Scientific and Regulatory Policy Committee Points to Consider: Nonclinical Research and Development of In Vivo Gene Therapy Products,Emphasizing Adeno-Associated Virus Vectors

Key Elements of rAAV Vector Design

During the discovery and early development of an rAAV-GTx product, initial design and selection of the optimal rAAV vector are the first steps to success, and basic knowledge of rAAV vector design (capsid serotype and DNA construct) is helpful in assessing the efficacy and safety of novel rAAV-GTx products in animal studies. Research pathologists should assist in the initial design and selection of the optimal rAAV vector by advising on serotype selection, promoter design and selection, interspecies sequence similarity of the transgene/gene product, comparative biology, and early morphology-based biodistribution and transgene expression characterization. Accordingly, research pathologists should have some familiarity with principles of vector design.

Cell tropism of naturally occurring AAVs is often broad, but each serotype has a biased tropism profile depending on multiple factors such as localization of different cell surface AAV receptors on host cells, AAV serotype-specific interactions with the AAV receptor, and AAV capsid interactions with the transgene promoter. ^{37 –39} To date, 12 naturally occurring AAV serotypes from both human and nonhuman primates (NHPs), and over 100 AAV variants, have been identified. ^23,40 Wild-type AAVs have a single rep gene encoding proteins necessary for viral replication, a single cap gene generating viral capsid proteins (VP1, VP2, VP3), and assembly-activating proteins necessary for viral capsid formation. The capsid proteins define the AAV serotype and help determine cell and tissue tropism, biodistribution, and intracellular trafficking following in vivo injection (Figure 2). The DNA of rAAV-GTx vectors retains the wtAAV inverted terminal repeats (ITR) yet is devoid of the rep and cap genes, which are replaced by the engineered transgene expression cassette placed between the ITRs. The choice of a serotype for a specific disease indication is guided by the tropism of the vector capsid for the cell types or tissue(s) involved in the disease process, the prevalence of preexisting antibodies against capsid proteins in the patient population (ie, seroprevalence) that might prevent tissue transduction by the vector, ^41,42 and the ease of manufacture and scale-up for that particular serotype. Consequently, considerable efforts have been geared toward the identification of naturally occurring capsid serotypes or designer-engineered novel capsid variants with optimal cell and tissue tropism and improved immune evasion profiles. ^{43 –45}

OPEN IN VIEWER

Figure 2.

Adeno-associated viruses (AAV) are composed of an icosahedral capsid (composed of VP1, VP2, and VP3 subunits) enclosing a DNA construct capable of producing a transgene product (either protein or RNA, such as a microRNA [miRNA] that inhibits or modifies endogenous mRNA). The AAV capsid can be generated from preexisting wild-type AAVs (wtAAV) or engineered into mosaic variants to improve cellular tropism, transduction efficiency, or to evade preexisting immunity. For recombinant AAV (rAAV), DNA constructs are single-stranded DNA molecules (ssAAV); however, double-stranded, self-complementary DNA constructs (scAAV) are occasionally used to avoid the requirement of second-strand synthesis and to permit quicker transgene expression in vivo. All components of rAAV can induce different reactions by the host immune system such as humoral antibody response against capsid proteins, innate immune response against foreign DNA, or cytotoxic T-cell response against cells expressing transgene protein product.

Once a capsid serotype with a favorable profile is selected, effort is directed toward the design of a suitable transgene-bearing DNA construct to be packaged within the capsid of choice. Naturally occurring AAV contains single-stranded DNA with packaging capacity limited to 4.7 kilobases (Kb). Consequently, rAAV transgene-bearing DNA construct designs generally should be within this size range (∼4.5-4.8 Kb), inclusive of upstream promoter/enhancer elements, introns, and downstream polyadenylation (poly (A)) sequences required to drive optimal transgene expression and persistence (Figure 2). ³¹ For transgenes that are larger than 4.7 kb, carefully designed truncated transgenes that conserve crucial functional domains of the targeted gene can be used instead. For example, the ∼14 Kb dystrophin gene may be spliced down to ∼5 Kb to design microdystrophin constructs currently used in several rAAV-GTx clinical trials. ⁴⁶ Alternatively, oversized vectors or regular-sized dual rAAV vectors have been used. ^47,48 Single-stranded AAV (ssAAV) constructs that rely on cellular replication factors to synthesize the complementary strand of DNA, or self-complementary AAV (scAAV) constructs in which the complementary strand of DNA is provided in the construct, may be employed in the design of rAAV vectors (Figure 2). The advantages and disadvantages of ssAAV and scAAV are summarized in Table 1. It should be noted that multiple expression cassettes can be included in the same construct if more than one transgene product is needed and their combined sizes allows them to be accommodated within the rAAV packaging capacity.

OPEN IN VIEWER

Table 1.

Advantages and Disadvantages of Single-Stranded and Self-Complementary AAV Vectors.

Vector type	Advantages	Disadvantages
ssAAV	Can carry a larger transgene cassette. Can accommodate optimal (typically longer) promoter sequences that drive more specific or ubiquitous transgene expression.	Slow rates of ssAAV transport into the nucleus and/or uncoating from the capsid are rate-limiting, resulting in delayed protein expression due to the requirement for de novo synthesis of the complementary DNA and a transient period of vector genome instability after dsDNA conversion. 49
scAAV	Immediate transgene and protein expression upon transduction.	Reduction (50%) in transgene cassette size compared to ssAAV, which requires that minimal promoters, introns, and poly (A) elements be combined to maximize the remaining space available for the transgene-coding sequence. 49
	Stronger gene expression (10- to 50-fold) than is available using ssAAV vectors at comparable doses. 50	Harder to manufacture on the large scale needed for clinical use. 51,52

Abbreviations: AAV, adeno-associated virus; scAAV, self-complementary AAV; ssAAV, single-strand AAV.

Although AAV capsid structure and target cell receptors determine cell and tissue tropism, the choice of promoter plays a critical role in controlling the specificity of transgene expression within a target cell type, the robustness of the transgene mRNA transcription and translation, and the persistence of transgene expression. Transgene expression in the desired target cells can be made more specific or enhanced in magnitude by modifying promoters. ⁵³ For example, the 1.8 Kb neuron-specific enolase (NSE) promoter, the 470 base pair (bp) human synapsin-1 (SYN1) promoter, the 229 bp methyl CpG-binding protein 2 (MeCP2) promoter, and the 2 Kb herpes simplex virus 1 latency-associated transcript (HSV1-LAT) promoter can all drive neuron-specific transgene expression, in contrast with stronger promoters such as chicken β-actin (CBA) or cytomegalovirus (CMV), which drive more ubiquitous expression. Properties of selected ubiquitous and cell type-specific promoters are detailed elsewhere. ⁵⁴

The design of the transgene can differ depending on the underlying disease mechanism. Transgenes may code for secreted proteins; nonsecreted membrane proteins; or intracellular proteins (eg, structural proteins), peptides, oligonucleotides, or inhibitory RNAs. For GTx aimed to correct monogenic disorders (eg, SMA or Friedreich’s ataxia), the transgene is composed of an open reading frame (ORF) expressing the gene product of interest in its native form or modified by codon optimization (to improve translation and reduce immunogenicity) or by deletions to allow the larger coding regions of the gene to fit within rAAV packaging capacity (such as microdystrophin for Duchenne muscular dystrophy). ⁵⁵ For diseases with toxic gain-of-function mutations where gene suppression is needed to correct the disease phenotype (eg, Huntington disease), the transgene delivered may be a dominant negative transgene, an artificial primary microRNA (miRNA), or a short hairpin RNA (shRNA) that leads to the suppression of gene expression (ie, “knockdown”). ⁴³ Post-transcriptional control elements can be engineered into the rAAV DNA construct to further fine-tune transgene expression, including regulatory sequences in untranslated regions or by codon optimization. ⁵⁶ The poly(A) tail of a transcript is also an important element to consider as it is critical for nuclear export, translation, and mRNA stability. ⁵⁴

Early Considerations Independent of rAAV Vector Design

Factors related to the host (patient or nonclinical animal species) and treatment regimen may influence the efficacy and safety profiles of rAAV-GTx products. Toxicologic pathologists are well versed in the importance of many parameters, including individual demographics (such as breed/strain, age, and sex), dose formulation, and selection of feasible doses and the optimal ROA. For example, sex has been shown to impact expression of transgenes following systemic delivery of rAAV vectors (eg, with greater expression in the liver of male mice). ^{57 –59} Similarly, host age can play a role. For instance, AAV serotypes 9, rh8, rh10, and rh43 (where “rh” stands for “rhesus” monkeys from which these serotypes were isolated) have both neuronal and glial tropism when administered systemically; however, differences in transgene expression occur, with predominantly neuronal expression in neonatal animals and glial cell expression in adults. ^{60 –64} Other factors such as basic concepts of methods of vector production are generally less familiar to the pathologist but should still be understood in principle. ⁵³

Dose selection for GTx encompasses several additional nuances relative to other test article types. The dose of rAAV for in vivo administration is typically expressed as vector genomes (vg), given on a “per animal” (vg/animal) basis, particularly for intracranial or intrathecal (IT) ROAs; “per tissue/compartment” (vg/tissue) basis, as for intraarticular or intravitreal ROAs; or per unit weight (vg/kg) basis, as is commonly the case for intramuscular (IM) and intravenous (IV) ROAs. Depending on the animal size, ROA, and maximal feasible concentration of vector that can be achieved without causing vector aggregation, the dose can be formulated in small-volume sizes (μL) for systemic delivery in small animals or localized administration in large animals, or in large-volume sizes (mL) for systemic delivery in large animals. Dose extrapolation across species for weight-based doses (ie, vg/kg) is converted based on per kg body weight; however, tissue/compartment-based doses (ie, vg/eye or vg/brain volume) need to consider other factors such as differences in organ anatomy, tissue composition (eg, fluid or solid), and compartment volume. The tropism of rAAV may be influenced by the injected dose as well as the injection rate and volume. For local injections (brain [intraparenchymal], eye, etc), increasing the concentration of vector genomes in an injected dose increases the number of rAAV particles that can be endocytosed by all cells local to the injection site, thereby driving stronger and more focal gene expression within both intended and unintended cell populations. In contrast, lower vector genome concentrations may instead increase transduction specificity by reducing the number of vector particles endocytosed per cell. ⁵³

The choice of ROA may offer some overall benefits for in vivo GTx. For example, when targeting the brain, direct brain intraparenchymal injection offers several advantages relative to either systemic delivery or central delivery directed into the CSF. Key benefits to intraparenchymal delivery include reduced exposure of peripheral organs, thus diminishing the likelihood of peripheral toxicity, and decreased amounts of vector needed to achieve high-level expression at the desired location. ⁶⁵ Intraparenchymal injection or exposure of nerve termini (eg, brain, cornea, footpad) may have associated anterograde or retrograde axonal transport resulting in transgene expression in neurons distal to the site of exposure. ^{66 –68} Bolus administration of the same rAAV dose to mice by IT versus IV injections provides higher transduction efficiency in sensory neurons (of dorsal root and trigeminal ganglia) and central nervous system (CNS) tissues when given IT compared to IV, while comparable transduction efficiency is observed in peripheral organs for both ROAs. ⁶⁹ In contrast, GTx treatments that require systemic target organ delivery typically will be administered by the IV route to ensure wider tissue transduction.

Initial in vivo pilot studies may be considered to assess rAAV transduction efficiency, vector DNA biodistribution, and cell tropism as well as differences in response based on animal species, age, ROA, disease-associated pathology, and approximate dose level, and so on. Pathologists will be essential players in evaluating such endpoints. These studies can be performed using a reporter transgene to provide data agnostic of the therapeutic transgene to speed data collection and interpretation. Localization of the transduced cells can be assessed quickly by standard molecular biology detection protocols performed on homogenized tissues or tissue sections to detect the products that are either reporter-tagged or that are specific for transduced cells. Common reporter genes for this purpose include green fluorescent protein (GFP) and β-galactosidase (lacZ). Expression of these reporter genes is typically driven by a ubiquitous promoter (eg, CBA, CMV) to assess the broadest potential of vector biodistribution. Although the distribution of vector DNA may not be impacted by the transgene selected for these pilot studies, the abundance of transgene expression may be influenced by the efficiency of transgene mRNA and protein expression; consequently, direct extrapolation of transgene expression levels between a reporter gene and an intended therapeutic transgene should be done with caution. ⁷⁰ Additionally, some reporter transgenes express nonendogenous proteins that may also result in toxicity, immunogenicity, or both. ⁷¹ Some pilot studies may need to be conducted in large animal species such as dogs, mini-pigs, or NHPs to test novel delivery devices or anatomically constrained ROAs that are challenging or not feasible to assess in rodents due to body size limitations. The use of large animal species, more importantly, may help in assessing the potential for interspecies differences in transduction efficiency, promoter biology, and transgene expression.

Methods of vector preparation and purification may play a role in impacting the in vivo attributes of a particular rAAV (eg, tropism, biodistribution, and immunogenicity). For example, one report documented that an intraparenchymal brain injection of CsCl-purified AAV8 resulted in strong astroglial tropism of a CMV-driven GFP reporter gene, while intraparenchymal brain injection of iodixanol-purified AAV8 carrying the same CMV-GFP construct exhibited only neuronal transduction. ⁷² To our knowledge, no details have been discovered to explain the mechanism for such process-related differences in cell tropism.

Dose Selection

Pathologists may be involved in dose selection for rAAV-GTx nonclinical studies by evaluating the semiquantitative endpoints of IHC or ISH (eg, percentage positive area or cells within the tissue, possibly by image analysis programs) and evaluating the efficacy in animal models, so the pathologist’s understanding of GTx dose selection principles is important to using these data to inform dose range estimations for toxicity studies. Nonclinical dose range-finding and efficacy studies for GTx products need to define the therapeutic (ie, pharmacologically effective) dose range. ⁶⁴ The lower end of this range is referred to as the minimally effective dose (MED), which is the smallest dose with a discernable useful (efficacious) effect. The upper limit of the range is the maximum dose beyond which no additional therapeutic benefit is observed. The MED from nonclinical studies is often used to estimate the starting dose in clinical trials. Unlike other modalities where dose escalation can be done in the same individual, GTx can only be given once per patient because of the immune response that will neutralize a subsequent vector administration and sustained effect of the therapy. ²³ Consequently, for ethical reasons, human clinical trials are initiated at a dose that is believed to have the potential for therapeutic benefit. The upper end of the therapeutic dose range in nonclinical studies provides a guide for clinical development regarding where the clinical dose will need to be raised to reach the optimal biological dose (ie, the dose that offers maximal therapeutic benefit with manageable risk). The nonclinical therapeutic dose range is only a guide for clinical development as the efficacious dose range may not directly extrapolate to humans. Nonclinical toxicity studies should be designed around the therapeutic dose range and a multiple thereof and should support the safety of the starting dose and the range within which the dose can be safely raised in determining the optimal biological dose in patients.

Like other therapeutic modalities, nonclinical toxicity studies for GTx should identify the potential toxicities that will need to be monitored in clinical studies and provide an assessment of the safety margin for the anticipated dose range that will be explored in clinical studies. A typical nonclinical toxicity study for GTx will have a low dose generally equivalent to the MED in efficacy studies. The middle dose is chosen based on the maximum effective dose determined in proof-of-concept studies; however, if there are known toxicity concerns, it may be set lower to define the optimal biological dose. The high dose is chosen to produce a toxic effect or provide an adequate safety margin (eg, 5- to 10-fold) relative to the estimated maximum dose in humans. The high dose may be limited to a maximum feasible dose, which is influenced both by what volume can actually be administered to an animal and the maximum concentration at which the vector can be formulated. Depending on risk–benefit considerations for a given disease condition and toxicity findings, a narrow therapeutic index (eg, 2-fold) may be acceptable for in vivo GTx products. In some instances, pivotal toxicity studies may be performed using only 2 doses. In these cases, the low dose is selected to mirror the maximum anticipated clinical dose that will be used. Studies using only 2 doses may be performed when there is a high degree of confidence that the low dose (maximum clinical dose) has an acceptable nonclinical safety profile or the resources for a 3-dose toxicity study are lacking. Prior experience with highly similar GTx products and modes of administration can help build confidence in predicting the likely outcomes of nonclinical studies once the GTx platform is well characterized.

Initial dose selection for human trials is defined using both empirical measurements for the novel GTx test article and interspecies scaling approaches. Following initial studies in rodents (usually mice but sometimes rats), the MED can generally be translated to larger animals (typically NHPs) and then to humans using a variety of scaling approaches such as dose/body weight, dose/surface area, vector concentration and/or volume/tissue area, dose/compartment volume, and so on. ^76,77 Other factors that should be considered include differences in dose-related tropism of the vector in different species (rodents and nonrodents), the impact of the disease state on transduction, the comparative biology of the promoter, and so on. Biologically relevant principles and a well-considered scientific rationale should be applied in dose extrapolation across species and for first-in-human dose proposals. For example, administration of AAV8 expressing human coagulation Factor IX (AAV8-hFIX) in C57BL/6 mice at 4 × 10¹² vg/kg resulted in approximately 2-log higher transgene expression compared with that achieved in rhesus macaques administered the same vector at a slightly higher dose of 5 × 10¹² vg/kg. ⁷⁸ Similarly, in a separate study, administration of AAV8 expressing hFIX in C57BL/6 mice and Wistar rats at 5 × 10¹² vg/kg resulted in greater than 6-fold higher transgene expression in mice compared with rats. ⁷⁹ Consequently, if vector tropism is similar in a rodent disease model and wild-type rodent but the transduction efficiency of the vector in an NHP is 10-fold less, then this difference needs to be taken into account in designing pivotal nonclinical safety studies and estimating the clinical dose and dose escalation plan.

The maximum concentration at which a vector can be formulated without precipitation or aggregation during production and shipment may be a limiting factor when choosing high doses for a nonclinical study and subsequent clinical use. The small size of rodents may necessitate dividing the dose volume to permit simultaneous administration at multiple sites (eg, IM injection). Vector concentration ultimately impacts the strength, specificity, and distribution of transgene expression, particularly when administered locally.

When determining doses for GTx products, the method used for calculating genome copies to be delivered is critical. The PCR-based assay (qPCR or droplet digital PCR [ddPCR]) is typically utilized to measure the total viral genomes. ⁸⁰ However, PCR results can vary dramatically among laboratories, and efforts must be made to standardize and validate the nonclinical PCR methods with those that will be used to quantify viral genomes in the product destined for the clinic. Other appropriate analytical methods should be in place to ensure comparability of the vector employed in pivotal nonclinical efficacy and safety studies with that used in clinical trials. The FDA has provided guidance for this purpose that should be considered when validating a PCR assay. ³⁴

Species Selection and Animal Models of Disease

As with other therapeutic modalities, the choice of nonclinical animal species plays a significant role in the development of rAAV-GTx. ⁸¹ The species of choice is selected on a case-by-case basis as determined by animals in which the test article is biologically active and are expected to be most sensitive to expected toxicological effects. ^20,21 Animal models of disease commonly are employed for proof-of-concept studies and may serve, if the test article is pharmacologically active, as a stand-alone system for nonclinical safety testing of rAAV-GTx products. ^20,21 There are no regulatory requirements that mandate the use of 2 nonclinical species (eg, rodents and nonrodents) in developing GTx products, ^20,21 which is a substantial difference relative to the traditional 2-species approach generally required for nonclinical testing of small molecule drugs as well as nucleic acid-based products and (where relevant) large molecules (eg, protein and antibody-based therapies). The scientific rationale for the use of a particular nonclinical species, including the availability of wild-type (“normal”) versus disease models for evaluating efficacy and safety, and knowledge of rAAV biodistribution across species is often discussed in advance with the regulatory agency when a GTx product is initially being considered for development.

The animal species selected for testing an rAAV-GTx product should be pharmacologically relevant. The target cells and tissues in animals should be transducible by the rAAV-GTx vector, and the typically human-derived transgene product should be active in the animal species. Where feasible, rodents (mice or rats) are used most often due to the number of available genetic backgrounds, including engineered and spontaneous animal disease models, as well as the large catalog of reagents suitable for IHC evaluation of cell type-specific antigens (including those needed to characterize any immune response). However, this expansive set of tools for rodents does not guarantee success in predicting human responses to the test article. For example, initiating anti-rAAV T-cell immune responses in several mouse models does not reproduce the elimination of transduced hepatocytes as observed in clinical trials. ^82,83 The same holds true for other nonclinical species, including NHPs. Another key consideration is that the animal species should be immunologically tolerant of the human transgene. Lack of immune tolerance typically necessitates the use of a transgene for an animal species-specific protein (ie, surrogate molecule) or immunosuppression. The prevalence of preexisting neutralizing antibodies to the AAV serotype used in the rAAV-GTx product is an important consideration in the selection of the nonrodent species and even individual animals as it may impact transduction efficiency. ⁸⁴ That said, the loss of sustained transgene product expression or activity is more likely predicted by antibodies directed against the transgene product itself rather than by the existence of anti-AAV antibodies. ⁸⁵ Importantly, the existences of anti-AAV and antitransgene antibodies have not been correlated with the extent of the cytotoxic lymphocyte response. ⁸⁵ Toxicity of AdV GTx vectors is reportedly enhanced in human patients if a prior infection with wild-type AdV induces an immune response prior to administration of AdV-based GTx, ⁸⁶ but a similar phenomenon has not yet been demonstrated for rAAV-GTx or other viral GTx vectors.

The animal species should be anatomically suitable for vector administration by the ROA intended for use in humans. Key elements to assess in comparing procedural details include factors such as the delivery rate (eg, bolus [“rapid”] injection vs infusion), injection volume, and number of injections (“dose splitting”). In some cases, a larger animal species (eg, dog or NHP) may be necessary if a delivery device (eg, chronically implanted catheter, large-bore hypodermic needle) will be employed clinically. In such cases, a comparison of the delivery device and administration procedures used in nonclinical studies should be made to the actual clinical delivery device and delivery procedure. This comparison should include any design modifications that are needed to accommodate anatomic differences between animals and humans (eg, for CNS indications, relatively smaller IT spaces and CSF volumes in rodents vs larger IT cisterns and CSF volumes in primates and especially in people ⁸⁷ ). Such size constraints may dictate that nonclinical studies be performed in juvenile or adult animals, even for GTx products destined for pediatric use. Very young rodents at postnatal days 7 to 10 are approximately equivalent in physiologic terms to human neonates and, depending on the administration route, may be administered test article but are too small for many experimental manipulations. ^88,89 Immature NHPs (<1 year of age) are larger but the supply of such immature animals is limited, particularly when maternal transfer of antibodies needs to be considered and infants need to be derived from mothers that are seronegative to the vector that will be administered.

It is recognized that biological responses of animal disease models may diverge from those of conventional (ie, wild-type “normal”) animals of the same species. The use of an animal model of disease for safety assessment may show evidence of toxicity that does not manifest in conventional animal species or vice versa. For example, the potential for toxicity induced by the rapid release of long-stored cell breakdown products after treatment of individuals with inborn errors of metabolism recently has been demonstrated in the acid sphingomyelinase deficiency (ASMD) knockout (KO) mouse model after treatment with recombinant human acid sphingomyelinase. ⁹⁰ Treated ASMD KO mice display toxicity that is characterized by cardiovascular shock and death with liver inflammation, adrenal gland hemorrhage, and increases in serum ceramide and cytokine concentrations at rAAV vector doses 3-fold lower than those that do not elicit toxicity in wild-type mice, rats, and dogs. In contrast, overexpression of a transgene in conventional animals may result in apparent toxicity that would not manifest in species- and strain-matched animals with the disease. This possibility was recently illustrated by the development of severe neurologic signs and neuropathologic changes in cynomolgus monkeys after intracranial infusion of AAVrh8 vectors encoding α- and β-hexosaminidase transgenes. ⁹¹ The neurotoxicity was attributed to overexpression of β-N-acetyl hexosaminidase protein in healthy animals that already were expressing normal levels of hexosaminidases. In contrast, evidence of neurotoxicity was not observed in hexosaminidase-deficient mouse, cat, and sheep models of GM₂ gangliosidosis administered AAVrh8 vectors encoding species-specific hexosaminidase after direct intracranial injection. These reports indicate that care will be necessary in designing nonclinical programs for rAAV-GTx products to permit evaluation and discrimination of both routine toxicities (ie, effects related to delivery of the rAAV vector and/or transgene) and the potential consequences of exaggerated pharmacology (ie, the release of toxic byproducts following successful transgene activity).

Possible disadvantages of using animal models of disease should also be considered when designing nonclinical studies. ^30,92,93 One important consideration is the potential confounding effects of variable transgene transduction/expression and the inherent individual animal variability and divergent progression of the disease phenotype in the model. A second consideration may be the lack of full concordance of the disease phenotype in the animal models and human patients with the disease. A third drawback is the paucity of background pathology data for genetically engineered animals and species not commonly used in conventional toxicity studies, a deficit that may only be addressed through extensive phenotypic characterization with biostatistical support. A final consideration is the technical feasibility, increased costs, and extended timelines associated with generating sufficient diseased animals to perform a given study. These factors may require that safety assessment for an rAAV-GTx product be conducted in wild-type animals in parallel or instead of toxicity testing in an animal disease model and, as discussed previously, the decision on the number and types of studies should be made on a case-by-case basis that is determined by disease indication and product attributes.

Routes of Administration

The ROAs used in nonclinical studies should, whenever possible, reflect the route planned for use in the clinic. Evaluation of the preferred clinical route may not be feasible in all animal species due to size or other anatomical constraints. In such circumstances, utilizing more than one species or animal model will likely be necessary to demonstrate effective transduction of target cells using the chosen ROA. When evaluating the safety of GTx products administered locally, the tissues at the site of administration along with those of regional distribution (eg, lymph nodes via lymphatic drainage from the administration site, axonal transport) should be collected for histopathologic evaluation and biodistribution analysis to look for local expression and responses to the test article.

The target tissues and cell types of interest should be considered when choosing the ROA. For example, studies evaluating the ability of 4 different ROAs (intranasal, intratracheal, intubation, and modified intranasal) to effectively deliver pneumotropic rAAV6 vectors to respiratory tissues show that transgene expression depends on the choice of delivery method. ⁹⁴ Transgene expression is consistently visible in the nasal cavity, trachea, and all branches of lung airways for all 4 methods, whereas transgene expression is consistently observed in the most distal aspect of lung lobes (alveolar epithelial cells) only after rAAV6 vectors were introduced via the intubation and intratracheal injection techniques. ⁹⁴ In addition, rAAV vector genome copy numbers in lung tissues are approximately 4-fold lower in mice that received rAAV6 vectors via intranasal administration relative to the other 3 methods of vector delivery. ⁹⁴ Investigatory studies using reporter constructs with appropriate detection such as IF or IHC techniques to localize vector biodistribution are often used to determine targeted delivery to a particular organ or cell type during lead candidate selection. In addition to anatomical localization, the volume of injection, effect of diluent, and the position of the nonclinical species during dose administration can all play a role in determining whether a GTx product can be delivered successfully using a given ROA to a nonclinical species for safety testing.

The ROA is a major determinant of potential biodistribution and target organ toxicity. For example, systemic rAAV-GTx administration (typically by IV injection or infusion) results in the liver as a major site of transduction by several rAAV serotypes. ^95,96 Alternatively, local ROAs (eg, by intraocular, IM, and intraparenchymal brain injections) are intended to yield a high vector uptake at specific sites of disease with more constrained distribution to distant tissues. The primary benefits of localized dosing include a requirement for less vector as well as potentially decreased risk of systemic toxicity and/or immune response to the vector proteins. ⁹⁷ For example, local administration by either IM injection or isolated limb infusion via a regional blood vessel results in high uptake in skeletal muscle. ⁹⁸ Direct delivery into the CNS by either an intraparenchymal (eg, directly into a brain nucleus), IT (lumbar puncture into the subarachnoid space or into the cisterna magna), or intracerebroventricular (ICV) route primarily leads to higher transduction of brain and/or spinal cord neurons and the sensory neurons in the dorsal root ganglia (DRG). ^{99 –101} This increased CNS transduction capability reflects the direct delivery of the rAAV-GTx product via bypassing the existing blood–neural barriers that often limit or prevent neural transduction of IV-delivered agents. In addition, local delivery to more immunologically privileged sites (eg, CNS, eye) may generate a less robust systemic neutralizing antibody response, potentially allowing for repeat rAAV-GTx product dosing with a reduced need for immunosuppressive therapy. ^{102 –104} However, it is important to note that leakage from local ROA is common and often results in systemic unintended biodistribution of rAAV to remote organs. ⁹⁷ For example, vector delivered to the CSF may result in significant systemic exposure. Importantly, vector genome numbers in remote organs may be similar to or even exceed those achieved with systemic administration or in the intended local target organs. ¹⁰⁵ The ROA may also influence the location of rAAV delivery within an organ as well as the subsequent immune response. For example, when comparing ICV to intracisterna magna (ICM) inoculations of AAV9 vectors expressing GFP in dogs, both ROAs resulted in efficient transduction throughout the brain and spinal cord, but animals dosed by the ICV route developed encephalitis associated with a T-cell response to the transgene product possibly related to parenchymal exposure along the ICV injection tract. ¹⁰⁶

Biodistribution

Conventional pharmacokinetic (PK) studies may not be relevant for rAAV-GTx assessment as many transgenes encode for membrane-bound or intracellular proteins, peptides, oligonucleotides, or interfering RNAs rather than circulating products. Instead, biodistribution studies replace many of the traditional PK studies conducted for small and large circulating molecules when evaluating “exposure” to GTx products. ^20,107 The biodistribution study can be conducted as a stand-alone study, but biodistribution is often assessed in conjunction with combined efficacy and toxicity studies, where the time points for evaluation of biodistribution may be aligned with time points chosen for efficacy and toxicity evaluation in order to permit correlation of genome copy number and transgene expression levels with any pathology findings. The correlation of histopathologic findings to vector biodistribution is a common role for toxicologic pathologists who participate in nonclinical studies for GTx products.

Measuring the biodistribution of rAAV-GTx products is not as straightforward as measuring the PK profiles of small molecules and biologics. Assessments measure both viral copy numbers using a validated PCR-based assay (qPCR or ddPCR) as well as the extent of transgene expression utilizing RT-PCR for transgene mRNA, and enzyme-linked immunosorbent assay (ELISA) or liquid chromatography–mass spectrometry (LC-MS) for transgene protein. Quantification of vector amounts in circulation varies depending on whether the analyte is free vector in plasma or cell-associated vector in whole blood. In addition, different serotypes of rAAV can vary greatly in time to clearance. ^{41,108 –110} Once vector is taken up by tissues, it must undergo the processes of capsid uncoating and gene transcription within transduced cells, which, depending on the vector design (eg, ssAAV vs scAAV), can result in an additional delay in transgene expression. ⁴⁹ For example, IM administration of scAAV versus ssAAV serotype 2 vectors expressing GFP in the cranial tibial (or tibialis anterior) muscle of mice results in strong transgene expression with scAAV at 1 week postdose versus no detectable expression using ssAAV. More than 50-fold higher expression was found in the scAAV group compared with the ssAAV group at 2 weeks and maximum transgene expression at 6 weeks with scAAV but at 6 months with ssAAV. ¹¹¹ In the same study, delayed transgene expression in the liver was observed with ssAAV compared with scAAV. Species differences in receptor abundance and localization can affect the kinetics of transgene production and can influence species differences in response to the gene promoter. For example, an inverse zonation of hepatocyte transduction can be seen in mice versus NHPs transduced with AAV8-based GTx products, likely due to differences in receptor localization on target cells. ¹¹² Total protein expression and protein localization resulting from vector administration also should be measured, in addition to viral genome copies, when calculating comparable dosages for humans. If the vector-derived protein cannot be distinguished from endogenous protein, then assessing mRNA expression may be an alternative. Such determinations typically are made using quantitative procedures (eg, spectrophotometric measurements) using fresh or frozen homogenized tissue rather than by digital image analysis of tissue sections assessing the intensity of IHC or ISH signals. In some instances, knowing the proportion of the target cell population and uniformity of distribution of the vector in the target cell population can be important for assessing the potential efficacy. In these cases, image analysis to assess the distribution of an IHC or ISH signal in tissue sections can be used (Figure 3). For example, an essential efficacy endpoint in SMA disease models is to demonstrate the percentage transduction of motor neurons in the spinal cord gray matter (ie, the target cell population for Zolgensma). Such cell type-specific assessment typically is conducted either on frozen or formalin-fixed, paraffin-embedded tissue sections by conventional IHC or ISH or alternatively on cells isolated by laser capture microdissection and subsequently homogenized for PCR-based assessment to demonstrate the presence and degree of transgene expression. These assays may also serve to identify and characterize the potential toxic findings in the target cells. Pathologists are essential participants in such biodistribution assessments.

OPEN IN VIEWER

Figure 3.

Morphology-based assessment of biodistribution in a mouse following systemic delivery of a recombinant adeno-associated virus (rAAV) vector. Immunohistochemistry (IHC) demonstrating transgene expression exclusively present within (A) neurons of the myenteric plexus in the colon (20× objective); (B) distal convoluted tubular epithelium of kidney, with sparing of cells in the macula densa of the juxtaglomerular apparatus (20× objective); and (C) hepatocytes, predominantly in the centrilobular regions with sparing of cells within portal triads (20× objective).

The FDA suggests a minimum of nine tissues for evaluating biodistribution for nonclinical studies of GTx products. ³⁴ The recommended tissues include injection site(s), blood, brain, gonads, heart, kidney, liver, lung, and spleen. This minimum number may need to be supplemented if it does not include the specific target organs of interest (ie, where the transgene product is expected to exert its therapeutic effect). It may also need to be adapted to survey tissues related to the ROA by including organs such as draining lymph nodes (eg, cervical lymph nodes for intra-CNS injections) and tissues near the delivery site (eg, meninges and spinal nerve roots for IT injections, subcutis for subcutaneous injections, and vessel walls and perivascular tissues for IV injections). Although biodistribution does not necessarily predict safety, understanding the biodistribution of an rAAV-GTx product can guide the tissue selection in toxicity studies. For example, DRGs have not been collected historically in routine GLP toxicity studies, but recent experience indicates that DRG may be an important organ to evaluate for toxicity after CSF (ICM, ICV, or IT ⁹⁹ ) or IV ^100,113 administration of rAAV vectors.

At necropsy, special consideration needs to be given to tissue collection procedures for biodistribution studies in order to prevent cross-contamination among samples. Pathologists and necropsy personnel will share responsibility in ensuring this procedure is performed correctly. Animals should be exsanguinated to remove the blood and any residual circulating viral vector to allow for more accurate counts of successfully transduced vectors in organs. Procedures that minimize and ideally eliminate cross-contamination might include the use of a separate set of disposable sterile instruments for collection of each organ, or the use of reusable instruments that have been thoroughly cleaned, followed by inactivation of DNases and RNases and autoclaving prior to their use. Gloves should be changed between each animal or in the event of contamination. The organ collection order during the necropsy should be specified such that organs with low expected levels of vector are collected prior to organs with high-expected levels of vector.

Nonclinical Safety Endpoints for rAAV-GTx Studies

Ideally, animal toxicity studies should be devised to assess relevant endpoints of the planned clinical trial. The nonclinical studies are designed to identify, characterize, and quantify potential local and systemic toxicities associated with exposure to vector nucleic acid (vector DNA and transgene mRNA) and transgene protein. Toxicologic pathologists will be instrumental in this endeavor and often will generate and interpret critical data needed to define key target tissues and set the threshold values (eg, no observed adverse effect level [NOAEL]) used in risk assessment.

As with other therapeutic modalities, histopathologic evaluation coupled with clinical pathology analysis often is the “gold standard” for safety assessment in nonclinical studies for toxicity of rAAV-GTx products. ^99,100,114 Key features to evaluate include the onset of dose-related toxic events (ie, acute vs delayed), the effect of dose level on the incidence and severity of these findings, the feasibility of the proposed GTx delivery system and procedure, and the nature of the immune responses (both humoral- and cell-mediated immune reactions to capsid proteins and the transgene product). Terminal time points generally are designed to capture the time of peak transgene expression (eg, 2-6 weeks postinjection) and a later time point (eg, 13 weeks postinjection or later) to measure vector persistence and monitor any long-term effects of transgene expression or immune response to the vector or transgene proteins. Potential risk mitigation strategies, such as the use of one or more immunosuppressive agents to control anti-rAAV immune responses that may limit rAAV-GTx efficacy or promote immunotoxicity, or mechanistic studies to distinguish various aspects of anti-rAAV immunogenicity, may be prospectively included in the nonclinical safety study design or set aside for evaluation in additional studies. ¹¹⁵

The endpoints and timing of the safety evaluations will be determined on a case-by-case basis that is determined by the factors such as the disease indication and product attributes and will depend upon prior characterization of the vector presence, persistence, and clearance from target cells and tissues. Safety evaluation of rAAV-GTx products generally occurs in conjunction with biodistribution studies to permit correlations between vector presence in a tissue and evidence of toxicity. As with other biopharmaceutical modalities, in-life endpoints typically include clinical signs, physical examinations, intermittent body weight measurements, and assessment of food consumption. Safety pharmacology endpoints may also be considered (depending on the target gene function, ROA, and known biodistribution) and may require a stand-alone study or additional animal species. Endpoints examined in specimens at or after necropsy typically include routine clinical pathology analytes (Table 2) and other biomarkers, organ weights, and macroscopic (gross) and microscopic pathology observations. Collection of a comprehensive battery of tissues is recommended (Table 3). Microscopic evaluation should include the standard list of tissues plus any other tissues that are known targets of the rAAV test article; since systemic function is integrated, transgene expression in a target organ should lead to evaluation of any downstream sites under its control (eg, pituitary gland and endocrine organs, respectively). Additionally, expanded tissue collections and specialized trimming techniques may be required depending on a priori knowledge of potential toxicities in organs infrequently collected or preferential vector tropism/biodistribution in such tissues. For example, recent reports have highlighted the potential of toxicity to occur in DRG, a tissue not typically examined in routine general toxicity studies. ^{99 –101} In the authors’ experience, 4- and 13-week-long, GLP-compliant, IND-enabling nonclinical toxicity studies may be sufficient to initiate clinical trials, although other time points may be added as needed at the sponsoring institution’s discretion. Retention of some animals for 26 weeks or longer after dosing may be considered, or requested by regulatory authorities, in some cases to assess transgene persistence and potential progression or attenuation of efficacy and toxicity over time. As with other therapeutic modalities, clinical pathology analysis and safety pharmacology assessment can be used at multiple time points in nonrodent studies to detect and monitor potential biomarkers of toxicity earlier during the in-life phase. Regional health authority variations are encountered with advanced therapeutic modalities where harmonized global regulatory guidance is not yet established. Sponsors should discuss the design and duration of the nonclinical program with regulatory health authorities as the nature of the rAAV-GTx product, disease indication, patient age, unmet medical need, and overall risk:benefit considerations will influence nonclinical study design, including study duration and dose levels.

OPEN IN VIEWER

Table 2.

Recommended Clinical Pathology Sampling in Developing rAAV Test Articles for In Vivo Gene Therapy.

Hematologic evaluation	Complete blood count
	Leukocyte differential (automated)
	Platelet count
	Blood smear (cytological evaluation of cell morphology, performed if warranted)
	Bone marrow smear (cytological evaluation of cell morphology and numbers, performed if warranted)
Serum chemistry analysis (minimum list of tests)	Albumin
	Globulin
	Total protein
	Albumin/globulin (A/G) ratio
	Alkaline phosphatase (ALP)
	Alanine aminotransferase (ALT)
	Aspartate aminotransferase (AST)
	γ-Glutamyl transferase (nonrodents; GGT)
	Blood urea nitrogen (BUN)
	Creatinine
	Phosphorus
	Calcium
	Glucose
	Total bilirubin
	Triglycerides
	Total cholesterol
	Sodium
	Chloride
	Potassium
	Tissue-specific biomarkers (as warranted)—creatine kinase (CK) for skeletal muscle, etc
Coagulation	Activated partial thromboplastin time (APTT)
	Prothrombin time (PT)
	Fibrinogen
Urinalysis	Volume
	Color/clarity
	pH
	Specific gravity
	Glucose
	Protein
	Ketones
	Blood
	Bilirubin

Abbreviation: rAAV, recombinant adeno-associated virus.

OPEN IN VIEWER

Table 3.

Recommended Tissue Sampling in Developing Test Articles for rAAV-Based In Vivo Gene Therapy.

Target organs (italicized organs = minimal tissue set for exploratory studies)	Administration site a
	Adrenal glandb
	Bone marrow a (femur, with marrow exposed, or sternum)
	Brain a,b
	Dorsal root ganglia (cervical, thoracic and lumbar, multiple per segment)c
	Gonad a,b (ovary or testis)
	Heart a,b
	Intestine, large—colon
	Intestine, small—jejunum
	Kidney a,b
	Liver a,b
	Lung a
	Lymph node (mesenteric and [if warranted] draining)
	Pancreas
	Skeletal muscle
	Skin
	Spleen a,b
	Thymus
	Gross lesions a
Ancillary organs (organs comprising a comprehensive tissue list for GLP toxicity studies)	Accessory sex glands, male—prostate gland, seminal vesicle
	Aorta
	Bone (vertebrae from cervical, thoracic, and lumbar segments)
	Connective tissue—brown and white adipose tissue
	Esophagus
	Intestine, small—duodenum, ileum (with Peyer’s patch)
	Intestine, large—cecum, rectum
	Eye/optic nerve
	Gall bladder
	Ganglion (trigeminal)
	Joint (femorotibial [“knee”])
	Mammary gland
	Nerve (sciatic ± tibial)
	Pituitary gland
	Reproductive organs, female—cervix, vagina, uterus
	Reproductive organs, male—epididymis, ductus deferens
	Salivary glands—mandibular (± parotid and sublingual)
	Skeletal muscle—diaphragm, gastrocnemius
	Stomach
	Spinal cord (cervical and lumbar intumescences ± mid-thoracic)
	Thyroid gland (± parathyroid gland)
	Urinary bladder

Abbreviations: FDA, U.S. Food and Drug Administration; GLP, Good Laboratory Practice; rAAV, recombinant adeno-associated virus; STP, Society of Toxicologic Pathology.

^a Minimal list of tissues (italicized) that is recommended by the FDA for evaluating biodistribution (along with blood) of gene therapy test articles. ³⁴

^b Minimal list of organs to be weighed.

^c Dorsal root ganglia (DRG) should be collected per STP recommended best practice for peripheral nervous system sampling ¹¹⁶ in all nonclinical studies for rAAV test articles—although histopathologic evaluation may be performed at the sponsor discretion—since this organ is a sensitive target for the rAAV platform. Sacral DRG may be considered for collection as well, especially if the test article is administered by intrathecal (IT) injection into the lumbar cisterna.

Nonimmune Manifestations of rAAV-GTx Target Organ Toxicity

Target organ toxicity can be observed in any organ system, but certain target organs have become more commonly associated with AAV gene therapies. It is important, however, to first consider whether any particular finding is likely or plausibly related to the transgene biology and therefore may be considered pharmacologically mediated. Such toxicities may be interpreted as exaggerated pharmacology in nonclinical studies, and their safety implications should be considered in the context of the intended clinical use, as one would approach any other therapeutic modality. For example, exaggerated pharmacology in genetically replete animals with a gene replacement therapy intended to treat a genetic deficiency disorder should be interpreted in this specific context. On the other hand, a pattern of AAV-associated toxicities is emerging independent of the transgene products, potentially warranting a common rAAV platform/class effect such as those observed in liver and DRGs as well as immune-mediated findings (discussed in detail below). Findings may be observed at these sites following administration of the optimal efficacious dose. As noted above, the ROA may be a key determinant of target organ toxicity. Administration of rAAV-GTx by both the systemic (IV) route and direct local delivery to a specific CNS compartment (IT) can lead to transduction of hepatocytes and DRG neurons, with subsequent degeneration of hepatocytes, neurodegeneration, and axonal degeneration in the peripheral nerves, spinal nerve roots, and dorsal white matter tracts of the spinal cord. ^99,100 Consequently, local rAAV delivery does not by default preclude occurrence of toxicities in distant tissues.

The liver is frequently highly transduced by rAAV-GTx products. High-dose exposures to rAAV-GTx test articles or high transgene expression driven by a strong promoter at moderate doses may induce hepatocellular degeneration and variable degrees of single-cell or multi-cell and focal or multifocal necrosis (Figure 4). These effects are mediated by vector DNA overload per hepatocyte, toxic transgene overexpression, or immune responses against rAAV components. ^26,100,117 The extent of liver injury typically is transient and asymptomatic and is usually evident within days or weeks after rAAV-GTx administration as elevated serum levels of hepatocyte leakage enzymes (eg, alanine aminotransferase [ALT] and aspartate aminotransferase [AST]). ^{100,118 –120} That said, liver toxicity induced by high-dose rAAV-GTx administration occasionally has been linked to clinical illness in animals and human patients and can be sufficiently severe to cause death. ^121,122

OPEN IN VIEWER

Figure 4.

Recombinant adeno-associated virus (rAAV) hepatotoxicity in a mouse as shown by dose-dependent increase in cytokaryomegaly, with the additional findings of single-cell hepatocellular necrosis, mixed cell infiltrate, and increased mitoses in the high dose group. A, Saline control (20× objective). B, Low dose vector (20× objective). C, High dose vector (20× objective).

The DRGs may be highly transduced by rAAV-GTx products. Dorsal root ganglia lie outside the blood–brain barrier but appear to have direct communication with the CSF compartment and are preferentially exposed following CSF versus systemic delivery, but rAAV transduction is extensive following either central delivery or systemic administration. ^104,113 Findings in DRGs (Figure 5) may include mononuclear cell infiltration (ie, accumulation of leukocytes without damage to DRG sensory neurons) or mononuclear cell inflammation (ie, aggregation of leukocytes with damage to the DRG parenchyma); neuronal degeneration and necrosis; and/or satellite glial cell proliferation. ^113,123 Neuron numbers may or may not be decreased once the leukocyte influx recedes. In one review, the peak extent of the DRG pathology findings in NHP was reported to occur between 21 and 169 days postdose followed by reduced severity or progression by ≥180 days. ¹¹³ The occurrence of DRG toxicity at a later time point has been reported once in mice. Late-onset DRG neurodegeneration in SMA KO and wild-type mice occurred at 90 and 300 days postinjection of an rAAV GTx vector; this finding has been postulated to result from chronic transgene overexpression and progressive accumulation of the survival motor neuron protein at supraphysiological levels, related to transgene expression. ¹²⁴ In a recent report, incorporation of target sequences for miRNA 183 in AAV DNA constructs resulted in reduced transgene expression in DRGs as well as reduced DRG toxicity in NHP. ¹²⁵ In contrast, concurrent administration of steroids with vectors lacking the microRNA target sequences did not mitigate DRG toxicity. These reports suggest that DRG toxicity is primarily due to transgene overexpression, but an immune response against the vector and/or transgene product may also play a role. In-life neurological signs linked to rAAV-induced DRG pathology have generally not been observed in nonclinical or clinical studies (although in rare cases, ataxia and/or tremor has been reported following high-dose exposure in animals ^99,100,113 and in one small clinical study in amyotrophic lateral sclerosis ¹²⁶ ).

OPEN IN VIEWER

Figure 5.

Inflammatory response in lumbar dorsal root ganglia (DRG) from a cynomolgus monkey is a class-specific finding following delivery of recombinant adeno-associated virus (rAAV)-based test articles. The lesion is characterized by multifocal neuronal degeneration/necrosis (shrunken, hypereosinophilic, and fragmented cells); infiltration by macrophages, lymphocytes, and fewer plasma cells that are sometimes removing necrotic neurons (ie, neuronophagia); and proliferation of satellite glial cells (multifocal nodules surrounding or replacing damaged neurons). Treatment: single injection of rAAV-mCherry at a dose of 1.5 × 10¹⁴ genome copies via the intracisternal (intracisterna magna [ICM]) route followed by a 14-day observation period. A, Dorsal root ganglion (10× objective). B, Dorsal root ganglion (20× objective).

Transgene-related toxicities with GTx may be associated with overexpression of the transgene or result from either reduction in the expression of an endogenous gene or rapid release of a metabolic product in response to correction of an inborn error in metabolism. Such toxicities are not unique to the use of GTx as a means of correcting the results of a genetic defect and might also be observed after repeated administration of a therapeutic protein.

Immune Responses to rAAV-Based GTx Products

Despite the relatively low immunogenicity of AAV and recent advancements with rAAV-GTx products, ^25,127 the immune response to treatment remains a major limitation for both persistence of transgene expression following a single dose and the capacity for vector readministration. ^44,120 In vivo rAAV-GTx can result in immune responses against capsid proteins, the vector DNA molecule, and/or transgene product (mRNA or protein), any of which may represent a source of potential immunotoxicity. Since all these components will be recognized as “non-self” by animal species, a variable but sometimes robust innate or adaptive immune response to the vector or human-derived transgene product is an expected occurrence in nonclinical studies.

Many reported toxicities after administration of rAAV-GTx products include stimulation of the innate and adaptive immune responses. ^99,128 Affected target organs may exhibit acute inflammation or infiltration by mononuclear leukocytes, including numerous CD3+ T-cells with fewer macrophages and sometimes plasma cells. Inflammation may lead to degeneration or necrosis of transduced or bystander cells. Sentinel immune cells in target organs also may respond, leading to increased cell size (ie, hypertrophy, indicative of cell activation) and/or number (ie, reactive hyperplasia). Other common manifestations of tissue-specific immune reactions include more prominent sinusoidal macrophages (Kupffer cells) in liver, microgliosis and to a lesser extent astrogliosis in the brain and spinal cord, and increased satellite glial cellularity in DRG. ^99,101 Pathologists are instrumental in assessing tissue responses indicative of these innate and acquired immune reactions to rAAV-GTx products.

The importance of potential anti-GTx immune responses cannot be overemphasized. For instance, liver toxicity (as shown by elevated serum activities of ALT and AST) has been reported in several clinical trials with concomitant increases in interferon-γ-positive, T-cell-mediated responses and decreased transgene expression. ²⁶ This hepatotoxicity was reversible by glucocorticoid (prednisolone) administration, and the reversal coupled with rescued transgene expression implicates immune responses as potential, although inconsistently expressed, factors in the pathogenesis of rAAV-mediated liver toxicity in humans. ²⁶

Innate Immune Responses to rAAV-GTx Products

The nonspecific capability of the innate immune system to attack microbial pathogens through the recognition of pathogen-associated molecular patterns (PAMPs), including rAAV elements, constitutes the first line of host defence. ¹²⁹ Components of rAAV are recognized by 2 of the toll-like receptor (TLR) family members that are instrumental in mounting an antiviral innate immune response. The cell surface receptor TLR2 recognizes vector capsid proteins, while upon internalization into endosomes the vector DNA genome is recognized by TLR9. ¹³⁰ In mice, activation of TLR9 in plasmacytoid dendritic cells allows the adapter molecule MyD88 to initiate an activation cascade that launches an interferon type I immune response that in turn drives CD8+ cytotoxic T-cell responses as well as the B-cell-mediated generation of neutralizing antibodies to both the transgene product and the AAV capsid proteins. ¹³¹ Elements of vector design, such as high CpG (cytosine and guanine linked by a phosphate group) content ¹³² and scAAV genomes, ¹³³ may further enhance the TLR9-mediated innate immune response. Capsid recognition by TLR2 on the membranes of Kupffer cells and sinusoidal endothelial cells upregulates the production of inflammatory cytokines through activation of nuclear factor κB, as has been shown in primary human cell cultures. ¹³⁴ These molecular mechanisms help drive the initial innate immune response that is mounted against a newly introduced GTx product.

In addition to TLR2 and TLR9, emerging data indicate that rAAV transduction may trigger the innate immune response via production of additional PAMPs. ¹³⁵ For example, transgene expression at later time points following rAAV transduction activates the cytosolic double-stranded RNA sensor MDA5 in human hepatocytes grown in vitro and also in vivo in human hepatocytes engrafted into a chimeric mouse model. ¹³⁶ In vivo, AAV capsid proteins can interact with components of the complement system and enhance vector uptake by macrophages without measurably impacting transgene expression in the target cell populations. ¹³⁷ Recent data demonstrate the presence of AAV capsid-specific natural killer cells in AAV-seronegative individuals, and these have been posited to play a role in anti-AAV immune responses. ¹³⁸ Although innate immune responses to vector nucleic acid and capsid proteins may be asymptomatic, signals generated by the innate immune system likely play a major role in shaping and potentiating the subsequent adaptive immune responses to vector capsid and transgene proteins. To that end, mice deficient in complement (C) receptor 1/2 and component C3 exhibit delayed humoral immunity against AAV2 vectors, thereby resulting in significantly lower neutralizing antibody titers compared with those of wild-type mice. ¹³⁷ Similarly, 2 recent clinical trials by different sponsors for the treatment of Duchenne muscular dystrophy using a rAAV-based IV GTx product reported complement activation associated with reduced platelet and red blood cell counts as well as transient renal impairment. ^139,140 Despite the successful removal of the clinical hold following inclusion of a modified steroid regimen and a limited course of eculizumab, a humanized monoclonal antibody inhibitor of complement activation, the trial was placed on a second clinical hold after another patient was reported with a similar serious adverse event secondary to complement activation. ¹⁴¹

Adaptive Immune Responses to rAAV-GTx Products

The onset of the adaptive (acquired) immune response against viruses and GTx viral vectors is delayed compared with the innate immune response. This later-onset reaction is mounted specifically against antigenic epitopes on the viral capsid proteins and/or transgene-derived protein and is undertaken in parallel by the humoral (antibody-based, B-cell-driven) and cell-mediated (T-cell-driven) branches of the adaptive immune system. Importantly, the initiation of the acquired reaction promotes the production of long-lived memory B-cells and memory T-cells that may limit the long-term efficacy of the transgene product and potentially pose a safety risk if subsequent therapeutic attempts are made using the same GTx product (eg, one constructed with an identical rAAV serotype).

Nonclinical Assessment of Immune Responses Against rAAV-GTx Products

Assessment of humoral and cell-mediated responses to capsid and/or transgene may be useful for interpretation of findings in nonclinical studies. ²⁶ However, it is difficult to predict whether immune responses mounted against rAAV capsid and/or transgene proteins interfere with the assessment of adverse pharmacology or result in adverse immune-mediated findings after vector administration. Therefore, a risk assessment strategy for evaluating nonclinical immune responses to rAAV-GTx products must be established on a case-by-case basis that is determined by disease indication and product attributes. Appropriate samples should be collected and retained to evaluate these immune responses.

This strategy typically employs the regulatory guidance principle that the study design should “obtain appropriate samples during the course of the study, which can subsequently be analyzed when warranted to aid in interpretation of the study results.” ¹⁵⁹ In the event that immune response-related findings are observed in toxicity studies, previously collected and retained samples will aid in the interrogation of a potential immune response impacting pharmacology parameters (eg, peak expression level and persistence of transgene expression) or toxicity findings. In the authors’ experience, appropriate sampling and analysis for nonclinical efficacy and safety studies for GTx products should include retention and archiving of frozen and fixed tissues (including blood and serum) from a comprehensive list (Table 3). Pathologists are instrumental in understanding potential anti-rAAV-GTx immune responses via their roles in the analysis and interpretation of routine pathology endpoints (eg, clinical chemistry, hematology, microscopic tissue examination) and, in some cases, special immunopathology endpoints (eg, flow cytometric data, IHC, and ISH preparations).

To determine whether a nonclinical toxicity study has adequately evaluated the potential for adverse transgene pharmacology, the antibody response to the transgene-derived protein as well as the impact of such antibodies on pharmacologic activity of the transgene should be evaluated. Therefore, serum samples typically should be retained in nonclinical efficacy and toxicity studies so that the presence of antitransgene antibodies may be measured if nonclinical findings suggest the presence of an antigen–antibody complex mechanism or progressive loss of pharmacologic activity.

Cell-mediated immune responses to the rAAV capsid and transgene-derived proteins can occur in nonclinical studies (and clinical trials). Therefore, consideration should be given to retaining frozen lymphocytes, especially in nonrodents, so that the development of capsid- or transgene-specific, cell-mediated immune responses may be evaluated later if warranted. For example, PBMCs should be collected at a minimum of 2 time points: prior to vector administration (ie, baseline) and at necropsy. Paired PBMC samples allow each animal to serve as its own control. Additionally, tissue lymphocytes harvested at necropsy from freshly collected, unfixed specimens of bone marrow, lymph nodes (typically mesenteric and/or near the rAAV administration site), and spleen may aid in the identification of a cell-mediated immune response as antigen-specific responses may not always be identified in PBMCs. ²⁶

Additional testing may be utilized, where feasible and warranted, to assess the potential for adverse innate immune responses that may occur shortly after rAAV-GTx administration. For this purpose, serum may be collected to measure such biomarkers of acute-phase immune responses as complement, proinflammatory cytokines, and routine clinical chemistry endpoints of acute hepatocyte damage (eg, ALT and AST activities), while whole blood may be obtained to evaluate hematologic parameters (especially counts of neutrophils, monocytes, and other cells of the innate immune response). The endpoints should be assessed prior to (baseline) and within hours to a few days (typically 2-7) following vector administration. The assessment of the innate immune responses to rAAV may have greater importance for indications where high doses of rAAV are required for efficacy. ^100,139 In general, such additional testing is employed only to explain unexpected outcomes of prior studies rather than as a prospective part of all nonclinical studies for rAAV-GTx products. A reasonable compromise is to routinely collect and retain additional serum samples in case such mechanistic studies become desirable after the initial microscopic screening of tissues has been completed. These approaches are more commonly used in nonrodent species due to blood volume and tissue sample size limitations although inclusion of additional animals dedicated to such ancillary molecular testing may be useful in rodent studies.

Although immune responses against rAAV may occur in nonclinical species, the predictive value and translatability of immune-mediated findings in nonclinical studies should be critically evaluated as part of the risk assessment strategy. For example, the first T-cell-mediated immune response to rAAV capsid with parenchymal damage and loss of transgene expression observed in patients during a hemophilia B clinical trial was not observed in nonclinical safety studies in rodents or NHPs. ¹⁶⁰ Similarly, a dog model of hemophilia B demonstrated sustained expression of the transgene for more than 8 years without evidence of an immune response. ¹⁶¹ Moreover, several mouse models induced to mount an anti-AAV T-cell immune response failed to eliminate transduced hepatocytes. ^82,83 The interactions among rAAV components (capsid proteins and DNA molecules), transgene (mRNA and protein), and host genetics (eg, human leukocyte antigen [HLA] haplotype) are complex. Therefore, the implications of immune-mediated findings in nonclinical studies on safety assessment and translatability of animal-derived data to predicting human responses to rAAV remain a major area of investigation in developing gene therapies. ¹⁶²

Minimizing Nonclinical Immune Responses Against rAAV-GTx Products

In the risk assessment strategy for developing rAAV-GTx products, several factors should be considered as a means of mitigating potential innate and/or adaptive anti-rAAV immune responses. Key parameters include basic biological attributes (both similarities and differences) of animals and humans, procedural elements of vector design and administration, the nature of the anti-rAAV immune response, and options for modifying the extent of this immune response. These factors may be viewed as individual components when designing the development program, but in fact they act in combination to influence the type and degree of anti-rAAV immunity as well as extent and persistence of transgene expression.

Biological characteristics represent the first key element to consider in devising a nonclinical strategy for mitigating immune response to rAAV-GTx. In principle, these considerations are identical to those encountered in developing any biologic-based therapeutic. For example, the degree of sequence and conformational similarity between the human transgene-derived protein and the corresponding protein for the nonclinical animal species can impact the incidence and magnitude of the immune response. Assessment of transgene product similarity between the human and nonclinical animal species is necessary in instances where immune responses against the transgene product are specific to the nonclinical animal species and thus prevent the evaluation of potential adverse pharmacology. In general, the amino acid sequence and conformation of proteins in NHPs are most closely conserved relative to human proteins; therefore, nonclinical efficacy and safety testing of GTx test articles often proceed in NHPs only, especially when the program cannot use a different nonrodent species (eg, a mutant dog model of disease does not exist). If warranted, a surrogate transgene that will express the animal species-specific orthologous protein can be used to preclude animal-specific immune responses to the human protein, thereby permitting protein function to be assessed in a setting where long-term transgene expression permits evaluation of the potential risk posed by exaggerated pharmacology. Alternatively, immunosuppression with one or several pharmacologic agents has been used in nonclinical studies. ¹⁶³ The goal of immunosuppression is to transiently suppress the immune system in the particular host to allow the assessment of a pharmacological effect produced by an rAAV-GTx in the absence of a robust immune response. The nonclinical immunosuppression regimen is generally not being qualified for use in the clinical setting, so immunosuppressive regimens may be tailored to the nonclinical animal species as the specific experimental design but that do not have to directly transfer to the clinic. In some instances, immunomodulatory regimens are modeled in nonclinical studies to support the clinical trial design.

Second, prequalification of experimental animals may be used to mitigate anti-rAAV immune responses in nonclinical studies. The prevalence and titer of any preexisting anti-AAV immune response will impact enrollment of animals for nonclinical studies. Characterizing the antibody status (particularly the anti-rAAV neutralizing antibody titer) of nonrodent animal species is generally expected for selecting subjects that are seronegative for neutralizing antibodies. This step is crucial since seropositive individuals will potentially be resistant to full transduction and sustained transgene product levels necessary for nonclinical assessments of efficacy and safety. Prevalence of antibodies differs among animal species (for rodents and nonrodents) and among colonies from different animal suppliers (study by Wang et al ¹⁴³ and authors’ unpublished data). Consequently, understanding the prevalence of antibodies in a particular animal species and for a specific animal supplier will suggest the number of subjects that should be prescreened for the study; neutralizing antibodies to the vector capsid greater than 1:5 have been shown to abrogate transduction when a vector is administered intravenously. ^78,149,150 Finally, other aspects of the anti-AAV humoral immune response may have additional impact on the risk assessment strategy. For instance, animals and people may also have binding antibodies that do not neutralize cellular transduction by the vector. Indeed, binding antibodies recently have been shown to enhance vector uptake and transgene expression in hepatocytes. ¹⁶⁴ The role of such binding (but non-neutralizing) antibodies in altering the biology or toxicity of the vector is poorly understood. Therefore, evaluation of binding antibodies and neutralizing antibodies as separate entities may be beneficial when developing rAAV-GTx products.

Finally, procedural factors represent another major element to consider in devising a nonclinical strategy for mitigating immune responses to rAAV-GTx. The design of the vector genome can impact the immunogenicity of rAAV-GTx. For instance, high CpG content ¹³² and self-complementary rAAV genomes ¹³³ may enhance the TLR9-mediated innate immune response. The ROA can also influence the nature of the immune response. Preexisting neutralizing antibodies in blood can reduce rAAV transduction efficiency following IV and intra-articular injections where antibodies are readily available in the fluid phase. ¹⁴⁵ On the other hand, subretinal, IM, IT, and intraparenchymal brain injections have barriers that limit vector exposure to circulating neutralizing antibodies and thus typically improve transduction following GTx (reviewed in detail in the study of Masat et al ¹⁴⁵ ). However, studies cannot assume that rAAV administration to such barrier-protected sites will mitigate the impact of neutralizing antibodies on transduction efficiency. Indeed, neutralizing antibody titers in blood are strongly correlated to weak or absent transgene expression in primate retina, ⁸⁴ and neutralizing antibody titers in CSF as low as 1:1 to 1:3 may impact rAAV transduction efficiency in the CNS of dogs. ¹⁶⁵