Enhancing the efficiency of nuclease-based prime editing in rice with the Tf1 reverse transcriptase

Abstract

The development of efficient and precise genome-editing tools is crucial for advancing functional genomics and improving crops. Our previously established nuclease-mediated prime editing (NM-PE) system, which combines the SpCas9 nuclease with prime editing based on microhomology-mediated end joining, enables the seamless insertion of small DNA fragments into plant genomes to add tags to genes of interest. However, the efficiency of 3 × FLAG sequence insertion via NM-PE requires further improvement. Here, we report a significant optimization of this system by replacing the M-MLV reverse transcriptase (RT) with evolved variants of the retrotransposon RT Tf1 derived from the mammalian PE6 system. Through codon optimization, we generated the evoTf1M4 variant, which substantially enhanced the efficiency of NM-PE. The optimized construct rPE20aV3 achieved up to 18.75% precise insertion of a 66-bp 3 × FLAG sequence at endogenous loci, representing a three-fold improvement over the original NM-PE system. Our results demonstrate that Tf1-aided optimization of NM-PE serves as an efficient platform for seamless insertion of a 3 × FLAG sequence in rice, offering broad potential for advanced genome engineering in plants.

Dear Editor,

The seamless integration of DNA fragments into plant genomes is essential for advancing crop functional genomics and molecular breeding. In our previous work, we established an efficient one-step DNA insertion approach that combines the SpCas9 nuclease from the bacterium Streptococcus pyogenes with a microhomology-mediated end joining (MMEJ)-based prime editing (PE) strategy, termed NM-PE [1]. Although NM-PE can insert sequences encoding protein tags such as FLAG, HA, and c-Myc to endogenous loci, the efficiency of inserting larger fragments, such as the 3 × FLAG tag, remains suboptimal [1]. Therefore, we aimed to improve NM-PE to enable insertion of longer fragments and at a higher editing efficiency. Recent studies on the PE6 series of PEs in mammalian systems reported substantial improvements in editing efficiency through phage-assisted evolution and rational redesign of reverse transcriptases (RTs) [2]. Notably, these PE6 variants also markedly enhance the efficiency of conventional PE systems in rice (Oryza sativa) [3,4]. Building upon these advances, we investigated whether variants of the retrotransposon RT Tf1 would further enhance the efficiency of the NM-PE system for precise and seamless insertion of the 3 × FLAG sequence.

Since the RT evoTf1M4 from the PE6c system, harboring 16 rationally designed and evolutionarily optimized mutations, demonstrated superior performance in complex editing outcomes, we codon-optimized its sequence for expression in rice [2]. We then replaced the truncated RT from Moloney murine leukemia virus (M-MLV RT) in rPE14b with codon-optimized evoTf1M4, generating the construct rPE20a (Fig. 1A). We designed a prime editing guide RNA (pegRNA) for the insertion of a 3 × FLAG tag sequence immediately upstream of the initiation codon (ATG) of CALCIUM-DEPENDENT PROTEIN KINASE 3 (OsCPK3). This pegRNA contained an 8-nucleotide (nt) primer-binding site (PBS) with a melting temperature of 30 °C and an 87-nt reverse transcription template (RTT). The RTT incorporated one additional nucleotide to maintain the 3 × FLAG sequence in-frame with OsCPK3, as well as a 15-nt homology arm (Fig. 1B). We designed additional pegRNAs targeting OsCPK4 or OsCPK6 using the same strategy (Fig. S1 and S2). We subcloned each pegRNA into the pENTR:sgRNA41 vector, enabling transfer of the pegRNA expression cassette into rPE20a or rPE14b via Gateway recombination for subsequent Agrobacterium (Agrobacterium tumefaciens)-mediated rice transformation [5]. Following transformation, we PCR-amplified the genomic region encompassing the target site from independent primary (T₀) transgenic plants and analyzed the amplicons by high-throughput sequencing, with selected plants further validated by Sanger sequencing. The original rPE14b construct from the NM-PE system achieved precise and in-frame epitope tagging in 3 of 48 independent primary transformants (6.25% efficiency) at OsCPK6, but no precise tagging events were detected at the other two loci tested (Fig. 1C and D; Fig. S3). By contrast, the use of the rPE20a construct harboring evoTf1M4 failed to produce any precise insertions at any of the target sites, and all regenerated plants remained wild type at the target loci (Fig. 1C and D; Fig. S3).

Fig. 1.

Testing the efficiency prime editors incorporating variants of the Tf1 reverse transcriptase in rice. A Diagrams of the prime editors (PE) rPE20a, its codon-optimized versions rPE20aV2 and rPE20aV3, rPE20b, and rPE20c. The constructs contain the SpCas9 nickase (nuclease), the reverse transcriptase evoTf1, evoTf1M4, evoTf1M4V2, or evoTf1M4V3, a Kozak sequence upstream of the coding sequence, and multiple copies of a nuclear localization signal (NLS). evoTf1, reverse transcriptase from PE6b; evoTf1M4, an evoTf1 variant from PE6c; the sequences of evoTf1M4V2 and evoTf1M4V3 are given as Supplemental Sequences 2 and 3. B Diagram of the OsCPK3 locus and sequences of the epegRNA and nicking guide RNA used for the insertion of a 3 × FLAG sequence. PAM, protospacer-adjacent motif; PBS, primer-binding site; RTT, reverse transcription template; MH, microhomology arm. C Editing Efficiency of precise 3 × FLAG sequence insertion by rPE14b, rPE20a, rPE20aV2, and PE20aV3 at the indicated loci. D Distribution of genotypes in T₀ plants following transformation with pUbi:rPE14b, pUbi:rPE20a, pUbi:rPE20aV2, or pUbi:rPE20aV3 for the precise insertion of a 3 × FLAG sequence. Ho, homozygote; He, heterozygote; Bi, bi-allelic; WT, wild type; By, byproduct. E Representative Sanger sequencing chromatograms showing the insertion of the 3 × FLAG sequence. F Examples of sequences at OsCPK3 following precise or inprecise editing for the insertion of the 3 × FLAG sequence in OsCPK3. G Diagram of the OsACCase locus and design of a pegRNA for the introduction of a point mutation in OsACCase. Genotypes of T₀ plants transformed by pUbi:rPE20 variants. Representative Sanger sequencing chromatogram of point mutation created by pUbi:rPE20 variants. Pre, precise editing; By, byproduct. H Editing efficiency of base conversion induced by pUbi:rPE14a, pUbi:rPE20b, and pUbi:rPE20c at the indicated loci. I Percentage of mutations induced by pUbi:rPE14a, pUbi:rPE20b, and pUbi:rPE20c in T₀ plants at the indicated loci. Ho, homozygote; He, heterozygote; Bi, bi-allelic; WT, wild type; By, byproduct.

Given that codon optimization is crucial for efficient protein translation, and that previous studies have shown that codon usage influences CRISPR/Cas-mediated editing efficiency [6], we attempted to improve PE performance by generating two evoTf1M4 variants: one using codon optimization for rice, and one with codon optimization for maize (Zea mays), both based on the online tool by GenScript. We then replaced the evoTf1M4 sequence in rPE20a with that of each variant to produce rPE20aV2 (codon-optimized for rice) and rPE20aV3 (codon-optimized for maize) (Fig. 1A). We used the same pegRNAs as with rPE20a to assess these two new constructs. Following rice transformation, rPE20aV3 achieved precise epitope tagging in 9 of 48 independent primary transformants (18.75% efficiency) at the OsCPK6 locus, representing a three-fold increase over rPE14b-mediated PE (Fig. 1C and D; Fig. S3). By contrast, rPE20aV2 exhibited markedly lower efficiency at the same locus (2.1%, 1 of 48 primary transformants) (Fig. 1C and D; Fig. S3). At the OsCPK3 locus, we identified accurate insertions in 3 out of 48 T₀ plants for rPE20aV2, representing a 6.25% editing efficiency, and in 7 out of 48 T₀ plants for rPE20aV3, representing a 14.6% efficiency (Fig. 1C and D; Fig. S3). For OsCPK4, rPE20aV2 achieved an editing efficiency of 4.2% (2 of 48 T₀ plants), whereas rPE20aV3 reached an efficiency of 10.4% (5 of 48 T₀ plants) (Fig. 1C and D; Fig. S3). TA cloning and Sanger sequencing of individual PCR products confirmed the precise insertion of the 3 × FLAG tag sequence at all three loci (Fig. 1E).

We performed immunoblots to determine whether the endogenous gene tagged with a 3 × FLAG sequence undergoes normal translation. Indeed, we detected the FLAG epitope in edited plants using rPE20aV3, with a molecular mass consistent with that of each tagged protein, indicating successful translation of transcripts produced from the tagged loci (Fig. S4). Overall, rPE20aV2 yielded accurate insertions in the endogenous rice genes OsCPK3, OsCPK4, and OsCPK6, and rPE20aV3 further enhanced efficiency by 2.3- to 9-fold compared to rPE20aV2. These findings suggest that both codon optimization approaches outperformed the original codon optimization proposed by Integrated DNA Technologies (IDT) for rice, with the maize-optimized variant (rPE20aV3) showing the highest efficiency. In addition to precise insertions, we detected unintended editing outcomes at all three loci, including small insertions or deletions (InDels) near the cleavage sites and imprecise insertions, such as truncated fragments of the 3 × FLAG sequence or insertions accompanied by InDels at the 5′ end or 3′ end of the cleavage site (Fig. 1F; Fig. S1B and S2B). These results suggest that our system can concurrently generate epitope-tagged alleles and knockout mutants.

Encouraged by the higher efficiency observed with rPE20aV3-mediated insertions of the 3 × FLAG sequence, we wondered whether its RT variant evoTf1M4V3 would also enhance the efficiency of introducing point mutations in the rice genome. To evaluate this possibility, we replaced M-MLV RT in rPE14a with either evoTf1 or evoTf1M4V3, generating the constructs rPE20b and rPE20c, respectively (Fig. 1A). We designed four pegRNAs to introduce point mutations into different endogenous rice genes: ACETYL-COA CARBOXYLASE (OsACCase), ACETOLACTATE SYNTHASE (OsALS), 5-ENOLPYRUVYLSHIKIMATE-3-PHOSPHATE SYNTHASE (OsEPSPS), and GRAIN SIZE 3 (OsGS3) (Fig. 1G; Fig. S5). In the T₀ generation, we detected precise editing at the OsACCase locus in 11.1% (5 of 45) independent transgenic plants transformed with rPE20c, whereas no mutation was detected when using rPE14a or rPE20b (Fig. 1H and I; Fig. S6). At the OsALS locus, rPE20c attained a 22.9% (11 of 48) editing efficiency, with all edited plants harboring one mutated copy. By contrast, rPE14a yielded an editing efficiency of 16.7% (8 of 48), comprising one homozygous plant and seven heterozygous plants, but rPE20b resulted in a lower efficiency of 6.25% (3 of 48). For OsEPSPS, rPE14a-mediated editing reached an efficiency of 18.75% (9 of 48), rising to 23.4% (11 of 47) with rPE20b, and to 26.7% (8 of 30) with rPE20c (Fig. 1H and I; Fig. S6). At the OsGS3 locus, rPE20b achieved an editing efficiency of 8.3% (4 of 48), whereas rPE14a and rPE20c both reached 14.6% (7 of 48) (Fig. 1H and I; Fig. S6). Collectively, these results demonstrate that the evoTf1M4V3-based prime editor rPE20c substantially enhances the efficiency of point mutations across multiple genomic loci in rice.

In this study, we optimized our previously established NM-PE system by incorporating variants of the Tf1 RT, leading to the development of the highly efficient prime editor rPE20 and its codon-optimized derivatives. Through RT evolution andcodon optimization, the seamless integration, which was induced by NM-PE, of DNA fragments of up to 66 bp. The best-performing PE variant, rPE20aV3, achieved a three-fold higher tagging efficiency than the original NM-PE system. Moreover, the evoTf1M4V3-based editor rPE20c markedly enhanced the efficiency of introducing point mutations (insertions or deletions) across multiple genomic loci (Supplemental Table 1). These findings highlight the utility of Tf1-based RTs for improving nuclease-mediated PE systems. Although further refinements are warranted to maximize efficiency and minimize unintended edits, this work establishes a robust and versatile platform for complex genome engineering in rice, thereby expanding the applicability of NM-PE in functional genomics and crop improvement.

1. Materials and methods

1.1. Plasmid construction

The coding sequences of evoTf1 and its variants were individually synthesized by Tsingke Biotech (Beijing, China) and inserted into the pMV vector, generating the pMV-evoTf1, pMV-evoTf1M4, pMV-evoTf1M4V2, and pMV-evoTf1M4V3 constructs. For prime editor construction, the full-length evoTf1M4, evoTf1M4V2, and evoTf1M4V3 sequences were first PCR-amplified using specific primers and then inserted in-frame with the sequence encoding the SpCas9 nuclease in the pUC57:rPE14b vector [1], yielding pUC57:rPE20a, pUC57:rPE20aV2, and pUC57:rPE20aV3. Similarly, the full-length evoTf1 and evoTf1M4V3 sequences were inserted in-frame with the SpCas9 sequence in the pUC19:rPE14a vector [1], resulting in the pUC19:rPE20b and pUC19:rPE20c constructs. Finally, the prime editor cassettes were inserted into the binary vector via BamHI/SpeI digestion, producing pUbi:rPE20a variants for 3 × FLAG insertion and pUbi:rPE20b and pUbi:rPE20c for point mutations.

The pegRNAs were synthesized by Tsingke Biotech (Beijing, China). Each pegRNA was inserted into the pENTR:gRNA41 vector via BsaI digestion and ligation, generating the pegRNA expression cassettes. Each cassette was recombined into the respective prime editor construct through Gateway recombination.

The identities of evoTf1, evoTf1M4 variants, all pUbi:rPE20 constructs, and inserted oligonucleotides were confirmed by Sanger sequencing.

1.2. Agrobacterium tumefaciens-mediated rice transformation

The rice cultivar ‘Kitaake’ was used in this study. Rice plants were cultivated in a paddy field under natural conditions during the normal rice growing season. Immature seeds were harvested from expanded panicles for use in rice transformation.

The editing constructs were individually introduced into the Agrobacterium tumefaciens strain EHA105. Rice callus derived from immature seeds of Kitaake was used for stable transformation following a previously described protocol [5].

1.3. Genotyping of transgenic rice plants

For the identification of editing events and types induced by prime editing, rice genomic DNA was extracted from each T₀ transgenic plant using the CTAB method [5]. PCR amplification of the target region was carried out with specific primers listed in Supplemental Table 2. Equal amounts of PCR products with different barcodes were pooled together and subjected to deep sequencing. 1 GB of sequencing data was generated for each sample using an Illumina NextSeq 500 platform to produce 150-bp paired-end reads with a 10% threshold. Subsequently, plants were classified based on the remaining types of reads in each line (two types or fewer). The classification criteria are as follows: When only one type of read is present, the plant is genotyped as wild-type if no editing is detected, and as homozygous if the read corresponds to the precise edit. When two types of reads are present, the plant is genotyped as heterozygous if one read is wild-type and the other is a precise edit, and as biallelic if one read is a precise edit and the other is an imprecise edit (e.g., insertion, deletion, or substitution not matching the intended design). All other scenarios are classified as byproducts (or complex editing outcomes). The genotypes of all transgenic plants are listed in Supplementary Table S3.

1.4. Total protein extraction and immunoblot analysis

Total protein was extracted from 0.1 g of roots of T₀ rice plants subjected to drought stress treatment using an extraction buffer containing 100 mM Tricine (pH 7.8), 10 mM KCl, 1 mM MgCl₂, 1 mM EDTA, 10% (w/v) sucrose, 2% (v/v) Triton X100, 1 mM DTT, and 1 × protease inhibitor cocktail (RM02916, ABclonal Tech, China). Following extraction, the samples were centrifuged at 16,300 rpm for 10 min at 4 °C. Subsequently, 8 μL of each protein extract mixed with 2 μL loading dye was separated on an 8% (w/v) SDS-PAGE gel and electrophoretically transferred to a PVDF membrane at 22 V for 30 min. The membranes were blocked with 5% (w/v) skim milk prepared in Tris-buffered saline with Tween 20 (TBST, 20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.1% [v/v] Tween 20). Immunoblotting was performed using an anti-DDDDK primary antibody (AE005, ABclonal Tech, China) diluted at 1:10,000 in TBST with 5% (w/v) skim milk. After overnight incubation with the primary antibody at 4 °C, the membranes were washed three times with TBST (5 min per wash), followed by incubation with a horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG secondary antibody (AS003, ABclonal, China; 1:10,000 dilution) in TBST containing 5% (w/v) skim milk for 1 h at room temperature. The membranes were then washed three times with TBST (10 min each). Signals were visualized using Super ECL Detection Reagent (36208ES60, Yeasen, China) and detected with a Touch Imaging System (e-BLOT, China).

CRediT authorship contribution statement

Guigen Ma: Writing – original draft, Investigation, Data curation. Hao Xu: Writing – original draft, Investigation, Data curation. Sujie Zhang: Writing – original draft, Investigation, Data curation. Xueqi Li: Investigation. Jiaqiang Liu: Investigation. Jiayue Xie: Investigation. Fang Yan: Supervision, Funding acquisition, Conceptualization. Huanbin Zhou: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is/are the supplementary data to this article:

Data availability

Data will be made available on request.