Plant base editing: a decade of progress and future applications

Abstract

Base editors derived from clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) systems are widely used for genomic studies in both plants and animals. The broad applicability and safety of base-editing approaches have garnered considerable attention from the research community, and there could be ways to further enhance targeting efficiency and target window range. However, the complex classification and diverse functionalities of base editors pose challenges to their effective utilization and improvement. In this review, we discuss technical principles characterizing various types of base editors, including cytosine base editors (CBEs), adenine base editors (ABEs), dual base editors (DBEs), thymine base editors (TBEs), and guanine base editors (GBEs), among others, which employ distinct mechanisms and DNA repair pathways. We also describe current optimization strategies to assist researchers in improving the deployment of these tools under specific conditions. Finally, we comprehensively analyze the practical applications and advantages of base editors, offering a clear view of their development, their previous and potential applications, and how to select the appropriate tools for specific purposes.

1. Introduction

The CRISPR (clustered regularly interspaced short palindromic repeats)/Cas (CRISPR-associated protein) system is a heritable RNA-mediated immune system prevalent in bacteria and archaea [1]. Modifications to this system by genetic engineering have led to the establishment of the widely utilized CRISPR/Cas9, CRISPR/Cas12, and CRISPR/Cas13 base-editing systems. Guided by a single guide RNA (sgRNA), the Cas9 protein uses its HNH and RuvC-like nuclease domains to cleave target DNA, generating double-strand breaks (DSBs). DSBs activate recipient cells to initiate repair pathways, including non-homologous end joining (NHEJ) or homologous recombination (HDR), resulting in heritable mutations, such as deletions, insertions, or replacements [[2], [3], [4]] (Fig. 1A).

Fig. 1.

Basic principles of base editing. A Fundamental principle of gene knockout based on clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 (CRISPR/Cas9). B Fundamental principle of base editor without a double-strand break based on CRISPR.

Although this process holds great potential for precise gene knockout, primarily dependent on NHEJ, it presents several challenges, including the uncontrollable generation of different types of mutations and considerable risks of off-target changes [5,6]. Addressing the need for more precise and multifunctional gene-editing tools, researchers developed Cas9 as a protein chassis [7]. Subsequently, CRISPR-based activation, inhibition, base editing, and prime editing technologies emerged [[8], [9], [10], [11], [12]], initiating the era of precise genome editing [13,14]. Base editing (BEs) offers three key advantages over previous genome editing approaches: enhanced precision compared to traditional CRISPR/Cas9 systems through the generation of single-nucleotide modifications without DSB formation; the ability to generate permanent heritable mutations, unlike epigenetic tools, which require continuous intervention; and higher efficiency when using simpler expression cassettes compared to prime editing systems. These characteristics make base editing particularly suitable for precision breeding and therapeutic applications.

BEs are engineered by fusing single-strand DNA (ssDNA) Cas9 nickase (Cas9n) with base deaminases, enabling base modifications within a target window without forming DSBs [3] (Fig. 1B). The enhanced safety profile of BEs makes them valuable for both animal disease models and therapeutic gene editing [15,16]. BEs have been used to establish animal disease models for pathological research and drug development. For example, Yang et al. [17] used adenine base editors (ABEs) to establish a rat model of Pompe disease, Chen et al. [18] used ACBE-Q to generate a mouse model of Duchenne muscular dystrophy, and Xue et al. [19] used hyperactive ABE in zebrafish to simulate β-hemoglobinopathy treatment.

In plant research, base editing is primarily used to generate mutant materials for studying genome function or for molecular breeding. Studies on plant genomes and gene functions have identified numerous single-nucleotide polymorphism (SNP) sites and promoter motifs related to continuous trait mutations, which are challenging to modify using CRISPR/Cas DSBs [20,21]. Jin-Soo Kim's laboratory used a RPS5A promoter-driven ABE to achieve efficient editing in the dicotyledonous plants Arabidopsis (Arabidopsis thaliana) and rapeseed (Brassica napus ssp. napus). The authors produced early-flowering ft mutants in Arabidopsis and achieved alternative mRNA splicing of the PDS gene in rapeseed [22]. BEs have also been widely used for molecular breeding in various other plants, including rice (Oryza sativa), wheat (Triticum aestivum), potato (Solanum tuberosum), soybean (Glycine max), and cotton (Gossypium hirsutum), generating herbicide-resistant and other superior crop varieties [[23], [24], [25], [26]]. Although recent reviews provide valuable insights into plant base0editing systems [27,28], a systematic mechanistic classification of BEs is still needed.

In this review, we describe the development of BEs in plants through a mechanistic lens, systematically categorizing base editors on the basis of their underlying apurinic/apyrimidinic (AP) and non-AP pathway mechanisms and the different substrate bases bound by the editors, including cytosine base editors (CBEs), ABEs (adenine base editors), TBEs (thymine base editors), and GBEs (guanine base editors), with an emphasis on their developmental trajectory and enzymatic pathways. Furthermore, we examine their potential applications in plant breeding and present optimization strategies derived from extensive scientific research to provide a deeper understanding of these applications. This review describes the development and application of base editing technology from a developmental and mechanistic perspective, supporting the further development of this technology and future applications for gene editing.

2. Major types of base editors

2.1. Single-base editors

Single BEs specifically convert one type of base at the target site to another. These editors modify substrate bases using specific enzymes, and this is followed by mismatch repair or glycosylase digestion to produce apurinic/apyrimidinic (AP) sites, thereby achieving precise base modifications. There are multiple types of single BEs. Single BEs are classified into AP site-dependent or AP site-independent types based on their primary repair pathways.

2.1.1. AP site-independent specific base editors

AP site-independent base editors generate deaminated base intermediates through substrate-base deamination and rely on the relaxed base-recognition capacity of translesion synthesis (TLS) polymerase. This mechanism can generate specific base replacements, including C-to-T replacements by CBEs and A-to-G replacements by ABEs [29]. CBEs and ABEs represent the most advanced single BEs and have achieved widespread use in animal and plant research because of their high efficiency, broad applicability, and precise editing outcomes.

2.1.1.1. Cytosine base editors (CBEs)

CBEs were the first single-base editors employed in scientific research. BE1, the first such base editor, was developed by David Liu and colleagues at Harvard University by fusing dCas9 with the highly active rat cytidine deaminase rAPOBEC1 [8]. The dCas9-rAPOBEC1 fusion protein, guided by sgRNA, converts cytosine (C) to uracil (U) through deamination within the editing window at a target position. During DNA replication or repair, U is interpreted as thymine (T), and the corresponding guanine (G) is converted to adenine (A), resulting in the directed conversion of C•G to A•T [[30], [31], [32]] (Fig. 2).

Fig. 2.

Primary components of diverse base editing systems. The green sector depicts non-apurinic/apyrimidinic (AP)-dependent base editors. The orange-yellow sector indicates AP-dependent base editors.

To enhance editing efficiency, the research team subsequently developed BE2 through BE4 based on BE1, substantially improving CBE performance [8,33]. BE2 was developed by incorporating a uracil glycosylase inhibitor (UGI) into BE1 to protect the uracil produced by deamination from hydrolysis by endogenous uracil DNA glycosylase (UDG/UNG), thereby improving editing efficiency. In BE3, dCas9 was replaced by Cas9n, which cleaves only ssDNA. This change ensured that while C at the target point undergoes deamination to form U, the complementary strand with the original sequence is eliminated, thus preventing the restoration of U back to C by the HDR repair pathway. BE4 contains more copies of UGI, enhancing the protection of U, thereby substantially improving editing efficiency and reducing byproduct generation. The establishment of both BE3 and BE4 provided crucial frameworks for further optimization of CBEs.

The evolution of TadA-8e, an adenine-specific deaminase, into a variant capable of cytosine deamination represents a major research breakthrough. Researchers used the phage-assisted continuous evolution (PACE) system to screen TadA variants (TadA-CDã ​TadA-CDe) that deaminate cytosine. The authors introduced N46 and Y73 point mutations and developed CBE6, achieving enhanced editing efficiency and reducing the generation of non-C-to-T base replacement by-products [34,35]. TadA8e-evolved cytosine deaminases offer several advantages: they are smaller than natural cytosine deaminases, maintain high editing efficiency, and have considerably reduced off-target rates. Moreover, their narrower editing window facilitates high-precision cytosine base editing.

2.1.1.2. Adenine base editors (ABEs)

ABEs include the adenine deaminase TadA, which does not occur naturally. Natural systems contain adenine deaminases that act on free adenine, adenosine, and RNA adenosine; however, they do not act on adenine bases in double-stranded DNA or ssDNA. Gaudelli et al. constructed ABE7.10 using the directed evolution of the tRNA-specific adenine deaminase in Escherichia coli through error-prone PCR and selection on multiple antibiotics, ultimately developing the ssDNA adenine deaminase TadA7.10. ABE7.10 is an adenine base editing system in mammalian cells that generates A-to-G conversions [36]. ABEs show similar modes of action to CBEs: adenosine deaminase converts A to inosine (I) through deamination, and I is transformed to G during repair. The corresponding T on the complementary strand is converted to C, achieving directed A•T to G•C conversion. The primary advantage of ABEs lies in the structural similarity between I and G, which results in minimal excision-induced insertions or deletions (indels), leading to more precise adenine base editing (Fig. 2).

2.1.2. AP site-dependent random base editors

BEs that depend on AP sites operate primarily through the base-excision repair (BER) pathway. They utilize glycosylases to eliminate intermediate bases produced by deamination or act directly on substrate bases to generate AP sites. Subsequently, TLS polymerase or DNA ligase repairs these AP sites, randomly replacing them with other bases. Alternatively, AP lyase removes AP sites, causing DSBs and introducing indels [37]. The editing products generated through AP site repair depend primarily on the cellular repair mechanisms and repair-related enzymes utilized. In eukaryotic cells, the tendency of specific bases to be replaced follows the order C ​> ​G ​> ​A ​> ​T. Currently, the purity of the editing products remains relatively low, and the type of base conversion exhibits high randomness that is difficult to control artificially. Nonetheless, the development of this type of base editor expanded the application scope of base editing technologies.

2.1.2.1. C-to-G base editors (CGBEs)

Unlike CBEs, C-to-G base editors (CGBEs) function through the exogenous expression of uracil glycosylase, which excises the uracil produced by cytosine deaminase from cytosine. This excision generates AP sites, which are repaired by DNA polymerases or ligases, resulting in base conversion [38]. In 2021, Kurt et al. fused E. coli homologous UNG (eUNG) with BE4max, excluding UGI, to develop CGBE1. This system produces numerous C-to-G mutations in mammalian cells [39]. Similarly, Zhao et al. used UNG to excise U and initiate the BER pathway repair. By fusing E. coli UNG with APOBEC and nCas9, they created APOBEC-nCas9-Ung, which predominantly generates C-to-G mutations, along with minor C-to-A and C-to-T mutations [40]. Researchers have transformed a UNG variant into a glycosylase variant with C as the substrate. This led to the development of CBEs, such as DAF-CBE and gCBE, that do not require a deaminase. In plants, Zeng et al. applied CGBEs to rice and developed CGBE-rUNG and CGBE-hUNG. CGBEs primarily generate C-to-T mutations in plants, with C-to-G and C-to-A mutations being less frequent [41]. Building upon these developments, another approach emerged through structural analysis. Chen et al. [42] analyzed the structure of the ABE8e complex and its interaction with substrates, strategically mutating key amino acids in TadA8e, including V28, V30, N46, and F84. The authors generated a C-to-G base editor (Td-CGBE) derived from the TadA8e (N46L) variant. This variant was fused with UGI and others, producing a Td-CBE series with editing activity comparable to that of BE4max.

2.1.2.2. Adenine-exchange base editors (AYBEs/AKBEs)

In contrast to ABEs, AYBEs (Y ​= ​T/C, K ​= ​T/G) operate by exogenously expressing purine glycosylase to direct cell repair toward the BER pathway. This mechanism excises the deoxyinosine induced by TadA, generating AP sites and achieving diverse base editing conversions [43]. In 2023, Tong et al. [44] fused the hypoxanthine excision protein N-methylpurine DNA glycosylase (MPG) and ABE8e to develop AYBE, which produces A-to-Y mutations. The authors optimized this protein to obtain AYBEv3, with editing products distributed as follows: A-to-C ​> ​A-to-K ​> ​indels (K ​= ​G/T). This editing preference can be attributed to the involvement of TLS polymerases such as Pol η, which preferentially catalyzes non-Watson-Crick base pairing. When AYBE introduces unnatural base intermediates (inosine derivatives from adenine deamination), Pol η tends to recognize them as purine analogs, leading to increased A-to-C/T transversion frequencies. Tong et al. [44] demonstrated that overexpressing Pol η enhances both the efficiency and purity of A-to-T conversions. Notably, AYBE editing exhibits different characteristics in plants. Upon fusing MPGv3 with ABE, the editor generated A-to-G conversions, with minimal A-to-T and indels. A-to-C mutations were virtually absent [45].

2.1.2.3. Thymine base editors (TBEs, TSBEs)

Based on the uracil-binding pocket and key amino acid groups that interact with the uracil substrate [46], Ye et al. [47] used error-prone PCR and bacterial screening to directionally transform human UNG, obtaining a DNA glycosylase with T as the substrate, i.e., thymine-DNA glycosylase (TDG3). By fusing this enzyme with Cas9n, they constructed deaminase-free TBE, achieving T-to-A base-exchange editing in E. coli and T-to-G base-exchange editing in mammalian cells [47]. Simultaneously, He et al. [48] used a protein language model to guide UNG evolution, creating an enhanced thymine-targeting variant (eTDG). This variant was employed to develop an efficient TSBE (S ​= ​G/C), facilitating T-to-S (G/C) editing.

2.1.2.4. Guanine base editors (GBEs)

Damaged G and A bases in genomic DNA frequently undergo depurination and generate AP sites through DNA glycosylases. This step initiates the BER pathway, leading to mutations [49,50]. Tong et al. [51] evolved MPG into a glycosylase that recognizes G as its substrate. The engineered MPG variant (T199R/S230R/Q294R/D295R) was used to develop GBE, which produces G-to-Y (Y=C/T) base conversions. This advancement completed the final component of the base editing system. Wu et al. [52] developed deaminase-free BEs targeting thymine (pTGBE) and cytosine (pCKBE) by combining engineered human uracil DNA glycosylase variants with nCas9. These editors achieved editing efficiencies of up to 78.05 ​% for T-to-G conversions and up to 61.11 ​% for C-to-G/T conversions in rice, further expanding the base editing toolbox in plants.

In summary, editing products generated from substrate bases modified through mismatch repair are predictable. By contrast, base conversions produced by repaired AP sites remain unpredictable. The primary distinction between AP site-independent and AP site-dependent BEs lies in product purity. AP site-independent BEs demonstrate higher purity and are more suitable for precise single-base editing. By contrast, AP site-dependent BEs produce diverse editing products, facilitating the construction of rich mutant libraries for studying gene function. However, the BER pathway results in the unavoidable production of indel by-products, rendering it unsuitable for precise editing.

2.2. Dual base editors (DBEs)

In addition to identifying gene differences and studying gene functions, researchers often explore the domestication and diversification of genes via mutation. This often requires the generation of multiple base changes within the target site. Single-base editing is constrained by the use of deaminase substrates, limiting the number of base conversions within the target site and impeding the generation of complex amino acid modifications. Dual base editors (DBEs) were developed to address these requirements [[53], [54], [55], [56], [57]]. DBEs can simultaneously modify two or more different base types within the target site, producing multiple base conversions. The design and use of these editors require knowledge about multiple substrate bases and the corresponding repair mechanisms in order to achieve diverse mutations at the target site.

Research on DBEs was first conducted in plants in 2020. Li et al. [57] fused the cytosine deaminase APOBEC3A, 2 ​× ​UGI, and ecTadA-ecTadA7.10 with the CRISPR/Cas9n system, resulting in simultaneous C-to-T/A-to-G base editing at the target site. This type of DBE, termed saturated targeted endogenous mutagenesis editors (STEME), was used to examine the OsACC gene in rice, leading to the identification of key sites affecting herbicide resistance. Hence, DBEs serve as powerful tools for saturated mutagenesis and directed evolution [57]. Additionally, Liang et al. integrated CGBE and ABE to develop a multifunctional dual base editor, i.e., AGBE, capable of producing four types of base conversions: A-G, C-G, C-T, and C-A. This approach greatly increased the diversity of the editing products, making it suitable for constructing saturated mutation libraries [58].

In contrast to typical DBEs based on protein fusion, Neugebauer et al. utilized a phage-assisted continuous evolution system to mutate TadA into the TadA-dual (R26G/V28A/A48R/Y73S/H96 ​N) variant, which simultaneously deaminates cytosine and adenine at its target sites. Fan et al. used this variant for plant gene editing to confirm its efficiency for dual base replacement, developing a compact plant dual editor termed TadDE [34,59]. These DBEs increase the abundance of mutations within a single target site and could be adapted to additional application scenarios.

2.3. Precise deletion genome editors (AFID)

Precise deletion genome editors derived from BEs utilize their functional elements to recognize and act on specific bases within the target site, generating predictable small fragment deletions.

In 2020, Wang et al. established a novel multinucleotide-targeted deletion system termed the APOBEC-Cas9 fusion-induced deletion system (AFID). The authors combined wild-type SpCas9 with the cytosine deaminase APOBEC, UDG, and AP lyase. This system, which is based on cytosine deamination and BER principles, generates multinucleotide deletions from different 5′-cytosines to the Cas9 cleavage site [60]. The researchers enhanced predictable deletion efficiency using a truncated variant of APOBEC3B deaminase (A3Bctd). Subsequently, in 2022, Zeng et al. [41] developed a system based on the selective excision repair pathway by incorporating endonuclease V (EndoV), which facilitates the hydrolysis of 3′-inosine phosphodiester bonds within nucleic acid chains. The authors integrated EndoV into ABE and achieved precise deletions between two as at the target site.

Precise deletion genome editors primarily function through cutting and repairing the target site. These genome editors can be used to analyze various regulatory elements, functional motifs, and non-coding DNA segments within the genome.

3. Optimization strategies for base editors

3.1. Selection based on CRISPR chassis effectors

A better understanding of gene-editing tools has led to the discovery of additional CRISPR effector proteins, some possessing unique properties and essential functions. Small nucleases can be effectively used to reduce the dimensions of BEs when used as chassis proteins.

Cas proteins currently utilized in BEs include Cas9 (SpCas9-NGG, ScCas9-NNG, SaCas9-NNGRRT, CjCas9-NNNNRYAC) [61,62] and Cas12 proteins (Cas12a-TTTV, Cas12b-TTTV, Cpf1-TTTV, Cas12j-TTTV, and Cas12f-TTN) [[63], [64], [65], [66], [67]]. These proteins broaden the range of protospacer-adjacent motif (PAM) types used for base editing and allow the sizes of base editing constructs to be reduced through the use of Cas12 (Fig. 3A). Furthermore, the targeting range can be expanded by mutating key amino acids in the BUC lobe domain, which is responsible for PAM recognition. This approach has been successfully utilized with SpCas9 variants, including SpCas9-NG, SpG, and SpRY [[68], [69], [70], [71], [72]]. However, the broader targeting capabilities of BEs are associated with increased off-target rates and self-targeting effects, requiring careful selection of Cas proteins based on specific experimental objectives [73].

Fig. 3.

Optimization strategies for base editors. A Selection based on clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 chassis effectors. The evolutionary tree in the Fig. comes from: http://caspedia.org/phylogeny_viewer.html. B Mining and enhancement of modifying enzymes (AP, accessory plasmid; MP, mutagenesis plasmid; SP, selection plasmid). C Selection of repair pathways. ∗Indicates that GBE can implement these paths, but with different efficiencies. D Auxiliary component optimization and additional improvement strategies.

3.2. Mining and improvement of base-modifying enzymes

BEs rely on substrate base-specific modifying enzymes that act on ssDNA. These enzymes select and modify bases within the target site through deamination and glycosylation, generating unnatural base intermediates that are recognized and repaired by cellular repair systems. Additionally, these enzymes determine the target base for editing and the editing window, commonly referred to as sequence preference and the editing activity window, respectively [74]. The primary goals in optimizing these enzymes, including deaminases and glycosylases, are to enhance their catalytic activity and alter their substrate base recognition.

Advances in machine learning and graph neural networks have provided novel insights into protein and DNA language models [75,76], enabling data-driven rational protein engineering. The categorization of proteomics data and evolutionary relationships, coupled with the development of protein structure clustering tools, such as Foldseek, has enhanced protein mining and modification [77]. These technologies have helped expand and optimize the BE toolbox, facilitating the discovery of new chassis proteins and improving their efficiency, thereby enhancing the performance of effector proteins (Fig. 3B).

To enhance deaminase performance, Thuronyi et al. used PACE and ancestral reconstruction systems involving rat-derived rAPOBEC and the sea lamprey homologous protein PmCDA1. This approach yielded high-activity mutants, such as evorAPOBEC1, evoFERNY, and evoCDA1, substantially reducing sequence preferences [78]. Gehrke et al. engineered the human cytidine deaminase hAPOBEC3A and developed eA3A, which significantly reduced unintended editing by BE3 while narrowing the editing window [79]. In separate studies, Gaudelli et al. [80] and Richter et al. [81] optimized TadA7.10 through direct evolution and PACE, respectively, producing high-activity adenine deaminase variants with substantially improved ABE efficiency. Yan et al. [82] assessed multiple TadA variants and developed ABE9 by combining mutated amino acids from different variants. Zhang et al. [83] performed deaminase optimization by modifying TadA8e-V106W via V28F transformation. This approach enhanced target editing efficiency while reducing transcriptome off-target effects 52-fold. Furthermore, TadA-derived cytosine deaminases, such as CBE6d, offer superior editing efficiency, narrower windows, and minimal off-target effects compared to APOBEC1.

By contrast, the maturation of E. coli phage-assisted protein evolution systems, including PACE and E. coli Orthogonal Replication frameworks, has led to innovations across various base editing tools [81,84] (Fig. 3B). Therefore, rational protein design serves as a foundation for gene engineering in both animals and plants and for expanding gene-editing toolboxes.

To improve glycosylase performance, Koblan et al. evaluated various uracil-modifying enzymes for CGBE development, revealing the superior performance of UdgX compared with UNG in animal cells [85]. Tong et al. conducted structural analysis and biochemical characterization of the MPG enzyme, achieving directional mutation and screening of DNA-interacting amino acids. The authors produced high-activity MPGv3, which enhanced the editing efficiency of AYBE and product purity [51].

Beyond using artificial mutations to enhance enzyme activity, modifying substrate recognition has emerged as a promising strategy for expanding the application of BEs. Liu and colleagues pioneered this approach by converting the adenine deaminase TadA into the cytosine deaminase TadA-CD [34]. Subsequently, Yang and team utilized mismatch PCR and bacterial screening to alter the substrate bases of glycosylases UNG and MPG, extending their target bases from uracil and hypoxanthine to other pyrimidine (C and T) and purine (G) groups. This led to the development of gCBE, gTBE, and gGBE [51,86].

3.3. Selection of repair pathways

Whereas gene-editing tools require DNA donors, BEs depend primarily on repair pathways to generate the desired editing outcomes. Specific mutations can be achieved by adding relevant proteases to induce the corresponding repair pathways. Among the available BE tools, fundamental tools for interchanging any two bases have attained maturity, meeting researchers’ requirements for base-to-base conversion (Fig. 3C). This advancement represents a major milestone in precision genome editing.

Cytidine deaminase converts C to U, which is readily recognized and excised by UNG, thus inducing the BER pathway and producing indels. The addition of UGI inhibits UNG activity, thereby inducing mismatch repair and facilitating C-to-T conversion [8,46]. Unlike CBEs, inhibiting endogenous alkyl adenine DNA glycosylase (AAG) does not enhance the editing efficiency of ABEs or the purity of the editing products, suggesting that AAG fails to efficiently recognize inosine intermediates. This strategy eliminates the need for additional disruption of inosine damage repair mechanisms [81,87].

Conversely, enhanced MPG activity promotes the recognition of I, a product of A deamination, generating AP sites. However, DNA repair methods vary across species. In maize (Zea mays), fusing an ABE with TLS polymerase promoted AP site repair through translesion synthesis, producing base conversions and reducing the production of indel by-products. Nonetheless, this strategy yielded no substantial improvement when applied in rice [88].

Koblan et al. [85] used a cellular DNA repair-related screening system to identify proteins that enhance CGBE efficiency and purity, including POLD2 and RBMX. Similarly, Chen et al. [89] enhanced C-to-G editing efficiency and purity by incorporating the BER-related proteins rXRCC1 and rPB (the DNA binding domain and DNA lysis domain, respectively, of DNA polymerase β).

3.4. Optimization of auxiliary elements

Beyond substrate base-modifying enzymes, it remains crucial to optimize other key components of editing systems, including nuclear localization signals (NLS) and linkers (Fig. 3D). Modifying the nuclear localization of a protein or increasing copy number can enhance fusion protein expression, thereby improving editing efficiency. Koblan et al. [90] developed BE4max and ABEmax by replacing the SV40 NLS with a 2 ​× ​bipartite NLS, achieving improved editing performance.

GS-rich flexible connectors are commonly used linkers in BEs; their lengths affect editing performance. Komor et al. reported that longer linkers in BE3 allow deaminases to access ssDNA within the R-loop more effectively, resulting in uniform deamination [33]. By contrast, Xie et al. [55] demonstrated that reducing the length of the linker between the adenine deaminase TadA and Cas9n from 32 to 16 amino acids significantly improved the efficiency of the DBE ACBE [55].

3.5. Other universal strategies for improving base editors

Improving BE efficiency using functional elements: Fusion with the bacteriophage Mu Gam protein reduced indel frequency by protecting DSBs caused by AP lyase and inhibiting NHEJ repair, thus improving CBE product purity [33]. Zhang et al. [91] fused multiple human-derived ssDNA binding domains (ssDBD/DBD) with CBEs. Fusion with the RAD51 DBD enhanced CBE editing activity, leading to the development of hyBE4max for efficient C-to-T replacement in animal cells. This technique achieved up to 80 ​% editing efficiency and expanded the editing window from C4–C8 to C4–C12 (indicating the positions of cytosines within the protospacer sequence that can be efficiently edited) [91]. Tan et al. [92] enhanced ABE efficiency in plants by fusing RAD51 DBDs to ABE8e. Furthermore, Zheng et al. [93] investigated the impact of DBD fusion formats on DBE efficiency and developed PhieDBEs, a set of high-efficiency plant DBEs (Fig. 3D).

Reducing off-target effects using split deaminase for safe editing (SAFE): This approach involves splitting the base editor into its N- and C-terminal parts by inserting the deaminase domain into nCas9 to deactivate both the deaminase and nCas9. Ultimately, this technique prevents gRNA-independent off-target effects caused by constitutive deaminase activity (Fig. 3D). At the target site, gRNA functions as a molecular scaffold to reconstruct the split parts into a fully functional base editor, enabling on-target editing. This method reduces both gRNA-dependent off-target editing of DNA and the CBE-induced generation of indel mutations [94]. Similarly, He et al. [48] reduced indel frequency by embedding eTDG between P1249 and E1250 of nCas9.

Co-selection or proxy selection strategy: Wei and colleagues developed HABE and ABEH by connecting the hygromycin resistance gene (HPT) to the termini of the ABE system using a 2A peptide (a polycistronic transgene linker), eliminating the requirement for separate HPT expression. This approach improved co-selection efficiency and the likelihood of base replacement and was later extended to prime editors, demonstrating its universal applicability [95,96] (Fig. 3D). Additionally, Yang and colleagues introduced a proxy selection strategy in 2020. Using this technique, the authors restored an artificially mutated hygromycin resistance gene in the T-DNA region through base editing, enriching efficient editing events on hygromycin-containing medium and improving CBE efficiency [97].

Bai et al. [98] improved chromatin accessibility by co-expressing the human RNA m6A demethylase in gene editing vectors. These improvements enhanced the efficiency of Cas9 and Cas12 gene editing in soybean and prime editors in rice [98].

4. Applications of base editors

Precision base editing can be achieved using both BEs and prime editing tools, each offering complementary advantages for different genome editing applications. Prime editing represents an advanced technology for the precise insertion of base sequences, incorporation of protein tags and functional elements, and accurate base replacement, addition, and deletion [[99], [100], [101], [102]]. However, prime editing in plants faces challenges including low efficiency, complex pegRNA design, and limitations on simultaneous multi-target editing.

Base editors, representing established precision editing tools, offer complementary advantages that address different requirements for plant genome modification. BEs have distinctive operational advantages, including simple target construction, ease of implementation, and the feasibility of achieving multiple base-type changes within individual target sites [103]. We systematically summarize the potential value of CRISPR toolsets for agriculture in Table 1 and present current applications of base editing tools across different crops in Table 2; collectively these demonstrate the widespread use and diverse applications of base editing for creating plant mutant models from an agricultural perspective. Similar to traditional CRISPR systems, sgRNA construction for BEs remains easy, enabling high-throughput multi-target base editing and generating high-abundance base mutations in promoter regions or within genes to precisely engineer plant phenotypes.

Table 1.

The potential value of CRISPR toolsets in agriculture.

Toolset Type	System name	Species	Target gene	Function	Potential applications in agriculture	References
PEs	PPEs	Arabidopsis thaliana, rice (Oryza sativa), wheat (Triticum aestivum), tomato (Solanum Lycopersicum), maize (Zea mays), and potato (Solanum tuberosum)	ALS, OsCDC48	All 12 kinds of bases substitutions can be produced in human cells with accurate insertions of up to 44 bp and deletions of 80 bp.	Precise analysis of gene function. Accelerate the creation of gene-edited germplasm resources.	[[131], [132], [133], [134], [135], [136], [137], [138]]
	PRIME-Del	human cell (Homo sapiens) HEK293T	HPRT1, FMR1	Causes deletions within 10 ​kb of the genome.	Apply large fragment deletion to study Cis-regulatory elements (CREs). Contribute to the study of quantitative trait loci (QTL).	[139]
	PEDAR	mouse (Mus musculus)	Fah	Direct the replacement of a genomic fragment ranging from ∼1 kilobases (kb) to ∼10 ​kb with a desired sequence (up to 60 bp) in the absence of an exogenous DNA template.		[140]
BEs	BE3, CBE, A3A-PBE, NG-CBE, SpRY-CBE, Target-AID	Arabidopsis thaliana, rice (Oryza sativa), wheat (Triticum aestivum), tomato (Solanum Lycopersicum), maize (Zea mays), potato (Solanum tuberosum), soybean (Glycine max Merr.), strawberry (Fragaria vesca), watermelon (Citrullus lanatus), cotton (Gossypium hirsutum), rape(Brassica napus)	OsAAT, OsCDC48, OsDEP1, OsNRT1.1B, OsOD, OsEV, OsHPPD, TaALS, TaMTL, TaLOX2, TaDEP1, TaHPPD, TaVRN1-A1, Ta ACC, ZmALS, ZmCENH3, StALS, StGBSS, GmFT2a, GmFT4, ClALS, GhCLA GhPEBP, BnAlS	Precisely convert cytosine (C) to thymine (T) or Guanine (G) to Adenine (A) in the genome.	Precise tools for single nucleotide polymorphisms (SNPs) studies. Contribute to the creation of excellent germplasm resources without adaptability costs. Efficient research tools for RNA alternative splicing. Effective research tools for upstream open reading frames (uORFs). Protein saturation mutations in plants.	[24,[141], [142], [143], [144], [145], [146], [147], [148], [149], [150], [151], [152], [153], [154], [155], [156], [157]]
	ABE, ABE-P2, ABE-P3, ABE-P4, ABE-P5, ABE-NG, SpRY-ABE, ABE-7, ABE7.10, ABE8e	Arabidopsis thaliana, rice (Oryza sativa), wheat (Triticum aestivum), tomato (Solanum Lycopersicum), maize (Zea mays), strawberry (Fragaria vesca), rape (Brassica napus)	OsSPL14, OsSPL16, OsSPL17, OsSPL18, OsACC, OsALS, OsCDC48, OsAAT, OsEV, OsOD, OsDEP1, OsNRT1.1B, TaDEP1, TaGW2 BnALS, BnPDS	Precisely convert Adenine (A) to Guanine (G) or thymine (T) to cytosine (C) in the genome.		[22,147,[158], [159], [160], [161], [162], [163], [164], [165], [166]]
	CGBE	human cells (Homo sapiens), mouse (Mus musculus), E. coli, rice (Oryza sativa), tomato (Solanum Lycopersicum), poplar (Populus tremula)	Tyr, RP11-177B4, PSMB2, EMX1, OsEPSPS, OsALS, OsCGRS55, SAgo7, PtPDS1	Guanine (G) and cytosine (C) reverse each other		[39,40,85,89,[167], [168], [169], [170]]
	STEME-1, STEME-NG,	rice (Oryza sativa)	ACC	It can introduce two mutations (C to T and A to G) simultaneously.		[171]
	AFID	rice (Oryza sativa)	CDC48, SPL14, SWEET14, MiR396, GASR6	It can generate predictable base deletions across the genome.		[60]

Table 2.

Applications of BEs.

Species	Editors	Target gene	Editing efficiency (%)	Delivery methods	References
Rice (Oryza sativa L.)	BE3-ΔUGI	OsNRT1.1B, OsSLR1	2.7–13.3	Agrobacterium	[146]
Rice (Oryza sativa L.)	BE3 (PBE/CBE-P1)	OsCDC48, OsNRT1.1B, OsSPL14, OsSLR1, OsPDS, OsSBEIIb, OsSNB	2.7–43.48	Agrobacterium	[144,146,147,155]
Rice (Oryza sativa L.)	Targeted-AID	OsALS, OsWaxy, OsFTIP1e	4–90	Agrobacterium	[24,148,152]
Rice (Oryza sativa L.)	CBE-P3 (VQR-BE3)	PMS3	61.1	Agrobacterium	[159]
Rice (Oryza sativa L.)	A3A-PBE	OsAAT, OsCDC48, OsDEP1, OsNRT1.1B, OsOD, OsEV, OsHPPD	44.1–82.9	Agrobacterium	[156]
Rice (Oryza sativa L.)	rBE9 (hAID∗Δ-XTEN-Cas9n-NLS)	OsAOS1, OsJAR1, OsJAR2, OsCOI2	11.8–69.4	Agrobacterium	[172]
Rice (Oryza sativa L.)	xBE3/xCas9(D10A)-PmCDA1	OsGS3, OsDEP1, LF, IAA13, SPL7, SPL4, MADS57	3.2–89.7	Agrobacterium	[143,157]
Rice (Oryza sativa L.)	CBE-NG/rBE22/nSpCas9-NGv1-AID/Cas9-NG(D10A)-PmCDA1-UGI	OsDEP1, OsEPSPS, OsPDS, OsCERK1, GSK4, ETR2, MPK11, MPK7, MPK10, MPK8, SERK1, SERK2, RLCK185, BZR1, Os03g020040, LF, IAA13, SPL7, SNB, PMS3	4.5–72.2	Agrobacterium	[142,143,147,157]
Rice (Oryza sativa L.)	ABE7.10	OsSPL14, SLR1, OsSPL16, OsSPL18, LOC_Os02g24720, OsMPK6, OsMPK13, OsSERK2, OsWRKY45	4.3–62.26	Agrobacterium	[158,166]
Rice (Oryza sativa L.)	PABE-7	OsACC, OsALS, OsCDC48, OsAAT, OsEV, OsOD, OsDEP1, OsNRT1.1B	3.2–59.1	Agrobacterium	[161]
Rice (Oryza sativa L.)	ABE-P2 (ABEsa)	OsSPL14, OsSPL17	26–61.3	Agrobacterium	[158]
Rice (Oryza sativa L.)	ABE-P3 (VQR-ABE)	OsSPL14, OsSPL16, OsSPL17, OsSPL18	30–74.3	Agrobacterium	[159]
Rice (Oryza sativa L.)	ABE-P4 (VRER-ABE)	OsTOE1, OsIDS1	2.6	Agrobacterium	[159]
Rice (Oryza sativa L.)	ABE-P5 (SaKKH-ABE)	SNB	6.5	Agrobacterium	[159]
Rice (Oryza sativa L.)	ABE-NG	OsSPL14, LF1, OsIAA13, OsSPL7	2.9–11.9	Agrobacterium	[143]
Rice (Oryza sativa L.)	ABE-NG-S	OsSPL14, LF1, OsIAA13, OsSPL7	2.0–7.7	Agrobacterium	[143]
Rice (Oryza sativa L.)	APOBEC1-XTEN-Cas9(D10A)	NRT1.1B, SLR1	2.7–13.3	Agrobacterium	[146]
Rice (Oryza sativa L.)	ABE	ALS	21.4	Agrobacterium	[161]
Rice (Oryza sativa L.)	ABE	OsTubA2	12.7	Agrobacterium	[173]
Rice (Oryza sativa L.)	pH-nCas9-PBE	OsCDC48	43.48	Agrobacterium	[155]
Rice (Oryza sativa L.)	CBE	NRT1.1B, SLR1	2.7–13.3	Agrobacterium	[146]
Rice (Oryza sativa L.)	CBE	PDS, SBEIIb	20	Agrobacterium	[144]
Rice (Oryza sativa L.)	CBE	CDC48	43.48	Agrobacterium	[155]
Rice (Oryza sativa L.)	CBE	ALS, FTIP1e	85.7	Agrobacterium	[24]
Rice (Oryza sativa L.)	CBE	CDC48, NRT1.1B	6.5–82.9	Agrobacterium	[156]
Rice (Oryza sativa L.)	CBE	ALS	19.2–50	Agrobacterium	[174]
Rice (Oryza sativa L.)	CBE	IPA1-DD		Agrobacterium	[175]
Rice (Oryza sativa L.)	ABE	ACC, ALS, CDC48, DEP1	6.9–46.5	Agrobacterium	[161]
Rice (Oryza sativa L.)	ABE	SPL14, SPL16, SLR1	26–61.3	Agrobacterium	[158]
Rice (Oryza sativa L.)	ABE	ZEBRA3, WSL5	1.34–38.92	Agrobacterium	[163]
Rice (Oryza sativa L.)	CBE, ABE	ALS	10	Agrobacterium/particle-bombardment	[160]
Rice (Oryza sativa L.)	CBE, ABE	ACC	93.8	Agrobacterium	[162]
Rice (Oryza sativa L.)	NG-CBE	SNB, SPL7, PMS3	4.5–72.2	Agrobacterium	[143]
Rice (Oryza sativa L.)	NG-ABE	SPL14, LF1, IAA13, SPL7	2.9–11.9	Agrobacterium	[143]
Rice (Oryza sativa L.)	NG-CBE	BZR1, SERK2	2.06–54.16	Agrobacterium	[147]
Rice (Oryza sativa L.)	NG-ABE	SERK2	93.33	Agrobacterium	[147]
Rice (Oryza sativa L.)	NG-CBE	DEP1	38.5–56.3	Agrobacterium	[157]
Rice (Oryza sativa L.)	x-CBE	SBEIIb	6.7–21.1	Agrobacterium
Rice (Oryza sativa L.)	NG-CBE	Waxy, EUI1, CKX2	5.6–50	Agrobacterium	[154]
Rice (Oryza sativa L.)	NG-ABE	CKX2	6.5–55.6	Agrobacterium	[154]
Rice (Oryza sativa L.)	SpRY-CBE	ACC1, IPA1, TAC1	8.3–52.1	Agrobacterium	[145]
Rice (Oryza sativa L.)	SpRY-ABE	ACC1, IPA1, SPL3	6.3–47.9	Agrobacterium	[145]
Rice (Oryza sativa L.)	SpRY-CBE	DEP1, ALS	10.0–72.2	Agrobacterium	[176]
Rice (Oryza sativa L.)	SpRY-ABE	PDS	79	Agrobacterium	[176]
Rice (Oryza sativa L.)	SpRY-CBE	CPK1, CPK4, MPK3	2.13–74.19	Agrobacterium	[72]
Rice (Oryza sativa L.)	SpRY-ABE	MPK13, GS1, GSK4	29.79–93.75	Agrobacterium	[72]
Rice (Oryza sativa L.)	SpRY-CBE	SERK2, MPK5, ALS	7.5–55.2	Agrobacterium	[177]
Rice (Oryza sativa L.)	SpRY-ABE	MPK2, NRT1.1B, Waxy	4.8–46.7	Agrobacterium	[177]
Rice (Oryza sativa L.)	AFID	CDC48, SPL14, SWEET14	50.0–93.6	PEG	[60]
Rice (Oryza sativa L.)	AFID	MiR396, GASR6	0.8-3.5	PEG	[60]
Rice (Oryza sativa L.)	STEME-1, STEME-NG	OSACC	15.1	Agrobacterium	[171]
Rice (Oryza sativa L.)	STCBE-2	OsEPSPS	25.3	Agrobacterium	[125]
Wheat (Triticum aestivum L.)	PBE	TaLOX2	0.39-7.07	Agrobacterium	[155]
Wheat (Triticum aestivum L.)	A3A-PBE	TaALS, Ta MTL, TaLOX2, TaDEP1, TaHPPD, TaVRN1-A1, TaACC	16.7–22.5	Agrobacterium	[156]
Wheat (Triticum aestivum L.)	PABE-7	TaDEP1, TaEPSPS, TaGW2	0.4-1.1	Agrobacterium	[161]
Wheat (Triticum aestivum L.)	CBE	LOX2	0.39-7.07	Agrobacterium	[155]
Wheat (Triticum aestivum L.)	CBE	ALS, ACCase	22	Agrobacterium	[178]
Wheat (Triticum aestivum L.)	CBE	ALS, MTL	16.7–22.5	Agrobacterium	[156]
Wheat (Triticum aestivum L.)	ABE	DEP1, GW2	0.4-1.1	Agrobacterium	[161]
Maize (Zea mays L.)	CBE	ALS	13.8	Agrobacterium	[179]
Maize (Zea mays L.)	pBnCas9-PBE	CATD	10.1	Agrobacterium	[155]
Maize (Zea mays L.)	PBE	ZmCENH3	0.39-4.03	Agrobacterium	[155]
Maize (Zea mays L.)	CBE	CENH3		Agrobacterium	[155]
Arabidopsis thaliana	ABE7.10	AtALS, AtPDS, AtFT, AtLFY	4.1	Agrobacterium	[22]
Arabidopsis thaliana	CRISPR/Cas9	uORF of AtBRI1, AtVTC2		Agrobacterium	[180]
Arabidopsis thaliana	CBE	eIF4E	50	Floral Dip Agro-transformation	[181]
Arabidopsis thaliana	ABE	PDS3, FT	84	Agrobacterium	[22]
Arabidopsis thaliana	CBE	ALS	1.7	Agrobacterium	[182]
Arabidopsis thaliana	BE3	AtALS, AtHAB1, AtT30G6.16, AtRS31A, AtAct2		Agrobacterium	[182,183]
Potato (Solanum lycopersicum.)	NG-CBE	ALS1	32	PEG	[150]
Potato (Solanum tuberosum.)	NG-CBE	GBSSI, DMR6	9–64	PEG	[150]
Potato (Solanum lycopersicum.)	CBE	ALS1		PEG	[150]
Potato (Solanum tuberosum.)	CBE	GBSSI, DMR6	10	PEG	[150]
Potato (Solanum tuberosum.)	CBE	DELLA, ETR1	26.2–53.8	Agrobacterium	[24]
Potato (Solanum tuberosum.)	CBE	GBSS	6.5	Agrobacterium	[156]
Potato (Solanum tuberosum)	A3A-PBE	StALS, StGBSS	6.5	Agrobacterium	[156]
Potato (Solanum tuberosum)	BE3	GBSS		Agrobacterium	[155]
Potato (Solanum tuberosum)	Target-AID	StALS	37	Agrobacterium	[25]
Tomato (Lycopersicon esculentum)	Target-AID	DELLA, ETR1, SlALS	71	Agrobacterium	[24,25]
Soybean (Glycine max Merr.)	CBE	GmFT2a, GmFT4	6.0–18.2	Agrobacterium	[141]
Strawberry (Fragaria vesca)	A3A-PBE	uORF of FvebZIPs1.1	100	Agrobacterium	[153]
Strawberry (Fragaria vesca)	CBE	ebZIPs1.1		Agrobacterium	[153]
Watermelon (Citrullus lanatus L.)	BE3	ALS	23	Agrobacterium	[149]
Cotton (Gossypium hirsutum L.)	BE3	GhCLA, GhPEBP	26.67-57.78	Agrobacterium	[184]
Rape (Brassica campestris L.)	ABE7.10	BnALS, BnPDS	8.8	Agrobacterium	[22]
Rape (Brassica napus)	CBE	ALS	1.8	Agrobacterium	[151]

4.1. Different applications based on editing window width

The width of the editing window significantly influences the application scenarios for BEs. In addition to linker length, various types of deaminases (such as CBE6 and TadDE) and auxiliary elements (such as DBD) affect the width of the editing window [93]. Narrow-window BEs, targeting three to five bases, are suitable for introduction of stop codons, replacement of key amino acids, genetic disease repair in animals, and in situ protein modification in crops [104]. Furthermore, narrow-window BEs are used in plant breeding to create precise genetic improvements that enhance both crop quality and yield [105,106].

When selecting appropriate editors based on specific requirements, the need for precise site-specific base substitutions with few by-products favors non-AP pathway-dependent single BEs. Narrow windows allow precise mutations to be achieved, specifically for the site-specific introduction of stop codons and the alteration of amino acids or splice sites [107]. This strategy minimizes changes to gene structure and reduces the likelihood of producing unknown polypeptides compared with knockout systems.

By contrast, wide-window BEs (such as hyBE4max and ABE8e-RAD51) target editing windows spanning 8–15 bases, making them suitable for editing multiple candidate sites or generating diverse mutations. When the need for editing purity is low, highly efficient AP pathway-dependent BEs or multiple BEs might be used. This is because AP pathway-dependent base mutations randomly generate diverse mutants. These wider editing windows are beneficial for constructing saturated mutation libraries, studying gene or promoter function, performing protein domain mutagenesis, or simultaneously editing multiple base-dense regions. Such editors efficiently disrupt functional sequences and enable extensive editing in complex genomes, enhancing editing efficiency [78,108,109] (Fig. 4A). For plant genome engineering, these editors facilitate the creation of complex traits, including improved disease resistance and metabolic characteristics [105].

Fig. 4.

Comparison of Base Editing Applications. A Structural comparison and implementation scenarios of narrow-window (3–5 bp) versus wide-window (8–15 bp) base editors. B Base editing in plant protein evolution, exemplified through the saturated editing of the EPSPS gene. C Fundamental principles of RNA base editing systems, illustrating dCas13-ADAR-mediated A to I conversion. D Comparative analysis of plant and animal base editing systems, encompassing effector protein structures, apurinic/apyrimidinic repair preferences, and distinct delivery methodologies.

4.2. Application of base editing for plant protein evolution

Base editing tools provide a robust research platform for the directed evolution of proteins in plants. Compared with traditional random mutation or in vitro evolution methods, base editing can be used for in situ protein structure modification and functional screening within plants, significantly enhancing efficiency (Fig. 4A). Wang et al. [110] utilized the diverse mutation characteristics of BEs to establish an efficient lineage-specific evolution system in Arabidopsis, successfully identifying variants of the genes EPSPS, ALS, and HPPD conferring herbicide resistance. This approach facilitated gene evolution and efficient screening (Fig. 4B). Gao and colleagues used structural alignment to identify novel deaminases from natural sources; they incorporated near-saturated mutations in the OsEPSPS gene and identified new alleles. The newly developed Sdd7-CBE demonstrated substantial editing efficiency in soybean [23].

In animals, researchers established the first Continuous Directed Evolution system in mammalian cells, addressing limitations om gene delivery and adaptability issues in prokaryotic systems (such as phage-based delivery) [111]. In prokaryotes, the efficient CRISPR-Cas9-based base editing system CRISPR-CDA-nCas9-UGI was developed for multi-target editing of the SecY, SecE, and SecG genes, which was successfully screened mutants with a 3.6-fold improved transport efficiency (such as SecE V36I/SecG A62T-V63I). By constructing a mutant library using an sgRNA pool, mutant strains with significantly enhanced antimicrobial peptide resistance (such as B9) and key active sites (such as G552 and T624) were identified. This technique serves as an efficient in vivo evolution tool for difficult-to-express bacterial endotoxin proteins or membrane proteins [112]. These studies demonstrate the substantial potential of base editing technology for protein engineering, establishing its efficacy for both directed evolution and in situ protein modification in animals and plants.

4.3. Other applications of BEs

BEs do not rely on DSB repair mechanisms, substantially reducing the risk of genome instability and providing distinct safety advantages. For the precise regulation of gene expression levels, BEs enable the fine regulation of gene expression through modifications to the promoter, 5′ untranslated region (5′ UTR), intron, and other non-coding regions of a gene [113] (Fig. 4A). By modifying transcription start sites or core promoter elements, researchers can attain quantitative reductions rather than the complete knockout of gene expression. This approach is essential for studying gene dosage effects and essential gene functions. For example, Wang et al. [114] targeted non-coding regions (such as promoters and introns) of the cotton GhTFL1 gene, generating 300 allelic variant lines. This technique affected mRNA splicing efficiency and yielded early-flowering phenotypes. Kang et al. [22] modified the splice site of the PDS gene in rapeseed using ABE, modifying mRNA splicing patterns and providing novel tools for studying the regulation of gene expression.

RNA base editing tools have several major applications. RtABE is a novel RNA editing system that enhances safety by inhibiting base editing activity before RtABE binds to its target RNA [115]. Similarly, Cas13-based RNA base editing tools expand the scope of base editing by utilizing adenosine deaminase acting on RNA type 2 to achieve programmable A-to-I conversion in target RNA transcripts [116] (Fig. 4C). Unlike permanent DNA editing, RNA base editing has temporary effects, enabling the reversible regulation of gene expression, which is particularly valuable for studying lethal genes and the functions of development-related genes. Miniaturized RNA base editing technology, such as that using Cas13j as the chassis protein, holds promise [117]. When combined with tissue-specific promoters, these systems can achieve post-transcriptional control of gene function in specific tissues, providing effective tools for studying tissue-specific differences in gene function.

A prominent advantage of base editing is the simple sgRNA design, which makes it suitable for high-throughput research. For example, TadDE, a single deaminase domain that simultaneously achieves C-to-T and A-to-G edits, is suitable for multi-target high-abundance base editing [59]. Additionally, the high-abundance mutations generated by BEs facilitate the development of numerous single-gene mutants. Researchers used AGBE to create mutations in the human diphtheria toxin receptor (hDTR) gene and successfully identified variants that confer resistance to this disease. Through deep sequencing, the authors identified 59,269 variants with different sensitivities to diphtheria toxins using 20 hDTR-targeting sgRNAs [58].

4.4. Comparison of base editing applications in animals and plants

The applications of base editing technology differ markedly between animals and plants due to distinct delivery constraints and inheritance patterns. Plant genome editing relies on Agrobacterium-mediated stable transformation, where editing effects are stable and heritable across generations. However, the large cargo capacity (15–20 ​kb) of Agrobacterium accommodates full-length base editing systems without size restrictions [20]. By contrast, animal systems utilize various delivery methods such as adeno-associated viruses (AAV) and liposomes for in vivo gene therapy applications [118,119]. While AAV-delivered base editing can theoretically create heritable edits, this remains highly controversial due to ethical concerns around germline modification. The preference for smaller effector proteins (such as SaCas9 and Cas12f) in animal systems stems from AAV's strict packaging limitation (∼4.7 ​kb) [120], whereas Agrobacterium-mediated delivery in plants is less constrained by cargo size, enabling the delivery of the complete editing machinery with multiple regulatory elements.

In plants, base editing is primarily utilized for crop improvement and gene function research to develop beneficial traits, such as disease resistance, stress tolerance, and increased yield. In animals, base editing primarily focuses on disease model development and treatment, including the correction of base mutations that cause genetic disorders, such as sickle cell anemia. Based on current reports regarding random base substitutions generated through translesion repair at AP sites in animals and plants, the trend in animal cells is C ​> ​G ​> ​A ​> ​T. For example, animal CGBE produces large amounts of C-to-G and small amounts of C-to-A and C-to-T substitutions, while plant CGBE mainly produces C-to-T, with small amounts of C-to-G and C-to-A substitutions [47,86]. The proportions of editing products of animal AYBEs are in the order A-to-C ​> ​A-to-K ​> ​indels (K ​= ​G/T), while plant AYBEs produce more A-to-G, some A-to-T and indels, and almost no A-to-C. This phenomenon may be caused by differences in intracellular repair mechanisms. Furthermore, in animals, when using the TadA system, the original ecTadA and its evolved variant ecTadA∗ are both required and must dimerize in order function effectively, whereas in plants, ecTadA alone can perform efficient gene editing. In animals, BEs are readily operational and can be delivered via AAV to repair pathogenic base mutations by replacing smaller effectors [44,118,121] (Fig. 4D).

5. Conclusion and future prospects

Base editing technology has been widely utilized for functional genomics in plants because of its high efficiency and precision. For instance, researchers have used CBEs and ABEs to develop Arabidopsis and rapeseed mutants with early flowering and disease resistance, shedding light on plant development and stress-resistance traits [22]. Furthermore, the successful use of ABEs to enhance drought resistance in rice illustrates the potential of base editing technology for plant breeding [23]. Beyond functional research and trait improvement, base editing has led to the development of herbicide-resistant crops. For example, researchers successfully obtained superior rice and wheat varieties resistant to glyphosate and sulfonylurea by editing key genes using ABEs [24,25]. These applications demonstrate the essential roles of base editing in mutant library construction and trait enhancement in plants, providing efficient approaches for molecular breeding.

The evolution of base editing technology has led to the development of novel editing tools, further expanding plant genome editing. The development of DBEs, such as the STEME editor, significantly improved the efficiency of screening for mutants of the rice ACC gene [57]. This tool enables the precise, simultaneous editing of multiple bases, advancing plant genomics and efficient breeding. Moreover, researchers have developed broad-targeting BEs for various crops by optimizing the dependence of Cas proteins on PAMs and editing windows. These techniques reduce off-target effects while enhancing editing efficiency and are gradually being utilized for economically significant crops [62,68]. Consequently, the key criteria for selecting the appropriate BEs include high efficiency, a broad editing window, a good targeting range, and high product purity, particularly when constructing mutant libraries, looking for key SNP sites, or conducting directed evolution of target genes.

However, editing efficiency is significantly influenced by chromatin accessibility at the target loci, as chromatin structure can limit the access of Cas-based systems to their genomic targets [122,123]. Open chromatin regions exhibit higher editing efficiencies, suggesting that future optimization should consider both improving editing tools and enhancing target site accessibility. For example, Wang, He [124] used ABE8e, evoCDA1, and evoAPOBEC1 to perform saturated mutagenesis on the CT domain of the rice ACC gene and identified SNPs linked to herbicide resistance. DBE development improved and accelerated this process. Zhang et al. [125] used STCBE to edit 35 sites in the rice EPSPS gene and identified a novel glyphosate-resistance allele (OsEPSPS-D213N; Fig. 4B). DBEs both improve screening efficiency and expand the application potential of base editing in plant research.

Much research has focused on continuous advancements in base editing. The directed transformation of the substrate specificity of base-modifying enzymes is crucial for enhancing the diversity of editing tools. For instance, using TadA dual variants, researchers modified the recognition pockets of MPG/AAG enzymes, developing variants capable of simultaneous glycosylation of multiple bases, thereby enabling nearly saturated base editing. Additionally, following the use of UGI, researchers identified novel proteases to inhibit the BER pathway and reduce indel formation. Recent studies have further expanded the base editing toolkit: Contiliani et al. [126] identified highly efficient cytidine deaminases from animal sources for robust multiplexed editing; Hu et al. [127] developed CyDENT, a CRISPR-free strand-selective editor for organellar genomes; and Jiang et al. [128] engineered improved C-to-G editors providing up to 52 ​% pure editing efficiency across multiple plant species. Such BEs improve high-efficiency editing, substrate selection, and the ability to edit broad targets, showing promise as the future standards [34,42,129,130]. Such improvements require highly efficient editing domains and the identification of active regions and specific sites, along with broad-targeting recognition ranges beyond PAMs. With advances in plant transformation methods, the application of BEs should extend to previously challenging crops such as sugarcane, which are difficult to transform using conventional approaches.

CRediT authorship contribution statement

Ruixiang Zhang: Writing – review & editing, Writing – original draft, Investigation, Formal analysis, Conceptualization. Zhiye Zheng: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Guangzhou Li: Writing – review & editing, Writing – original draft. Xinglei Zheng: Writing – original draft, Data curation, Conceptualization. Liying Su: Data curation, Conceptualization. Xudong Yuan: Writing – review & editing, Formal analysis, Data curation. Tie Li: Writing – review & editing. Jiantao Tan: Writing – review & editing, Writing – original draft, Resources. Dongchang Zeng: Writing – review & editing, Writing – original draft, Supervision. Shaocun Zhang: Writing – review & editing, Writing – original draft, Methodology. Jialin Liu: Writing – original draft. Haochun Shen: Writing – original draft. Nan Chai: Writing – review & editing, Writing – original draft, Resources, Project administration, Methodology. Yao-Guang Liu: Validation, Supervision, Resources, Project administration, Methodology, Investigation, Conceptualization. Qinlong Zhu: Writing – review & editing, Writing – original draft, Resources, Project administration, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no conflict of interest.

Data availability

Data sharing is not applicable to this article as no datasets were generated or analyzed as part of this study.