Increasing fidelity and efficiency by modifying cytidine base-editing systems in rice

Ruiying Qin, Shengxiang Liao, Juan Li, Hao Li, Xiaoshuang Liu, Jianbo Yang *, Pengcheng Wei *

Key Laboratory of Rice Genetics & Breeding of Anhui Province, Institute of Rice Research, Anhui Academy of Agricultural Science, Hefei 230031, Anhui, China

Keywords:CRISPR-Cas9 Base editing BE3 CDA Oryza sativa

ABSTRACT The efficiency of plant cytidine base-editing systems is limited, and unwanted mutations frequently occur in transgenic plants. We increased the cytidine editing frequency and fidelity of the plant base editor 3(BE3)and targeted activation-induced cytidine deaminase(CDA) (target-AID) systems by coexpressing three copies of free uracil-DNA glycosylase(UDG)inhibitor(UGI).The editing efficiency of the improved BE3 and CDA systems reached as high as 88.9% and 85.7%, respectively, in regenerated rice plants, with a very low frequency of unwanted mutations. The low editing frequency of the BE3 system in the GC context could be overcome by the modified CDA system. These results provide a highfidelity and high-efficiency solution for rice genomic base editing.

1. Introduction

Precise genome modification is highly desired in crop breeding and in plant genomic research. Considerable efforts have been made to establish gene targeting systems in plants; however, the efficiency of these systems is limited in higher plants, especially crops, mainly because of the very low frequency of homologous recombination(HR). Recently developed base-editing systems permit direct nucleotide substitution at target genomic loci without requiring DNA double-strand breaks and homologydirected repair (HDR). C?G to T?A base editors such as BE3,are normally composed of an APOBEC1/AID member of the cytosine deaminase family,a catalytically defective Cas9 or Cpf1, and a uracil-DNA glycosylase (UDG) inhibitor (UGI)[1,2], while A?T to G?C editors (adenine base editors, ABEs)employ a fusion protein composed of a laboratory-evolved tRNA adenine deaminase (TadA) and a SpCas9 or SaCas9 nickase variant [3,4]. In mammals, the efficiency and precision of base editors could be enhanced by optimizing nuclear localization signals (NLSs) and codon usage, engineering nCas9 fusion proteins, and coexpressing Mu Gam protein or free UGI [5-7].In plants,three types of C?G to T?A base editors are most commonly used: BE3, which uses rat APOBEC1 for C-to-T conversion [8-12]; a target-AID system using Petromyzon marinus CDA1 (PmCDA) [13]; and rBE5,which uses a mutated variant of human AID [14]. These systems have successfully induced targeted base conversion in different plant species,including rice,wheat,maize,Arabidopsis, tomato, and watermelon [8,12,15]. However,the base-editing frequency induced by these systems is lower than the mutagenesis efficiency of the CRISPR/SpCas9 system in plants. Furthermore, in addition to targeted base conversion, these systems often generate indels and unwanted nucleotide substitutions. These infidelity mutations lead to great inconvenience during the genotyping and phenotyping of transgenic plants,compromising the application of these systems in plants.We set out to optimize the plant BE3 and target-AID systems to achieve cleaner and more efficient base editing of multiple sites in rice.

2. Materials and methods

2.1. Vector construction

Fig.1-Base editing in transgenic rice plants using BE3 variants.A.Expression cassette of BE3 and eBE3 in binary vectors.PUBI,promoter of maize ubiquitin 1;35S-ter,terminator of the CaMV 35S promoter.B.Base-editing efficiencies of different BE3 and eBE3 vectors in regenerated populations.The number of lines carrying targeted base substitutions(which may also carry alleles with unwanted mutations)was used to calculate the substitution efficiency(left),while the number of regenerated plants with an exclusive mutation type of targeted C-to-T conversions(and no unwanted mutations)was used to calculate the clean editing yield efficiency; C.Frequencies of unwanted mutations in BE3/eBE3 plants.The regenerated lines carrying unwanted mutations,including InDels,a base conversion of C-to-A or C-to-G(Non-T),or both in a single line(Non-T + InDels)are indicated separately. D.Frequencies of targeted C-to-T conversions in regenerated populations treated with BE3/eBE3 vectors at the indicated position of the sgRNA target region.

Fig.2- Base editing in BE3 lines using native sgRNA and esgRNA.A.Base-editing efficiencies of the BE3-sgRNA and BE3-esgRNA vectors in the regenerated population.B. Frequencies of unwanted mutations in BE3-sgRNA and BE3-esgRNA plants.C.Frequencies of targeted C-to-T conversions in the BE3-sgRNA and BE3-esgRNA regenerated populations at the indicated position of the sgRNA target region.

The previously described sequences of rice codon-optimized SpCas9, APO-XTEN, and UGI-NLS were synthesized separately [11,16] (GENEWIZ, Suzhou, China). The D10A mutation was induced in SpCas9 by PCR. The fragments were seamlessly assembled using a HiFi DNA Assembly Cloning Kit (NEB, Beverly, MA, USA) to produce BE3. To construct eBE3, the sequence of UGI-3X2A-UGI (eGUI) was codon-optimized for rice and synthesized [6]. Te eGUI was then attached to the 3′ terminus of the APO-XTEN-nSpCas9 fragment by seamless cloning. To construct CDA, the Arabidopsis codon-optimized sequence of SH3-FLAGPmCDA [13] was reoptimized for rice expression, and the synthesized fragment was linked to nSpCas9 to generate CDA.The 3X2A-UGI sequence was then amplified from eGUI and directly added to the 3′end of CDA to construct eCDA.The native sgRNA and esgRNA were synthesized and placed into an OsU3-derived expression cassette with a spectinomycin selection marker [17]. The base editors and OsU3-SpR-sgRNA or OsU3-SpR-esgRNA cassette were cloned into the pHUN411 backbone to generate binary vectors. To construct the genomic editing constructs, protospacers were annealed and inserted to replace SpR in the vector. Following the protocol of the pHUN system,the clones were positively selected with kanamycin and negatively selected with spectinomycin. All primers used in this study are listed in Table S1.

2.2. Rice transformation

The vector was individually introduced into the Agrobacterium strain EHA105-pSoup. Mature seeds of the japonica rice cultivar Nipponbare were used for callus induction for three weeks.The embryonic calli were then transfected by Agrobacterium following a previously described procedure [18]. Resistant calli were selected with 50 μg mL?1hygromycin for four weeks.Transgenic plants were then regenerated under selection with 25 μg mL?1hygromycin. Only one plant from each transfected callus (a single event)was selected for rooting and further examination.

2.3. DNA extraction and genotyping

In each transgenic plant, one leaf from each tiller was collected. Genomic DNA samples were prepared by the CTAB method and diluted to the same concentration with a NanoDrop spectrophotometer (Thermo Fisher, Waltham,MA, USA). For Hi-TOM detection, specific primers containing barcodes were used to amplify the target region.Following the manufacturer's instructions, the products of two rounds of PCR were pooled and sequenced with the Illumina HiSeq platform (Novogene, Tianjin, China). Approximately 1 Gb of raw data was generated for each of 96 samples. Mutations were identified with Hi-TOM(http://www.hi-tom.net/hi-tom).To confirm the next-generation sequencing(NGS)genotyping,at least 16 samples for each target were randomly selected,amplified and Sanger sequenced. All results from the firstgeneration sequencing perfectly matched the genotypes identified by the Hi-TOM method. The efficiencies of vectors were compared with Fisher's exact test.The NGS data used for the Hi-TOM analysis can be accessed under NCBI BioProject number PRJNA481881.

3.Results

Recombinant BE3 was first generated by fusion of plant codon-optimized rat APOBEC and UGI to the N-and C-termini of the SpCas9 nickase (D10A). Then, enhanced BE3 (eBE3) was developed by addition of triplet copies of 2A-UGI to the 3′ end of the BE3 sequence (Fig. S1) to coexpress an additional three free copies of UGI (Fig. 1-A).The BE3 and eBE3 genes separately replaced SpCas9 in a previously modified pHUN411 binary vector[19,20], resulting in pHUN411-BE3 and pHUN411-eBE3, respectively. The base editing of the two vectors was tested at five different genomic sites (ALS-T1, CHL9, IPA, NRT1.1, and SLR1) in transgenic rice plants using Agrobacterium-mediated stable transformation (Fig. S2). All independent regenerated lines (24 to 48 lines for each vector) were genotyped by Hi-TOM [21], a NGS method.In agreement with findings of previous reports in plants, the efficiencies of BE3-mediated targeted base conversions varied from 25.0% to 58.8% (Fig. 1-B). Base targeting was not statistically increased by eBE3 at most of the tested genomic sites (Fisher's exact test, P ＜0.05), except in the SLR1 target region (achieving 70.8% efficiency compared to 25.0% by BE3).However, the editing fidelity was different in the BE3 or eBE3 regenerated populations with the same target sgRNA. As indicated in Fig. 1-C, the indel frequency in BE3 plants was 12.5%-25.0%, possibly as a consequence of the base excision repair of apurinic/apyrimidinic (AP) sites transformed from APOBEC1-converted U by UDG. In contrast, no indels were detected in plants carrying the eBE3 construct(Fig.1-C).AP sites could also lead to undesired C-to-A or C-to-G conversion instead of C-to-T substitution. The eBE3 with increased UGI expression showed a substantially lower unwanted base conversion frequency (0-3.5%) than the BE3 vectors (2.5%-14.7%) (Fig. 1-C).At all five sites, the percentage of clean editing yield (ratio of lines carrying only the C-to-T substitution to total lines) in the regenerated population was 1.14- to 3.81-fold higher for eBE3 than for BE3(Fig.1-B).

Fig.3-Base editing in transgenic rice plants using CDA variants.A.Expression cassette of CDA and eCDA in binary vectors.B.Base-editing efficiencies of different CDA and eCDA vectors in regenerated populations.C.Frequencies of unwanted mutations in CDA/eCDA plants.D.Frequencies of C?G to T?A conversions in CDA/eCDA regenerated populations at the indicated position of the sgRNA target region.

Previous reports [17,22,23] indicated that optimization of the sgRNA sequence (esgRNA) with a mutated potential terminator sequence and extended duplex length can improve the efficiency of SpCas9 and its variants(Fig.S3).To test whether optimized sgRNA could increase the frequency of the plant base editors, the esgRNA scaffold sequences were synthesized and individually fused with each of the five above-described protospacers in the eBE3 vector. The BE3-esgRNA combination was also tested with the CHL9,IPA1,and SLR1 targets as controls (Fig. 2). In plants treated with the eBE3-esgRNA vectors, base editing was highly effective(varying from 53.5%to 88.9%)(Fig.1-B).The editing frequency of esgRNA was 1.9- and 2.1-fold higher than that of native sgRNA at the ALS-T1 and IPA1 targets,respectively.Moreover,eBE3-esgRNA still produced a much lower frequency of unwanted mutations (0-7.0%) than BE3-esgRNA (25.0%-50.0%). Producing a perfectly matched 20 bp guide sequence using the tRNA-sgRNA expression system is another strategy to enhance CRISPR editing efficiency [24,25]. Because the rice U3 promoter in the vectors attaches an additional adenine to the 5′end of the mature sgRNA,the 20-bp spacers of the ALST1, IPA1, and NRT1.1 targets, which did not start with “A”,were individually cloned downstream of the tRNA to generate a precisely matched guide sequence.The clean editing yield in the regenerated populations varied from 48.4% to 63.2% (Fig.1-B), and the unwanted mutation rate varied from 0 to 3.1%(Fig. 1-C). These results suggested that eBE3 in combination with esgRNA or tRNA-sgRNA can efficiently generate baseedited plants without unwanted mutations. Consistent with the BE3 system, the most effective editing window of eBE3 tools is still positions 4 to 8(Fig.1-D).

Fig.4- Base editing in CDA lines using native sgRNA and esgRNA.A.Base-editing efficiencies of the CDA-sgRNA and CDAesgRNA vectors in the regenerated populations.B.Frequencies of unwanted mutations in CDA-sgRNA and CDA-esgRNA plants.C.Frequencies of targeted C-to-T conversions in CDA-sgRNA and CDA-esgRNA regenerated populations at the indicated position of the sgRNA target region.

Target-AID showed limited editing efficiency and frequently showed undesired mutations, restricting its use for generating precise gain-of-mutations [14]. To improve the efficiency of this system in crops, the coding sequence of PmCDA1 (Os-PmCDA) was specifically codon-optimized for rice expression. Os-PmCDA was linked to the 3′ terminus of nSpCas9 (D10A) with or without the addition of the 3X2A-UGI fraction to form the CDA or eCDA gene. Similarly to the method described above for BE3, SpCas9 in the binary vector pHUN411 was replaced by CDA or eCDA, leading to pHUN411-CDA or pHUN411-eCDA, respectively (Fig. 3-A). The editing effect of CDA was first tested at the ALS-T2 target (Fig. S4).Targeted mutations occurred in 19.4% and 25.0% of regenerated CDA plants with native sgRNA and esgRNA(Figs.3-B, 4),respectively.However,many mutated lines carried InDels and unwanted base conversions (19.4%-30.0% of total lines), and the clean editing yield was lower than 5.6% (Fig. 3-B). As expected, eCDA provided a much higher clean editing yield(18.6% and 18.8% using native sgRNA and esgRNA,respectively) and a much lower unwanted mutation rate(3.1% and 9.3%) than CDA (Fig. 3-B). The combination of tRNA-sgRNA and eCDA showed much higher editing efficiency. Of the 40 regenerated lines, 30 lines were mutants that carried only a targeted C-to-T substitution(s)(representing 75.0% efficiency), including seven biallelic mutants. Moreover, no undesired mutations occurred (Fig.3-C).Interestingly,efficient editing occurred even at position 7, suggesting that eCDA might have a wider editing window than the previously reported 3-bp highly-mutated region(positions 2 to 4) of CDA (Fig. 3-D) [13]. The members of the APOBEC/AID family have different sensitivities to the 5′adjacent nucleotide of target C. The BE3 system has limited GC editing efficiency. We confirmed that the editing frequency of the position 4 C in the GC context of the NRT1.1 target was substantially lower than that of the C at the same position in the CC context of the CHL9 and SLR1 targets (Fig.1-D). We accordingly tested the GC preferences of the CDAs.A 20-bp Pi-d2 region with a G located at position 4 in the GC context (reverse complement strand) was selected for CDA/eCDA-mediated base editing. The clean position 4 G-to-A conversion occurred in 24 lines of 28 tRNA-eCDA plants(85.7% efficiency) (Fig. 3-B) and 17 of these lines were identified as having biallelic or homozygous mutations.These findings implied that, similar to the rBE5 system [14],the CDA/eCDA tool would be more useful in targeted editing of plant genomes with a high GC context.

4. Discussion

Limited editing fidelity and efficiency have greatly compromised the application of base editing in the plant genome. The absent or insufficient UGI activity of the target-AID or BE3 system normally leads to a high frequency of undesired mutations in addition to the targeted C-to-T conversions. The unwanted mutations,especially InDels, interfere with the sequence determination of base conversions in plants. Although some unwanted mutations can segregate in the progeny, robust production of clean base conversion in the T0generation will greatly reduce time and economic costs. Although increasing the fidelity of base editing by adding UGI has been reported in mammalian cells [5,6,26], it had not been tested in the plant genome. In this study, we developed plant base-editing systems with additional UGI activity.Combined with the optimized sgRNA expression cassette,our modified eBE3 and eCDA tools showed as high as 86.1%and 85.7% clean editing efficiency (Fig. 1-B), respectively,4.9- and 6.9-fold of the corresponding BE3 and CDA vectors with the conventional structure. In addition, the eBE3 and eCDA tools have different editing scope and target preferences. The improved base-editing toolkits demonstrated in this study will expand the scope of targeted single-base substitutions in the rice genome. We propose that these high-fidelity and high-efficiency base-editing tools, together with the previous reported base substitution systems in plants, can accelerate the application of precise mutagenesis to plant fundamental research and trait improvement.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2019.04.007.

Declaration of Competing Interest

Authors declare that there are no conflicts of interest.

Acknowledgments

This work was funded by the Genetically Modified Breeding Major Project (2016ZX08010-002-008), the National Natural Science Foundation of China (31701405), and the Natural Science Foundation of Anhui Province,China(1708085QC60).