Targeted deaminase-free T-to-G and C-to-K base editing in rice by fused human uracil DNA glycosylase variants

Base editors (BEs), a groundbreaking class of genome editing tools, enable precise single-nucleotide alterations at target genomic sites, leading to mutations that either disable or enhance gene functions, thus significantly advancing plant functional genomics research and crop enhancement (Li et al., 2023). In plants, significant advancements have been made in DNA base editors that can directly modify adenine (A), cytosine (C) and guanine (G) (Li et al., 2018; Zong et al., 2017). Nevertheless, a direct base editor for thymine (T) remains elusive. Recently, two innovative deaminase-free glycosylase-based base editors were developed: the gTBE for direct T editing (T-to-S conversion, S = G or C) and the gCBE for direct C editing (C-to-G), enabling orthogonal base modifications in mammalian cells (Figure 1a; Tong et al., 2024). These base editors utilized the fusion of Cas9 nickase (nCas9) with engineered variants of human uracil DNA glycosylase (UNG), allowing for the direct excision of T or C to generate apurinic/apyrimidinic (AP) sites. However, such direct T base editor has not been developed in plants to date. In this study, we developed a deaminase-free direct T base editor (pTGBE) and direct C base editor (pCKBE, K = G or T) in rice, marking a substantial step forward in expanding genetic manipulation capabilities in plants.

Figure 1

Open in figure viewerPowerPoint

Development of pTGBE and pCKBE base editors in rice. (a) Schematic diagram showing the editing mechanisms of deaminase-free glycosylase-based thymine base editor (gTBE) and deaminase-free glycosylase-based cytosine base editor (gCBE) in mammalian cells. mhUNG, engineered human uracil DNA glycosylase variants. PAM, Protospacer adjacent motif. AP, apurinic/apyrimidinic sites. Star (*) in magenta indicates the nick generated by nCas9. TLS, translesion synthesis. DSB, double-strand break. InDel, insertion and deletion. (b) Plasmid constructs of pTGBE and pCKBE in this study. Short vertical lines represent mutated amino acids. Δ1-88: 1–88 amino acid residue truncation of human UNG2. (c) Summary of the editing efficiencies of pTGBE and pCKBE at the endogenous genes in stable T₀ transgenic lines. Ho, homozygous mutation; He, heterozyous mutation; Bi, biallelic mutation; Chi, chimeric mutation. The PAM sequence is highlighted in bold and the edited bases are marked in red. (d, e) Average frequencies of T conversions by pTGBE (d) and C conversions by pCKBE (e) across the protospacer positions 1–20 (where PAM is at positions 21–23) from the edited sites in (c). (f) Schematic diagram illustrating the retention of intron 1 of OsARF24 caused by pTGBE-mediated disruption of the splicing donor site (from GU in WT to GG in line #45). (g) RT-PCR analysis of OsARF24 mRNAs in WT and mutant #45. ‘M’, DNA Marker. ‘WT’, wild type. (h) The sequence of the RT-PCR amplicons from (g). (i) Diagram illustrating desired base conversion with available base editors in rice.

To establish a pTSBE in rice, we fused the rice-codon-optimized human uracil DNA glycosylase variant UNG2Δ88-Y156A/A214T/Q259A/Y284D (mhUNGv3) (Tong et al., 2024) to nCas9 with a 32-amino-acid linker. A bipartite nuclear localization signal peptide was fused to UNG variant to increase nuclear entry efficiency, resulting in the pTSBE construct (Figure 1b). We chose ten endogenous sites targeting five genes in rice to test the editing activities and windows. A total of 400 T₀ stable edited plants were obtained and Hi-TOM results showed that T-to-S base conversion in transgenic rice plants with up to 78.05% efficiency (Figure 1c), but essentially no C or G editing and A editing with 1.85% efficiency at all sites (Figure S1a–c). We found that pTSBE also induced insertions or deletions (InDels) with frequencies ranging from 20.00% to 75.32% at the ten edited sites (Figure 1c). Notably, the proportion of T-to-G edits (up to 78.05%, averaging 39.21%) in the products was 13.38-fold higher on average than that of T-to-C edits (up to 3.70%, averaging 2.93%). The T-to-G is the predominant editing type generated, with the purity exceeding 80% (Figure S1d), showing a quite different editing pattern from that in mammalian cells. In mammalian cells, gTBEv3 exhibited T-to-S base editing activity with average editing efficiencies of 27.26% and 18.75% for T-to-G and T-to-C, respectively (Tong et al., 2024). Thus, we designated this BE as pTGBE to better reflect its editing characteristics in plants. Furthermore, the editable range was positions T2–T12, T14 and T18, and the optimal editing window at positions T3–T5 with the highest editing efficiency at T3 (PAM position as 21–23) (Figure 1d). In contrast, gTBEv3 in mammalian cells typically produced T-to-S transversions at positions T2-T11 and optimal editing window at T5 (Tong et al., 2024). The T₀ events included homozygous, heterozygous, biallelic or chimeric-edited alleles (Table S3; Figure 1c). Homozygous base conversions were observed at 60.00% (6/10) for all the ten sgRNA sites, with a maximum efficiency of 30.77% (4/13) at OsNRT1.1B-SG3 site, while heterozygous base conversions reached up to 27.78% (15/54) at OsARF24-SG2 site. Phytoene desaturase (PDS) is a key enzyme involved in carotenoid biosynthesis, possessing a crucial single-domain (amino acids 106–556). T₀ plant #6 underwent homozygous T-to-G editing via sgRNA OsPDS-SG2, resulting in the alteration of Leucine at amino acid position 114 to Valine, leading to the observed albino phenotype with white stripes in the leaves (Figure S3).

We further explored the potential application of pTGBE in modulating gene expression through alternative splicing (AS). As pre-mRNA transcripts undergo processing, AS can lead to intron retention (IR), alternative 5′ splicing, alternative 3′ splicing and exon skipping, offering different gene expression patterns (Liu et al., 2024). Notably, both the splicing donor (SD) site and the complementary strand of the splicing acceptor (SA) site harbour a T. To illustrate this application, we designed sgRNAs specifically targeting SD or SA sites of the OsARF24 gene (Figure S4). We identified a homozygous mutant #45 with a T-to-G conversion at the desired target T within the 5′ splice site of intron 1, which was targeted by OsARF24-SG1. We performed RT-PCR using a forward primer in exon 1 and a reverse primer in exon 3. A 240 bp fragment was generated from wild-type (WT) plants, whereas a 319 bp fragment was amplified from mutant plant #45 (Figure 1f,g). Sequencing of this fragment revealed intron 1 was retained, which completely prevented the production of normal splicing isoform (Figure 1h). Additionally, we generated 12 heterozygous mutants targeting the SA site of intron 7 using OsARF24-SG2 in T₀ plants, which will produce homozygous lines in T₁ plants for identifying the AS isoforms (Figure S4). Overall, our results demonstrate that pTGBE can program AS by mutating T within SD or SA sites, enabling the production of desired mature transcripts.

To explore the editing type and efficiency of gCBE in plants, we engineered a construct by fusing the rice-codon-optimized human uracil DNA glycosylase variant UNG2Δ88-K184A/N213D/A214V (mhUNGv2) (Tong et al., 2024) to nCas9 for evaluating the editing characteristics in rice (Figure 1b). We chose eight endogenous sites targeting three genes in rice to test the editing activities and windows. Hi-TOM sequencing of a total of 255 T₀ transgenic plants showed that gCBE in rice caused highly efficient C base editing with frequencies ranging from 26.09% to 61.11%, including predominant C-to-G editing efficiency up to 58.33% as well as C-to-T conversions up to 40.91%, but essentially no A, T or G editing at all examined sites (Figure S2a–c). The percentage of purity for C-to-G/T conversions almost exceeded 85%, and there were very few C-to-A conversions detected (Figure S2d). Hence, we designated this BE as pCKBE. The editable range of pCKBE was positions C2-C7, C9-C11, C13 and C15-C16 (Figure 1e), with InDel efficiencies from 13.04% to 72.22% at the eight edited sites (Figure 1c). The T₀ events included homozygous (up to 8.51%), heterozygous (up to 22.58%), biallelic (up to 50.00%) and chimeric (up to 22.58%) edited alleles (Figure 1c; Table S4).

To evaluate the specificity of pTGBE and pCKBE in stable rice lines, we selected potential off-target sites based on the predictions made by Cas-OFFinder (http://www.rgenome.net/cas-offinder/) for all targets. Minimal off-target effects were observed. Only one off-target site of OsNRT1.1B-SG3-OFF1 by pTGBE and two off-target sites of OsLCY-SG3-OFF1 and OsNRT1.1B-SG3-OFF1 by pCKBE exhibited detectable edits (Table S5).

In this study, we have developed a novel deaminase-free base editor pTGBE that directly excised the T with an engineered DNA glycosylase (UNG), producing T-to-G editing with efficiencies up to 78.05% in stable rice lines. We also generated a new deaminase-free base editor pCKBE for C-to-K transversion editing in rice, producing C-to-G and C-to-T editing events. The pTGBE and pCKBE greatly broadened the targeting scope of base editors by breaking the narrow editing window, thus increasing the opportunity to obtain an efficient strategy for further research. By utilizing pTGBE to edit splicing sites, alternative splicing (AS) isoforms were generated, providing a novel approach to modulating gene expression patterns. However, pTGBE and pCKBE induced higher frequencies of InDels compared with the well-developed pABEs or pCBEs. The pABEs or pCBEs facilitate base editing through DNA repair following a deamination reaction. In contrast, pTGBE and pCKBE, as well as AYBE and CGBE, enabled base editing after the generation of AP sites. InDels are likely caused by DNA double-stranded breaks generated during the repair of these AP sites. Recently, two studies showed that the suicide enzyme HMCES could reduce the InDel byproducts by shielding AP sites and thus safeguarding the DNA from breaks during the editing by CGBE and TSBE base editors (He et al., 2024; Huang et al., 2024). In addition, introducing the Gam proteins, which bind to the ends of DSBs to prevent their degradation, reduced InDels (Komor et al., 2017). In summary, the engineered pTGBE and pCKBE will enable diverse base conversions in plants, expanding the plant base-editing toolbox. By combining pTGBE with other previously reported base editors, all 12 types of base conversions can be achieved in rice, especially for direct T base editing in future (Figure 1i).