A truncated reverse transcriptase enhances prime editing by split AAV vectors

Abstract

Prime editing is a new CRISPR-based, genome-editing technology that relies on the prime editor (PE), a fusion protein of Cas9-nickase and M-MLV reverse transcriptase (RT), and a prime editing guide RNA (pegRNA) that serves both to target PE to the desired genomic locus and to carry the edit to be introduced. Here, we make advancements to the RT moiety to improve prime editing efficiencies and truncations to mitigate issues with adeno-associated virus (AAV) viral vector size limitations, which currently do not support efficient delivery of the large prime editing components. These efforts include RT variant screening, codon optimization, and PE truncation by removal of the RNase H domain and further trimming. This led to a codon-optimized and size-minimized PE that has an expression advantage (1.4-fold) and size advantage (621 bp shorter). In addition, we optimize the split intein PE system and identify Rma-based Cas9 split sites (573-574 and 673-674) that combined with the truncated PE delivered by dual AAVs result in superior AAV titer and prime editing efficiency. We also show that this minimized PE gives rise to superior lentiviral vector titers (46-fold) over the regular PE in an all-in-one PE lentiviral vector. We finally deliver the minimized PE to mouse liver by dual AAV8 vectors and show up to 6% precise editing of the PCSK9 gene, thereby demonstrating the value of this truncated split PE system for in vivo applications.

Keywords: AAV vectors; CRISPR-Cas9; PASTE; gene editing; gene therapy; in vivo delivery; prime editing; reverse transcriptase.

Graphical abstract

Figure 1

RT variant screening and codon optimization (A) Schematic of the PE gene cassette employed in this study. PE consists of Cas9-nickase (Cas9 [H840A]) and M-MLV RT. The bipartite SV40 nuclear localization signals (bpNLSs) are marked in orange and a linker in gray. (B) Screening of prime editing activity of 11 RT variants codon optimized by GenScript showed activity only for XMRV RT is shown (all variant data shown in Figure S1). PE constructs were transfected with plasmids encoding pegRNA and ngRNA into HEK293T cells, and prime editing results were analyzed after 3 days. (C) The effect of RT codon optimization on PE expression using algorithms from different companies is shown. Same amounts of PE plasmids were transfected into HEK293T cells, and western blotting was conducted after 3 days. PE protein expression levels were normalized to β-actin protein levels, and wild-type (WT) PE protein expression level was arbitrarily set to 1.0 (indicated below the PE blot). The image shows a representative blot from two independent experiments with similar results. (D) Comparison of prime editing frequencies of the original PE and the codon-optimized PE^CO by plasmid delivery is shown. HEK293T cells were transfected with varying amounts of PE or PE^CO (1,500, 300, and 60 ng) together with fixed amounts of plasmid encoding pegRNA and ngRNA. After 3 days, PCR products were subjected to Sanger sequencing and ICE analysis to evaluate prime editing frequencies. (E) Comparison of prime editing frequencies of PE and PE^CO by all-RNA delivery is shown. PE mRNA and synthetic pegRNA and ngRNA were electroporated into HEK293T cells. Cells were subjected to ICE analysis 3 days post-transfection. Bars represent mean values ± SD, and all data points for individual replicates are shown. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

Figure 2

Generation of a minimal PE by truncating M-MLV RT (A) Schematic of PE^CO and truncated PE^CO variants used in this study. The RT is composed of a polymerase and an RNase H domain. Initially, a truncated PE was made with the RNase H domain deleted (471 bp) to form PE^CO-ΔR. Subsequently, serial 30-bp truncations at both ends were tested with the aim of generating a minimal PE^CO (PE^CO-Mini). (B) Minimizing M-MLV RT by deleting the RNase H domain (PE^CO-ΔR) and by further N- and C-terminal trimming is shown. The same molar amounts of PE^CO-ΔR and trimmed PE^CO-ΔR plasmids were transfected into HEK293T cells together with fixed amounts of plasmids encoding pegRNA and ngRNA. Cells were subjected to ICE analysis for evaluating prime editing activity 3 days post-transfection. (C) Combinatorial N- and C-terminal truncations were analyzed for prime editing efficiencies as in (B). Based on the activity, we selected RT-ΔR with ΔN60 + C90 as the minimal RT, which is 1,410 bp long. (D) Comparison of PE^CO and the two truncated variants PE^CO-ΔR and PE^CO-Mini performed as in (B) is shown. Bars represent mean values ± SD with all data points from independent experiments shown. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

Figure 3

The engineered minimal PE variant enhances prime editing when delivered by dual AAV vectors (A) Schematic of the split PE system where the full-length PE is reconstituted by intein protein-mediated trans-splicing. (B) Screening of split PE systems with different inteins and Cas9 split positions by plasmid transfection of HEK293T cells is shown. (C) Schematic of AAV vector genomes encoding the C-terminal part of the split PEs (PE^CO-C or PE^CO-Mini-C) is shown. ITRs, inverted terminal repeats; rbGlob, rabbit beta globin poly(A) signal. (D) Comparable activity of the split full-length and the two split truncated PE systems (PE^CO-ΔR split and PE^CO-Mini split) by plasmid transfection in HEK293T cells is shown. The single plasmid non-truncated PE^CO was included for comparison. (E) AAV6 vectors encoding the C-terminal part of the split PE^CO-Mini (PE^CO-Mini-C) were produced in parallel with vectors encoding the C-terminal part of the non-truncated PE^CO (PE^CO-C). AAV titers were determined by ddPCR with the titers of PE^CO-C set to 1. (F) HEK293T cells were transduced with equal amounts of AAV vectors (5 × 10⁵ vg/cell) of the generated split AAV vectors, and prime editing frequencies were evaluated 5 days post-transduction by Sanger sequencing and ICE analysis. Bars represent mean values ± SD. For (B), (D), and (E), data points are from independent experiments. For (F), each data point is an independent transduction with one AAV preparation. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

Figure 4

Delivery of PE^CO-Mini by dual AAVs for in vivo editing (A) Screening of pegRNAs and ngRNAs by plasmid transfection of the Hepa 1–6 cell line. Cells were co-transfected with a plasmid encoding a puromycin resistance gene (PAC), and cells were transiently selected for transfection-positive cells by applying puromycin for 2 days, starting at 24 h post-transfection. Prime editing frequencies were quantified by Sanger sequencing and ICE analysis. (B) Schematic figure of the two AAV genomes carrying the split PE^CO-Mini system is shown. (C) Schematic overview of animal studies is shown. (D) Quantification of editing efficiency in mouse livers is shown. Four weeks after AAV vector injection into six mice (three mice for each dose), mice livers were isolated for genomic DNA extraction. Editing in the PCSK9 target gene was quantified by NGS. Bars represent means ± SD with individual data points from biological replicates shown.