Introduction

Mitochondria, essential organelles responsible for energy production and cellular homeostasis, possess their own genome distinct from the nuclear one. Mitochondrial DNA (mtDNA) alterations have been implicated in various human diseases, making precise genome editing tools invaluable for both basic research and potential therapeutic interventions [1]. Base editing has emerged as a transformative technology for precise genome manipulation, offering unprecedented capabilities for targeted nucleotide alterations without inducing double-strand breaks. While extensively explored in the nuclear genome, the application of base editing to mtDNA presents unique opportunities, but also challenges [2].

The inability to efficiently deliver nucleic acids into mitochondria has prompted the development of genome modification technologies based on custom DNA-binding protein. These technologies included mitochondrially-targeted zinc finger nucleases (mtZFNs), transcription activator-like effector nucleases (mitoTALEs) and meganuclease (mitoARCUS) and have been used for site-specific cleavage of mtDNA [3], [4], [5], [6], [7]. Combining ZF or TALE programmable DNA-binding proteins with deaminase enzymes, such as cytidine or adenine deaminases, allows for C:G to T:A or A:T to G:C conversions, respectively. The first major breakthrough was DddA-derived cytosine base editors (DdCBEs) which allowed for site-specific C:G to T:A mutations in the mtDNA for the first time [8]. The initial dimeric DdCBE architecture contained the catalytic domain of DddAtox bacterial cytidine deaminase acting on double-stranded DNA, split into two inactive fragments (at positions G1333 or C1397), which were brought together by TALE DNA binding domains [8]. The technology was rapidly used by our lab and others for many applications including; knocking out each mtDNA-encoded protein-coding gene [9], reverse genetics approaches [10] and in vivo applications in zebrafish[11], [12], rats [13] and mice [14], [15]. The technology was also evolved to improve efficiency and widen compatibility for different sequence contexts and DNA-binding proteins[16], [17], [18], [19], [20]. The discovery of mitochondrial cytidine base editing provided foundation for the development of adenine base editors in mitochondria using the TadA8e enzyme, deemed transcription-activator-like effector (TALE)-linked deaminases (TALEDs), majorly increasing the capabilities of the mitochondrial gene editing toolkit [21]. Further developments included adenine editors containing a nickase to generate the single-stranded DNA required for deaminase activity (mitoBEs) [22]. In this study, we sought to apply the adenine base editing technologies, TALEDs and mitoBEs, in vivo using adeno-associated viral delivery to mouse somatic tissues.

Materials and methods

Plasmid construction and viral vectors

The TALE sequences and plasmids used were designed as reported in [9]. To construct the plasmids used in the cell screen, the functional domain was changed to varying architectures of TadA8e, DddAtox(E1347A) or BspD61(C) using AvrII and XbaI for pTracer or BamHI and KpnI for pcDNA3.1. Vector construction intended for AAV production was achieved by restriction digest of the transgenes using 5’ NotI and 3’ BamHI sites, allowing cloning into a rAAV2-CMV backbone, previously reported in [15]. The cloned plasmids were used to generate recombinant AAV9 viral particles at the UNC Gene Therapy Center, Vector Core Facility (Chapel Hill, NC) and VectorBuilder.

Cell culture and transfections

NIH/3T3 cells (CRL-1658TM, obtained from the American Type Culture Collection (ATCC)) were maintained at 37 °C in a 5% (vol/vol) CO2 atmosphere in complete Dulbecco’s Modified Eagle Medium (DMEM) containing 4.5 g/L glucose, 2 mM glutamine, and 110 mg/ml sodium pyruvate, supplemented with 10% calf bovine serum with iron and 5% penicillin/streptomycin (all sourced from Gibco). Regular mycoplasma tests performed on the culture medium yielded negative results.

For the adenine base editor pair screening, NIH/3T3 mouse cells were seeded in six-well tissue culture plates at a 70% confluency and transfected with 3200 ng of each monomer (L and H), using 16 µl of FuGENE-HD (Promega) according to the manufacturer’s instructions. After 24 h, the cells were harvested for Fluorescence-activated cell sorting (FACS) and sorted for GFP and mCherry double-positive cells employing a BD FACSMelody^TM Cell sorter. The collected double-positive cells were allowed to recover for an additional 6 days before being utilised for DNA extraction, as detailed below. For cell survival experiments, AAV plasmids were transfected with empty GFP plasmid in a 1:1 ratio, sorted using BD FACSMelody^TM Cell sorter and replate for 3 days. Cells were then analysed using Countess™ 3 Automated Cell Counter (Invitrogen) to determine cell survival.

Animals

(Ethics statement) All animal experiments were approved by the local Animal Welfare Ethical Review Body (AWERB) at the University of Cambridge and carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 (Procedure Project Licence: PP1740969) and EU Directive 2010/63/EU). Mice on C57BL/6J background were sourced from Charles River Laboratories. Animals were housed in a temperature- and humidity-controlled environment, maintained under a 12-hour light/12-hour dark cycle, and continuous access to food and water ad libitum. Euthanasia was carried out via cervical dislocation. Newborn pups [both males and females on Postnatal day 1(P1)] were injected with a dose of 5 × 10¹¹ vg per monomer per animal (1 x 10¹² vg total) via the temporal vein (systemic administration), using a 30 G, 30° bevelled needle syringe. Control P1 pups were injected with equivalent volumes of vehicle buffer (1× PBS, 230 mM NaCl, and 5% w/v D-sorbitol). A total of 3 animals per group were used.

Genomic DNA isolation and Sanger sequencing

NIH/3T3 mouse cells underwent trypsinisation for collection, followed by a single wash in PBS and extraction using the DNeasy Blood & Tissue Kit (QIAGEN) following manufacturers guidelines. The lysates were then extracted using DNeasy Blood & Tissue Kit (QIAGEN) following manufacturers guidelines. Genomic DNA extraction from mouse samples was also extracted using the DNeasy Blood & Tissue Kit (QIAGEN) following manufacturers guidelines. For Sanger sequencing, the edited region underwent PCR amplification using GoTaq G2 DNA polymerase (Promega) and using primers and PCR protocol for each site as listed in [9]. PCR purification and subsequent Sanger sequencing were conducted by GENEWIZ/AZENTA (UK).

High-throughput targeted amplicon mtDNA sequencing, processing and mapping

mtDNA-wide sequence analysis was performed, sequenced and analysed as in [9]. Briefly, two overlapping long amplicons (8,331 bp and 8,605 bp) spanning the entire mtDNA molecule were generated via long-range PCR utilising PrimeSTAR GXL DNA polymerase (TAKARA). Tagmentation and indexing PCR were performed using the Nextera XT index kit (Illumina, FC-131-1096) as per the manufacturer’s instructions. Subsequently, libraries were subjected to high-throughput sequencing on the Illumina NovaSeq platform (PE250), with demultiplexing performed using the respective manufacturer’s software.

For processing and mapping of the high-throughput data related to mtDNA-wide analysis, sequenced reads underwent quality trimming and 3′ end adaptor clipping simultaneously using Trim Galore! (--paired). Reads were aligned to ChrM of the mouse reference genome (GRCm39) with Bowtie2 (--very-sensitive; --no-mixed; --no-discordant). Count tables were then generated using samtools mpileup (-q 30) and varscan.

Quantification of viral genomes copy number and relative mtDNA content by quantitative real-time PCR

Quantitative real-time PCR analyses were conducted in a QuantStudioTM 3 system (Thermo Fisher) using TaqMan™ Gene Expression Master Mix (Thermo Fisher), following the manufactureŕs instructions, and Ct values were obtained using the system’s built-in software. Each reaction was performed in a final volume of 20 µl in technical triplicates, with each dataset representing samples from an individual mouse.

To assess the relative mtDNA content in mouse tissue samples, mtDNA levels were quantified using primers and probes targeting MT-ND1: Mmu_Nd1_Fw: 5’-GAG CCT CAA ACT CCA AAT ACT CAC T −3’; Mmu_Nd1_Rv: 5’-GAA CTG ATA AAA GGA TAA TAG CTA TGG TTA CTT CA −3’; Mmu_Nd1_probe: 5’-/56-FAM/ CCG TAG CCC /ZEN/ AAA CAA T /3IABkFQ / −3’. MT-ND1 levels were normalized to the nuclear gene Actin using the following primers and probe: Mmu_Actin_Fw: 5’-CTG CTC TTT CCC AGA CGA GG −3’; Mmu_Actin_Rv: 5’-AAG GCC ACT TAT CAC CAG CC −3’; Mmu_Actin_probe: 5’-/56-TAMN/ ATT GCC TTT CTG ACT AGG TG /3BHQ/ −3’. ΔΔCt analysis was employed to calculate the relative mtDNA copy number, with vehicle-injected mice serving as the control.

Immunoblotting of adenine base editors in mouse tissues

Mouse tissue samples (∼50 mg) underwent homogenization in 200 uL of ice-cold RIPA buffer (consisting of 150 mM NaCl, 1.0% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mM Tris, pH 8.0) supplemented with 1X cOmplete^TM mini EDTA-free Protease Inhibitor Cocktail (Roche, UK), utilizing a gentleMACS^TM dissociator. Following homogenization, the samples were incubated on ice for 20 mins and then clarified by centrifugation (20,000 × g for 20 min at 4 °C). Protein lysates (∼20 μg) were combined with 10 X NuPAGE™ sample reducing agent and 4X NuPAGE™ LDS sample buffer (Invitrogen), followed by incubation for 5 min at 95 °C. The denatured protein samples were subsequently separated in a Bolt 4–12% Bis-Tris pre-cast gel (Thermo Fisher) and transferred onto a PVDF membrane using an iBlot 2 gel transfer system (Thermo Fisher), following the manufactureŕs guidelines. Residual proteins remaining in the gel were visualized using SimpleBlue SafeStain (Thermo Fisher) and served as a loading control. The membrane was blocked in 5% milk in PBS with 0.1% Tween 20 (PBS-T) for 1 h at room temperature (RT) before being probed with either rat anti-HA-tag antibody (Roche, 11867423001) diluted 1:1000 in 5% milk in PBST, mouse anti-FLAG-M2-tag antibody (Sigma–Aldrich, F3165) diluted 1:2000 in 5% milk in PBST or mouse OXPHOS cocktail (abcam, ab110412) diluted 1:300 in 5% milk in PBST. After three washes with PBS-T for 10 min each at RT, the membranes were incubated with HRP-linked secondary antibodies, either anti-rat IgG (Cell Signaling, 7077 S) or anti-mouse IgG (Promega, W4021), diluted 1:5000 in 5% milk in PBST. Following another three washes as before, the membranes were digitally imaged using an Amersham Imager 680 blot and gel imager (GE Healthcare) upon exposure to Amersham ECLTM Western Blotting Detection Reagents (GE Healthcare).

Results and discussion

Adenine base editing in vitro

The originally reported adenine base editors (TALEDs) were shown to work within varying efficacies dependent on the architecture used [21]. The authors used either: (1) dimeric TALEDs (dTALED), which combines two sequence specific TALE arrays binding to opposite strands, with either the catalytically inactive (E1357A) DddAtox or TadA8e at their C-terminus, (2) split DddA TALED (sTALED) which uses the dimeric approach but with the DddAtox(E1357A) split at the G1397 position or (3) monomeric TALED (mTALED) which uses a single TALE-array attached to both the DddAtox(E1357A) and the TadA8e adenine deaminase (Fig. 1a; Supplementary Fig. 1). In order to test the efficiency of the TALED technology, we used validated TALE arrays for mtDNA-encoded mouse Mt-Cytb, Mt-CoII and Mt-Atp6 from our previous work [9]. These TALEs are all known to bind the desired mtDNA position and induce high levels of cytosine base edits, and therefore acted as a good starting point to test the adenine base editing technology in vitro. Each of the TALE pair was cloned into plasmids containing the different DddAtox(E1357A)/TadA8e variations. This resulted in a total of 6 combinations which were transfected into WT NIH-3T3 mouse cells, sorted and replated to grow for 6 days. The levels of detectable editing from Sanger sequencing varied significantly (Supplementary Figs. 2-4), which was quantified using next generation sequencing (NGS) (Fig. 1b). For Mt-Cytb, low editing levels were observed in one of the dimeric DddAtox(E1357A) and TadA8e combinations, and also for the monomeric H strand binding construct (Fig. 1b). In the case of Mt-CoII, the dimeric DddAtox and TadA8e version showed editing alongside the split DddAtox(E1357A) dimeric combination (Fig. 1b). Finally, Mt-Atp6 showed similar editing in both dimeric, split DddAtox(E1357A) and monomeric architectures (Fig. 1b). The highest editing efficiency we achieved after 7 days in vitro was 17%, comparable to previous reports for mouse NIH-3T3 cells, albeit lower than that observed in human cell A-to-G editing. [21], [22], [23]. In the cases where editing was observed, it was present at multiple A:T sites within/across the editing window rather than being specific to predominantly one position as with the DdCBE technology [9]. The NGS analysis also allowed for mtDNA-wide off-target editing frequency to be calculated. The off-target editing was approximately 0.1% for all constructs which falls within the variation in unedited WT cells (Fig. 1c), suggesting generally lower activity of TALEDs, as compared to DdCBEs [9]. The off-targets for these constructs are therefore not concerning, although the on-target levels are relatively low and if they increased, the off-targets would likely follow.

Fig. 1

We next decided to focus on Mt-Atp6, as we previously successfully edited this gene in vitro and in vivo using cytosine base editors [9]. We tested the strand selective mitochondrial base editors (mitoBEs), which use nickase enzymes to create the single stranded DNA required for TadA8e activity (Fig. 2a, Supplementary Fig. 1) [22]. We used the Nt.BspD61(C) nickase due to its lack of sequence specificity, unlike the MutH* nickase used in the previous study, which requires a 5’-GAT-3’ sequence not present in the Mt-Atp6 target window. Following the NIH-3T3 cells transfection and selection (see above), the NGS analysis revealed up to 4% editing at multiple sites withing the targeting window between the TALE biding sites when using the mitoBEs (Fig. 2b, Supplementary Fig. 5), whilst mtDNA-wide off-targets were not significantly above control levels (Fig. 2c). Taken together, our study replicated the outcomes documented for both TALEDs and mitoBEs, with the efficiencies and editing patterns observed in mouse cells being largely consistent with those reported previously [21], [22], [23].

Fig. 2.

Adenine base editing in vivo

Next, we intended to investigate if adenine base editing could be used to edit mtDNA in the post-mitotic tissues of neonate mice, as reported by us previously for various cytosine editors [15], [16]. Informed by our in vitro data, we decided to test the best performing Mt-Atp6 constructs in dimeric architecture using TadA8e with the H-strand binding TALE and either the inactive DddAtox(E1357A) (the TALED architecture) or nickase Nt.BspD61(C) (the mitoBE architecture) with the L-strand binding TALE. We cloned the respective DNA sequences in a AAV-compatible vector and proceeded with encapsidation according to our standard AAV production approach [15], [24]. However the titres for the H-strand TadA8e construct were approximately 1000-fold lower than for the L-strand DddAtox(E1357A) or L Nt.BspD61(C), which proved insufficient for any mouse injections. We repeated the procedure at two different facilities with similar results. During the course of these experiments a study was published addressing the significant RNA off-targets of the TadA8e enzyme greatly affecting cell viability upon transfection, therefore, we hypothesized that this was likely the reason for the failure to produce AAVs containing the TadA8e enzyme [23]. The study engineered the TadA8e protein to reduce RNA off-target edits, with the V28R mutation preventing >99% RNA-induced off-targeting and leading to increased cell viability [23]. We installed the V28R mutation into the TadA8e AAV plasmid and tested the effect of TadA8e(V28R) on cells following transfection of HEK293T and comparing it to the unmodified TadA8e AAV plasmid. Consistent with the previously reported data, we observed substantial (up to ∼90%) cell death with the unmodified TadA8e, which was alleviated following transfection with the TadA8e(V28R) mutant (Supplementary Fig. 6). The generation of TadA8e(V28R) enabled us to produce sufficient AAV titres for subsequent in vivo experiments. We then injected neonatal subjects (postnatal P1) with Mt-Atp6 H-strand TadA8e(V28R) AAV9 with either DddAtox(E1357A) or BspD61(C) AAV9 via temporal vein injection at 5 × 10¹¹ vg per monomer per animal (Fig. 3a). Upon sacrificing the animals at 4-weeks post-injection, we confirmed successful AAV delivery through robust expression in mouse cardiac and skeletal muscle tissue (Supplementary Fig. 7 and 8). At this time-point we observed minimal editing of mtDNA in mouse heart, with NGS revealing up to 0.5% of A:T to G:C edits in 3 replicate mice (Fig. 3b-c). However, we did not detect any editing in either skeletal muscle or liver tissues, as compared to control animals (Fig. 3b-c). We did not detect any mtDNA-wide off-targets above background, which was consistent with the low levels of edits observed (Fig. 3d). We further did not observe any significant changes in mtDNA copy number in the AAV-treated mice as compared to the vehicle-injected controls (Fig. 3e). We also did not detect any changes in OXPHOS protein steady state levels (Supplementary Fig. 9 and 10). Taken together, these findings confirm that adenine base editing of mtDNA in vivo in post-mitotic tissues is feasible upon AAV delivery, but the observed levels are notably low after a 4-week exposure period.

Fig. 3

Conclusions

In this study, we tested the effect of adenine base editors first in cells using our TALEs targeted to Mt-CytB, Mt-CoII and Mt-Atp6 and confirmed editing within the target window. We then tested the effect of Mt-Atp6 adenine base editors in mice using AAV delivery to neonate mice and observed negligible editing in mouse hearts, with no editing in other tissues. In comparable experimental conditions, TALE-based DdCBE and ZF-DdCBEs installed up to 30% and 50% of C:G to T:A edits, respectively [15], [16]. Considering the limited editing efficiency and the absence of single-nucleotide specificity within the editing window, current adenine base editors need further optimization to become a valuable tool for generating or correcting pathogenic in vivo point mutations associated with mitochondrial diseases.

Data availability

The data supporting the findings of this study are available within the paper and its supplementary information files, apart from proprietary scripts which are available upon request by contacting the authors. The NGS files generated in this study have been deposited in GEO: GSE276406: Source data are provided with this paper. The materials used in this study are available upon request from M.M.

Supplementary material

Supplementary Figure 1:

Supplementary Figure 2:

Supplementary Figure 3:

Supplementary Figure 4:

Supplementary Figure 5:

Supplementary Figure 6:

Supplementary Figure 7:

Supplementary Figure 8:

Supplementary Figure 9:

Supplementary Figure 10:

Acknowledgements

We would like to acknowledge the members of the Mitochondrial Genetics Group (MRC-MBU, University of Cambridge) for useful discussion during the course of this research. We would also like to thank the CIMR Flow facility for their support in this work.

Additional information

Funding

This work was supported by core funding from UKRI Medical Research Council UK (MRC) (MC_UU_00028/3), the UKRI MRC award MC_PC_21046 to the National Mouse Genetics Network Mitochondria Cluster (MitoCluster), AFM-Téléthon (#24237) and a research grant from the CureMito Foundation. P.S.-P. is additionally supported by The Champ Foundation.

Authors’ contributions

C.D.M., P.S.-P. and M.M. planned and designed experiments; C.D.M. performed most of the experiments; L.V.H. analysed the NGS experiments; L.L.G. was involved in the NGS experiments, P.S.-P. performed the AAV in vivo experiments; K.T. managed the animal work. C.D.M., P.S.-P. and M.M. drafted the manuscript. P.S.-P. and M.M. supervised the study and acquired funds. All authors revised the manuscript.