A miniature alternative to Cas9 and Cas12: Transposon‐associated TnpB mediates targeted genome editing in plants

The two popular genome editor nucleases, Cas9 and Cas12a, hypothetically evolved from IscB and TnpB, respectively (Altae‐Tran et al., 2021). Recent reports showed that TnpBs also function as RNA‐guided nucleases in human cells (Karvelis et al., 2021). TnpB proteins are much smaller (~400 aa) than Cas9 (~1000–1400 aa) and Cas12a (~1300 aa). The large cargo size of Cas9 and Cas12a hinders their delivery into cells, particularly through viral vectors. Hence, TnpB offers an attractive candidate that can be adopted as a new type of genome editing tool for eukaryotes. However, it is unknown whether TnpB can mediate genome editing in plant systems. In this study, we developed and optimized hypercompact genome editor based on TnpB protein from Deinococcus radiodurans ISDra2 and achieved editing efficiency as high as 33.58% on average in the plant genome.

To develop a TnpB genome editing system in plants, we first codon optimized the ISDra2TnpB and cloned it under the OsUbi10 promoter. The right end element (reRNA), which forms an RNP complex with TnpB protein, is required for target DNA recognition and cleavage (Karvelis et al., 2021; Figure 1a). We used a protoplast system workflow for evaluating TnpB‐mediated editing (Figure 1b; Panda et al., 2024). We cloned the reRNA component under the OsU3 promoter to construct pK‐TnpB1 (Figure 1c; Figure S1). Analogous to the PAM requirement of Cas12, TnpB cleavage is dependent on the presence of transposon‐associated motif (TAM) 5′ to the target sequence. For ISDra2TnpB (only TnpB from this point onward), the TAM sequence is 5′‐TTGAT‐3′. Genome‐wide analysis revealed a 0.35% TTGAT TAM coverage in rice, highlighting TnpB's unique targetability to regions not accessible by Cas9 or Cas12a. TnpB cleaves targets at 15–21 bp from TAM, generating staggered patterns (Karvelis et al., 2021; Figure 1a). We have designed guide RNAs for six rice genomic loci, with five containing a recognition sequence for a specific restriction enzyme at the expected cleavage site. To assess effectiveness, we transfected rice protoplasts with these six constructs and cloned amplified target loci into pGEM‐T‐Easy vector. We digested the colony PCR products with target‐specific restriction endonucleases (REs). On average, screening 100 colonies per guide revealed 1.5–7.15% of undigested bands due to disruption of RE sites. Sanger sequencing of the undigested bands confirmed the result. We observed mostly deletions ranging 7–53 bp across the targets (Figure S1). To assess editing efficiency in the whole protoplasts population, we repeated transfection and performed targeted deep amplicon sequencing. pK‐TnpB1 induced mutations at all target loci, exhibiting the highest indel efficiency (average 14.84 ± 4.88%) at the HMBPP locus (Figure 1d).

Figure 1.

TnpB system developed as a hypercompact plant genome editor. (a) Schematic of the IsDra2TnpB attached with target dsDNA. TAM, Transposon associated motif; reRNA, right end RNA. (b) Plant protoplast assay system workflow. (c) Schematic of the TnpB1 vector. HDV, hepatitis delta virus ribozyme; NLS, nuclear localization signal. (d) TnpB‐mediated indel efficiency in six rice loci. (e) TnpB activity in targets with noncanonical TAM (TCGAT). (f, g) Multiple reRNA‐guide expression system and indel generation through multiplexing. (h) TnpB vector systems for Arabidopsis genome editing. (i) Indel efficiency in the target sites in Arabidopsis with TnpB‐D1 and TnpB‐D2. (j) Schematic of TnpB2, TnpB3 and TnpB4. HH, Hammerhead ribozyme. (k–p) Comparison of indel generation efficiency of TnpB vectors across six rice genomic loci. Protoplast transfections were done simultaneously for this comparative analysis. (q) Representative mutation spectrum generated by TnpB2 in OsHMBPP site. (r) TnpB binary vector used for Agrobacterium‐mediated transformation. (s) Albino mutants showing loss‐of‐function phenotypes for OsSLA4. Sanger chromatogram showing the causal 53 bp deletion. (t) T₁ plants from T₁ seeds collected from T₀ plant with HMBPP monoallelic edit. Albino mutants are due to complete loss‐of‐function of OsHMBPP. Representative chromatogram showing biallelic homozygous editing. All data represented as the mean of three biologically independent experiments (shown in dots) ± SEM.

To verify TAM specificity, we assessed TnpB activity in two loci with the noncanonical TCGAT TAM. TnpB caused <1% average indels, confirming its high specificity for the canonical TTGAT TAM (Figure 1e). Then, we checked whether TnpB is suitable for multiplex genome editing by targeting the OsBSRK1 and OsWAXY genes, using a polycistronic‐tRNA‐gRNA (PTG) system (Xie et al., 2015; Figure 1f). We obtained simultaneous indel efficiencies of 5.41 ± 1.85% and 5.31 ± 2.03% on average for BSRK1 and WAXY, respectively (Figure 1g; Figure S2).

Next, to examine the TnpB system in a dicot model, Arabidopsis, we constructed the TnpB‐D1 vector (Figure 1h). TnpB was expressed under AtUbi10 promoter and the reRNA+guide component was expressed with AtU6‐26 promoter. We measured the efficiency of editing at three loci in Arabidopsis protoplasts. TnpB‐D1 was found to induce 0.2–0.46% editing on average across the loci (Figure 1i). Encouraged by our earlier study, where eCaMV35S promoter showed superior performance to AtUbi10 in Arabidopsis (Panda et al., 2024), we constructed TnpB‐D2 by replacing the AtUbi10 promoter with eCaMV35S promoter. TnpB‐D2 exhibited enhanced editing efficiency, averaging 1.37 ± 0.39% at the AtGAT site; however, not much difference was observed at the AtABP and AtTMPK sites (Figure 1i; Figure S3).

We, then, constructed TnpB D191A to generate catalytically deactivated TnpB (dTnpB). Base editors can install precise base substitutions in the genome without creating double‐strand breaks (Molla et al., 2021). To generate an adenine base editor (ABE), we constructed two types of TnpB‐ABE by fusing TadA‐8e (Richter et al., 2020), separately to the N‐terminus and C‐terminus of dTnpB (Figure S4a,b). When we targeted five rice loci having multiple A's within each protospacer, dTnpB‐ABE8e‐N and dTnpB‐ABE8e‐C exhibited minimum base editing activity (0.42–1.12%) (Figure S4d). Interestingly, no indels were detected across the loci, indicating D191A alone is sufficient in making a dTnpB (Figure S4c). This low‐base editing outcome could be partially due to the inability to generate a nick in the unedited strand.

We next generated a transcription activator system with dTnpB. We fused a TV activation domain (6XTAL‐VP128) (Pan et al., 2021) to dTnpB to generate dTnpB‐Act (Figure S5a,b). We tested dTnpB‐Act for transcriptional activation of three genes, with a vector control. On average, a 9.24‐, 8.62‐ and 7.89‐fold increase in gene expression was observed for OsCHS, OsDXS and OsPDS, respectively, over the vector control (Figure S5c). This result suggests that dTnpB is a potent candidate for developing a miniature gene activation system for plants.

To further optimize the TnpB system, we have constructed three additional versions of TnpB vectors: TnpB2, TnpB3 and TnpB4 (Figure 1j; Figure S6). Since the guide RNA component is crucial in determining editing efficiency, we have manipulated the reRNA+guide expression cassette in these three versions.

The Cas12a system was previously shown to have better editing efficiency when crRNA was expressed with a Pol‐II promoter (Zhang et al., 2021). In TnpB2, OsU3 (Pol‐III) promoter was replaced with ZmUbi (Pol‐II) promoter and reRNA+guide was flanked with HH and HDV ribozymes. TnpB3 has an additional tRNA sequence upstream to reRNA+guide over the TnpB1. The tRNA gene contains internal promoter and is likely to increase the transcription of downstream sequences. Recently, a composite promoter (CaMV35S‐CmYLCV‐U6) has been shown to increase prime editing efficiency (Jiang et al., 2020). We used the composite promoter to drive reRNA+guide expression in TnpB4. We assessed the efficiency of TnpB2, TnpB3 and TnpB4 by targeting the same six loci as we did with TnpB1. Remarkably, when using TnpB2, the indel frequency for the OsHMBPP target reached 33.58 ± 18.54% on average, whereas it was 11.21 ± 2.91% and 6.59 ± 4.19% when employing TnpB3 and TnpB4, respectively (Figure 1k). Hence, TnpB2 offered almost a 2.5‐fold improvement at the OsHMBPP site over TnpB1. A representative schematic showing the mutation spectrum at the HMBPP locus is shown in Figure 1q. On average, the highest editing efficiency at the OsSLA4g1 site with TnpB1 was 2.09 ± 0.19%, which increased to 8.22 ± 1.53% with TnpB2, 5.53 ± 1.07% with TnpB3 and 3.94 ± 0.63% with TnpB4 (Figure 1l). TnpB2 demonstrated enhanced editing at the OsSLA4g2 test site, achieving an average efficiency of 3.63 ± 1.14%, marking a threefold improvement compared with TnpB1 (Figure 1m). TnpB3 and TnpB4 exhibited even greater enhancement at this site, registering average editing efficiencies of 4.31 ± 2.67% and 6.53 ± 0.58%, respectively. For the OsPi21 site, the performance of different vectors followed the sequence: TnpB1 < TnpB4 < TnpB2 < TnpB3 (Figure 1n). TnpB2 and TnpB4 exhibited enhanced editing efficiency comparable to TnpB1 at the OsCKX2 site (Figure 1o). The OsCAF2 guide showed inefficiency with all vectors (Figure 1p). Deletion length ranged from 1 to >90 bp across the sites (Figures S7–S12). Taken together, on average across the six loci, TnpB2 outperformed the other three versions of TnpB systems (Figure S13).

Finally, to evaluate TnpB's ability to perform genome editing in regenerated rice plants, we constructed a binary pKb‐TnpB2 vector (Figure 1r). We strategically chose the HMBPP and SLA4 genes, pivotal for chloroplast development, as their disruption yields readily observable albino phenotypes (Liu et al., 2020; Wang et al., 2018). Agrobacterium‐mediated transformation generated 53 and 65 T₀ plants for HMBPP and SLA4g2 targets, respectively. Locus‐specific genotyping, through Sanger sequencing, revealed both monoallelic and biallelic editing. In the case of SLA4g2, 12 T₀ plants displayed biallelic editing, resulting in albino phenotypes (Figure 1s; Tables S1 and S2). The majority of albino plants exhibited a 53‐bp deletion in the SLA4g2 site. However, for the HMBPP locus, we achieved monoallelic editing in T₀ plants with an efficiency of 38% (20 plants out of 53) (Tables S1 and S2). Subsequently, T₁ plants were generated from the T₁ seeds obtained from T₀ plants, yielding hmbpp homozygous and albino mutants in the T₁ generation with high efficiency (Figure 1t). Sanger sequencing unveiled a 23 bp deletion in majority of the hmbpp mutants.

In conclusion, our study pioneers the use of TnpB as a compact genome editor in both monocot and dicot plant species. By optimizing the expression system, we successfully increased editing efficiency. Repurposing dTnpB for transcriptional activation and base editing showcases TnpB's versatile potential. Future endeavours will refine reRNA structure and length, and explore TnpB's natural diversity. Our findings position TnpB as a versatile and promising tool for plant genome engineering.

Conflict of interest

The Indian Council of Agricultural Research (ICAR), New Delhi, filed a patent application related to this study.

Author contributions

K.M. and M.J.B. conceived and designed the study and supervised the entire project. K.M. and S.K. planned the experiments. S.K., D.P., M.D., S.P., R.S. and S.P.A. performed experiments and collected data. D.P., S.K., K.M., S.P., R.S., P.D., S.P.A. and J.S. analysed the data. Y.Y. provided advice on various experiments. K.M., S.K. and D.P. wrote the manuscript. Y.Y., J.S., A.K.N. and M.J.B. edited the manuscript.

Funding information

Indian Council of Agricultural Research (ICAR), New Delhi, in the form of the Plan Scheme—‘Incentivizing Research in Agriculture’ project.

Supporting information

Figure S1 Schematic diagram of pK‐TnpB1 vector and genome editing in rice.

Figure S2 pK‐TnpB multiplexing vector and frequency of deletion types in different loci in rice.

Figure S3 pK‐TnpB1 mediated genome editing in Arabidopsis (as a model dicot) and frequency of deletion types for different target genes.

Figure S4 pK‐dTnpB‐ABE8e mediated base editing and indel status at different target loci in rice.

Figure S5 dTnpB mediated gene activation in rice.

Figure S6 Schematic diagrams of different versions of TnpB constructs used for genome editing in rice.

Figure S7 Rice HMBPP (OsHMBPP) gene‐targeted with different versions of the TnpB vectors.

Figure S8 Rice SLA4 (OsSLA4) gene‐targeted with different versions of the TnpB vectors with guide sequence (1).

Figure S9 Rice SLA4 (OsSLA4) gene‐targeted with different versions of the TnpB vectors with guide sequence (2).

Figure S10 Rice Pi21 (OsPi21) gene‐targeted with different versions of the TnpB vectors.

Figure S11 Rice CKX2 (OsCKX2) gene‐targeted with different versions of the TnpB vectors.

Figure S12 Rice CAF2 (OsCAF2) gene‐targeted with different versions of the TnpB vectors.

Figure S13 Indel efficiency comparison of TnpB systems across six loci.

Table S1 Details of type of mutations in T0 plants.

Table S2 Details of the type of mutations in T1 plants.

Table S3 Different vectors, polynucleotide and polypeptide sequences, and their sequence IDs.

Table S4 Target genes with locus IDs and guide sequences.

Table S5 List of primers used for making different TnpB constructs.

Table S6 List of primers used for cloning different guides.

Table S7 List of primers used for screening of mutants and qRT‐PCR analysis.

Table S8 List of primers used for deep amplicon sequencing (NGS).

Data availability statement

All data supporting the findings of this study are accessible within the article and its supplemental file. The sequencing data for all target loci have been deposited in the National Center for Biotechnology Information (NCBI) under the Sequence Read Archive (SRA) with the BIOPROJECT ID: PRJNA1066105. Plasmids, pKb‐TnpB1 (plasmid #215333) and pKb‐TnpB2 (plasmid #215334), will be available through Addgene.