such that each position in the randomized region had a 58% chance of being the WT nucleotide at that position and a 14% chance of being any of the other 3 nucleotides. The synthesized library was purified, extended, and transcribed as described in the Methods.

Figure 2.. TdT-based capture of sgRNA cleavage.

(A) Variant or WT sgRNAs are complexed to SpCas9 and a radiolabeled or fluorescently labeled DNA substrate. Upon cleavage by Cas9, TdT adds a poly(A) chain to the PAM-distal DNA’s 3’ end. A biotinylated Oligo(dT) probe binds the poly(A) tail, and the whole complex can be captured with magnetic streptavidin coated beads. Captured complexes are analyzed by scintillation counting or flow cytometry. (B) The WT sgRNA was complexed with Cas9 variants and then incubated with a radiolabeled substrate DNA and the components of an A-tailing assay as described in the Methods. As negative and positive controls, the WT sgRNA was complexed with inactive “dead” Cas9 or active Cas9 with (as indicated with ‘(TdT)’) or without TdT. Complexes containing cleaved DNA targets were bound by magnetic streptavidin beads, washed several times, and the amount of labeled DNA in the bead fraction and wash fractions was determined. The DNA only control does not contain either Cas9 or sgRNA and serves as a non-specific bead-binding background control. Error bars represent 4 independent replicates.

Figure 3.. BLADE SELEX generates cleavage capable sgRNA scaffolds.

(A) A library based on the WT SpCas9 sgRNA sequence was synthesized with 5’ and 3’ constant regions and a 60 nucleotide partially randomized region. BLADE SELEX was performed with this library and consisted of 3 phases: RNA pool-RNP formation, RNP binding to DNA substrate, and a TdT-based screen for cleavage. (B) DNA cleavage by the pools of sgRNA variants isolated following various rounds of BLADE SELEX (blue) was assessed by A-tailing and flow cytometry and compared to the WT sgRNA (red) complexed with inactive “dead” Cas9 or active Cas9. In this assay, functional RNPs cleave a Cy5-labeled DNA target, TdT adds a poly(A) tail to the newly exposed 3’ end-cleavage product allowing for capture by a biotinylated oligo(dT) probe and magnetic streptavidin beads.

Figure 4.. Sequence differences between selected sgRNAs and WT sgRNA.

The divergence of 647 selected cleavage-capable variant sgRNAs is shown compared to the WT sgRNA sequence. Positions where the base does not change from WT (“native”) are shown in black. Changes in base identify from the WT sequence are shown in color as indicated. Certain variants have an additional nucleotide between positions 71 and 72 or 73 as marked in gray

Figure 5.. Structure-based clustering of sgRNA variants displaying a range of cleavage efficiencies in vitro and editing efficiencies in cells.

(A) Selected sgRNAs were clustered based on their structures, as described in the Methods. A subcluster with > 50% of its members having mean in cell editing efficiency ≥ 50% is colored in green (WT subcluster). Leaves representing sgRNAs with mean in cell editing efficiency ≥ 50% are highlighted by yellow circles. The in vitro cleavage efficiency of sgRNAs is shown with a colormap with Complete cleavage (> 50%) in dark blue and Partial cleavage (< 50%) in light blue. The mean in cell editing is shown as a barplot, with efficiencies obtained from each replicate scattered with dots. Absence of dots implies that the sgRNA was not tested in cell. The number of differences to the WT obtained from the aligned scaffold sequences is shown with a stacked barplot, which differentiates mutations based on their role in interacting with Cas9. (B) Base pairing probability matrices of alternative functional sgRNA structures (upper triangle) and of the WT sgRNA structure (bottom triangle) are displayed as dotplots. The sgRNA IDs, corresponding to the IDs in the clustering (A), are given along the diagonal. The Tetraloop Stem, Stem 1 (S1), Stem 2 (S2), and Stem 3 (S3) previously reported are annotated for compatibility. Arrows point to the major changes between each sgRNA scaffold and the WT scaffold structure.

Figure 6.. Structure differences between the members of the WT subcluster.

(A) Base pairing probability matrices of sgRNA structures that are part of the WT subcluster (upper triangle) and of the WT sgRNA structure (bottom triangle) are displayed as dotplots. The sgRNA IDs are given along the diagonal and colored based on the efficiency category (low ≤ 10%, medium between 10%, and 50%, high ≥ 50%). The Tetraloop Stem, Stem 1 (S1), Stem 2 (S2), and Stem 3 (S3) previously reported ( are annotated for compatibility. Arrows point to the major changes between each sgRNA scaffold and the WT scaffold structure. (B) Structure logo of sgRNAs in the WT subcluster is shown. The structure logo annotates a sequence logo with mutual information of RNA base pairs, shown with the “M” symbol. Nucleotides that appear less than expected are shown upside down. (C) Structure-based sequence alignment of sgRNAs in the WT subcluster. The consensus structure is shown in the last row. Columns corresponding to base pairs in the consensus are colored based on the number of different base-pair types in the sequences.

Figure 7.. Selected variants demonstrate altered efficiencies when targeted to other sites.

The wild type (WT) and ten selected scaffolds were retargeted to five other target sites on the GFP gene, and GFP knockout was assessed in cells. GFP Target 1, the DNA site utilized during selection, is provided for comparison. The degree of GFP knockout is shown for each combination of scaffold variant with each DNA target domain. White colored boxes represent no knockout, black indicates 50% knockout, and red represents >80% reduction in GFP 10 expressing cells (significance values are reported in Data.S2; n= 2–5 independent experiments for each of the 66 GFP targeted sgRNAs in each of the two cell types).