A split and inducible adenine base editor for precise in vivo base editing

Abstract

DNA base editors use deaminases fused to a programmable DNA-binding protein for targeted nucleotide conversion. However, the most widely used TadA deaminases lack post-translational control in living cells. Here, we present a split adenine base editor (sABE) that utilizes chemically induced dimerization (CID) to control the catalytic activity of the deoxyadenosine deaminase TadA-8e. sABE shows high on-target editing activity comparable to the original ABE with TadA-8e (ABE8e) upon rapamycin induction while maintaining low background activity without induction. Importantly, sABE exhibits a narrower activity window on DNA and higher precision than ABE8e, with an improved single-to-double ratio of adenine editing and reduced genomic and transcriptomic off-target effects. sABE can achieve gene knockout through multiplex splice donor disruption in human cells. Furthermore, when delivered via dual adeno-associated virus vectors, sABE can efficiently convert a single A•T base pair to a G•C base pair on the PCSK9 gene in mouse liver, demonstrating in vivo CID-controlled DNA base editing. Thus, sABE enables precise control of base editing, which will have broad implications for basic research and in vivo therapeutic applications.

Introduction

As an emerging class of precision genome-editing tools, DNA base editors consist of a deaminase fused to a programmable DNA-binding protein, enabling targeted nucleotide conversions without introducing double-stranded DNA breaks^1,2. Adenine base editors (ABEs) utilize an evolved Escherichia coli tRNA adenosine deaminase (TadA) and act on a single-stranded DNA substrate for A•T to G•C base conversions², which have been tested in animal models^{3,4,5,6,7,8,9} and primary human cells^4,8,9,10,11 with various applications, including site-directed mutagenesis¹², gene silencing¹³, gene knockout¹⁰, gene isoform discovery¹⁴, functional screens of epigenetic markers¹⁵ or pathogenic mutations¹⁶, and molecular recording¹⁷. ABEs are particularly useful for investigating or therapeutically correcting human pathogenic alleles because nearly half of the disease-causing point mutations could be corrected by reversing the pathogenic A•T base pair to a G•C base pair^18,19. Recently, TadA-ABEs have been re-engineered to achieve other types of base editing, including C-to-T^20,21,22, C-to-G²¹, C/A-to-T/G^20,22, or A-to-Y²³ conversions.

However, the lack of precise control over the deamination activity of ABE limits its application in research and therapy. The current TadAs in ABEs are constitutively active, and the uncontrolled deaminase can cause undesirable genomic and transcriptomic off-target effects^24,25,26,27, raising concerns for ABEs’ application for the production of genetically modified organisms and gene therapy. For instance, BEs lead to both genomic and transcriptomic off-target due to long-term expression in vivo in transgenic mice, and mice zygotes injected with ABE7.10 encoded AAV exhibit low birth rates²⁸. Although inducible promoters can be used to regulate the expression of ABEs^29,30, the leaky expressions and the delayed response from transcription to translation are highly undesirable. Post-translational inducible control of Cas proteins^31,32,33,34 can potentially regulate ABE recruitment to the genome but still cannot directly control the deaminase activity of ABE, which does not curtail its off-target effects^25,26,27,35. Thus, precision control of the deaminase activity of ABEs would greatly expand its applications.

Here, we present a split ABE (sABE) design with inducible deaminase activity by integrating the chemically induced dimerization (CID) system³⁶. We demonstrate that ABE8e²⁴ can be split into two inactive parts in the TadA-8e deaminase domain: one fused to FK506-binding protein 3 (FKBP3) and the other to FKBP-rapamycin binding (FRB) protein³⁷. These two ABE components can reassemble into an active form upon rapamycin-induced FRB-FKBP3 heterodimerization. Through extensive engineering and optimization, we engineer sABE v3.22, which shows efficient and precise on-target single adenine editing upon rapamycin induction and significantly reduced genomic and transcriptomic off-target effects. Using dual adeno-associated viral (AAV) vectors to deliver sABE v3.22 in mice, we perform inducible editing of the PCSK9 gene and demonstrate high precision on the targeted adenine, showcasing in vivo CID-controlled DNA base editing.

Results

Chemically inducible split ABE (sABE) with tightly regulated deaminase activity

To monitor the DNA deaminase activity, we created a fluorescence reporter by introducing a premature stop codon into the EYFP gene via a C•G to T•A base pair conversion, rendering it dysfunctional (EYFP*) (Fig. 1a). ABE guided by a single guide RNA (sgRNA) can edit the adenine on the antisense strand of the EYFP* gene and convert the A•T base pair back to the G•C base pair, thereby restoring the original glutamine codon and resulting in the expression of full-length, functional EYFP (Fig. 1a). We validated the response of EYFP fluorescence to ABE8e in HEK293T cells, without detectable background fluorescence in the absence of ABE8e (Fig. 1a).

To achieve inducible control over ABE deaminase, we split TadA-8e into two inactive parts (TadA-8e_N and TadA-8e_C) and fused each part to FRB and FKBP3, respectively. In the presence of rapamycin, FKBP3 and FRB will heterodimerize, bringing the two parts of TadA-8e into proximity and enabling their assembly into a functional unit (Fig. 1b). We constructed sABE v1 and sABE v2 by splitting the TadA-8e³⁸ deaminase into two fragments. The split sites occurred at loop-25 for sABE v1 and loop-74 for sABE v2 (Fig. 1d). An FKBP3-FRB dimer insertion into these peripheral flexible loop regions is unlikely to alter the TadA-8e core catalytic domain or the reassembly of TadA-8e_N and TadA-8e_C. In sABE v1, TadA-8e_N contains the first 24 amino acids of the TadA-8e, which is linked to an FRB via a flexible linker to its C-terminus (Supplementary Fig. 2a). We also fuse a bipartite SV40 nuclear localization signal (NLS) at the N-terminus of TadA-8e_N. TadA-8e_C contains the remaining 142 amino acids of the TadA-8e and is fused to an FKBP3 at its N terminus and a Streptococcus pyogenes Cas9 nickase (nSpCas9, D10A) at its C terminus. Each terminus has a monopartite SV40 NLS. These two components, sABE(N) and sABE(C), are expressed separately from two plasmids under a cytomegalovirus promoter (pCMV). The constructs in sABE v2 are similar, except that the split site occurs after Arginine 74 of TadA-8e. We co-transfected HEK293T cells with plasmids encoding sABE(N), sABE(C), EYFP*, and sgRNA, using EBFP as a negative control, followed by induction of the sABE activity with 100 nM rapamycin 12 h after transfection (Fig. 1c). Forty-eight hours after induction, we quantified the normalized fluorescence intensity and the percentage of EYFP-positive cells by flow cytometry (Fig. 1c, Supplementary Fig. 1). sABE v1 and v2 successfully activated the EYFP reporter upon rapamycin induction, with sABE v2 showing higher EYFP activation but also higher background (Fig. 1h, Supplementary Fig. 2b). To further investigate split sites adjacent to Arginine 74, we created sABEs v2.1 to v2.4 by shifting the split site one amino acid at a time (Supplementary Fig. 2c). We found that sABE v2.3, with the split occurring after Isoleucine 76 of TadA-8e, had higher EYFP activation upon rapamycin induction and lower background compared to sABE v2 (Supplementary Fig. 2d).

To further improve the rapamycin-induced deaminase activity and reduce the background activity under the non-induced condition, we optimize components in the sABE construct (Fig. 1e). First, we developed sABE v2.7 by adding a nucleoplasmin NLS to the C terminus of sABE v2.3(N) while keeping the same sABE v2.3(C). Although this modification enhanced editing efficiency, we also observed a higher background activity, probably due to the auto-reassembly of the TadA-8e fragments in the nucleus when they are abundant (Fig. 1h, Supplementary Fig. 3b). Next, we characterized the effect of the dimerization domain copy number on sABE activity. We constructed sABE v2.8, v2.9, and v3.11 by introducing an additional copy of the dimerization domains based on sABE v2.3 (Supplementary Fig. 2e). We found that sABE v3.11, harboring two copies of FRB domain at the C terminus of sABE(N) and two copies of FKBP3 domain at the N terminus of sABE(C), led to a comparable level of EYFP reporter activation with a reduced background (Supplementary Fig. 2f). We then tested different types of linkers with varying lengths between TadA-8e_N and 2×FRB domain in sABEv3.11(N) and between 2×FKBP3 domain and TadA-8e_C in sABE v3.11(C), creating four versions of sABE(N) and four versions of sABE(C). We transfected different combinations of resulting sABE(N) and sABE(C) constructs and screened a total of 16 sABEs (v3.11 to v3.44) using our fluorescence reporter assay (Supplementary Fig. 3a). We chose sABE v3.22 as the final version since it showed comparable EYFP activation with sABE v3.11 while exhibiting significantly reduced background activity under the non-induced condition (Fig. 1f).

Further, after evaluating a range of rapamycin concentrations, we found that 100 nM rapamycin effectively activated sABE v3.22 (Fig. 1g, Supplementary Fig. 3d). We decided to use this concentration for subsequent experiments. In addition, we selected five sABEs (v1, v2, v2.3, v2.7, and v3.11) to compare reporter assay responses and endogenous gene editing efficiencies at three genomic sites in HEK293T cells. The results showed a strong correlation between the sABE activities in these two assays (Fig. 1h, i, Supplementary Fig. 3b, c). We also examined whether the sABE system could be deactivated. Cells transfected with sABE v3.22 were treated with 10, 25, 50, or 100 nM rapamycin for 2 h, after which the culture medium was changed to remove the rapamycin. Both the reporter assay and genomic editing data showed sABE v3.22 activation in rapamycin-treated groups. The group from which rapamycin was removed showed decreased deaminase activity compared to the rapamycin-sustained group (Supplementary Fig. 4). This effect was less significant when the initial concentration of rapamycin was increased beyond 50 nM, likely due to the residual intracellular rapamycin and the inefficient excretion and degradation of rapamycin in HEK293T cells in vitro³⁹. In sum, we successfully split the ABE8e into two inactive parts at the TadA-8e deaminase domain and rendered its deaminase activity chemically inducible using the FKBP3-FRB CID. Through engineering approaches and fluorescence reporter screening, we developed sABE v3.22, which has a high level of induced base editing activity and a low level of non-induced background activity.

sABE v3.22 achieves high DNA on-target editing efficiencies and enhanced precision

We compared the performance of sABE v3.22 to the intact ABE8e by targeting 19 human genomic loci that span different sequence contexts (Fig. 2a, Supplementary Fig. 5a). ABE8e achieved A-to-G conversions ranging from 7.2% to 72% in the conventional A₄-A₈ activity window, with a mean of 56% at the A₄-A₅ positions. In the absence of rapamycin, sABE v3.22 showed very low background A-to-G conversions in the A₄-A₈ window ranging from 0.1% to 3.1%, with a mean of 0.7%. The deaminase activity of sABE v3.22 was induced by an average of 89-fold (ranging from 15-fold to 389-fold), reaching a mean of 80% (ranging from 53% to 97%) of the activities of intact ABE8e at A₄-A₅ positions (Fig. 2b). Additionally, sABE v3.22 exhibited a narrower activity window of A₄ and A₅, with reduced activity on A₆ and A₇ and minimum activity elsewhere in the protospacer (Fig. 2c, Supplementary Fig. 5b)….