The CRISPR/Cas9 system is an RNA-guided adaptative immune system in prokaryotes that targets foreign DNA where CRISPR/Cas stands for clustered regularly interspaced short palindromic repeats associated with Cas nuclease. This system emerged during the evolution of archaea and bacteria to prevent the invasion of these organisms by viruses (Barrangou et al., 2007; Jinek et al., 2012; Chylinski et al., 2014). There are two classes and 6 types of CRISPR/Cas systems known to date. Class 1 has effector modules composed of multiple Cas proteins, whereas the class 2 CRISPR mechanism requires a single Cas protein (CRISPR-associated protein) (Jinek et al., 2012; Chylinski et al., 2014). In class 2, Cas9 and Cas12 are DNA nucleases, whereas Cas13 is an RNA nuclease. In this review, we focus mainly on the widely used Cas9-based system; for further information about other CRISPR/Cas systems, see, for example, (Hille et al., 2018). See Figure 1 for a Chronological overview of major genome editing innovations.

In prokaryotes, the CRISPR repeat array is transcribed into a precursor RNA, which contains multiple CRISPR RNAs (crRNAs). Each of these crRNAs contains a single 20-base pair sequence that is complementary to invading DNA (Mojica et al., 2009; Gasiunas et al., 2012; Jinek et al., 2012; Chylinski et al., 2014), and repeats of conserved sequences that are complementary to a section of a transactivating CRISPR RNA called tracrRNA. The primary transcript is then processed into individual crRNAs by ribonuclease III (RNase III). The crRNA-tracrRNA complex interacts with the Cas9 protein to form an active RNA-guided nuclease. A protospacer adjacent motif (PAM) sequence, NGG, where “N” is A, T, C or G, is required for the binding of the Cas9 protein to a target sequence complementary to the spacer sequence.

The PAM acts as a sequence that distinguishes “self” from “non-self”, and PAMs are absent from bacterial chromosome targets (Mojica et al., 2009). Cas9 endonuclease cleaves the target DNA (Barrangou et al., 2007; Gasiunas et al., 2012; Jinek et al., 2012) adjacent to PAM sequence in this case called protospacer. Alternative type II systems (Chylinski et al., 2014), such as Cas12a (previously known as CPF1), recognize a different PAM sequence, i.e., TTTV, where “V” is A, C, or G, and induce double-strand breaks with cohesive ends (Zetsche et al., 2015; Safari et al., 2019; Alok et al., 2020). Other Cas proteins identified subsequently also recognize alternative PAMs (Shah et al., 2013; Leenay et al., 2016). Among type II proteins, Streptococcus pyogenes Cas9 (SpCas9) has been extensively used and modified for biotechnological applications.

A major challenges of the development of this technology lies in reducing unintended DNA cleavage events, the off-target phenomena (Hsu et al., 2013; Tsai et al., 2015; Doench et al., 2016; Cameron et al., 2017; Lazzarotto et al., 2020; Pacesa et al., 2022b), to enhance genome editing specificity. Table 1 summarizes all of the GE specificity and efficiency approaches described in the following paragraphs. See also Figure 3a for a schematic view of the mode of action of the SpCas9 sgRNA complex.

There are many orthologs to SpCas9 from different prokaryotic organisms that have been used in plants, such as SaCas9 (Staphylococcus aureus Cas9) (Steinert et al., 2015), iSpyMacCas9, a hybrid between the PAM interacting (PI) domain of SpCas9 and the PI domain of Cas9 SmacCas9 (Streptococcus macacae Cas9) (Sretenovic et al., 2021b), St1Cas9 (Streptococcus thermophilus Cas9) (Steinert et al., 2015), Nm1Cas9 and Nm2Cas9 (Neisseria meningitidis Cas9) (Xu R. et al., 2022) or ScCas9 (Streptococcus canis Cas9) (Xu et al., 2020b). They offer certain advantages over SpCas9, such as recognition of alternative PAMs (Table 2), but have variable efficiencies and fidelity. SaCas9 is the most interesting alternative to SpCas9 in plant genome editing, with comparable or even superior editing efficiency than SpCas9 in many plant species (Steinert et al., 2015; Kaya et al., 2016; Jia et al., 2017; Qin et al., 2019; Zhang et al., 2022). Unlike SpCas9, SaCas9 recognizes a more specific PAM sequence (5′-NNGRRT-3′) which may limit the range of editing target regions. To broaden the number of targetable sites, the SaCas9-KKH variant, incorporating E782K/N968K/R105H mutations, recognizes an expanded PAM (5′-NNNRRT-3′) (Kleinstiver et al., 2015a) and appears to be as effective as the original nuclease in rice (Qin et al., 2019). Additionally, a high-fidelity variant of SaCas9 (N260D mutation) has been developed to minimize off-targets but has not yet been used in plants (Xie et al., 2020).

Knock-in consists of introducing complex modifications such as, for instance, HA tags, introduction of a reporter gene such as GFP, insertion of an enhancer into a promoter. The use of SpCas9 has significantly advanced targeted insertion, by enabling precise genome editing coupled with the activation of cellular repair systems such as NHEJ and HDR (homologous DNA repair), leading to insertion.

The NHEJ-KI technique in plants requires codelivery of SpCas9 complex, which targets the inserted zone, and a DNA template to be inserted (ssDNA oligonucleotides, dsDNA, plasmids, PCR products, etc.). Efficient insertion requires the simultaneous delivery of a large quantity of donor DNA with SpCas9 to reduce indel formation and increase the likehood of template’s presence near the DSB site. Bringing the matrix to be inserted close to the target also improves the insertion rates (Aird et al., 2018; Ali et al., 2020).

Lu et al. used this approach to insert tags into the rice genome (Lu et al., 2020). They first reported that it was possible to insert tags using dsDNA oligos but not ssDNA. Modification of oligonucleotides at the 5′end by phosphorylation to promote NHEJ and at the 5′and 3′ends by phosphorothioate linkage to protect against endogenous exonucleases strongly improved KI rates. Using 60-bp tags, they achieved insertion efficiencies of approximately 25%, i.e., a 5–6-fold improvement over unmodified oligonucleotides for several targets (Lu et al., 2020). To test the insertion of larger fragments, they produced matrices for insertion via PCR of fragments with protected oligonucleotides (Lu et al., 2020). The efficiency decreased with increasing size of the inserted fragment (5% with 2 kb), and the rate of deletions at the junction increased significantly, probably because only one strand of the PCR products was protected, in contrast with oligonucleotides protected on both strands.

Similar strategies, without end protection, were developed, and fragments of several kilobases were successfully inserted, with efficiency rates of 2%–3%, suggesting that it is indeed possible to insert long fragments by biolistic techniques via NHEJ but at the cost of low efficiency (Li et al., 2016; Xu et al., 2020c). Finally, to achieve seamless insertions and sequence replacements, Lu et al. introduced the tandem repeat HDR (TDR-HDR) method, which combines initial NHEJ-mediated insertion of a first sequence followed by HDR-recombination facilitated by a second sRNA. The second sgRNA is used to cleave the sequence at the first inserted oligonucleotide, stimulating recombination via HDR between the two homologous fragments (Lu et al., 2020). This technique can be used to insert any sequence with an efficiency of approximately 15%.

While effective for sequence insertion or replacement, biolistics methods as opposed to Agrobacterium tumefaciens delivery, raises known problems, including extensive genomic rearrangements (deletions, duplications) and multiple inserted transgenes (see, for example, (Liu J. et al., 2019; Banakar et al., 2019), for a discussion of biolistic drawbacks). Currently, most technological developments for knock-in revolve around the use of prime editing (PE) and dual pegRNA.

One strategy to improve PE is to introduce specific mutations into SpCas9 that enhance its editing efficiency. Simultaneous introduction of the R221K and N394K mutations have been shown to double editing efficiency, across eight distinct human genomic targets (Spencer and Zhang, 2017). Located at the interface of the REC1 and REC2 domains, these mutations probably facilitate HNH positioning and SpCas9 cleavage activity. Incorporating these mutations into the prime editor, led to a fourfold increase in editing efficiency in plants systems (Li J. et al., 2022; Ni et al., 2023; Qiao et al., 2023). These mutations may affects nickase’s cleavage efficiency and/or enhance pegRNA binding to its target. Unlike the D10A nickase, the H840A nickase which cut the nontargeted DNA strand has residual DSB activity (Lee et al., 2023). This residual activity is responsible for the significant rate of mutations induced by the NHEJ repair system. The mutation rate depends on the region targeted (Lee et al., 2023). Introduction of N854A and N863A mutations into the H840A nickase eliminate its residual DSB activity while maintaining editing efficiency, thus significantly enhancing specificity (Lee et al., 2023). This modification is obviously of interest for improving the specificity of PE for use in gene therapy but would also be useful in plants and use D10A nickase in prime editors thus represent also an interesting alternative to reduce indel formation but to our knowledge, it has never been tested in plants.

The RNAse H domain of MMLV degrades viral RNA in heteroduplexes after retrotranscription. Its deletion have been shown to stabilize the pegRNA/DNA heteroduplex during PE, leading to a threefold increase editing efficiency as demonstrated in (Zong et al., 2022; Ni et al., 2023). Removing of both the RNAse H and the adjacent connection domain from MMLV reverse transcriptase completely abolished PE, suggesting the connection domain is necessary for the proper function of MMLV (Ni et al., 2023). Furthermore, the same authors demonstrated that the incorporation of a nucleocapsid protein, functioning as a chaperone for MMLV reverse transcriptase, also improved PE efficiency. Additional studies suggest that while the RNAse H-deficient version generally improves prime editing efficiency, it may lead to increased indel mutations with highly structured reverse transcriptase templates (RTT) (Doman et al., 2023). To improve pegRNA reverse transcription, researchers analyzed the effects of specific mutations described to improve MMLV reverse transcriptase activity in wheat PE (Ni et al., 2023). Introducing the V223A mutation resulted in an average sixfold increase in editing efficiency. This mutation has been associated previously with increase processivity and faster reverse transcription compared to the wild type enzyme (Paliksa et al., 2018). Other retrovirus-derived reverse transcriptases have been tested, such as those derived from cauliflower mosaic virus (CaMV) to enhance prime editing in rice and wheat (Lin et al., 2020). Although the CaMV-based prime editor works with an efficiency comparable to MMLV (Lin et al., 2020), none have been found to be superior. MMLV reverse transcriptase comes from animal systems and has been optimized for prime editing (Anzalone et al., 2019). CaMV-derived reverse transcriptase, which comes from plants, has never been optimized. We can only speculate that introducing mutations analogous to those in MMLV (e.g., D200N, L603W, T306K, W313F, and T330P) could potentially improve the efficiency of CaMV-based prime editors in plant systems (Anzalone et al., 2019). Introducing these or structurally analogous mutations, guided by predicted reverse transcriptase protein structures has been shown to significantly enhance PE efficiency of alternative RTs (Doman et al., 2023). Despite extensive efforts, involving mutagenesis and phage-assisted evolution, none of the reverse transcriptases developed have surpassed MMLV’s efficiency in human cells prime editing applications (Liu et al., 2022; Doman et al., 2023). Interestingly, Cao et al. (2024) reported that PE6c, incorporating an engineered RT from the yeast Tf1 retrotransposon, and PE6d, a MMLV variant, achieved 2–3.5-fold higher editing efficiency compared to PE3. Conversely, Xu et al. (2024) reported that PE6c was less efficient and PE6d equivalent to PE2 for small edits insertion in rice. Further experiments are thus needed to provide a final conclusion on the efficiency of these new versions of PE in plants. These novel, more compact RTs with efficiencies similar to MMLV, are particularly promising for applications where vector size is a limiting factor, such as RNA virus-mediated delivery systems for plant prime editing.

PegRNA secondary structure can lead to misfolding, which negatively impacts editing efficiency by promoting unfavorable intramolecular interactions. Although sgRNA are less prone to misfolding, their secondary structures can still influence editing. For example, in a study testing the structural determinants of the editing efficiency of many sgRNAs, the self-folding energy and Tm of the sgRNA were among the factors that were found to most strongly influence editing (Wang et al., 2019). Key factors influencing pegRNA functionality include: its availability and stability, the resistance of its 3′-end to exonucleases degradation, and its secondary structure, which affects interactions with the nCas9-MMLV complex and the hybridization efficiency of the PBS to the non-targeted strand. While sgRNAs are largely shielded from 3′exonuclease activity upon binding to SpCas9, the addition of 3′extensions in pegRNAs exposes them to degradation. This degradation results in formation of competing truncated pegRNAs that can still associate with nCas9 but are ineffective for prime editing (Nelson et al., 2022). Incorporating structured RNA motifs, such as the 42-nucleotides evopreQ1, at the of 3′-end of pegRNA enhances their stability and has led to significant improvements in PE efficiency (Nelson et al., 2022) in human cells. These improved pegRNA were termed enhanced pegRNA (epegRNA). To minimize unintended interactions between the structured motif and the pegRNA, the authors suggest inserting an 8-nucleotide random linker designed using the pegLIT software. Finally, they also demonstrated that incorporation of evopreQ1 could influence pegRNA transcription leading to a recommendation for using enhanced promoters to avoid a trade-off between pegRNA protection and transcription (Nelson et al., 2022). Additional 3′modifications to pegRNAs have been shown to improved PE efficiency (Liu et al., 2021; Li et al., 2022b), probably by providing increase resistance to exonuclease degradation. Implementing the evopreQ1 motif at the 3′ end of pegRNA, i.e., using epegRNA, significantly improved the PE efficiency in several plant species (Jiang et al., 2022; Li J. et al., 2022; Zong et al., 2022; Ni et al., 2023; Qiao et al., 2023).

An intrinsic feature of pegRNA design is the potential for intramolecular base pairing between the primer binding site (PBS) and the spacer sequence, as both target overlapping regions. This intramolecular interaction is responsible for an autoinhibitory effect that affects target binding and initiation of reverse transcription (Liu et al., 2021; Ponnienselvan et al., 2023). This autoinhibitory effect was demonstrated by substituting PE with Cas9 and using pegRNA to cleave the target, revealing reduced activity (Liu et al., 2021; Vu et al., 2022; Ponnienselvan et al., 2023). Using pegRNA instead of sgRNA has been shown to abolish or diminish editing efficiency in both mammalian and plant cells (Ponnienselvan et al., 2023). Historically, the most efficient PBSs were 11–13 nucleotides long (Anzalone et al., 2019; Kim et al., 2021), while recent studies found that shortening the PBS to 7–8 nucleotides, alleviated autoinhibition, restoring editing with Cas9 and pegRNA (Liu et al., 2021; Ponnienselvan et al., 2023).

How can the conflicting findings regarding optimal PBS lengths be reconciled? Historically, pegRNAs lacked 3′protected structures probably leading to partial degradation by endogenous exonucleases necessitating longer PBS regions to maintain functionality. Integration of 3′protectives structures makes it possible to use shorter PBSs and limiting or even eliminating the autoinhibitory effect associated with longer PBSs in mammalian cells (Liu et al., 2021; Ponnienselvan et al., 2023). In plants, the impact remains unclear, since reducing the PBS size did not enhance editing efficiency in tomato (Ponnienselvan et al., 2023). However, the pegRNAs used in these experiments were not 3′-protected and rendering them vulnerable to partial degradation by exonucleases, as noted by authors. The melting temperature of the PBS is another critical parameter influencing prime editing efficiency. The optimal melting temperature of PBS for a large set of pegRNAs in rice and mammalian cells corresponded to optimal growth temperatures of the host cells: 30 °C for rice (Lin et al., 2021) and 37 °C for mammalian cells (Ponnienselvan et al., 2023), respectively.

Another strategy to improve PE for indels is to modify the hairpin 1 of the pegRNA. Unlike shorter sgRNA, pegRNA carry additional RTT and PBS sequences at their 3′end, which can disrupt hairpin 1 stability and lead to overall pegRNA misfolding. Replacing the G/A pair at the base of hairpin 1 with a C/G pair yielded the so-called ‘apegRNA,’ enhancing prime editing efficiency roughly threefold (Li et al., 2022c). Interestingly, this approach echoes the GOLD strategy, developed to improve genome editing by increasing stability of hairpin 1 (Riesenberg et al., 2022). Authors have shown that suboptimal sgRNA suffer from unwanted 3′-spacer base pairing with the spacer and locking hairpin 1 restored their activity (Riesenberg et al., 2022). GOLD stabilization strategy to pegRNA may similarly enhance prime editing efficiency and warrants experimental evaluation. To the best of our knowledge, these simple yet elegant strategies have not yet been evaluated in plants.

Multiple studies have reported that employing composite promoters (Li J. et al., 2022; Ni et al., 2023; Qiao et al., 2023) or utilizing viral amplicons (Vu et al., 2024) can substantially enhance prime editing efficiency in plants. There are many reasons for this: increased pegRNA transcription levels, as RNA polymerase II promoters are more effective at transcribing long RNAs; and improved pegRNA folding facilitated by incorporating a 5′tRNA and a 3′HDV ribozyme, which are cleaved during maturation. This is particularly true for low-efficiency sgRNAs, indicating that higher expression levels of pegRNA/sgRNA are necessary to achieve effective edition (Yuen et al., 2017).

The mismatch repair (MMR) system corrects errors during replication by detecting mismatches and identifying the newly synthesized strand through nearby DNA nicks. Anzalone et al. introduced the PE3 system, which employs an additional sgRNA to nick the unedited strand at a distance from the edited strand, thereby enhancing the likelihood that the MMR system will replace it using the edited strand as a template (Anzalone et al., 2019). However, simultaneous nicking of both DNA strands can lead to DSBs, which may be repaired by the non-homologous end joining (NHEJ) pathway, potentially resulting in indel mutations. To mitigate this issue, Anzalone et al. developed the PE3b system, where the additional sgRNA targets the edited strand, enabling for sequential nicking limiting the incidence of unintended indel (Anzalone et al., 2019). Unfortunately, the PE3 and PE3b versions did not improve PE in plants for unknown reasons (Lin et al., 2020; Xu R. et al., 2020).

The 3′ ssDNA flap carrying the edited sequence competes with the 5′flap derived from the unedited strand for integration into the genome (Anzalone et al., 2019). Two approaches have therefore been explored to enhance 3′flap incorporation during prime editing. The first approach is to inhibit repair pathways that are deleterious for PE and 3′flap degradation. In bacteria, deletion of three exonucleases has been shown to enhance prime editing efficiency by up to 100-fold (Zhang et al., 2024). Combining PE with Cas12a-mediated CRISPR interference of exonucleases further boosts editing efficiency in bacterial models (Zhang et al., 2024). In mammalian cells, suppression of specific MMR components has also been found to increases prime editing efficiency (Chen P. J. et al., 2021; Ferreira da Silva et al., 2022). Unfortunately, attempts to replicate MMR inhibition strategies in plants, such as coexpressing dominant-negative forms of MLH1 dh and MutS, have not significantly improved PE in rice and tomato (Li J. et al., 2022; Vu et al., 2022). However, RNAi-mediated knockdown of OsMLH1 improved PE in rice (Liu X. et al., 2024). While the precise genetic factors influencing PE in plants remain unclear, comparisons across species suggest that the repair pathways involved may differ significantly between bacteria, humans, and plants. In any case, of MMR inhibition must be temporary as prolonged suppression can elevate mutation rate and compromise genomic stability.

An alternative strategy involves facilitating the removal of the 5′DNA flap by incorporating a 5′to 3′exonuclease into the prime editing system. In human cells, the recruitment of bacteriophage T5 exonuclease through PP7 (Pseudomonas bacteriophage) RNA aptamers inserted at the pegRNA’s tetraloop has proven to be an effective system in human cells (Truong et al., 2024). This approach generally increases PE efficiency in comparison with the PEmax system, but above all, it improves PE specificity for insertion ranging from 30 to 60 bp pairs, which is consistent with the idea that this system favors integration of longer 3′flaps. A comparable strategy was applied in rice, where fusing the same to the N-terminus of the prime editor led to a 1.7- to 2.9-fold increase in editing efficiency varying with the target site (Liang et al., 2023). Interestingly, using a similar aptamer mediated exonuclease strategy as described in (Truong et al., 2024), resulted in reduced prime editing efficiency in his context (Liang et al., 2023). The authors utilized MS2 aptamers inserted in the 3′end of the pegRNA, which may have negative effect on pegRNA binding to its target compared to tetraloop insertions (Truong et al., 2024), thereby reducing PE efficiency (Liang et al., 2023).

Prime editing efficiency depends on favoring the integration of the edited DNA strand over the original wild-type strand. The Mismatch repair (MMR) system typically targets edited strand for correction, as it preferentially recognizes nicks introduced during editing. Enhancing the incorporation of the edited strand, requires to bias the MMR to favor its integration. A strategy involves inhibiting MMR by introducing multiple synonymous mutations in and around the protospacer adjacent motif (PAM). Altering the PAM or the adjacent seed region through mutations prevent nCas9 from re-binding and re-cutting the edited strand, thereby reducing MMR recognition and enhancing prime editing efficiency (Anzalone et al., 2019). Introducing multiple synonymous mutations can hinder MMR recognition, since MMR complex such as Msh2-Msh6 primarily detect single-base mismatches and small indels, whereas Msh2-Msh3 targets larger indels which is consistent with the idea that inhibiting of the MMR enhance prime editing efficiency (Ferreira da Silva et al., 2022). Consequently, multiple substitutions are less efficiently recognized by MMR, reducing the likelihood of the edited strand being corrected back to the wild-type sequence. Finally, targeting the coding strand (non-RNApolymerase template strand) of actively transcribed regions for editing may be advantageous as it is not used as a template for correction during transcription. Indeed, transcription-coupled repair, along with mismatch repair, preferentially monitors and corrects the template (non-coding) strand when heteroduplexes are present (Georgakopoulos-Soares et al., 2020). Therefore, editing the coding strand can also therefore theoretically increase the rate of prime editing, particularly when combined with introduction of multiple synonymous mutations.

This strategy of introducing multiple synonymous mutations has been effectively applied in both animal and plant cells. For instance, introducing same-sense mutations (SSMs) or silent mutations at positions +1, 5, 6, 2/5, and 3/6, relative to the nicking site, has been shown to strongly enhance PE efficiency, by an average of 350-fold, particularly for pegRNAs with very low initial efficiency (Li et al., 2022c). These modified pegRNAs, termed spegRNAs are compatible with PE2 or PE3 systems, albeit demonstrating enhanced synergistic effects when used with PE3 (Li et al., 2022c). In rice, Xu, et al. demonstrated that introducing mutations within the RTT, at the PAM or PAM-proximal region strongly enhanced prime editing efficiency (Xu W. et al., 2022). Li, X. et al. proposed guidelines for introducing SSMs in single-base PE, suggesting placement at +3/+6 when substituting the first base after nicking, at +1 for the second base, and at +2/+5 the third base (Li et al., 2022c). This strategy maintains the reading frame and ensures that only the targeted amino acid is modified. Combining spegRNA and apegRNA, where apegRNA correspond to a substitution of the G/A pair at the hairpin 1 base by a C/G, have shown synergistic effects (Li et al., 2022c). Interestingly, sapegRNAs also enhance PE efficiency in the PE2 system by approximately threefold, although this improvement is less pronounced compared to their effect in the PE3 system (Li et al., 2022c).

An interesting approach to enhance prime editing efficiency is to use a dual pegRNA system, where two pegRNA are designed to introduce identical modifications on both the 5′and 3′strands [see, for example, (Lin et al., 2021; Choi et al., 2022)]. This strategy therefore requires the design of two pegRNAs along with compatible nicking sites to facilitate simultaneous editing on both strands. Tools like PlantPegDesigner (Lin et al., 2021) provide design assistance for dual pegRNA strategies when applicable, aiming to optimize PE efficiency. This strategy is interesting, although its applicability depends on the availability of suitable target sites.

Improving the efficiency and specificity of base editing (BE) and prime editing (PE) requires leveraging improvements made in native SpCas9 alongside technology-specific modifications. A rational approach to SpCas9-derived editing technologies should incorporate these optimizations upstream to develop more efficient constructs. For instance, incorporating introns into the SpCas9 has been shown to significantly boost expression and editing efficiency in dicotyledonous plants (Grutzner et al., 2021), suggesting potential benefits for both BE and PE. Similarly, composite promoters have been demonstrated to enhance the expression of inefficient sgRNAs and markedly increase PE efficiency in plant systems (Li J. et al., 2022; Ni et al., 2023; Qiao et al., 2023). Composite promoters should be particularly important for long RNA transcription and multiplexing strategies, which will be crucial for breeding applications. Further, increasing SpCas9 editing efficiency directly enhances PE efficiency (Li J. et al., 2022; Ni et al., 2023; Qiao et al., 2023), as demonstrated in PEmax, which incorporates double mutations in nCas9 (Spencer and Zhang, 2017) or a bipartite nuclear localization signal (BP-NLS) to improve nuclear targeting and editing rates compared to single NLSs (Develtere et al., 2024). The development of SpCas9-derived technologies can therefore benefit from advances in SpCas9’s efficacy and specificity, and must therefore be taken into account in any current or future editing technology. GE and PE technologies have proven valuable in plant breeding, contributing to traits such as enhanced grain quality (Zhou et al., 2019) and broad-spectrum resistance to bacterial blast (Gupta et al., 2023) in rice. Notably, genome-edited crops like GABA-enriched tomatoes (Waltz, 2022) and high-oleic acid soybean (Demorest et al., 2016) have reached commercial markets. Beyond efficiency, specificity will become a critical factor for the routine application of genome editing technologies in breeding programs, particularly within Europe. The European Commission has proposed a threshold of 20 genetic modifications, mirroring changes achievable through conventional breeding, to classify such genome-edited plants equivalently to traditional bred counterparts (Organisms et al., 2024). This includes the targeted insertion of a contiguous DNA sequence already present within the gene pool.

Genome editing technologies rely on endogenous DNA repair mechanisms, with numerous modifications designed to modulating specific pathways to increase editing efficiency and precision. For example, MMEJ can be exploited to induce targeted deletions by carefully selecting sgRNA that promote this repair pathway (Martinez-Galvez et al., 2021). A novel Cas9 variant, vCas9, introduces staggered DNA cuts, thereby favoring repair via MMEJ and HDR pathways over NHEJ (Chauhan et al., 2023). Base editing (BE) techniques modulates DNA repair by inhibiting BER through the use uracil DNA glycosylase inhibitors (UGIs). Enhanced versions like BE4 and BE4max employ dual UGI system to strengthen this inhibition, while nicking the unedited DNA strand stimulate MMR to favor insertion of the desired edit (Komor et al., 2017; Koblan et al., 2018). Prime Editing (PE) efficiency can be enhanced by suppressing of MMR, either through conditional RNAi targeting OsMLH1 in rice or by introduction of multiple silent mutations (SSMs) within the RTT of pegRNA (Xu W. et al., 2022). PE3b enhances editing efficiency by introducing a second sgRNA to nick the unedited strand, facilitating precise repair stimulating MMR of the unedited strand (Anzalone et al., 2019). Beyond direct genome editing, CRISPR-based transcriptional modulation techniques such as CRISPR activation (CRISPRa) and interference (CRISPRi) can be also use to targeting key repair genes using catalytically inactive sgRNA termed dead sgRNAs (dsgRNAs) (Dahlman et al., 2015; Ye et al., 2018), a promising avenue to enhance efficiency and specificity of PE, BE and emerging editing technologies. Furthermore, variations in DNA repair pathway regulation across species may account for observed differences in editing efficiencies, especially between monocotyledonous and dicotyledonous plants. This is not new, the efficiency of stable T-DNA transformation in plants has long been associated with the activity of NHEJ and MMEJ repair pathways (Qi et al., 2013; Saika et al., 2014). Overexpression of Ku80, a pivotal protein in the NHEJ pathway, has been shown to improve transgene integration (Li et al., 2005). A deeper understanding of DNA repair mechanisms and their interplay with genome editing technologies is crucial to overcome existing limitations for applications of plant genome editing technologies.

One of the main limitations to the use of genome editing technologies in plants is genetic transformation, which is still restricted to certain species and/or genotypes. For a recent review on this topic, see (Wang et al., 2025). Among transformation approaches, protoplast-based and biolistic methods offer the advantage of generating transgene-free edits, but they remain limited by laborious regeneration protocols and genotype dependency in most species. Biolistic delivery enables transformation of a wide range of tissues but the approach often causes extensive DNA rearrangements. By contrast, A. tumefaciens mediated transformation remains the most widely used and reliable system for stable integration with fewer somaclonal variants, although many agronomically important species remain recalcitrant and genotype dependent, underscoring the need to expand its applicability to diverse tissues and resistant genotypes.

To overcome this, strategies have been developed to improve transformation efficiency by modulating dedifferentiation regulators (such as GRF, WUSCHEL (WUS) and BABY BBOL (BBM) (Lowe et al., 2016; Debernardi et al., 2020). However, the constitutive expression of these pivotal developmental genes can lead to adverse effects, highlighting the need of alternative methods to enhance transformation and regeneration. An alternative promising strategy involved CRISPR activation (CRISPRa) using dead single-guides RNA (dsgRNAs) with inducible promoters enabling the temporal activation of these regulators within specific time frames (Dahlman et al., 2015). Furthermore, the same strategy can be used to facilitate stable insertions by enhancing the activity of the non-homologous end joining (NHEJ) repair pathway (Li et al., 2005).

In vegetatively propagated crops, the inability to eliminate transgenes through segregation necessitates the development of transgene-free genome editing approaches. This challenge is mainly being addressed by transient transformation of protoplasts, which avoids stable integration of transgenes while still allowing precise genetic modification (Gu et al., 2021). Another approach involves grafting wild-type shoots onto transgenic donor rootstocks that express mobile RNA versions of SpCas9 and sgRNA (Yang et al., 2023). However, efficient regeneration remains a significant hurdle for many species; the use of dedifferentiation regulators has been shown to enhance this process (Lowe et al., 2016; Debernardi et al., 2020). Moreover, the editing efficiency achieved trhough grafting strategies remains too low, often below 0.1%, rendering them impractical for routine breeding applications (Yang et al., 2023). Ultimately, for genome editing to be routinely applied in the breeding of vegetatively propagated crops, it is essential to optimize both transformation and editing efficiency. Without selection markers, the frequency of regenerated plants harboring the desired edit depends on both the success of transformation success and the efficiency of the editing process. Enhancing both transformation and editing efficiencies will minimize the screening required to identify desired mutations, thereby accelerating the adoption of genome editing in crop improvement.

A key challenge in crop improvement is identifying alleles of agronomic interest and tailoring genome editing strategies accordingly. The majority of beneficial alleles have been discovered in model species, limiting their direct application to a wide range of crops species. Addressing this limitation, requires a comprehensive catalogue of beneficial alleles and the identification of their orthologs in target crop species through translational biology, thereby expending the repertoire of potential genetic improvements (Inze and Nelissen, 2022). The specific nature of the desired genetic alteration dictates the choice of genome editing technology.

For example, CRISPR/Cas9 can efficiently generate simple gene knockouts, exemplified by the yield-enhancing GS2 alleles in rice (Wang W. et al., 2022). However, intricate genetic modifications require advanced technologies: for example, base editing (BE) facilitates the introduction of the NRT1.1B allele to enhance nitrogen use efficiency (NUE) (Lu and Zhu, 2017), whereas prime editing (PE) enables the precise insertion of heat stress-responsive elements (HSEs) into invertase promoters in rice and tomato to maintain yield at high temperatures (Lou et al., 2025).

Addressing structural variations, including presence-absence variants (PAVs), like the SUB1 (Xu et al., 2006) or PSTOL1 (Gamuyao et al., 2012) genes in rice, remain challenging; however, emerging PE-based technologies capable of integrating kilobase-scale DNA sequences offer promising solutions (Sun et al., 2024). Combining the effects of these favorable alleles across various target species and genetic backgrounds is a crucial step. In rice, introduction yield-enhancing mutations across various genotypes has resulted in variable outcomes, likely due to unpredictable complex epistatic interactions (Shen et al., 2018). These findings highlight the need for comprehensive field trials and the continuous refinement of genome editing strategies.

This review offers an overview of recent progress in plant genome editing, from SpCas9 to prime editing. Genome editing technologies are becoming increasingly efficient and precise, thereby accelerating plant breeding and facilitating functional gene analysis. Among the most significant advances, multiplexing is pivotal to expediting breeding programs, as it allows the concurrent introduction of multiple advantageous mutations (Zhou et al., 2019; Zhou et al., 2024). Moreover, due to its versatility in facilitating both simple and complex genetic modifications, prime editing holds promise for large-scale, multi-target genome alterations, potentially revolutionizing crop improvement strategies (Gupta et al., 2024). Nevertheless, important limitations remain: the difficulty of achieving efficient transformation in many agronomically relevant species, restrictions on knock-in size, and heterogeneous prime editing efficiencies across targets and species. Future efforts should focus on improving transformation efficiency and delivery systems, as well as developing more robust and efficient prime editors for plant breeding and functional genomics. Finally, we have not yet reached the end of the golden path of genome editing with the advent of new emerging technologies like the bridge RNA (Durrant et al., 2024).