Multiplex base- and prime-editing with drive-and-process CRISPR arrays

Abstract

Current base- and prime-editing technologies lack efficient strategies to edit multiple genomic loci simultaneously, limiting their applications in complex genomics and polygenic diseases. Here, we describe drive-and-process (DAP) CRISPR array architectures for multiplex base-editing (MBE) and multiplex prime-editing (MPE) in human cells. We leverage tRNA as the RNA polymerase III promoter to drive the expression of tandemly assembled tRNA-guide RNA (gRNA) arrays, of which the individual gRNAs are released by the cellular endogenous tRNA processing machinery. We engineer a 75-nt human cysteine tRNA (hCtRNA) for the DAP array, achieving up to 31-loci MBE and up to 3-loci MPE. By applying MBE or MPE elements for deliveries via adeno-associated virus (AAV) and lentivirus, we demonstrate simultaneous editing of multiple disease-relevant genomic loci. Our work streamlines the expression and processing of gRNAs on a single array and establishes efficient MBE and MPE strategies for biomedical research and therapeutic applications.

Fig. 1. Cas12a multiplex strategy for MBE.

Heat maps show 3 independent biological replicates at each condition, counting the most PAM-distal position as 1. The dashed line in scatters dot plots shows the mean value of all desired BE efficiencies within the displayed activity windows of each site. Single-site, using CBE-dCas12a or ABE-dCas12a with the relevant hU6-driven gRNA for single-site base editing. Detailed constructs of CBE-dCas12a (2xUGI-rAPOBEC1-dCas12a) and ABE-dCas12a (TadA-8e-dCas12a) are available in Supplementary Table 1. Protospacer and amplicon information can be found in Supplementary Table 5.

Fig. 2. Development of DAP strategy for MBE.

a Schematic of 2-loci hCtRNA-gRNA array driven by hU6 and using hCtRNA as Pol III promoter. hU6, human U6 promoter (RNA Pol III). b 5′ leader engineering of hCtRNA. c Adding 3 nt 5′ leader to mature hCtRNA (72 nt) showing efficient 2-loci hU6-driven MBE. P, pooled gRNAs; S, single gRNA; CBE used, NBE4max (R33A + K34A). d, Secondary structure of hCtRNA, gray shadow showing 3 nt 5′ leader. e, hCtRNA in comparison with hU6 as gRNA promoter for single-guide BE. CBE used, NBE4max (R33A + K34A); ABE used, ABE7.10 (F148A). f, g Efficient 5-loci MBE using hCtRNA-M without additional upstream RNA Pol II (EF1a) or Pol III (hU6). hU6-M w/o 5′ leader, using mature hCtRNA (72 nt) in the hU6-M array. bGHpA, bovine growth hormone poly-A termination signal. h Comparison of nSpCas9-MBE and dLbCas12a-MBE. Data of dLbCas12a-MBE is from Fig. 1. Heat maps show 3 independent biological replicates at each condition, counting 1 as the most PAM-distal position. Dashed line showing mean value. c, e, g were analyzed using Sanger sequencing, h was analyzed using NGS. Error bars represent mean ± s.e.m. from n = 3 replicates (e, h use unpaired two-tailed t-test; g uses unpaired multiple t-test; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). All tests were performed in HEK293T cells. Multiple t-tests information is available in Supplementary Table 4. Detailed protospacers, edited bases, and amplicons of the relevant figures are available in Supplementary Table 5.

Fig. 3. DAP strategy enables large-scale MBE with minimal off-target effect.

a–c Schematic of RNA-seq results mapped onto the hCtRNA-gRNA multiplex arrays. d CBE multiplex editing of 10 loci using array shown in a with dash line indicating 50% editing. CBE architecture of each variant in d is shown in Supplementary Fig 12. e ABE8e multiplex editing of 20 loci using array shown in b with dash line indicating 40% editing. f 31-loci MBE using dual-deaminase base editor ACME and two 16-loci arrays in c, with dash line indicating 40% editing. gRNA 14 and gRNA 28 are identical and exhibit the same results in f. g Violin plot showing the value distribution of all replicates of d with solid lines representing quartiles, 57.4%, 67.3%, and dash line indicating median, 62.3%. h Violin plot showing the value distribution of all replicates in 20-loci MBE, with solid lines representing quartiles, 37.78%, 60.69%, and dash line indicating median, 48.68%. i Violin plot showing the distribution of all replicates of 31-loci MBE, with solid lines representing quartiles, 44.61%, 59.86%, and dash line indicating median, 50.12%. All tests were performed in HEK293T cells and analyzed using NGS. j Cas9-independent off-target editing evaluation of MBE at five Sa (SaCas9) sites (1,2,4,5,6) used in orthogonal R-loop assay. k Cas9-dependent off-target editing evaluation of MBE at seven GUIDE-seq identified EMX1 off-target sites. Detailed protospacers, edited bases, amplicons, primers, and plasmid maps of the relevant figures are available in Supplementary Tables 1 and 5. Error bars represent mean ± s.e.m. from n = 3 replicates.

Fig. 4. MPE with DAP arrays.

a Schematic of MPE development. The size of PE3 gRNA arrays and single-site MPE array were compared using gRNA sequences designed for RNF2 (+5G to T). b Efficient single-site and 3-loci MPE. The w/o I and w/ I arrays were compared for 3-loci MPE. c Schematic illustration of the observed low 3-loci MPE efficiency (w/o I) in b. After the endogenous processing of the MPE array without interval sequence (w/o I), partial or complete 5′ leader sequences will remain at the 3′ end of pegRNA, which causes undesired primer binding and decreases the PE efficiencies. By adding the interval sequence (w/ I) at the 3′ end of pegRNA, the poly-T termination signal will isolate the original 3′ end of pegRNA from additional sequences, achieving desired binding and expected PE efficiencies. All tests were performed in HEK293T cells and analyzed using NGS. Detailed protospacers, edited bases, amplicons, primers, and plasmid maps of the relevant figures are available in Supplementary Tables 1 and 5. Error bars in b represent mean ± s.e.m. from n = 3 replicates (unpaired two-tailed t-test; ns not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). P = 0.0146 (prime edited value comparison between MPE and PE3 for HEK3 (+5G to C)). P = 0.0444 (prime edited value comparison between MPE and PE3 for RNF2 (+5G to T)). P = 0.0166 (Indel value comparison between MPE and PE3 for FANCF (+5G to T)). P = 1.87E-06 (prime edited value comparison between w/o I and w/ I for HEK3 (+5G to C)). P = 0.0113 (Indel value comparion between w/o I and w/ I for HEK3 (+5G to C)). P = 4.44E−06 (prime edited value comparison between w/o I and w/ I for RNF2 (+5G to T)). P = 0.0006 (Indel value comparison between w/o I and w/ I for RNF2 (+5G to T)). P = 0.002 (Indel value comparison between w/o I and w/ I for FANCF (+5G to T)).

Fig. 5. MBE of polygenic disease-relevant loci, adapting MBE and MPE to viral delivery for collective installation of disease-suppressing mutations.

a MBE of polygenic disease-relevant loci in HEK293T cells. b Dual AAV vectors encoding optimized split ACME and hCtRNA-M array. SFFV, spleen focus-forming virus promoter; B and H represent gRNAs of BCL11A target 2 and HBG1/2 protospacer, respectively (Supplementary Fig 22); GY, extein amino acid residues Gly (G) Tyr (Y); SSS, extein amino acid residues Ser (S) Ser (S) Ser (S); gp41-1-N/C, N- or C-terminal gp41-1 trans-splicing intein; ITR, inverted terminal repeat. c Heat maps showing MBE in HEK293T cells using dual AAV vectors depicted in b, packaged as AAV1 pseudotype. Values represent the mean of three biological replicates. d Schematic of lentiviral Cas9 constructs with the hCtRNA-M cassette in forward and reverse orientations. LTR, long terminal repeat; B and H represent gRNAs of BCL11A target 1 and HBG1/2 protospacer (Supplementary Fig 22); EFS, elongation factor-1α short promoter. e Comparing multiplex editing between the two lentiviral constructs depicted in d, showing (+) strand RNA copy with a recognizable forward hCtRNA-M array is susceptible to RNase P and Z. f, g Optimization of lentiviral transduction for multiplex Cas9 nuclease editing in HEK293T cells (Methods). Lentiviruses containing reverse hCtRNA-M array depicted in d were used. h Multiplex Cas9 nuclease editing of BCL11A and HBG1/2 in HEK293T, K562, Jurkat cells, and primary human CD34 + HSPC, followed by 27 days of puromycin selection for three cell lines and 20 days for HSPC. i Schematic of lentiviral MPE constructs non-susceptible to endogenous tRNA processing. N, nicking gRNA; P, pegRNA. j The representative on-target MPE editing reads with desired deletion editing highlighted in red. k MPE at BCL11A site in HEK293T, K562 cells and HSPC using lentivirus encoding PE2 and a reverse single-site MPE array assembling pegRNA and nicking gRNA, followed by 20 days of puromycin selection. Detailed protospacers, edited bases, amplicons, primers, and plasmid maps of the relevant figures are available in Supplementary Tables 1 and 5. Error bars represent mean ± s.e.m. from n = 3 replicates.