In vivo genome editing at the albumin locus to treat methylmalonic acidemia.

Methylmalonic acidemia (MMA) is a metabolic disorder most commonly caused by mutations in the methylmalonyl-CoA mutase (MMUT) gene. Although adeno-associated viral (AAV) gene therapy has been effective at correcting the disease phenotype in MMA mouse models, clinical translation may be impaired by loss of episomal transgene expression and magnified by the need to treat patients early in life. To achieve permanent correction, we developed a dual AAV strategy to express a codon-optimized MMUT transgene from Alb and tested various CRISPR-Cas9 genome-editing vectors in newly developed knockin mouse models of MMA. For one target site in intron 1 of Alb, we designed rescue cassettes expressing MMUT behind a 2A-peptide or an internal ribosomal entry site sequence. A second guide RNA targeted the initiator codon, and the donor cassette encompassed the proximal albumin promoter in the 5' homology arm. Although all editing approaches were therapeutic, targeting the start codon of albumin allowed the use of a donor cassette that also functioned as an episome and after homologous recombination, even without the expression of Cas9, as an integrant. Targeting the albumin locus using these strategies would be effective for other metabolic disorders where early treatment and permanent long-term correction are needed.

Introduction

Methylmalonic acidemia (MMA) is a prototypical autosomal-recessive metabolic disease, most commonly caused by mutations in the gene coding for methylmalonyl-CoA mutase (MMUT), the terminal enzyme in the catabolic pathway of isoleucine, valine, odd-chained fatty acids, cholesterol, and propionate oxidation. As with other severe metabolic disorders of infancy and childhood, treatment options remain limited, and long-term outcomes guarded. The crux of disease management continues to be the dietary restriction of metabolic pathway precursors, and aggressive management of intercurrent illnesses., However, even with vigilant monitoring, MMA is associated with increased morbidity and mortality due to acute and chronic multi-systemic disease manifestations. The guarded prognosis of MMA has led to the use of elective liver transplantation increasingly at younger ages, to improve survival and delay long-term disease complications.5, 6, 7, 8 Successful liver transplant eliminates the propensity toward metabolic instability characteristic of MMA and related organic acidemias, but carries the need for lifelong immune suppression and the risk of developing transplant-related complications, including post-transplant lymphoproliferative disorder. Thus, there is a great need for non-invasive therapies that could provide sufficient hepatic enzyme activity, that ideally would persist for the life of the affected individual.

In an effort to establish new treatments for MMA, we previously developed a number of new genomic therapies, including recombinant adeno-associated viral (rAAV) gene therapy and systemic MMUT mRNA therapy.11, 12, 13, 14 Both approaches have intrinsic limitations related to the transient nature of expression, untoward immune reactions that might limit re-administration and, in the case of rAAV, the potential for genotoxicity caused by random integration of vectors containing strong promoter elements. A distinct strategy to circumvent the limitations of transient and ectopic MMUT expression would be to use genome editing to target the insertion of a MMUT cDNA to a safe harbor locus, an area of the genome that can be manipulated without adverse effects, allowing for stable transgene expression in the liver. This strategy would provide the corrective benefit of conventional rAAV gene therapy, while enabling permanent correction after a single treatment.

To examine genome editing as a treatment for MMA in newly constructed knockin mouse models that recapitulate severe (muto) and partial (mut−) deficiency forms of the disorder, we targeted the safe harbor locus albumin (Alb). AAV delivery of therapeutic MMUT donor cassettes, either into the first intron of Alb, or at the initiator codon in exon 1 of Alb with or without Cas9 endonuclease, provided pronounced clinical and biochemical effects. Targeting the 5′ end of albumin using the strategies described here can be readily applied to other metabolic disorders, and the nuclease-free insertion approach targeting at the albumin start codon may fully harness the potential for episomal expression of the donor cassette, circumvent genotoxic off-target activity, and protect against untoward effects from pre-existing immunity to Cas9 orthologs.

Results

Murine models of MMA recapitulate patient phenotypes and respond to conventional AAV gene therapy

To model the diversity of pathogenic human mutations and phenotypes, and their corresponding murine phenotypes, we used CRISPR-Cas9 genome editing to knock in characteristic muto (MMUT p.R108C) or mut− (MMUT p.G717V) missense mutations (Figure 1A) at the orthologous positions in the Mmut gene. We then characterized the clinical and biochemical phenotypes of mice with homozygous missense Mmut alleles, and established the response to conventional AAV gene addition therapy.

Figure 1

Phenotypic and biochemical characterization of MMA mouse models recapitulating patient missense mutations before and after treatment with therapeutic AAV8 or AAV9 vectors

(A) Murine Mmut alleles generated using CRISPR-Cas9 genome editing. MMUTp.R108C (murine Mmutp.R106C), located in the region coding for the substrate binding domain, is classified as mut0: there is no detectable MMUT activity, resulting in the more common, and severe, form of the disorder. MMUTp.G717V (murine Mmutp.G715V) is located in the putative cofactor binding pocket and is classified as mut−, exhibiting partial activity in vitro. (B) Survival of Mmutp.G715V/p.G715V mice. There was no significant difference in the mean survival between healthy Mmut+/p.G715V and diseased Mmutp.G715V/p.G715V mice. (C) Plasma methylmalonic acid concentrations in Mmutp.G715V/p.G715V mice. Plasma methylmalonic acid was significantly elevated in Mmutp.G715V/p.G715V compared with age-matched Mmut+/p.G715V controls at 16 weeks. Methylmalonic acid concentrations in Mmutp.G715V/p.G715V at 16 weeks treated with 5 × 1012 GC/kg AAV8.EF1a.PI.MMUT.RBG dropped significantly, by over 75%, 3 weeks post-treatment. (D) Growth trends in Mmutp.G715V/p.G715V. Throughout the first 6 months of life, untreated Mmutp.G715V/p.G715V only reach ∼70% of the weight of their Mmut+/p.G715V littermates. Treatment on DOL41 with 5 × 1012 GC/kg AAV8.EF1a.PI.MMUT.RBG led to increased body weight and growth comparable with Mmut+/p.G715V controls by DOL60. (E) Survival of Mmutp.R106C/p.R106C mice. Only 6% of Mmutp.R106C/p.R106C survived to DOL57; however, treatment of Mmutp.R106C/p.R106C mice at birth via retro-orbital injection with 1e11 GC/pup AAV9.EF1a.PI.MMUT.RBG rescued mutant mice from neonatal lethality. (F) Plasma methylmalonic acid concentrations in Mmutp.R106C/p.R106C mice on DOL0 are massively elevated compared with Mmut+/p.R106C littermates. After treatment at birth with 1e11 GC/pup AAV9.EF1a.PI.MMUT.RBG, methylmalonic acid concentrations in Mmutp.R106C/p.R106C were significantly reduced at age 5 months. (G) Growth trends in Mmutp.R106C/p.R106C mice for which untreated controls do not survive past 60 days. Untreated Mmutp.R106C/p.R106C mice that survived past weaning were severely growth impaired compared with Mmut+/p.R106C littermates at age 1 month, but treatment with 5 × 1012 GC/kg AAV8.EF1a.PI.MMUT.RBG on DOL52 improved growth to levels not significantly different from healthy Mmut+/p.R106C controls by DOL60. In (B) and (E), data were analyzed by log rank Mantel-Cox and Gehan-Breslow-Wilcoxon test. ∗∗p < 0.01, versus the untreated Mmutp.R106C/p.R106C group. In (C) and (F), values are expressed as mean ± SD and analyzed by unpaired Student's t test. In (C), methylmalonic acid concentrations for untreated Mmutp.G715V/p.G715V (802.5 ± 218.7 μM), age-matched Mmut+/p.G715V (5.66 ± 1.60 μM, n = 4), and Mmutp.G715V/p.G715V treated with AAV8.EF1a.PI.MMUT.RBG (71.57 ± 22.09 μM, n = 3). In (F), methylmalonic acid concentrations in untreated Mmutp.R106C/p.R106C at birth (2,931 ± 820.5 μM, n = 5) compared with Mmut+/p.R106C littermates (11.07 ± 1.63 μM, n = 5), and 5 months post-treatment with AAV9.EF1a.PI.MMUT.RBG (588 ± 49.0 μM, n = 4). ∗∗∗p < 0.001. In (D) and (G), weight values are expressed as mean ± SD and analyzed by unpaired Student's t test. At age 1 month, untreated Mmutp.R106C/p.R106C (7.35 ± 0.79 g) versus healthy Mmut+/p.R106C controls (18.72 ± 2.66 g, ∗∗∗∗p < 0.0001). Adult Mmutp.R106C/p.R106C mice treated with 5 × 1012 GC/kg AAV8.EF1a.PI.MMUT.RBG on DOL52 improved growth to levels not significantly different from Mmut+/p.R106C littermates by DOL60 (17.7 ± 1.32 g for treated Mmutp.R106C/p.R106C versus 22.5 ± 3.30 g for Mmut+/p.R106C).

Mmutp.R106C/p.R106C and Mmutp.G715V/p.G715V mice were observed from birth and their survival compared with unaffected Mmut+/p.R106C and Mmut+/p.G715V littermates (Figures 1B and 1E). Less than 10% of Mmutp.R106C/p.R106C (n = 37) survived to day of life 57 (DOL57), with the greatest mortality observed during the first 2 weeks of life: more than half of the homozygous mutant mice perished (Figure 1E). After treatment at birth with an AAV serotype 9 vector expressing a human codon-optimized human MMUT from a ubiquitous promoter, the majority of the treated mutant Mmutp.R106C/p.R106C survived normally compared with carrier controls (Figure 1E). We similarly followed the survival of Mmutp.G715V/p.G715V and compared it with wild-type Mmut+/p.G715V littermates (Figure 1B). At 111 days, no significant difference in the mean survival was observed between healthy Mmut+/p.G715V and affected Mmutp.G715V/p.G715V mice, indicating that this MMA model does not exhibit disease-related mortality, similar to what we observed in a transgenic overexpression mouse model of this mutation.

Because Mmutp.R106C/p.R106C mice manifested a more severe phenotype, with disease-associated lethality occurring predominantly in the first week of life, we euthanized a litter of Mmutp.R106C/p.R106C on DOL1 to measure methylmalonic acid concentrations. Indeed, we observed that plasma methylmalonic acid concentrations were massively elevated, with values of 2,931 ± 820.5 μM (Figure 1F), close to the values seen in Mmut exon 3 knockout mice sampled at a similar age. After neonatal treatment with a dose of 1e11 GC/pup AAV9.EF1a.PI.MMUT.RBG, metabolite concentrations in Mmutp.R106C/p.R106C were significantly reduced, comparable with concentrations observed at age 1 year in Mmut exon 3 knockout mice treated at birth with a dose of either 1e11 GC/pup or 2e11 GC/pup of an AAV8 vector expressing murine Mmut from a chicken beta-actin promoter (Figure 1F). In comparison, methylmalonic acid levels in the blood of Mmutp.G715V/p.G715V were significantly elevated compared with age-matched Mmut+/p.G715V controls, but much lower than the levels measured in younger Mmutp.R106C/p.R106C mice, with concentrations of 802.5 ± 218.7 μM at 16 weeks (Figure 1C).

Growth, as assessed by weight in Mmutp.R106C/p.R106C mice from age 1 month and in Mmutp.G715V/p.G715V from weaning, was measured to determine whether the clinical phenotype aligned with observations made from MMA patients, many of whom exhibit growth retardation. As expected, based on the mortality and biochemical parameters, Mmutp.R106C/p.R106C that survive beyond weaning were severely growth impaired, with weights that were 50% less than age-matched controls (Figure 1G).

During the first 4 months of life, Mmutp.G715V/p.G715V display retarded growth compared with wild-type Mmut+/p.G715V controls when fed a regular chow diet (Figure 1D). To assay the effects of traditional AAV gene therapy, a small cohort of Mmutp.G715V/p.G715V (n = 3) was treated with 5 × 1012 GC/kg of AAV8.EF1a.PI.MMUT.RBG gene, delivered via the retro-orbital plexus, at age 1 month. AAV8-treated Mmutp.G715V/p.G715V mice increased their body weight by an average of 45% within a week of treatment, and grew to be not significantly different in size from Mmut+/p.G715V littermates (Figure 1D). A brisk metabolic response also was noted, with plasma methylmalonic acid concentrations dropped by over 75% to less than 100 μM in Mmut p.G715V/p.G715V 3 weeks after treatment with AAV8.EF1a.PI.MMUT.RBG (Figure 1C).

Due to the high rate of mortality, only two Mmutp.R106C/p.R106C male mice were treated with systemic AAV8.EF1a.PI.MMUT.RBG gene therapy after weaning. By 15 days post-treatment, both MMA mice had doubled in weight to sizes not significantly different from unaffected littermates (Figure 1G). Because Mmutp.R106C/p.R106C further displayed pronounced lethality before weaning, we additionally performed a neonatal gene therapy study using an identical vector pseudotyped with an AAV9 capsid (Figure 1E). Rescue to weaning and growth correction beyond was observed, with mitigation of biochemical perturbations and a sparse pattern of MMUT mRNA expression in hepatocytes of treated Mmutp.R106C/p.R106C mice compared with mutant mice receiving genome-editing components (Figures 4E, 5E, and S4). In summary, these new MMA mouse models recapitulate the spectrum of severity of clinical and biochemical phenotypes seen in patients with the corresponding mutations, and respond to conventional AAV gene therapy, as neonates, juveniles, or adults, as has been noted in other mouse models of MMA.,,,

Figure 4

AAV8.MMUT.SV40 targeted to the Alb ATG leads to increased MMUT mRNA and MMUT protein expression and rescues Mmutp.R106C/p.R106C mice from disease lethality

(A) Donor construct corresponding to the gRNA targeting the Alb start codon (Sa08). The left homology arm contains the Alb 5′ UTR, including part of the albumin promoter. (B) Survival of untreated Mmutp.R106C/p.R106C and Mmutp.R106C/p.R106C administered 5e10 GC/pup AAV8.HLP.SaCas9.Sa08 + 2 × 1011 GC/pup AAV8.MMUT.SV40 at birth. The survival curve depicts the same control cohort of Mmutp.R106C/p.R106C and Mmutp.R106C/+ mice presented in Figure 1E. (C) Plasma methylmalonic acid concentrations for Mmutp.R106C/p.R106C treated with 5e10 GC/pup AAV8.HLP.SaCas9.Sa08 + 2 × 1011 GC/pup AAV8.MMUT.SV40 at 1, 2, 3, 6, and 9 months (terminal); concentrations from untreated Mmutp.R106C/p.R106C euthanized at birth were used as controls. (D) Weights of untreated Mmutp.R106C/p.R106C and untreated Mmut+/p.R106C compared with Mmutp.R106C/p.R106C treated with the gRNA targeting the Alb start codon and donor. Treated Mmutp.R106C/p.R106C were significantly larger than untreated Mmutp.R106C/p.R106C at both time points, and were not significantly different in size from Mmut+/p.R106C by age 1 month. (E) Liver sections from treated Mmutp.R106C/p.R106C mice euthanized at 9 months post-treatment were stained for MMUT mRNA. (F) Expression of the MMUT mRNA at 9 months in treated Mmutp.R106C/p.R106C mice was evenly distributed throughout the liver. The treated Mmut+/p.R106C control exhibited fewer cells with positive MMUT mRNA staining at the same time point. In (B), data were analyzed by log rank Mantel-Cox and Gehan-Breslow-Wilcoxon test. ∗∗p < 0.01, versus the untreated Mmutp.R106C/p.R106C group. In (C), plasma methylmalonic acid concentrations were analyzed by unpaired Student's t test as mean ± SD. Methylmalonic acid values were significantly reduced at 1 month in treated Mmutp.R106C/p.R106C mice compared with untreated neonatal Mmutp.R106C/p.R106C controls, and were further reduced for all later time points. ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, versus the untreated Mmutp.R106C/p.R106C group. In (D), weight values are expressed as mean ± SD. In (F), MMUT mRNA expression was quantified as percentage of hepatocytes stained positive and expressed as mean ± SEM. At 9 months edited Mmutp.R106C/p.R106C mice exhibited 52.89% ± 1.46% MMUT-positive cells in the liver. The treated Mmut+/p.R106C control exhibited 19.57% of hepatocytes with positive mRNA staining at this time point.

Figure 5

Nuclease-free editing with AAV8.MMUT.SV40 results in MMUT mRNA expression and improves survival, growth, and methylmalonic acid concentrations in Mmutp.R106C/p.R106C mice

(A) Donor construct corresponding to the gRNA targeting the Alb start codon (Sa08). (B) Survival of Mmutp.R106C/p.R106C mice treated at birth with either the dual 5e10 GC/pup AAV8.HLP.SaCas9.Sa08 + 2 × 1011 GC/pup AAV8.MMUT.SV40 (blue), with 2 × 1011 GC/pup of the AAV8.MMUT.SV40 donor vector only (purple), or with 1e11 GC/pup AAV9.EF1a.PI.MMUT.RBG (green) compared with untreated Mmut+/p.R106C and Mmutp.R106C/p.R106C mice. Mmutp.R106C/p.R106C mice treated with only the AAV8.MMUT.SV40 donor survived to DOL78 and did not show significantly different survival from healthy Mmut+/p.R106C controls. To compare treated Mmutp.R106C/p.R106C mice to untreated controls, which do not survive beyond DOL60, only time points for survival and weights were included for which control data were available. The survival curve depicts the same control cohort of Mmutp.R106C/p.R106C and Mmutp.R106C/+ mice presented in Figure 1E. (C) Plasma methylmalonic acid concentrations for Mmutp.R106C/p.R106C (n = 4) treated with AAV8.MMUT.SV40 donor-only (purple) at 2 months post-treatment, compared with Mmutp.R106C/p.R106C treated with Cas9 + donor at the same time point (n = 7). Untreated Mmutp.R106C/p.R106C euthanized at birth (n = 5) and untreated Mmut+/p.R106C (n = 5) were used as controls. Methylmalonic acid values in donor-only-treated mice were significantly reduced compared with untreated Mmutp.R106C/p.R106C, and were not significantly different from Mmutp.R106C/p.R106C treated with Cas9 + donor. (D) Weights of Mmutp.R106C/p.R106C treated with Cas9 + donor vectors (blue), or AAV8.MMUT.SV40 donor-only (purple) at age 2 months, compared with untreated Mmut+/p.R106C and surviving untreated Mmutp.R106C/p.R106C. Donor-only-treated Mmutp.R106C/p.R106C were significantly larger than untreated Mmutp.R106C/p.R106C at 2 months, but were smaller in size than Mmutp.R106C/p.R106C treated with Cas9 + donor. (E) To assess genome editing following treatment with the donor-only vector, liver sections from treated Mmutp.R106C/p.R106C mice euthanized at 5 months post-treatment were stained for MMUT mRNA. (F) Expression of the MMUT transgene at 5 months in Mmutp.R106C/p.R106C (n = 4), with discreet clusters of positively stained cells dispersed throughout the liver. One treated Mmut+/p.R106C control exhibited almost no cells positive for mRNA staining at 5 months. In (B), 85.71% (n = 7) Mmutp.R106C/p.R106C mice treated with only the AAV8.MMUT.SV40 donor survived to DOL78 and did not show significantly different survival from healthy Mmut+/p.R106C controls (log rank [Mantel-Cox] and Gehan-Breslow-Wilcoxon tests). In (C), plasma methylmalonic acid concentrations were analyzed by unpaired Student's t test as mean ± SD. Methylmalonic acid values in donor-only-treated mice were significantly reduced compared with untreated Mmutp.R106C/p.R106C (p = 0.0006, unpaired t test), and were not significantly different from Mmutp.R106C/p.R106C treated with Cas9 + donor (unpaired t test). ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. In (D), weight values are expressed as mean ± SD. Donor-only-treated Mmutp.R106C/p.R106C were significantly larger than untreated Mmutp.R106C/p.R106C at 2 months (p < 0.05), but were smaller in size than Mmutp.R106C/p.R106C treated with Cas9 + donor (p < 0.01). ∗∗p < 0.01, n.s., not significant. In (F), MMUT mRNA expression was quantified as percentage of hepatocytes stained positive and expressed as mean ± SEM. Expression of the MMUT transgene at 5 months in Mmutp.R106C/p.R106C (n = 4) averaged 24.05% ± 4.65%, with rosettes of positively stained cells throughout the liver. One treated Mmut+/p.R106C control exhibited almost no cells positive for mRNA staining at 5 months (0.81%).

Therapeutic editing into the albumin locus

We identified target sites for the Staphylococcus aureus Cas9 guide RNAs in the 5′ end of the albumin locus (Figure S1A), tested in vitro (Figures S1D and S1E), and then in vivo as described previously. Next-generation sequencing across the native target demonstrated on-target indel formation at the Sa37 and Sa08 targets of ∼30% and ∼20%, respectively (Figure S1B), as well as rare evidence of integration of the AAV vector at the Sa37 site (Figure S1C). Next, we delivered dual vectors, including the SaCas9/gRNA at a dose of 5e10 GC per pup, and rescue cassettes containing a codon-optimized MMUT transgene at a dose of 2e11 GC per pup, to Mmutp.R106C/p.R106C and Mmutp.G715V/p.G715V mice and littermate controls treated as neonates. Treated MMA mice and wild-type heterozygote controls were aged to 9 months prior to sacrifice, and the clinical (weight and survival) and biochemical (plasma methylmalonic acid concentration) phenotypes were assessed. Cohorts of treated mutant mice were also allowed to age beyond the 9-month time point to examine the long-term safety profile after genome editing.

Correction with donor constructs targeting albumin intron 1 (Sa37)

Donor cassettes were designed using a human codon-optimized human MMUT cDNA in cis with either a 2A peptide or internal ribosomal entry site (IRES) sequence (Figures 2A and 3A), with Alb homology arms of ∼400 bp on either side of the insert.

Figure 2

AAV8.2A.MMUT.SV40 targeted to Alb intron 1 leads to MMUT mRNA expression and reduced methylmalonic acid concentrations in Mmutp.G715V/p.G715V

(A) Schematic of the donor construct corresponding to the Sa37 gRNA, targeting intron 1 of Alb. 2A-peptide precedes a full-length, human codon-optimized methylmalonyl-CoA mutase cDNA. (B) Plasma methylmalonic acid concentrations for Mmutp.G715V/p.G715V treated with 5e10 GC/pup AAV8.HLP.SaCas9.Sa37 + 2 × 1011 GC/pup AAV8.2A.MMUT.SV40 at birth, compared with untreated Mmutp.G715V/p.G715V at age 1, 2, 3, 6, and 9 months (terminal). At 1 month, no reduction in plasma methylmalonic acid was observed in the treated mice. At 2 months, a significant difference was observed in methylmalonic acid in untreated versus treated mice, which persisted until sacrifice at 9 months. (C) Weights of Mmutp.G715V/p.G715V treated with AAV8.HLP.SauCas9.Sa37 + AAV8.2A.MMUT.SV40 compared with untreated Mmutp.G715V/p.G715V and untreated Mmut+/p.G715V mice at 1 month and 45 days, and 2, 3, and 6 months. At 1 month, weights of treated Mmutp.G715V/p.G715V were significantly larger than untreated Mmutp.G715V/p.G715V, but by age 2 months were not significantly different than untreated Mmutp.G715V/p.G715V mice, and were significantly smaller in size than heterozygous Mmut+/p.G715V controls at all subsequent time points. The growth data shown from these experiments include time points for which we had relevant untreated Mmutp.G715V/p.G715V and Mmut+/p.G715V control data. (D) Liver sections from treated Mmutp.G715V/p.G715V mice and one treated Mmut+/p.G715V control were stained for MMUT mRNA using a probe specific to the transgene. (E) Expression of MMUT mRNA at 9 months was in the liver of edited Mmutp.G715V/p.G715V mice. Positively stained cells were evenly distributed throughout the liver. A Mmut+/p.G715V littermate control treated with AAV8.HLP.SauCas9.Sa37 + AAV8.2A.MMUT.SV40 exhibited only a small number of cells with positive mRNA staining at 9 months. In (B), plasma methylmalonic acid concentrations were analyzed by unpaired Student's t test as mean ± SD. Methylmalonic acid values at 1 month in treated Mmutp.G715V/p.G715V mice (848.3 ± 99.38 μM) were comparable with untreated Mmutp.G715V/p.G715V (820.3 ± 92.6 μM), but were significantly reduced in treated mice for later time points. ∗∗p < 0.01, ∗∗∗p < 0.001, versus the untreated Mmutp.G715V/p.G715V group. In (C), weight values are expressed as mean ± SD and analyzed by unpaired Student's t test. At 1 month, treated Mmutp.G715V/p.G715V were significantly larger than untreated Mmutp.G715V/p.G715V (p = 0.0001). In (E), MMUT mRNA expression was quantified as percentage of hepatocytes stained positive and expressed as mean ± SEM. At 9 months edited Mmutp.G715V/p.G715V mice exhibited 41.19% ± 3.56% MMUT-positive cells in the liver. The treated Mmut+/p.G715V control exhibited 8.40% of hepatocytes with positive mRNA staining at this time point.

Figure 3

Targeting an AAV8.IRES.MMUT cassette into Alb intron 1 results in MMUT mRNA expression and improves clinical outcomes in Mmutp.G715V/p.G715V

(A) Schematic of the donor construct corresponding to the Sa37 gRNA targeting intron 1 of Alb. (B) Plasma methylmalonic acid concentrations for Mmutp.G715V/p.G715V administered AAV8.HLP.SaCas9.Sa37 + AAV8.IRES.MMUT.SV40 at birth, compared with untreated Mmutp.G715V/p.G715V controls at 1, 2, 3, 6, and 9 months (terminal). At 1 month, no reduction in plasma methylmalonic acid was observed in treated versus untreated mice. At 2 months, a significant difference was observed in methylmalonic acid in untreated versus treated mice which persisted. (C) Weights for Mmutp.G715V/p.G715V treated with 5e10 GC/pup AAV8.HLP.SauCas9.Sa37 + 2 × 1011 GC/pup AAV8.IRES.MMUT.SV40 compared with untreated Mmutp.G715V/p.G715V and untreated Mmut+/p.G715V mice at 1 month and 45 days, and 2, 3, and 6 months. Treated Mmutp.G715V/p.G715V were comparable in size with heterozygous Mmut+/p.G715V controls for all time points. The growth data shown from these experiments include time points for which we had relevant untreated Mmutp.G715V/p.G715V and Mmut+/p.G715V control data. (D) Liver sections from treated Mmutp.G715V/p.G715V were stained for MMUT mRNA using a transgene-specific probe. (E) Expression of MMUT mRNA in treated Mmutp.G715V/p.G715V at 9 months. Treated Mmutp.G715V/p.G715V exhibited an even distribution of positively stained cells throughout the liver. One Mmut+/p.G715V littermate control treated with 5e10 GC/pup AAV8.HLP.SauCas9.Sa37 + 2 × 1011 GC/pup AAV8.IRES.MMUT.SV40 exhibited significantly fewer positively stained hepatocytes at 9 months. In (B), plasma methylmalonic acid concentrations were analyzed by unpaired Student's t test as mean ± SD. Methylmalonic acid values at 1 month in treated Mmutp.G715V/p.G715V mice (690.5 ± 130.9 μM) were comparable with untreated Mmutp.G715V/p.G715V (820.3 ± 92.6 μM), but were significantly reduced in treated mice for later time points. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, versus the untreated Mmutp.G715V/p.G715V group. In (C), weight values are expressed as mean ± SD and analyzed by unpaired Student's t test. In (E), MMUT mRNA expression was quantified as percentage of hepatocytes stained positive and expressed as mean ± SEM. At 9 months, edited Mmutp.G715V/p.G715V mice exhibited 41.85% ± 4.90% MMUT-positive cells in the liver. The treated Mmut+/p.G715V control exhibited 9.69% of hepatocytes with positive mRNA staining at this time point.

Plasma methylmalonic acid concentrations and growth in treated Mmutp.G715V/p.G715V remained elevated at age 1 month, but became significantly reduced beginning at the 2-month time point, and continued to decrease during subsequent time points (Figure 2B). Some of the treated mutant mice exhibited elevated levels of plasma metabolites during the terminal bleed at 9 months due to the stress of euthanasia, an effect observed for all genome-editing strategies (Figures 2B, 3B, and 4C). One month after editing, treated Mmutp.G715V/p.G715V mice were significantly larger in size than untreated Mmutp.G715V/p.G715V controls, and similar in size to heterozygous littermates (Figure 2C). However, as the mice aged, they did not achieve the weight of the heterozygous-treated control littermates.

To validate on-target integration of the AAV8.2A.MMUT.SV40 donor, we developed a PCR assay of the 5′ integration junction from liver genomic DNA of treated Mmutp.G715V/p.G715V mice and untreated controls. Successful amplification indicated that integration had occurred in the treated, but not the untreated mice (Figure S2D). RNA in situ hybridization assay using a specific probe to detect the human codon-optimized MMUT mRNA showed significantly less staining in treated Mmut+/p.G715V livers compared with treated Mmutp.G715V/p.G715V mice (Figures 2D and 2E), consistent with previous observations suggesting that corrected hepatocytes display a growth advantage in the environment of an MMA liver. Similar studies with an AAV8.IRES.MMUT.SV40 donor construct were performed, and likewise documented a biochemical and growth response seen only in the treated Mmutp.G715V/p.G715V mice (Figures 3 and S2E). In summary, Cas9-mediated insertion of either an AAV8.2A.MMUT.SV40 or an AAV8.IRES.MMUT.SV40 donor in the first intron of albumin resulted in targeted integration of the rescue cassette, leading to a sustained reduction in methylmalonic acid concentrations, and an improvement in growth in treated Mmutp.G715V/p.G715V, with a concomitant expansion of the edited hepatocytes.

MMUT expression after editing at the start codon of albumin

We next assessed an alternate strategy targeting integration of a MMUT cDNA to the albumin start codon (gRNA designated Sa08) in the more severe MMA mouse model, Mmutp.R106C/p.R106C (Figure 4A and S2F). Because the 5′ homology arm of the donor contains part of the albumin promoter, immediate and substantial levels of expression from episomes of the AAV donor cassette were expected prior to MMUT integration. We were therefore able to assess the potential for this genome-editing strategy to rescue mice with more severe MMA phenotypes requiring immediate MMUT expression.

Because Mmutp.R106C/p.R106C exhibit disease-associated mortality early in life, we first assessed efficacy of editing at the start codon by looking for improvement in survival (Figure 4B). Strikingly, 75% of mutant Mmutp.R106C/p.R106C treated at birth survived to the experimental endpoint of 9 months. Plasma metabolites were significantly reduced in treated Mmutp.R106C/p.R106C compared with control values (Figure 4C), and treated Mmutp.R106C/p.R106C were significantly larger than untreated Mmutp.R106C/p.R106C controls at age 1 month (Figure 4D), reaching the size of untreated Mmut+/p.R106C littermates, an effect that was sustained for the duration of the study.

Untreated Mmut and Mmutp.G715V/p.G715V mice produce immunoreactive Mmut protein in their livers (Figures S3A and S3B), yet treated Mmut exhibited a relative increase in protein expression for 9 months after treatment (Figure S3C). RNA in situ hybridization results indicated that ∼50% of hepatocytes were expressing MMUT mRNA in treated Mmutp.R106C/p.R106C at 9 months post-treatment, greatly increased compared with the treated Mmut+/p.R106C littermate (Figures 4E and 4F), and even ∼10% higher than the corresponding values seen in Mmutp.G715V/p.G715V mice treated with the AAV8.SaCas9.Sa37 and either the AAV8.2A.MMUT.SV40 or AAV8.IRES.MMUT.SV40 constructs. On-target integration of the MMUT donor was supported by integration junction amplification noted only in the Sa08-treated mice (Figure S2F), with MMUT genome copy numbers higher in treated Mmutp.R106C/p.R106C compared with mice treated with AAV8.SaCas9.Sa37 and either the AAV8.2A.MMUT.SV40 or AAV8.IRES.MMUT.SV40 constructs at 9 months post-treatment (Figures S2A–S2C).

Nuclease-free genome editing at the albumin start codon in Mmutp.R106C/P.R106C

To examine whether the significant therapeutic effects of genome editing at the albumin ATG were due to episomal donor vector expression, we treated a cohort of Mmutp.R106C/p.R106C mice at birth with the AAV8.MMUT.SV40 donor-only, at the dose used in dual-vector studies (2e11 GC/pup), but without the Cas9 vector (Figure 5A). We observed significant rescue from disease-related neonatal mortality in Mmutp.R106C/p.R106C treated with AAV8.MMUT.SV40 donor-only compared with untreated Mmutp.R106C/p.R106C controls, with 85% of treated mice surviving normally compared with Mmut+/p.R106C controls (p < 0.001 compared with untreated Mmutp.R106C/p.R106C controls, log rank Mantel-Cox test) (Figure 5B). RNA in situ hybridization results indicated that ∼24% of hepatocytes were expressing the MMUT mRNA in treated Mmutp.R106C/p.R106C at 5 months post-treatment, compared with <1% in a treated Mmut+/p.R106C at 5 months (Figures 5E and 5F). Corresponding with the improvement in survival, plasma methlymalonic acid concentrations were significantly reduced at age 2 months in the donor-only-treated Mmutp.R106C/p.R106C and were not significantly different than values from Mmutp.R106C/p.R106C treated with the dual Cas9 and donor vectors, at the same time point (Figure 5C). Donor-only-treated Mmutp.R106C/p.R106C were significantly larger in size than untreated Mmutp.R106C/p.R106C controls at 2 months, but were smaller in size than Mmutp.R106C/p.R106C treated with the dual-vector strategy (Figure 5D).

Discussion

To expand upon existing mouse models of MMA, we used CRISPR-Cas9 genome editing to knock in missense mutations observed in mut0 and mut− MMA patients. These new models recapitulate key disease parameters identified in existing knockout and transgenic MMA mouse models, such as growth impairment, massively elevated levels of disease metabolites, and substantial lethality in the mut0 missense model.,,,25, 26, 27, 28, 29, 30 While most previous mouse models harbor selection cassettes or synthetic elements, the new alleles presented here are transgene and deletion free. In addition to recapitulating features of the patients with corresponding mutations, these mice respond to gene therapy using conventional AAV vectors, showing improvements in survival and growth, and reduction of metabolites, similar to what we have observed from studies performed in knockout and transgenic MMA mouse models.,,,

Due to the transient therapeutic effects associated with episomal transgene expression when AAV is delivered in the neonatal period, and the potential for genotoxicity when AAV vector integration occurs randomly, we relied upon editing into albumin, which has been employed for several metabolic disorders with variable success. Targeting the 5′ end of albumin using zinc finger nucleases (ZFNs) has entered human trials for Hunter syndrome, and efficiency of the ZFN strategy has been enhanced using CRISPR-Cas9 in treatment of mucopolysaccharidosis type I. Editing the 3′ end of albumin preserves albumin expression, but current applications rely upon the fusion of albumin to the inserted transgene via a 2A peptide, which adds foreign 2A amino acids to the C terminus of albumin, and a proline to the gene of interest., Some proteins will tolerate such alteration, but whether the resultant Alb-2A produced by editing at the 3′ end of albumin might engender an antibody formation or elicit a T cell response remains unknown, and potentially, a limiting toxicology consideration. Furthermore, the low efficiency of homologous recombination with nuclease-free approaches, and very high AAV doses required for efficacy, may further present a hurdle as has been demonstrated by the need for Cas9 to improve the efficacy of 3′ transgene insertion into the albumin locus in treatment of Crigler-Najjar syndrome mouse models., For these reasons, we selected the CRISPR-Cas9 genome-editing system to enable efficient transgene integration near the site previously targeted by ZFN editing, and then further focused our efforts at the albumin start codon. The programmability of the CRISPR-Cas9 system allowed the design and testing of several guides, with two then selected for further in vivo application.

After in vitro and in vivo characterization of the lead guides, we designed several DNA donor cassettes to edit the albumin locus that utilized either a 2A peptide or an IRES sequence preceding the MMUT coding sequence in the donor construct, inserted into intron 1 of albumin. While the correction with both cassettes was substantial, and durable, the biochemical parameters in the treated MMA mice showed less correction than seen with conventional AAVs (Figure 1), suggesting that the editing approach might be further improved. We then targeted a MMUT cDNA to the start codon of albumin, encouraged by the lack of off-target effects seen with the Sa08 guide (Figure S1F) and the potential for immediate transgene expression before integration due to the presence of the albumin promoter in the 5′ homology arm of the donor, and used a more severely affected mouse model to assess therapeutic benefit. Indeed, the rescue from lethality and reduction of metabolites was comparable between nuclease plus donor- and donor-only-treated Mmutp.R106C/p.R106C mice. Furthermore, the morphologic pattern of MMUT mRNA expression in livers harvested from donor-only-treated mice revealed clusters of cells that further yielded a PCR product that could only be generated after successful homologous recombination, is consistent with the clonal expansion after homologous recombination, and similar to what has been seen in other studies,,,35, 36, 37, 38 suggesting that AAV-mediated, nuclease-free homologous recombination was responsible for the long-term therapeutic effects. While we were unable to measure homology-directed repair (HDR) versus non-homologous end joining (NHEJ) end capture as the mechanisms underlying correction, the fact that our studies were conducted in neonatal mice, where HDR is permissive and favored as a repair pathway, suggests that the functional repair of hepatocytes is most likely the result of HDR with minimal NHEJ end capture. In aggregate, our results demonstrate that genome editing at the albumin start codon has the potential to correct the disease phenotypes with immediacy due to the promoter contained in the AAV donor 5′ homology arm, followed by subsequent durability from integration into albumin. Patients with both severe and intermediate MMUT deficiencies could be treated using an integrating, nuclease-free approach, with a perhaps a preference toward treating more severe mut0 patients very early in life given the potential for substantial subsequent expansion in vivo, which is supported by the striking efficacy in the Mmut R106C mouse model.

Although promising, our studies possess limitations. Off-target and inactivating cleavage by the Cas9 nuclease with any guide targeting albumin remains possible. Although the physiological consequences of extreme analbuminemia are well tolerated in some species, even humans, cleavage without recombination of the donor construct could lead to loss of expression of Alb, but would require extensive, biallelic inactivation via editing to fully disable a targeted cell's ability to produce albumin. Future studies should aim to improve the design of the donor cassette to restore albumin expression after successful targeting. In addition, alternative means to deliver the Cas9, such as RNA lipid nanoparticles,42, 43, 44 to avoid immune complications and to avoid potential toxicity due to the high doses of AAV required, seems prudent. Nevertheless, our proof-of-concept studies demonstrate that nuclease-enhanced, as well as nuclease-free genome editing at the 5′ end of the albumin locus, in multiple positions, including the start codon, are promising treatments for MMA, and might be generally extended to other related metabolic disorders where hepatic correction is needed.

Materials and methods

Generation of MMA mouse models using genome editing

Benchling (http://www.benchling.com) was used to identify protospacer-adjacent motif (PAM) sequences (NGG) from the Streptococcus pyogenes (SpyCas9) within ten nucleotides of the desired site for targeted mutagenesis in Mmut. gRNAs were synthesized as custom synthetic single-guide RNAs (sgRNAs) (Edit-R predesigned synthetic sgRNA, Horizon Discovery):

Mmut p.R106C (5′-TACTAAAGCCTGCATACTGA -3′)

Mmut p.G715V(5′-TATGAATTTCTGTATGAAGT -3′)

Single-stranded donor oligo sequences containing the desired mutation were co-injected with the Cas9 and sgRNA. For the Mmut p.R106C mutation, a repair template was designed with 80 bp homology arms on either side of the mutation, which was located in the PAM. For the Mmut p.G715V mutation, an asymmetric donor oligo was designed using DeskGen software, with 92 bp left HA and 43 bp right HA. A silent C > A mutation was introduced in the PAM to prevent Cas9 cleavage of the donor construct.

Injected zygotes were cultured at 37°C under 5% CO2 until blastocyst stage then transferred into the uterus of pseudopregnant females. Murine experiments were approved and performed according to the regulations and standards of the National Human Genome Research Institute (NHGRI) (Bethesda, MD) Animal Care and Use Committee.

Founder genotyping

DNA was extracted from tail clips using the DNeasy Blood & Tissue Kit (QIAGEN) for PCR amplification. ddPCR probes (Bio-Rad) were designed to identify the Mmut p.R106C and Mmut p.G715V mutation sequences. DNA samples from founders were screened by fPCR (for gene knockout) and ddPCR (for HDR of mutation sequences), and crossed with wild-type mice to generate heterozygous carriers of the mutation, which were further bred.

Fluorescent PCR primers

p.R106C_Fwd_fl (5′-TGTAAAACGACGGCCAGTTTCCAGGAGTGAAGCCATTC-3′), p.R106C_Rev_fl (5′- GTGTCTTGCATCCACTTGTTTCACAGC-3′), p.G715V_Fwd_fl (5′- TGTAAAACGACGGCCAGTCAGCTCCACATACAGTGTTCC-3′), and p.G715V_Rev_fl (5′-GTGTCTTTCATCAAGCACTTGGACAGCA-3′).

Identification of SaCas9 guide sequences and in vitro/in vivo NHEJ testing

To identify sites amenable to genome editing in Alb, the first exon and first intron of Alb were searched for PAM sequences (NNGRRT) of S. aureus Cas9 (SauCas9). The in vitro activity of the output 20-nucleotide protospacer guide RNA (gRNA) sequences was measured by transfection of plasmids into mouse carcinoma cell lines. Multiple cell lines, including NIH 3T3 (murine embryonic fibroblastoma), Neuro2A (neuroblastoma), and hepatoma (Hepa1-6) were tested to determine which could provide optimal delivery efficiency by standard transfection and nucleofection methods. Lipofectamine 2000 for standard lipofection was used. For Neuro2A transfections, a ratio of 2:1 (μL of lipofectamine 2000 to μg of DNA) was used, and for Hepa1-6 transfections, a ratio of 3:1 was used. Cells were seeded in 24-well plates 24 h prior to transfection; 1 μg of plasmid DNA was transfected per well.

For the T7E1 assay, primers were designed to amplify a 500- to 800-bp region centered on the gRNA target. High-fidelity Accuprime Taq from Life Technologies was used for all PCR experiments. PCR was run for 40 cycles to ensure sufficient product. The PCR reaction was purified and normalized to 12 ng/μL; 16.2 μL (∼200 ng) of the 12 ng/μL PCR product was mixed with 1.8 μL of 10X NEB buffer 2 for a final volume of 18 μL. DNA was melted and re-annealed by placing in a thermocycler for 10 min at 95°C, then decreasing at 0.1°C/s down to 25°C. Two microliters of an enzyme master mix (0.5 μL T7EI, 0.2 μL NEB buffer 2, and 1.3 μL water) was added to each reaction and immediately placed in a thermocycler at 37°C for 60 min. After 60 min, the reactions were immediately removed from the thermocycler and quenched by adding EDTA to a final concentration of 45 mM. A quenching loading dye composed of 0.5 M EDTA and 6X DNA loading dye at a ratio of 1:2 EDTA:dye was used. The full content of the samples was immediately loaded into a 2% agarose gel and run until the bands were well separated.

Off-target analysis and deep sequencing

The bioinformatic tool CRISPR off-target sites with mismatches, insertions, and/or deletions (COSMID) was used to search the Mus musculus (Mm10 build) genome for chromosomal locations similar to the guide RNA with a maximum of three mismatches and an NNGRRT PAM. Mouse liver genomic DNA was extracted using the QIAGEN Blood & Tissue Kit, and target sequences were amplified using site-specific primers containing common adaptor sequences. A second round of PCR was used to add custom Illumina sample indexes. Both rounds of PCR were performed using five touchdown cycles with the annealing temperatures decreasing each round by 1°C from 65°C to 60°C. Target amplicons were purified using magnetic beads, pooled in equimolar amounts, and sequenced using the Illumina MiSeq platform. Alignment of sequence reads to reference sequences was performed as described previously, and indels overlapping a ±2-base pair window around the cut site were quantified.

SaCas9 expression cassettes

The AAV.HLP.SauCas9.U6.gRNA construct used for in vivo experiments was described previously. Oligos containing the gRNA sequence and restriction enzyme overhangs were resuspended in distilled water to a concentration of 10 μM. Oligos (0.8 μL) were phosphorylated in a 25-μL reaction, then mixed before annealing for 1 h (50 μL). One microliter of annealed oligos plus 12.5 ng of vector were ligated in a 10-μL ligation reaction (25 μL E. coli, 175 μL SOC, and all 200 μL were plated). The gRNA oligo sequences for sites targeting intron 1 of albumin (Sa37) and the ATG of albumin (Sa08) are as follows (the 5′ base was replaced by a “g” for U6 promoter expression):

Sa_08_T:CACCgCTAGCCTCTGGCAAAATGAA

Sa_08_B:AAACTTCATTTTGCCAGAGGCTAGc

Sa_37_T:CACCgTGCACAGATATAAACACTTA

Sa_37_B:AAACTAAGTGTTTATATCTGTGCAc

Donor construct design

Donor constructs were designed for Sa37 and Sa08 gRNA targets. A full-length, human codon-optimized methylmalonyl-CoA mutase cDNA (MMUT) was preceded by a 2A peptide or an IRES sequence. The 2A peptide sequence was preceded by the two nucleotides coding for histidine, the 27th amino acid in Alb, to maintain the reading frame. A polypyrimidine tract and splice acceptor were additionally added to the 5′ end of the insert. A synthetic SV40 poly(A) tail was added to the 3′ end of MMUT. Homology arms were designed to be at least 400 bp when possible; the albumin intron 1 target (Sa37) 3′ homology arm was truncated due to a complex repeat region in the mouse genome in intron 2 of Alb. For the donor construct corresponding to the gRNA that targets the Alb transcriptional start site (Sa08), the MMUT cDNA was full-length except for the first two nucleotides (the AT of the ATG of Alb), which were contained in the 5′ homology arm containing the Alb 5′ UTR. The final cassettes were sequenced using NGS to validate before use.

Recombinant AAV vector production

A dual-vector system was adopted to deliver genome-editing components using AAV serotype 8 vectors. One vector (AAV8.HLP.SauCas9) co-expressed the SauCas9 gene from a human liver-specific promoter and the sgRNA scaffold sequence expressed from the U6 promoter; the second vector (AAV8.donor) contained a human codon-optimized MMUT cDNA. Pups were injected on DOL0 or DOL1 with a ratio of 1:4 of the AAV8.HLP.SauCas9 (5e10 GC/g) to the AAV8.donor (2e11 GC/g). AAV8 vectors were produced using standard methods at either Baylor College of Medicine (Dr. William Lagor's lab) or by the Vector Core at University of Massachusetts Medical School (Dr. Guangping Gao's lab).

Another set of AAV vectors, AAV8.EF1a.PI.MMUT.RBG or AAV9.EF1a.PI.MMUT.RBG were prepared by University of Pennsylvania's Vector Core as described previously.,

Methylmalonic acid measurement

Plasma was isolated from blood collected by orbital bleeding. Samples were stored at −80°C until analysis by gas chromatography-mass spectrometry to measure methylmalonic acid concentration, as described previously.

Animal studies

Animal work was performed in accordance with the guidelines for animal care at NIH. Both mouse models were on a mixed C57BL/6 × 129SV/Ev × FvBN genetic background. Experimental groups contained equal numbers of both genders. Treated mice received AAV via intrahepatic injection on day 0 or day 1. Mice were bled via retro-orbital sinus plexus sampling using a sterile glass capillary tube and weighed monthly.

Western blot analysis

Samples were prepared and run according to standard protocols. The following antibodies were used for the detection of MMUT (Abcam, ab134956), β-actin (Abcam, ab8227) in conjunction with the following secondary antibodies (LI-COR, 925-32211) and (LI-COR, 926-68072).

Hepatic MMUT RNA in situ hybridization

RNAscope probes were designed against the human codon-optimized MMUT RNA by ACDBio using their proprietary technology and do not cross-react with other transcripts in the murine transcriptome, including Mmut. Livers were fixed in 4% PFA and processed into paraffin blocks. Five-micron sections were cut and stained with the RNAScope 2.5 HD Assay-Brown (ACDBio 322300) following the manufacturer's instructions. Slide images were captured using the Zeiss AxioScan Z1 slide scanner and analyzed using Image-Pro premier 3D version 9.3 by Media Cybernetics.

Vector genome quantitation

Genomic DNA from liver samples was extracted using DNeasy Blood & Tissue Kit (QIAGEN; 69506) following the manufacturer's procedures. ddPCR was performed according to the manufacturer's recommendations for the Bio-Rad QX200 AutoDG ddPCR system using 10 or 20 ng of DNA as input and the following probes: Bio-Rad ddPCR CNV assay Gapdh (cat no. 10042961, dMumCNS300520369) and MMUT (cat no. 10042958 dCNS513322846).

Albumin integration assay

Genomic DNA from liver samples was extracted using DNeasy Blood & Tissue Kit (QIAGEN; 69504) following the manufacturer's procedures. A PCR product representing the genomic integration between the Alb locus and 5′ end of the MMUT donor construct indicated on-target integration. gDNA (200 ng) from various mouse liver samples was used as template for the DNA integration assay.

Statistical analyses

Prism 8 by GraphPad was used to analyze all data. Results are expressed as mean ± SD/SEM. Values of p < 0.05 were considered statistically significant. Where animal injections were unsuccessful, as determined by vector genome copies in liver, data were excluded. Depending on the experimental design, an unpaired t test, or log rank (Mantel-Cox) test was used as indicated in the legends to the figures. The statistical analysis performed for each dataset is indicated in the figure legends. For all figures, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, except where different symbols were used.