Identification of optimal adenine and cytosine base editors for genome editing in Arabidopsis and soybean

Abstract

Base editors, including adenine base editors (ABEs) and cytosine base editors (CBEs), are widely used in numerous organisms to introduce site-specific sequence modifications in genomic DNA without causing double-strand breaks (DSBs). However, these editors exhibit low editing efficiencies, particularly in dicot plants, thereby limiting their application in dicot plant genome engineering. In this study, we assessed the editing efficiencies of various base editors to identify those optimal for base editing in dicot plants. We discovered that ABE8e, an ABE variant, demonstrated superior A-to-G base editing efficiency within A5-A8 windows, and A3A/Y130F-V04, a CBE variant, exhibited the highest C-to-T base editing efficiency within C4-C15 windows in both Arabidopsis and soybean protoplasts. Overall, we recommend these two base editors as prime choices for efficient genome engineering in a range of crop plants.

INTRODUCTION

Clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 genome editing tools are extensively used to generate site-specific mutations. Utilizing this tool involves introducing double-stranded DNA breaks (DSBs) into the target DNA sequence, repaired subsequently by non-homologous end joining (1). Despite this tool’s efficacy in introducing site-specific mutations, CRISPR-induced DSBs can lead to undesired outcomes such as large chromosomal deletions and translocations (2). Accordingly, Cas9-derived base editors, which comprise a Cas9 nickase fused to a base deaminase, have been designed to enable precise nucleotide changes at target genomic sites without generating DSBs or requiring donor DNA templates (3).

Two major types of base editors have been developed: adenine base editors (ABEs) that convert adenine to guanine (A-to-G) and cytosine base editors (CBEs) that convert cytosine to thymine (C-to-T) (3, 4). All currently available ABEs originate from the prototype ABE7.10, consisting of a Cas9 nickase fused with two adenosine deaminases: a wild-type tRNA-specific adenosine deaminase (TadA) from Escherichia coli [wtTadA] and an engineered version, TadA7.10 (4). To enhance A-to-G editing efficiency, iterations of the TadA7.10 protein have produced variants like TadA8.17 (V82S/Q154R) (5), TadA8.20 (I76Y/V82S/Y123H/Y147R/Q154R) (5), and TadA8e (A109S/T111R/D119N/H122N/Y147D/F149Y/T166I/D167N) (6). Additionally, TadA8e was further modified into TadA8eWQ (V106W/D108Q) (7) and TadA9 (V82S/Q154R) (8), which demonstrate higher editing efficiencies with reduced unintended cytosine editing and RNA deamination. TadA7.10 and its derivatives have been successfully utilized for base editing in various plant species (8, 9).

Cytidine deaminase orthologs, which exhibit varying editing efficiencies and windows, are widely distributed across living organisms. These include rat APOBEC1 (3), human APOBEC3A (A3A) (10), A3A/Y130F (11), human AID (12), Petromyzon marinus cytidine deaminase 1 (PmCDA1) (13), and rAPOBEC1/W90Y+R126E (YE1) (14). Many of these have been utilized for C-to-T base editing in various plant species (15-17). Despite the widespread use of DNA base editors, their efficiency in dicot plants remains low, highlighting the urgent need for systematic evaluations to identify the most effective base editors for future use in dicot plants.

RESULTS AND DISCUSSION

Here, we utilized a streamlined protoplast transient expression system with Arabidopsis and soybean mesophyll protoplasts to efficiently screen and evaluate the editing efficiencies of two major types of base editors (ABEs and CBEs). We employed the cauliflower mosaic virus (CaMV) 35S promoter to express Arabidopsis codon-optimized ABE7.10, ABE8.17, ABE8e, ABE8eWQ, and ABE9, and the AtU6 promoter to express guide RNAs (gRNAs) (Fig. 1A). For Arabidopsis genome editing, three target genes (AtGL2, AtPDS3, and AtSKL1) were selected to assess A-to-G base editing efficiency (Fig. 1B). Transient expression of each construct in Arabidopsis protoplasts was followed by high-throughput sequencing analysis of genomic DNA isolated from the transfected protoplasts, revealing that ABE8e had higher A-to-G base editing efficiency at all target sites compared to other ABE variants, although the editing efficiency varied depending on the sequence context (Fig. 1C). Additionally, we investigated the editing window and the purity of products of A-to-G editing by ABEs. Results indicated that the editing window was roughly similar for all five ABEs examined (Fig. 1D), with mostly A-to-G editing and negligible A-to-C and A-to-T editing in the preferred editing window (Supplementary Fig. 1).

Fig. 1.

Adenine base editing by five ABEs in protoplasts of A. thaliana and G. max. (A) The architectures of nCas9 (D10A)-fused adenosine deaminases used for A-to-G editing in protoplasts of A. thaliana and G. max are illustrated. ABE, adenine base editor; nCas9, Cas9 nickase; Ade, adenosine deaminase; PAM, protospacer adjacent motif; gRNA, guide RNA; wtTadA, wild-type tRNA-specific adenosine deaminase; NLS, nuclear localization sequence. (B) Five target sites (AtGL2, AtPDS3, AtSKL1, GmFT2a, and GmFT4) were selected to evaluate the editing efficiencies of various ABEs. Adenines within the protospacer target sequence are highlighted in red. (C) Evaluation of five ABEs at three different target sites in A. thaliana protoplasts. (D) The editing windows of five ABEs across three different target sites in A. thaliana protoplasts. (E) Adenine base editing efficiencies of five ABEs at two independent target sites in G. max protoplasts are displayed. (F) A-to-G editing windows of five ABEs within two target sites in G. max protoplasts are presented. In (C, E), each dot represents a biological replicate. In (C to F), data represent means ± SEM from three independent biological replicates. Different letters indicate significant differences at P < 0.05 (one-way ANOVA with Tukey’s post hoc test).

We also tested the A-to-G editing efficiency of each base editor in soybean protoplasts, using the same constructs as for Arabidopsis, except for the gRNAs. The gRNAs target the GmFT2a or GmFT4 gene (18). Among the five ABEs, ABE8e demonstrated the highest A-to-G editing at the target sites as in Arabidopsis protoplasts (Fig. 1E), and ABE9 also exhibited an editing efficiency comparable to that of ABE8e at some target sites (Fig. 1E). The editing windows and product purity of ABEs in soybean protoplasts appeared similar to those in Arabidopsis protoplasts (Fig. 1F, Supplementary Fig. 2). Notably, while several ABEs have been reported to exhibit target-dependent off-targeting efficiency (19), no significant off-target editing by any of the ABEs used in our study was detected (Supplementary Fig. 3). Overall, these results indicate that ABE8e is the most efficient ABE within the editing window (protospacer positions 5-8) for dicot plants.

Next, we assessed the C-to-T editing efficiency of five CBE variants, including rAPOBEC1-V01, A3A/Y130F-V01, hAID-V01, PmCDA1-V01, and YE1-V01 (14, 16). To generate recombinant plasmids for Arabidopsis base editing, each base editor was controlled by the CaMV 35S promoter and paired with AtU6-driven gRNAs targeting AtGL2, AtSKL1, and AtVAR2 (Fig. 2A, B). In transient expression analyses using Arabidopsis protoplasts, the A3A/Y130F-V01 base editor demonstrated the highest C-to-T editing efficiency across all assessed sites (Fig. 2C). Editing window analysis revealed that the five CBEs had differing ranges for C-to-T editing, with A3A/Y130F showing the largest window (protospacer positions 5-16) among the tested CBEs (Fig. 2D). An analysis of the product purity for A3A/Y130F-V01 indicated that C-to-T editing predominantly occurred within the specified editing windows (Supplementary Fig. 4). We concluded that, although base editing efficiency varied slightly with sequence context, A3A/Y130F-V01 could be a primary choice for cytosine base editing in dicot plants.

Fig. 2.

Cytosine base editing by five CBEs in protoplasts of A. thaliana and G. max. (A) The architecture of nCas9 (D10A)-fused cytidine deaminases utilized for C-to-T editing in protoplasts of A. thaliana and G. max is outlined. CBE, cytosine base editor; UGI, uracil DNA glycosylase inhibitor; nCas9, Cas9 nickase; Cde, cytidine deaminase; PAM, protospacer adjacent motif; gRNA, guide RNA; rAPOBEC1, rat APOBEC1; A3A, APOBEC3A; hAID, human AID; PmCDA1, Petromyzon marinus cytidine deaminase 1; YE1, rAPOBEC1/W90Y+R126E; NLS, nuclear localization sequence. (B) Five target sites (AtGL2, AtSKL1, AtVAR2, GmFT2a, and GmFT4) served to evaluate the editing efficiencies of various CBEs. Cytosines within the protospacer target sequence are highlighted in red. (C) Evaluations of five CBEs at three distinct target sites in A. thaliana protoplasts are documented. (D) The editing windows for five CBEs at three different target sites in A. thaliana protoplasts are delineated. (E) Cytosine base editing efficiencies of five CBEs at two independent target sites in G. max protoplasts are shown. (F) C-to-T editing windows of five CBEs within two target sites in G. max protoplasts are depicted. In (C, E), each dot signifies a biological replicate. In (C to F), the data are presented as means ± SEM. Distinct letters indicate a significant difference at P < 0.05 (one-way ANOVA with Tukey’s post hoc test).

We also examined C-to-T editing efficiency in soybean protoplasts by expressing each base editor along with a gRNA targeting the GmFT2a or GmFT4 locus (Fig. 2B). Similar to the base editing in Arabidopsis protoplasts, A3A/Y130F exhibited the highest C-to-T editing efficiency at the targeted sites among the five tested CBEs (Fig. 2E). The editing window and product purity of C-to-T editing were also comparable to those observed in Arabidopsis protoplasts (Fig. 2F, Supplementary Fig. 5). These results suggest that A3A/Y130F-V01 is the optimal cytosine base editor within protospacer positions 4-15 in dicot plants.

Uracil DNA glycosylase inhibitor (UGI) is used alongside cytidine deaminases for efficient base editing (3). The MS2-MS2 coat protein (MCP) interaction-based CBE-V04 system enhances C-to-T editing efficiency by recruiting four additional MCP-UGI copies to a single gRNA scaffold, thereby achieving a high focal concentration of UGIs (16). We also designed the A3A/Y130F-V04 system (Fig. 3A) and observed higher C-to-T editing efficiencies compared to the A3A/Y130F-V01 system at all target sites in both Arabidopsis and soybean protoplasts (Fig. 3B, D), with both systems showing similar editing windows and product purity (Fig. 3C, E, Supplementary Fig. 6, 7) while avoiding off-target mutations (Supplementary Fig. 8).

Fig. 3.

Cytosine base editing by A3A/Y130F-V01 and A3A/Y130F-V04 in protoplasts of A. thaliana and G. max. (A) Structure of the A3A/Y130F-V04 cytosine base editor is illustrated. CBE, cytosine base editor; UGI, uracil DNA glycosylase inhibitor; nCas9, Cas9 nickase; Cde, cytidine deaminase; PAM, protospacer adjacent motif; gRNA, guide RNA; A3A, APOBEC3A; NLS, nuclear localization sequence; T2A, Thosea asigna virus 2A; MCP, MS2 coat protein. (B) C-to-T editing efficiencies of A3A/Y130F-V01 and A3A/Y130F-V04 at three target sites in Arabidopsis protoplasts are presented. (C) Editing windows of A3A/Y130F-V01 and A3A/Y130F-V04 at three target sites in A. thaliana protoplasts are displayed. (D) C-to-T editing efficiencies of A3A/Y130F-V01 and A3A/Y130F-V04 at two independent target sites in G. max protoplasts are shown. (E) Base editing windows of A3A/Y130F-V01 and A3A/Y130F-V04 within two target sites in G. max protoplasts are depicted. In (B, D), each dot signifies a biological replicate. In (B to E), data are rendered as means ± SEM. Distinct letters indicate a significant difference at P < 0.05 (one-way ANOVA with Tukey’s post hoc test).

In conclusion, ABE8e and A3A/Y130F-V04 emerged as the most effective ABE and CBE for A-to-G and C-to-T base editing, respectively, in dicot plants. While other base editors may be optimal in different sequence contexts, our results indicate that these editors should be primary choices for efficient and precise plant genome engineering and crop molecular breeding. We anticipate that these base editors could be further enhanced by fusing Cas9 variants with relaxed PAM specificity to small dsDNA-binding proteins like Sso7d and HMG-D (20-24).

MATERIALS AND METHODS

Plasmid construction

All PCR fragments for plasmid construction were amplified with Primestar GXL DNA polymerase (Takara Bio, R050A), and the amplified fragments were assembled by Gibson assembly. gRNA design and cloning were performed based on previously described protocols (25). All reconstructed plasmids were confirmed by Sanger sequencing.

The ABE variant expression plasmid was constructed based on pBAtC (backbone plasmid) (Addgene #78097) (25) and pJY-RpABE (TadA7.10) (Addgene #112872) (26). The coding sequences of the A. thaliana codon-optimized adenosine deaminase dimer construct and nickase Cas9-NLS fragment in pJY-RpABE were amplified and directly inserted into the backbone of pBAtC by replacing the Cas9-NLS-HA fragment, resulting in p35S::wtTadA-TadA7.10-nCas9 (D10A). To construct ABE8.17 and ABE8e with the TadA variant monomer, the wtTadA fragment was removed from the ABE7.10 backbone by one-fragment Gibson assembly. Mutations were inserted in TadA7.10 sequence by PCR-based site-directed mutagenesis using a primer containing the desired mutations, resulting in ABE8.17 (V82S/Q154R in TadA7.10) and ABE8e (A109S/T111R/D119N/H122N/Y147D/F149Y/T166I/D167N in TadA7.10). ABE8eWQ and ABE9 variants were constructed by site-directed mutagenesis in ABE8e, resulting in ABE8eWQ (V106W/D108Q in TadA8e) and ABE9 (V82S/Q154R in TadA8e).

The CBE variant plasmids were constructed using pBAtC (backbone plasmid) (Addgene #78097) (25), pYPQ265 (rAPOBEC1-CBE-V01) (Addgene #164712) (16), pYPQ265E2 (A3A/Y130F-CBE-V01) (Addgene #164719) (16), pYPQ265C (hAID-CBE-V01) (Addgene #164715) (16), pYPQ266 (PmCDA1-CBE-V01) (Addgene #164713) (16), pYPQ269E2 (A3A/Y130F-CBE-V04) (Addgene #164721) (16), and pYPQ141D2.0 (gRNA2.0 scaffold) (Addgene #99906) (16). The coding regions of cytidine deaminases and maize codon-optimized nCas9 (D10A) fused to UGI-NLS fragments were amplified and inserted into the backbone of pBAtC by replacing the Cas9-NLS-HA fragment, resulting in p35S::rAPOBEC1-NLS-nCas9 (D10A)-NLS-UGI-NLS (rAPOBEC1-V01), p35S::A3A/Y130F-NLS-nCas9 (D10A)-NLS-UGI-NLS (A3A/Y130F-V01), p35S::hAID-NLS-nCas9 (D10A)-NLS-UGI-NLS (hAID-V01), and p35S:: NLS-nCas9 (D10A)-NLS-PmCDA1-NLS-UGI (PmCDA1-V01). The YE1-V01 variant was constructed by site-directed mutagenesis in rAPOBEC1-V01, resulting in YE1-V01 (W90Y/R126E in rAPOBEC1-V01). To construct the A3A/Y130F-V04 expression plasmid, the gRNA 2.0 scaffold was cloned into pBAtC by replacing the gRNA scaffold fragment and A3A/Y130F-CBE-V04 derived from pYPQ269E2 was inserted into pBAtC (gRNA2.0 scaffold version) by replacing the Cas9-NLS-HA fragment, resulting in p35S::A3A/Y130F-NLS-nCas9 (D10A)-NLS-UGI-NLS-T2A-NLS-MCP-UGI- NLS (A3A/Y130F-V04).

Plant materials and growth conditions

The Wassilewskija (Ws-2) accession of A. thaliana and Williams 82 cultivar of G. max were used in all experiments. A. thaliana seeds were surface-sterilized and plated on half-strength Murashige and Skoog (MS) medium containing vitamins supplemented with 1% (w/v) sucrose and 0.8% (w/v) plant agar, and then stratified for 3 days at 4°C in continuous darkness. The G. max plants were grown in plastic pots (100 mm × 85 mm) filled with sterilized horticultural soils. Seedlings were grown under long-day conditions (16-h light/8-h dark cycles) using white fluorescent lamps (120 μE m⁻²s⁻¹) at 23°C.

Plasmid preparation

All plasmids were amplified in E. coli DH5α cells. Single colonies of DH5α carrying different plasmids were picked up from Luria–Bertani (LB) agar plate and grown overnight at 37°C with shaking (250 rpm) in a test tube containing 2 ml of LB medium supplemented with 100 mg/L spectinomycin. Then, 0.5 ml of the culture was transferred to 40 ml fresh Terrific Broth medium supplemented with 100 mg/L spectinomycin in 250 ml baffled flasks and cultured at 37°C for 16 h with shaking (250 rpm). The overnight culture was transferred to a 50 ml tube. The bacterial cells were harvested by centrifugation at 6,500 g for 10 min, and the pellet was resuspended in 4 ml TEG buffer (25 mM Tris-HCl, 10 mM EDTA, and 50 mM glucose, pH 8.0). Then, 8 ml lysis solution (0.2 N NaOH and 1.0% (w/v) SDS) was added, mixed gently by inverting 5-7 times, and incubated at room temperature for 10 min. The lysate was neutralized with 6 ml potassium-acetate solution (3M KOAc and 11.5% (v/v) acetic acid) and incubated at room temperature for 10 min. Genomic DNA, proteins, and cellular debris were removed from the lysate by centrifugation at 14,000 g for 10 min. The supernatant was filtered through two layers of Miracloth (Millipore, 475855), and the filtrate was collected in a new 50 ml tube. The filtrate was mixed with 10 ml isopropyl alcohol and incubated at room temperature for 10 min. The pellet containing nucleic acids was collected by centrifugation at 14,000 g for 10 min, washed with 70% (v/v) ethanol, resuspended in 0.6 ml TE buffer (25 mM Tris-HCl and 10 mM EDTA, pH 8.0), and transferred to a 2 ml tube. For removal of RNA, the nucleic acid suspension was mixed with 0.3 ml of 4.2 M CaCl₂ solution and incubated at room temperature for 5 min. The tube was centrifuged at 12,000 g for 10 min. After transfer of the supernatant to a new 2 ml tube, the supernatant was precipitated with 70 μl 3 M NaOAc and 0.5 ml isopropyl alcohol. After incubation at room temperature for 10 min, the tube was centrifugated at 12,000 g for 10 min. The pellet was washed twice with 70% (v/v) ethanol and resuspended in 0.5 ml TE buffer. For selective precipitation of plasmid DNA, an equal volume of PEG-NaCl solution (20% (w/v) PEG 8000 and 0.5 M NaCl) was added and mixed gently by inverting 5-7 times. Immediately after mixing, the tube was centrifugated at 12,000 g for 10 min and the purified plasmid DNA pellet was washed twice with 70% (v/v) ethanol and resuspended in TE buffer. Plasmid DNA concentrations were determined using BioDrop (Biochrom, BD1776) and adjusted to 2 μg/μl.

Protoplast transfection assays

Protoplast isolation, purification, and transfection were performed as previously described with the following modifications (27). Briefly, 10-day-old seedlings of A. thaliana and 7-day-old unexpanded unifoliate leaves of G. max were soaked in 20 ml 0.5 M mannitol solution (pH 5.8) at 23°C for 1 h. The plasmolyzed seedlings or leaves were incubated for 6-8 h on a rotating shaker (60 rpm) with 20 ml of enzyme solution containing 2% (v/v) Viscozyme L, 1% (v/v) Celluclast 1.5 L, and 1% (v/v) Pectinex ultra SP-L. After filtration through a 40 μm cell strainer, protoplasts were purified using sucrose-density gradient centrifugation, followed by washing twice with MMC solution (0.47 M mannitol, 10 mM MES, and 10 mM CaCl₂, pH 5.8). Purified protoplasts were resuspended in MMG solution (0.4 M mannitol, 15 mM MgCl₂, and 4 mM MES, pH 5.7) at 4 × 10⁶ protoplast/ml for A. thaliana or 1 × 10⁶ protoplast/ml for G. max. Twenty microgram of plasmid DNA was mixed with 300 μl of protoplasts, and an equal volume of 40% PEG solution (40% (w/v) PEG 4000, 0.1 M CaCl_2, 0.2 M mannitol) was added and mixed, followed by incubation for 5 min at room temperature. After incubation, 1 ml of W5 solution (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, 2 mM MES, and 5 mM glucose, pH 5.7) was added and mixed, and protoplasts were collected by centrifugation at 80 g for 5 min. Protoplasts were washed in 1 ml W5 buffer followed by centrifugation at 50 g for 5 min and resuspension in 100 μl PIM medium (Gamborg B5 medium containing vitamins, 20 g/L sucrose, 60 g/L myo-inositol, 2 mg/L 6-BAP, and 0.5 mg/L α-NAA, pH 5.8). Transfected protoplasts were incubated at 23°C in darkness for 3 days. After incubation, protoplasts were collected and genomic DNA was extracted using the cetyltrimethylammonium bromide (CTAB) method for targeted deep sequencing.

High-throughput sequencing

To perform targeted amplicon sequencing, genomic regions containing the on-target and potential off-target sites were PCR amplified with primers, including adapters that are compatible with Illumina index barcodes, and KOD multi & epi DNA polymerase (TOYOBO, KME-101) according to the manufacturer’s instructions. Amplified DNAs were purified using Expin PCR SV mini (GeneAll, 103-102) and purified libraries were then sequenced using the Miniseq sequencing system (Illumina). After sequencing, paired-end data were analyzed using a BE-analyzer (http://www.rgenome.net/be-analyzer/) by comparing BE-treated and non-treated protoplasts. Sequencing data analysis was mostly performed using the computing server at the Genomic Medicine Institute Research Service Center.