Hemophilia B
Individuals most severely affected by hemophilia B exhibit circulating levels of blood clotting FIX below 1% of normal, leading to sporadic hemorrhaging from the time of birth [
32,
33]. Restoration of FIX to more than 1% of normal, however, converts the disease to a more mild form, making hemophilia a powerful model to assay correction using genome editing approaches. A foundational study for in vivo genome editing used zinc-finger nucleases (ZFNs) to target the first intron of the
FIX gene in neonatal mice, with the goal of inducing sufficient levels of HDR into the
FIX locus [
31]. A humanized murine model of hemophilia B was generated by expressing a mutant human FIX (hFIX) knocked-into the Rosa26 locus on the background of a homozygous
FIX deletion mouse. ZFNs were delivered by a single AAV serotype 8 vector designed to target the liver, the major site of synthesis for FIX. A second AAV8 delivered a promoterless cDNA encoding exons 2–8 preceded by a splice acceptor site; this rescue cassette was engineered to recombine into the first intron of the
FIX gene after ZFN cleavage. Upon perfect recombination, splicing of the
FIX minigene behind the endogenous
FIX exon 1 led to expression of functionally active FIX. In mice treated as neonates, HDR was measured at 1–3%, and FIX protein expression measured at 3–7%. Although indels produced by the ZFNs in the intron did not interfere with the creation of functional protein after HDR, the frequency of off-target cleavage by the ZFN pair, measured at up to 4%, and the lower overall efficiency of correction required that this approach be further optimized for clinical translation. Nevertheless, it was the first study to prove that genome editing could be used in vivo to correct a mammalian model of a disease.
Two studies sought to improve upon the design by directing the therapeutic transgene into albumin, an ideal safe-harbor locus due to its high transcriptional activity, and corresponding increased efficiency of integration by homologous recombination. Furthermore, a transgene correctly inserted into this locus would be expressed from the endogenous and liver-specific Alb promoter. Both studies aimed to introduce promoterless, partial cDNAs of hFIX into the albumin locus in mouse models of hemophilia B using the liver tropic AAV serotype 8 as the delivery vector, but different regions of the Alb gene were targeted, and different editing strategies employed.
In an extension of the genome editing study that aimed to correct hemophilia B at the disease locus, the first intron of the albumin gene was selected for insertion of a partial h
FIX cDNA [
34]. The ZFN pair targeting intron 1 was delivered on a single vector, with a second vector carrying the
hFIX cDNA rescue cassette of exons 2 to 8. Adult C57Bl6 mice treated with 1 × 10
11vg AAV8-ZFNs or 5 × 10
11vg AAV8-h
FIX-donor exhibited high circulating hFIX levels despite a low level of integration; the m
Alb-h
FIX mRNA corresponded to 0.5% of total wild-type m
Alb mRNA transcripts. Although HDR commonly occurs during the S and G2 phases of the cell cycle, when cells are dividing, these results suggest that insertion of a donor sequence via NHEJ and ligation is an effective correction method in the adult liver, when cells are quiescent and more prone to DNA repair by NHEJ. Additionally, only a small number of hepatocytes need to be modified at the albumin locus in order to achieve clinically relevant levels of hFIX.
In the second study, the disruption to albumin expression was minimized by mediating HR of a corrected partial cDNA donor flanked by arms of homology without the use of site-specific endonucleases [
35]. Additionally, the h
FIX cDNA was expressed as a 2A-fusion at the end of the
Alb reading frame, 5’ to the
Alb stop codon. Thus, both albumin and FIX were co-expressed from a single mRNA transcript and processed into two separate proteins, both containing a signal peptide for secretion. Because neither the
FIX nor the 2A peptide contained a methionine residue, off-target integration of the construct did not lead to hFIX expression. Targeted
Alb alleles versus wild-type
Alb alleles were measured at 0.5% on average for mice injected as neonates or adults at the highest dose and both neonatal and adult mice treated the rAAV8-donor showed hFIX protein levels in plasma at 7–20% of normal, with clotting activity similar to wild-type mice 2 weeks post-treatment. Targeted integration without the use of site-specific endonucleases may therefore mitigate off-target effects and the possibility of genotoxicity, but the accompanying increase in vector dose required (by up to an order of magnitude in adult mice) may pose concerns related to immune activation and manufacturing. Whether mice treated as neonates develop HCC by any of the aforementioned vectors remains to be determined, but will be critical to examine. Taken as a whole, the genome editing studies using hemophilia models demonstrate the potential for genome editing to be applied to other IEM for which a low level of corrected gene expression can lead to amelioration of the disease phenotype, such as lysosomal storage disorders.
Hereditary tyrosinemia type I (HT-I)
HT-I is caused by mutation of fumarylacetoacetate hydrolase (FAH), the last enzyme in the tyrosine catabolic pathway. FAH deficiency causes accumulation of fumarylacetoacetate in hepatocytes and results in severe liver damage. Affected individuals fail to gain weight and grow at the expected rate, which is recapitulated by the
Fah
–/– mouse model [
29,
36]. Current therapies for HT-I include nitisinone (NTBC), which inhibits hydroxyphenylpyruvate dioxygenase (HPD), the enzyme catalyzing the second step of tyrosine catabolism. Removal of NTBC from the diet of mutant mice results in the expansion of repaired hepatocytes that repopulate the liver, a unique characteristic of this disease model which makes it useful for testing the efficacy of novel genome editing therapies [
36], because it allows selection of cells that have been corrected in vivo. In the first study using CRISPR to correct a disease model in vivo [
37], CRISPR components were systemically delivered via hydrodynamic tail vein injection in a mouse model of hereditary tyrosinemia type I. The Cas9 enzyme from
S. pyogenes and one of three gRNAs (FAH1, FAH2 or FAH3) were co-injected with a short single-stranded DNA oligo donor template into adult
Fah
–/–
mice. Animals receiving the FAH2 guide targeting in the intron just downstream of the point mutation did not lose weight after NTBC-containing water was withdrawn, while untreated
Fah
–/–
mice experienced rapid weight loss after NTBC withdrawal, and then death. Due to the selective advantage conferred after successful editing at the
Fah locus, a correction of 0.4% was sufficient to rescue the treated mutant mice from lethality; 30 days after NTBC withdrawal, 33.5% of hepatocytes were Fah
+ in treated mice. Follow-up studies aimed to improve the efficacy and mode of delivery for clinical implementation. By coupling transient non-viral delivery methods with pharmacological selection, a follow-up study was able to achieve a more than ten-fold increase in corrected hepatocytes [
38].
In a complementary approach using AAV gene delivery, HDR-mediated rescue of a point mutation in the
Fah gene was accomplished using an AAV8 vector to deliver the FAH2 guide RNA and a HDR template (AAV-HDR) [
38]. The
SpCas9 mRNA was packaged into a lipid nanoparticle (nano.Cas9) for short-term expression, reducing the chance of genotoxic off-target effects. Fourteen days post-treatment, a greater than 6% correction of Fah
+ hepatocytes was detected in treated
Fah
–/–
mice. Highly efficient delivery vectors led to a significant increase in the initial correction of Fah over the initial treatment, and completely restored the liver by 1 month.
An alternative approach has leveraged the potent gene inactivating capability of the CRISPR system to target an enzyme upstream of FAH in the tyrosine catabolic pathway, HPD, as an alternative method of correcting the
Fah
–/–
mouse model. Metabolic reprogramming by knockout of the HPD enzyme would convert HT-I to a milder disorder, HT-III. As predicted,
Fah
–/–
; Hpd
–/–
mice show reduced risk of HCC compared to
Fah
–/–
mice treated with NTBC; thus, bypassing this pathway via deletion of the
HPD gene could be an effective treatment for HT-I patients [
39]. To test this premise, CRISPR components designed to inactivate HPD were delivered to adult
Fah
–/–
mice [
40]. Assessment of editing efficiency by staining cells for Hpd expression showed that bi-allelic knockout at 1 week post-treatment was 8%. At 4 weeks, the efficiency increased to 68% due to the positive selection of
Hpd
–/–
hepatocytes, and treated mice phenotypically resembled wild-type mice at 25–27 days post withdrawal of NTBC. In some mice, healthy hepatocytes completely replaced diseased cells by 8 weeks post-treatment. As with direct correction at the
Fah locus, the success of this strategy was contingent upon the positive selection inherent to the HT-I model, and provided the incentive for testing artificial methods of induced selection to enrich for desired editing events in modified hepatocytes.
Improving the efficacy of genome editing is a pervasive problem independent of targeting strategy and delivery method. Although the previous examples are encouraging, they are applicable only in the rare setting in which healthy hepatocytes have a selective growth advantage, such as in HT-I. For most liver diseases, this is not the case, and a method of selecting for cells would allow the amplification of less efficient gene correction. To this end, a novel approach using a small molecule and gene editing has been used to induce a transient state of superior fitness in corrected hepatocytes. Critical to the success of this study has been the use of a transition state analogue and potent inhibitor of FAH, CEPHOBA, which can induce a transient HT-I [
41], and the previous observation that shRNA-mediated knockdown of an alternate enzyme in the pathway, Hpd, can positively select hepatocytes in the HT-I mouse model [
42]. To recapitulate positive selection in wild-type mice, an rAAV8 cassette was designed with an
Hpd shRNA (sh
Hpd) embedded in a miRNA so that it would be expressed only after perfect recombination into the albumin gene. Co-delivery of the hFIX cDNA fused to a 2A peptide on the same vector allowed for facile assaying of HDR events at the
Alb locus
. The dual expression construct was administered to neonatal C57Bl6 mice, and at 4 weeks of age, CEPHOBA was injected daily for 4 weeks to enrich for gene targeted cells. In two of three mice, hFIX levels steadily increased and were stable in the therapeutic range but declined in a third. With a second round of CEPHOBA, however, hFIX was expressed in the therapeutic range for all three mice. shRNA-mediated knockdown of
Hpd thus protected hepatocytes against the effects of the FAH inhibitor CEPHOBA. Although efficiency of HR with this method was less than 1%, after selection, transgene-expressing cells constituted 50% of the liver. Any transgene delivered in cis to the protective shRNA could be co-selected, making the “GeneRide” construct a malleable instrument applicable to other metabolic diseases, with fewer concerns of off-target integration.
Ornithine transcarbamylase (OTC) deficiency
OTC deficiency, an X-linked disease, is the most common urea cycle disorder. Affected individuals generally possess little or no functional OTC enzyme, but restoration of as little as 3% activity can significantly improve the disease phenotype [
43]. The
spf
ash
OTC mouse model recapitulates the hyperammonemia commonly seen in patients when stressed with a high protein diet, due to residual levels of OTC activity (~5%) in these mice [
30]. To accomplish therapeutic editing in vivo at the
Otc locus, an SaCas9/gRNA system was used to correct a disease-causing mutation in the
spf
ash
mice [
44]. SaCas9 under expression of the liver-specific TBG promoter was packaged into a serotype-8 capsid and co-delivered with a second AAV8 carrying the gRNA cassette and a donor template. In mice treated as neonates, 10% of
Otc alleles were corrected and Otc enzyme activity was restored to 20% and 16% at 3 and 8 weeks, respectively, in liver homogenates. All of the mice treated as neonates survived a 1-week high protein stress test, and manifested significantly lower ammonia levels compared to untreated
spf
ash
mice, where one-third of control
spf
ash
mice died or had to be euthanized when treated in an identical fashion. In comparison to the neonatal mice,
spf
ash
mice treated as adults exhibited higher levels of NHEJ, with deletions large enough to extend into the adjacent exon. Although all treated adults displayed effects of Otc deficiency, the higher doses of AAV were more severe, culminating in termination of the experiment at 2 weeks. These animals showed elevated levels of plasma ammonia, suggesting that editing presumably led to increased mutations in the Otc gene, causing substantial loss of Otc activity and leading the mice to succumb to the resultant hyperammonemia. Such an outcome emphasizes the importance of further characterizing DNA repair in vivo after genome manipulation.
Lysosomal storage disorders
The proof-of-concept studies reviewed above clearly demonstrate the therapeutic potential of genome editing approaches. Future success will depend on identifying and implementing correction strategies applicable to a broad range of disorders, such as insertion of a therapeutic transgene at a safe harbor locus. The same system using ZFNs to insert a cDNA into the albumin locus also led to a similar approach to treat two lysosomal storage disorders [
34]. Preclinical studies performed in Hurler and Hunter syndrome mouse models demonstrated high activity levels of alpha-L-iduronidase and iduronate sulfatase, respectively, in liver and plasma post-treatment with a single dose of two AAVs – one to deliver the
Alb ZFNs, and the other a cDNA donor to correct the enzyme defect. A unique feature of certain lysosomal storage disorders is that cross correction, i.e., the restoration of enzyme activity in uncorrected tissues that take up the secreted enzyme from the circulation, can be observed. The data suggests that therapeutic levels of alpha-L-iduronidase and iduronate sulfatase produced in the liver could have widely beneficial effects.
In addition to albumin, there is the potential for integration into other safe harbor loci. Targeting of the glucose-6-phosphatase (
G6PC) cDNA into the
Rosa26 locus using ZFNs led to increased survival in a mouse model of glycogen storage disease type Ia (GSD1a) [
45]. A dual AAV system was employed, with one vector delivering ZFNs and a second the
G6PC transgene. While 8-month-old knockout mice showed improved survival, from 43% to 100%, when components were delivered with serotype 8 vectors, correction of standard GSD1a biomarkers was observed only when components were delivered via AAV serotype 9. Interestingly, there appeared to be a selective advantage for stably transduced G6P-ase expressing hepatocytes, as determined by an increase in allele modification for treated knockout mice versus treated wild-type littermates, a phenomenon not previously observed. This study establishes
Rosa26 as a safe harbor locus for transgenesis via gene targeting in alternate murine disease models. In human-derived models, parallel efforts have successfully integrated therapeutic cDNA into the AAVS1 site in stem cells [
46] and T cells [
47]; the ongoing transcriptional activity of transgenes inserted in AAVS1 makes it a plausible target [
48]. Murine and rodent models have been generated carrying the AAVS1 locus [
49] and could serve to test future AAVS1-targeting genome editing platforms. Table
1 summarizes the studies performed to date that have successfully employed in vivo genome editing to correct or ameliorate disease manifestations in mouse models of IEMs.
Table 1
Summaries of genome editing studies performed on preclinical models of inborn errors of metabolism
Hemophilia B | Targeting FIX disease locus Dual AAV8 vectors: Zinc-finger nucleases (ZFNs) Partial FIX cDNA Neonatal, hemophilia B mouse model | First in vivo study using therapeutic genome editing Low rates of homology-directed repair (HDR) detectable off-target events (<4%) | |
Hemophilia B | Targeting Alb safe harbor Dual AAV8 vectors: ZFNs Partial FIX cDNA Adult C57BL/6+ hemophilia B mouse models | Low rate of NHEJ-mediated correction (0.5% fused mRNA transcripts) Due to integration in Alb, secretion of corrected protein led to therapeutic levels of circulating FIX Duration of effect 12 weeks after single treatment | |
Hemophilia B | Targeting Alb safe harbor No endonuclease Single AAV8 vector targeting FIX donor as 2A fusion to Alb
Neonatal + adult hemophilia B mice | No off-target Low rate of HDR (0.1% fused mRNA transcripts) Addition of amino-terminal proline to FIX due to 2A peptide | |
Hereditary Tyrosinemia Type I | Targeting Fah disease locus CRISPR: SpCas9/gRNA + ssDNA oligonucleotide Naked DNA | Positive selection of hereditary tyrosinemia type I (HT-I) mouse model Low HDR (0.4%); increased to 33.5% after 30 days Off-target for Fah gRNA < 0.3% (NIH3T3 cells) | |
Hereditary Tyrosinemia Type I | Targeting Fah disease locus CRISPR: SpCas9 mRNA, LNP delivery (nano.Cas9) FAH2 gRNA + donor (AAV-HDR) | Transient expression SpCas9 (LNP) HDR 6% without selection Low level off-target cutting (Hepa1-6 cells) | |
Hereditary Tyrosinemia Type I | Targeting Hpd (disease locus) CRISPR: SpCas9 + 2 gRNAs non-homologous end joining (NHEJ) Naked DNA | Positive selection of HT-I mouse model Metabolic reprogramming Multiplex editing (2 guides) 8% NHEJ efficiency at both cut sites 1-week post 68% efficiency 4-weeks post (+ selection) | Pankowicz et al., 2016 [ 39] |
Hereditary Tyrosinemia Type I | Targeting Alb safe harbor No endonuclease rAAV8: Hpd shRNA + hFIX C57Bl/6 wild-type mice | Inducible positive selection using CEPHOBA Initial homologous recombination < 1%; after selection 50% | Nygaard et al., 2016 [ 41] |
Ornithine Transcarbamylase Deficiency | Targeting Otc disease locus CRISPR: SaCas9 + gRNA Dual AAV8 vectors: SaCas9 + gRNA-donor Otc mouse model | Smaller Cas9 orthologue
S. aureus
10% HDR in neonatal mice Large deletions in adult mice ➔ hyperammonemia | |
Lysosomal Storage Disorders (MPSI, MPSII) | Targeting Alb safe harbor Dual AAV8 vectors: ZFNs + donor C57Bl/6 wild-type mice | Therapeutic protein detectable by Western blot Phase I clinical trial MPSI: 3 AAV6 vectors: ZFN + ZFN + donor | |
Glycogen Storage Disorder Type Ia | Targeting Rosa26 safe harbor Dual AAV8/AAV9 ZFNs + donor GSD1a mouse model | AAV8: Improvement in survival AAV9: Improvement in biochemical phenotype Positive selection of corrected hepatocytes | |