Using temporary gene editing to treat metabolic liver disease in mice

In a previous study, we attempted to correct the pathogenic mutation in the Pah^enu2 mouse model for PKU using AAV-mediated delivery of intein-split PE2 and a pegRNA targeting Pah^enu2 (mPKU-2.1)¹⁷. However, despite applying AAV doses of 1 × 10¹⁴ vector genomes (vg) per kg, correction rates remained below 1%, and treatment was only successful when prime editing components were delivered from adenoviral vectors at doses that are not viable in a clinical context. This prompted us to test whether recent improvements in the prime editing technology could increase correction rates of the Pah^enu2 mutation.

We exchanged PE2 with PEmax²³—a PE variant with optimized codon usage and nuclear localization signal-linker designs—and incorporated pseudo-knot structures to the 3′ end of the pegRNA for protection from exonucleases (epegRNAs)^24,25. In addition, we modified the RTT sequence of the pegRNA to co-introduce a silent mutation in the GG sequence of the protospacer adjacent motif (PAM), preventing retargeting of the locus once the edit is installed (Supplementary Fig. 1b). When transfected into HEK293T cells that contain an ectopic copy of the murine Pah^enu2 locus (Supplementary Fig. 1a), the adapted editing components increased correction rates and reduced insertion–deletion (indel) rates to background (Fig. 1a–c).

Next, we assembled AAV vectors for in vivo delivery of the optimized PE components (PEmax + tevopreQ₁-mPKU-SM) into Pah^enu2 mice. As the size of the PE exceeds the packaging capacity of AAV, we used the intein-split system to split Cas9 and express the PE from two separate AAVs^14,15,17. Testing two intein-split designs for PEmax (1153/1154 and 713/714), we observed comparable activity in HEK293T cells (Supplementary Fig. 1c). However, we selected the 713/714 variant for our in vivo experiments as it facilitates the generation of AAV constructs that are more equal in size when the non-essential RNaseH domain of the reverse transcriptase is removed and the pegRNA is positioned on the vector containing the N-intein. The tevopreQ₁-mPKU-SM pegRNA was cloned downstream of the human U6 promoter, and N- and C-terminal fragments of PEmax were cloned between the liver-specific P3 promoter²⁶ and a 3′ UTR containing the W3 post-transcriptional regulatory element²⁷ and the simian virus (SV40)-polyA tail (Fig. 1d and Supplementary Fig. 1d). Recombinant AAV2 genomes were then packaged into hepatotropic AAV serotype 9 capsids (AAV2/9) and systemically administered to Pah^enu2 mice via the tail vein in a 1:1 ratio at a final dose of 1 × 10¹⁴ vg kg⁻¹ (Fig. 1e). Analysis of isolated hepatocytes by next-generation sequencing (NGS) after a period of 7 weeks revealed 4.6% correction of the Pah^enu2 mutation (Fig. 1f), which resulted in a reduction of blood Phe levels from over 1,500 µmol l⁻¹ to 623.7 µmol l⁻¹ (Fig. 1g and Supplementary Fig. 2a). In line with previous genome editing studies that used AAV9 vectors in combination with the hepatocyte-specific P3 promoter, we found that editing was largely limited to hepatocytes^8,26,28 (Supplementary Fig. 2b). Furthermore, we did not detect unintended editing in the liver of treated mice at the top five experimentally identified off-target binding sites of the mPKU-SM pegRNA¹⁷ (Supplementary Fig. 2c,d).

While Pah^enu2 correction rates with optimized AAV vectors and prime editing components were substantially increased compared to our initial study²², they were not sufficient to reduce Phe levels below the therapeutic threshold (600 µmol l⁻¹ for adult PKU patients and 360 µmol l⁻¹ for children and pregnant women)²⁹. In addition, AAV doses of 1 × 10¹⁴ vg kg⁻¹ can lead to severe immune reactions^30,31, and prolonged PE expression from AAV vectors could lead to an accumulation of off-target mutations or induction of immune responses to the bacterial Cas9 or viral reverse transcriptase.

In previous studies, LNP-mediated mRNA and sgRNA delivery has been used for transient genome editing with Cas9 nucleases and base editors^8,28,32–36. To assess whether a similar approach is feasible for prime editing, we generated nucleoside-modified mRNA^37,38 encoding PE2 and PEmax and packaged it into LNP. Confirming transient liver expression after systemic delivery of 2 mg kg⁻¹ mRNA–LNP, we observed a peak in PE mRNA levels at 6 h post injection (h.p.i.) and a peak in protein levels at 24 h.p.i., whereas at 45 h.p.i. neither PE mRNA nor PE protein levels were detectable anymore (Fig. 2a and Supplementary Fig. 3a,b). Next, we chemically synthesized the tevopreQ₁-mPKU-SM pegRNA targeting the Pah^enu2 locus, and a pegRNA that installs a G-to-C edit at the Dnmt1 locus. This pegRNA has previously been used for in vivo prime editing in the liver using AAV- and adenoviral-mediated delivery¹⁷. To protect pegRNAs from exonucleases, they were modified with 2′-O-methyl-3′-phosphorothioate at the 5′ end and 2′-O-methyl-3′-phosphonoacetate at the 3′ end³⁹. In addition, we generated pegRNAs in which the RTT was substituted with deoxyribose nucleotides for protection from endonucleases (DNA-mod pegRNAs)⁴⁰ (Supplementary Table 4). Confirming functionality of the synthesized pegRNAs, co-electroporation with PEmax mRNA into HEK reporter cells resulted in 25% editing at the Pah^enu2 site and 37% editing at the Dnmt1 site using non-DNA modified pegRNAs, and 3% editing at the Pah^enu2 site and 21% editing at the Dnmt1 site using DNA-modified pegRNAs (Fig. 2b).

We then administered PEmax mRNA and pegRNA into mice using a repeated dosing scheme, in which a dose of 2 mg kg⁻¹ LNP containing PEmax mRNA was followed 14 h later by a dose 2 mg kg⁻¹ LNP containing the respective pegRNA. Although this dosing scheme ensured the presence of PE protein at the time of pegRNA delivery (Fig. 2a), editing rates remained at 1.4% at the Dnmt1 locus (after two dosing rounds) and at 0.4% at the Pah^enu2 locus (after three dosing rounds; Fig. 2c,d). Hypothesizing that toxicity or innate immune stimulation from the initial mRNA–LNP dose might have prevented hepatocytes from taking up the pegRNA–LNP¹⁹, we tested additional dosing regimens delivering 2′-O-methyl-3′-phosphorothioate end-protected pegRNA–LNP targeting Dnmt1 simultaneously, 4 h or 14 h after the mRNA–LNP dose. It is worth noting that NGS analysis showed highest editing in mice injected simultaneously with mRNA–LNP and pegRNA–LNP (3.5% editing after a single dose and 10.5% editing when mice are re-dosed after 5 days; Fig. 2e,f). Therefore, we next also treated PKU mice with this dosing regimen. However, even though mice were re-dosed three times in a 5 day interval, editing rates remained marginal (0.1%; Fig. 2e,f).

We hypothesized that prime editing efficiency in the RNA–LNP approach could have been limited by the low abundance and rapid degradation of the pegRNA. Therefore, we also tested an alternative strategy where only the PE was transiently delivered via mRNA–LNP and the pegRNA was expressed from a DNA construct delivered via AAV. PKU mice were first administered with 2.5 × 10¹³ vg kg⁻¹ self-complementary AAV2/9 (scAAV2/9) encoding the tevopreQ₁-mPKU-SM pegRNA (Supplementary Fig. 4a), followed by a treatment with PE2 mRNA–LNP or PEmax mRNA–LNP at a dose of 3 mg kg⁻¹ (Fig. 3a). In addition, wild-type mice pretreated with an scAAV2/9 encoding the tevopreQ1-modified Dnmt1 pegRNA were injected with a single dose of 3 mg kg⁻¹ mRNA–LNP (Fig. 3a). While editing rates were relatively low at Pah^enu2 locus (1.1% with PE2 and 1.0% with PEmax), at the Dnmt1 locus we observed 17.7% editing with PE2 and 26.2% editing with PEmax (Fig. 3b,c). As expected for a treatment that installs a base change on genomic DNA, editing rates were not decreased when mice were analysed after 4 months compared to 1 week after PE delivery (Supplementary Fig. 4b). Moreover, editing efficiency did not decrease when the scAAV2/9 dose was reduced from 2.5 × 10¹³ vg kg⁻¹ to 1 × 10¹² vg kg⁻¹ (26.2% versus 20%; Fig. 3d) or when mRNA–LNP levels were reduced from 3 mg kg⁻¹ to 2 mg kg⁻¹ (26.2% versus 24%; Fig. 3b,e). It is worth noting that recent studies suggested that accessibility of deoxyribonucleoside triphosphates could be a limiting factor for prime editing in slowly dividing cells such as resting T cells or haematopoietic stem and progenitor cells (HSPCs)^41,42, prompting us to also test an approach where we co-delivered PEmax mRNA with an mRNA encoding VPX—a lentiviral factor that increases deoxyribonucleoside triphosphate levels in quiescent cells by targeting SAMHD1 for degradation. However, in our experimental set-up, we did not observe an increase in prime editing efficiency at the Dnmt1 locus compared to control mice where mRNA encoding inactive VPX^Q76A was co-delivered with the PE (Fig. 3e).

As correction rates at the Pah^enu2 locus were not sufficient to substantially reduce Phe levels, we next re-dosed PKU mice that were pretreated with pegRNA-AAV three times with 2 mg kg⁻¹ mRNA–LNP (Fig. 4a). It is worth noting that this redosing regimen increased editing rates up to 6.2% with PE2 (average 2.9%) and up to 8.7% with PEmax (average 4.3%; Fig. 4b), leading to a reduction of blood Phe levels below 600 µmol l⁻¹ and a corresponding increase in PAH enzyme activity in the liver (Fig. 4c–e). Similar to the Dnmt1 locus, also at the Pah^enu2 locus editing rates did not substantially decrease when the scAAV2/9 dose was lowered from 2.5 × 10¹³ vg kg⁻¹ to 5 × 10¹² vg kg⁻¹ (Supplementary Fig. 4c).

While in adults Phe levels below 600 µmol l⁻¹ are considered therapeutic, in children and pregnant women Phe levels should be kept below 360 µmol l⁻¹ to avoid neurotoxicity and cognitive impairment. Therefore, we attempted to further increase the correction efficiency of the Pah^enu allele by (1) designing a pegRNA that encodes for additional silent bystander mutations and that was computationally predicted by PRIDICT2.0 (ref. ⁴³) to result in higher editing than pegRNA mPKU-SM (Supplementary Fig. 5a) and by (2) using a novel PE, PE7, in which the small RNA-binding factor La is fused to the PE to protect pegRNAs from exonucleases⁴⁴. While the pegRNA encoding an additional silent mutation did not increase editing rates in vitro in cell lines (Supplementary Fig. 5b), or in vivo after three doses of 2 mg kg⁻¹ mRNA–LNP with the full LNP-delivery approach (4.4% (Fig. 5b) versus 4.3% with pegRNA mPKU-SM (Fig. 4b)), the use of PE7 had a strong effect on editing efficiencies. We pretreated the mice with scAAV2/9 at a dose of 2.5 × 10¹³ vg kg⁻¹ expressing the pegRNA tevopreQ1-Dnmt1 or the pegRNA tevopreQ1-mPKU-SM4 for targeting the Dnmt1 locus or the Pah^enu2, respectively. We obtained 47.4% editing at the Dnmt1 locus after a single dose of 2 mg kg⁻¹ mRNA–LNP (Fig. 5a) and 20.7% editing at the Pah^enu2 locus after three doses of 2 mg kg⁻¹ mRNA–LNP (Fig. 5b), which was sufficient to reduce Phe levels below the therapeutic threshold of 360 µmol l⁻¹ (Fig. 5c).

To next evaluate the PE7 system in a full RNA–LNP approach, we combined PE7 mRNA with synthetic La-accessible pegRNAs containing a polyU stretch (U*mU*mU*mUU) for La binding at their 3′ end. While this approach was less efficient than the dual AAV–LNP approach, treatment of mice with 2 mg kg⁻¹ PE7 mRNA–LNP and 2 mg kg⁻¹Dnmt1 pegRNA–LNP (2 doses with a 5 day interval between dosing) resulted in 35.9% editing (Fig. 6a), representing a 3.4-fold increase compared to the full RNA–LNP approach with PEmax (Fig. 2f). Moreover, in PKU mice treatment with 2 mg kg⁻¹ PE7 mRNA–LNP and 2 mg kg⁻¹ mPKU-SM pegRNA–LNP (three doses with a 5 day interval between dosing) resulted in 8.0% correction of the Pah^enu2 mutation (Fig. 6b). This marks a substantial increase compared to the less than 0.5% editing observed with PEmax (Fig. 2d,f) and effectively reduced Phe levels to 342 µmol l⁻¹.

PKU can be caused by various mutations in the PAH gene. We therefore next determined whether other pathogenic mutations can be corrected with similar editing efficiencies than the Pah^enu2 allele. Using PRIDICT2.0, we predicted the most effective pegRNA designs for the 11 most common PKU mutations (Supplementary Fig. 5c) and selected the top three pegRNA designs per mutation for testing in K562 reporter cells, which are mismatch repair proficient and show pegRNA efficiency patterns similar to hepatocytes⁴³. It is worth noting that for 8 out of the 11 tested mutations we observed editing rates that surpassed those achieved for the Pah^enu2 mutation (Fig. 6d and Supplementary Fig. 5c), suggesting that in vivo prime editing using PE7 mRNA–LNP may offer a promising avenue for therapeutic application in PKU patients.

To assess whether prime editing using the dual AAV–LNP or full RNA–LNP approach was limited to the liver, DNA of treated mice was isolated from different organs and analysed by NGS. Consistent with previous mRNA–LNP biodistribution studies⁴⁵, we did not observe substantial editing in any of the analysed tissues in Pah^enu2-targeted or Dnmt1-targeted mice with the dual AAV–LNP or the full RNA–LNP approach (Fig. 7a,d and Supplementary Fig. 6a,b). Next, we assessed whether editing occurred at other sites in the genome in the liver of treated mice. However, when we performed targeted amplicon sequencing at the top five off-target binding sites of the two pegRNAs (identified by circularization for high-throughput analysis of nuclease genome-wide effects by sequencing (CHANGE-seq)¹⁷; Supplementary Figs. 2c and 6c), we did not observe editing above background—neither in the AAV–LNP approach nor in the full RNA–LNP approach (Fig. 7b,c,e,f).

Finally, we examined whether delivery of mRNA–LNP, pegRNA–LNP or pegRNA-AAV triggered liver toxicity or innate immune responses. A mild elevation of alanine transaminase (ALT) was observed 6 h after administration of 2 mg kg⁻¹ PE mRNA–LNP or pegRNA–LNP, but levels returned to baseline levels at 45 h.p.i. or 1 week after administration, respectively (Fig. 7g,h). Similarly, transient increases in proinflammatory cytokines, such as MCP1 and IFNγ, were detected but normalized at later time points (Fig. 7i–l). When administering scAAV2/9 encoding the pegRNA at a dose of 2.5 × 10¹³ vg kg⁻¹, we did not observe significant increases in ALT, aspartate aminotransferase (AST), or proinflammatory cytokines at any of the tested time points (Fig. 7g–l). In line with these observations, histological examination of the liver of treated mice did not reveal any obvious signs of tissue necrosis (Supplementary Fig. 6d).

To generate epegRNA plasmids, annealed spacer, scaffold and 3′ extension oligos (Supplementary Table) were cloned into pU6-tevopreq1-GG-acceptor (Addgene number 174038) by Golden Gate Assembly. For the generation of split-intein PEmax constructs, inserts were PCR amplified from the epegRNA plasmids and from pCMV-PEmax (Addgene number 174820) and inserted into the respective backbones (Addgene numbers 187181 and 117182) using HiFi DNA Assembly Master Mix (New England Biolabs, NEB). To generate PiggyBac PKU reporter plasmids, inserts with homology overhangs for cloning were ordered from IDT and cloned into the pPB-Zeocin backbone using HiFi DNA Assembly Master Mix (NEB). All PCRs were performed using Q5 High-Fidelity DNA Polymerase (NEB). 2

HEK293T (American Type Culture Collection, CRL-3216) or K562 (American Type Culture Collection, CCL-243) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) plus GlutaMAX (Thermo Fisher Scientific) or Roswell Park Memorial Institute 1640 medium (Thermo Fisher Scientific), respectively, and supplemented with 10% (v/v) fetal bovine serum and 1% penicillin/streptomycin (Thermo Fisher Scientific) at 37 °C and 5% CO₂. Cells were maintained at confluency below 90% (HEK293T) or below a density of 1.8 × 10⁶ cells per ml (K562) and seeded into 48-well cell culture plates (Greiner). About 1 × 10⁵ cells were transfected using 1.5 μl of Lipofectamine 2000 (Thermo Fisher Scientific) with 375 ng of PEmax and 125 ng of epegRNA or 250 ng of each AAV construct according to the manufacturer’s instructions. Unless otherwise noted, cells were incubated for 3 days, and genomic DNA was isolated by direct lysis. For nucleofections 0.5 pmol of the PEmax mRNA were used in a 1:1 molar ratio with the pegRNAs. Nucleofections of mRNA were carried out with one pulse of 1,400 mV and 20 ms pulse width using the Neon Nucleofector electroporation system by Invitrogen. After nucleofection, cells were cultured in 200 μl of DMEM plus GlutaMAX for 48 h before isolation of genomic DNA by direct lysis.

The unmodified pegRNA was generated by in vitro transcription as previously described, using a synthetic DNA fragment (Supplementary Table). The mRNA transcription template was prepared by PCR using a forward primer to introduce the correct T7 promoter sequence and a reverse primer to incorporate the poly(A) tail, with the synthetic DNA fragment serving as the PCR template. The mRNA was then transcribed from this PCR-derived template using the HiScribe T7 High Yield RNA Synthesis Kit (NEB). 4

For generation of the PKU reporter cell lines with the PiggyBac transposon, 75,000 HEK293T or K562 cells were seeded into a 24-well culture plate (Greiner) and transfected the next day using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer’s instructions. In brief, 1.5 μl of Lipofectamine was mixed with 23.5 μl of Opti-MEM medium, incubated at room temperature for 10 min and added to 225 ng of transposon plasmid and 25 ng of transposon helper plasmid (filled up to 25 μl of Opti-MEM). After 30 min of incubation at room temperature, cells were transfected. Three days after transfection, cells were enriched for 10 days using Zeocin (InvivoGen, 150 μg ml⁻¹) selection with a final concentration of 0.1 μg ml⁻¹.

Pseudo-typed AAV9 vectors (AAV2/9) were produced by co-transfection of packaging, capsid and helper plasmids (Addgene numbers 112865 and 112867), which were incubated for 5 days until collection. The cells were pelleted by low-speed centrifugation (1,500g, 15 min). The medium was then decanted, and the resulting cell pellet was resuspended in resuspension buffer (150 mM NaCl, 50 mM Tris–HCl), transferred into a Lysis Kit tube (Precellys Evolution) and mechanically disrupted according to the manufacturer’s instructions. Subsequently, the samples were lysed with benzonase (Sigma-Aldrich) at 37 °C for 30 min, followed by centrifugation at 5,000g for 1 h at 4 °C. The supernatant from this centrifugation step was combined with the supernatant from the previous step and precipitated using PEG8000 and NaCl solutions (1–2 days, 4 °C). The precipitated material was then pelleted by medium-speed centrifugation (5,000g, 1 h). The pellet was resuspended in resuspension buffer (final volume of 6 ml), and 1.5 ml of 5 M NaCl was added. The AAV particles were further purified by gradient centrifugation using OptiPrep (Sigma-Aldrich) according to the manufacturer’s instructions. Finally, the purified AAVs were concentrated using Vivaspin 20 centrifugal concentrators (VWR) following the manufacturer’s protocol.

Physical titres (vector genomes per millilitre) were determined using a Qubit 3.0 Fluorometer. In brief, the Qubit Fluorometer 3.0 (Life Technologies) was used to measure the concentrations (ng ml⁻¹) of the extracted genomes by denaturation at 95 °C for 5 min, after which the readings were converted to vector genomes per millilitre using the genome’s molecular mass and Avogadro’s constant. The identity of the packaged genomes of each AAV vector was confirmed by Sanger DNA sequencing by Microsynth AG, testing 500 ng of denatured AAV using an AAV-genome-specific sequencing primer. AAV2/9 viruses were stored at −80 °C until use and diluted with phosphate-buffered saline (PBS; Thermo Fisher Scientific) if necessary.

mRNA production was performed as previously described⁵¹. In short, the coding sequences of the PEs were cloned into the mRNA production plasmid using HiFi DNA Assembly Master Mix (NEB). mRNAs were transcribed to contain 101 nucleotide-long poly(A) tails. m1Ψ-5′-triphosphate (TriLink) instead of UTP was used to generate modified nucleoside-containing mRNA. Capping of the in vitro transcribed mRNAs was performed co-transcriptionally using the trinucleotide cap1 analogue, CleanCap (TriLink). mRNA was purified by cellulose (Sigma-Aldrich) purification as previously described⁵². All mRNAs were analysed by agarose gel electrophoresis and were stored frozen at −20 °C. Synthetic pegRNAs were ordered and synthesized by Axolabs (peg-mod_1) or Agilent (peg-mod_2-4). LNP were formulated as previously described⁵³. In short, an ethanolic solution of 1,2-distearoyl-sn-glycero-3-phosphocholine, cholesterol, a PEG lipid and an ionizable cationic lipid was rapidly mixed with an aqueous solution (pH 4) containing PE mRNA using an in-line mixer. The PEG and ionizable lipid used in this study are described in patent application WO 2017/004143. The resulting LNP formulation was dialysed overnight against 1× PBS, 0.2 μm sterile filtered and stored at −80 °C at a concentration of 1 μg μl⁻¹ of total RNA. Encapsulation efficiencies of mRNA in the LNP were >97% as measured by the Quant-iT Ribogreen Assay (Life Technologies), and LNP sizes were below 80 nm as measured by a Malvern Zetasizer (Malvern Panalytical). Acuitas Therapeutics will provide the LNP formulation used in this study to any academic researcher who wishes to test and replicate our findings.

Animal experiments were performed in accordance with protocols approved by the Kantonales Veterinäramt Zürich and in compliance with all relevant ethical regulations. B6.BTBR-Pah^enu2 (strain number 029218) and C57BL/6J (strain number 000664) mice were housed in a pathogen-free animal facility at the Institute of Pharmacology and Toxicology of the University of Zurich. Mice were kept in a temperature- and humidity-controlled room on a 12 h light–dark cycle. Mice were fed a standard laboratory chow (Kliba Nafag number 3437 with 18.5% crude protein) and genotyped at weaning. Heterozygous Pah^enu2 littermates were used as controls for physiological Phe concentrations in the blood (<120 μmol l⁻¹). For sampling of blood for Phe determination, mice were fasted for 3 to 4 h, and the blood was collected from the tail vein. Adult mice were injected with 5–10 × 10¹³ vg kg⁻¹ (AAV) or with 1–3 mg kg⁻¹ RNA–LNP in a maximal volume of 150 µl via the tail vein. The selected AAV and RNA–LNP doses were based on the maximum injection volume for adults (150 μl of undiluted viral vectors via the tail vein). Mice were injected with AAVs and/or LNPs via the tail vein from 6 to 12 weeks of age.

Primary hepatocytes were isolated using a two-step perfusion method. Briefly, pre-perfusion with Hanks’ buffer (supplemented with 0.5 mM EDTA and 25 mM Hepes) was performed by inserting the cannula through the superior vena cava and cutting the portal vein. Next, livers were perfused at low flow for about 10 min with digestion buffer (low-glucose DMEM supplemented with 1 mM HEPES) containing freshly added Liberase (32 μg ml⁻¹; Roche). Digestion was stopped using isolation buffer (low-glucose DMEM supplemented with 10% fetal bovine serum), and cells were separated from the matrix by gently pushing with a cell scraper. The cell suspension was filtered through a 100 μm filter (Corning), and hepatocytes were purified by two low-speed centrifugation steps (50g for 2 min).

Genomic DNA from mouse tissues were isolated by direct lysis. Locus-specific primers were used to generate targeted amplicons for deep sequencing. First, input genomic DNA was amplified in a 10 μl reaction for 26 cycles using NEBNext High-Fidelity 2× PCR Master Mix (NEB). PCR products were purified using Sera-Mag magnetic beads (Cytiva) and subsequently amplified for six cycles using primers with sequencing adapters. Approximately equal amounts of PCR products from each sample were pooled, gel purified and quantified using a Qubit 3.0 fluorometer and the dsDNA HS Assay Kit (Thermo Fisher Scientific). Paired-end sequencing of purified libraries was performed on an Illumina MiSeq.

Sequencing reads were demultiplexed using MiSeq Reporter (Illumina). The NGS data were collected and demultiplexed by Illumina NovaSeq Control software (v1.7) and MiSeq Control software (v3.1 and v4.0). Data analysis and visualization was performed using Python v3.9 and v3.10. Cutadapt (v3.1) was used to trim sequencing reads. For characterization of endogenous editing and unintended edits, CRISPResso2 (v2.2.7) was used⁵⁴. Prime editing efficiencies were calculated as percentage of (number of reads containing only the desired edit)/(number of total reads). Indel yields were calculated as percentage of (number of indel-containing reads)/(total reads).

Amino acids were extracted from a 3.2 mm dried blood sample using the Neomass AAAC Plus newborn screening kit (Labsystems Diagnostics). An ultra-high performance liquid chromatography system Nexera X2 coupled to an LCMS-8060 triple quadrupole mass spectrometer with electrospray ionization (Shimadzu) was used for the quantitative analysis of phenylalanine. LabSolutions v.6.115 and Neonatal Solution software (Shimadzu) were used for data acquisition and data analysis.

Whole-liver extracts were analysed using isotope-dilution liquid chromatography–electrospray ionization–tandem mass spectrometry⁵⁵. In brief, Phe (1 mmol l⁻¹) was pre-incubated at 25 °C for 4 min, then Fe(NH₄)₂(SO₄)₂ (100 μmol l⁻¹) was added, and incubation was continued for one more minute. After 5 min total pre-incubation time, BH₄ (75 μmol l⁻¹ for mouse tissue) was added to start the reaction. Between 2.5 and 20 μl (2–165 μg) of total protein lysate extracted from mouse liver tissue was used for activity measurements. The applied total protein amount depended on mouse liver tissue samples. It was generally higher for sample types with low activity where only little amounts of Tyr were produced. Initially, this was determined empirically by measuring a series of increasing total protein amounts for each sample type to determine the measurable linear range. Reaction time was 2 min; short incubation time reduces any possible chaperone-like effect of BH₄. The amount of Tyr produced was determined by liquid chromatography with electrospray ionization–tandem mass spectrometry.

Proteins were isolated from liver samples of treated and untreated mice. In brief, cells were lysed in RIPA buffer, supplemented with protease inhibitors (Sigma-Aldrich). Protein amounts were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher). Equal amounts of protein (80 μg) were separated by SDS–PAGE (Thermo Fisher) and transferred to a 0.45 μm nitrocellulose membrane (Amersham). Membranes were incubated with mouse anti-Cas9 (1:1,000; catalogue number 14697T; Cell Signaling) and rabbit anti-GAPDH (1:10,000; catalogue number ab181602; abcam). Signals were detected by fluorescence using IRDye-conjugated secondary antibodies; IRDye 800CW Goat anti-Rabbit IgG Secondary Antibody (catalogue number 926-32211, LI-COR bio, 1:15,000) and IRDye 680RD Goat anti-Mouse IgG Secondary Antibody (catalogue number 926-68070, LI-COR bio, 1:15,000).

RNA was isolated from shock-frozen liver samples using the RNeasy Mini kit (Qiagen) according to the manufacturer’s instructions. RNA was reverse transcribed to cDNA using random primers and GoScript Reverse Transcriptase kit (Promega). Reverse transcription quantitative PCR (RT–qPCR) was performed using Firepol qPCR Master Mix (Solis BioDyne) and analysed by 7900HT Fast Real-Time PCR System (Applied Biosystems). Fold changes were calculated using the delta C_t method. Used primers are listed in Supplementary Table 3.

During liver perfusion, one liver lobe was tightened off using a silk suture thread (Fine Science Tools). Tissues were transferred to a 30% sucrose solution overnight at 4 °C and embedded in optimal cutting temperature (OCT) compound in cryomolds (Tissue-Tek) and frozen at −80 °C for at least 30 min. Frozen tissues were sectioned at 7 µm at −20 °C and mounted directly on Superfrost Plus slides (Thermo Fisher Scientific). Cryosections were counterstained with DAPI (Thermo Fisher Scientific) and mounted in VECTASHIELD mounting medium (Vector Labs). Two frozen sections per mouse per tissue were analysed. Mouse tissue was imaged using Zeiss Axioscope and Colibri 7 LED Illumination lighting system. Imaging conditions and intensity scales were matched for all images. Images were taken using Zeiss software Zen2 and analysed by Fiji ImageJ software (v1.51n)⁵⁶.

Livers were fixed using 4% paraformaldehyde (Sigma-Aldrich), followed by ethanol dehydration and paraffinization. Paraffin blocks were cut into 5-μm-thick sections, deparaffinized with xylene and rehydrated. Sections were haematoxylin–eosin-stained and examined for histopathological changes.

Blood was collected from the inferior vena cava using lithium–heparin-coated 0.5 ml tubes (MiniCollect) before liver perfusion. Samples were centrifuged at 2,000g for 10 min, and the supernatant was collected and stored at −20 °C until measurement. AST and ALT levels from all mouse samples were measured as routine parameters at the Division of Clinical Chemistry and Biochemistry at the University Children’s Hospital Zurich using Alinity ci-series. Pro-inflammatory cytokines were detected using LEGENDplex Mouse Inflammation panel (13-plex; Biolegend, catalogue number 740446, lot number B354399), a bead-based multiplex assay, according to the manufacturer’s instructions.

Statistical analyses were performed using SciPy (1.6.3) amd GraphPad Prism 9.0.0 for macOS. Data are represented as biological replicates and are depicted as mean ± s.d. or s.e.m. as indicated in the corresponding figure legends. Likewise, sample sizes and the statistical tests used are described in detail in the respective figure legends. For all analyses, P < 0.05 was considered statistically significant.

Further information on research design is available in thelinked to this article. Nature Portfolio Reporting Summary

Nature Biomedical Engineering thanks Hyongbum Kim and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

All data associated with this study are present in the paper or the Supplementary information. Illumina sequencing data are available in the Sequence Read Archive under the accession number PRJNA947564↗ and PRJNA1258720↗.