Current Status on Clinical Development of Adeno-Associated Virus-Mediated Liver-Directed Gene Therapy for Inborn Errors of Metabolism

Introduction

Inborn errors of metabolism (IEM) are defined as disorders in which impairment of a biochemical pathway is intrinsic to their pathophysiology, and include more than 1,000 conditions. ¹ Although each of these conditions is individually rare, as a group, IEM have an overall prevalence of about 1:2,000 at birth. ² Improvements in dietary and pharmacologic interventions have significantly reduced morbidity and mortality. However, most IEM are still lacking effective treatments. The lack of therapies is particularly relevant because expanded newborn screening based on tandem mass spectrometry can detect several IEM at birth, thus allowing early diagnosis and potential treatment before the onset of invalidating complications. ³

In several IEM, the liver is a central organ and the main target for treatment because most metabolic reactions take place in the hepatocyte. For this reason, orthotopic liver transplantation (OLT) has been used for treatment of many IEM. Although progresses in surgical procedures and perioperative management have significantly improved patients' and graft survival in the last decades, OLT is still affected by several limitations, including donor organ availability, risks related to the surgical procedure, stress-related metabolic decompensation, and toxicity associated with lifelong immunosuppression. IEM that can be treated by OLT are virtually targets also for liver-directed gene therapy.

For disorders affecting biochemical pathways expressed in other organs besides liver, OLT provides only partial benefit. For example, in classic organic acidurias, such as methylmalonic and propionic acidemia, which are complicated by multiorgan involvement and risk of acute metabolic decompensation, OLT can improve metabolic stability and arrest the progression of some complications, but it does not appear to completely abolish the risk of metabolic stroke. ^4,5 Therefore, liver-directed gene therapy in these disorders might not be completely effective.

Liver Gene Therapy with Adeno-Associated Virus Vectors

Adeno-associated virus (AAV) vectors have emerged as excellent tools for liver-directed gene therapy based on their safety and efficacy. In vivo administration of AAV vectors has shown efficacy in clinical trials for some inherited diseases with a major milestone being the liver-directed gene therapy clinical trial for hemophilia B by intravenous AAV8 administrations. In this trial, long-term expression (>4 years) of the factor IX in several patients resulted in reduction of bleeding episodes and need for prophylaxis. ^6,7 Other trials have confirmed the efficacy in both hemophilia A and B patients, using AAV5 or bioengineered AAV vectors, ^{8 –10} and a number of additional trials are currently ongoing.

However, with growing preclinical and clinical experience, AAV gene therapy has also showed a number of shortcomings requiring further investigation and optimization. These issues include risk of acute toxicity at extremely high doses, pre-existing immunity against AAV vectors, immune response against transduced hepatocytes, loss of transgene expression due to liver cell division, and risks of long-term complications.

In contrast to adenoviral vectors, AAV vectors have never been associated with acute toxicity in clinical trials. However, an acute and life-threatening hepatotoxicity has been recently observed after systemic administration of very high doses (2 × 10¹⁴ gc/kg) of an AAV9-like vector in nonhuman primates (NHP), ¹¹ raising concerns on the use of these vectors at extremely high doses in humans. The same dose of AAV9 has been administered in pediatric patients in a clinical trial for spinal muscular atrophy type 1 (SMA), ¹² without reported serious treatment-related adverse events besides increased transaminases ¹² and based on results from a phase 3 clinical trial (NCT03461289), the Food and Drug Administration has recently approved the AAV-based gene therapy product for SMA. ¹³ A threshold of toxicity for systemic AAV presumably exists and intravenous infusion of extremely high doses of AAV could eventually result in acute toxicity if this threshold is overcome. Nevertheless, based on available clinical data, it is likely that doses currently used in most clinic trials are below this threshold.

AAV-mediated gene therapy results in the introduction of two types of potentially immunogenic antigens, which are the viral capsid proteins and the transgene product. Being placed between the gastrointestinal system and the systemic circulation, the liver has unique immunologic properties and can induce immune tolerance to foreign antigens, likely through a regulatory T cell-mediated mechanism. ¹⁴ For this reason, transgenes expressed in hepatocytes under the control of liver-specific promoters typically do not elicit an immune response and induce tolerance, which can even eradicate pre-existing antibodies to the transgene product. ¹⁵ Systemic injections of AAV vectors result in the formation of neutralizing antibodies (NAbs) against the viral capsid.

The same antibodies are also produced after natural exposure to the wild-type virus. Although NAbs are specific for different serotypes, there is often cross-reactivity. ¹⁶ Subjects who are positive for NAbs are excluded from gene therapy trials because NAbs prevent vector-mediated liver transduction. ¹⁷ Moreover, following vector administration in naive subjects, formation of NAbs prevents readministration of the same vector serotype, which could be required if transgene expression is lost as a consequence of liver growth.

To overcome the obstacle of NAbs, AAV serotypes with a lower prevalence of NAbs in the general population can be used. Other strategies are under investigation and include the following: (1) administration of immunosuppressive drugs at time of vector administration, (2) removal of NAbs by plasmapheresis, ¹⁸ (3) AAV preparations with empty capsids that function as decoys for binding to NAbs, ¹⁹ and (4) generation of AAV vectors that are resistant to NAbs by targeted mutagenesis of capsid epitopes or by directed evolution in the presence of high levels of NAbs. ²⁰ Moreover, coadministration of nanoparticles carrying the immunomodulatory drug rapamycin at the time of systemic AAV delivery was recently found particularly effective in preventing the formation of NAbs in both mice and NHP. ²¹

The cytotoxic T cell (CTL)-mediated immune response against hepatocytes expressing capsid antigens has only been observed in humans so far and not in preclinical models. It was first seen in hemophilia B subjects after systemic administration of AAV2 through the hepatic artery. About five weeks after vector infusion, in these subjects, an increase in factor IX activity up to 11% was followed by loss of expression ²² and asymptomatic increase in transaminases. CD8⁺ cells reactive against viral capsid proteins were also detected. ²² A similar response was also observed in four of six hemophilia B subjects treated with the highest dose (2 × 10¹² gc/kg) of an AAV8, but a short course of oral prednisolone appeared to blunt the immune response and prevent complete loss of transgene expression. ⁷

Based on this evidence, transaminase levels have been carefully monitored and steroids have been used in subjects receiving AAV vectors in following clinical trials. However, the timing of intervention, dose and duration of steroids, and the use of other immunosuppressive drugs remain open issues and, to date, there is neither an optimal protocol nor consensus on the most effective immunosuppressive strategy.

Vector-related factors, including DNA conformation, presence of CpG dinucleotides or product-related impurities, which could trigger innate immunity, and the amount of empty capsids in the final formulation, can possibly affect the CTL response against AAV-transduced cells, ²³ but these factors are not yet included in designing the optimal immunosuppressive regimen.

Besides the CTL-immune response, an important limitation of AAV administration in children is the loss of vector genomes secondary to liver growth that requires vector readministration, hampered by NAbs, to maintain sustained transgene expression. This is particularly relevant in IEM that often have childhood onset and require early treatments. Strategies to prevent the formation of NAbs and serotype switching (discussed earlier) have the potential to overcome this important limitation. Alternatively, persistent transgene expression can be achieved through genome integration by genome editing strategies (discussed in the next section) or integrating vectors, such as lentiviral vectors. ²⁴

AAV vectors are largely episomal, but a low-rate integration in the host genome has been observed in preclinical models and it has been associated with a potential risk of genotoxicity. An increased frequency of hepatocellular carcinoma (HCC) was observed for the first time in mucopolysaccharidosis type VII mice treated in the newborn period with systemic administration of an AAV vector expressing the human GUSB gene under the control of a ubiquitous promoter. ²⁵ HCC formation has been related to AAV integration at the Rian locus within murine chromosome 12, specifically in mir341 that has no human orthologs, causing overexpression of adjacent microRNAs involved in hepatocyte proliferation. ²⁶ Factors influencing the risk of carcinogenesis in mice include high vector doses, enhancer/promoter elements included in the vector, and administration in the neonatal period. ²⁶ However, no evidence of insertional mutagenesis has been observed in long-term studies with adult-injected mice and large animal models. ^{27 –29} Moreover, AAV5 integration analysis in NHP and patients enrolled in a gene therapy clinical trial for acute intermittent porphyria (AIP) ³⁰ did not reveal integration clusters or events in genes previously reported in HCC. Taken together, these observations suggest that if present, the risk of HCC in humans is very low in the context of AAV gene therapy with vectors administered beyond the neonatal period.

Clinical Studies of AAV-Mediated Liver-Directed Gene Therapy in IEM

Besides hemophilia A and B, there are an increasing number of AAV vector clinical trials for IEM that are now under investigation. In the following section, we provide an overview of the most representative examples of liver-directed gene therapy clinical trials, which have either been completed or are under development.

A clinical trial with an AAV5 vector has been completed in 8 patients with severe AIP, a disorder caused by a defect of the porphobilinogen deaminase enzyme presenting with recurrent acute attacks of severe abdominal pain, nausea, vomiting, tachycardia, hypertension, peripheral neuropathy and other neurological symptoms. Subjects enrolled in this dose escalation trial, including four cohorts (5 × 10¹¹ gc/kg; 2 × 10¹² gc/kg; 6 × 10¹² gc/kg; and 1.8 × 10¹³ gc/kg), did not show significant evidence of clinical efficacy, neither on biochemical (urinary levels of δ-aminolevulinic acid and porphobilinogen) nor on clinical endpoints. ³¹ In conclusion, this trial showed safety but failed to show signs of efficacy.

Urea cycle disorders have long been considered targets for liver-directed gene therapy because urea cycle enzymes are mainly expressed in the liver and the risk/benefit ratio is favorable. Most preclinical research focused on the most common disorder, that is ornithine transcarbamylase (OTC) deficiency. Preclinical studies ^{32 –34} have supported a phase I/II clinical trial based on the intravenous administration of AAV8 expressing human OTC that is currently ongoing in adult subjects with late-onset disease (NCT02991144). However, the disease onset of urea cycle disorders is frequently in the neonatal period, and the efficacy of early gene therapy is limited by the need of achieving therapeutic expression levels with sufficient rapidity ³⁵ and the progressive loss of transgene expression due to the liver growth.

Similarly, clinical trials with escalating doses of AAV8 are ongoing in subjects with the following IEM: (1) homozygous familial hypercholesterolemia (NCT02651675), a severe disease caused by a defect in the gene encoding the low-density lipoprotein (LDL) receptor, associated with very high levels of plasma LDL cholesterol leading to increased risk of atherosclerosis and premature death; (2) Crigler–Najjar syndrome, a disorder caused by a defect in bilirubin conjugation manifesting at birth with persistent and severe unconjugated hyperbilirubinemia leading to risk of kernicterus and irreversible neurological damage (NCT03466463); and (3) glycogen storage disease type Ia, caused by deficiency of glucose-6-phosphatase and presenting with hypoglycemia and glycogen accumulation in the liver and kidneys (NCT03517085).

In contrast to the disorders described above, in which the primary goal is to correct the hepatocyte metabolic defect by delivering the therapeutic gene to liver cells, other liver-directed gene therapy approaches focused on transducing the liver to express and secrete in the systemic circulation therapeutic proteins that can cross-correct other diseased body districts. Examples of this approach include mucopolysaccharidosis type VI (MPSVI) due to deficiency of arylsulfatase B (ARSB gene) and glycogen storage disease type II, also known as Pompe disease (PD) due to deficiency of the lysosomal acid α-glucosidase (GAA).

For both disorders, enzyme replacement therapy (ERT) consisting in repeated intravenous administrations of the enzyme is clinically available. The mode of action of ERT is based on the principle of cross-correction, that is the cellular uptake of circulating lysosomal enzymes via mannose-6-phosphate receptors followed by delivery into lysosomes. However, this treatment has important limitations, such as the need for lifelong infusions, high costs, inability to target all affected tissues, and immune reactions to the recombinant proteins. ³⁶ Gene therapy aiming at expressing the protein in hepatocytes to convert the liver into a “factory” organ for enzyme production and secretion has the potential to overcome these limitations and has provided encouraging preclinical data.

MPSVI is a multisystem disorder due to accumulation of dermatan sulfate in multiple organs and tissues, in the absence of primary brain involvement. Disease manifestations include dysostosis, joint limitation, organomegaly, cardiac valve disease, obstructive/restrictive respiratory disease, and corneal clouding. In this disorder, long-lasting therapeutic levels of circulating enzyme achieved by systemic AAV8 infusion resulted in normalization of glycosaminoglycan levels, reduction of cardiac valve thickness, significant improvement of motor ability, and even partial improvement of femur length in multiple disease animal models. ^{37 –39} Moreover, the efficacy of AAV8 infusions was found to be similar to weekly ERT in mice. ⁴⁰ Based on these encouraging preclinical results, a clinical trial with escalating doses of AAV8 expressing human ARSB is currently ongoing (NCT03173521).

In PD, deficiency of GAA results in accumulation of glycogen in cardiac and skeletal muscles, and the main clinical features include cardiomyopathy and progressive respiratory failure. As muscle is the most affected tissue, intramuscular administrations of AAV vectors with muscle tropism have been largely investigated. However, the first clinical trial using intradiaphragmatic administration of an AAV vector expressing human GAA in subjects with PD showed efficacy limited to the injected muscle. ⁴¹ Formation of anti-GAA NAbs with high frequency is an important issue in PD as they limit the efficacy of ERT and increase the risk of severe immune reactions. Liver-directed gene therapy has the potential both to provide a source for sustained endogenous GAA production and to induce immune tolerance due to liver-specific expression of the transgene. Preclinical studies have indeed shown that hepatic GAA expression by AAV can partially correct the skeletal muscle, ^42,43 induce immunologic tolerance, and prevent NAb formation also in the context of ERT. ^44,45 Based on these premises, AAV8-mediated liver gene transfer is currently under investigation in a clinical trial for subjects with late-onset PD (NCT03533673).

Because high circulating levels are needed to achieve therapeutic levels of GAA and biochemical correction, an engineered codon-optimized human GAA resulting in increased secretion has been investigated for AAV8-mediated liver-directed gene therapy and this vector resulted in higher circulating levels compared with the vector expressing the native protein, correction of glycogen storage in heart and multiple skeletal muscles, improvement of muscle function, reduced immunogenicity, and long-term survival. ⁴⁶

Liver Genome Editing for IEM

Genome editing can either correct the genetic defect at its chromatin locus (“gene repair”) or insert the therapeutic transgene in a different locus (e.g., under the control of a highly active promoter, typically the albumin promoter in hepatocytes). Genome editing allows stable genomic integration of the transgene and overcomes the limitation of loss of transgene expression due to cell division occurring in gene replacement approaches. This is generally achieved by AAV-mediated delivery of the transgene flanked by sequences (donor DNA) that are homologous to the target locus of integration. Then, endogenous homology-directed repair (HDR) operates targeted integration and the efficiency of HDR can be greatly enhanced if delivery of the donor DNA is coupled to delivery of a nuclease [Zn-finger (ZFN) or CRISPR/Cas9] that induces double-strand breaks. When genome editing is performed at the albumin locus, the therapeutic gene is expressed at supraphysiological levels ⁴⁷ and is particularly effective in noncell autonomous IEM due to defects of secreted proteins, such as lysosomal enzymes that can be released at high levels in the circulation to correct multiple affected body sites. However, targeting the brain remains challenging even if high circulating levels of enzyme are achieved because the bloodborne enzyme does not cross the blood/brain barrier. A ZFN-based approach has recently reached clinical stage and is currently under investigation in subjects with MPSI and MPSII (NCT02702115 and NCT03041324). To overcome the risks related to off-target effects induced by DNA nucleases, promoter-less therapeutic genes have been inserted in-frame just upstream of the albumin stop codon, resulting in a fused messenger RNA that is translated into albumin and into the therapeutic gene. ⁴⁸ This strategy has also been investigated in a mouse model of Crigler–Najjar syndrome resulting in rescue of the lethal phenotype and partial reduction of bilirubin levels. ⁴⁹

In hereditary tyrosinemia type I (HT1) due to deficiency of fumaryl acetoacetate hydrolase (FAH gene), correction of the genetic defect by homologous recombination provides a selective advantage to corrected hepatocytes that can proliferate and repopulate the liver. Integration of the correct sequence delivered by AAV into the genome was achieved in HT1 mice through spontaneous homologous recombination without introducing a DNA break, followed by proliferation and hepatic repopulation by corrected hepatocytes. ⁵⁰ A further study used CRISPR/Cas9 and a donor template from the genomic Fah sequence to increase the efficiency of gene repair by HDR in HT1 mice. ⁵¹

Because the whole-genome editing system exceeds the AAV cargo capacity, the single-guide RNA (sgRNA) and the donor template were delivered by an AAV vector, while lipid nanoparticles were used to deliver the Cas9. This approach resulted in transient Cas9 expression that prevents nonspecific DNA breaks and undesirable immune reactions. Despite the rate of hepatocytes correction was still modest (∼6%), a substantial phenotypic correction and a significant reduction of liver damage were observed. ⁵¹

In adult nonproliferating hepatocytes, nonhomologous end joining (NHEJ) is the most frequent repair mechanism intervening after DNA double-strand breaks and can result in insertions/deletions (indels) in the coding sequence of target genes leading to gene disruption. This issue became evident in the adult OTC-deficient spf^ash mouse model that showed low levels of correction and dose-dependent toxicity after delivery of CRISPR/Cas9 by two AAV8, one expressing Cas9 and the other the sgRNA and donor sequence. ⁵² Deep sequencing of the Otc target region revealed frequent and extensive on-target indels, possibly leading to a worsening of the disease phenotype in this mouse model. ⁵² The efficiency of NHEJ has been exploited to insert a donor DNA with the therapeutic transgene at a given locus in the absence of homology arms. In this approach, one AAV vector is used to deliver CRISPR/Cas9 and a second AAV vector to deliver both the sgRNA expression cassette and the donor DNA flanked by CRISPR/Cas9 sgRNA target sites. Upon delivery of both AAV vectors, CRISPR/Cas9 cleaves both the genomic locus where the insertion should occur and the donor DNA allowing NHEJ to insert the donor DNA at the double-strand break in the genomic locus. This homology-independent targeted integration occurred in about 4% of mouse hepatocytes. ⁵³

Another application of gene editing is targeted disruption of an upstream metabolic reaction to reduce production and accumulation of toxic metabolites. This strategy exploits NHEJ to induce indels in target genes and it has been investigated in primary hyperoxaluria type I (PH1), ⁵⁴ a disorder resulting in oxalate overproduction and systemic tissue damage due to deposition of calcium oxalate stones. Inhibition of glycolate oxidase (GO), an upstream enzyme in the glyoxylate pathway, results in a mild phenotype with asymptomatic increase of glycolate urinary excretion, and has been proposed as a strategy to reduce glyoxylate production in PH1. ⁵⁵ An AAV8 expressing CRISPR/Cas9 targeting the murine gene encoding GO and sgRNA under a liver-specific promoter was effective in PH1 mice at reducing toxic oxalate levels. ⁵⁴

However, long-lasting expression of the editing machinery can lead to DNA damage, improper gene editing, and deleterious cellular effects that ultimately might predispose to cancer. ⁵⁶ Although not observed in preclinical animal models, this possibility and its genotoxic consequences are serious concerns. Another major obstacle for therapeutic applications is the specificity of the genome editing. Off-target cleavage is a known limitation of genome editing tools, and indels at off-target sites have been repeatedly reported. ^57,58 In addition, a better understanding of the factors that influence CRISPR/Cas9-mediated genome editing is needed. Two independent reports have showed that CRISPR/Cas9-mediated genome editing leads to p53-mediated stress response and cell cycle arrest. ^59,60 These findings raise concerns that CRISPR/Cas9 genome editing might be biased toward tumor-prone cells with p53-deficient activity. When cultured ex vivo for genetic correction, cells with p53 dysfunction might undergo selection and while such clonal selection could be identified and eliminated by screening before infusions into patients, it could not be avoided during in vivo gene correction.

In addition, humans harbor pre-existing adaptive immune responses to the Cas9 orthologs derived from Streptococcus pyogenes and Staphylococcus aureus and CRISPR/Cas9 can induce in animal models an immune response with elimination of edited cells expressing Cas9 antigens. ⁶¹ Nevertheless, further work is needed to determine whether these findings have an impact on in vivo genome editing.

Conclusions

Clinical applications of AAV vectors for liver-directed gene therapy have increased tremendously in the last few years. We are learning from each clinical study and expanding our knowledge on both therapeutic efficacy and drawbacks of AAV vectors in humans. Although clinical evidence confirms their overall safety, important issues remain to be addressed. The main limitations currently include the immune responses and the loss of expression with liver growth. Moreover, several IEM pose specific problems related to their early onset, multisystemic involvement, or biologic aspects (e.g., cell autonomous disorders). Different strategies are under investigation to overcome present limitations and they have the potential to further expand the therapeutic applications of AAV in IEM.