Massively targeted evaluation of therapeutic CRISPR off-targets in cells

Abstract

Methods for sensitive and high-throughput evaluation of CRISPR RNA-guided nucleases (RGNs) off-targets (OTs) are essential for advancing RGN-based gene therapies. Here we report SURRO-seq for simultaneously evaluating thousands of therapeutic RGN OTs in cells. SURRO-seq captures RGN-induced indels in cells by pooled lentiviral OTs libraries and deep sequencing, an approach comparable and complementary to OTs detection by T7 endonuclease 1, GUIDE-seq, and CIRCLE-seq. Application of SURRO-seq to 8150 OTs from 110 therapeutic RGNs identifies significantly detectable indels in 783 OTs, of which 37 OTs are found in cancer genes and 23 OTs are further validated in five human cell lines by targeted amplicon sequencing. Finally, SURRO-seq reveals that thermodynamically stable wobble base pair (rG•dT) and free binding energy strongly affect RGN specificity. Our study emphasizes the necessity of thoroughly evaluating therapeutic RGN OTs to minimize inevitable off-target effects.

Fig. 1. Design of SURRO-seq.

An overview of the nine major steps of SURRO-seq is schematically presented. MM, mismatches; ON, on-target; OT, off-targets; PBS, primer binding sites; BB, BsmBI binding site; Sp, spacer; Scr, SpCas9 gRNA scaffold; BC, barcodes; PS, protospacer; PAM, protospacer adjacent motif; 4r, 4 bp downstream sequences; GGA, Golden Gate Assembly; MOI, multiplexity of infection; WT, wildtype; Results presented in Step 9 are indel frequencies captured by SURRO-seq for RGN VEGFA T3.

Fig. 2. Validation of RGN OTs detection between T7E1, GUIDE-seq, and/or CIRCLE-seq by SURRO-seq.

a Overview of RGN gRNAs and OTs selected for validated with SURRO-seq. b–d Comparison of the OT detection concordance rate between SURRO-seq and T7E1 (b), GUIDE-seq (c) and CIRCLE-seq (d). Numbers are total OTs for each RGN (upper) evaluated with the compared method and OTs agreed with SURRO-seq (lower). e Venn diagram comparison of OTs with significant significantly detectable off-targets (SURRO-seq) and OTs with deep sequencing reads detected by GUIDE-seq or CIRCLE-seq. Numbers are OT sites. f–g Comparison of VEGFA-T1 (f) and VEGFA-T2 (g) OT detections between T7E1, GUIDE-seq, CIRCLE-seq and SURRO-seq. Full results are showed in Supplementary Data 2. RGI, relative gel intensity; RE%, percentage of relative efficiency, calculated by % reads in OT per reads in ON; IF, indel frequency; I/T reads, indel/total reads; P values for comparison between SpCas9 and MOCK IF% are calculated with Benjamini and Hochberg (BH)-adjusted Fisher’s exact test (two-sided). *, represents OT with significantly detectable indels (adj. P value < 0.05, FC (IF% SpCas9/ IF% MOCK) > = 2). NS, represents OTs with not significantly detectable indels.

Fig. 3. High throughput evaluation of gene therapy RGN OTs with SURRO-seq.

a Overview of gene therapy RGN selection and number of OTs captured. b Quantification of indel frequencies for the 110 RGNs by SURRO-seq. R, RGN edited, M, MOCK control. c Overview of the number of RGNs OTs with not significantly detectable indel (NSOT, outer circle, adj. P value > 0.05 or FC (IF% SpCas9/ID MOCK < 2)), with significantly detectable indels (adj. P value < 0.05 and FC (IF% SpCas9/ID MOCK > = 2)) but low indel frequency (<3%, LIOT, middle circle), and significantly detectable indels with high indel frequency (> = 3%, HIOT, inner circle). P values are derived from Benjamini and Hochberg (BH)-adjusted Fisher’s exact test (two-sided). d Bar plot of total number (left) and fraction (right) of LIOTs and HIOTs for the RGNs. e Violin plot of the gene and genomics location of the HIOTs and indel frequency. OTs in cancer genes are highlighted in red. Data are presented as mean values +/− SD.

Fig. 4. Validation of endogenous OTs in five human cell lines by deep sequencing.

a Overview of RGNs, SURRO-seq identified Sig. OTs and NS. OTs selected for validation. b Schematic illustration of the experiments. RNP, ribonucleoprotein; HEK, HEK293T cell; Fib, human skin-derived fibroblasts. c Dot plot of on-target indel frequencies (indel reads/total reads %) in the CRISPR RNP edited cells. Indel frequency values were showed in Supplementary Data 4. d Heatmap summary of the RGN OTs evaluated by SURRO-seq and deep sequencing in five human cells lines. Indel frequency values were showed in Supplementary Data 4. e Example of indel frequencies of four OTs for RGN11189 measured by SURRO-seq and by deep sequencing of RGN edited human cell lines. f Scatter plots of indel frequencies for 7 on-target and 23 off-targets (referred to 4a), measured by SURRP-seq and by amplicon sequencing of the corresponding endogenous loci in RNP nucleofected cells (HEK293T and Fibroblasts). Extended plots for all cells can be found in Supplementary Data 4 (sheet 4.9).

Fig. 5. Effects of mismatch number, position and type on RGN off-target activity.

a Box-and-whisker plot of log 2 indel frequency for RGN OTs evaluated by SURRO-seq in LibB. Data are presented as values representing the median (line within the box), the interquartile range (length of the box), the 75 and the 25th percentiles (whiskers above and below the box) of the indel frequencies. Sites were grouped based the number of mismatches (MM), plotted according to significance and log10 adj. p-values (Benjamini and Hochberg (BH)-adjusted Fisher’s exact test (two-sided)). NS, RGN OTs with not significantly detectable indels; Sig. RGN OTs with significantly detectable indels. One mismatch (NS, N = 0 biologically independent RGN OTs; Sig, N = 6 biologically independent RGN OTs), two mismatches (NS, N = 26 biologically independent RGN OTs; Sig, N = 49 biologically independent RGN OTs), three mismatches (NS, N = 501 biologically independent RGN OTs; Sig, N = 140 biologically independent RGN OTs), and four mismatches (NS, N = 5860 biologically independent RGN OTs; Sig, N = 558 biologically independent RGN OTs). b Heatmap presentation of the fraction of mismatches occurred in each position of the gRNA for OTs in LibB, grouped based on total OTs, NSOTs, LIOTs and HIOTs. c Bar plot of appearance frequencies of each type of mismatches occurred in the different groups of RGN OTs in LibB. d–g Dot plots of indel frequencies for OTs with one mismatch measured in LibC. One-way pair-wise ANOVA analysis was performed for A type mismatches (d, N = 30 biologically independent mismatch sites), T type mismatches (e, N = 20 biologically independent mismatch sites), C type mismatches (f, N = 18 biologically independent mismatch sites), and G type mismatches (g, N = 28 biologically independent mismatch sites). Data are presented as mean values +/− SD. Indel frequency values can be found in Supplementary Data 5 (5.4–5.7).