SunTag-PE: a modular prime editing system enables versatile and efficient genome editing

N-1×GCN4 SunTag-PE was the most efficient configuration

We initially designed SunTag-PE2 by fusing the scFv to the N terminus of Moloney murine leukemia virus reverse transcriptase (M-MLV RT)¹⁹ and fusing different copies of GCN4 to the N terminus or C terminus of nCas9-H840A²⁰. In SunTag-PE system, the scFv-RT was recruited by n×GCN4-nCas9 (the n in n×GCN4 = 1, 2, 3, 5, 10). We designated them as SunTag-PE (n×GCN4-nCas9) and PE-SunTag (nCas9-n×GCN4) based on the domain order (Fig. 1a, b).

a SunTag-PE consists of GCN4 fused to the N-terminus of nCas9. b PE-SunTag consists of GCN4 fused to the C-terminus of nCas9. c Two-plasmid system of SunTag-PE, one plasmid expressing pegRNA and scFv-RT and the other expressing n×GCN4 tethered nCas9. d Different configurations of SunTag-PE and PE-SunTag (the copy number of GCN4 was 1, 2, 3, 5, and 10). e Comparison of CTT insertion efficiencies of different SunTag-PE2 and PE2-SunTag configurations in HEK3 locus by deep sequencing. f Comparison of CTT insertion efficiencies of SunTag-PE2 with N-1×GCN4, deficient SunTag-PE2, Split-PE2, MS2-PE2, and canonical fused PE2 in HEK3 locus. Data and error bars indicate the mean and s.d. of three independent biological replicates. Negative controls represent groups only with pegRNAs, untreated with PE components. The numbers above the bars represent fold change between different treatment groups.

Based on the above design principles and reasonable size distribution, we constructed a two-plasmid system, one plasmid expressing pegRNA and scFv-RT (Fig. 1c), and the other expressing n×GCN4 tethered nCas9 (Fig. 1d). To test its feasibility in PE editing, we first co-transfected the above plasmids into HEK293T cells and HeLa cells to introduce a 3-nucleotide (nt) CTT insertion in HEK3 endogenous locus. The HEK293T cells’ editing efficiency and indel rate were detected by Sanger sequencing (Supplementary Fig. 1a) and deep sequencing (Fig. 1e). We observed that the editing efficiency was the highest when there was only one copy of GCN4 tethered to nCas9. The editing efficiency of GCN4 tethered to the N-terminus of nCas9 was generally better than that of the C-terminus. As shown in Fig. 1e, when n×GCN4 was configured at the N-terminal of nCas9, the CTT insertion average rate was 13.8% for 1×GCN4, 11.44% for 2×GCN4, 8.3% for 3×GCN4, 10.16% for 5×GCN4, and 9.64% for 10×GCN4. When n×GCN4 was configured at the C-terminal of nCas9, the CTT insertion average rate was 12.44% for 1×GCN4, 9.77% for 2×GCN4, 8.87% for 3×GCN4, 8.25% for 5×GCN4, and 8.27% for 10×GCN4. The GCN4 tethered to the N-terminus of nCas9 with fold-changes of 1.11× for 1×GCN4, 1.17× for 2×GCN4, 1.23× for 5×GCN4, and 1.17× for 10×GCN4 than C-terminus. Consistently, we also observed the highest editing efficiency with N-terminal 1×GCN4 in HeLa cells (Supplementary Fig. 1b). The above results were contrary to the previous understandings^17,20 that the more copies of GCN4, the better efficiency. This phenomenon may be caused by the steric hindrance of neighboring peptide binding sites^17,21.

SunTag-PE2 exhibited higher editing efficiency than previous different split Pes

After confirming that the most potent SunTag-tethered configuration was N-1×GCN4-nCas9 (hereinafter referred to as SunTag-PE), we selected it for further investigation. We first compared the editing efficiency of SunTag-PE with canonical fused-PE³, MS2-PE¹⁵, and Split-PE (untethered RT)^15,16 in the HEK3 locus of HEK293T cells in PE2 form. The data showed that, SunTag-PE2 achieved comparable levels of CTT insertion editing as canonical fused-PE2 (up to 14.58%) and significantly higher than those of Split-PE2 (7.63%) and MS2-PE2 (8.57%) with no increase in indel byproducts (Fig. 1f and Supplementary Fig. 2). Furthermore, we confirmed that the predominant editing ability of SunTag-PE2 was derived from the enrichment effect of GCN4-scFv by comparing with the deficient SunTag-PE2 (nCas9 + scFv-RT), which was lack of GCN4 but otherwise identical to SunTag-PE2 (Fig. 1f). Of note, we observed that deficient SunTag-PE2 also achieved comparable editing levels to Split-PE2, consistent with previous studies that RT may be able to function as untethered form refs. ^15,16.

Efficient editing by SunTag-ePE2 at endogenous and exogenous sites

To further improve the editing efficiency of SunTag-PE2, we constructed SunTag-ePE2 by using engineered pegRNAs (epegRNAs) containing a 3′ evoPreQ1 motif²². We tested its editing efficiency at four endogenous genomic loci (HEK3, EMX1, FANCF, and RNF2) to install nucleotide transversion, insertion, and deletion in HEK293T cells. We observed 1.5-fold increased efficiency of SunTag-ePE2 compared to SunTag-PE2 in installing CTT insertion in HEK3 locus (21.86% vs. 14.58%, Figs. 1e and 2a; Supplementary Fig. 3). As expected, SunTag-ePE2 could efficiently install desired mutations across all four genomic sites tested with editing level comparable to ePE2 and no increase in indel byproducts (Fig. 2a–d).

Comparison of SunTag-ePE2 and ePE2 at four endogenous loci, including HEK3 locus (a), EMX1 locus (b), FANCF locus (c), and RNF2 locus (d). Each locus contains three types of prime editing, nucleotide transversion, insertion, and deletion. Negative controls represent SunTag-ePE2 or ePE2 components without pegRNAs. Off-target analysis of SunTag-ePE2 and Fused-ePE2 systems at the EMX1 (e) and FANCF (f) target sites. The percentage of total sequencing reads with desired edits or indels is shown for the on-target and off-target site (OT-1 to OT-3). Spacer sequences for the on-target sites are displayed, with mismatched bases highlighted in red. OT-1: off target site 1; OT-2: off target site 2; OT-3: off target site 3. g The illustration of GFP mutant reporter. It contains a premature stop codon resulting from a 4 nt deletion frameshift mutation in the GFP coding sequence, indicated by red pentagram. Once the GTTC sequence is inserted to the mutation site, it will convert into GFP. h, i The percentage of GFP-positive cells after SunTag-ePE2 and ePE2 editing. Data and error bars indicate the mean and s.d. of three independent biological replicates. The numbers above the bars represent fold change between different treatment groups.

To determine whether SunTag-PE2 alters off-target editing levels, we first assessed the off-target activity of SpCas9 at the EMX1 and FANCF loci using GUIDE-seq. As shown in Supplementary Fig. 4, the results identified 10 off-target sites for EMX1 and 3 off-target sites for FANCF. We selected the top 3 off-target sgRNAs from EMX1 and the top 1 off-target sgRNA from FANCF to serve as off-target detection sites for the spacer regions of the SunTag-PE2 epegRNAs. Deep sequencing results revealed that both SunTag-ePE2 and Fused-ePE2 systems exhibited extremely low off-target primer editing efficiencies at all evaluated sites. These results suggest that the SunTag-ePE2 system does not significantly increase off-target editing levels at most target sites compared to the standard Fused-ePE2 system (Fig. 2e, f).

Further, we validated the editing efficiency of SunTag-ePE2 in exogenous site. We first used the piggyBac transposase tool²³ to establish a green fluorescent protein (GFP)-mutated (GFPm) reporter cell line, containing a premature stop codon resulting from a 4 nt deletion frameshift mutation in the GFP coding sequence^14,24 (Fig. 2g). Then using SunTag-ePE2 and ePE2 to insert GTTC sequence into the GFP mutation site can convert GFPm into GFP. The rate of GFP positive cells was further calculated by flow cytometer (FCM) to evaluate the PE editing efficiency. The data showed that the GFP positive cells of SunTag-ePE2 (13.4%) were 1.14 fold-change more than ePE2 (11%), suggesting that the precise editing efficiency of SunTag-ePE2 could be higher than ePE2 (Fig. 2h, i and Supplementary Fig. 5).

SunTag-ePE3 achieved comparable levels of editing in HEK293T and higher levels in HeLa cells than ePE3

A previous study showed that PE3 achieves about 3-fold editing efficiency compared with PE2³. To demonstrate whether SunTag-PE system was also feasible to PE3, we constructed SunTag-ePE3 system and tested it in four endogenous loci to install desired substitution, insertion, and deletion in both HEK293T and HeLa cells. The data showed that the editing efficiency of SunTag-ePE3 was 2-5-fold higher than that of SunTag-ePE2 across the four genomic loci in HEK293T (Figs. 2 and 3), indicating that the SunTag system was also feasible in PE3 form.

Comparison of SunTag-ePE3 with canonical fused ePE3, Split-ePE3, and MS2-ePE3 at four endogenous loci, including HEK3 locus (a), EMX1 locus (b), FANCF locus (c), and RNF2 locus (d) in HEK293T cells. Each locus contains three types of prime editing: nucleotide transversion, insertion, and deletion. Data and error bars indicate the mean and s.d. of three independent biological replicates. Negative controls represent four kinds of PE3 components without pegRNAs. The numbers above the bars represent fold change between different treatment groups.

We next compared the SunTag-ePE3 with canonical ePE3, Split-ePE3, and MS2-ePE3. As expected, the data in HEK293T cells showed that SunTag-ePE3 achieved comparable editing levels as canonical ePE3 (up to 39.68%) and higher editing levels than Split-ePE3 and MS2-ePE3 across all the tested sites in HEK293T cells (Fig. 3). Notably, SunTag-ePE3 showed higher editing efficiencies compared to Split-ePE3 at multiple loci: 1.21-fold for single nucleotide deletion at HEK site3; 1.18-1.31-fold across various edit types at EMX1; 1.23-fold for insertion at FANCF; and 1.18-fold for insertion at RNF2. When compared to MS2-ePE3, SunTag-ePE3 exhibited 1.12–1.23-fold higher efficiency for substitution and deletion at EMX1 and 1.21-fold higher efficiency for deletion at FANCF. Since the efficiency of current prime editors varies by cell types²⁵, we also tested our SunTag-ePE3 system in HeLa cells at four genomic loci. The editing levels of SunTag-ePE3 were comparable to canonical ePE3 across all four tested sites for all desired mutations, with marginally higher efficiency observed in the SunTag system (Fig. 4). For substitution edits, SunTag-ePE3 showed enhanced efficiency compared to canonical ePE3 at site3 (32.69% vs. 29.48%, 1.21-fold), FANCF (39.31% vs. 33.37%, 1.18-fold), and RNF2 (36.68% vs. 20.43%, 1.80-fold). Similar improvements were observed for trinucleotide insertions, with SunTag-ePE3 outperforming canonical ePE3 at site3 (51.41% vs. 44.16%, 1.16-fold), FANCF (55.86% vs. 51.05%, 1.09-fold), and RNF2 (46.82% vs. 31.28%, 1.50-fold). For single nucleotide deletions, SunTag-ePE3 maintained superior performance at site3 (33.27% vs. 26.23%, 1.27-fold), FANCF (44.99% vs. 40.01%, 1.12-fold), and RNF2 (26% vs. 16.67%, 1.56-fold). The above results indicated that SunTag-PE system was a versatile platform that can accommodate different prime editor versions and exhibited slightly higher editing efficiency than canonical fused-PE in specific mammalian cells.

Comparison of SunTag-ePE3 with canonical fused ePE3, Split-ePE3, and MS2-ePE3 at four endogenous loci, including HEK3 locus (a), EMX1 locus (b), FANCF locus (c), and RNF2 locus (d) in HeLa cells. Each locus contains three types of prime editing: nucleotide transversion, insertion, and deletion. Negative controls represent SunTag-ePE3 or ePE3 components without pegRNAs. Off-target analysis of SunTag-ePE3 and Fused-ePE3 systems at the EMX1 (e) and FANCF (f) target sites. The percentage of total sequencing reads with desired edits or indels is shown for the on-target and off-target sites. Spacer sequences for the on-target sites are displayed, with mismatched bases highlighted in red. OT-1: off target site 1; OT-2: off target site 2; OT-3: off target site 3. Data and error bars indicate the mean and s.d. of three independent biological replicates. The numbers above the bars represent fold change between different treatment groups.

Similarly, we also evaluated the off-target effects of the SunTag-ePE3 and Fused-ePE3 systems. Deep sequencing data at the EMX1 and FANCF loci revealed that both the SunTag-ePE3 and Fused-ePE3 systems exhibited extremely low off-target editing efficiencies at all evaluated sites. This further indicates that the SunTag-ePE3 system does not significantly increase off-target editing levels at most target sites compared to the standard Fused-ePE3 system (Fig. 4e, f).

SunTag-PE system is transferable to other orthologs of Cas9

To further validate the modularity of SunTag-PE system, we constructed SauCas9-SunTag-PE and FrCas9-SunTag-PE to explore whether SunTag-PE was applicable to other orthologs of Cas9. SauCas9 is a smaller Cas9 for all-in-one AAV delivery²⁶ and holds tremendous therapeutic potential^27,28. FrCas9 is a type II-A Cas9 with high cutting efficiency and high fidelity discovered in our laboratory²⁹. Notably, we employed active nucleases (SauCas9 and FrCas9) rather than nicking nuclease to construct the DSB-SunTag-PE system. DSB formation followed by NHEJ repair inevitably introduces indels, leading to increased editing byproducts (Supplementary Fig. 6b–e). For SauCas9 DSB-SunTag-PE (Supplementary Fig. 6d), three different types of edits were attempted at the EMX1 locus: transversion (+3CtoT and +8GtoT), insertion (+1AAGCTTins), and deletion (+1GGCdel). The precise editing efficiencies varied substantially among these editing types: the substitutions showed the lowest efficiency at 0.66%, while insertion and deletion demonstrated higher efficiencies of 3.7% and 9.10%, respectively. However, regardless of the editing type, SauCas9-mediated indel frequencies remained consistently high at ~28–30%, suggesting a strong preference for NHEJ-mediated repair across all editing strategies. Besides, previous studies revealed that DSB repair will generate undesired consequences, such as pegRNA scaffold integration through nonhomologous end joining (NHEJ) repair pathway³⁰ (Supplementary Fig. 6a). To improve the precise editing efficiency of DSB-SunTag-PE, we optimized FrCas9-SunTag-PE by removing the RT homology tail of pegRNA and retaining PBS and intended insertions. By using this system, we successfully inserted a short sequence containing a stop codon in the exon 9 of AKT1 gene. The results demonstrated that FrCas9 achieved a precise editing (+1GATGAins) efficiency of 20.29%, while the indel frequency was lower at 11.51% of total sequencing reads (Supplementary Fig. 6e). During short-sequence insertion editing by DSB-PE, we observed that editing byproducts typically emerge following precise target sequence insertion. This feature makes DSB-PE particularly valuable for gene knockout applications through stop codon insertion, as the desired gene disruption can still be achieved even with additional sequence alterations. These data suggested that the SunTag-PE system is a modular and easy-to-use platform for constructing PEs with different Cas9 orthologs. Compared with canonical fused-PE, this kind of technology transfer is easier and faster due to modularization, providing more flexibility to prime editing experimental design.

SunTag-ePE3 delivered by dual-AAV

Having validated the efficient PE editing of SunTag-PE, we further test whether this system can be delivered with two AAV vectors. First, we used SunTag-PE2 system to install a A•T-to-T•A transversion mutation in HEK293T HBB gene to construct an HBB mutant cell line³¹, mimicking the mutation in sickle cell disease³². Then, we encoded the N-1×GCN4-nCas9 (4.67 kb) in one AAV vector and the pegRNAs/ngRNA combination for HBB (T to A) together with scFv-RT in the other (4.36 kb) (Fig. 5a). Delivery of both vectors to the HBB mutant cells yielded a mean HBB T-to-A substitution frequency of nearly 25.47%, whereas delivery of only the pegRNA-ngRNA-scFv-RT did not yield detectable desired editing (Fig. 5b, c). This modular SunTag-PE addresses a limitation imposed by size-constrained AAV vectors and avoids additional step of protein reconstitution imposed by intein sequence used previously (Split-intein PE2)⁸.

a Schematic of SunTag-ePE3 delivered by dual-AAVs. b The HBB mutation repair results of dual-AAV SunTag-ePE3 analyzed by CRISPResso2. c The HBB mutation editing efficiency SunTag-ePE3 in HBB mutant cell model. Data and error bars indicate the mean and s.d. of three independent biological replicates.