Key sequence features of CRISPR RNA for dual-guide CRISPR-Cas9 ribonucleoprotein complexes assembled with wild-type or HiFi Cas9

Abstract

Specific sequence features of the protospacer and protospacer-adjacent motif (PAM) are critical for efficient cleavage by CRISPR-Cas9, but current knowledge is largely derived from single-guide RNA (sgRNA) systems assessed in cultured cells. In this study, we sought to determine gRNA sequence features of a more native CRISPR-Cas9 ribonucleoprotein (RNP) complex with dual-guide RNAs (dgRNAs) composed of crRNA and tracrRNA, which has been used increasingly in recent CRISPR-Cas9 applications, particularly in zebrafish. Using both wild-type and HiFi SpCas9, we determined on-target cleavage efficiencies of 51 crRNAs in zebrafish embryos by assessing indel occurrence. Statistical analysis of these data identified novel position-specific mononucleotide features relevant to cleavage efficiencies throughout the protospacer sequence that may be unique to CRISPR-Cas9 RNPs pre-assembled with perfectly matched gRNAs. Overall features for wild-type Cas9 resembled those for HiFi Cas9, but specific differences were also observed. Mutational analysis of mononucleotide features confirmed their relevance to cleavage efficiencies. Moreover, the mononucleotide feature-based score, CRISPR-kp, correlated well with efficiencies of gRNAs reported in previous zebrafish RNP injection experiments, as well as independently tested crRNAs only in RNP format, but not with Cas9 mRNA co-injection. These findings will facilitate design of gRNA/crRNAs in genome editing applications, especially when using pre-assembled RNPs.

Figure 1.

Establishing adequate quantities of CRISPR-Cas9 RNP complexes for crRNA evaluation. (A, B) Indel frequencies (in %) generated by different amounts of RNP complex with WT Cas9. 1 nl and 2 nl of 1.5 μM RNP complex (1.5 and 3 fmol, respectively) assembled using WT Cas9 with the indicated crRNAs were microinjected into 1-cell stage embryos, and indel frequencies were assessed with TIDE (A) and ICE (B). (C, D) Indel frequencies (in %) generated by different amounts of RNP complex with HiFi Cas9. TIDE (C) and ICE (D) were used for the assessment.

Figure 2.

Cleavage efficiencies of the 51 crRNAs in the dgRNA RNP complex assembled with WT or HiFi Cas9. (A) A scatter plot of indel frequencies (in %) of the 51 crRNAs assessed with TIDE and ICE with the use of WT Cas9. (B) A scatter plot of indel frequencies of the 51 crRNAs assessed with TIDE and ICE using HiFi Cas9. (C) Comparison of indel frequencies generated with WT and HiFi Cas9 RNP complexes and assessed with ICE. Pearson correlation coefficients (r_p) are shown in each panel (A–C). (D) Indel frequencies (in %) assessed with ICE for the 51 crRNAs in dgRNA RNP complexes assembled with WT Cas9 and HiFi Cas9 are plotted on the left and right sides, respectively. Means of replicates are shown with the standard error bars. The experimental dataset for (A)–(D) is shown in Supplementary Table S4.

Figure 3.

Dose-dependent cleavage efficiencies of selected crRNAs from low, medium, and high activity groups. Two crRNAs each from high (A), medium (B) and low (C) activity groups were injected as RNP at four different amounts of 0.75, 1.5, 3 and 4.5 fmol. Indel frequencies (in %) assessed with ICE for RNP complexes assembled with WT Cas9 and HiFi Cas9 are plotted on the left and right sides, respectively. Means of replicates are shown with standard error bars. The experimental dataset is shown in Supplementary Table S5.

Figure 4.

Assessment of cleavage efficiency of crRNA in an in vitro cleavage assay. (A) Representative image of agarose gel electrophoresis for the in vitro cleavage assay using HiFi Cas9. Uncleaved substrate fragments and cleaved fragments are indicated by brackets. Cleavage efficiencies for 51 crRNAs were calculated by quantifying band intensities of uncleaved fragments, as summarized in Supplementary Table S6. -, no RNP control; AC, AL, etc., suffixes of the respective crRNA names. (B) There was only weak correlation between cleavage efficiencies obtained with the in vitro cleavage assay and the ICE indel frequencies obtained by injection of HiFi Cas9 RNP into embryos. A scatter plot with a Pearson correlation coefficient (r_p) is shown.

Figure 5.

Position-specific mononucleotide features determined with kpLogo. (A) Nucleotide frequency at each position of input sequences. Position 1 is assigned to the 5′-terminal nucleotide of the 20-nt protospacer target sequence (positions 1–20) and positions 21–23 correspond to PAM. (B) A kpLogo plot of mononucleotide features obtained for WT Cas9. Favored and disfavored nucleotides are shown in the upper and lower sides, respectively, with the height scaled to its P-value (–log₁₀ transformed) derived from Student's t tests of whether crRNAs with a particular nucleotide at a specific position were more or less efficient than other crRNAs. (C) A kpLogo plot of mononucleotide features obtained for HiFi Cas9. Significant nucleotides after Bonferroni correction done at each position are marked in (B) and (C) (**P < 0.01/4 = 0.0025; *P < 0.05/4 = 0.0125).

Figure 6.

Experimental validation of mononucleotide features associated with crRNA efficiency. (A) Establishment of the in vivo plasmid assay. Plasmid substrates containing target sequences for otx2b_AA/AB, pax2a_AJ and sox19a-KO_4 crRNAs were first injected into 1-cell stage zebrafish embryos and corresponding HiFi Cas9 RNP complexes were subsequently injected to ensure in vivo cleavage reaction. Plasmid and genomic DNAs were extracted from isolated nuclei of injected embryos at 24 hpf for the TIDE analysis. Plasmid substrates were cleaved in a similar manner to genome targets, albeit less efficiently. Means of two injection experiments are shown with standard errors. (B−D) In vivo plasmid assays for mutant crRNAs of the otx2b_AA (B), otx2b_AB (C), and pax2a_AJ (D) using HiFi Cas9. In target sequences on the plasmid, nucleotides with high P-values were mutated into nucleotides with the opposite effect, as indicated by red squares. H and L denote mutations with disfavored to favored nucleotides and vice versa, respectively. Means of two injection experiments are shown as values relative to wild-type crRNA (Mutant/Wild-type), which were obtained by dividing indel frequencies of mutant plasmid targets induced by mutant crRNAs by those for corresponding wild-type crRNAs. Standard errors are also shown. (E) In vivo plasmid assays for mutant crRNAs of sox19a-KO_4 using WT Cas9 and HiFi Cas9. Means of two (WT Cas9) or three (HiFi Cas9) injection experiments are shown as values relative to WT crRNA with standard errors.

Figure 7.

Predictive power of current CRISPR-Cas9 gRNA design tools for dgRNA/Cas9 RNPs. The heat map shows Spearman rank correlation coefficients (r_s) between prediction scores of major gRNA design tools for 51 crRNAs tested in this study (Supplementary Tables S3 and S7) and ICE indel frequency values for WT (left) and HiFi (right) Cas9 (Supplementary Table S4). Corresponding scatter plots are shown in Supplementary Figure 3. For DeepSpCas9variants, tRNA-N20 sgRNA values were used.

Figure 8.

Prediction of gRNA efficiency by CRISPR-kp, position-specific mononucleotide feature-based scoring. (A−F) Performance of CRISPR-kp on independent crRNA efficiency data in comparison with CRISPRscan and Rule Set 2. CRISPR-kp scores were obtained by summing the respective mononucleotide feature [−log₁₀(P-value)] values, in which negative values were given for disfavored mononucleotide features (Supplementary Table S8). Cleavage efficiencies of the second set of 27 crRNAs were determined using the dgRNA RNP complex assembled with WT or HiFi Cas9 with ICE (Supplementary Figure S4). Indel frequencies (in %) obtained with WT Cas9 (A−C) and HiFi Cas9 (D−F) were compared to scores obtained with CRISPR-kp (A,D), CRISPRscan (B,E) and Rule Set 2 (Doench/Fusi 2016) (C, F) by scatter plots. (G−I) Performance of CRISPR-kp on published data in comparison with CRISPRscan and Rule Set 2. Reported gRNA efficiencies assessed with RNP injections into zebrafish embryos (9,24,54,55) are compared with their CRISPR-kp (G), CRISPRscan (H), and Rule Set 2 (Doench/Fusi 2016) (I) scores. gRNA sequences and their CRISPR-kp scores are shown in Supplementary Table S8. Spearman correlation coefficients (r_s) are indicated in each panel. In panels G-I, Spearman correlation coefficients for data including Hoshijima's dgRNAs are shown in parentheses.

Figure 9.

Comparison of dgRNA efficiency between injection of pre-assembled RNP and a dgRNA/Cas9 mRNA mixture. (A) Selected crRNAs from the first set of 51 crRNAs were tested for their activity through injection of 3 fmol dgRNA with 200 pg HiFi Cas9 mRNA. Indel frequencies (in %) assessed with ICE were compared to those obtained with HiFi Cas9 RNP, using bar graphs. Means of replicates are shown with standard error bars. The experimental dataset for dgRNA/Cas9 mRNA mixture is shown in Supplementary Table S9. HiFi Cas9 RNP data were adopted from Figure 2. (B) A second set of 27 crRNAs were tested for their activity through injection of 3 fmol dgRNA with 200 pg HiFi Cas9 mRNA. Indel frequencies (in %) assessed with ICE were compared to those obtained with HiFi Cas9 RNP using bar graphs. Means of replicates are shown with standard error bars. The experimental dataset for dgRNA/Cas9 mRNA mixture is shown in Supplementary Table S9. HiFi Cas9 RNP data were adopted from Supplementary Figure S4. (C−E) Performance of CRISPR-kp on efficiency of 27 crRNAs in dgRNA/HiFi Cas9 mRNA format in comparison with CRISPRscan and Rule Set 2 scores. Indel frequencies (in %) obtained with dgRNA/HiFi Cas9 mRNA were compared to scores obtained with CRISPR-kp (C), CRISPRscan (D) and Rule Set 2 (Doench/Fusi 2016) (E) using scatter plots. Spearman correlation coefficients (r_s) are indicated in each panel.