Synergistic engineering of Casδ nuclease for robust genome editing

ABSTRACT

Casδ is a recently identified evolutionary transitional CRISPR system characterized by its compact size (~900 amino acids), broad temperature tolerance, and guidance by a short crRNA without the requirement of a tracrRNA. However, the low editing efficiency of Casδ in eukaryotic cells limits its application. Here, we have developed a hierarchical engineering strategy to improve the genome editing activity of Casδ‐1, with optimization focused on enhancing its interactions with the crRNA, the protospacer adjacent motif (PAM) duplex, the single‐stranded DNA substrate, and the RNA–DNA heteroduplex. Through this strategy, we successfully generated an activity‐enhanced Casδ‐1 variant, designated enCasδ, which harbors 9 amino acid substitutions that synergistically augment its editing efficiency. In human cell lines, enCasδ showed 1.3‐ to 29.3‐fold higher editing activity than the wild‐type Casδ‐1 across ten tested genomic loci, with an average editing efficiency of 54.6%. In addition, enCasδ also mediated robust genome editing in maize; its editing efficiency increased by an average of 5.3‐fold relative to Casδ‐1, and reached up to an average of 80% at the TS4 and PSY1 loci in stable transgenic lines. The overall editing performance of enCasδ was comparable to that of Streptococcus pyogenes Cas9 (SpCas9) and other Cas12 nucleases. Collectively, enCasδ represents a highly optimized Casδ‐1 variant that broadens the applicability of the Casδ CRISPR system and facilitates robust genome editing in both animal cells and plants.

A hierarchical engineering strategy was used to enhance the editing activity of Casδ, yielding an optimized variant termed enCasδ. This variant enables robust genome editing in animal cells and plants, with overall editing performance comparable to that of SpCas9 and other Cas12 nucleases.

INTRODUCTION

The Clustered regularly interspaced short palindromic repeats‐Cas (CRISPR‐Cas) systems are adaptive immune systems widely distributed in prokaryotes (Jansen et al., 2002) that provide valuable resources for the development of a variety of versatile gene editing tools in eukaryotes (Li et al., 2013; Shan et al., 2013; Wang et al., 2013; Niu et al., 2014; Xing et al., 2014). Beyond the well‐established type II Cas9 system, an increasing number of previously unexplored Cas systems have been discovered, which has enabled effective expansion of the gene editing toolkit and has contributed significantly to basic biological research, gene therapy, and breeding of plants and animals (Zetsche et al., 2015; Harrington et al., 2018; Strecker et al., 2019; Pausch et al., 2020; Altae‐Tran et al., 2021; Özcan et al., 2021; Xu et al., 2021; McGaw et al., 2022; Wang et al., 2023; Yan et al., 2019). However, the natural gene editing activity of most newly discovered CRISPR‐Cas systems is poor in animals and plants (Liu et al., 2019; Pausch et al., 2020). Optimization of their activity, particularly through rational protein engineering, has emerged as a highly important and powerful strategy to address this limitation (Kleinstiver et al., 2019; Xu et al., 2021; Zhu et al., 2021; Chen et al., 2022; McGaw et al., 2022; Han et al., 2023; Duan et al., 2024a; 2024b; Bai et al., 2025; Li et al., 2025; Wang et al., 2025a).

Class 2 systems, usually characterized by a single effector protein, are particularly suitable for genetic engineering applications, including type II Cas9 and type V Cas12 systems (Jiang et al., 2013; Makarova et al., 2019). Due to the high diversity of Cas12 proteins, a lot of the type V systems were identified after the first reported protein, Cpf1 (Cas12a) (Zetsche et al., 2015), in this family (Yan et al., 2019; Pausch et al., 2020; Al‐Shayeb et al., 2022; Duan et al., 2024b). Just recently, a new type V CRISPR‐Cas effector superfamily, designated Casδ, was identified (Yang et al., 2025). Casδ has a relatively smaller size (867–936 aa) than Cas9 and Cas12a, and has a 5′‐RYR‐3′ protospacer adjacent motif (PAM) preference (Yang et al., 2025). In addition, Casδ requires only a single mature crRNA for DNA targeting and displays a wide temperature adaptability for its activity (Yang et al., 2025). These characteristics endow Casδ with excellent application potential as a valuable gene editing tool in both plants and animals. However, the gene editing activity of natural Casδ is relatively low, which consequently limits its broad application.

Here, we developed a hierarchical engineering strategy to improve the genome editing activity of Casδ‐1, a Casδ nuclease with editing activity in both animals and plants (Yang et al., 2025), by enhancing its interactions with the crRNA, the PAM duplex, the single‐stranded DNA substrate, and the RNA–DNA heteroduplex. Through this strategy, we successfully generated an activity‐enhanced Casδ‐1 variant, designated enCasδ, which harbors 9 amino acid substitutions that synergistically augment its editing function. Compared to natural Casδ‐1, enCasδ showed 1.3‐ to 29.3‐fold enhancement of editing activity across ten tested genomic loci, with an average editing efficiency of 54.6%. In addition, enCasδ also mediated robust genome editing in maize; its editing efficiency increased by an average of 5.3‐fold relative to Casδ‐1, and reached up to an average of 80% at the TS4 and PSY1 loci in stable transgenic lines. The overall editing performance of enCasδ was comparable to that of Streptococcus pyogenes (SpCas9), LbCas12a, and a recently engineered Cas12i3 nuclease (Cas‐SF01) (Jiang et al., 2013; Zetsche et al., 2015; Duan et al., 2024a). Collectively, enCasδ represents a highly optimized Casδ‐1 variant that could enable wider application of the Casδ CRISPR system for robust genome editing in animals and plants.

RESULTS

Identifying improved Casδ‐1 variants with single amino acid substitutions in human cells

We used a hierarchical engineering strategy to improve the editing activity of Casδ‐1 (Figure 1A). First, we attempted to identify single amino acid‐substituted Casδ‐1 variants to enhance the interaction of Casδ‐1 with the PAM duplex (EIP), the interaction of Casδ‐1 with crRNA (EIR), the interaction of Casδ‐1 with the RNA–DNA heteroduplex (EIH), and the interaction of Casδ‐1 with the ssDNA substrate (EIS), and thus to facilitate both target DNA recognition and hydrolysis. Then, a combination assay was performed for the amino acid substitutions associated with EIP, EIR, EIH, and EIS to further improve the editing activity of Casδ‐1. Finally, we aggregated the effects of EIP, EIR, EIH, and EIS to obtain an optimal Casδ‐1 variant with improved editing activity.

Figure 1.

Identification of single amino acid substitutions that improve Casδ‐1 editing activity

(A) Schematic representation of the hierarchical engineering strategy used for improvement of Casδ‐1 gene editing activity. AlphaFold3‐based prediction of the ternary complex structure of Casδ‐1 is shown. EIP: Enhanced interaction with the PAM duplex. EIR: Enhanced interaction with crRNA. EIH: Enhanced interaction with the RNA–DNA heteroduplex. EIS: Enhanced interaction with the ssDNA substrate. (B) Editing activity of Casδ‐1 variants with single amino acid substitutions. Pink and blue dots represent the results obtained in HEK293T cells based on the targets of B2M‐sg1 and CⅡTA‐sg1, respectively. The relative editing efficiencies shown refer to the efficiency of a particular Casδ variant compared with that of the wild‐type Casδ‐1 at the same target. (C) Presentation of the 24 Casδ‐1 variants with editing efficiency more than 1.2‐fold that of Casδ‐1. Error bars represent SD of means, n = 2 or 3 biological replicates.

Structure‐guided substitutions of amino acid residues with the positively charged arginine (R) have been demonstrated to enhance the editing efficiency of CRISPR‐Cas proteins in both mammalian cells and plants (Xu et al., 2021; Zhu et al., 2021; Chen et al., 2022; Duan et al., 2024a; Bai et al., 2025; Wang et al., 2025a). According to the predicted protein structure of Casδ‐1 by AlphaFold3 (Abramson et al., 2024), we roughly identified a total of 200 potential non‐arginine amino acid residues that interact with or surround the PAM duplex, crRNA, the RNA–DNA heteroduplex, and the ssDNA substrate (Figure S1A). We thus further mutated these non‐arginine amino acid residues into arginine to construct Casδ‐1 variants. Besides mutating into an arginine, there were 7 potential amino acids interacting with the PAM duplex that were also substituted with residues bearing larger, flat aromatic rings, including phenylalanine (F), tryptophan (W), and/or tyrosine (Y) (Figure S1B), which have been shown to enhance the stacking interaction with the PAM duplex (Chen et al., 2022; Spasskaya et al., 2023).

To screen for single substitutions of Casδ‐1 with enhanced editing activity, we cloned each variant into a protein expression vector. In this vector, the engineered Casδ‐1 protein was fused with NLS (nuclear localization signal) at both the N‐terminus and the C‐terminus, and was connected with a puromycin‐resistant gene via T2A for the selection of positive cells (Figure S1C). The construct was then co‐transfected with the crRNA expression vector into Human Embryonic Kidney 293T (HEK293T) cells. The editing efficiency was determined by combining the polymerase chain reaction (PCR) and next‐generation sequencing (NGS). Based on B2M and CⅡTA targets, where the editing activities were generally positively correlated among different variants (Figure S2), most variants showed reduced activity compared to the wild type (Figure 1B). Even so, 24 variants showed more than a 20% increase in editing activity, including 3, 3, 6, and 12 variants associated with EIP, EIR, EIH, and EIS, respectively (Figure 1B, C).

Generating candidates for the optimization of EIP, EIR, EIH, and EIS through combinations of single amino acid substitutions

Next, we attempted to determine whether the editing activity of Casδ‐1 can be further enhanced through efficiency optimization of EIP, EIR, EIH, and EIS by combining the aforementioned single amino acid substitutions. Three target sites at the B2M, CⅡTA, and TRAC genes of human were used for genome editing activity evaluation with HEK293T cells. For EIP efficiency optimization, we selected the two variants from the initial screen, including D120F and K129R, both of which are located in the PAM interacting (PI) domain of Casδ‐1 (Figures 1B, S3A). Subsequently, we combined amino acid alterations of D120F and K129R into one Casδ‐1 variant and observed further enhanced editing activity based on the evaluation in HEK293T cells. On average, the D120F/K129R variant showed a 1.60‐fold enhancement in gene editing efficiency compared to wild‐type Casδ‐1 at the B2M, CⅡTA, and TRAC gene loci (Figure 2A). Therefore, it was chosen as the candidate for EIP optimization.

Figure 2.

Editing activity of Casδ‐1 variants with a combination of single amino acid substitutions in HEK293T cells

(A–E) Editing activity of Casδ‐1 variants with combinations of single amino acid substitutions associated with the optimization of EIP (A), EIR (B), EIH (C), and EIS (D, E). The relative editing efficiencies shown refer to the efficiency of a particular Casδ variant compared with that of the wild‐type Casδ‐1 at the same target. Error bars represent SD of means, n = 2 biological replicates. The mutations selected as the candidates for corresponding optimization are highlighted in red.

For the EIR efficiency optimization, we also selected two variants (K413R and K420R) from the initial screen, both of which were located in the Wedge (WED) domains of Casδ‐1 (Figures 1B, S3B), for the next combination. However, no additive effect was observed when K413R and K420R were combined, as the editing efficiency was lower than that of the single K420R substitution (Figure 2B). Therefore, we used the K420R variant as the candidate for EIR optimization.

For EIH efficiency optimization, the top four variants from the initial screen were selected. These included K651R, K661R, and E664R located in the Recognition lobe (REC‐Ⅲ) domain, and L672R located in the bridge helix (BH) domain (Figures 1B, S3C). Since the L672R variant demonstrated the best editing activity performance in the initial screen (Figure 1C), we decided that it was essential to combine it with the other variants. A total of seven combinations with at least two amino acid substitutions were constructed, among which the K651R/L672R variant showed the highest average editing efficiency at the target sites of B2M, CⅡTA, and TRAC (Figure 2C). Additionally, we observed that although the editing efficiency varied among different targets, the average editing efficiency of the K661R variant at the three target sites was similar to that of the K651R/L672R variant (Figure 2C). Consequently, both the K661R and K651R/L672R variants were selected as candidates for EIH optimization.

According to the initial screening, there were a total of 12 EIS optimization‐associated substitutions with over 1.2‐fold improvement of editing activity relative to wild‐type Casδ‐1, far more than that of the substitutions identified for EIR, EIP, and EIH efficiency optimization (Figures 1B, S3D). To maximize the utilization of these substitutions, we selected nine variants located in the RuvC‐Ⅱ and Nuclease lobe (NUC) domains and randomly divided them into two groups for combinatorial analysis. Group I included the A710R, F701R, S702R, G716R, and V830R variants, while group II included the Q717R, P760R, D814R, and V817R variants. For group I, the A710R variant showed the most significant improvement in editing efficiency during the initial screening (Figure 1C). Therefore, it was considered essential to combine it with the other variants. In total, 18 combinations were constructed. According to the average editing efficiency at B2M, CⅡTA, and TRAC gene loci, nine combinations were identified with further improvement in activity. Among these, the top four combinations were chosen as candidates, including A710R/S702R, A710R/G716R, A710R/S702R/V830R, and A710R/G716R/V830R (Figure 2D). Similarly, we constructed 7 combinations for group II, all of which harbored the Q717R substitution, which showed the best performance in the initial screen (Figure 1C). Among these, the combinations of Q717R/D814R, Q717R/P760R/D814R, and Q717R/P760R/D814R/V817R had higher editing efficiencies (Figure 2E). These combinations, together with the P760R single substitution, were selected as candidates for group II.

Obtaining the activity‐enhanced Casδ‐1 variants by combination of EIP, EIR, EIH, and EIS optimization

Subsequently, we attempted to further engineer Casδ‐1 by integrating EIP, EIR, EIH, and EIS optimizations. First, we constructed 51 combinations based on 4 candidates of EIP, EIR, and EIH optimizations and 4 candidates from group I of EIS optimization. Robust and significant synergistic effects were observed for some combinations (Figure 3A). Among these, the combination of D120F/K129R/K420R/S702R/A710R/V830R/K651R/L672R, named Casδ‐8M, outperformed other combinations, with an average of 2.9‐fold gene editing efficiency enhancement compared to wild‐type Casδ‐1 at the B2M, CⅡTA, and TRAC gene loci (Figure 3A). Next, we added the corresponding mutations of 4 candidates from group Ⅱ of the EIS optimization to Casδ‐8M, respectively. We found that the addition of P760R to Casδ‐8M further enhanced the editing performance at the TRAC gene (Figure 3B). This Casδ‐1 variant with significantly enhanced editing activity, which harbored a total of nine amino acid substitutions (Figure 3C), was designated as enCasδ.

Figure 3.

Engineering of Casδ‐1 by combination of the EIP, EIR, EIH, and EIS optimization

(A) Editing efficiency of Casδ‐1 variants with combinations of mutations associated with EIP, EIR, EIH, and EIS optimization. Error bars represent SD of means, n = 2 biological replicates. The variant, named Casδ‐8M, outperformed other combinations. (B) Editing efficiency of Casδ‐1 variants with further combination of mutations associated with EIS optimization into Casδ‐8M. Error bars represent SD of means, n = 2 biological replicates. The variant with the addition of P760R at Casδ‐8M was called enCasδ. (C) Structural positions of nine amino acid substitutions present in enCasδ. (D) Comparison of editing activity of Casδ‐1, Casδ‐8M, enCasδ, SpCas9, and LbCas12a in HEK293T cells. Their editing efficiencies at 10 target sites in the B2M, CⅡTA, TRAC, PDCD, VEGFA, and EMX1 genes were used for analysis. Significant differences were determined by one‐way ANOVA.

Both Casδ‐8M and enCasδ were selected for further evaluation of editing activity and comparison with SpCas9 and LbCas12a based on 10 target sites in the B2M, CⅡTA, TRAC, PDCD, VEGFA, and EMX1 genes. All target sites were designed with overlapping spacers among different nucleases, including 5 sites where the spacer sequences of SpCas9 and/or LbCas12a completely overlapped with those of Casδ (Table S4). A similar design strategy was also widely used previously to minimize the effect of differences in spacer sequence on editing efficiency for comparison of different Cas nucleases (Bernabé‐Orts et al., 2019; Al‐Shayeb et al., 2022; Chen et al., 2023; Huang et al., 2023; Bai et al., 2025). The results indicated that enCasδ showed average gene editing efficiencies of 54.6% at the 10 target sites, which was significantly higher than that of the wild‐type Casδ‐1, with an average of 8.5‐fold (ranging from 1.3‐ to 29.3‐fold) enhancement (Figures 3D, S4). A significant increase in editing activity was also observed for Casδ‐8M, which showed average gene editing efficiencies of 55.1%, similar to that of enCasδ (Figures 3D, S4). Additionally, we found that the editing efficiencies of Casδ‐8M and enCasδ are generally comparable to those of SpCas9 and LbCas12a, two widely used Cas nucleases, at the tested sites (Figures 3D, S4). Taken together, these results suggested that the engineered Casδ‐8M and enCasδ are both highly efficient Cas nucleases for genome editing in mammalian cells.

enCasδ shows high genome editing efficiency in maize

The high genome editing activity of Casδ‐8M and enCasδ in HEK293T cells prompted us to evaluate their in vivo gene editing performance in maize. We transfected maize protoplasts with a construct expressing the optimized Casδ‐1 variant and target crRNA, and determined the indel rates after incubating them for 48 h. A total of five target sites in PSY1, GAR2, TS1, TS3, and BX9 genes were designed for evaluation. The editing efficiencies of enCasδ were higher than that of Casδ‐8M at all five target sites, with enhancement ranging from 1.1‐ to 6.5‐fold as compared to Casδ‐8M (Figure 4A). Taken together, enCasδ was selected as the best‐performing combination variant based on the evaluation in human cells and maize protoplasts.

Figure 4.

enCasδ displays robust genome editing in maize

(A) Comparison of editing activities of Casδ‐1, Casδ‐8M, and enCasδ at five target sites in maize protoplasts. Error bars represent the SD of the means, n = 3 biological replicates. Significant differences were determined by one‐way ANOVA. For a given gene, different letters indicated above the bar represent significant differences (P < 0.05), whereas the same letter represents no significant difference. (B) Comparison of editing activity of Casδ‐1, enCasδ, SpCas9, and Cas‐SF01 in maize protoplasts. Their editing efficiencies at 10 target sites in the TS1, TS3, TS4, GAPDH, GAR2, PSY1, DGK1, and BX9 genes were used for analysis. Significant differences were determined by one‐way ANOVA. (C) Profile of deletion sizes introduced by enCasδ at TS4 locus in maize protoplasts. (D) Schematic representations of the genome editing constructs with enCasδ for maize transformation. ZmU6‐2: maize U6‐2 promoter; ZmUbi: maize ubiquitin promoter; E9 ter: rbcS E9 terminator; DR: Direct repeat; and NLS: nuclear localization signal. (E) Genotypes of six T0 heterozygous plants edited by enCasδ at the TS4 locus. (F) Genotypes of two T0 heterozygous plants edited by enCasδ at the PSY1 locus.

Next, we attempted to further determine the editing activity of enCasδ in maize using 10 target sites on TS1, TS3, TS4, GAPDH, GAR2, PSY1, DGK1, and BX9 genes. According to the results in maize protoplasts, enCasδ displayed 1.4‐ to 8.8‐fold higher efficiencies than the wild‐type Casδ, with an average enhancement of 5.3‐fold (Figures 4B, S5). The editing efficiencies of enCasδ at TS1, TS4, GAPDH, PSY1, DGK1, and BX9 targets ranged from 12.8% to 22.7%, although these were still relatively low at the remaining four targets (Figure S5). Simultaneously, we compared the in vivo editing efficiencies of enCasδ to Cas9 and Cas‐SF01; the latter is an engineered Cas12i3 nuclease reported recently (Duan et al., 2024a). All 10 target sites used had overlapping spacers among different nucleases, including 6 and 3 sites designed with completely overlapping spacer sequences between Casδ and SpCas9, and between Casδ and Cas‐SF01, respectively (Table S5). In general, the average editing efficiencies of enCasδ (11.2%) at the 10 tested target sites were comparable to those of Cas‐SF01 (7.9%) and Cas9 (10.0%) (Figures 4B, S5). Particularly, enCasδ showed significantly higher editing efficiencies than Cas9 at BX9‐sg2, TS4, and PSY1 target sites, and significantly higher editing efficiencies than Cas‐SF01 at BX9‐sg2, GAPDH, DGK1, and PSY1 target sites (Figure S5). We further characterized the sizes of deletions generated by editing enCasδ at the TS4 locus, which was selected as an example. The next‐generation sequencing data‐based analysis demonstrated that most of the deletions were larger than 10 bp (Figure 4C).

Then, we characterized the editing fidelity of the enCasδ system based on three randomly selected target sites in maize, including TS1, BX9‐sg2, and GAPDH. No detectable indel mutations were observed for all 11 potential off‐target loci analyzed in maize protoplasts as compared with the negative control (Figure S6A). Similarly, no off‐target events were also detected for the enCasδ system at all 14 potential off‐target sites for three representative human cell targets, including B2M‐sg1, PDCD1‐sg2, and TRAC‐sg1 (Figure S6B). These results demonstrated that enCasδ is highly specific in genome editing (Figures 3D, 4B). We also performed an in vitro depletion assay by using the PAM library containing eight random nucleotides at the 5’ end of the spacer sequence for both Casδ‐1 and enCasδ, and found no remarkable difference in PAM preference between them (Figure S7). Similar to Casδ‐1 (Yang et al., 2025), enCasδ also specifically recognized 5′‐RYR‐3′ PAMs (where R = A or G, Y = T or C).

Following the assessments of enCasδ in maize protoplasts, we further selected two target sites, TS4 and PSY1, to evaluate the editing performance of enCasδ in stable transgenic lines of maize. The enCasδ system was constructed into an all‐in‐one vector, pBUE411, in which the enCasδ nuclease and the crRNA targeting endogenous TS4 or PSY1 gene were driven by the maize ZmUbi and ZmU6‐2 promoters, respectively (Figure 4D). In total, 25 and 16 T0 stable transgenic maize lines were obtained for TS4 and PSY1 targets, respectively. The genotypes of these transgenic lines were analyzed by Sanger sequencing of PCR amplicons, with at least 10 clones sequenced for each line. Among 25 stable lines of TS4, 21 (84%) were identified to have targeted deletions, including 6 lines in which the genotypes of all mutated clones were identical and accounted for about half of the clones sequenced (Figure 4E; Table S8). Further targeted deep sequencing suggested that these 6 lines were heterozygous mutants, as the dominant types of mutants took up about half of the sequencing reads generated (an average of 45.66%), similar to the proportion of wild‐type reads (an average of 43.04%) (Figure S8), while the remaining lines were chimeras. Similarly, 12 (75%) of the edited lines were identified to have targeted deletions for PSY1 by Sanger sequencing and NGS, including 2 heterozygous mutants (Figure 4F; Table S8). In line with the observation in protoplasts, the sizes of DNA fragment deletions mainly ranged from 10 to 30 bp in these transgenic maize lines.

Overall, both maize protoplast‐ and transgenic line‐based results indicated that enCasδ could mediate robust in vivo gene editing in maize.

DISCUSSION

Taking advantage of the rapid growth in genomic and metagenomic sequence databases, an increasing number of CRISPR‐Cas systems have been discovered (Harrington et al., 2018; Strecker et al., 2019; Pausch et al., 2020; Altae‐Tran et al., 2021; Özcan et al., 2021; Xu et al., 2021; McGaw et al., 2022; Wang et al., 2023; Yan et al., 2019). However, as all CRISPR‐Cas systems originate from bacteria, archaea, or bacteriophages (Makarova et al., 2019), and have not evolved for efficient genome editing in plants and animals, the majority of the newly discovered CRISPR‐Cas systems show poor activity as gene editing tools (McGaw et al., 2022; Chen et al., 2023; Han et al., 2023; Duan et al., 2024a; 2024b). In this study, we developed a hierarchical engineering strategy to improve the genome editing activity of the CRISPR‐Cas system. Through identification of efficient single amino acid substitutions, separate optimization of the regions interacting with crRNA, the PAM duplex, the ssDNA substrate, and the RNA–DNA heteroduplex, and the combination of optimization of different regions, we successfully engineered natural Casδ‐1 into a robust genome editing tool named enCasδ. Compared to the wild‐type Casδ‐1, enCasδ harbors a total of 9 amino acid substitutions that synergistically augment its editing activity. Our results suggested that the hierarchical engineering strategy used here is powerful for aggregating the effects of as many amino acid substitutions as possible, thereby achieving maximum editing activity optimization, especially when no large effect amino acid substitution was identified. Notably, the optimization strategy used here is part of structure‐guided protein‐directed evolution, which essentially involves searching for the optimal solution of protein fitness within the protein sequence space. Therefore, the combinatorial effects of different single mutation sites are not a simple linear superposition (Soskine and Tawfik, 2010). In other words, the optimal variant is not achieved by simply stacking as many single beneficial mutations as possible.

Casδ‐1 is a recently identified evolutionary transitional CRISPR system characterized by its compact size (~900 amino acids) and guidance with a short crRNA without the requirement of a tracrRNA (Yang et al., 2025). Besides, Casδ‐1 displayed broad temperature tolerance (10°C–65°C, as tested in vitro), which is conducive to meeting the various temperature requirements for application in different species. Here, we generated enCasδ, a highly optimized Casδ‐1 variant, which showed an average 8.5‐ and 5.3‐fold enhancement in genome editing efficiencies at the tested targets in human cells and maize protoplasts, respectively. The overall editing performance of enCasδ was comparable to that of SpCas9, LbCas12a, and a recently engineered Cas12i3 nuclease (Cas‐SF01) (Zetsche et al., 2015). We noted that, according to the tested targets in HEK293T cells and maize protoplasts, a certain degree of target dependency was still found for enCasδ relative to SpCas9, although it was reduced as compared to Casδ‐1. Certainly, there is clear room for further improvement of the editing efficiency and reduction of the target dependency of enCasδ via other approaches, including crRNA optimization, species‐specific codon optimization, and exonuclease fusion, which have been demonstrated to be effective for optimization of CRISPR‐Cas genome editing systems (Kleinstiver et al., 2019; Zhang et al., 2022; Wang and Doudna, 2023; Wang et al., 2024; Duan et al., 2024a; 2024b; He et al., 2025; Meng et al., 2025; Wang et al., 2025a; 2025b). In line with the observation of Casδ‐1, enCasδ was found to specifically recognize 5′‐RYR‐3′ PAMs, which differs markedly from the 5′‐NGG‐3′ PAM of SpCas9 and the T‐rich PAM motifs observed in typical type V Cas12 family nucleases (Jiang et al., 2013; Zetsche et al., 2015), and thus increases toolkit diversity so as to expand the targeting range. In addition, similar to that of Casδ‐1, the size of dominant deletions induced by enCasδ is also larger than 10 bp, which is favorable for complete gene knockout and removal of regulatory elements, particularly for enhancers and TF binding sites in promoter regions. In summary, we obtained an activity‐enhanced Casδ‐1 variant, enCasδ, which will serve as a robust tool for genome editing applications in both animal cells and plants.

MATERIALS AND METHODS

Prediction of the ternary complex structure of Casδ‐1 and identification of candidate mutation sites

The ternary complex structure of Casδ‐1 with crRNA and DNA was predicted using AlphaFold3 (https://alphafoldserver.com/) (Abramson et al., 2024), with the random seed parameter set to 1. Protein–nucleic acid interaction sites within the ternary complex were then identified using PLIP (Schake et al., 2025), and residues located within 4–8 Å of these interaction sites were selected as candidate positions for mutagenesis. The visualization of the ternary complex structure of Casδ‐1 was performed using PyMOL (Rosignoli and Paiardini, 2022).

Plasmid construction

The human codon‐optimized Casδ‐1, SpCas9, and LbCas12a, and maize codon‐optimized Casδ‐1, SpCas9, and Cas‐SF01 coding sequences were synthesized by Tsingke Biotechnology (Beijing, China). The amino acid sequences of Casδ‐1, SpCas9, LbCas12a, and Cas‐SF01 are presented in Table S1. For HEK293T cells' editing, human codon‐optimized Casδ‐1, SpCas9, and LbCas12a were cloned into the BspDI (NEB)‐ and FseI (NEB)‐digested pZG vector backbone. Utilizing the constructed pZG‐Casδ‐1 vector, we further constructed the vectors of Casδ‐1 variants via PCR to introduce mutations at the given sites. The crRNA vectors corresponding to Casδ‐1, Casδ‐1 variants, and LbCas12a editors, as well as the sgRNAs corresponding to the spCas9 editor, were designed by annealing multiplex complementary oligos unidirectionally via their sticky‐end design and then ligating them into the BbsI (NEB)‐digested pCRISPR‐sg6 plasmid using T4 ligase.

For maize protoplast editing, the maize codon‐optimized Casδ‐1, SpCas9, and Cas‐SF01 were cloned into the AscI (NEB)‐digested pUC19 vector backbone. Based on the constructed pUC19‐Casδ‐1 vector, Casδ‐8M and enCasδ vectors were constructed via PCR to introduce mutations at the given sites. The crRNA corresponding to Casδ‐1, Casδ‐8M, and enCasδ editors, as well as the sgRNAs corresponding to the SpCas9 editor, were then introduced into the resulting pUC19‐Cas constructs to obtain the all‐in‐one vectors. The same strategy as that used for the construction of crRNA and sgRNA vectors for human cell gene editing was used.

For maize stable transgenic line editing, the codon‐optimized enCasδ was cloned into the StuI (NEB)‐digested pBUE411 vector (Xing et al., 2014). Next, the crRNAs targeted to the TS4 and PSY1 genes were then inserted into the resulting construct. All PCRs were performed with Phanta® Max Master Mix (Vazyme). The endogenous target sequences of human cells and maize used are provided in Tables S2 and S3, respectively.

Human cell culture and transfection

HEK293T cells were grown at 37°C with 5% CO₂ in Dulbecco's modified Eagle medium (DMEM, Thermo Fisher) containing 10% (v/v) heat‐inactivated fetal bovine serum (FBS, Gibco), 1% (v/v) penicillin–streptomycin (1 × 10⁴ units/mL penicillin and 1 × 10⁴μg/mL streptomycin, Thermo Fisher), and 1% (v/v) non‐essential amino acids (100 × NEAA, Thermo Fisher).

For transient transfection, HEK293T cells were seeded into 24‐well plates and transfected at ~80% confluency. After about 20 h, cells were co‐transfected with 0.6 μg of Cas protein expression pZG vectors carrying the puromycin‐resistant gene and 0.4 μg of crRNA expression pCRISPR‐sg6 vectors using LipoFiter 3.0 lipid‐based transfection reagent (HANBIO) following the manufacturer's recommended protocol. Two or three biological replicates were set for each transfection.

After puromycin (1.5 μg/mL, InvivoGen) selection for 48 h, the HEK293T cells were collected and used for genomic DNA extraction using 1× DNA lysis buffer (10 mmol/L Tris‐HCl, pH 8.0–8.5; 0.45% NP‐40; 0.45% Tween‐20; and 0.1 mg/mL proteinase K) at 55°C for 120 min and at 95°C for 15 min. The lysate was then used for genome editing efficiency analysis.

Maize protoplast isolation and transformation

The maize protoplast isolation and transformation were performed as described previously (Coy et al., 2022). For protoplast isolation, the seeds of the maize inbred line LH244 were grown at 28°C in the dark for 2 weeks. Then, etiolated leaves were cut to about 1.0 mm strips and digested with an enzymatic hydrolysis solution at 28°C for 3.5 h in the dark. The mixture was then filtered using a 70 μm cell strainer to collect the protoplasts. The resulting protoplasts were resuspended in a mannitol–MgCl₂–MES (MMG) solution at a concentration of about 1 × 10⁵ cells/mL. For PEG‐mediated protoplast transformation, 60 μg of vector DNA was mixed with 200 μL of protoplasts gently, and then 240 μL of PEG‐Ca²⁺ solution was added. The mixture was incubated at room temperature for 18 min in the dark. The reaction was next stopped by adding 960 μL of W5 buffer, followed by centrifugation to remove the supernatant. The protoplast pellet was resuspended in 900 μL of W5 buffer and incubated at 28°C in the dark for 48 h. Three or four biological replicates were set for each of the transformation. The genomic DNA of maize protoplasts was extracted using the CTAB (Cetyltrimethylammonium bromide) method (Springer, 2010) and used for further genome editing efficiency analysis.

Determination of genome editing efficiency by next‐generation sequencing (NGS)

NGS was applied to determine the genome editing efficiency of CRISPR‐Cas systems in HEK293T cells and maize protoplasts. The genomic DNA extracted from the HEK293T cells and maize protoplasts were used as templates. The target sequences were first amplified using M5 SuperFast Taq PCR MasterMix (Mei5 Bioservices Co., Ltd) with target‐specific primers. Common sequences (F: 5′‐CTGCAAGGCGATTAAGTTGGGTA‐3′; R: 5’‐GCAGTGAGCGCAACGCAATT‐3’) were added to the target‐specific primers, and thus the resulting PCR products served as templates for the next amplification step to further introduce sample‐specific barcodes and sequencing adapters. The primers used for constructing NGS libraries are listed in Tables S6 and S7. The resulting libraries were sequenced by the Illumina NovaSeq platform. Sequencing data were analyzed using CRISPResso2 (Clement et al., 2019).

Maize plant stable transformation

The pBUE411 binary vectors containing the enCasδ system were transformed into Agrobacterium tumefaciens strain EHA105. A transformation‐efficient elite inbred line of maize, LH244, was used for transformation (Tan et al., 2025). The embryos peeled from 10 to 13 d after pollination (DAP) seeds of maize LH244 were used for Agrobacterium‐mediated transformation as in a previous study (Li et al., 2021). Positive transgenic events were identified via herbicide resistance. Further verification of the transgenic events and analysis of the genotype of target site were performed by a combination of PCR amplification and Sanger sequencing at the seedling phase.

PAM depletion and validation assay

The PAM preferences of wild‐type Casδ‐1 and enCasδ were analyzed using an in vitro depletion assay described previously (Yang et al., 2025). Briefly, a randomized PAM plasmid library containing eight random nucleotides at the 5’ end of the spacer sequence was constructed. The PAM library consumption experiment was performed at 37°C for 8 h, with the addition of purified Cas protein and crRNA. The resulting products were further used for the construction of sequencing libraries and sequencing on the Illumina platform. Reads containing the PAM region were extracted, quantified, and normalized to total reads. Sequences with a depletion ratio greater than four‐fold relative to the control were used to generate the PAM sequence logo via WebLogo (Crooks et al., 2004).

Off‐target analysis

Potential off‐target sites for enCasδ at endogenous targets in HEK293T cells and maize protoplasts were predicted using Cas‐OFFinder software (Hwang et al., 2021), with a threshold at which the number of mismatches in spacer sequences was lower than 6 nucleotides relative to that of on‐target sites. An exception was that as only two potential off‐target sites were identified for the target of BX9‐sg2 with mismatches lower than 6 nucleotides, we added 2 extra potential off‐target sites with 7 mismatched nucleotides for analysis. The sequences around potential off‐target sites were amplified and sequenced on the Illumina platform for further editing efficiency assay. The On‐target site was set as a positive control. Three biological replicates were set. The spacer sequences of potential off‐target sequences are provided in Table S9. The primers used for off‐target analysis are listed in Table S10.

CONFLICTS OF INTEREST

The authors declare no competing interests.

AUTHOR CONTRIBUTIONS

J.L., J.C., Z.Y., and C.B. designed the project. F.G., C.P., Y.D., and Y.C. performed the experiments and editing efficiency analysis. Z.Z., M.Y., H.F., Y.X., and S.S. provided assistance with bioinformatics analysis and experiments. S.L. performed the maize transformation. Z.Y., C.B., B.X., H.Z., S.W., and J.L. contributed to discussion of the results. J.C., F.G., and C.P. wrote the manuscript. All the authors have read and approved the contents of this paper.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article: http://onlinelibrary.wiley.com/doi/10.1111/jipb.70222/suppinfo

Figure S1. Candidate mutation sites for engineering of Casδ‐1 nuclease

Figure S2. Editing activity of different Casδ‐1 variants

Figure S3. Structural positions of amino acids in Casδ‐1

Figure S4. Editing efficiency of different Cas systems in HEK293T cells

Figure S5. Editing efficiency of different Cas systems in maize protoplasts

Figure S6. Editing efficiency of potential off‐target sites

Figure S7. PAM preference of Casδ‐1 (A) and enCasδ (B)

Figure S8. Characterization of editing events in heterozygous mutants of the T0 generation

Table S1. Amino acid sequences of Casδ‐1, enCasδ, SpCas9, LbCas12a, and Cas‐SF01

Table S2. Endogenous target sequence information of different Cas systems in human cells

Table S3. Endogenous target sequence information of different Cas systems in maize

Table S4. Spacer sequences of targets designed for editing activity comparison of different Cas proteins in HEK293T cells

Table S5. Spacer sequences of targets designed for editing activity comparison of different Cas proteins in maize

Table S6. Primers used for target amplification

Table S7. Primers for constructing NGS libraries by fusing genomic amplicons

Table S8. Sanger sequencing‐based editing activity analysis at the TS4 and PSY1 loci in maize T0 generation lines

Table S9. Spacer sequences of potential off‐target sites of enCasδ in human cells and maize

Table S10. Primers used for amplifying sequences around potential off‐target sites

Biographies

Ge, F. , Peng, C. , Du, Y. , Chen, Y. , Zhao, Z. , Yu, M. , Feng, H. , Xie, Y. , Sun, S. , Liu, S. , et al. (2026). Synergistic engineering of Casδ nuclease for robust genome editing. J. Integr. Plant Biol. 68: 3231–3242.