Prime editing: a new precise gene-editing method for improving rice

CRISPR refers to clustered regularly interspaced short palindromic repeats of DNA sequences derived from a prokaryotic adaptive immune system that protects them from invaders by using CRISPR-RNAs (crRNAs) to silence viral nucleic acids (Jinek et al., 2012). The system comprises two main components: Cas9 endonuclease protein that is responsible for inducing DSBs at specific DNA sites (Anzalone et al., 2020), and a single guide RNA (sgRNA) that guides Cas9 to the complementary target DNA sequence by base pairing (Asmamaw and Zawdie, 2021). The CRISPR/Cas9 editing process includes three steps: target recognition, DNA cleavage, and repair. As shown in Figure 1, the sgRNA guides Cas9 to bind to the target sequence, introducing a DSB that is then repaired by either non-homologous end joining (NHEJ) or homology-directed repair (HDR) (Zhao et al., 2023).

The discovery of the CRISPR/Cas9 system has played a pivotal role in genome editing of eukaryotes and has revolutionized the field of genetic engineering and enabled a wide range of potential applications in medicine (Li et al., 2023) and agriculture (Gan and Ling, 2022). For example, CRISPR/Cas9 technology has revolutionized plant breeding, enabling precise genetic changes that improved rice yield and quality (Abdelrahman and Zhao, 2020; Liu et al., 2022), disease resistance (Wei et al., 2021; Talakayala et al., 2022; Shawky et al., 2024), and environmental adaptability (Romero and Gatica-Arias, 2019), alleviating issues with global food security and sustainable agriculture. Moreover, CRISPR/Cas9 has been used to improve plant disease resistance to various diseases by modifying host susceptibility genes or the regulatory genes that are known to be involved in viral infection in rice (Talakayala et al., 2022). Furthermore, promoter-targeted CRISPR editing has been utilized to modulate gene expression and develop quantitative trait variation in rice crops (Wu et al., 2024). The applications of CRISPR/Cas9 technology in rice crops are summarized in Table 1.

However, despite its enormous capabilities, several limitations restrict the precision and scope of CRISPR/Cas9-mediated genome editing. One major constraint is the requirement for a protospacer adjacent motif (PAM) near the target site, which limits the range of editable genomic loci. In addition, CRISPR/Cas9 relies on the induction of double-strand breaks (DSBs), which are primarily repaired through the error-prone non-homologous end joining (NHEJ) pathway, often resulting in random insertions or deletions (indels) rather than precise nucleotide substitutions (Hussain and Ahmad, 2022). Although homology-directed repair (HDR) can be used to introduce precise edits, this pathway occurs at very low efficiency in plants and requires the delivery of an exogenous donor template, making precise genome modification challenging. Furthermore, DSB formation can lead to unintended mutations, large genomic rearrangements, and off-target effects, which may compromise genome integrity and editing specificity. These limitations highlight the need for next-generation genome editing technologies capable of introducing precise genetic modifications without relying on DSBs or donor DNA templates. To overcome the limitations associated with DSB-dependent editing systems, alternative CRISPR-based technologies such as base editing and prime editing have been developed to enable more precise and predictable genome modifications.

BE, an innovative genome editing tool, has shown high efficiency in rice crop improvement (Yarra and Sahoo, 2021). BE is a CRISPR-derived genome editing approach that enables precise single-nucleotide substitutions without introducing DSBs or requiring donor DNA templates. This method employs adenosine (ABEs) or cytidine base editors (CBEs) to precisely edit at the desired target location (Rees and Liu, 2018).

CBEs use a catalytically dead Cas protein fused with cytidine deaminase (CDA) to convert cytosine (C) to uracil (U), recognized as thymine (T) after replication, causing C•G to T•A substitutions and amino acid changes such as arginine (CGA) to stop codon (TGA). Conversely, ABEs employ an adenosine deaminase that converts adenine (A) to inosine (I), read as guanine (G), leading to A•T to G•C substitutions and amino acid alterations like lysine (AAA) to glutamic acid (GAA) (Biswas et al., 2022) (Figure 1).

BE has been highly efficient by enabling point mutations in rice plants for improving agronomic traits as shown in Table 2. For example, to improve disease resistance, the CBE rBE5, was used to target the OsPi-d2, a gene that controls rice defense mechanisms against blast fungus, and demonstrated a 71.9% efficiency for mono- or bi-allelic mutations (Ren et al., 2018). Furthermore, increasing the cytidine editing frequency of plant base editor 3 (BE3) and CDA systems, achieved an editing efficiency that reached 88.9% and 85.7% in regenerated rice plants (Qin et al., 2020). On the other hand, yield and plant has successfully improved plant architecture substantially by utilizing an ABE. For example, targeted editing of the regulatory binding site in OsSPL14 (Oryza Sativa squamosa promoter binding protein-like14) had achieved an editing efficiency of 23% (Hua et al., 2018), the combination of CBE/ABE was successfully used to fine-tune of the Amylose Content (AC) through BE of the OsWaxy gene and generated a series of mutants with ACs ranging from 1.4% to 11.9% (Xu et al., 2021b).

Furthermore, BE enables efficient management of agronomic input–related traits, successfully creating stacked resistance to multiple herbicide classes by simultaneously editing four distinct genes, including Acetolactate Synthase (OsALS1), Cytosolic Glutamine Synthetase (OsGS1), alpha-Tubulin (OsTubA2), and Acetyl-CoA Carboxylase (OsACC). The co-editing efficiencies achieved for dual-gene editing reached up to 72.92% for OsALS1 and OsGS1, and 73.81 for co-editing OsTubA2 and OsACC. Additionally, the system achieved 56.25% co-editing efficiency when targeting all four genes simultaneously, with editing efficiency ranging from 56.25% to 89.58% for individual loci in the multiplex editing experiment (Yan et al., 2021).

Although base editing enables efficient point mutations without inducing DSBs, it still has several important limitations. First, base editors are restricted to transition mutations, primarily C•G to T•A and A•T to G•C substitutions, and therefore cannot generate transversion mutations, insertions, or deletions (Anzalone et al., 2019). Another major limitation is the presence of a defined editing window, where multiple cytidine or adenine bases within the window can undergo simultaneous deamination, often resulting in bystander mutations and unintended edits. In addition, base editors have been reported to exhibit off-target activity at both DNA and RNA levels due to the promiscuous activity of deaminase enzymes (Slesarenko et al., 2022). Furthermore, similar to CRISPR/Cas9, base editors still rely on PAM sequences for target recognition, which restricts the range of editable genomic loci (Hua et al., 2019). These limitations reduce the flexibility of base editing for precise genome engineering. To overcome these constraints, PE was developed as a versatile genome editing technology capable of introducing all types of base substitutions as well as small insertions and deletions without requiring DSBs or donor DNA templates, thereby expanding the scope of precise genome modification in crops.

PE tool was developed in 2019 as next-generation CRISPR-based genome editing technology (Anzalone et al., 2019). This innovative system enables targeted substitutions, insertions, and deletions with high efficiency and minimal off-target effects. Unlike conventional CRISPR/Cas9 system, PE operates without the need for DSBs or exogenous donor DNA or, resulting in greater accuracy, flexibility, and fewer indel by-products. Moreover, compared with base editing, it exhibits robust editing capabilities by enabling all types of base substitutions.

The PE system consists of a Cas9 nickase (nCas9, H840A) fused to a reverse transcriptase (RT) together with a pegRNA (Figure 1). The pegRNA contains both a spacer sequence for target recognition and an RNA template encoding the desired edit. This configuration allows PE to precisely introduce genetic modifications at specific loci without relying on error-prone DNA repair pathways (Scholefield and Harrison, 2021; Chen and Liu, 2023b). Shuto et al. (2024) developed a modified prime editor by fusing Streptococcus pyogenes Cas9 (spCas9) H840A nickase mutant with the wild type of from Moloney murine leukemia virus RT enzyme (M-MLV RT) to facilitate efficient DNA synthesis at the target site. An engineered prime editor (ePPE) significantly improved editing efficiency in plants by combining reverse transcriptase optimization with nucleocapsid protein fusion, resulting in an average ~5.8-fold increase in editing efficiency compared to the original PE system. In rice, this translated to 11.3% editing efficiency versus 2.1% with conventional PE, without increasing off-target effects or unintended by-products (Zong et al., 2022). These findings demonstrate that systematic optimization of PE components can substantially enhance editing performance while maintaining high specificity.

PE functions through a search-and-replace strategy mediated by the pegRNA and nCas9-RT fusion protein. the nCas9 component first creates a single strand break SSB also called as a nick at the target DNA strand forming an R-loop structure at the target strand to allow the pegRNA spacer to bind to the complementary genomic sequence. Thus, the SSB of the non-targeted strand causes the release of 3’ single-strand end. The resulting 3’ end hybridizes with the primer binding site (PBS) of the 3’ end of the pegRNA (Vu et al., 2024).

Following the hybridization, the RT extends the 3' of the nicked DNA strand using the pegRNA template, producing a 3' DNA flap that incorporates the intended edit and is homologous to the downstream genomic sequence. Subsequently, the 3' flap displaces an adjacent strand of genomic DNA via flap interconversion. The excision of the displaced 5' flap, followed by the ligation of the remaining nick, results in a heteroduplex where one genomic strand contains the edit. Finally, DNA ligation and mismatch repair processes incorporate the desired edits into the complementary strand, thereby making the prime edit permanent (Figure 2) (Anzalone et al., 2020). Compared to CRISPR/Cas9 and BE, PE offers several key advantages, including higher precision, reduced by-product formation, and greater flexibility in editing outcomes. However, challenges such as variable editing efficiency, pegRNA design optimization, and delivery of large PE constructs remain areas of ongoing research. A comparative summary of CRISPR/Cas9, base editing, and prime editing is presented in Table 3.

PE1, the first-generation of prime editor, was developed with a minimum number of components (Anzalone et al., 2019). The system comprises of the pegRNA containing the guide sequence, a primer binding site (PBS), and an RT template, which together can direct the edit, along with a modified Cas9, Cas9 nickase (H840A). The modified Cas9-M-MLV RT fusion and pegRNA directs the nicking and RT mediated insertion of the desired edit (Figure 3) (Fu et al., 2023). Moreover, PE1 has several advantages, including its simple design and ability to mediate small insertions, deletions, and single-base substitutions (Chen and Liu, 2023b). Anzalone et al. tested PE1’s ability to cause transversion point mutations at about five genomic sites designed in the pegRNA and reported that the efficiency of the editing depends on the length of the PBS, with the maximum efficiency of 0.7 to 5.5% (Anzalone et al., 2020). Despite its ability to create diverse base edits, PE1 remains limited by low editing efficiency (Chen and Liu, 2023b).

To address PE1’s low efficiency, PE2 was developed through engineering of the RT domain. Five amino acid substitutions (D200N, L603W, T330P, T306K, and W313F) were introduced into the M-MLV RT domain (Fu et al., 2023). These modifications, shown in Figure 3, improved RT thermostability, processivity, and DNA-binding efficiency. The three amino acid substitutions of D200N, L603W, and T330P (in M-MLV RT, termed M3) yielded an average of 6.8-fold increase in insertion and transversion editing efficiency compared to the wild-type RT in PE1 (Anzalone et al., 2019). Furthermore, additional mutations (T306Kand W313F) resulted in a further increase in the transversion or insertion efficiencies.

In addition to optimization of the RT, modifications to the pegRNA design were important to enhance the efficiency of PE2. The length and GC content of the PBS influence editing outcomes, as lower GC content generally requires a longer PBS to ensure stable DNA hybridization (Ponnienselvan et al., 2023). A PBS length of about 13 nucleotides is generally used, but differences in GC content often require testing multiple lengths to achieve optimal editing (Anzalone et al., 2019; Ponnienselvan et al., 2023). The RT template, usually 10–20 nucleotides long, should also be optimized for each genomic target. Longer templates (>15 nt) may enhance editing at some sites, while shorter ones (10–16 nt) can perform better at others (Anzalone et al., 2019; Tang et al., 2020). Moreover, a cytosine (C) as the first nucleotide next to the 3′ hairpin of the pegRNA can disrupt RNA structure and lower editing efficiency; thus, a non-C base at this position is recommended (Anzalone et al., 2019; Tang et al., 2020; Ponnienselvan et al., 2023).

PE2 markedly enhanced editing efficiency, showing a 1.6–5.1-fold improvement over PE1 and improved the efficiency of insertions, deletions, and transversions as illustrating compatibility with shorter PBS sequences (Anzalone et al., 2019; Daliri et al., 2024). Moreover, the improved performance of PE2 was demonstrated in rice by restoring fluorescence in nonfunctional GFP mutants, the PE2-mediated editing yielded over 50% GFP-positive calli (Hua et al., 2020). In addition, Li et al. (2022c) achieved up to 66.7% scarless precise editing efficiency in the OsWaxy locus in rice, which had a substantially higher efficiency in comparison to the best PE3-mediated precise editing (13%) reported in the same study as well as avoiding the accompanying NHEJ/indel formation often observed in PE3. These studies support the potential of using PE2 as a robust and precise option for plant genome editing.

PE3 was developed to further increase editing efficiency by introducing an additional single-guide RNA (sgRNA) that can cause a nick in the non-edited DNA strand. Directing nickase to cut the non-edited strand PE3, thereby ensuring the likelihood that the edited strand will be used as a template, enabling the cell to utilize the new sequence across the duplex (Anzalone et al., 2019).

However, this approach increases the risk of indel formation, as nicking of both DNA strands may result in unintended double-strand breaks (Chen and Liu, 2023b). To address this challenge, the PE3b system was introduced in which the sgRNA is designed to match only the edited strand (Figure 2). This ensures that the non-edited strand is nicked only after successful editing, reducing indel formation by 13-fold compared to PE3 without sacrificing efficiency (Anzalone et al., 2019).

DNA mismatch repair (MMR) pathways play a critical role in maintaining the integrity of the genome by recognizing and correcting genetic mutations that may arise during DNA replication (Bateman, 2021). While essential, these proteins can also interfere with PE by reversing desired edits (Ramchandani et al., 2023). Research by Chen et al. (2021) used CRISPR interference (CRISPRi) screens to understand how the genes involved in DNA repair affect PE; it was shown that specific MMR genes significantly suppress PE efficiency and increase indel formation.

To modulate MMR activity, a dominant negative MLH1 (MLH1dn) MMR protein was integrated into PE2 and PE3 to suppress MMR activity, thus creating PE4 and PE5, respectively. In comparison to PE2, a 7.7-fold increase in editing efficiency was reported in PE4 (PE2+MLH1), On the other hand, PE5 (PE3+MLH1dn) showed a two-fold increase in efficiency compared PE3 while significantly decreasing indel formation by 3.4-fold (Chen et al., 2021).

Similarly, Liu et al. (2024)a attempted to suppress the expression of OsMLH1 by RNAi knockdown in a modified PE system (ePE5c) that resulted in increased PE efficiency by up to 2.11-fold across six targets in transformed rice cells. Moreover, the study also reported that in the T0 generation, 87.15% of the rice plants were edited whereas 85% were homozygous mutants (Liu et al., 2024a).

PEmax is an optimized PE architecture designed to enhance editing efficiency through multiple modifications to the PE2 protein. Starting with human codon-optimized RT, in addition to introducing a 34-amino acid sequence linker between the RT and the Cas9 nickase, including a bipartite SV40 nuclear localization signal (NLS) with a C-terminal c-Myc NLS added, thus further enhancing nuclear localization. Two new Cas9 mutations were introduced (R221K and N394K), which had a significant impact on improving Cas9 nuclease activity (Spencer and Zhang, 2017). When tested at seven target sites of the HeLa cells, PEmax had outperformed other versions of prime editors. The integration of this architecture with PE2, PE3, PE4, and PE5 systems has shown significant increase in the frequency of intended editing in both HeLa cells and in HEK293T cells (Chen et al., 2021).

In rice, PE has been employed to address specific agricultural challenges by enabling precise genetic modifications that will enhance traits such as stress tolerance, disease resistance, and nutritional quality. For instance, the development of a new germplasm resistant to X. oryzae pathovar oryzae (Xoo), which causes bacterial blight, became possible through PE (Gupta et al., 2023). Moreover, a study explored a PE system called the PrimeRoot system which enables rapid breeding and disease resistance by precisely integrating large DNA fragments (Tingting et al., 2023). In another study, Zhong et al. (2024) introduced NEPE, an engineered plant PE system that comprises of N-terminal region of M-MLV RT fused with SpCas9 (H840A) and a pegRNA (epegRNA) that includes an additional 3’ motif (EvopreQ1) in order to prevent degradation. This modified system has shown significant improvement over PE2. Employment of this system in rice protoplasts resulted in an editing efficiency of up to 18% with rates of insertion and deletion that were less than 0.2%. Moreover, in T0 plant lines, this system achieved an editing efficiency of 30% in comparison to PE2. The study highlighted how pegRNA stabilization and modification of editor architecture can result in versatile and highly efficient PE editors (Zhong et al., 2024). The comparison among the PE systems with their efficiency and limitations is mentioned in Table 4.

Prime editors have been used to make modifications in the rice Waxy gene (Os06g04200), responsible for amylose biosynthesis. Amylose content strongly influences rice texture, cooking quality, and market preference. Li et al. (2022c) modified a PE system that can applied to monocots by using a modified SpCas9 (KF33) (Li et al., 2022c). They constructed DNA vectors specifically designed for monocot genome editing, using the maize U6 promoter driving the expression of pegRNA and the ZmUBI1 promoter driving codon-optimized SpCas9 expression. Two sites within the Waxy gene were targeted (OsWx-G3 and OsWx-G4), validation of the modified system’s precision was carried with conventional CRISPR/Cas9 constructs KF133 and KF134. They further enhanced the system by using PE2 and PE3 variants; each variant was modified in a specific manner that enabled precise editing.

Additionally, the Agrobacterium tumefaciens-mediated transformation of rice plants (O. sativa L. japonica) was performed according to protocols, with minor modifications, using strain EHA105. Mature seeds from a proprietary rice cultivar, Kendao-32, were used to induce embryogenic callus for transformation. The two compared PE strategies were PE2 and PE3. Genotyping analysis showed that the PE2 system achieved a high of 66.7% in precise gene editing at targeted sites in the Waxy gene that were critical in amylose synthesis. PE3 on the other hand, achieved only 13% with 17.4% unwanted non-homologous end joining (NHEJ) edits in the same study. The discrepancy between the PE2 and PE3 systems most likely stem from the fact that PE3 systems depend on a dual-nicking strategy which would likely result in DSBs if the nicks are near each other. Moreover, DSBs often would result in a tendency towards a NHEJ repair pathway which was observed by Li et al. Their results detail that when enhancing template bias, the genomic stability must also be considered (Li et al., 2022c). In contrast to the precise nucleotide-level modifications enabled by prime editing, CRISPR/Cas9-mediated mutagenesis of the Waxy gene in rice has been shown to produce complex and unpredictable outcomes. For example, CRISPR/Cas9-induced knockout mutations resulted in only partial reduction of GBSS activity (61–71% of wild-type levels), due to compensatory upregulation of GBSSII, accompanied by altered starch composition, abnormal endosperm structure, and broader transcriptional reprogramming of starch biosynthetic genes (Pérez et al., 2019). These findings highlight that DSB-induced gene disruption may trigger unintended metabolic feedback effects, limiting the predictability of phenotypic outcomes compared to precise editing approaches such as prime editing.

Bacterial blight, caused by X. oryzae pv. oryzae (Xoo), is a bacterial disease that affects rice crops that are responsible for major losses in production and with yield losses as high as 70% (Ritbamrung et al., 2025). Gupta et al. (2023) described an improved PE-system approach aimed at enhancing broad-spectrum resistance in rice against bacterial blight (Gupta et al., 2023).

The pathogen utilizes transcription activator-like effectors (TALEs) to regulate host susceptibility genes, including the SWEET gene family which are essential for the virulence of the pathogen, thus serving as targets for TALEs binding to their specific effector binding elements (EBEs) within their promoter regions (Figure 4). The authors used the upgraded PE system, referred to as PE5max and employed two approaches in the rice cultivar. The first approach was to introduce 30-bp EBEs from the SWEET14 gene in the promoter of the non-functional executor gene xa23 to reconstitute a dominant resistance allele of Xa23, termed Xa23SW14 (Figure 4). This approach attained 47.2% editing efficiency, with an 18% rate of biallelic edits in the T0 generation. The second approach involved the use of the single amino acid substitution, V39E, in the TFIIAc5 gene for susceptibility to Xoo, which was used to generate the recessive xa5 allele. This achieved a very impressive 88.5% editing efficiency, with 30% of T0 lines having biallelic edits. Both strategies have produced efficient, heritable edits that conferred broad-spectrum resistance. Both the Xa23SW14 and xa5-engineered alleles were inheritable and conferred strong resistance against multiple strains of Xoo in successive generations. Whole-genome sequencing demonstrated high specificity without any off-target effects associated with the random mutations linked to the PE system (Gupta et al., 2023).

While CRISPR/Cas9 has been successfully applied to engineer resistance against bacterial blight in rice, its efficiency and precision remain constrained by the underlying repair mechanisms. For example, targeted disruption of effector binding elements (EBEs) in the OsSWEET14 promoter using CRISPR/Cas9 resulted in bacterial blight-resistant lines, but editing efficiency was relatively low, reaching only ~9% in Super Basmati rice (Zafar et al., 2020). Moreover, strategies relying on homology-directed repair (HDR) to introduce precise modifications, such as insertion of EBE elements into the Xa23 promoter, are technically challenging and often inefficient in plants due to the low activity of HDR pathways and the requirement for donor DNA templates (Wei et al., 2021). In contrast, prime editing enables precise nucleotide substitutions and targeted insertions without requiring donor DNA or DSBs, offering a more efficient and predictable approach for engineering disease resistance alleles. For instance, PE-mediated editing of susceptibility and resistance-related genes has achieved editing efficiencies of up to 88.5% in rice while maintaining high specificity and heritability (Gupta et al., 2023). These comparisons highlight the advantage of PE in achieving precise and efficient genetic modifications for disease resistance compared to conventional CRISPR/Cas9 approaches.

PE has demonstrated strong potential in developing herbicide-resistant rice through precise modification of genes involved in herbicide target pathways. These applications include engineering resistance to glyphosate and other herbicides, as well as enabling simultaneous modification of multiple agronomic traits.

PE has also been utilized to address abiotic stress such as heat stress in crops. Lou et al. (2025) demonstrated a novel application of PE for improving rice resilience to high temperatures. Using the PEmax system, a 10-bp heat-shock element (HSE) was inserted into the promoter regions of cell-wall invertase (CWIN) genes in elite rice cultivars. This small regulatory “switch” enabled the plants to rapidly activate CWIN expression during heat stress, maintaining sugar transport from leaves to developing grains. Increased CWIN activity boosts sugar transport from leaves to grains (“source–sink” dynamics), counteracting the typical yield losses from heat stress. In multi-location field trials, these engineered plants achieved up to 25% higher yields under heat stress compared to controls, with some lines rescuing over 40% of heat-induced yield losses, without affecting grain quality. This work highlights the potential of PE for fine-tuning gene regulation, offering a precise and non-disruptive strategy for developing climate-resilient rice cultivars (Lou et al., 2025).

Approximately 94.6% of white rice varieties carry a 14-bp deletion in the Rc allele which causes reduced pigmentation and antioxidant properties. Nguyen et al. (2025) proposed employing PE to reintroduce the 14-bp missing sequences (Nguyen et al., 2025) using the Kitaake variety, the researchers incorporated recombinatorial attL sequences for Gateway cloning into a modified pH-ePPE prime editor system (Cao et al., 2024), and developed a TwinPE strategy employing paired engineered pegRNAs with complementary RT templates to enhance insertion efficiency. In comparison, TwinPE had an efficiency of 44.2%, whereas Single PE had an efficiency of 11.4%. All altered plants had brown pericarps and heterozygous insertions. Heritability confirmed stable transmission of the 14-bp insertion in the T₁ generation, with restored Rc transcription and induction of flavonoid biosynthetic genes, although proanthocyanidin accumulation remained low due to a mutation in the Rd gene. These results establish TwinPE as an efficient and precise genome editing approach for small DNA insertions in plants, offering a promising platform for improving rice grain nutritional quality through targeted genetic restoration (Cao et al., 2024; Nguyen et al., 2025).

The regulatory classification of prime-edited (PE) rice varies considerably across jurisdictions, yet a clear global trend is emerging toward product-based frameworks that assess the nature of the genomic change rather than the technology used. Most nations have adopted a site-directed nuclease (SDN) classification system: SDN-1 (small indels), SDN-2 (defined substitutions or small insertions), and SDN-3 (foreign gene insertion). PE-generated edits, which are typically precise substitutions or small insertions without donor DNA, are considered from the SDN-2 category and, in many countries, are subject to simplified or expedited review rather than full transgenic regulation as shown in Table 3. China (USDA, 2022) and Japan (USDA, 2021) have established notification-based systems under which PE edits indistinguishable from natural variation may be exempt from GM oversight. Brazil and Argentina, through CTNBio Resolution 16 and Resolution 173 respectively, already treat transgene-free PE events as non-GMO, providing the fastest commercial pathways globally (Normative Resolution No. 16, of January 15, 2018 - Normative Resolutions - National Technical Biosafety Commission - CTNBio, 2018). The European Union remains the most restrictive context; however, the proposed NGT Regulation (COM/2023/411) would create a two-tier system under which SDN-2 PE events could follow a streamlined risk-based pathway rather than a full GMO dossier (EUR-Lex - 52023PC0411 - EN - EUR-Lex,2023). In South and Southeast Asia, where rice food security is most critical, national frameworks in India, the Philippines, Vietnam, and Indonesia are still under active development.

Achieving regulatory exemption in most jurisdictions requires that the final plant material contain no detectable exogenous DNA. Three principal strategies are available to meet this requirement in rice. First, and most practically accessible, is T0 Mendelian segregation, in which plants stably transformed with the PE cassette are self-ed across one to two generations; approximately 25% of T1 progeny inherit the intended edit while segregating away the transgene, which can be confirmed by PCR screening (Lin et al., 2020; He et al., 2022). This approach is efficient and broadly applicable but requires additional breeding generations. Second, ribonucleoprotein (RNP) delivery of preassembled PE protein and pegRNA directly into rice protoplasts avoids any DNA integration event, yielding truly transgene-free edited plants in a single generation (Toda et al., 2019). While the large size of PE proteins (~200 kDa) currently limits RNP transfection efficiency compared to Cas9, optimization of protein formulation, chemical stabilization of pegRNAs, and electroporation parameters is rapidly improving outcomes. Third, transient plasmid delivery, without selection pressure, allows PE expression over a short window after which the episomal construct is diluted through cell division, combining reasonable editing efficiency with a low integration risk that can be excluded by routine PCR genotyping (Lin et al., 2020). Across all strategies, the use of engineered pegRNAs (epegRNAs) bearing 3’ structured motifs is strongly recommended, as these significantly reduce unintended pegRNA scaffold sequence incorporation at the target locus and improve edit fidelity (Nelson et al., 2022). As delivery technologies continue to evolve including lipid nanoparticle and virus-based platforms currently in early development for rice, 100% transgene-free PE without protoplast culture may become feasible, further simplifying both the technical and regulatory pathway to PE-improved rice varieties.

In addition to these limitations, genome-wide specificity is a critical aspect of PE safety that needs to be carefully evaluated. Although PE is widely regarded as a high precision genome editing technology, whole genome sequencing (WGS) analysis of rice studies has shown a more complex picture. Several studies exhibit very low to undetectable off-target mutation frequencies, especially with PE2 or PEmax systems, which support the high specificity of the system (Hua et al., 2020; Tang et al., 2020). However, the risk profile is dependent on the PE system used. PE3-based approaches, based on the use of dual nicking, have been associated with increased frequencies of unintended insertions and deletions (indels), probably because of the formation of double-strand break-like intermediates (Anzalone et al., 2019; Chen et al., 2021). In addition to small indels, emerging evidence suggests that low-frequency large deletions or genomic rearrangements may occur, although these events remain insufficiently characterized in rice. Another key consideration is possible integration of pegRNA scaffold sequences into the genome. While the unintended inclusion of reverse-transcribed pegRNA fragments has been reported in some editing systems, creating issues in biosafety and specifically for regulatory approval, as well as for downstream applications in breeding (Nelson et al., 2022).

However, three optimization strategies have been reported to further enhance the efficiency of PE. The first approach involves engineering PE guide RNAs (epegRNAs), by adding structured motifs to the 3′ terminus of pegRNAs. By doing this, epegRNAs are less likely to be damaged or degraded (Nelson et al., 2022). The second approach involves the improvement of the PE2 protein, an essential component of the editing process, by incorporating an additional nuclear localization signal and a new linker between nCas9 and reverse transcriptase (Chen et al., 2021). The third approach proposes inhibiting DNA mismatch repair mechanisms to prevent correction of the introduced edits by the cell (Chen et al., 2021). Moreover, Jiang et al. (2022) tested an optimized PE system that conferred glyphosate resistance by targeting the EPSPS gene. The ePE3max system incorporates the PE2 protein, an epegRNA, and a nicking sgRNA. This system was experimented on rice protoplasts to generate rice mutant lines that harbored the TAP-IVS mutation, resulting in a significant improvement in editing efficiency in the transgenic lines compared to conventional PE (Jiang et al., 2022). Improvements in pegRNA design, RT engineering, and delivery technologies, such as nanoparticle-based or virus-based systems. Recent studies have detailed that extending pegRNAs from RNA polymerase II has the potential to overcome the poly-T-associated transcriptional termination, resulting in broadened editable sequences and improving the efficiency of large insertions. This improvement addresses the sequence limitation of traditional PE. However, delivery and efficiency of large insertions have yet to be addressed, advances in engineering pegRNA architecture and delivery methods would significantly advance the applicability of PE in rice (He et al., 2025). In addition, multiplex PE approaches are a promising approach for simultaneous enhancement of multiple agronomic traits, shortening breeding schedules and decreasing costs, as has been shown for other applications of genome editing (Cao et al., 2024).

Next-generation prime editing technologies, including twin prime editing (twinPE) and Programmable Addition via Site-specific Targeting Elements (PASTE) have great potential in the engineering of plant genomes (Anzalone et al., 2022; Chen and Liu, 2023a). twinPE enables large-scale genomic modifications, including targeted insertions, deletions and inversions, that may be useful in the construction of complex traits and gene stacking in rice (Anzalone et al., 2022). Similarly, PASTE enables the incorporation of large DNA fragments without inducing double-strand breaks, thereby creating new possibilities for the introduction of whole biosynthetic pathways or a cluster of resistance genes (Chen and Liu, 2023a).

Recent studies have demonstrated the applicability of the technique of prime editing (PE) for the precise insertion of epitope or fluorescent tags into the gene, as well as the modification of the non-coding sequences of the gene, including the promoters, enhancers, and long non-coding RNAs, thus broadening the applications of the technique in functional genomics as well as crop improvement (Li et al., 2023; Qiu et al., 2025; Van Vu et al., 2025). However, their current applications remain constrained by several technical and biological limitations especially in monocot plants such as rice because of the difficulty in the delivery efficiency, accuracy, and regeneration system of plants which together increase the likelihood of suboptimal editing outcomes and necessitates extensive empirical optimization (Doman et al., 2022; Wang et al., 2022; Chen and Liu, 2023a). Future research should therefore focus on optimizing these technologies for plant-specific contexts, improving delivery platforms, and evaluating their performance under field conditions.

The primary bottleneck in genome editing for plants is the dependency on conventional methods of delivery. The most commonly used approach nowadays for the delivery of T-DNA via Agrobacterium mediated delivery with ‘transient’ expression or using gold bombardment and polyethylene glycol (PEG) mediated transfection to deliver ribonucleoproteins (RNPs) and messenger RNA (mRNA) (Lu et al., 2025). However, these types of delivery mechanisms are typically highly genotype-specific, as well as requiring lengthy phases of tissue culture. Accordingly, the goal of ongoing research is to find transgene-free delivery platforms that can allow for the ability to perform deliveries regardless of genotype instead of going through cell wall barriers, such as nanoparticle-based and virus-based systems.

The actual effectiveness of PE in the field is quite variable. This is due to genotypic and target locus-dependent variability (Cao et al., 2024; Hua et al., 2020). Also, cellular MMR actively counteracts PE by restoring the edits that were introduced (Ramchandani et al., 2023). Researchers are working to address this reduction in efficiency through structural and biochemical engineering. The use of NGS technologies (such as PE4 and PE5) that employ a MLH1dn protein to transiently inhibit MMR has proven to be very effective; they dramatically increase the retention of intended edits and decrease the number of unintended indels (Chen et al., 2021). Similarly, engineered pegRNAs (epegRNAs) with folded 3’ ends eliminate RNA degradation, enhancing efficiency (Nelson et al., 2021).

For agricultural biotechnology to be commercially possible throughout the world, we have to move from single-locus proof of concept edits to the capacity for high-throughput multiplexing of many genes. Multiplexing has been successfully demonstrated to permit simultaneous enhancement of multiple agronomic traits; thus, reducing the amount of time needed to breed a new variety and reducing the cost to do so (Cao et al., 2024). Furthermore, reliability of target design is important for scalability. Advanced machine learning tools such as Easy-Prime, which uses previously published data from all of the published pegRNA sequences created using and OPED, that is a deep-learning algorithm utilizing deep transfer learning to provide researchers with an efficient prediction of the optimal pegRNA design for their target. Consequently, these tools enable us to complete automated high-throughput screening of pegRNA constructs (Li et al., 2021; Liu et al., 2023). By reducing the need for trial and error associated with developing pegRNA constructs, these computational models can ensure reproducible and scalable editing across an essentially unlimited number of rice genomes.

GRAND systems, which rely on engineered pegRNAs (epegRNAs) and extended RT templates, offer enhanced flexibility for simultaneous insertions and substitutions, yet remain highly dependent on sequence context, pegRNA stability, and secondary structure, all of which can significantly affect editing efficiency (Wang et al., 2022). Despite these advances, both TwinPE and GRAND remain largely constrained in their ability to introduce large DNA insertions. While TwinPE can facilitate moderately sized insertions more efficiently than single pegRNA approaches, editing efficiency declines sharply with increasing insert length, limiting its applicability for large gene integration without the use of additional recombinases (Anzalone et al., 2022). Likewise, GRAND-mediated insertions are typically restricted to short to intermediate sequences, with peak efficiency at approximately 250 bp, as longer RT templates reduce processivity and increase the probability of incomplete or erroneous DNA synthesis (Wang et al., 2022).

In addition to size constraints, these systems also present an elevated risk of unintended recombination events or byproduct formation (Chen and Liu, 2023a). The generation of multiple DNA flap intermediates during paired editing or extended reverse transcription can lead to misalignment, partial integrations, or byproduct formation, including indels and hybrid sequences (Doman et al., 2022). These challenges are particularly important in plant and mammalian systems where DNA repair pathways may variably resolve these intermediates (Huang and Liu, 2023).Collectively, although TwinPE and GRAND represent important innovations toward more versatile and precise genome editing, their increased architectural complexity, limited scalability for large insertions, and susceptibility to recombination-related byproducts underscore the need for further optimization. Improvements in pegRNA design algorithms, reverse transcriptase engineering, and control of DNA repair pathways will be essential to enhance their robustness and broaden their applicability in both basic research and applied biotechnology (Doman et al., 2022; Chen and Liu, 2023a).