Optimization of a hypercompact Fanzor2 system for improved genome editing performance in plants

Abstract

Native Fanzor2 nucleases exhibit weak editing activity in plants. An optimized Fanzor2 system engineered via ωRNA optimization and structure‐based protein mutagenesis achieves robust editing at recalcitrant genomic sites and enables cytosine base editing, supporting versatile crop genome modification via transformation or viral delivery.

Fanzors (Fz), as the eukaryotic orthologs of prokaryotic TnpB, are the programmable RNA‐guided DNA endonucleases that belong to the obligate mobile element guided activity (OMEGA) system (Saito et al., 2023). Recently, an engineered enNlovFz2, which recognizes an expanded TAM of 5’‐NMYG‐3’, exhibited an approximately 11‐fold increase in editing efficiency in mammalian cells (Wei et al., 2025), but its editing activity has not been investigated in plants yet. In addition, the performances of the Fanzor systems in plants remain largely unexplored and exhibit relatively lower editing activities (Zhang et al., 2024; Ji et al., 2026).

In this study, we first systematically assessed the genome editing efficiencies of NlovFz2/ωRNA and enNlovFz2/ωRNA in soybean hairy roots. We then optimized the ωRNA expression by adding a self‐cleaving ribozyme (ωRNA‐Rz) to generate enNlovFz2/ωRNA‐Rz in order to improve its editing activity. Furthermore, through AlphaFold3 structural alignment, rational design, and in vivo testing in soybean hairy roots, we discovered that a key amino acid change at K416R in enNlovFz2 significantly enhanced its editing activity and engineered an improved version, enNlovFz2(K416R)/ωRNA‐Rz. Moreover, we also primarily show the potential of the Fanzor2‐based cytosine base editor, which has not been previously documented in plant base editing.

We first benchmark the editing performances of NlovFz2 and enNlovFz2 systems (NlovFz2/ωRNA and enNlovFz2/ωRNA) in soybean hairy roots. We constructed the all‐in‐one expression vectors containing either the plant codon‐optimized NlovFz2 or enNlovFz2 driven by the CaMV35S promoter and its cognate ωRNA expressed under the Arabidopsis U6 (AtU6) promoter, respectively (Figure 1A). We then targeted eight endogenous sites located in GmBADH2, GmCCD4a, GmSweet15a, and GmFAD2‐1A genes in soybean, respectively. These vectors were subsequently introduced into soybean hairy roots via Agrobacterium rhizogenes‐mediated transformation to evaluate their editing activities. Our next‐generation sequencing results revealed that whereas NlovFz2/ωRNA enabled detectable editing only at two of the eight endogenous targeted loci (GmSweet15a‐T1 and GmBADH2‐T3), with efficiencies ranging from 1.07% to 1.23% (Figure 1B), the enNlovFz2/ωRNA system enhanced the editing activity at various degrees, reaching up to 5.45‐ and 6.75‐fold at GmSweet15a‐T1 and GmBADH2‐T3, respectively (Figure 1B). Notably, the enNlovFz2/ωRNA successfully edited four inaccessible loci (GmBADH2‐T1, GmBADH2‐T2, GmFAD2‐1A‐T1, and GmFAD2‐1A‐T2) by NlovFz2/ωRNA, albeit still no editing activities were detected at the rest of two loci (GmFAD2‐1A‐T3 and GmCCD4a‐T1) (Figure 1B).

Figure 1.

Engineering NlovFz2 system for improved genome editing performance in plants

(A) Schematic representation of the NlovFz2 and two enNlovFz2 systems for genome editing in soybean hairy roots. (B) Genome editing efficiency at eight endogenous loci in soybean hairy roots. (C, D) Deletion positions and distribution of deletion sizes at the GmSweet15a‐T1 site. (E) Structural alignments of enNlovFz2 with ApmFz2 and IsDra2, respectively. The conserved residue in the RuvC domain of IsDra2 TnpB (I304) and its corresponding residue in that of enNlovFz2 (K416) are highlighted in red. (F) Domain organization of NlovFz2 and the editing efficiencies of the respective enNlovFz2 variant at the GmSweet15a‐T1 site. (G) Genome editing efficiency of enNlovFz2(K416R) at the remaining seven target sites. (H) Schematic representation of the enNlovFz2‐based cytosine base editor. (I) Editing efficiencies of denNlovFz2(K416R)/ωRNA‐Rz/CBE system at three endogenous target sites in the soybean hairy root. (B, F, G, I) Data are presented as mean values ± SD (n > 3). Data were analyzed using one‐way ANOVA, assuming a Gaussian distribution of residuals and equal standard deviations among groups (ns, not significant, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). (J) Schematic representation of enNlovFz2(K416R)/ωRNA‐Rz constructs for genome editing in rice. (K) The editing efficiencies induced by enNlovFz2(K416R)/ωRNA‐Rz across eight endogenous sites in T₀ rice plants.

To further improve the editing performance of enNlovFz2/ωRNA, we coupled a self‐cleaving HDV ribozyme at the 3’‐terminus of ωRNA to generate enNlovFz2/ωRNA‐Rz (Figure 1A). This modification significantly enhanced editing activity, enabling the editing of two target sites (GmFAD2‐1A‐T3 and GmCCD4a‐T1) that could not be edited by enNlovFz2/ωRNA, with a 1.79‐ to 63.02‐fold increase in efficiency at the rest of the target sites (Figure 1B). Our next‐generation sequencing results revealed that the enNlovFz2/ωRNA‐Rz mainly produced the deletions ranging from 1 to 30 bp, which primarily centered around the 14th nucleotide position of the spacer region, counting from the proximal of TAM as position 1 (Figure 1C, D).

Moreover, rational design and protein engineering have been documented as a useful strategy for enhancing the editing activity of various genome editing tools. Our sequence alignment and AlphaFold3 structural analyses revealed that, despite low amino acid (aa) sequence homology (Figure S1), enNlovFz2 shares high structural similarity to ApmFz2 and a TnpB ortholog, IsDra2, with RMSD values of 4.15 and 3.82, respectively (Figure 1E). Deep mutational scanning and rational design of ApmFz2 and IsDra2 have identified numerous key amino acids, which are located in different functional domains and significantly boost editing efficiencies either alone or in combinations in mammalian cells (Schargel et al., 2025). We hypothesized that mutating the corresponding residues in enNlovFz2 could confer similar enhancement of its genome editing activity. To test this hypothesis, we generated 24 enNlovFz2 variants, including 16 inferred from structural alignment of enNlovFz2 vs IsDra2, and 8 from that of enNlovFz2 vs ApmFz2, respectively (Figure 1E; Table S1). The editing efficiencies of these enNlovFz2 variants in an architecture of enNlovFz2 variant/ωRNA‐Rz were assessed individually at the GmSweet15a‐T1 site in soybean hairy roots via next‐generation sequencing. Our results indicated that, among the 24 variants, two variants (S58R and R214K) maintained editing activities comparable to the wild‐type enzyme, while 21 variants showed reduced average genome editing efficiency to different extents. The remaining variant carrying the single substitution K416R yielded markedly enhanced editing activity, with editing efficiency elevated by 42% relative to wild‐type enNlovFz2 (Figure 1F). We subsequently designated this system as enNlovFz2(K416R)/ωRNA‐Rz. It is worth mentioning that the residue K416R corresponds to I304R in IsDra2 TnpB, a known mutational hotspot located in the RuvC domain that enhances editing activity in the previous engineered variants such as TnpB‐VGIRL and Ymu1‐WFR (Thornton et al., 2026; Zhou et al., 2026).

To further validate its broad utility, we tested the performance of enNlovFz2 (K416R) at the remaining seven target loci. Impressively, except for GmFAD2‐1A‐T1, the editing efficiencies across six of seven target sites were significantly improved compared to those of the NlovFz2/ωRNA (Figure 1G). Furthermore, enNlovFz2(K416R) enhanced the average editing efficiency ranging from 5% to 110% at the remaining six target sites, albeit statistical significance was not reached at some individual sites when compared with enNlovFz2/ωRNA‐Rz (Figure 1G). In addition, we also observed that the efficiencies varied among different target sites, even within the same gene context, such as three target sites of GmBADH2 (Figure 1G), indicating a target‐site dependency like most of the members in the Cas12 family. This target‐site dependency highlights the necessity of carefully screening for more effective target sites in practical gene editing experiments.

To date, there have been no reports of Fanzor‐based plant base editor yet. To explore its potential, we generated a catalytically inactive variant, named denNlovFz2(K416R), by introducing a D294A mutation in enNlovFz2(K416R), based on the D191A mutation that inactivates IsDra2 (Luo et al., 2025) and structural alignment (Figure S2). Then, the denNlovFz2 was fused with the mini‐Sdd7 cytidine deaminase and uracil glycosylase inhibitor (UGI) to generate a cytosine base editor designated as denNlovFz2(K416R)/ωRNA‐Rz/CBE (Figure 1H). We further evaluated the base editing activities of denNlovFz2(K416R)/ωRNA‐Rz/CBE at three target sites (GmSweet15a‐T1, GmBADH2‐T1, and GmBADH2‐T3). Our preliminary results show that the editing windows were predominantly located at positions + 2 to + 13 bp within the spacer sequence, with the average editing efficiencies ranging from 0.94% to 1.81%, and the highest C‐to‐T editing efficiency of 8.27% was observed at GmSweet15a‐T1 (Figures 1I, S4).

To assess the genome editing efficiency of enNlovFz2(K416R)/ωRNA‐Rz in monocot species, we constructed an all‐in‐one expression vector and evaluated its performance at eight endogenous loci in rice stable lines (Figure 1J, K). The enNlovFz2(K416R)/ωRNA‐Rz successfully enabled targeted mutagenesis with editing efficiencies ranging from 17.6% to 93.3% (Figure 1K). Among the edited lines, biallelic or homozygous mutations were detected at all target sites in the T₀ generation, with the highest frequency observed at the Os11g0508600 locus, where 28 out of 30 transgenic plants were successfully edited, and 15 of these were biallelic or homozygous in the T₀ generation (Figure 1K). The THYLAKOID FORMATION1 gene (OsTHF1) is involved in regulating chlorophyll degradation during leaf senescence in rice. Phenotypic analyses of the T₀ homozygous and biallelic knockout mutants under dark‐induced leaf senescence conditions revealed a stay‐green phenotype of the mutant lines (Figure S5).

In summary, we engineer a more robust enNlovFz2(K416R)/ωRNA‐Rz system, which exhibits superior performance over NlovFz2/ωRNA for plant genome editing through rational design and protein engineering, as well as the optimization of ωRNA expression. We also develop a first enNlovFz2(K416R)‐based plant cytosine base editor, which exhibits potential in C‐to‐T base transitions.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest. A patent application has been filed on the NlovFz2‐mediated genome editing in plants.

AUTHOR CONTRIBUTIONS

Y.S and L.X. conceived the study and designed the experiments. X.C., S.B., H.L., R.W., and Y.S. performed the experiments. Y.S., L.X., X.C., and S.B. wrote and revised the manuscript. All authors read and approved the final manuscript.

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.70346/suppinfo

Figure S1. Multiple sequence alignment of enNlovFz2, ApmFz2, and IsDra2 TnpB.

Figure S2. Alignment of IsDra2 TnpB with enNlovFz2.

Figure S3. Genome editing efficiency of enNlovFz2(K416R) at 18 non‐CCG TAM loci.

Figure S4. Base editing profiles at the GmSweet15‐T1 site.

Figure S5. Genotypes and phenotypes of the different Osthf1 mutants.

Table S1. NlovFz2 mutation corresponds to gain‐of‐function mutations in IsDra2 TnpB and ApmFz2, respectively.

Table S2. Analysis of potential off‐target effects.

Table S3. Sequences of the expression cassette for enNlovFz2, denNlovFz2/CBE, and ωRNA.

Table S4. Amino acid sequence of NlovFz2 variants.

Table S5. Target sites in this study