==== Front
Natl Sci Rev
Natl Sci Rev
nsr
National Science Review
2095-5138
2053-714X
Oxford University Press

10.1093/nsr/nwad143
nwad143
Research Article
Molecular Biology & Genetics
Nsr/8
AcademicSubjects/MED00010
AcademicSubjects/SCI00010
Programmable deaminase-free base editors for G-to-Y conversion by engineered glycosylase
https://orcid.org/0000-0002-9988-1818
Tong Huawei HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Liu Nana HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Wei Yinghui HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Zhou Yingsi HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Li Yun HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Wu Danni HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Jin Ming Department of Neurology and Institute of Neurology of First Affiliated Hospital, Institute of Neuroscience, and Fujian Key Laboratory of Molecular Neurology, Fujian Medical University, Fuzhou 350004, China

Cui Shuna HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Li Hengbin HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Li Guoling HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Zhou Jingxing HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Yuan Yuan HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Zhang Hainan HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Shi Linyu HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

Yao Xuan HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China

https://orcid.org/0000-0003-3590-722X
Yang Hui HuidaGene Therapeutics Co., Ltd., Shanghai 200131, China
Institute of Neuroscience, Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai 200031, China
Shanghai Center for Brain Science and Brain-Inspired Intelligence, Shanghai 200031, China

Corresponding author. E-mail: huaweitong@huidagene.com
Corresponding author. E-mail: huiyang@huidagene.com
Equally contributed to this work.

8 2023
16 5 2023
16 5 2023
10 8 nwad14324 4 2023
13 5 2023
13 5 2023
03 7 2023
© The Author(s) 2023. Published by Oxford University Press on behalf of China Science Publishing & Media Ltd.
2023
https://creativecommons.org/licenses/by/4.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT

Current DNA base editors contain nuclease and DNA deaminase that enables deamination of cytosine (C) or adenine (A), but no method for guanine (G) or thymine (T) editing is available at present. Here we developed a deaminase-free glycosylase-based guanine base editor (gGBE) with G editing ability, by fusing Cas9 nickase with engineered N-methylpurine DNA glycosylase protein (MPG). By several rounds of MPG mutagenesis via unbiased and rational screening using an intron-split EGFP reporter, we demonstrated that gGBE with engineered MPG could increase G editing efficiency by more than 1500 fold. Furthermore, this gGBE exhibited high base editing efficiency (up to 81.2%) and high G-to-T or G-to-C (i.e. G-to-Y) conversion ratio (up to 0.95) in both cultured human cells and mouse embryos. Thus, we have provided a proof-of-concept of a new base editing approach by endowing the engineered DNA glycosylase the capability to selectively excise a new type of substrate.

glycosylase
base excision repair
base editor
depurination
deaminase-free
transversion
==== Body
pmcINTRODUCTION

Base editing is a powerful technology for basic research and therapeutic applications [1,2]. Current base editors mainly contain programmable DNA-binding proteins, such as a catalytically impaired CRISPR-associated (Cas) nuclease that was fused with a single-stranded DNA deaminase enzyme and sometimes an additional protein that could modulate DNA repair machinery [3,4]. In the past several years, two main classes of DNA base editors, adenine base editors (ABEs) [5] and cytosine base editors (CBEs) [4] have been developed and widely used for A-to-G and C-to-T conversions, respectively (Fig. 1a). Recently, C-to-G base editors (CGBEs) [6–10] and adenine transversion base editor (AYBE) [11] were also constructed by fusing existing CBE or ABE with DNA glycosylase variants to generate new tools for achieving more versatile base editing outcomes, including C-to-G, A-to-C and A-to-T editing (Supplementary Fig. S1). CRISPR-free CBEs (DdCBEs) were reported for performing C-to-T base editing in mitochondria DNA, by fusing two halves of a double-strand DNA cytidine deaminase (DddA) variants with two separate TALE (transcription activator-like effector) proteins [12–14]. So far, these base editing methods all began with deamination of C and A as the key step to produce deoxyuridine (U) and deoxyinosine (I) intermediates, respectively, which in turn were transformed into other bases by endogenous DNA repair or replication mechanisms [4–11]. Although G and T in the non-edited strand could be modified along with the event of editing of C and A in the edited strand, respectively, no existing base editor is capable of directly editing G or T.

Figure 1. Design and mechanisms of base editing by conventional and glycosylase-based base editors. (a) Schematic diagrams of ABE (left) and CBE (middle) and deaminase-free glycosylase-based guanine base editor (gGBE, right). An nCas9-sgRNA complex creates an R-loop at the target site in the DNA. In ABE and CBE, the evolved adenine deaminase (tRNA adenosine deaminase, TadA) and AID/APOBEC-like cytidine deaminase converts the exposed adenine (A) into deoxyinosine (I) and cytosine (C) into deoxyuridine (U), respectively. In CBE, an additional linked protein, uracil glycosylase inhibitor (UGI), protects U from uracil DNA N-glycosylase (UNG). After deamination, the resulting I is recognized as G, and U as T by DNA polymerase during DNA repair or replication. In gGBE, a prototype version of glycosylase-based guanine base editor (gGBE) designed to remove G, and the AP site generated is repaired by TLS and/or DNA replication, leading to G-to-C or G-to-T conversion. PAM, protospacer adjacent motif. AP, apurinic/apyrimidinic sites. (b) A screening reporter system for detecting G-to-T conversion by gGBE. P2A, 2A peptide from porcine teschovirus-1. (c) Percentage of EGFP+ cells for G editing activity evaluation of gGBE with various MPG variants (mean ± s.e.m., n = 3). WT: wild-type.

Base excision proteins in base excision repair (BER) pathways have been successfully harnessed for developing new base editing tools [4–11,15]. The mechanism of BER involves a damage-specific step performed by one of the DNA glycosylases, followed by base lesion events, most frequently spontaneous DNA depurination and deamination [16,17]. Mutagenesis by deamination is unlikely for G (due to spontaneous remediation) [18] and impossible for T (due to the absence of amine), making the development of G and T base editors a challenging task. Depurination of damaged G and A occurs frequently in genomic DNA and generates apurinic/apyrimidinic (AP) sites by DNA glycosylases, initiating BER that readily remediates the loss of these bases [19,20].

There are many types of DNA glycosylases, including N-methylpurine DNA glycosylase (MPG), 8-oxoguanine DNA glycosylase (OGG1) and mutY DNA glycosylase (MUTYH), that could excise various damaged purine bases [19,21–23]. The MPG (also known as alkyladenine DNA glycosylase, AAG) is one of the first responders to DNA base damage and capable of recognizing a wide range of damaged purine bases, including alkylpurines like 3-methyladenine (3meA) and 7-methylguanine (7meG), oxidized adenine 1, N6-ethenoadenine (ϵA) and deaminated adenine hypoxanthine (Hx) [24–29]. Although MPG may also remove normal G at very low rates in vitro [30], overexpression of MPG in transgenic mice did not result in additional point mutations as compared to that found in WT mice, supporting the notion that AP sites due to MPG excision are quickly processed by apurinic/apyrimidinic endonuclease 1 (APE1) into single-strand breaks (SSBs), which could not be mutated by translesion synthesis (TLS) [31]. This implies that depurination-initiated BER could be used for developing novel base editing tools that may prove to be reliable and free of side effects. We speculated that excision of canonical G could be achieved by engineering the MPG protein moiety, leading to efficient base editing through the BER pathway.

Here, we developed a novel tool, namely glycosylase-based guanine base editor (gGBE), to achieve high editing efficiency for G-to-C and G-to-T conversion in mammalian cells. After four rounds of mutagenesis, we obtained marked enhancement of G base editing activity, as compared with that obtained by using wild-type (WT) MPG. We characterized the editing profile of gGBE by targeting dozens of endogenous genomic loci in cultured mammalian cells as well as mouse embryos, demonstrating its high G-to-Y (Y = C or T) base editing efficiency. Thus, DNA glycosylase could be engineered into a new base editing tool.

RESULTS

Development of glycosylase-based base editors

We have attempted to develop a novel base editing system by harnessing the BER pathway triggered by depurination of G or A. By fusing human MPG at the C-terminus of Cas9 D10A nickase (nCas9), we generated a prototype version of deaminase-free glycosylase-based base editor (gBE), in which MPG may remove normal G or A, two purines with similar structure of Hx, to generate AP sites, with nCas9 nicking the opposite non-edited strand. This allowed the damaged DNA to preferentially use the edited strand as a template for DNA repair and/or DNA replication. Therefore, nucleotide incorporation opposite the AP site via TLS would lead to diverse editing outcomes (Fig. 1a). To conveniently evaluate the gBE editing activity, we developed a simple intron-split EGFP reporter system as previously reported [11]. Disruptive point mutations (AG to CG or TG) were introduced in the intron boundary to generate inactive splicing acceptor (SA) signals, and G-to-T or A-to-T conversion (C-to-A or T-to-A conversion in the opposite strand) was required to correct the mutation for proper splicing of the EGFP-coding sequence, thus activating EGFP expression (Fig. 1b and Supplementary Fig. S2a). The EGFP fluorescence intensity could be detected with flow cytometry (Supplementary Fig. S2d). The G-to-T reporter is to be used for evaluating the activity of glycosylase-based guanine base editor (gGBE), and the A-to-T reporter for evaluating the glycosylase-based adenine base editor (gABE). We first constructed gBE vectors with different versions of MPG (MPGv0.2 to MPGv3) that were reported in our previous study [11]. After co-transfecting the G-to-T or A-to-T reporter vector with the gBE vector containing the single-guide RNA (sgRNA) that targets the intronic mis-splicing mutation, we found that all the gBE showed G-to-T but not A-to-T conversion activity (Fig. 1c and Supplementary Fig. S2b). The differences between normal G and A substrate recognition by MPG might account for the failure of A editing in this system. Among them, gBE containing MPGv3 (hereafter referred to as gGBEv3, carrying G163R, N169S, S198A, K202A, G203A, S206A and K210A mutations) exhibited the highest G-to-T editing activity (4.33%) in cultured HEK293T cells (Fig. 1c) as compared to that of gGBE with WT MPG (0.03%), a striking 144-fold enhancement in the editing efficiency. Furthermore, no conversion activity was found for gGBE with dead MPG (dMPG, carrying E125A, Y127A and H136A mutations) or gGBEv3 together with non-targeting sgRNA (Supplementary Fig. S2c). Thus, the G-to-T conversion by gGBEv3 depends on the catalytic G excision domain of MPG and specific sgRNA.

Further enhancement of G editing activity of gGBE

To further increase the G-to-T activity of gGBEv3, we performed protein engineering by several rounds of rational mutagenesis of the MPG moiety, using the G-to-T reporter to evaluate the editing activity (Fig. 2a). Based on structural analysis of MPG [32,33], we selected the 17-aa-long region from R163 to V179 (R163–V179) of MPGv3, which forms a pocket around the targeted G base in a model of MPG-DNA complex (Supplementary Fig. S3a). First, we mutated gGBEv3 at G174R or D175R, two mutations known to improve AYBE activity [11], and generated two variants gGBEv3.1 (v3 with G174R) and gGBEv3.2 (v3 with D175R). We found that the G-to-T conversion activity for gGBEv3.2 was 1.78-fold of that found for gGBEv3 (Fig. 2a and Supplementary Fig. S3b). We also attempted to scan the R163–V179 region with sequential substitutions of asparagine (X > N), which is critical for substrate recognition by MPG [29,34]. Interestingly, we obtained a variant gGBEv3.3 (v3 with C178N), which substantially elevated the G-to-T conversion activity by 5.5-fold (Fig. 2a and Supplementary Fig. S3c). Furthermore, in a variant gGBEv4 (v3 with both D175R and C178N), we found a synergistic enhancement of G-to-T editing activity by 6.9-fold in comparison with the G editing activity of gGBEv3 (Fig. 2a and Supplementary Fig. S3b). In addition, the gGBEv4 with MPG fused at the C-terminus showed slightly higher editing activity than that at the N-terminus (34.6% vs. 25.9%, Supplementary Fig. S3d).

Figure 2. Mutagenesis of the MPG moiety in gGBEs. (a) Schematic diagram of mutagenesis and screening strategy for the engineered gGBE. The EGFP reporter plasmids were transiently co-transfected into cultured cells along with the gGBE plasmids. (b) Genotypes of a subset of engineered gGBEs, with percentage of EGFP+ cells for each gGBE variant on the far-right column (more engineered variants listed in Supplementary Fig. S5). Different steps of mutagenesis (see text) are marked by different shaded colors. (c and d) G base editing outcomes with different gGBE variants at the edited G7 position in site 3 (c) and site 10 (d) in transfected HEK293T cells by target deep sequencing (mean ± s.e.m., n = 3).

We then performed another round of mutagenesis and screening based on gGBEv4 to further elevate G editing activity of gGBE. We mutated the R163–V179 region of MPGv4 by sequential replacement with amino acids of distinct properties, including glutamic acid (with negative charged side chain), valine (with small hydrophobic side chain), glycine (with no side chain), or tyrosine (with large hydrophobic side chain) (X > E, V, G or Y). Although most of these mutations reduced the G editing activity, we found three variants, gGBEv4.1 (v4 with I170V), gGBEv4.2 (v4 with S169G) and gGBEv4.3 (v4 with R163Y), that showed elevated editing efficiency by 1.06-, 1.28- and 1.09-fold, respectively (Supplementary Fig. S4), as compared with that of gGBEv4. We also attempted to generate two additional variants (gGBEv5.1 and gGBEv5.2) by combining the above three effective mutations, but found no further enhancement (Supplementary Fig. S5). Encouraged by the above finding on the variant gGBEv3.2, we further introduced T199R, S230R, Q294R or D295R individually into gGBEv4.2, mutations previously found to improve conversion activity of AYBE [11]. We found that gGBEv6.3 (v4.2 with Q294R) further increased the editing efficiency by 1.02- and 1694-folds, as compared with gGBEv4.2 and gGBE with WT MPG, respectively (Fig. 2b and Supplementary Fig. S5).

We next validated the enhancement of G editing activity of gGBE variants obtained above at two endogenous genomic sites in cultured HEK293T cells. The cells were transfected with a construct encoding each gGBE variant, together with mCherry and sgRNA that targeted site 3 (or site 10), and mCherry+ cells were FACS-sorted for target deep sequencing analysis. We obtained a gradual elevation of overall G editing efficiency at G7 from 6.4% to 78.5% for site 3, and from 7.5% to 80.3% for site 10, respectively (Fig. 2c and d). The G-to-C and G-to-T conversions were the predominant events at these two sites, and only a low percentage of G-to-A conversion (4.6% for site 3, 5.0% for site 10) and a few insertions and deletions (indels, 8.5% for site 3; 4.3% for site 10) were detected for gGBEv6.3. These results indicate that the four rounds of mutagenesis described above had effectively optimized gGBE activity for G-to-C and G-to-T base editing. Taken together, the engineered version of gGBEv6.3 (carrying G163R, N169G, D175R, C178N, S198A, K202A, G203A, S206A, K210A, Q294R mutations) had the highest G editing efficiency and was used for the following studies.

Characterization of gGBEv6.3 at human genomic DNA sites

We further characterized the editing profiles of gGBEv6.3 by targeting 24 endogenous genomic loci, most of which were used in previous base editing studies [9,35,36]. We found that gGBEv6.3 achieved efficient G base editing activity (ranging from 27.6% to 76.5%), with predominately G-to-C and G-to-T conversions, but essentially no A, C or T editing at all 24 sites examined (Fig. 3a, c and Supplementary Fig. S6a–c). Among all cells examined, the ratio of G-to-C/T (G-to-Y, Y = C and T) to G-to-A/T/C conversion was high (ranging from 0.72 to 0.94, Fig. 3b). Only a low percentage of G-to-A conversion was detected (Fig. 3a and Supplementary Fig. S6e–g), consistent with previous results with AYBE [11] and CGBEs [6–10]. We found that gGBEv6.3 also induced indel at the 24 edited sites with frequency ranging from 4.7% to 30.0% (Supplementary Fig. S6h). Furthermore, we found that the editable range of gGBEv6.3 was positions 1 to 14, and the optimal editing window with high efficiency of G conversion covered protospacer positions 6 to 11, with the highest editing efficiency at position 7 (Fig. 3c, Supplementary Fig. S6d and i). The analysis of on-target editing and sequences of all the sites showed that gGBEv6.3 performed no obvious NG motif preference for G conversions (Supplementary Fig. S6j).

Figure 3. Characterization of editing profiles of gGBE via target deep sequencing. (a) Bar plots showing the on-target DNA base editing at positions with the highest G conversion frequencies at each genomic site in HEK293T cells (mean ± s.e.m., n = 3). G#: G position with highest on-target base editing frequencies across protospacer positions 1–20; site #: genomic site number. (b) The ratio of G-to-C/T to G-to-A/C/T conversion frequency by gGBEv6.3 editing at the sites shown in a. (c) Frequencies of G conversions by gGBEv6.3 across protospacer positions 1–20 at the edited sites in a (in which PAM was at positions 21–23). Single dots represent individual data point from three independent replicates per site. Boxes span the interquartile range (25th to 75th percentile); horizontal line in the box indicates the median (50th percentile); small horizontal bars mark the minimal and maximal values. (d) The sgRNA-dependent off-target analysis for gGBEv6.3 editing efficiency at site 7 (mean ± s.e.m, n = 3). OT: off-target. (e) The sgRNA-independent off-target editing efficiency detected by the orthogonal R-loop assay at each R-loop site (mean ± s.e.m, n = 3).

We have analyzed the guide-dependent off-target activity of gGBEv6.3 at several previously reported [11,35] and in silico-predicted [37] guide-dependent off-target sites, and characterized the ability of gGBEv6.3 to mediate guide-independent off-target DNA editing using orthogonal R-loop assay in five dSaCas9 R-loops, reported in previous studies [11,35]. We found similar or lower percentage of editing at the guide-dependent off-target loci (Fig. 3d and Supplementary Fig. S7a), as compared with that of adenine base editors found previously [11,35]. Moreover, among five guide-independent off-target sites, we detected very low frequencies (1.1% on average) at four sites (Fig. 3e and Supplementary Fig. S7b), and a slightly higher frequency in one site, which harbored 12 Gs across the entire protospacer. Taken together, we found that gGBEv6.3 represents a highly efficient G-to-Y base editor with a low off-target effect in mammalian cells.

Potential gene-editing applications of gGBE

The G-to-Y conversion ability of gGBE allows for a variety of gene-editing applications, including editing splicing sites, introduction of premature termination codons (PTCs), as well as editing that bypasses PTCs (Fig. 4a). The inactive splicing acceptor (SA) signal with disruptive point mutations, exemplified by intron-split EGFP reporter system used above, could be remediated with gGBE (Fig. 1b). Conversely, gGBE could be used for disrupting the splicing signal by converting G within a splicing donor site (‘GT’) or splicing acceptor site (‘AG’) to other bases, resulting in exon skipping. To illustrate this application, we targeted the splice acceptor site of DMD (Duchenne muscular dystrophy) exon 45 with gGBEv6.3, and achieved a high efficiency of G editing (up to 30.3%) with a high G-to-Y ratio (up to 0.88) when targeting DMD site 1 (Fig. 4b, c and Supplementary Fig. S8a).

Figure 4. Gene editing applications of gGBE. (a) Application of gGBEv6.3 for editing splicing sites, introduction of premature termination codons (PTCs), as well as editing that bypasses PTCs. (b) Schematic diagram illustrating gGBE-indued skipping of DMD exon 45. (c) Bar plots showing the on-target DNA base editing frequencies of G editing and the ratio of G-to-C/T to G-to-A/C/T editing frequencies by gGBEv6.3 at two DMD sites in HEK293T cells (mean ± s.e.m, n = 3). SpG, a PAM-flexible Cas9 variant, was used for targeting DMD site 2. (d) Schematic diagram illustrating the introduction of PTCs in the mouse Tyr gene by gGBE. (e) Bar plots showing the on-target G editing frequencies and G-to-Y Ratio by gGBEv6.3 at three Tyr sites in N2a cultured cells (mean ± s.e.m, n = 3). (f) Percentages of G conversion-induced PTCs by gGBEv6.3 at two Tyr sites in mouse pups (mean ± s.e.m., n = 25 and 21 pups for site 2 and site 3, respectively). (g) Phenotype of F0 mice generated by gGBE editing in mouse zygotes. The image showing the presence of edited P6 mice. Red arrowhead, albino; blue arrowhead, mosaic pigmentation. (h) Bar plots showing the on-target G editing frequencies for individual mouse pups, with gGBEv6.3 targeting Tyr site 3. (i) Genotyping of representative F0 pups from (h). The frequencies of mutant alleles were determined by high-throughput sequencing. Red arrowhead, albino pups.

Another application of gGBE is to introduce PTCs to disrupt gene expression by converting TCA, TAC and GAA codons into stop codons TGA, TAG or TAA. Note that PTC by GAA to TAA conversion could be introduced only by using gGBEv6.3, no other current base editor could induce this type of PTC. By targeting three sites in the mouse Tyr (Tyrosinase, associated with coat color) gene with gGBEv6.3 for creating PTCs (Fig. 4d) in cultured N2a cells, we achieved a high efficiency of G editing (up to 46.3%) with high G-to-Y ratio (up to 0.95) (Fig. 4e and Supplementary Fig. S8b). To further illustrate the potential in vivo application, we co-injected gGBEv6.3 mRNA and Tyr-targeting sgRNA into mouse zygotes of C57BL6 background (black coat color), with 20 mouse embryos used for each of three Tyr-targeting sgRNAs. We found highly efficient G editing (Supplementary Fig. S9a) for two of three sgRNAs, with an average of 50.9% PTC introduction efficiency when targeting the Tyr site 3 (Supplementary Fig. S9b and c). Similar to the data obtained above in cultured cell lines (Fig. 4e), gGBE induced very few indels in mouse embryos (Supplementary Fig. S9d and e). When targeting the Tyr site 3 with gGBEv6.3, all the 21 F0 pups had G conversions with an average of 59.4% efficiency (Fig. 4f–h). This gGBEv6.3-induced G-to-Y editing resulted in the albino or mosaic phenotype in the coat color of F0 mice (Supplementary Fig. S10), suggesting efficient disruption of the tyrosinase activity (Fig. 4i). Thus, gGBEv6.3 was an efficient G base editor not only in cell lines but also in mouse embryos.

DISCUSSION

Two major classes of deaminase-based base editors (dBE), ABE and CBE, as well as their derivatives (such as AYBE and CGBE), perform base editing with deamination of A or C as the first key step [3–11]. In this study, we designed deaminase-free editors based on engineered MPG, and generated a gGBE editor that could achieve highly efficient G-to-C and G-to-T conversion in both cultured human cells and mouse embryos. This high editing efficiency could be attributed to mutations in the MPG protein that facilitate its specific substrate selection or DNA-binding activity, or both, which needs to be elucidated by structural and biochemical experiments in the future.

In all human pathogenic SNPs (60 372 total), there are about 10% C-to-G and 5% T-to-G SNPs [11]. Although C-to-G SNPs could be corrected by CGBE [6–10], we could not usually find an efficient sgRNA due to the PAM limitation and narrow editing window. The gGBE could increase the opportunity to find an efficient sgRNA by targeting the opposite strand compared with CGBE. For T-to-G SNPs, no current base editor could efficiently induce G-to-T (or C-to-A in the opposite strand) conversion. Therefore, gGBE greatly broadens the targeting scope of base editors.

Although we have found a low off-target effect of the gGBE editor on several targeted genes in culture mammalian cells, a comprehensive evaluation of gGBEs through high-throughput whole-genome sequencing methods, such as GOTI [38,39], SAFETI [40] and Detect-seq [41], is required for a thorough assessment of off-target effects. Although the editable range (positions 1 to 14) is wide with the current version of gGBE (Fig. 3c), a more accurate gGBE with a refined editing window, like ABE9 [42] or YE1-BE3 [43], might be achieved through further engineering of the glycosylase moiety or architectures of gGBE. Moreover, further structural fine-tuning of gGBE could potentially achieve more specific G-to-C and G-to-T editors by co-expression of deaminases or modifying DNA repair machinery (e.g. co-expression of a specific TLS polymerase) [11,44,45]. We note that indels generated by gGBE, as well as by AYBE and CGBEs that created AP sites, were several folds higher than those generated by ABE or CBE [3–11]. Thus, additional effort is required to reduce the indel frequency. Encouraged by the previous studies on CGBE [7,10] and AYBE [11], the level of off-target editing or indels could be reduced through protein or sgRNA engineering approaches in the future.

In summary, we have shown that DNA glycosylases could be engineered into proteins that selectively excise a specific nucleotide base. Thus, besides G base editing illustrated here, glycosylase-based base editor could also be developed by similar protein engineering methods for base editing of A, C and T, providing a complete set of base editing tools for gene editing studies.

MATERIALS AND METHODS

Molecular cloning

Base editor constructs used in this study were cloned into a mammalian expression plasmid backbone under the control of a EF1α promoter by standard molecular cloning techniques. Intron-split EGFP reporters were engineered as previously described [11]. In brief, corresponding mutations at the splice acceptor site were made to construct A-to-T reporter or G-to-T reporter via site-directed mutagenesis by PCR, respectively. Mutations at the splice acceptor site led to inactive EGFP production by non-spliced EGFP transcripts. Encouraged by the findings from previous base editors [7,10], modification of the 68th base (G-to-C) was made in the intron sequence for introducing artificial PAM on the template strand, thus the corresponding mutations at the splice acceptor site were at position 6 across the protospacer. Transversion corrections in A-to-T reporter or G-to-T reporter were required for proper splicing of EGFP-coding sequence. MPG mutagenesis libraries were designed and generated as previously described [46]. The 17aa-long region from R163 to V179 (R163–V179) of MPGv3 was selected for protein engineering. BpiI-harboring mutant, MPG- G174R/D175R/T199R/S230R/Q294R/D295R mutants or corresponding combinations were constructed via site-directed mutagenesis by PCR. Sequential asparagine/glutamic acid/valine/glycine/tyrosine substitutions (X > N, E, V, G or Y) were designed, with oligos coding for the mutants annealed and ligated into corresponding BpiI-digested backbone vectors. The gRNA oligos were annealed and ligated into BpiI sites. The amino-acid sequence for gGBEv6.3 is supplied in Supplementary Table 1. The MPG mutants and corresponding codon substitutions used are listed in Supplementary Table 2.

Analysis of target sequencing data

Target sequencing data analysis has been previously described [11]. In brief, the targeted amplicon sequencing reads were processed using fastp with default parameters [47]. The cleaned pairs were then merged using FLASH v1.2.11. The amplified sequences from individual targets were demultiplexed using fastx_barcode_splitter.pl from fastx_toolkit (0.0.14). Further amplicon sequencing analysis was performed by CRISPResso2 [48]. A 10-bp window was used to quantify modifications centered around the middle of the 20-bp gRNA. Otherwise, the default parameters were used for analysis. G-to-C purity was calculated as G-to-C editing efficiency/(G-to-C editing efficiency + G-to-T editing efficiency + G-to-A editing efficiency). G-to-Y conversion ratio was calculated as (G-to-C editing efficiency + G-to-T editing efficiency)/(G-to-C editing efficiency + G-to-T editing efficiency + G-to-A editing efficiency).

DATA AVAILABILITY

Expression plasmids used in this study have been deposited at Addgene and will be available at https://www.addgene.org/Huawei_Tong/ (Addgene plasmid nos. 202629-202630). All data are available in the main text or the Supplementary Materials. Targeted amplicon sequencing data have been deposited at the Sequence Read Archive and can be accessed at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA971099. All relevant original data are available from the corresponding authors upon request.

Supplementary Material

nwad143_Supplemental_Files Click here for additional data file.

ACKNOWLEDGEMENTS

We thank M.-m. Poo for helpful discussions and comments on this manuscript.

FUNDING

This work was supported by HuidaGene Therapeutics Co., Ltd. (H.T.).

AUTHOR CONTRIBUTIONS

H.T. designed and conducted experiments. N.L., Y.L. and D.W. performed experiments with the help of S.C. and J.Z.; Y.W., M.J. and H.L performed mouse experiments; Y.Z. performed sequencing data analysis. G.L., Y.Y. and H.Z. provided key advice for this project. H.Y., L.S., X.Y. and H.T. supervised the whole project. H.Y. and H.T. wrote the manuscript, and all authors contributed to the editing of the manuscript.

Conflict of interest statement. H.T. is the inventor of patent applications related to this manuscript. H.T., N.L., Y.Z., Y.L., D.W., S.C., H.L., G.L., J.Z., Y.Y. and H.Z. are employees of HuidaGene Therapeutics Co., Ltd. H.Y., L.S. and X.Y. are cofounders of HuidaGene Therapeutics Co., Ltd. The remaining authors declare no competing interests.
==== Refs
REFERENCES

1. Porto EM , KomorAC, SlaymakerIMet al. Base editing: advances and therapeutic opportunities. Nat Rev Drug Discov 2020; 19 : 839–59.10.1038/s41573-020-0084-6 33077937
2. Rees HA , LiuDR. Base editing: precision chemistry on the genome and transcriptome of living cells. Nat Rev Genet 2018; 19 : 770–88.10.1038/s41576-018-0059-1 30323312
3. Komor AC , ZhaoKT, PackerMSet al. Improved base excision repair inhibition and bacteriophage Mu Gam protein yields C:G-to-T:A base editors with higher efficiency and product purity. Sci Adv 2017; 3 : eaao4774.10.1126/sciadv.aao4774 28875174
4. Komor AC , KimYB, PackerMSet al. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature 2016; 533 : 420–4.10.1038/nature17946 27096365
5. Gaudelli NM , KomorAC, ReesHAet al. Programmable base editing of A*T to G*C in genomic DNA without DNA cleavage. Nature 2017; 551 : 464–71.10.1038/nature24644 29160308
6. Zhao D , LiJ, LiSet al. Glycosylase base editors enable C-to-A and C-to-G base changes. Nat Biotechnol 2021; 39 : 35–40.10.1038/s41587-020-0592-2 32690970
7. Kurt IC , ZhouR, IyerSet al. CRISPR C-to-G base editors for inducing targeted DNA transversions in human cells. Nat Biotechnol 2021; 39 : 41–6.10.1038/s41587-020-0609-x 32690971
8. Koblan LW , ArbabM, ShenMWet al. Efficient C*G-to-G*C base editors developed using crispri screens, target-library analysis, and machine learning. Nat Biotechnol 2021; 39 : 1414–25.10.1038/s41587-021-00938-z 34183861
9. Chen L , ParkJE, PaaPet al. Programmable C:G to G:C genome editing with CRISPR-Cas9-directed base excision repair proteins. Nat Commun 2021; 12 : 1384.10.1038/s41467-021-21559-9 33654077
10. Yuan T , YanN, FeiTet al. Optimization of C-to-G base editors with sequence context preference predictable by machine learning methods. Nat Commun 2021; 12 : 4902.10.1038/s41467-021-25217-y 34385461
11. Tong H , WangX, LiuYet al. Programmable A-to-Y base editing by fusing an adenine base editor with an N-methylpurine DNA glycosylase. Nat Biotechnol 2023; doi: 10.1038/s41587-022-01595-6.
12. Mok BY , de MoraesMH, ZengJet al. A bacterial cytidine deaminase toxin enables CRISPR-free mitochondrial base editing. Nature 2020; 583 : 631–7.10.1038/s41586-020-2477-4 32641830
13. Lei Z , MengH, LiuLet al. Mitochondrial base editor induces substantial nuclear off-target mutations. Nature 2022; 606 : 804–11.10.1038/s41586-022-04836-5 35551512
14. Cho SI , LeeS, MokYGet al. Targeted A-to-G base editing in human mitochondrial DNA with programmable deaminases. Cell 2022; 185 : 1764–76.10.1016/j.cell.2022.03.039 35472302
15. Ladstatter S , Tachibana-KonwalskiK. A surveillance mechanism ensures repair of DNA lesions during zygotic reprogramming. Cell 2016; 167 : 1774–87.10.1016/j.cell.2016.11.009 27916276
16. Marrone A , BallantyneJ. Hydrolysis of DNA and its molecular components in the dry state. Forensic Sci Int Genet 2010; 4 : 168–77.10.1016/j.fsigen.2009.08.007 20215028
17. Lindahl T . Instability and decay of the primary structure of DNA. Nature 1993; 362 : 709–15.10.1038/362709a0 8469282
18. Lee SJ , ChoiMY, MillerRE. Vibrational spectroscopy of xanthine in superfluid helium nanodroplets. Chem Phys Lett 2009; 475 : 24–9.10.1016/j.cplett.2009.05.016
19. Hindi NN , ElsakrmyN, RamotarD. The base excision repair process: comparison between higher and lower eukaryotes. Cell Mol Life Sci 2021; 78 : 7943–65.10.1007/s00018-021-03990-9 34734296
20. Thompson PS , CortezD. New insights into abasic site repair and tolerance. DNA Repair (Amst) 2020; 90 : 102866.10.1016/j.dnarep.2020.102866 32417669
21. Bauer NC , CorbettAH, DoetschPW. The current state of eukaryotic DNA base damage and repair. Nucleic Acids Res 2015; 43 : 10083–101.10.1093/nar/gkv1136 26519467
22. Jacobs AL , SchärP. DNA glycosylases: in DNA repair and beyond. Chromosoma 2012; 121 : 1–20.10.1007/s00412-011-0347-4 22048164
23. Robertson AB , KlunglandA, RognesTet al. DNA repair in mammalian cells: base excision repair: the long and short of it. Cell Mol Life Sci 2009; 66 : 981–93.10.1007/s00018-009-8736-z 19153658
24. Gallagher PE , BrentTP. Partial purification and characterization of 3-methyladenine-DNA glycosylase from human placenta. Biochemistry 1982; 21 : 6404–9.10.1021/bi00268a013 7150564
25. Asaeda A , IdeH, AsagoshiKet al. Substrate specificity of human methylpurine DNA N-glycosylase. Biochemistry 2000; 39 : 1959–65.10.1021/bi9917075 10684645
26. Lau AY , WyattMD, GlassnerBJet al. Molecular basis for discriminating between normal and damaged bases by the human alkyladenine glycosylase, AAG. Proc Natl Acad Sci USA 2000; 97 : 13573–8.10.1073/pnas.97.25.13573 11106395
27. Wyatt MD , SamsonLD. Influence of DNA structure on hypoxanthine and 1,N6-ethenoadenine removal by murine 3-methyladenine DNA glycosylase. Carcinogenesis 2000; 21 : 901–8.10.1093/carcin/21.5.901 10783310
28. Saparbaev M , LavalJ. Excision of hypoxanthine from DNA containing dIMP residues by the Escherichia coli, yeast, rat, and human alkylpurine DNA glycosylases. Proc Natl Acad Sci USA 1994; 91 : 5873–7.10.1073/pnas.91.13.5873 8016081
29. O’Brien PJ , EllenbergerT. Dissecting the broad substrate specificity of human 3-methyladenine-DNA glycosylase. J Biol Chem 2004; 279 : 9750–7.10.1074/jbc.M312232200 14688248
30. Berdal KG , JohansenRF, SeebergE. Release of normal bases from intact DNA by a native DNA repair enzyme. EMBO J 1998; 17 : 363–7.10.1093/emboj/17.2.363 9430628
31. Kay JE , CorriganJJ, ArmijoALet al. Excision of mutagenic replication-blocking lesions suppresses cancer but promotes cytotoxicity and lethality in nitrosamine-exposed mice. Cell Rep 2021; 34 : 108864.10.1016/j.celrep.2021.108864 33730582
32. Vallur AC , MaherRL, BloomLB. The efficiency of hypoxanthine excision by alkyladenine DNA glycosylase is altered by changes in nearest neighbor bases. DNA Repair (Amst) 2005; 4 : 1088–98.10.1016/j.dnarep.2005.05.008 15990363
33. Lau AY , SchärerOD, SamsonLet al. Crystal structure of a human alkylbase-DNA repair enzyme complexed to DNA: mechanisms for nucleotide flipping and base excision. Cell 1998; 95 : 249–58.10.1016/S0092-8674(00)81755-9 9790531
34. Connor EE , WyattMD. Active-site clashes prevent the human 3-methyladenine DNA glycosylase from improperly removing bases. Chem Biol 2002; 9 : 1033–41.10.1016/S1074-5521(02)00215-6 12323378
35. Richter MF , ZhaoKT, EtonEet al. Phage-assisted evolution of an adenine base editor with improved Cas domain compatibility and activity. Nat Biotechnol 2020; 38 : 883–91.10.1038/s41587-020-0453-z 32433547
36. Doman JL , RaguramA, NewbyGAet al. Evaluation and minimization of Cas9-independent off-target DNA editing by cytosine base editors. Nat Biotechnol 2020; 38 : 620–8.10.1038/s41587-020-0414-6 32042165
37. Bae S , ParkJ, KimJS. Cas-OFFinder: a fast and versatile algorithm that searches for potential off-target sites of Cas9 RNA-guided endonucleases. Bioinformatics 2014; 30 : 1473–5.10.1093/bioinformatics/btu048 24463181
38. Zuo E , SunY, WeiWet al. GOTI, a method to identify genome-wide off-target effects of genome editing in mouse embryos. Nat Protoc 2020; 15 : 3009–29.10.1038/s41596-020-0361-1 32796939
39. Zuo E , SunY, WeiWet al. Cytosine base editor generates substantial off-target single-nucleotide variants in mouse embryos. Science 2019; 364 : 289–92.10.1126/science.aav9973 30819928
40. Yan N , FengH, SunYet al. Cytosine base editors induce off-target mutations and adverse phenotypic effects in transgenic mice. Nat Commun 2023; 14 : 1784.10.1038/s41467-023-37508-7 36997536
41. Lei Z , MengH, LvZet al. Detect-seq reveals out-of-protospacer editing and target-strand editing by cytosine base editors. Nat Methods 2021; 18 : 643–51.10.1038/s41592-021-01172-w 34099937
42. Chen L , ZhangS, XueNet al. Engineering a precise adenine base editor with minimal bystander editing. Nat Chem Biol 2023; 19 : 101–10.10.1038/s41589-022-01163-8 36229683
43. Kim YB , KomorAC, LevyJMet al. Increasing the genome-targeting scope and precision of base editing with engineered Cas9-cytidine deaminase fusions. Nat Biotechnol 2017; 35 : 371–6.10.1038/nbt.3803 28191901
44. Wang Y , ZhaoD, SunLet al. Engineering of the translesion DNA synthesis pathway enables controllable C-to-G and C-to-A base editing in Corynebacterium glutamicum. ACS Synth Biol 2022; 11 : 3368–78.10.1021/acssynbio.2c00265 36099191
45. Sun N , ZhaoD, LiSet al. Reconstructed glycosylase base editors GBE2.0 with enhanced C-to-G base editing efficiency and purity. Mol Ther 2022; 30 : 2452–63.10.1016/j.ymthe.2022.03.023 35381364
46. Tong H , HuangJ, XiaoQet al. High-fidelity Cas13 variants for targeted RNA degradation with minimal collateral effects. Nat Biotechnol 2023; 41 : 108–19.10.1038/s41587-022-01419-7 35953673
47. Chen S , ZhouY, ChenYet al. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics 2018; 34 : i884–90.10.1093/bioinformatics/bty560 30423086
48. Clement K , ReesH, CanverMCet al. CRISPResso2 provides accurate and rapid genome editing sequence analysis. Nat Biotechnol 2019; 37 : 224–6.10.1038/s41587-019-0032-3 30809026