An Overview of Nonclinical and Clinical Liver Toxicity Associated With AAV Gene Therapy

Abstract

This article reviews the presentation given at the 2023 annual meeting of the Society of Toxicologic Pathology (STP) on liver toxicity observed with adeno-associated viral vector (AAV) gene therapy. After decades as a therapeutic modality largely confined to the academic research environment, gene therapy has emerged in recent years as a rapidly expanding therapeutic approach in the biopharmaceutical industry with AAV as the most commonly used viral vector for gene delivery. This interest in the field of gene therapy by industry has been enhanced by the recent success of approved therapies for curing genetic diseases such as ZOLGENSMA for spinal muscular atrophy and LUXTURNA for Leber congenital amaurosis. However, recently reported clinical and nonclinical toxicities highlight the challenges in safely developing AAV gene therapies that require high dose systemic administration. The presentation reviewed general attributes of AAV as a gene therapy vector, clinical and nonclinical liver toxicity associated with AAV gene therapy and the potential for a multimodal immune suppression strategy that may mitigate toxicities.

Gene therapy is a therapeutic intervention to genetically alter or modify living cells using either viral or nonviral vectors to replace a missing or defective gene in order to correct a disease of genetic origin or one associated with altered gene expression.^1,21 Numerous gene delivery approaches are being evaluated including viral vectors (eg, adenovirus, adeno-associated virus, herpesvirus, and lentivirus / retroviruses), as well as nonviral vectors (eg, naked DNA/plasmid DNA delivered via liposomes, nanoparticles, or exosomes).^{7,12,14,26,38} Associated viral vector (AAV) has emerged as the most frequently utilized viral vector for in vivo use because there are (1) multiple naturally occurring serotypes of AAV with different tissue tropisms, (2) biotechnology approaches for engineering novel capsids with unique tissue tropism and other biologic properties, (3) widespread infection of wild type AAV in mammalian populations without evidence of pathogenicity, and (4) AAV is able to express a transgene without integrating into the host cell genome (ie, episomal) compared with lentiviral / retroviral vectors which require integration. However, recently reported clinical and nonclinical toxicities highlight the challenges in safely developing AAV gene therapies. This presentation at the 2023 Society of Toxicologic Pathology (STP) annual meeting reviewed key attributes of AAV gene therapy vectors, liver toxicities observed in clinical and nonclinical studies, and the potential for a multimodal immune suppression strategy to mitigate some of these toxicities.

Wildtype AAV is a single-stranded DNA parvovirus that is replication defective and requires the help of adenovirus or herpesviruses for replication. ^7,12,32,38 Wildtype AAVs have a single replication (rep) gene encoding proteins necessary for viral replication, a single capsid (cap) gene generating three viral capsid proteins (VP1, VP2, VP3), and assembly-activating proteins necessary for viral capsid formation. The capsid proteins define the AAV serotype and determine cell and tissue tropism, (eg, biodistribution), and intracellular trafficking following in vivo injection. The DNA of AAV gene therapy vectors retains the wild type AAV inverted terminal repeats (ITRs) which are required for packaging of the vector DNA within the capsid during vector production and second strand synthesis when the vector enters a cell. The rep and cap genes of the wild type AAV are replaced by the engineered transgene expression construct of that is placed between the ITRs.

The three capsid proteins form 1 of 60 icosahedral units that are assembled to make up the 20- to 25-nM particle that encapsulates DNA construct that is expresses the gene of interest.^32,38 The DNA expression construct may be single or double-stranded (also called self-complementary). The double stranded form enhances the rapidity of expression but decreases the size of the construct that can be packaged from approximately 5000 to 2500 kilobases of DNA. In its simplest form, the expression construct placed between the ITRs consists of a promoter that drives expression of the transgene mRNA and a polyA RNA termination signal. Other elements may be added to regulate the expression or stability of the transgene mRNA expression.

Associated viral vectors can be made by several different production methods ³² with the two most common approaches being transfection of the human HEK cell line or insect cell line (baculovirus) with plasmids that contain the expression construct of interest, genes encoding the capsid protein, and genes required for production of the intended viral particle. Regardless of the production method, the therapeutic product is a complex mixture of viral particles that contain the intended DNA expression construct as well as process-related impurities that consist of capsids that lack DNA (eg, empty capsids) and partially filled capsids.^4,6,35-38,42 In addition to the intended DNA expression construct, the encapsulated DNA is a mixture of unintended DNA forms consisting of truncated expression construct, chimeric DNA that is a mixture of the DNA from the plasmids required for vector production, and the production cell line DNA as well as free fragments of DNA from the same sources. Other process-related impurities that are in the final product may also include proteins from the production cell line as well as reagents used in the purification process. One of the major efforts in developing AAV therapeutic products is optimizing the vector design and production process to minimize the content of unintended DNA process-related impurities, empty capsids and other impurities from the production process. Additionally, the production process needs to be standardized so that test material used for pivotal nonclinical pharmacology and toxicity studies is comparable with material used in clinical trials. Some of the DNA process-related impurities in addition to interacting with foreign DNA recognition pathways that initiate innate immune and antiviral response may also be transcriptionally active and integrate into host cell DNA. ⁶ The role that various types and quantities of process-related impurities have in the toxicity profile of AAV gene therapy is poorly understood and is an area for further investigation.

Associated viral vector gene therapy vectors are complex products and their interactions with humans and animal species are also complex. Associated viral vector gene therapy vectors are complex products and their interactions with humans and animal species are also complex. Associated viral vector does not. Associated viral vector does not transduce hepatocytes uniformly within the hepatic lobule and there can be marked differences in species and serotypes in hepatic transduction efficiency that may contribute to the differences in toxicity profile between species and vector constructs. Two examples were provided. AAV-9 and a closely-related capsid variant (PHP.b) have a pronounced preference for periportal distribution in the cynomolgus monkey liver whereas in the mouse AAV-8 may be more uniformly distributed. Sex-related differences in hepatic tropism, occurring with multiple serotypes, have been observed in mice where males typically show greater liver transduction than females. When using in situ hybridization to track the cell distribution of AAV DNA and mRNA expression in the liver, a frequent observation is that there is more diffuse distribution of hepatocytes that contain vector DNA compared with those expressing transgene mRNA. The intracellular pathway from cellular uptake of an AAV particle to transgene expression is complex ³⁸ and the underlying biology that results in variable expression of the transgene between individual hepatocytes has not been elucidated. The variable expression of transgene mRNA between adjacent cells of the same type that contain vector DNA is not unique to hepatocytes.

Liver toxicity has been observed in both animal and humans following AAV administration. There is poor concordance between the liver toxicities observed in animals and humans. In humans, liver injury following AAV administration was first observed in early clinical trials in hemophiliac patients that were administered AAV that delivered the factor IX gene in hemophilia B patients.^{10,22-24,27-29} The liver injury was evidenced by an increase in the serum liver transaminases and a decrease in factor IX activity several weeks after vector administration. The liver injury was clinically asymptomatic and preservation of factor IX expression could be maintained if corticosteroids were used when there was an initial increase in liver enzymes and / or reduction in coagulation factor expression. Prophylactic use of corticosteroids is also effective in minimizing the potential for liver injury. This presentation has been a common finding in clinical trials of gene therapy for hemophilia B or A. The liver injury has been linked to a cell-mediated response to the AAV capsid but not the transgene protein. ²⁴ The reader is referred to multiple references that review the clinical presentation and mechanistic hypothesis for the liver injury observed in hemophilic patients.^{10,23,24,27-29} Nonclinical studies in rodents, dogs and nonhuman primates that were conducted to initiate clinical trials with AAV in hemophiliac patients did not identify liver injury as a potential liability. ²²

There have been two clinical indications where severe symptomatic liver injury has been reported in humans following AAV gene therapy. These include X-linked myotubular myopathy (MTM1) and spinal muscular atrophy (SMN). In the trial for MTM1 there were 4 patients that died following treatment and 5 others that developed severe nonfatal adverse events related to hepatobiliary disfunction.^3,5,15,22,25 There are no peer-reviewed publications describing these patients, but the reader is referred to several public disclosures on these patients.^2,3,5,15,25 As in the hemophilia example, nonclinical studies in multiple species and disease models did not identify hepatic injury as a potential liability. It is also important to note that MTM1 is associated with preexisting hepatobiliary disease that may have potentiated the direct hepatic injury induced by AAV. The liver injury was characterized clinically as progressive severe cholestatic liver failure that was identified as severe adverse event (SAE) in patients starting 37 to 142 days post AAV administration, although liver enzyme increases were first noted 1 to 4 weeks post treatment. The available description of the liver histology indicates the process was not immune mediated as there was no immune cell infiltrate. At the FDA advisory meeting on AAV therapeutics the sponsor indicated that the current mechanistic hypothesis is a disruption of hepatocellular metabolism due to high vector genome exposure. ²

There are peer reviewed publications and public disclosures on liver injury associated with AAV gene therapy (ZOLGENSMA) in SMA patients,^8,9,11 with 2 reported fatalities associated with liver failure. Despite prophylactic glucocorticoid treatment, a large proportion of patients treated with ZOLGENSMA develop an asymptomatic transient increase in serum liver transaminases 1 to 4 weeks post-dose that is not generally associated with an increase in bilirubin. A clinical case report describes two patients that developed severe, nonfatal liver injury associated with a marked increase in serum transaminases and bilirubin 2 to 7 weeks post AAV administration that was responsive to glucocorticoid treatment. The liver biopsy histology described in this case report is consistent with a cell-mediated immune (CMI) response and consisted of hepatocyte degeneration / necrosis in zone 3, peri-portal CD8+ T cells, some neutrophils, eosinophils, and plasma cells, with marked bile ductular reaction. The CD8 positive T cell infiltrate and response to glucocorticoid treatment are suggestive of a CMI response. Two fatalities 5 to 6 weeks post vector administration associated with liver failure have also been reported in Russia and Kazakhstan following treatment with ZOLGENSMA. The ZOLGENSMA product label indicates that hepatic findings in mouse toxicity studies were characterized as hepatocellular hypertrophy, Kupffer cell activation, perinuclear vacuolization, and scattered hepatocellular necrosis. In nonhuman primates the product label indicates there was oval cell proliferation in the liver at 6 weeks with a decrease in severity at 6 months.

In nonclinical studies with AAV, liver injury has been reported in rodents and nonrodents. Injury associated with overexpression of the transgene, CMI repose to the transgene, and acute injury that appears to be a direct effect on hepatocytes have been reported. Liver toxicity observed may occur early within the first few days and may or may not be associated with transgene expression or an adaptive immune response. There are also marked differences in nonclinical species susceptibility to hepatic toxicity.

While CMI responses to the vector capsid observed in hemophilia clinical trials were not observed in nonclinical studies, there are examples of CMI to the transgene resulting in liver toxicity when a strongly immunogenic transgene is used. These responses typically occur 2 to 4 weeks post dose. The reader is referred to several articles that discuss differences in cell-mediated response to AAV vectors in animals and humans.^13,17

Two examples of hepatic toxicity associated with transgene overexpression were provided. Hepatic toxicity related to overexpression of short hairpin RNA (shRNA) has been observed in rodents and nonrodents. ³⁴ The underlying mechanism of this has been shown to be disruption of normal microRNA processing due to the overexpression of shRNA that inhibits transport of endogenous pre-microRNAs through the nuclear envelope protein, exportin-5, into the cytoplasm. Decreasing the strength of the promoter that drives the expression of the shRNA or converting the shRNA structure to microRNA hairpin structure that is more efficiently processed by the endogenous microRNA processing pathway mitigates this type of toxicity. A second example of hepatic toxicity in mice was overexpression of frataxin protein which is involved in the intracellular respiration and iron metabolism. ²⁰ Abrogating the biologic activity of the frataxin protein with a point mutation without altering the expression of the protein mitigated the toxicity. Toxicity related to overexpression of the transgene is not unique to the liver. ¹⁶

A common finding in both monkeys and rats following AAV administration is oval cell and / or bile duct proliferation (unpublished personal observations). This is generally mild and resolves over time. However, one example was presented in rats where marked bridging of portal bile duct and fibrovascular proliferation was observed 6 weeks following AAV administration. Resolution was not evaluated in this rat study.

Acute hepatocellular injury that occurs within the first 3 to 10 days post vector administration is frequently seen in cynomolgus and rhesus monkeys, but not in other nonprimate species when the dose of vector is greater than 1e13 vg/kg. ^18,19,30 This response is not related to transgene expression as it has been observed with vectors that do not have promoters that are active in hepatocytes (unpublished personal observations). In most cases the increase in transaminases are variable in magnitude, infrequently associated with an increase in bilirubin, and asymptomatic. However, with selected vectors at a dose in the mid 10¹³ vg/kg or greater range, there can be marked acute liver injury that progresses to hepatic failure and death or morbidity requiring euthanasia within 3 to 7 days post vector administration. There also appears to be individual animal susceptibility with only select animals in a treatment group progressing to liver failure with others presenting with transient asymptomatic liver injury. Histologically there is widespread to coalescing areas of individual cellular necrosis without a zonal distribution. In some cases, there are prominent intracytoplasmic eosinophilic / hyaline droplets in hepatocytes that contain plasma proteins. Platelet aggregates in hepatic sinusoids have also been reported but are not consistently seen. Complement split products in serum can also be increased, but usually only at doses where there is marked hepatocellular injury. Wide spread complement deposition is not demonstrated in the liver suggesting that complement activation is not the cause of liver injury but secondary to acute tissue damage. Dose-related neutral lipid accumulation in hepatocytes may also occur, as demonstrated by oil-red O staining, and there is a lack of leukocyte infiltrates, suggesting that the injury to hepatocytes is a direct effect of the vector. The reader is referred to several publications that describe the morphologic and clinical pathology findings associated with acute liver failure observe in monkeys following AAV administration.^18,19,30 Monkeys appear to be particularly sensitive to acute liver injury as when the same vector that induced liver failure in monkeys is administered to mice, rats or pigs, at equivalent doses, liver injury was not observed. The translatability of this acute severe liver injury in monkeys to humans is unknown.

While the acute nature of the liver injury and histologic appearance is suggestive of direct injury to hepatocytes by the vector, a study in cynomolgus monkeys was presented where the treatment of monkeys with methylprednisolone, sirolimus and rituximab prevented the rise in transaminases as well as a rise in anti-capsid IgG and IgM antibodies that occurred within the first 8 days post dose. ³¹ Methylprednisolone by itself or in combination with sirolimus did not prevent the rise in liver enzymes. Methylprednisolone had no effect on post-treatment rise in immunoglobulin, but the addition of sirolimus inhibited an increase in IgG but not IgM that occurred in the first 8 days post AAV administration. Adding rituximab to methylprednisolone and sirolimus inhibited an increase in both IgG and IgM. The presentation also highlighted an observation that has been previously reported in the literature^39-41,43 and that is the marked uptake of AAV in lymphoid germinal centers within the first few days post AAV administration. Taken together these observations suggest that B-cells may be playing an important role in the acute liver injury observed in monkeys. The mechanism by which the addition of rituximab prevents this is unknown and requires further investigation.

Associated viral vector induced hepatocellular cancer observed in neonatal mice was not discussed in the presentation but the reader is referred to a recent publication that reviews the literature and discusses the relevance to human risk assessment. ³³

Associated viral vector gene therapy is a promising therapeutic modality that is in the early stages of development as a therapeutic approach. As with other novel therapeutic modalities that have emerged from early research with promising results, there are challenges that are identified with broader application. The direct effect of AAV vectors and associated process-related impurities as well as adaptive and innate immune response on liver injury is an area for future research that will enhance the application of AAV gene therapy as a therapeutic modality.