Prime editing-mediated precise knockin of protein tag sequences in the rice genome

Dear Editor,

Accurately labeling proteins in living plant cells has long been a challenge and can be addressed by targeted insertion of tag sequences in a given locus. Recent optimized plant prime editors (PEs) enable efficient programmable installation of small insertions or deletions, including insertions of short sequences (Li et al., 2022a, 2022b; Jiang et al., 2022; Xu et al., 2022; Zong et al., 2022; Zou et al., 2022). To investigate whether prime editing can be used to tag endogenous proteins in rice, we made use of the enpPE2 system described in our previous report (Li et al., 2022b). At the site 9 bp upstream of the stop codon of the ubiquitin gene OsUBQ10 and the site 12 bp upstream of the stop codon of the α-tubulin gene OsTubA1, 6XHis and hemagglutinin (HA) sequences were separately designed in the reverse transcriptase template (RTT) of engineered prime-editing guide RNAs (epegRNAs) with the evopreQ₁ modification in the enpPE2 binary vector for use in Agrobacterium-mediated transformation (Figures 1A and 1B and Supplemental Figure 1). A PCR assay with one primer on the inserted fragment and the other on the flanking sequence (supplemental materials and methods) suggested that knockins of the tags were produced at both targets in stable transgenic plants (Figures 1C–1F). The genomic region across the target site was analyzed by sequencing to identify individual tagged T₀ lines. We found that 18-bp 6XHis was precisely knocked in to OsUBQ10 and OsTubA1 in 43.75% and 41.67% of plants, respectively (Figures 1G–1J; Supplemental Table 1). For the 27-bp HA, the frequency of plants with inserts was comparable to or slightly lower than that for 6XHis edits at the OsUBQ10 site (43.75% for 6XHis vs. 37.50% for HA), but the frequency was much lower at the OsTubA1 site (41.67% for 6XHis vs. 10.42% for HA). Although the targeted genes were tagged, we noticed that on average, 5.21% of plants harbored homozygous tag insertions at the OsUBQ10 site, whereas homozygous mutations were not detected at the OsTubA1 site, which might restrict the application of enpPE2-mediated protein tagging.

Figure 1.

Precise knockin of tag sequence in rice using prime editing.

(A) Schematic illustration of prime editing systems for tag knockin. Left, precise insertion of tags mediated by enpPE2 or ePE2. ePE2 was updated from enpPE2 by stacking an M-MLV RNase H domain truncation and viral NC protein fusion in pPEmax. The tag sequence (red) was embedded in the RTTs of evopreQ1-modified epegRNA and inserted into the target site in the rice genome by the enpPE2 system. To detect targeted insertions by PCR, a forward (F1) primer was designed at the tag sequence and paired with a reverse (R) primer on the 3′ flanking sequence. The target sites were amplified by another forward (F2) primer on the 5′ flanking sequence and the R primers for sequencing. Right, precise insertion of tags mediated by GRAND editing with duo epegRNAs. Paired pegRNAs were designed with partial (40 bp) or full (72 bp) 3XFLAG sequences as complementary regions (red) in RTTs.

(B) Design of tags knocked in to the 3′ terminus of OsUBQ10 and OsTubA1 by prime editing. The stop codons of the genes are labeled in red. The protospacer of epegRNAs is underlined. Arrows indicate the in-frame fusion sites of tags.

(C–F) Detection of targeted insertions of 6XHis (C and E) or HA (D and F) at the OsUBQ10 (C and D) or OsTubA1 (E and F) site in independent T₀ lines by PCR. For each type of edit, 11 out of 48 examined plants are presented. The genomic DNA of untransformed wild-type rice seedlings was amplified as a negative control. The insertions were identified with F1+R primers. PCR of the Cas9 region of the enpPE2 vector was used as an internal control.

(G–J) Sanger sequencing chromatograms of in-frame 6XHis (G and I) or HA (H and J) insertion at the OsUBQ10 (G and H) or OsTubA1 (I and J) site in representative T₀ lines. Clones of the PCR product of the targeted regions were sequenced. A representative clone carrying the accurate insertion is indicated. The inserted tag sequence is shadowed in gray.

(K) Efficiency of 6XHis or HA tag knockin in stably transformed callus cells with the enpPE2 and ePE2 systems. The ratios of reads carrying the desired edits and reads carrying unintended mutations or incomplete insertions to total clean reads were defined as the accurate insertion efficiency and byproduct efficiency, respectively. Data are presented as the mean value and standard deviation of three biological replicates from independent transformations. Statistical differences in accurate insertions were determined by two-tailed t-tests. ∗p < 0.05.

(L) Sanger sequencing chromatograms of in-frame 3XFLAG insertions in representative T₀ lines.

(M–O) Targeted knockin of 3XFLAG by GRAND editing at the OsUBQ10 (M), OsTubA1 (N), and OsACT2 (O) sites in transgenic rice. Illustrations of the deletion/insertion sites of GRAND editing are shown on the left. Right, Sanger sequencing chromatograms of representative clones confirmed 3XFLAG insertion mediated by GRAND editing in identified T₀ lines.

(P) Western blotting confirmed the presence of HA-tagged proteins and 3XFLAG proteins in representative T₀ transgenic lines. Total proteins from untransformed rice seedlings (wild type) were used as negative controls.

To optimize the editing efficiency of enpPE2, the RNase H domain of Moloney-murine leukemia virus reverse transcriptase (M-MLV RT) was removed, and a a viral nucleocapsid (NC) protein was fused between the Cas9 nickase and the M-MLV RT variant following the methods described in a recent report (Figure 1A and Supplemental Figure 1) (Zong et al., 2022). The resulting engineered enpPE2 (ePE2) was initially tested for editing of small insertions or deletions and 1–2-bp nucleotide substitutions in the rice genome. Amplicon next-generation sequencing of stably transformed calli showed that, compared with those of enpPE2, the efficiencies of ePE2 were significantly increased, ranging from 1.36- to 1.72-fold at five out of seven sites (p < 0.05, t-test; Supplemental Figure 2). Independent T₀ lines were produced to further evaluate editing efficiencies. The average frequency of edited plants at the seven targets improved from 52.08% with enpPE2 to 70.54% with ePE2 (Supplemental Table 2). Strikingly, ePE2 increased the average frequency of homozygous mutants to 32.74% from the 15.48% generated by enpPE2. We next evaluated the tag insertion efficiency of ePE2 in rice calli. As expected, ePE2 displayed a 1.90- and 1.40-fold improvement in efficiency of 6XHis insertion compared with enpPE2 at the OsUBQ10 and OsTubA1 sites, respectively (Figure 1K). For the HA tag, ePE2 resulted in an efficiency enhancement of 1.62-fold compared with that of enpPE2 at the OsTubA1 site and showed slightly higher, or comparable, knockin activity at the OsUBQ10 site. The tag vectors of ePE2 were retransformed to obtain transgenic plants. Consistent with results in calli, the frequencies of plants carrying the precise insertion were improved by 1.14- to 1.80-fold with ePE2 compared with enpPE2 (Supplemental Table 1). Importantly, an average of 15.63% and 5.21% of transgenic plants harbored homozygous 6XHis and HA tag insertions, respectively. Together, these results suggest that optimized ePE2 is an effective tool to precisely install tags for labeling plant genes.

Aside from the expected edit, prime editing may also produce undesired byproducts (Anzalone et al., 2019). For the enpPE2 and ePE2 systems, next-generation sequencing of stably transformed calli showed that the average edit:byproduct ratio for installation of single or multiple nucleotide(s) substitutions, deletions, and insertions (point editing) was 4.89- to 6.30-fold that of tag knockin (tag editing) (Supplemental Figure 3). Differences in the proportions of unintended products were also revealed in ePE2 transgenic lines. In point editing, all the edited plants were accurately mutated at two out of seven targets, and less than 8.33% of plants carried byproducts at the remaining four sites (Supplemental Table 2; Supplemental Figure 4). Although byproducts were induced in up to 47.92% of plants at the Ehd1-T site, all the plants harbored an unintended insertion of a single nucleotide derived from the pegRNA scaffold (Supplemental Figure 5), which might be prevented by avoiding specific nucleotides at the adjacent site when selecting RTT regions (Jiang et al., 2020, 2022). By contrast, an average of 39.06% of plants subjected to ePE2-mediated tag editing had byproducts (Supplemental Table 1; Supplemental Figure 4). These byproducts were typically inaccurate tag sequences with a series of 3′ truncations and/or oligonucleotide insertions (Supplemental Figure 6). Intriguingly, we noted that the unexpected insertion was derived from the sequence flanking the nicked site of prime editing. The unintended truncations and “duplications” implied the potential involvement of 3′-to-5′ removal of oligonucleotides by exonucleases and the mislocated 5′-to-3′ templated addition of oligonucleotides by the DNA polymerase of the mismatch repair system (Li, 2008). Therefore, it would be interesting to examine whether plant PEs optimized with mismatch repair manipulation produce more accurate tag editing. Some tagged plants with heterozygous and chimeric insertions were selected to produce seeds by strict self-pollination (Supplemental Table 3). Precisely tagged T₁ progenies were found in the examined T₀ lines, some of which were free of T-DNA. These results indicate that PE-mediated tag insertions are heritable in rice.

Encouraged by the reliable knockin frequency of ePE2 for 18- and 27-nt tags, we examined the insertion of a 3XFLAG sequence (66 nt) into the OsUBQ10 and OsTubA1 sites as well as a site 15 bp upstream of the OsACT2 stop codon (Supplemental Figure 7). Sanger sequencing of independent T₀ plants revealed precise knockin of 3XFLAG in two out of 48 lines (2/48) at the OsUBQ10 site and one out of 48 lines at the OsTubA1 site, but 3XFLAG was not detected at the OsACT2 site (Figure 1L; Supplemental Table 4). To improve the efficiency of targeted insertion, a GRAND (genome editing by RTTs partially aligned to each other but nonhomologous to target sequences within duo pegRNA) editing strategy (Wang et al., 2022) was used to simultaneously delete 57-, 90-, and 186-bp 3′ terminal and UTR sequences and to insert a 72-bp tag sequence (66-nt 3XFLAG plus two copies of a stop codon) at the OsUBQ10, OsTubA1, and OsACT2 sites, respectively (Figures 1A and 1M–1O). For “tag replacement” assays, dual epegRNA pairs containing 40-bp complementary sequences in RTTs were designed for GRAND editing (Figure 1A). Accurate edits were observed at the three targets in 8.33%–25% of independent T₀ lines (Figures 1M–1O; Supplemental Table 4), indicating that compared with single epegRNA-mediated prime editing, GRAND editing improved knockin frequency in plants. Next, the full 72-bp insertion sequence was designed as a complementary RTT for GRAND editing (Figure 1A). The average proportion of plants with precise edits decreased to 2.78% compared with 14.58% with 40-bp complementary RTTs (Supplemental Table 4), consistent with a previous report showing that longer complementary RTTs produced lower efficiencies (Wang et al., 2022). Substantial numbers of byproducts were found in GRAND-edited plants (Supplemental Table 4; Supplemental Figure 8). Most byproducts were knockin sequences with truncated ends or unintended mutations (Supplemental Figure 9). It has been suggested that avoiding microhomology of RTTs and target sites could increase the ratio of accurate editing. Thus, carefully designed RTTs with synonymous mutations could enable the efficient insertion of longer tag sequences.

In this study, three extensively used epitope tags were precisely inserted into the 3′ termini of targeted genes by the enpPE2 or optimized ePE2 system in rice plants. Our data suggested that longer tags, such as 3XFLAG, could be efficiently inserted by GRAND editing. Protein tagging with HA and 3XFLAG was confirmed in the edited plants by western blotting (Figure 1P). Collectively, our results provide a simple, easy-to-use, and efficient protein tagging strategy for plant research. A twin prime-editing method, which is similar to GRAND editing, has been reported to enable deletions, substitutions, and insertions of large fragments with high frequency (Anzalone et al., 2022). By combining prime editing-mediated efficient insertion of the attB sequence and the site-specific recombinase Bxb1 system, gene-sized DNA sequences can be precisely integrated into the mammalian genome (Anzalone et al., 2022). Recently, various large serine recombinases with much higher recombination efficiency in human cells than Bxb1 were identified (Durrant et al., 2022). Therefore, knockin of tags longer than 3XFLAG, such as fluorescent proteins, could be expected by integrating the ePE2 and large serine recombinase systems in the near future. Furthermore, editing used for protein tagging can enable the precise insertion of cis-elements, enhancers, and other engineering elements for crop breeding.

Funding

This work was funded by the National Key Research and Development Program (2022YFF1002803); the National Natural Science Foundation of China (U19A2022, 32270430, and 32000284); the Natural Science Foundation of Anhui Province (2108085Y07, 2208085Y11, and 2008085MC71); the Innovative Research Team of Anhui Education (2022AH010056); the Science and Technology Major Projects of Anhui Province (2021d06050002); and the Improved Varieties Joint Research (Rice) Project of Anhui Province (the 14th 5-year plan).

Author contributions

P.W. designed the experiments. P.W., R.Q., and M.L. supervised the project. P.W. and J.L. wrote the manuscript with input from all the authors. J.L., J.D., J.Z., R.X., and C.Q. performed vector construction and genotyping. D.G., X.L., J.L., and H.W. performed rice transformation and sampling. J.L., J.D., and J.Z. analyzed the data.

Published: March 7, 2023