Nature biomedical engineering

Base editing in living mice improves liver problems and cell cleaning defects in a model of Zellweger spectrum disorder

Updated

Abstract

Essence

Base editing corrected PEX1 disease alleles and improved peroxisome-related liver pathology in a Zellweger spectrum disorder mouse model.

Evidence

A preclinical in vivo and ex vivo base-editing study treated homozygous Pex1-p.G844D mice and patient-derived fibroblasts, measuring allele correction, peroxisome metabolites, liver transcriptomes and histology, body weight, and off-target editing.

Caveat

The findings are preclinical and centered on liver and fibroblast endpoints, so they do not show human clinical benefit or full neurologic rescue.

Simplified

Full Text

Results

The-p.G843D allele is associated with liver disease in individuals with ZSD PEX1

To better understand the natural course of ZSD liver disease caused by the PEX1-p.G843D allele(s), we analysed clinical data from 52 individuals (age of assessment 0.04–49.9 years) in a longitudinal ZSD natural history study (ClinicalTrials.gov ID NCT01668186) who were either homozygous or compound heterozygous for this allele. Evidence of liver disease was subclassified into liver dysfunction (including increased liver transaminases and hepatomegaly) and severe liver disease (including cirrhosis, portal hypertension, esophageal varices, gastrointestinal bleeding, ascites or hepatic cancer). There was frequent evidence of liver dysfunction or severe liver disease in patients homozygous for the PEX1-p.G843D allele (67%, liver dysfunction; 9%, severe disease) or compound heterozygous for the PEX1-p.G843D allele (61%, liver dysfunction; 29%, severe disease) (Fig. 1b). On average, PEX1-p.G843D homozygote patients, who have two PEX1 alleles with residual activity, showed less severe liver disease than compound heterozygous patients, who could also carry a PEX1 null allele (indeed, frameshift or nonsense variants were found in 24 such patients). Overall, these data support the unmet need for therapies for ZSD liver disease caused by PEX1-p.G843D allele(s).

Development of a base editing strategy to correct the-p. G843D allele in patient-derived fibroblasts PEX1

To correct the PEX1-p.G843D (c.2528G>A) variant, we designed a Streptococcus pyogenes Cas9 (SpCas9) gRNA that places the c.2528G>A adenine variant at the fourth position (A4) of the protospacer 5′-TTGATGGGTTACATGAAGTT-3′, located within the 5-nt base editing window (5′-N4NNNN8-3′) for most base editors that use SpCas941 (Fig. 1c). Next, we tested three different ABEs that use SpCas9 domains to recognize NGG protospacer-adjacent motif (PAM) sequences: ABE7.1042, ABE8e43 and ABE8e-V106W43. Laboratory-evolved deoxyadenosine deaminases ABE8e and ABE8e-V106W offer higher editing efficiencies than ABE7.10. ABE8e-V106W is a high-fidelity version of ABE8e that results in efficient on-target editing but reduced RNA editing and Cas-independent off-target DNA editing owing to a V106W substitution in the deaminase43. After nucleofecting patient-derived homozygous PEX1-p.G843D skin fibroblasts with ABE7.10 mRNA and synthetic gRNA, 20% of total sequencing reads contained the corrected base edit (from the pathogenic A4 to wild-type (WT) G4) and 7.6% of sequencing reads showed bystander editing (synonymous mutation) at protospacer position A11 (Fig. 1d). Using ABE8e and ABE-V106W to target this protospacer yielded a 4-fold improvement in editing efficiency in patient-derived fibroblasts compared with ABE7.10, achieving 80% correction with ABE8e and ABE8e-V106W. This large improvement in editing efficiency was accompanied by increased bystander editing at positions A11 and A13. Given the enhanced adenine deaminase activity, we expected some bystander editing at these neighbouring bases43. Bystander A•T-to-G•C editing at A11 results in a synonymous mutation (p.L845L, TTA>TTG) predicted not to impact PEX1 RNA splicing44 (Fig. 1d). Bystander editing at position A13 introduces a missense mutation (p.H846R) predicted to be benign by the variant classification algorithms tested44,45 and gnomAD46, and has not been reported in ClinVar47. We observed bystander editing (A13) at a frequency of 19% with ABE8e and 9.3% with ABE8e-V106W, respectively, and low levels of indels (<2%) within 20 bp of the protospacer in all ABE-treated fibroblasts. Further assessment of recently reported compact ABE editors in the homozygous PEX1-p.G843D skin fibroblasts showed largely reduced editing efficiency for this pathogenic allele compared with ABE8e (Supplementary Note 1).

These results demonstrate that adenine base editing can efficiently correct the PEX1-p.G843D variant in patient-derived cells while minimizing indels. We, therefore, chose to apply these ABE8e and ABE8e-V106W PEX1-p.G843D correction strategies in vivo using mouse models of ZSD.

In vivo adenine base editing efficiently corrects the mouse-p.G844D allele Pex1

To investigate the potential of an ABE-mediated therapeutic strategy for ZSD, we designed a SpCas9 gRNA to target the Pex1-p.G844D murine equivalent of the PEX1-p.G843D allele. The mouse allele differs from the corresponding PEX1-p.G843D human allele at the target protospacer by a single nucleotide at the eighth position (5′-TTGATGGATTACATGAAGTT-3′). We designed a dual-AAV9 vector system that encodes the ABE as two halves, each linked to a fast-splicing intein48 (Fig. 1e). Following the co-transduction of AAV encoding both ABE halves, the complete base editor can form upon protein splicing or association of the two halves48.

To determine the ability of ABE8 variants to correct the Pex1-p.G844D allele in the liver of mice, we treated P1Pex1G844D/+ heterozygous mice with one of three doses of AAV9 (4 × 1011 vg, 4 × 1010 vg or 4 × 109 vg per mouse) encoding ABE8e or ABE8e-V106W and the above gRNA by facial vein injection (Fig. 1f). Heterozygous mice did not tolerate the highest dose of ABE8e (4 × 1011 vg per mouse, equivalent to 2.8 × 1014 vg kg−1); all mice in this cohort died by day 15 after injection. By contrast, all five mice successfully injected with the ABE8e-V106W base editor at the same dose of 4 × 1011 vg per mouse survived. Six weeks after injection, the mice were killed, and liver tissue analysis revealed 52% bulk correction of Pex1-p.G844D pathogenic alleles, the highest editing efficiency among all surviving cohorts (Fig. 1g). At 4 × 1010 vg and 4 × 109 vg doses, ABE8e did not lead to any lethality and yielded 47% and 13% correction of the Pex1-p.G844D alleles. We observed 44% and 10% correction in ABE8e-V106W injected P1 mice dosed with 4 × 1010 vg and 4 × 109 vg, respectively. We also observed very low levels of undesired bystander editing (<0.1% for A at the minus one position relative to the protospacer, p.I843V) and indels (<0.55%) across all mice treated with either ABE variant.

Collectively, these data in heterozygous mice establish that AAV-mediated in vivo delivery of ABE can correct the Pex1-p.G844D allele in animals. Moreover, these results suggested an AAV9-ABE8e-V106W dose range to evaluate potential phenotypic rescue in the homozygous Pex1-p.G844D mouse model. As we observed high editing efficiency and no evident toxicity when using AAV9 as a delivery vehicle, we advanced the ABE8e-V106W base editor to in vivo phenotypic rescue studies with homozygous Pex1-p.G844D mice.

In vivo base editing in homozygous-p.G844D mice Pex1

Homozygous Pex1-p.G844D mice have a severe neonatal mortality when bred in the B6/J (B6.Cg-Pex1tm1.1Sjms/Mmjax, catalogue 25065)39 or B6/N37 genetic background strains. By contrast, mixed genetic background homozygous Pex1-p.G844D mice have been previously shown to have a more prolonged survival36,39. Here we used a more defined breeding strategy by crossing B6.Cg-Pex1tm1.1Sjms/Mmjax with 129S6. Cg-Pex1tm1.1Sjms/Mmjax. The use of F1 hybrid mice improves reproducibility and reduces phenotypic variation owing to segregating genetic modifiers. The resulting homozygous mice for the Pex1-p.G844D allele have normal survival, but show growth deficits, retinal degeneration38,49 and liver pathologies resulting from impaired peroxisomal function.

To test the potential of base editing to restore peroxisome and liver function, we delivered dual-AAV9-ABE8e-V106W to P1Pex1G844D/G844D pups systemically, via facial vein injection (Extended Data Fig. 1a). All vehicle (saline)-treated homozygous mice survived until 16 weeks post-injection, the endpoint of the study, confirming that the B6129S6 F1 hybrid background addressed the neonatal lethality observed in pure B6 or pure 129S6 mouse strains (Extended Data Fig. 1b). While the 4 × 1011 vg AAV9-ABE8e-V106W dose did not affect the viability of healthy heterozygous Pex1G844D/+ mice, Pex1G844D/G844D mice were less tolerant of this dose, most likely owing to their reduced weight at P1 (1.4 g homozygous versus 1.6 g heterozygous on average) and pronounced liver disease. Among the 19 (44 out of 63 mice died) surviving mice, we observed an average of 59% bulk liver correction of the G844D allele. Since hepatocytes account for the majority (~60%) of mouse liver cells50, this degree of bulk liver editing is consistent with robust hepatocyte editing (Extended Data Fig. 1c). The body weight of Pex1G844D/G844D mice treated with ABE8e-V106W improved compared with untreated controls (Extended Data Fig. 1d). All 19 mice showed a normalization of hepatic lipid profiles with decreased levels of free and esterified VLCFAs and BCFAs by gas chromatography/mass spectrometry (GC/MS) (Extended Data Fig. 1e). Further histopathology analysis showed a reduction of lipid accumulation in the liver (Extended Data Fig. 1f), suggesting restored peroxisome function and rescued liver steatosis.

To assess the relative contributions of viral dose or editor cargo to toxicity, we injected P1Pex1G844D/G844D mice with a 4 × 1011 vg dose of AAV9-GFP. AAV9-GFP-treated Pex1G844D/G844D mice survived at a similar rate (~20%; Extended Data Fig. 2a) compared with Pex1G844D/G844D mice treated with AAV9-ABE at the same dose (~30%; Extended Data Fig. 1b). These results suggest that a high viral dose is the major contributor to the observed toxicity, rather than the ABE8e-V106W cargo. Because the 4 × 1011 vg dose of AAV9-ABE8e-V106W achieved high editing but reduced survival, we lowered the AAV9-ABE dose to identify a safe yet effective level. We performed facial vein injection at P1 of 1 × 1011 vg, 4 × 1010 vg or 1 × 1010 vg (Extended Data Fig. 2b). These doses were chosen based on the dose-dependent editing that we observed in the heterozygous cohort. After injection, all three dosing conditions were well tolerated (more than 90% of treated mice survived) (Fig. 2a). The growth deficit in the Pex1G844D/G844D mice was effectively rescued at all doses of ABE8e-V106W in both males and females (Fig. 2b). In addition, high-throughput sequencing (HTS) analysis at 6 weeks of age showed 56% and 57% average correction in the bulk liver of the Pex1-p.G844D allele at the 1 × 1011 vg and 4 × 1010 vg doses, respectively, with average indel frequencies below 0.12%. Although the 1 × 1010 vg dose yielded 29% editing of pathogenic alleles at week 6, the frequency of corrected alleles increased 1.8-fold by week 16, resulting in an average of 53% editing efficiency, similar to what we observed in mice treated with the 1 × 1011 vg and 4 × 1010 vg doses (Fig. 2c). Higher doses led to faster correction, consistent with the durable expression of ABE (Extended Data Fig. 2c) from AAV among transduced cells. We also assessed base editing of the Pex1-p.G844D variant in central nervous system tissues such as cortex, cerebellum and retina 16 weeks post-injection. As expected37,39, we observed low editing efficiency of 0.35%, 0.14% and 0.51%, respectively, likely owing to the inefficiency of AAV9 transduction of mouse central nervous system (CNS) following systemic administration51 (Extended Data Fig. 3a–c).

ABE rescues liver pathologies and peroxisome metabolites in homozygous-p.G844D mice Pex1

To assess the impact of gene editing on peroxisome homeostasis, we measured ABCD3 (also called PMP70) peroxisomal membrane protein levels, a previously reported biomarker of hepatic peroxisome content52 that can be influenced by biogenesis or pexophagy activity. Consistent with the known peroxisome deficit in this mouse model, western blot analysis revealed little to no detected ABCD3 protein in Pex1G844D/G844D vehicle-treated mouse liver tissues (Fig. 2d). By contrast, liver tissue from ABE8e-V106W-treated Pex1G844D/G844D mice revealed a dramatic increase in ABCD3 protein levels at all tested doses (Fig. 2d). ABCD3 protein levels at all doses were similar to those in WT mice, suggesting that even the lowest tested dose (1 × 1010 vg) of AAV encoding ABE8e-V106W could begin to restore peroxisome homeostasis.

Next, we analysed levels of free and esterified fatty acids in plasma and liver and found a dose-dependent rescue of hepatic peroxisomal metabolic activities in Pex1G844D/G844D mice at early time points (6 weeks after injection). The 1 × 1011 vg and 4 × 1010 vg ABE doses normalized increased phytanic and pristanic acid (BCFA) in liver and plasma in Pex1G844D/G844D mice, consistent with rescue of peroxisomal alpha- and beta-oxidation activity, while the lowest dose tested (1 × 1010 vg) resulted in a partial reduction (Fig. 2e and Extended Data Fig. 3d). Likewise, at 6 weeks, the 1 × 1011 vg and 4 × 1010 vg doses resulted in near-complete normalization of increased hepatic VLCFA (including C26:0 and C26:1 n-9) levels in Pex1G844D/G844D mice, consistent with rescue of the hepatic peroxisomal beta-oxidation pathway, while the lowest dose tested resulted in partial reduction (Fig. 2e). Increased hepatic levels of the membrane lipid C26:0-LPC in Pex1G844D/G844D mice were unchanged after ABE-AAV9 treatment, suggesting heterogeneity in the rates of metabolism of specific C26:0-containing lipids (Supplementary Note 2). At 6 weeks, increased plasma C26:1 n-9 levels were partially reduced at the 1 × 1011 vg and 4 × 1010 vg doses, but increased plasma C26:0 levels were unchanged. Increased plasma C26:0-LPC levels were not reduced in treated Pex1G844D/G844D mice (Extended Data Fig. 3d and Supplementary Note 2).

At 16 weeks after injection, more robust rescue of hepatic peroxisome lipid metabolism in Pex1G844D/G844D mice was observed. Phytanic and pristanic (BCFA) levels in liver and plasma (Fig. 2e and Extended Data Fig. 3d) were lowered to WT levels for all doses tested. Hepatic VLCFA levels in Pex1G844D/G844D mice were near WT for all doses. Plasma C26:1 n-9 levels in Pex1G844D/G844D mice were substantially reduced for all doses tested, but plasma C26:0 levels were mildly reduced. No reduction in liver or plasma C26:0-LPC levels was observed for any dose tested in these mice (Extended Data Fig. 3d and Supplementary Note 2). Overall, although varying the ABE-AAV dose changes the kinetics of gene correction, the major outcome after 16 weeks—robust correction of free and esterified BCFA and VLCFA levels in liver—were achieved across all doses tested.

Histological analysis of lipid content in liver sections also revealed a dose-dependent efficacy at an early age. The liver tissue in the low-dose-treated neonatal mice was characterized by irregular hepatic Oil Red O (ORO) staining (Fig. 2f, top right). By contrast, by 16 weeks after injection, hepatic lipid content was reduced to near-WT levels (Fig. 2f, bottom right). These results demonstrate that ABE8e-V106W-mediated correction of pathogenic Pex1-p.G844D alleles rescues clinically relevant peroxisomal liver disease phenotypes when delivered to neonatal mice at all tested doses.

Transcriptome analysis on ABE8e-V106W-treatedneonatal mice Pex1 G844D/G844D

To further understand the mechanisms underlying the phenotypic correction in mice treated with AAV9-ABE8e-V106W, we performed RNA sequencing (RNA-seq) analysis of liver tissues from mice injected with 1 × 1011 vg, 4 × 1010 vg or 1 × 1010 vg at 6 and 16 weeks after injection (Supplementary Table 1). We applied a stringent false discovery rate (FDR < 1%) to identify differentially expressed genes (DEGs) among different treatment conditions. At 6 weeks, 691 DEGs were identified in livers from the homozygous Pex1G844D/G844D and WT mice. Even at this early time point, there were dramatic decreases in the numbers of DEGs in the base-edited Pex1G844D/G844D compared with saline-treated WT mice (105 DEGs at the 1 × 1010 vg dose, 13 at the 4 × 1010 vg dose and 38 at the 1 × 1011 vg dose) (Extended Data Fig. 4a and Supplementary Table 2). By the 16-week time point, a total of 352 DEGs were identified in livers from the Pex1G844D/G844D and WT mice. Remarkably, there were no DEGs detected between livers of Pex1G844D/G844D mice treated with the ABE8e-V106W at the 4 × 1010 vg and 1 × 1011 vg doses and WT mice at 16 weeks. Only one DEG was detected in the liver of the Pex1G844D/G844D mice treated with the ABE8e-V106W at the lower 1 × 1010 dose (Extended Data Fig. 4b and Supplementary Table 2). These data indicate that in vivo base editing correction of Pex1-p.G844D normalizes the liver transcriptomic profile 16 weeks after treatment.

To further explore, we performed hierarchical clustering analysis on all DEGs from ABE-treated versus vehicle-treated Pex1G844D/G844D mouse livers at 6 weeks (363 genes) and 16 weeks (250 genes) after injection (Fig. 3a and Supplementary Table 3). The lowest AAV9-ABE dose, 1 × 1010 vg (equivalent to 7 × 1012 vg kg−1), resulted in a partially rescued transcriptome state between WT and homozygous disease mice at 6 weeks after injection. In comparison, by 16 weeks, the mice showed near-complete reversal of transcriptome changes, suggesting that the increased correction of Pex1-G844D pathogenic variants from 29% to 53% during this time contributed to transcriptome normalization.

Gene-set-based KEGG and phenotype pathway analysis of DEGs between livers of saline-treated Pex1G844D/G844D and WT mice at 6 and 16 weeks revealed DEGs related to physiological responses to liver injury, fibrosis and hepatocellular carcinoma (Supplementary Table 4). Twenty severe liver disease-associated DEGs were more highly expressed in Pex1G844D/G844D mice compared with WT controls (Fig. 3b). Consistent with a rescue of liver health, these genes were downregulated to at or near-WT levels in the ABE-treated Pex1G844D/G844D mice (Fig. 3b). In addition, alterations in hepatic carbohydrate metabolism have been reported in mouse models with ZSD and Drosophila models with other peroxisomal biogenesis disorders39,53. As expected37,39, many DEGs found between mutant and WT mice linked to peroxisomal/mitochondrial lipid metabolic dysfunctions and carbohydrate metabolism were rescued upon ABE treatment (Fig. 3a, Extended Data Fig. 5 and Supplementary Table 5).

Furthermore, histologic analysis of liver tissues via periodic acid–Schiff (PAS) staining revealed restoration of normal glycogen levels upon ABE treatment (Fig. 3c,d), supporting the conclusion that phenotypic rescue is the result of ABE-mediated correction. We found that Cyp4a10, Cyp4a14 and Cyp4a31—genes associated with liver ω-oxidation54–56—were more highly expressed in Pex1G844D/G844D compared with WT mice, potentially reflecting PPAR-mediated activation of a compensatory mechanism to catabolize BCFAs. These same genes were downregulated upon ABE treatment at 16 weeks after injection compared with the vehicle-treated Pex1G844D/G844D mice, suggesting that ABE-enabled rescue minimized dependence on the compensatory ω-oxidation pathway owing to the restoration of peroxisomal α-oxidation activity. Together, the transcriptome analyses above support therapeutic benefit of AAV9-ABE-V106W for rescuing liver pathogenicity in neonatal Pex1G844D/G844D mice.

Therapeutic base editing in 4-week-oldmice Pex1 G844D/G844D

While neonatal treatment with ABE8e-V106W resulted in an almost complete rescue of the ZSD hepatic phenotypes, testing the effect of in vivo base editing in animals at a more advanced disease stage would inform the potential of treatment for patients of different ages. Therefore, we injected AAV9-ABE8e-V106W at a dose of 5 × 1011 vg per mouse (equivalent to the neonatal dose of 1 × 1011 vg per mouse or 7 × 1013 vg kg−1) via retro-orbital injection into 4-week-old Pex1G844D/G844D mice (Extended Data Fig. 6a). We tested the lipid profile in plasma at 6 and 12 weeks after injection and in the liver at 12 weeks after injection. The treatment was well tolerated, and all animals survived until the 16-week endpoint (Fig. 4a). The body weight of Pex1G844D/G844D mice treated with ABE8e-V106W improved compared with vehicle-treated controls from 8 weeks of age onwards (Fig. 4b). We assessed base editing efficiency at 16 weeks of age, revealing 50% bulk liver correction of the Pex1-p.G844D allele with no detected non-synonymous bystander editing (Fig. 4c). ABCD3 protein levels were restored in base-edited mice to levels similar to WT controls, consistent with improved peroxisome function (Fig. 4d). The levels of phytanic and pristanic BCFAs and most VLCFAs in plasma and liver were also restored to those comparable to WT controls (Fig. 4e and Extended Data Fig. 6b).

Consistent with the cholestatic liver disease present in people with moderate and severe ZSD, by 16 weeks Pex1G844D/G844D mice showed an accumulation of C27-bile acid intermediates (3α,7α-dihydroxycholestanoic acid (DHCA) and 3α,7α,12α-trihydroxycholestanoic acid (THCA)) thought to be involved in its aetiology16 (Fig. 4f). To determine whether restoration of PEX1 activity could address the accumulation of these C27-bile acid intermediates, we analysed DHCA, THCA and 7α-hydroxy-3-oxo-4-cholestenoic acid (7-HOCA) in the liver and found that they were restored to WT levels, reflecting the rescue of a biosynthetic pathway requiring peroxis omal beta-oxidation (Fig. 4f). Mature bile acid levels did not differ significantly in WT and mutant mice or in mice treated with the ABE (Supplementary Note 3). Histologic analysis of the liver suggested rescue of hepatic lipids and carbohydrate metabolism (Fig. 4g and Extended Data Fig. 6c). These findings indicate that base editing resulted in physiologically relevant rescue of liver function when administered to 4-week-old Pex1G844D/G844D mice.

LNPs offer a more transient, non-viral delivery modality for gene editing agents. LNP-mediated base editor delivery has been used effectively in non-human primates and in human patients57,58. In addition to being amenable to well-established, clinically accepted manufacturing methods, the transience of LNP-mediated base editor mRNA delivery minimizes the potential for off-target editing after the targeted edits are made59,60. To test the feasibility of LNP-mediated ABE mRNA delivery in Pex1G844D/G844D mice, we formulated OF-02 LNPs61 that have previously demonstrated robust and highly specific RNA delivery to the mouse liver following intravenous administration61 and delivered ABE8e-V106W mRNAs and modified sgRNA into 4-week-old mice (Extended Data Fig. 7a). Encouragingly, we observed, in the liver, up to 27% allele correction in a dose-dependent manner, with minimal unwanted bystander editing or indels (Extended Data Fig. 7b,c). Average ABCD3 protein levels increased 7.6-fold in Pex1G844D/G844D mice treated with the highest dose of ABE-LNP (3 mg kg−1) compared with saline-treated mice, indicating improved peroxisome function (Extended Data Fig. 7d). ORO staining showed that increased hepatic lipids in the Pex1G844D/G844D mice were reduced by 33% 6 weeks after treatment with 3 mg kg−1 ABE-LNP, approaching lipid levels in WT mice (Extended Data Fig. 7e). Furthermore, PAS staining revealed restoration of normal hepatic glycogen levels upon 3 mg kg−1 ABE-LNP treatment (Extended Data Fig. 7e). These findings demonstrate the compatibility of the base editing strategy developed in this work with clinically validated non-viral in vivo base editor delivery methods. Although systemic delivery of ABE8e-V106W resulted in low genome editing in the CNS of treated mice as expected, we observed a significant reduction in BCFAs (phytanic and pristanic acids), but not VLCFAs, in the cortex of AAV- and LNP-treated mice after injection (Extended Data Fig. 8). This observation suggests that restoring hepatic alpha-oxidation reduces the pool of circulating phytanic acid that would otherwise accumulate in the cortex.

Collectively, AAV9-ABE8e-V106W-mediated correction of the pathogenic Pex1-p.G844D allele in 4-week-old mice resulted in near-complete rescue at genomic, lipidomic and phenotypic levels. Importantly, our findings indicate that the therapeutic benefits of ABE-V106W observed in neonates can be extended to young mice with a more advanced stage of liver disease.

In vivo and in vitro off-target analysis of ABE8e-V106W targetingandpathogenic variants Pex1 PEX1

To assess Cas-dependent DNA and Cas-independent RNA off-target editing in ABE-V106W-treated mice, we performed CIRCLE-seq62, an unbiased and sensitive genome-wide off-target detection method that nominates genomic DNA sites engaged by SpCas9 nuclease complexed with the Pex1-p.G844D-targeting gRNA in vitro. When applied to mouse genomic DNA, CIRCLE-seq nominated 631 ranked candidate off-target sites. We evaluated the top 30 sites using the liver genomic DNA samples collected from the homozygous mice treated with 1 × 1011 vg of ABE-AAV at 6 weeks and 16 weeks after injection (Fig. 5a and Supplementary Table 6). Despite high dose of AAV9-ABE8e-V106W exposure, we observed minimal off-target editing, with only two sites showing editing above background levels from vehicle-treated controls: OT1 (≤0.35%) and OT15 (≤0.29%). Off-target editing did not increase between 6 and 16 weeks. Both off-target sites reside in intergenic regions and are not annotated as regulatory elements in the mouse genome. These data suggest a very low off-target editing rate obtained in the mice treated with even a high dose of AAV9-ABE8e-V106W (Fig. 5b). To identify Cas-independent RNA off-target editing in ABE-treated Pex1G844D/G844D mice, we analysed liver RNA-seq datasets from mice treated with 1 × 1011 vg, 4 × 1010 vg or 1 × 1010 vg of ABE-AAV. At 6 weeks, but not 16 weeks after injection, we observed a modest 1.2-fold increase in transcriptome-wide A-to-G RNA changes relative to vehicle-treated controls (Fig. 5c and Extended Data Fig. 9a), consistent with higher AAV9-ABE8e-V106W expression level at earlier time points and with the transient nature of cellular RNA. These findings further support that ABE8e-V106W base editing of the Pex1-p.G844D allele is highly site specific, with minimal detected genetic or transcriptomic changes. In juvenile mice treated with ABE-AAV9 or ABE-LNPs, we likewise detected no significant increases in OT1 and OT15 editing relative to saline-treated controls. It is possible that differences in viral vector transduction efficiency between the neonatal and juvenile cohort may influence ABE expression levels and/or target site accessibility could vary with age (Extended Data Fig. 9b).

To identify potential off-target editing associated with the ABE8e-V106W base editor and a gRNA targeting human PEX1-p.G843D, we performed CIRCLE-seq analysis using patient-derived compound heterozygous PEX1G843D/I700fs cultured skin fibroblasts, resulting in the nomination of six candidate off-target sites in total. In addition, we also performed in silico analysis, Cas-OFFinder63, and predicted 20 potential off-target sites with 3 or fewer mismatched nucleotides compared with the targeted protospacer (Extended Data Fig. 9c). To assess off-target editing associated with high and prolonged expression of ABE8e-V106W, we incubated two sets of patient-derived cultured fibroblasts (homozygous PEX1G843D/G843D and compound heterozygous PEX1G843D/I700fs, respectively) with lentivirus to deliver ABE-V106W and gRNA. We confirmed efficient on-target editing with up to 73% correction of the target PEX1-p.G843D alleles 2 weeks after transduction (Fig. 5d). Next, we amplified the lentivirus-treated genomic DNA (gDNA) from PEX1G843D/G843D cells at 26 nominated off-target sites (the 6 from CIRCLE-seq and 20 from Cas-OFFinder) and examined all possible A•T-to-G•C changes within the protospacers. We did not detect sequence modifications above the 0.1% detection threshold for HTS or above levels of genetic variation present in vehicle-treated samples for 25 of the 26 tested sites. One off-target site in an intron of FBXO38 showed 0.88% A•T-to-G•C substitution, which could potentially affect a splicing acceptor sequence (Fig. 5e).

Overall, these analyses suggest that the ABE8e-V106W editing strategies both in vivo in mice and in vitro in patient-derived cells result in high-specificity editing, with minimal observed DNA or RNA off-target editing, even following a high-dose AAV treatment that leads to prolonged expression of ABE.

Correction of-p.G843D variant in patient-derived fibroblasts restores peroxisome homeostasis PEX1

We assessed whether correction of PEX1-p.G843D variant restores peroxisome homeostasis in patient-derived fibroblasts. We nucleofected two patient-derived fibroblasts: homozygous PEX1G843D/G843D and compound heterozygous PEX1G843D/I700fs with ABE8e-V106W mRNA and a guide RNA targeting the PEX1-p.G843D (c.2528G>A) allele or a non-targeting guide RNA (sgNT) that does not match any human genome sequence64. We observed robust disease mutation correction efficiency exceeding 80% following treatment with ABE and the PEX1-p.G843D-targeting guide RNA (Extended Data Fig. 10a).

The PEX1-p.G843D allele, like other pathogenic PEX1 variants, impairs peroxisome biogenesis and results in mislocalization of peroxisome matrix proteins such as catalase5. To evaluate peroxisomal integrity, we stained catalase and the peroxisomal membrane protein ABCD3 using specific primary and fluorophore-conjugated secondary antibodies. Confocal microscopy revealed that sgNT-treated cells show diffuse cytosolic catalase signals alongside distinct ABCD3-positive foci, consistent with the known defects previously reported in ZSD patient-derived cells5,65 (Fig. 6, rows 2 and 4). By contrast, fibroblasts treated with the PEX1-p.G843D-targeting gRNA displayed catalase concentrated in punctate structures, co-localizing with ABCD3 foci, closely resembling healthy donor fibroblasts (Fig. 6, rows 1, 3 and 5). These findings indicate that correction of PEX1-p.G843D restores peroxisomal homeostasis.

To further assess peroxisomal assembly via live cell imaging, we transduced corrected and control PEX1G843D/G843D fibroblasts with a baculovirus vector expressing a GFP reporter protein fused to an N-terminal peroxisome targeting signal 1 (GFP–PTS1)66,67. The sgNT-treated cells showed diffuse cytosolic distribution of GFP fluorescence or background GFP signals, indicating impaired import. By contrast, PEX1-p.G843D-corrected fibroblasts displayed robust punctate GFP–PTS1 structures similar to those observed in healthy control fibroblasts consistent with restored peroxisomal import and assembly (Extended Data Fig. 10b). Altogether, these results demonstrate that targeted correction of PEX1-p.G843D variant with ABE8e-V106W re-establishes peroxisomal matrix protein import and peroxisomal assembly in patient-derived cells.

Discussion

Current treatments for ZSD are largely supportive and do not address the root cause that impairs the function of PEX genes required for normal peroxisome homeostasis17,23. Patients who are homozygous or compound heterozygous for the common PEX1-p.G843D allele frequently develop progressive liver disease associated with frequent hospitalizations and a shortened lifespan. Here we demonstrate that a Pex1-targeted base editing strategy to rescue liver functions in a dose- and time-dependent manner in a ZSD mouse model. Although further studies are needed to comprehensively assess safety and long-term efficacy, the translation of this approach to patients with liver disease has the potential to markedly improve quality of life and longevity16. Importantly, international newborn screening for ZSD21,22 provides an opportunity for deploying genetic therapies before irreversible liver and CNS damage occurs.

We identified ABE8e-V106W as a promising base editor to correct the common PEX1-p.G843D pathogenic variant in patient-derived fibroblasts and in homozygous Pex1-p.G844D ZSD mice68,69 (Fig. 1d and Supplementary Note 1). In patient-derived fibroblasts, mRNA delivery corrected >80% PEX1-p.G843D alleles, and in vivo AAV9 delivery produced up to 60% correction in bulk liver following a single treatment, enabling permanent and precise correction of the root cause of peroxisomal dysfunction. Since hepatocytes, the primary liver cell type transduced by AAV9, make up ~60% of the mouse liver50, these results suggest efficient base editing in the vast majority of hepatocytes70.

Dose–response studies in the neonatal ZSD mouse showed that even the lowest dose tested achieved near-maximal editing by 16 weeks after injection. Editing efficiency nearly doubled (1.8-fold increase) between weeks 6 and 16. Consistent with the increased efficiency in DNA correction over time, transcriptomic, lipidomic and histological analysis showed a robust rescue of hepatic gene expression and most fatty acid profiles normalizing to WT or near-WT levels by week 16. Despite these improvements, hepatic C26:0-LPC levels remained increased. This is because phospholipid and lysophospholipid levels reflect contributions from de novo synthesis in the endoplasmic reticulum and mitochondrial membranes71, remodelling via the Lands cycle72, degradation by lysophospholipases72 and biliary clearance73,74. Further investigation is needed to dissect how these processes contribute to a persistent C26:0-LPC increased. Moreover, the activities of extrahepatic tissues could influence circulating and hepatic C26:0-LPC levels. For example, hematopoietic stem cell transplantation reduces increased circulating C26:0-LPC levels in patients with cerebral adrenoleukodystrophy75, raising the possibility that hematopoietic-derived cells from healthy donors also could influence plasma C26:0-LPC levels. Overall, our results highlight heterogeneity in the metabolism of specific VLCFA-containing lipid species, such as those present in membranes, which should be considered when selecting biomarkers for future studies.

Regardless, this lipid heterogeneity did not prevent base editor-mediated rescue of animal growth, liver histology or liver transcriptome abnormalities in Pex1G844D/G844D mice. Indeed, base editing in Pex1G844D/G844D mice resulted in rescue of hepatic transcriptomic signatures relevant to lipid metabolism and peroxisome functions, as well as physiological responses to liver injury, fibrosis and fatty liver disease (Fig. 3b and Supplementary Table 5). The progressive increase in editing and near-complete phenotypic rescue between weeks 6 and 16 likely reflects sustained ABE expression from AAV9 (Extended Data Fig. 2c), hepatocyte proliferation and/or selective advantage among edited cells. Our findings collectively support the potential of ABE8e-V106W-mediated gene correction as a therapeutic strategy for ZSD-associated liver disease, even at comparatively low vector doses.

Despite the anticipated low to minimal gene editing in the CNS following systemic ABE-AAV9 delivery, we observed a significant reduction in brain BCFAs (phytanic and pristanic acids), but not VLCFAs (Extended Data Fig. 8). Unlike VLCFAs that are primarily made endogenously76,77, phytanic acid is exclusively diet-derived in mammals78. BCFA accumulation in the Pex1G844D/G844D mouse cortex most likely results from circulating phytanic acid77,78 that crosses the blood–brain barrier, but is poorly metabolized by peroxisomal alpha-oxidation in the brain. Restoring hepatic alpha-oxidation reduces the circulating pool of phytanic acid available to enter the brain. This finding has translational implications, since it suggests that an effective gene editing correction of the peroxisomal alpha-oxidation pathway in liver could help protect the CNS and other tissues from toxic accumulation of diet-derived BCFAs. This observation also has implications for the treatment of adult Refsum disease, in which impaired PHYH-mediated alpha-oxidation can lead to toxic accumulation of phytanic acid, resulting in neurological and sensory deficits79.

Finally, ABE8e-V106W treatment of 4-week-old ZSD mice, representing a more advanced liver disease stage, also demonstrated near-complete rescue of liver genomic, lipidomic and histological abnormalities. Given the progressive course of liver disease in our preclinical mouse model and the patient population, this result suggests that the therapeutic window for treating ZSD liver disease by gene correction strategies may be long enough to benefit a substantial proportion of patients. Collectively, these results support the potential of using a base editing therapeutic strategy to address chronic liver disease found in many patients with ZSD. In our mouse model, additional ABE dose-escalation studies could further elucidate how varying levels of hepatic editing influence the rescue of liver functions. Such findings could guide the design of future human clinical trials. Non-viral LNP delivery of base editor mRNA demonstrated the potential to correct pathogenic PEX1 alleles and restore peroxisomal function. We also expect that chemical modification of the mRNA80 and sgRNA81 will further improve RNA stability and editing efficiency. Although re-dosing remains a challenge for AAV delivery systems owing to neutralizing antibodies82, transient delivery systems such as LNPs59,83 or engineered virus-like particles84,85 in conjunction with adeno-associated vehicles that have brain and retina tropism could open new therapeutic opportunities for treating multisystemic genetic disorders such as ZSD in a variety of other tissues and organs, including the retina, brain and cochlea.

Methods

Clinical analysis of liver disease in individuals with ZSD

Patients with a diagnosis of a peroxisome biogenesis disorder were enrolled in a longitudinal, retrospective natural history study at the Research Institute of the McGill University Health Center with informed consent. Patients were recruited internationally, 87% from North America, 2.7% Australia, 1.3% Europe and 0.7% South America. Medical records were requested from birth to study entry, and annually thereafter from each participant’s healthcare institutions following authorization from the patient or parent/legal representative. De-identified longitudinal clinical data were collected from participants with at least one PEX1-c.2528G>A (p.G843D) allele. Patients without any liver-related abnormalities were classified as no liver disease (normal); patients reported with any of the following—hepatomegaly, jaundice, increased liver enzymes in blood or coagulopathy—were considered to have some hepatic dysfunction; patients diagnosed with cirrhosis, portal hypertension, esophageal varices, gastrointestinal bleeding, ascites or hepatic cancer were classified as severe liver disease. The study was registered on ClinicalTrials.gov (NCT01668186) where the study protocol and analysis plan are provided (also see ‘Reporting summary’). Sex, genotypes, age and the liver disease category are provided in Source Data Fig. 1b.

Mammalian cell culture conditions

ZSD patient-derived PEX1G843D/I700fs (catalogue ID GM16510) and healthy donor-derived primary skin fibroblasts (catalogue number GM03348) were purchased from the Coriell Institute for Medical Research. ZSD patient-derived homozygous PEX1G843D/G843D primary skin fibroblasts were obtained from the Peroxisomal Disease Laboratory at the Kennedy Krieger Institute. Fibroblasts were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with GlutaMAX (Thermo Fisher Scientific) and 15% (v/v) fetal bovine serum (FBS) (Thermo Fisher Scientific). HEK293T cells were purchased from the American Type Culture Collection (ATCC). All cells were cultured at 37 °C with 5% CO2. Cell lines were authenticated by their respective suppliers and tested negative for mycoplasma.

Molecular cloning

Plasmids were constructed using Gibson assembly or restriction enzyme cloning. PCR amplification was performed with PhusionU Green Multiplex PCR Master Mix (Thermo Fisher Scientific). Gene fragments and plasmid backbones were generated by PCR or restriction digestion, or purchased from IDT. Recombinant plasmids were assembled using NEBuilder HiFi DNA Assembly Master Mix or T4 DNA ligase (NEB). Mach1 (Thermo Fisher Scientific) and NEB Stable (High Efficiency) chemically competent Escherichia coli cells were used for plasmid propagation.

Electroporation of patient-derived fibroblasts

Patient-derived fibroblasts at 80–90% confluence in T-75 flask were washed with phosphate-buffered saline (PBS) (Thermo Fisher Scientific), trypsinized using TrypLE Express (Thermo Fisher Scientific) and resuspended in 10 ml of media. Cells were centrifuged at 110g for 5 min. During centrifugation, RNA reagents were prepared: for each sample, 1 μg base editor mRNA and 50 pmol of synthetic gRNA were mixed (2 μl total) and combined with 20 μl of P2 Primary Cell Nucleofector Solution and Supplement Solution mixture (Lonza, P2 Primary Cell 96-well Nucleofector Kit) per the manufacturer’s protocol. Pelleted cells were washed with PBS and resuspended in Lonza buffer. Twenty microlitres of cell suspension (~200,000 cells) was added to each editor/gRNA mixture, transferred to Nucleocuvette plates (Lonza) and electroporated using programme DS150 on a Lonza 4D nucleofector with X unit. After electroporation, 80 μl of media was added and incubated for 10 min at room temperature. Cells were then seeded into pre-equilibrated 24-well plates containing 1 ml medium per well and cultured for 3 days before lysis and sequencing.

HTS of genomic DNA

Seventy-two hours after nucleofection, cells were washed with PBS and lysed for 1 h at 37 °C in lysis buffer (10 mM Tris-HCl pH 8, 0.05% SDS and 25 μg ml−1 proteinase K (Thermo Fisher Scientific)). Lysate was then heat-inactivated at 80 °C for 30 min. Lysate (1 μl) was used as an input for PCR1. PCR1 reactions (25 μl total volume) used Phusion Hot Start II kit (Thermo Fisher Scientific) or PhusionU Green Multiplex PCR Master Mix, and 0.125 μl of each 100 μM primer (sequences listed in Supplementary Table 7). PCR1 was performed under the following cycle conditions: 98 °C for 2 min, (98 °C for 10 s, 61 °C for 20 s, 72 °C for 30 s) × 28 cycles, 72 °C for 2 min.

Samples were barcoded by a second PCR reaction (PCR2). PCR2 reactions were 25 μl total, using the PhusionU Green Multiplex PCR Master Mix, 1.25 μl each of 10 μM Illumina barcoding primers, and 1 μl of PCR1. All PCR2 reactions were performed using the following cycling conditions: 98 °C for 2 min, (98 °C for 10 s, 61 °C for 20 s, 72 °C for 30 s) × 10 cycles, 72 °C for 2 min. After PCR2, samples of similar lengths were pooled and gel-extracted from a 1.5% agarose gel using a QIAquick gel extraction kit (Qiagen). Concentrations of purified libraries were determined using a Qubit Double-Stranded DNA High Sensitivity kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Libraries were diluted to 4 nM and sequenced on a MiSeq (Illumina) using single-read cycles.

HTS data analysis

Samples were demultiplexed with MiSeq Reporter (Illumina). CRISPResso2 was used to analyse demultiplexed reads. Samples were aligned to the WT amplicon in batch mode, using the following parameters: ‘-q 30’ and ‘-qwc’. The value of the qwc parameter defined the portion of the sequence to be analysed for indels. The qwc interval covered the entire protospacer and adjacent nucleotides (26–30 bp quantification window). Percent indels were calculated as the sum of the reads containing insertions and deletions divided by ‘Reads_aligned’ in the CRISPResso_quantification_of_editing_frequency.txt output file. Percent editing was calculated from the specified base conversion in the genomic sequence using the Nucleotide_percentage_summary.txt output file.

In vitro transcription of editor mRNA

In vitro transcription (IVT) of editor mRNA was performed as previously described86 with minor modifications. Editor sequences were cloned into pT7 expression plasmids (example Addgene number 178113). Linear DNA templates were generated by PCR using the Phusion U Green Multiplex Master Mix (NEB) and purified using the QIAquick PCR Purification kit (Qiagen). IVT reactions used a T7 High Yield RNA Synthesis kit (NEB), following the manufacturer’s directions with two exceptions: CleanCap Reagent AG (TriLink BioTechnologies) was added and the uridine-5′-triphosphate in the kit was replaced with N1-methylpseudouridine 5′ triphosphate (TriLink BioTechnologies). Reactions were incubated at 37 °C for 2 h and RNA was purified using lithium chloride (Thermo Fisher Scientific) precipitation, followed by 70% ethanol wash, and resuspended in nuclease-free water. RNA quality was verified by 2% agarose gel electrophoresis, diluted to 2 μg μl−1 and stored at −80 °C.

Lentivirus production and transduction of patient-derived fibroblasts

Lentiviral particles encoding ABE8e-V106W and gRNA were generated in HEK293T using standard three-plasmid packaging (lentiviral transfer plasmid; psPAX2, Addgene number 12260; and pMD2.G, Addgene number 12259). Cells were transfected with Lipofectamine 2000 (Thermo Fisher Scientific) following the manufacturer’s recommended protocol, and 48 h later, viral supernatant was centrifuged at 3,000g for 15 min, filtered (0.45 μm) and stored at −80 °C. Lentiviral transduction of ZSD patient-derived fibroblasts was performed as reported previously87 with minor modifications. Fibroblasts were plated in 6-well culture plates at a density of 1 × 105 cells per well and incubated for 18 h. The lentiviral supernatant was added at high multiplicity of infection (MOI) in the presence of 6 μg ml−1 polybrene and switched to puromycin-containing medium (2 μg ml−1) after 8 h for selection. Cells were expanded under puromycin-selection media for 2 weeks and then lysed for gene-editing and off-target analysis.

AAV production

AAV9 vectors were produced by triple transfection method48 with minor modifications or obtained from the University of Massachusetts Chan Medical School Viral Vector Core. In brief, HEK293T/17 cells were transfected with PEI containing 5.7 μg of AAV genome, 11.4 μg of pHelper (Clontech) and 22.8 μg of AAV9 rep-cap plasmid per plate. The following day, media was exchanged for DMEM (5% FBS). Three days after, cells were collected and resuspended in 500 μl hypertonic lysis buffer (40 mM Tris base, 2 mM MgCl2, 500 mM NaCl and 100 U ml−1 salt-active nuclease, ArcticZymes Technologies) per plate and incubated at 37 °C for 1 h. The media was decanted and combined with 5× solution of polyethylene glycol (PEG) 8000 (Sigma-Aldrich) and NaCl to achieve a final concentration of 8% PEG and 500 mM NaCl. This solution was incubated on ice for 2 h or overnight to facilitate PEG precipitation and then centrifuged (3,200g, 30 min). The pellet was resuspended in 500 μl hypertonic lysis buffer per plate. This was added to the cell lysate, which was either immediately ultracentrifuged or stored at 4 °C overnight. Cell lysates were first clarified by centrifugation at 3,400g for 10 min and viral particles were purified by ultracentrifugation at 465,800g for 2 h 15 min at 4 °C with an iodixanol step gradient, followed by buffer exchange and concentration using 100 kDa MWCO columns (EMD Millipore). The concentrated viral solution was sterile-filtered using a 0.22 μm filter and stored at 4 °C until use. All viruses were titered via quantitative PCR using the AAVpro Titration Kit v.2 following the manufacturer’s protocol (Takara Bio).

LNP formulation and characterization

LNPs were prepared in a microfluidic chip device by mixing an aqueous phase containing the mRNA with an ethanol phase containing the lipids88. The aqueous phase was prepared in a 10 mM citrate buffer with either ABE8e mRNA or sgRNA targeting PEX1 or a non-targeting sgRNA control (5′-GAUCUCGCUUAUAUAACGAG-3′). The ethanol phase was prepared by solubilizing a mixture of OF-02 (Cayman Chemical), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE; Avanti Polar Lipids), cholesterol (Sigma-Aldrich) and 1,2-dimyristoyl-sn-glycero-3-p hosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (C14-PEG2000; Avanti Polar Lipids) at a molar ratio of 35:16:46.5:2.5 (OF-02/DOPE/cholesterol/C14-PEG2000) with an ionizable lipid/mRNA weight ratio of 10:1. The aqueous and ethanol phases were mixed in a microfluidic device at a 3:1 ratio by syringe pumps to a final mRNA concentration of 0.15 mg ml−1. The resultant formulation was dialysed overnight against PBS in a 20 kDa molecular weight cut-off dialysis cassette (Thermo Fisher Scientific) at 4 °C. Following dialysis, LNPs were concentrated using Amicon 100 kDa molecular weight cut-off centrifugation filters (Sigma-Aldrich) at 4 °C. Total RNA concentration in the LNP solution was measured with a Stunner spectrophotometer (Unchained Labs). LNPs were stored in a PBS solution with 10% (w/v) sucrose at −80 °C until further use.

ZSD mouse models

All animals were Mus musculus. Mixed genetic background mice (B6129F1-Pex1<tm1.1Sjms>/Lutzy/Mmjax) were generated by crossing heterozygous B6.Cg-Pex1tm1.1Sjms/Mmjax (JAX, stock number 25065) with heterozygous 129S6.Cg-Pex1tm1.1Sjms/Mmjax (JAX, stock number 30931). All mice (3–5 per cage) were housed in individually high-efficiency particulate air-filtered polysulfonate cages, under controlled conditions of 12 h/12 h light/dark cycles (6 am to 6 pm), a room temperature of 22 ± 4 °C and 50 ± 15% humidity, with 15 air exchanges per hour. Animals had ad libitum access to acidified water (pH 2.5–3.0) and standard chow. Animals were monitored daily for welfare and survival. All experiments followed National Institutes of Health (NIH) guidelines and The Jackson Laboratory IACUC (protocol 20029–1).

Animal treatments

For neonatal treatments, P1 mice were cryo-anaesthetized and injected intravenously via the facial vein with 5 μl of gene editors or vehicle (PBS). Pups were returned to their dam after recovery. For young adult treatments, P28 mice were anaesthetized with isoflurane and injected retro-orbitally with up to 200 μl of gene editors or vehicle. Mice were returned to their home cage after recovery and monitored daily for survival and adverse effects; those with >20% loss of maximum body weight or body conditioning score <2 were humanely euthanized. For LNP intravenous injections, a 1:1 RNA weight ratio of mRNA-LNP/sgRNA-LNP was administered via retro-orbital injection. Experimental groups were sex-balanced and randomly assigned at dosing, with exact n sizes reported in figure legends. Investigators were blinded to treatment groups during outcome assessment.

Mouse tissue sample preparation

All mice were euthanized at 6 or 16 weeks of age by CO2 asphyxiation and perfused with PBS. The liver was divided into three portions. The upper fifth was fixed in 10% neutral buffered formalin (NBF) for 24 h, rinsed in PBS, paraffin embedded and sectioned at 5 μm for histology. The largest lobe was fixed in 4% paraformaldehyde for 24 h, rinsed in PBS for 15 min and placed in 30% sucrose for 24 h before embedding in optimal cutting temperature solution and cryosectioning at 10 μm for histological staining. Remaining tissue was snap-frozen in liquid nitrogen and stored at −80 °C until use.

Blood collection

Mice were anaesthetized with 2% isoflurane in oxygen at 6 or 16 weeks of age. Blood was collected retro-orbitally into a K2EDTA tube (Thermo Fisher Scientific). Samples were centrifuged at 4 °C at 15,000g for 5 min to separate plasma from blood cells. Plasma and the cell fractions were stored at −80 °C until use.

Histology analysis

ORO staining was used to detect lipids in liver cryosections. Sections were stained with ORO stock solution saturated with isopropanol (Rowley Biochemical Institute), followed by 10% NBF, Mayer’s haematoxylin (Sigma-Aldrich) and lithium carbonate before coverslipping (Corning). PAS was used to detect glycogen in paraffin-embedded liver sections, treated with 0.5% periodic acid for 5 min, Schiff’s reagent (Poly Scientific) for 30 min, and counterstained with Mayer’s haematoxylin (Fisher Scientific). To minimize batch variation, slides for each time point were processed together. Slides were scanned using a Nanozoomer S210 (Hamamatsu) at 40× magnification. Three regions of interest per section were selected at 10× magnification for analysis. The average stained area relative to the total tissue was quantified using CellProfiler (version 4.2.4; pipeline provided in). Supplementary Information

Mouse tissue gDNA purification

Brain, eye and liver frozen tissues were processed using the DNAdvance kit (Beckman Coulter) according to the manufacturer’s instructions. Approximately 50 mg ground frozen tissue was placed in 1.5 ml tubes and lysed in 500–1,000 μl lysis buffer supplemented with proteinase K overnight at 55 °C in a benchtop shaker (800 rpm). Lysates were bead-purified and washed with 70% ethanol. Purified genomic DNA was quantified by spectrophotometry and used for HTS or droplet digital PCR.

RNA-seq and data analysis

Total RNA was extracted from liver using the RNeasy kit (QIAGEN) and quantified using Qubit HS (Thermo Fisher Scientific). Libraries were prepared from 20 ng RNA using a template-switching protocol with Maxima RT (Thermo Fisher Scientific) and KAPA HiFi HotStart ReadyMix (Roche), followed by SPRI bead cleanup (Beckman Coulter). Libraries were indexed with Nextera adapters (Illumina) and sequenced on an Illumina NovaSeq 6000 using S2 chemistry. Samples with <25 million reads were excluded.

FASTQ files were generated with bcl2fastq v2.20, trimmed with Trim Galore v0.6.7, aligned to GENCODE mouse reference genome M31 (GRCm39) using STAR (v2.7.10a), quantified with kallisto and refined to canonical coding sequences using CCDS release 21.

Differential expression was analysed with PyDESeq2 using the model: log treatment sex q ~ + × ( ) + × ( ) . β 0 β 1 β 2

Vehicle-treated WT or Pex1G844D/G844D mice served as the reference level for the treatment variable (depending on the analysis) and male mice as the reference level for the sex variable. Inferred log2 fold-change (FC) values were shrunk with a heavy-tailed Cauchy prior, and P values were calculated according to Wald’s test. Reported effect sizes correspond to post-shrinkage log2FC values.

Volcano plots are shown in Extended Data Fig. 4, and DEGs were further analysed via WebGestalt89 (https://www.webgestalt.org/). KEGG, Phenotype and other related pathway analyses were conducted and shown in Extended Data Fig. 5 and Supplementary Table 5, focusing on the Pex1G844D/G844D vehicle versus WT vehicle at 6 weeks and 16 weeks after injection. Enrichment analyses focused on categories meeting a 1% FDR and defined by 10–1,000 genes, thereby avoiding overly specialized or broad categories.

Analysis of the transcriptome-wide editing RNA-seq data was performed as follows. FASTQ files were generated using Bcl2fastq2, then trimmed using Cutadapt90 v.1.18 to remove adapter sequences, unpaired sequences and low-quality bases. We created sorted bam alignments using STAR v.2.7.11a to align paired reads from each of three biological replicates to the GRCm38/mm10 mouse reference genome (Ensembl). To calculate the average percentage of A-to-I editing among adenosines sequenced in transcriptome-wide sequencing analysis, we used REDItools2. We removed all nucleotides except adenosines from our analysis and then removed all adenosines with a read coverage less than 10 to avoid errors owing to low sampling; in addition, we removed positions with a mapping or read quality score below 25. Next, we calculated the number of adenosines that were converted to an inosine in each sample and divided this by the total number of adenosines in our dataset after filtering to obtain a percentage of adenosines that were edited to inosine in the transcriptome.

To investigate how transcriptome-wide A-to-I editing levels vary with AAV dose and time, we focused our investigation on a set of high-confidence A-to-I off-targets. This set of sites was determined by identifying adenosines with non-zero (≥0.1%) A-to-I editing in either both male high-dose mice at 6 weeks or both female high-dose mice at 6 weeks (22,131 As), then removing all overlap with sites that also exhibit non-zero (≥0.1%) A-to-I editing in at least 2 saline-treated control samples (96,097 As). This filtering procedure left 4,161 adenosines that we investigated as candidate off-target A-to-I editing sites. Jitter plots were generated for each sample by plotting each of the 4,161 off-target sites that showed non-zero editing.

Lipid analysis

Plasma and tissue fatty acids were hydrolysed from triglycerides and phospholipids by acidification with 90:10 (v/v) acetonitrile/6 N HCl, followed by neutralization with 90:10 (v/v) methanol/10 N NaOH. Samples were incubated at 104 °C for 45 min and then reacidified in 6 N HCl. Total fatty acids were extracted in hexane and derivatized91,92 to pentafluorobenzyl esters. Samples were analysed by GC/MS in negative ion chemical ionization mode using ammonia as reagent gas (source pressure, 1.6 Torr) on a SP2560 capillary column (50 m × 0.25 mm × 0.2 μm) with helium carrier gas (column pressure, 20 psi)93. After a 1 μl 1:35 split injection, oven temperature was programmed: 60 °C for 2 min; ramped at 10 °C min−1 to 180 °C; then at 2.5 °C min−1 to 250 °C and held for 14 min (total run time, 55 min). The MSD transfer line was held at 240 °C for 33 min, then increased at 2.5 °C min−1 to 250 °C and held for 32 min. Data were acquired in select ion monitoring mode with dwell times of 50–100 ms per ion. Each analyte was matched to 1 of 13 stable-isotope internal standards. Quantification was based on calibration curves from 50 analyte standards; when pure standards were unavailable, the closest analogue was used. Individual fatty acids were expressed as μg ml−1 in plasma and as μg mg−1 of total lipid content. The lipid FC relative to WT controls for each lipid moiety was calculated using RStudio (version 2024.12.1_563; Supplementary Information) and P values calculated using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data.

Bile acid analysis

Methanol, acetonitrile, water and formic acid (all Optima grade) were purchased from Fisher Scientific. d4-t-CA, d4-g-CA, d4-g-CDCA, d4-CA, d4-CDCA, t-CA, g-CA and CA were purchased from Cayman Chemical; 7-HOCA from Avanti Polar Lipids; and d3-THCA and d3-DHCA, THCA and DHCA from VUMC (Herman ten Brink). Liver tissue (25 mg) homogenized in 50 μl PBS was extracted with 0.5 ml methanol and sonicated for 30 min at room temperature (Branson 3510). Extracts were centrifuged at 600g for 10 min, and supernatants were dried under nitrogen and reconstituted in 240 μl of methanol/H2O (3:1 v/v) containing 100 ng internal standards. Samples were filtered through 0.22 μm nylon Spin-X filters (Costar) and transferred to injection vials (Phenomenex). Chromatographic separation used a Waters TQD with an Acquity UPLC system using an HSST3 column (1.8 μm, 50 × 2.1 mm) at 50 °C in negative-ion electrospray ionization mode. Flow rate was 0.5 ml min−1 with a gradient from 75% A (water/0.1% formic acid) and 25% B (acetonitrile/0.1% formic acid) to 10% A/90% B over 1 min, held for 1–2.25 min and then equilibrated to initial conditions by 3 min. Quantification used Masslynx 4.1 with TargetLynx integration; peak areas were exported to Excel; concentrations were expressed as pmol mg−1 protein and analysed in GraphPad Prism 10.

Western blotting

Liver lysates were prepared in radioimmunoprecipitation assay buffer containing Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific). Protein concentration was determined by detergent-compatible assay (Bio-Rad). Lysates were diluted to 0.75 mg ml−1 with 5× fluorescent master mix and heated at 95 °C for 5 min. ABCD3 protein levels were quantified using automated capillary western blotting (Simple Wes, ProteinSimple), with an anti-PMP70/ABCD3 antibody (Abcam, ab85550) at a 1:50 dilution. Chemiluminescent signals were analysed using Compass for Simple Western (version 6.2.0). ABCD3 peak area was normalized to total protein area94,95. FC values were calculated relative to WT controls.

Droplet digital PCR

Genomic DNA from liver tissue was analysed by droplet digital PCR (ddPCR) using the Bio-Rad QX ONE platform. Reactions were prepared with 100 ng gDNA, target-specific primers and probes (), 2× ddPCR Supermix (Bio-Rad), and processed according to the manufacturer’s guidelines. PCR cycling conditions were 95 °C for 10 min, 50 cycles of 94 °C for 30 s and 58 °C for 2 min, followed by 98 °C for 10 min. Data were analysed using QX ONE software (v1.3). Supplementary Table 2

Off-target editing analysis

CIRCLE-seq off-target analysis was performed as described previously62,96. In brief, gDNA from N2A mouse cells or ZSD patient-derived PEX1G843D/I700fs fibroblasts was isolated using the Gentra Puregene Kit (Qiagen) according to the manufacturer’s instructions. Purified gDNA was sheared with a Covaris S2 instrument to an average length of 300 bp. The fragmented DNA was end repaired, poly-A tailed and ligated to a uracil-containing stem-loop adaptor using the KAPA HTP Library Preparation Kit, PCR Free (KAPA Biosystems). Adaptor-ligated DNA was treated with Lambda Exonuclease (NEB) and E. coli Exonuclease I (NEB), then with USER enzyme (NEB) and T4 polynucleotide kinase (NEB). Intramolecular circularization of the DNA was performed with T4 DNA ligase (NEB), and residual linear DNA was degraded by Plasmid-Safe ATP-dependent DNase (Lucigen). In vitro cleavage reactions were performed with 250 ng of Plasmid-Safe ATP-dependent DNase-treated circularized DNA, 90 nM of SpCas9 nuclease protein, Cas9 nuclease buffer (NEB) and 90 nM of synthetic chemically modified gRNA (Synthego) in 100 μl. Cleaved products were poly-A tailed, ligated with a hairpin adaptor (NEB), treated with USER enzyme (NEB) and amplified by PCR with barcoded universal primers (NEBNext Multiplex Oligos for Illumina, NEB) using Kapa HiFi Polymerase (KAPA Biosystems). Libraries were sequenced with 150-bp paired-end reads on an Illumina MiSeq instrument. CIRCLE-seq data were analysed using open-source CIRCLE-seq analysis software and default recommended parameters (https://github.com/tsailabSJ/circleseq).

Cas-OFFinder analysis was performed online (http://www.rgenome.net/cas-offinder/). Potential off-targets were filtered to ≤3 mismatches for SpCas9. Twenty predicted off-target sites were nominated.

The top 30 CIRCLE-seq-nominated off-target sites for the mouse genome and Cas-OFFinder nominated sites were examined via amplicon sequencing using the gDNA from editor-treated liver tissues or patient-derived fibroblasts, as specified in the texts and figures.

Baculoviral peroxisome-GFP transduction of patient-derived fibroblasts

CellLight peroxisome-GFP, BacMam 2.0 kit was used (ThermoFisher Scientific, C10604). Healthy donor-derived fibroblasts, base editor-treated and non-target guide-treated control fibroblasts were plated onto 24-well plates (Greiner). On the second day, cells were transduced with baculoviral peroxisome-GFP vectors following the manufacture’s recommended protocol, incubated for 16 h and treated with far-red nuclear staining dye (MilliporeSigma) before imaging. Live-cell imaging used the MICA confocal microscope. Incubation temperature was set at 37 °C, and CO2 was set at 5%. The 24-well plate and imaging parameters (‘dynamics and living’) were set for live cell imaging. The absorbance and emission were set for the fluorophores specified in the product manual. A 63× water-immersion objective lens was used to capture images. Treated sample images for comparison were acquired using the same imaging settings. Images were processed further with ImageJ2 (version 2.14.0/1.54f).

Immunofluorescence microscopy of ABE-treated patient-derived fibroblasts

Nucleofected patient-derived and healthy donor-derived fibroblasts were grown on sterile glass coverslips. Cells were washed with PBS and fixed using 3% formaldehyde in PBS for 20 min at room temperature, and permeabilized with 1% Triton X-100 in PBS (5 min). After washing, cells were incubated overnight at 4 °C with primary antibodies: anti-catalase (Thermo Fisher Scientific) and anti-PMP70 (ABCD3) (Abcam). The next day, cells were washed and incubated for 20 min at room temperature with Alexa Fluor 594 goat anti-rabbit IgG (Thermo Fisher Scientific) and Alexa Fluor 488 goat anti-mouse IgG (Thermo Fisher Scientific). Cells on coverslips were washed and mounted onto glass slides with a drop of Prolong Diamond Antifade Mountant (Thermo Fisher Scientific). Images were acquired using Leica MICA-recommended settings for fixed cells and processed with ImageJ2.

Quantification and statistical analysis

The number of independent biological replicates and technical replicates for each experiment is provided in the figure legends or the ‘’ section. Statistical tests were performed using GraphPad Prism version 10.1.1 for Windows and 10.3.0 for Mac (GraphPad Software). Methods

Extended Data

Characterization ofneonatal mice treated with AAV9-ABE8e-V106W. Pex1 G844D/G844D , AAV9-ABE8e-V106W (4×10vg per mouse) was delivered via facial vein injection into Pmice (n = 63). Vehicle-injected WT () mice (n = 40) andmice (n = 52) served as controls. Animals were randomized from multiple litters., Kaplan-Meier survival curves for AAV9-ABE-treated and vehicle-treated mice (n as in)., Editing efficiencies in liver at 6–16 weeks post-treatment, measured by high-throughput sequencing (HTS) in vehicle-treated mice (6 weeks n = 20; 16 weeks n = 19) and AAV9-ABE -treated mice (6 weeks n = 11; 16 weeks n = 8). Adenines within and near the protospacer are plotted. Bars indicate mean ± s.d.; dots represent individual mice., Growth curves for treated and control mice shown as mean ± s.e.m. of body weight for females and males (n as in)., Lipid profile of plasma and liver at 6–16 weeks, analyzed by gas chromatography-mass spectrometry (GC-MS). Lipids fold-change relative to vehicle-treated WT mice is shown in volcano plots. Green and red dotted lines indicate two-fold decrease and increase; black dashed line marks P = 0.05. P values were calculated using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data., Liver lipid content by Oil-Red O (ORO) staining at 6–16 weeks in vehicle-treated WT mice (6 weeks n = 20; 16 weeks n = 15),mice (6 weeks n = 20; 16 weeks n = 18), and AAV9-ABE-treatedmice (6 weeks n = 11; 16 weeks n = 7). Images captured at 20x magnification; scale bar, 50 μm. Statistical analysis by two-way ANOVA with Sidak’s test; significance was set at P < 0.05. Bars represent mean ± s.e.m; dots represent individual mice. Numbers above bars indicate P-values for specified comparisons. Mouse diagram increated in BioRender; Piec, M.(2026). a b a c d a e f a 11 G844D/G844D +/+ G844D/G844D G844D/G844D G844D/G844D 1 Pex1 Pex1 Pex1 Pex1 Pex1 https://biorender.com/n8hnab4

Characterization ofneonatal mice treated with AAV9-ABE8e-V106W. Pex1 G844D/G844D , AAV9-ABE8e-V106W (4×10vg per mouse) was delivered via facial vein injection into Pmice (n = 63). Vehicle-injected WT () mice (n = 40) andmice (n = 52) served as controls. Animals were randomized from multiple litters., Kaplan-Meier survival curves for AAV9-ABE-treated and vehicle-treated mice (n as in)., Editing efficiencies in liver at 6–16 weeks post-treatment, measured by high-throughput sequencing (HTS) in vehicle-treated mice (6 weeks n = 20; 16 weeks n = 19) and AAV9-ABE -treated mice (6 weeks n = 11; 16 weeks n = 8). Adenines within and near the protospacer are plotted. Bars indicate mean ± s.d.; dots represent individual mice., Growth curves for treated and control mice shown as mean ± s.e.m. of body weight for females and males (n as in)., Lipid profile of plasma and liver at 6–16 weeks, analyzed by gas chromatography-mass spectrometry (GC-MS). Lipids fold-change relative to vehicle-treated WT mice is shown in volcano plots. Green and red dotted lines indicate two-fold decrease and increase; black dashed line marks P = 0.05. P values were calculated using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data., Liver lipid content by Oil-Red O (ORO) staining at 6–16 weeks in vehicle-treated WT mice (6 weeks n = 20; 16 weeks n = 15),mice (6 weeks n = 20; 16 weeks n = 18), and AAV9-ABE-treatedmice (6 weeks n = 11; 16 weeks n = 7). Images captured at 20x magnification; scale bar, 50 μm. Statistical analysis by two-way ANOVA with Sidak’s test; significance was set at P < 0.05. Bars represent mean ± s.e.m; dots represent individual mice. Numbers above bars indicate P-values for specified comparisons. Mouse diagram increated in BioRender; Piec, M.(2026). a b a c d a e f a 11 G844D/G844D +/+ G844D/G844D G844D/G844D G844D/G844D 1 Pex1 Pex1 Pex1 Pex1 Pex1 https://biorender.com/n8hnab4

AAV9-GFPstudy and AAV9-ABE8e-V106W treatment ofneonatal mice rescues ZSD phenotypes. in vivo Pex1 G844D/G844D , Kaplan-Meier survival curves for WT (n = 20), andvehicle-treated (n = 21); andmice treated with AAV9-GFP (4×10vg per mouse; n = 20)., Schematic of ABE treatment inmice. Homozygous mice were injected at P1 via facial vein with AAV9-ABE8e-V106W at total doses of 1×10vg (n = 18; 9 females, 9 males), 4×10vg (n = 13; 7 females, 6 males), or 1×10vg per mouse (n = 18; 8 females, 10 males). WT (n = 19; 9 females, 10 males) andmice (n = 21; 12 females, 9 males) injected with vehicle served as control. Animals were randomized from multiple litters., Quantification of N- and C-terminal components of dual AAV9 viral vectors carrying the ABE8e-V106W editor by droplet digital PCR (ddPCR) at 6 and 16 weeks using vector-specific fluorescent probes. Liver gDNA was extracted from vehicle-treated WT mice (6 weeks n = 9; 16 weeks n = 10), vehicle-treatedmice (6 and 16 weeks n = 10), and AAV9-ABE8e-V106W-treatedmice at doses of 1×10vg (6 weeks n = 9; 16 weeks n = 8), 4×10vg (6 weeks n = 4; 16 weeks n = 9), or 1×10vg (6 weeks n = 7; 16 weeks n = 9) per mouse. Bars represent mean ± s.e.m.; dots indicate individual mice. Mouse diagram increated in BioRender; Piec, M.(2026). a b c b Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 G844D/G844D 11 10 10 G844D/G844D G844D/G844D G844D/G844D 11 10 10 https://biorender.com/n8hnab4

AAV9-GFPstudy and AAV9-ABE8e-V106W treatment ofneonatal mice rescues ZSD phenotypes. in vivo Pex1 G844D/G844D , Kaplan-Meier survival curves for WT (n = 20), andvehicle-treated (n = 21); andmice treated with AAV9-GFP (4×10vg per mouse; n = 20)., Schematic of ABE treatment inmice. Homozygous mice were injected at P1 via facial vein with AAV9-ABE8e-V106W at total doses of 1×10vg (n = 18; 9 females, 9 males), 4×10vg (n = 13; 7 females, 6 males), or 1×10vg per mouse (n = 18; 8 females, 10 males). WT (n = 19; 9 females, 10 males) andmice (n = 21; 12 females, 9 males) injected with vehicle served as control. Animals were randomized from multiple litters., Quantification of N- and C-terminal components of dual AAV9 viral vectors carrying the ABE8e-V106W editor by droplet digital PCR (ddPCR) at 6 and 16 weeks using vector-specific fluorescent probes. Liver gDNA was extracted from vehicle-treated WT mice (6 weeks n = 9; 16 weeks n = 10), vehicle-treatedmice (6 and 16 weeks n = 10), and AAV9-ABE8e-V106W-treatedmice at doses of 1×10vg (6 weeks n = 9; 16 weeks n = 8), 4×10vg (6 weeks n = 4; 16 weeks n = 9), or 1×10vg (6 weeks n = 7; 16 weeks n = 9) per mouse. Bars represent mean ± s.e.m.; dots indicate individual mice. Mouse diagram increated in BioRender; Piec, M.(2026). a b c b Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 G844D/G844D 11 10 10 G844D/G844D G844D/G844D G844D/G844D 11 10 10 https://biorender.com/n8hnab4

Assessment of ABE8e-V106W editing in CNS tissues fromneonatal mice treated with three AAV9 doses at P1. Pex1 G844D/G844D , Base editing efficiencies were quantified via HTS and analyzed with CRISPResso2. Brain cortex tissues were collected from vehicle-treated WT andmice, or AAV9-ABE8e-V106W-treatedmice (1×10vg, 4×10vg, or 1×10vg per mouse) at 16 weeks. Adenines within and near the protospacer are shown., Quantification of base editing efficiencies in the cerebellum., Quantification of base editing efficiencies in whole eye tissue. For AAV9-ABE-V106W dosing groups, n = 8 mice (1×10vg), n = 9 mice (4×10vg), n = 9 mice (1×10vg). Vehicle-treatedmice, n = 11; vehicle-treated WT mice, n = 10., Lipid profile of plasma samples collected at 6–16 weeks (n as in), analyzed by GC-MS. Fold-change in lipid abundance was calculated relative to vehicle-treated WT mice and shown in volcano plots. Green and red dotted lines represent twofold decrease and increase in lipid abundance, respectively; black dashed line marks P = 0.05. P values were calculated relative to vehicle-treated WT samples using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data. a b c d b Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 10 10 11 10 10 G844D/G844D

Assessment of ABE8e-V106W editing in CNS tissues fromneonatal mice treated with three AAV9 doses at P1. Pex1 G844D/G844D , Base editing efficiencies were quantified via HTS and analyzed with CRISPResso2. Brain cortex tissues were collected from vehicle-treated WT andmice, or AAV9-ABE8e-V106W-treatedmice (1×10vg, 4×10vg, or 1×10vg per mouse) at 16 weeks. Adenines within and near the protospacer are shown., Quantification of base editing efficiencies in the cerebellum., Quantification of base editing efficiencies in whole eye tissue. For AAV9-ABE-V106W dosing groups, n = 8 mice (1×10vg), n = 9 mice (4×10vg), n = 9 mice (1×10vg). Vehicle-treatedmice, n = 11; vehicle-treated WT mice, n = 10., Lipid profile of plasma samples collected at 6–16 weeks (n as in), analyzed by GC-MS. Fold-change in lipid abundance was calculated relative to vehicle-treated WT mice and shown in volcano plots. Green and red dotted lines represent twofold decrease and increase in lipid abundance, respectively; black dashed line marks P = 0.05. P values were calculated relative to vehicle-treated WT samples using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data. a b c d b Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 10 10 11 10 10 G844D/G844D

Analysis of differentially expressed genes in vehicle- or AAV9-ABE8e-V106W-treatedneonatal mice compared to WT controls, related to. Pex1 G844D/G844D Fig. 3 , RNA-seq was performed and analyzed as indicated in the. A false discovery rate (FDR) < 1% was applied to identify differentially expressed genes (DEGs) among vehicle-treated WT mice, vehicle-treatedmice, and AAV9-ABE8e-V106W-treatedmice. Volcano plots display DEGs for the following comparisons (from top to bottom):vehicle versus WT vehicle;treated with 1×10vg AAV9-ABE versus WT vehicle;treated with 4×10vg AAV9-ABE versus WT vehicle; andtreated with 1×10vg AAV9-ABE versus WT vehicle. Cutoff thresholds are indicated by black dashed lines. RNA-seq was performed using liver tissue samples: vehicle-treated WT mice (6 weeks n = 4; 16 weeks n = 3), vehicle-treatedmice (6 weeks n = 3; 16 weeks n = 4), and AAV9-ABE8e-V106W treatedmice (1×10vg: 6 weeks n = 4, 16 weeks n = 4; 4×10vg: 6 weeks n = 4, 16 weeks n = 4; 1×10vg: 6 weeks n = 7, 16 weeks n = 4). Analyses of samples collected at 6 weeks post-injection are shown., Samples at 16 weeks post injection were analyzed and plotted as in. a b a Methods Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D G844D/G844D G844D/G844D 11 G844D/G844D 10 G844D/G844D 10 G844D/G844D G844D/G844D 11 10 10

Analysis of differentially expressed genes in vehicle- or AAV9-ABE8e-V106W-treatedneonatal mice compared to WT controls, related to. Pex1 G844D/G844D Fig. 3 , RNA-seq was performed and analyzed as indicated in the. A false discovery rate (FDR) < 1% was applied to identify differentially expressed genes (DEGs) among vehicle-treated WT mice, vehicle-treatedmice, and AAV9-ABE8e-V106W-treatedmice. Volcano plots display DEGs for the following comparisons (from top to bottom):vehicle versus WT vehicle;treated with 1×10vg AAV9-ABE versus WT vehicle;treated with 4×10vg AAV9-ABE versus WT vehicle; andtreated with 1×10vg AAV9-ABE versus WT vehicle. Cutoff thresholds are indicated by black dashed lines. RNA-seq was performed using liver tissue samples: vehicle-treated WT mice (6 weeks n = 4; 16 weeks n = 3), vehicle-treatedmice (6 weeks n = 3; 16 weeks n = 4), and AAV9-ABE8e-V106W treatedmice (1×10vg: 6 weeks n = 4, 16 weeks n = 4; 4×10vg: 6 weeks n = 4, 16 weeks n = 4; 1×10vg: 6 weeks n = 7, 16 weeks n = 4). Analyses of samples collected at 6 weeks post-injection are shown., Samples at 16 weeks post injection were analyzed and plotted as in. a b a Methods Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D G844D/G844D G844D/G844D 11 G844D/G844D 10 G844D/G844D 10 G844D/G844D G844D/G844D 11 10 10

KEGG pathway and phenotype analyses of differentially expressed genes in treated neonatal mouse liver tissues. , Differentially expressed genes between vehicle-treated WT and vehicle-treatedmice at 6 weeks post-injection were analyzed using WebGestalt for gene set analysis. All KEGG categories with FDR < 0.01 (Benjamini-Hochberg method) are shown. Categories were restricted to those defined by 10–1000 genes, to exclude overly specialized (< 10 genes) or overly broad (> 1000 genes) sets.Phenotype enrichment analyses, with enrichment ratio indicated on the x-axis., Differentially expressed genes between vehicle-treated WT and vehicle-treatedmice at 16 weeks post injection were analyzed using the same criteria as in. KEGG pathway analysis is shown (The number of tissue samples is the same as in). a b, d c a Pex1 Pex1 G844D/G844D G844D/G844D Extended Data Fig. 4

KEGG pathway and phenotype analyses of differentially expressed genes in treated neonatal mouse liver tissues. , Differentially expressed genes between vehicle-treated WT and vehicle-treatedmice at 6 weeks post-injection were analyzed using WebGestalt for gene set analysis. All KEGG categories with FDR < 0.01 (Benjamini-Hochberg method) are shown. Categories were restricted to those defined by 10–1000 genes, to exclude overly specialized (< 10 genes) or overly broad (> 1000 genes) sets.Phenotype enrichment analyses, with enrichment ratio indicated on the x-axis., Differentially expressed genes between vehicle-treated WT and vehicle-treatedmice at 16 weeks post injection were analyzed using the same criteria as in. KEGG pathway analysis is shown (The number of tissue samples is the same as in). a b, d c a Pex1 Pex1 G844D/G844D G844D/G844D Extended Data Fig. 4

Analysis of ABE-AAV-treated 4-week-oldmice, related to. Pex1 G844D/G844D Fig. 4 , Schematic of retro-orbital injection of 4-week-oldmice (n = 7) with AAV9-ABE8e-V106W (5×10vg per mouse) to correct the-p.G844D pathogenic allele. WT (n = 6) and(n = 5) injected with vehicle, served as controls. Mice randomized from multiple litters (mixed sex)., Lipid profile of plasma at 10 weeks by GC-MS; volcano plots show fold-change relative to WT. Green and red dotted lines indicates twofold decrease and increase; black dashed line marks P = 0.05. P values calculated using Dunn’s non-parametric many-to-one comparison test for Kruskal-type ranked data., Histological analysis of hepatic lipid content by Oil Red O (ORO) staining and glycogen content by Periodic Acid Schiff (PAS) staining at 16 weeks in vehicle-treated WT mice, vehicle-treatedmice, and ABE8e-V106W-treatedwith 5×10dew per mouse. Bars represent mean ± s.e.m.; dots indicate individual mice, n as in a. Mouse diagram increated in BioRender; Piec, M.(2026). a b c a Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D 11 G844D/G844D G844D/G844D G844D/G844D 11 https://biorender.com/l4insv3

Analysis of ABE-AAV-treated 4-week-oldmice, related to. Pex1 G844D/G844D Fig. 4 , Schematic of retro-orbital injection of 4-week-oldmice (n = 7) with AAV9-ABE8e-V106W (5×10vg per mouse) to correct the-p.G844D pathogenic allele. WT (n = 6) and(n = 5) injected with vehicle, served as controls. Mice randomized from multiple litters (mixed sex)., Lipid profile of plasma at 10 weeks by GC-MS; volcano plots show fold-change relative to WT. Green and red dotted lines indicates twofold decrease and increase; black dashed line marks P = 0.05. P values calculated using Dunn’s non-parametric many-to-one comparison test for Kruskal-type ranked data., Histological analysis of hepatic lipid content by Oil Red O (ORO) staining and glycogen content by Periodic Acid Schiff (PAS) staining at 16 weeks in vehicle-treated WT mice, vehicle-treatedmice, and ABE8e-V106W-treatedwith 5×10dew per mouse. Bars represent mean ± s.e.m.; dots indicate individual mice, n as in a. Mouse diagram increated in BioRender; Piec, M.(2026). a b c a Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D 11 G844D/G844D G844D/G844D G844D/G844D 11 https://biorender.com/l4insv3

Analysis of ABE-LNP-treated 4-week-old mice. , Schematic of retro-orbital (RO) injection of ABE-LNP into 4-week-old mice., Base editing efficiencies in liver quantified via HTS and CRISPResso2analysis in vehicle-treated mice (WT n = 7;n = 5), sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 6; 1.5 mg/kg n = 5) or sgNT-ABE-LNP-treated mice (non-targeting guide, 3 mg/kg n = 6). Adenines within and near the protospacer are shown., Base editing efficiencies in brain cortex tissue (n as in). Bars represent the mean ± s.d.; dots indicate individual mice (,)., Automated Western blot analysis of ABCD3 levels in liver homogenates from vehicle- or ABE-LNP-treated mice (1.5 mg/kg or 3 mg/kg; n as in). Peak area for ABCD3 was normalized to total protein, and fold-change was calculated relative to WT. Statistical analysis was performed by two-way ANOVA and Sidak’s multiple comparison test., Histological analysis of hepatic lipid content by Oil Red O (ORO) staining and glycogen content by Periodic Acid Schiff (PAS) staining. Representative liver images at 10 weeks of age for vehicle-treated WT mice (n = 7),mice (n = 5), or ABE8e-V106W-treatedmice injected with 3 mg/kg (n = 5) or 1.5 mg/kg (n = 5) LNP per mouse. Images captured at 20x magnification; scale bar, 50 μm. Bars represent mean ± s.e.m.; dots indicate individual mice (). Statistical analysis was performed using two-way ANOVA and Sidak’s multiple comparison test. P-values above bars indicate comparisons using the statistical method specified above. Mouse and LNP diagram increated in BioRender; Piec, M.(2026). a b c b b c d b e d, e a [99] G844D/G844D G844D/G844D G844D/G844D G844D/G844D Pex1 Pex1 Pex1 Pex1 https://biorender.com/1kdiodh

Analysis of ABE-LNP-treated 4-week-old mice. , Schematic of retro-orbital (RO) injection of ABE-LNP into 4-week-old mice., Base editing efficiencies in liver quantified via HTS and CRISPResso2analysis in vehicle-treated mice (WT n = 7;n = 5), sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 6; 1.5 mg/kg n = 5) or sgNT-ABE-LNP-treated mice (non-targeting guide, 3 mg/kg n = 6). Adenines within and near the protospacer are shown., Base editing efficiencies in brain cortex tissue (n as in). Bars represent the mean ± s.d.; dots indicate individual mice (,)., Automated Western blot analysis of ABCD3 levels in liver homogenates from vehicle- or ABE-LNP-treated mice (1.5 mg/kg or 3 mg/kg; n as in). Peak area for ABCD3 was normalized to total protein, and fold-change was calculated relative to WT. Statistical analysis was performed by two-way ANOVA and Sidak’s multiple comparison test., Histological analysis of hepatic lipid content by Oil Red O (ORO) staining and glycogen content by Periodic Acid Schiff (PAS) staining. Representative liver images at 10 weeks of age for vehicle-treated WT mice (n = 7),mice (n = 5), or ABE8e-V106W-treatedmice injected with 3 mg/kg (n = 5) or 1.5 mg/kg (n = 5) LNP per mouse. Images captured at 20x magnification; scale bar, 50 μm. Bars represent mean ± s.e.m.; dots indicate individual mice (). Statistical analysis was performed using two-way ANOVA and Sidak’s multiple comparison test. P-values above bars indicate comparisons using the statistical method specified above. Mouse and LNP diagram increated in BioRender; Piec, M.(2026). a b c b b c d b e d, e a [99] G844D/G844D G844D/G844D G844D/G844D G844D/G844D Pex1 Pex1 Pex1 Pex1 https://biorender.com/1kdiodh

Lipid profiles in ABE-AAV9-treated or ABE-LNP-treated mouse brain cortex and liver tissues. , Lipid profile of ABE-AAV9-treated P1 neonatal mice based on GC-MS analysis of brain cortex samples collected at 6 and 16 weeks. Cortex of vehicle-treated WT mice (6 weeks n = 9, 16 weeks n = 10) Vehicle-treatedmice (6 weeks n = 9, 16 weeks n = 11), AAV9-ABE8e-V106W-treatedmice at doses 1×10vg (6 weeks n = 9, 16 weeks n = 7)., Lipid profiles of ABE-AAV9-treated 4-week-old mice based on GC-MS analysis of brain cortex samples collected at 16 weeks., Lipid profiles of ABE-LNP-treated 4-week-old mice based on GC-MS analysis of liver and brain cortex samples collected at 16 weeks. Vehicle-treated WT mice (n = 6) Vehicle-treatedmice (n = 6), AAV9-ABE8e-V106W-treatedmice at 5×10vg (n = 7),. Liver of vehicle-treated mice (WT n = 8;n = 5), sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 5; 1.5 mg/kg n = 4). Cortex of vehicle-treated mice (WT n = 8,n = 5, sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 5; 1.5 mg/kg n = 5). For, fold-change in signal for each lipid was calculated relative to the average for vehicle-treated WT mice. Volcano plots show fold-change in lipid abundance. Green and red dotted lines indicate twofold decrease and increase, respectively; black dashed line marks P = 0.05. P values were calculated relative to vehicle-treated WT samples using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data. a b c c a-c Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 G844D/G844D G844D/G844D 11 G844D/G844D G844D/G844D G844D/G844D G844D/G844D

Lipid profiles in ABE-AAV9-treated or ABE-LNP-treated mouse brain cortex and liver tissues. , Lipid profile of ABE-AAV9-treated P1 neonatal mice based on GC-MS analysis of brain cortex samples collected at 6 and 16 weeks. Cortex of vehicle-treated WT mice (6 weeks n = 9, 16 weeks n = 10) Vehicle-treatedmice (6 weeks n = 9, 16 weeks n = 11), AAV9-ABE8e-V106W-treatedmice at doses 1×10vg (6 weeks n = 9, 16 weeks n = 7)., Lipid profiles of ABE-AAV9-treated 4-week-old mice based on GC-MS analysis of brain cortex samples collected at 16 weeks., Lipid profiles of ABE-LNP-treated 4-week-old mice based on GC-MS analysis of liver and brain cortex samples collected at 16 weeks. Vehicle-treated WT mice (n = 6) Vehicle-treatedmice (n = 6), AAV9-ABE8e-V106W-treatedmice at 5×10vg (n = 7),. Liver of vehicle-treated mice (WT n = 8;n = 5), sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 5; 1.5 mg/kg n = 4). Cortex of vehicle-treated mice (WT n = 8,n = 5, sgPex1-ABE-LNP-treatedmice (3 mg/kg n = 5; 1.5 mg/kg n = 5). For, fold-change in signal for each lipid was calculated relative to the average for vehicle-treated WT mice. Volcano plots show fold-change in lipid abundance. Green and red dotted lines indicate twofold decrease and increase, respectively; black dashed line marks P = 0.05. P values were calculated relative to vehicle-treated WT samples using Dunn’s nonparametric many-to-one comparison test for Kruskal-type ranked data. a b c c a-c Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 Pex1 G844D/G844D G844D/G844D 11 G844D/G844D G844D/G844D 11 G844D/G844D G844D/G844D G844D/G844D G844D/G844D

Off-target analyses of ABE8e-V106W-treated mouse tissues and human cells, related to. Fig. 5 , Jitter plots derived from RNA-seq experiments showing RNA adenines modified in ABE8e-V106W-AAV9-treated or saline-treatedmouse liver tissues. Each dot represents an A-to-G change detected and n indicates the average of total number of modified adenines per liver sample at the same dose., Amplicon sequencing analysis of CIRCLE-seq-nominated off-target (OT) sites. The top 30 OT sites andon-target site were amplified and sequenced from liver genomic DNA (gDNA) of four-week-oldmice treated AAV9-ABE8e-V106W (5×10vg), ABE-LNP, or vehicle. Editing efficiencies were measured by HTS and analyzed with CRISPResso2. n = 3 mice with the highest on-target editing efficiency were used., CIRCLE-seq analysis of ABE8e-V106W targeting the human-p.G843D pathogenic variant was performed using patient-derived fibroblast () gDNA. Six potential OT editing sites were nominated., Schematic diagram showing nominated OT editing sites from CIRCLE-seq and Cas-OFFinder, including 20 potential OT sites nominated by Cas-OFFinder. Mismatched nucleotides are shown in red. No overlapping OT sites were observed between CIRCLE-seq- and Cas-OFFinder-nominated sites, as illustrated in the Venn diagram. Twenty-six potential OT sites were further examined by amplicon sequencing of liver gDNA from AAV9-ABE8e-V106W-treated mice (1X10vg). a b c d Pex1 Pex1 Pex1 PEX1 PEX1 G844D/G844D G844D/G844D 11 G843D/I700fs 11

Off-target analyses of ABE8e-V106W-treated mouse tissues and human cells, related to. Fig. 5 , Jitter plots derived from RNA-seq experiments showing RNA adenines modified in ABE8e-V106W-AAV9-treated or saline-treatedmouse liver tissues. Each dot represents an A-to-G change detected and n indicates the average of total number of modified adenines per liver sample at the same dose., Amplicon sequencing analysis of CIRCLE-seq-nominated off-target (OT) sites. The top 30 OT sites andon-target site were amplified and sequenced from liver genomic DNA (gDNA) of four-week-oldmice treated AAV9-ABE8e-V106W (5×10vg), ABE-LNP, or vehicle. Editing efficiencies were measured by HTS and analyzed with CRISPResso2. n = 3 mice with the highest on-target editing efficiency were used., CIRCLE-seq analysis of ABE8e-V106W targeting the human-p.G843D pathogenic variant was performed using patient-derived fibroblast () gDNA. Six potential OT editing sites were nominated., Schematic diagram showing nominated OT editing sites from CIRCLE-seq and Cas-OFFinder, including 20 potential OT sites nominated by Cas-OFFinder. Mismatched nucleotides are shown in red. No overlapping OT sites were observed between CIRCLE-seq- and Cas-OFFinder-nominated sites, as illustrated in the Venn diagram. Twenty-six potential OT sites were further examined by amplicon sequencing of liver gDNA from AAV9-ABE8e-V106W-treated mice (1X10vg). a b c d Pex1 Pex1 Pex1 PEX1 PEX1 G844D/G844D G844D/G844D 11 G843D/I700fs 11

Assessment of ABE8e-V106W editing of patient-derived fibroblasts and live-cell imaging of peroxisomal assembly, related to. Fig. 6 , Two patient-derived fibroblast cell lines—homozygousand compound heterozygous—were nucleofected with ABE8e-V106W mRNA and a synthetic guide RNA targeting the-p.G843D (c.2528 G > A) pathogenic variant or a non-targeting control guide RNA (sgNT) that does not match any human genome sequence. Editing efficiencies were measured 72 h post-nucleofection by HTS. The indels and adenines modified near or within the protospacer were analyzed and plotted. n = 2 replicates were performed., Edited cells were confirmed by HTS, passaged, and seeded onto a 24-well plate. Cells were transduced with baculoviral vectors expressing a modified GFP fused to a C-terminal peroxisomal targeting signal 1 (PTS1). Live-cell imaging was performed using a MICA confocal microscope. Imaging was conducted under blinded conditions for sgNT and-p.G843D-targeting guide identities. Two biological replicates were performed and > 30 cells were imaged and analyzed. Scale bar, 20 μm. a b PEX1 PEX1 PEX1 PEX1 G843D/G843D G843D/I700fs

Assessment of ABE8e-V106W editing of patient-derived fibroblasts and live-cell imaging of peroxisomal assembly, related to. Fig. 6 , Two patient-derived fibroblast cell lines—homozygousand compound heterozygous—were nucleofected with ABE8e-V106W mRNA and a synthetic guide RNA targeting the-p.G843D (c.2528 G > A) pathogenic variant or a non-targeting control guide RNA (sgNT) that does not match any human genome sequence. Editing efficiencies were measured 72 h post-nucleofection by HTS. The indels and adenines modified near or within the protospacer were analyzed and plotted. n = 2 replicates were performed., Edited cells were confirmed by HTS, passaged, and seeded onto a 24-well plate. Cells were transduced with baculoviral vectors expressing a modified GFP fused to a C-terminal peroxisomal targeting signal 1 (PTS1). Live-cell imaging was performed using a MICA confocal microscope. Imaging was conducted under blinded conditions for sgNT and-p.G843D-targeting guide identities. Two biological replicates were performed and > 30 cells were imaged and analyzed. Scale bar, 20 μm. a b PEX1 PEX1 PEX1 PEX1 G843D/G843D G843D/I700fs

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41551-026-01651-5.

Funding

Competing interests

9 of 23
authors report competing interests
14 report none
PubMed

What Lands in Your Inbox Each Week:

  • 📚7 fresh studies
  • 📝plain-language summaries
  • ✅direct links to original studies
  • 🏅top journal indicators
  • 📅weekly delivery
  • 🧘‍♂️always free