Liver directed adeno‐associated viral vectors to treat metabolic disease

The liver is the metabolic center of the body and an ideal target for gene therapy of inherited metabolic disorders (IMDs). Adeno‐associated viral (AAV) vectors can deliver transgenes to the liver with high efficiency and specificity and a favorable safety profile. Recombinant AAV vectors contain only the transgene cassette, and their payload is converted to non‐integrating circular double‐stranded DNA episomes, which can provide stable expression from months to years. Insights from cellular studies and preclinical animal models have provided valuable information about AAV capsid serotypes with a high liver tropism. These vectors have been applied successfully in the clinic, particularly in trials for hemophilia, resulting in the first approved liver‐directed gene therapy. Lessons from ongoing clinical trials have identified key factors affecting efficacy and safety that were not readily apparent in animal models. Circumventing pre‐existing neutralizing antibodies to the AAV capsid, and mitigating adaptive immune responses to transduced cells are critical to achieving therapeutic benefit. Combining the high efficiency of AAV delivery with genome editing is a promising path to achieve more precise control of gene expression. The primary safety concern for liver gene therapy with AAV continues to be the small risk of tumorigenesis from rare vector integrations. Hepatotoxicity is a key consideration in the safety of neuromuscular gene therapies which are applied at substantially higher doses. The current knowledge base and toolkit for AAV is well developed, and poised to correct some of the most severe IMDs with liver‐directed gene therapy.


| INTRODUCTION
Currently, there are over 1450 inherited metabolic disorders (IMDs) known that result from loss of a specific enzymatic activity or transport function, or otherwise impair a particular metabolic pathway. 1 IMDs impose a severe burden of morbidity and mortality on patients and families, and many of those lack viable treatment options.Although these diseases are rare on an individual basis, they are collectively more common, occurring in as many as 1:784 new births depending on geographic location. 2Advances in newborn screening, genetics, and deep phenotyping will undoubtedly broaden the list of known IMDs and improve their diagnoses.As the number of IMD patients grows, so does the need for new therapies.After many twists and turns, the gene therapy field is reaching maturity and finally delivering lifesaving treatments to patients.][5] The liver is the metabolic hub of the body orchestrating many essential processes including amino acid, carbohydrate, peptide, lipid, nucleotide, metal, vitamin, and hormone metabolism.This organ also occupies the interface between the intestine and the rest of the body, playing a crucial role in filtration of toxins and xenobiotics entering from the portal circulation.In the opposite direction, the liver is the gatekeeper for biliary excretion of important metabolites including cholesterol, bile acids, bilirubin, and metal ions. 6The majority of proteins secreted into the bloodstream are also generated by the liver, which control everything from coagulation and fibrinolysis to lipid transport.Because of its central role in whole body metabolism, the liver is a critical target for treating IMDs.This organ is also particularly amenable to gene delivery with AAV vectors, which could provide lifelong expression of therapeutic proteins.Here we provide a brief overview about AAV vectorology, and highlight important factors that govern its effectiveness as a human gene therapy vector.

| AAV AS A VIRUS AND GENE DELIVERY SYSTEM
AAV is a dependovirus that was originally isolated as a contaminant in adenovirus preps in 1965, 7,8 hence the name adeno-associated virus.Wild type AAV cannot complete its own life cycle without a helper virus, and is not known to cause any specific pathology in humans.The AAV consists of a very small 25 nm nonenveloped protein capsid that encloses a 4.7 kb singlestranded deoxyribonucleic acid (DNA) genome. 9airpin-like inverted terminal repeats (ITRs) are present on both sides of the AAV genome, which play critical roles in virus replication and gene expression. 10he AAV genome contains two major genes Rep and Cap, and two other shorter open reading frames, the assembly activating protein (AAP) and membraneassociated AAV protein (MAAP) within Cap known as accessory proteins (Figure 1A).The Rep gene encodes four proteins involved in viral replication, transcription, and integration into the genome (Rep78/68/52/40). 11Cap encodes the three capsid viral proteins (VP) VP1, VP2, and VP3 that come together in a 1:1:10 molar ratio to form the 60 subunit icosahedral protein capsid of the virus 12 (Figure 1B).AAP promotes assembly of AAV capsid VP subunits in the nucleus. 13MAAP is a recently discovered protein also encoded within the Cap gene that facilitates AAV replication and is involved in controlling adenoviral infection. 14ecombinant AAV (rAAV) vectors can be produced by supplying the wild type AAV genes (Rep/Cap) and adenoviral genes required for packaging on different plasmids. 15These recombinant AAV particles contain no viral genes-only the transgene cassette flanked by the 145 nucleotide hairpin-like ITR structures (Figure 1C).In principle, any sequence can be packaged, provided it fits within the roughly 4.9 kb size between ITRs inclusive.In almost all cases, the Rep gene and ITR sequences from AAV2 are supplied, which can be used to package any other AAV serotype.AAV serotypes are named by the sequence of the Cap gene, which defines the structure of the capsid.Naturally occurring serotypes include: AAV1, AAV2, AAV3b, AAV4, AAV5, AAV6, AAV7, AAV8, AAV9, and others. 16There are also a wide variety of custom AAV capsids available that have been engineered to acquire new tissue targeting (tropism) 17 or immune evasion. 18Engineered AAV serotypes can be produced through a number of different approaches including rational design, 19 PCR shuffling, 20 mutagenic PCR, 21 library screens, 22 peptide display, 23 biopanning, 24 and directed evolution. 25,26While not the subject of this review, recombinant AAV vectors can be produced and purified at scale based on their physical properties using ion exchange, or affinity chromatography, and/or density gradient ultracentrifugation. 27

| TRANSDUCTION OF LIVER WITH RAAV VECTORS
Transduction refers to the entire process of delivery of the AAV genetic cargo to the cell, including initial receptor binding through functional transgene expression (Figure 2).The first step in this process involves the AAV virion interacting with receptors and co-receptors at the cell surface. 28Following binding, AAV particles are endocytosed and trafficked to an endosomal compartment.Upon endosome acidification, the N-terminal region of the VP1 capsid subunits become exposed, revealing a phospholipase A2 catalytic domain which facilitates endosomal escape through hydrolysis of membrane phospholipids. 29Next, basic clusters in VP1 and VP2 subunits act as nuclear localization sequences to direct the AAV capsid into the nucleus. 29Inside the nucleus, the AAV capsid uncoats releasing the single-stranded DNA (ssDNA) genome. 30The ssDNA genome is converted to a double-stranded intermediate by the host cell's DNA repair machinery, which then circularizes via the ITRs. 31hese circular AAV genomes are known as "episomes," a stable non-integrating form of the viral cargo that can exist as circular monomers or concatamers.These AAV episomes express therapeutic transgenes and persist for decades.
Nearly all AAV serotypes have a strong tropism for hepatocytes when injected into the systemic circulation.Initial AAV binding to the cell surface is mediated by interaction with molecular patterns, which are generally glycans (i.e., heparan sulfate for AAV2, N-linked sialic acid for AAV1/5/6, and terminal galactose for AAV9).For some AAV serotypes specific proteins have been identified as primary receptors such as αVβ5 integrin 32 and fibroblast growth factor receptor 33 for AAV2, epidermal growth factor receptor (EGFR) 34 for AAV6, plateletderived growth factor receptor (PDGFR) 35 for AAV5, and laminin receptor (LamR) 36 for AAV2,3,9.Since cognate glycans are present on a variety of plasma membrane proteins in different tissues, they have a tremendous impact on AAV tropism.Following the initial binding event, a secondary receptor promotes endocytosis and uptake of AAV particles into the cell.Most naturally occurring AAV serotypes, except for AAV4 and AAVrh32.33 37 utilize a recently identified common receptor called AAV receptor (AAVR) (KIAA0319L) 38 to enter cells.AAVR binds to conserved regions in the AAV capsid through Iglike polycystic kidney disease (PKD) domains in its extracellular regions, and facilitates endocytosis.
In mice, many AAV serotypes can target the liver with high efficiency, particularly AAV8, AAV7 and AAV9. 39However, this is not universal and may be very different from one species to another.While AAV8 has been used extensively for human liver gene therapy, it may not be the best serotype for human hepatocytes. 40tudies in human liver chimeric mice have shown that AAV3B 41 and the engineered variants AAV-LK-03 42 and AAV-NP-59 43 have dramatically improved selectivity for human hepatocytes.Conversely, AAV3B performs poorly in mouse liver.This creates challenges in designing preclinical studies to support investigational new drugs.Notwithstanding, several AAV serotypes have been very successful in clinical trials for liver gene therapy in humans-most notably the recent FDA approval of an AAV5 vector, etranacogene dezaparvovec (Hemgenix), for hemophilia B.

| SEX DIFFERENCES IN AAV TRANSDUCTION
Sex differences in efficiency of liver transduction with AAV vectors could have a major impact on their translation to human gene therapy.5][46] This appears to be a general phenomenon that is applicable to most, if not all, naturally occurring AAV serotypes.These sex differences are independent of mouse strain, age of injection, serotype, route of administration, dose, or the promoter used (Table 1).Some evidence indicates that injecting mice with high doses of AAV 47 or injecting AAV intrahepatically (IH) 46 are able to mitigate the impact of these sex differences by increasing overall hepatocyte transduction.These differences in AAV transduction impact not only transgene expression levels, but also genome editing efficiency. 48,49Even within the liver, there have also been regional sex differences reported in hepatocyte transduction with respect to zonation of the liver.Using AAV8 encoding GFP, Dane et al. found that both sexes had perivenous (around the central vein) GFP expression initially (2 weeks post-injection). 50owever, at later time points (up to 12 months) livers from male mice showed periportal GFP expression while females retained perivenous GFP expression.They concluded that the distribution differences observed between sexes arose from sexual dimorphism in spatial and temporal hepatocellular proliferation.
In terms of mechanism, Davidoff et al. first reported that androgens play an important role in AAV hepatic F I G U R E 2 Transduction of hepatocytes with recombinant AAV vectors.The basic steps of AAV transduction begin with binding of the capsid to primary glycan receptors, followed by AAVR interaction for most serotypes (1).The AAV virion is endocytosed (2) and trafficked to acidifying endosomes, where the phospholipase domain of VP1 mediates its escape from the lipid bilayer (3).AAV virions that escape proteosomal degradation are imported into the nucleus, where the capsid uncoats and releases the single-stranded DNA genome (4).The AAV genome is converted to circular double-stranded episomes by the host cell's DNA repair machinery.AAV episomal genomes support transcription (5) and protein expression of the therapeutic transgene (6) T A B L E 1 AAV Sex differences reported in mouse Liver.GC/mouse M > F transduction in mice. 45They showed that castrating mice prior to AAV injection significantly reduced liver transgene levels, while oophorectomizing mice resulted in no change in transduction.Moreover, administering the androgen 5α dihydrotestosterone (DHT) to oophorectomized mice 2 weeks prior to AAV injection led to increased liver transduction and transgene protein levels.Notably, although data was not reported, DHT administration did not increase transgene expression when given at time of injection or after.When comparing nuclear extracts from male and female mice, they found there were more hepatic nuclear proteins bound to the repbinding site of ITRs from male mice.Nietupski et al. replicated Davidoff et al.'s results by administering DHT to female mice, which increased circulating transgene protein to levels seen in male mice. 51In contrast with Davidoff's results, Nietupski's work showed that administering DHT post AAV injection could increase transgene protein levels, suggesting that androgen related proteins may influence transgene expression in mice.

Age
Most reported sex differences in AAV transduction involve the liver.However, there are cases in which some tissues transduce better in female mice.Brain has been reported to transduce better in females than males when AAV9 is injected systemically 52,53 or has at least trended that way. 54Lungs, although transduced at similar levels, 45 were reported to express significantly more recombinant protein in females 55 (Table 2).It is worth noting that the liver is a sink for AAV vectors of most serotypes, and that differences in transduction of peripheral organs by non-hepatic tissues may occur as a secondary effect of altered liver uptake.In the same way that capsid modifications can de-target hepatocytes, 56 allowing for improved delivery to non-hepatic tissues, so also there may be examples of improved transduction of other non-liver tissues in female mice relative to males.This merits further investigation.
There is minimal data available on potential sex differences in AAV transduction in humans and nonhuman primates (NHP).Binny et al. did not find differences in the amount of AAV5 in livers from rhesus macaques (Macaca mulatta) 4 weeks post-intravenous (IV) injection by southern blotting. 57Results from Pañeda et al. 58 coincided, and they determined that liver transgene genome copies and transgene expression to be similar between sexes.Nietupski et al. determined that juvenile male rhesus macaques, who have undetectable levels of testosterone, only experienced a minor effect on circulating transgene protein levels after administration of DHT. 51Even less information is available in humans due to ethical and logistical challenges in obtaining liver biopsies following AAV administration.The potentially richest datasets to examine AAV transduction in human
liver would be derived from clinical trials for hemophilia A and B, where these secreted coagulation factors could be tracked in the bloodstream over time.Unfortunately, these are both X-linked recessive genetic diseases that mainly affect males, precluding a comparison by sex. 59y analogy to mice, it seems likely that sexual dimorphism in cell surface receptors or nuclear proteins could have an important impact on AAV liver transduction in humans.This should be considered moving forward in human clinical trial data design and analysis.

| LIVER AAV GENE THERAPY
Hemophilia A and B are X-linked recessive hereditary bleeding disorders resulting from deficiency of the secreted coagulation Factors VIII and IX respectively.In both cases, standard of care involves expensive lifelong treatment with infusions of recombinant factor concentrate to reduce the risk of serious bleeding events.Hemophilia A and B have been pursued aggressively with AAV gene therapy due to the secretory capacity of the liver, the ability of AAV vectors to transduce this organ, and the low threshold of correction required.AAV gene therapy for hemophilia B in particular, has paved the way for other diseases, identifying some of the major challenges to address.The first liver-directed gene therapy for hemophilia B was initiated in 2006 using an AAV2 vector, achieving detectable FIX expression for about 2 months with a favorable long-term safety profile. 60The decline in FIX levels was accompanied by a transient elevation of liver transaminase levels, which coincided with destruction of transduced hepatocytes by cytotoxic T-cells.A subsequent trial in 2011 used AAV8 delivery of FIX with a liver-specific promoter, a capsid serotype with more efficient transduction of human liver and a lower frequency of pre-existing immunity.It was found that administration of short-term immunosuppression with glucocorticoids could normalize liver enzymes, and allowed for sustained FIX levels in the range of 3%-11%. 61Long-term monitoring of these patients continues, and many continue to have FIX expression beyond 8 years after therapy. 62Further improvements in vector design include a gain-of-function Factor IX variant (R338L Padua), the use of capsids with a lower seroprevalence (i.e., AAV5, AAV6) or higher transduction efficiency of human hepatocytes (AAVLK03, AAVhu37). 63ombined with higher doses in the range of 2-6 Â 10 13 genome copies/kg, remarkable results have been achieved.This has ultimately led to the recent FDA approval of Hemgenix, an AAV5 vector for hemophilia B. 64 Hemophilia A gene therapy is also progressing well, 65 although further work is needed to ensure durable levels of expression for this factor.Note: Greater than sign ">" indicates better transduction in one sex compared to the other.The promoter region, the first exon, and the first intron of chicken beta-actin gene, and the splice acceptor of the rabbit betaglobin gene.Abbreviations: CAGG, cytomegalovirus (CMV) early enhancer element; GC, genome copies; I.T., intratracheal; I.V., intravascular; S.G., salivary gland; T.V., tail vein; w, weeks.
In parallel, and informed by the experience with hemophilia, many AAV gene therapy trials are currently in progress for inherited metabolic diseases of the liver.A representative, but not exhaustive list is shown in Table 3.Like hemophilia, several of these diseases require secretion of the defective protein by hepatocytes: hereditary agioedema and Alpha 1-antitrypsin deficiency (AATD).Some of these diseases are metabolic disorders with cell autonomous defects within hepatocytes that may indirectly affect the rest of the body: glycogen storage disease type Ia, acute intermittent porphyria, Crigler-Najjar Syndrome, ornithine transcarbamylase deficiency (OCTD), Wilson disease, and homozygous familial hypercholesterolemia.In addition, there are numerous multiorgan inherited metabolic disorders being pursued, such as lysosomal storage diseases (LSDs), where restoration in liver would be expected to substantially improve disease progression and quality of life: glycogen storage disease type II, mucopolysaccharidosis type I, type II, type IIIA, type VI, and Fabry disease.LSDs are not liver specific conditions, but the liver can be used to express the corrected form of the enzyme and indirectly reduce accumulation of harmful substrates in other tissues.It should also be noted that a small fraction of these enzymes can be secreted, and then taken up by peripheral tissues and act in the lysosome of those cells. 66Addition of signal peptides and uptake signals to artificially boost secretion of lysosomal storage enzyme transgenes is a promising avenue to promote correction outside of liver. 67Nonetheless, addressing the neurological component of these diseases also requires the use of vectors that can cross the blood-brain barrier, 68 such as high dose AAV9 or AAVrh.10. 69Further improvements in transgene and vector design, capsid choice, and management of the immune response will likely also make these therapies a reality.
Many liver metabolic diseases are expected to have a higher threshold of correction than hemophilia B-which is as low as 1%-5% of normal protein. 62The threshold of correction requires consideration of both transgene copy number and expression per cell, as well as the fraction of hepatocytes that must express the corrected protein.These are related parameters that cannot be titrated completely independently with AAV delivery.The exact threshold varies depending on IMD and the underlying biology.For example, it has been suggested that 15%-20% hepatocyte correction could be sufficient to elevate AAT to therapeutic levels, but even less might prevent lung emphysema seen in AATD. 70Hepatocyte transduction of 10%-15% is deemed sufficient to improve disease burden in an OTC mouse model. 71,72For a small subset of diseases, hepatocyte viability is dramatically improved with expression of the therapeutic protein, offering a "selective advantage" to the corrected cells.Perhaps the most extreme example is hereditary tyrosinemia type I, which results in liver failure and early death due to accumulation of toxic tyrosine catabolites.Cell transplantation experiments in fumarylacetoacetate hydrolase deficient mice have shown that replacing as few as 1000 cells is sufficient to repopulate the liver and rescue the animals. 735][76][77] Despite the low degree of correction needed, residual liver cancer risk in tyrosinemia, and the availability of an approved drug nitisinone, has discouraged clinical development of AAV gene therapy.A selective advantage for corrected hepatocytes also exists for alpha1-antitrypsin deficiency, 78 Wilson disease, 79 methylmalonic academia, 80 and likely a few others.Strategies for these diseases would ideally use gene editing or targeted genomic integration of the AAVdelivered transgene, so that corrected cells can divide and repopulate the liver.
The interior space of the AAV capsid is essentially identical for all serotypes, 81 and is able to accommodate approximately 4.9 kb of DNA including ITRs.This restricts which IMD can be treated given that some genes exceed packaging capacity.However, different strategies have been developed to circumvent this problem such as minigene versions that retain enzymatic activity. 82Dual AAV systems can also be used to deliver the gene in two parts.These two parts can come together through transsplicing mRNA across ITRs, 83 reconstitution at the level of episomes by overlapping homology regions (HR), 84 or at the protein level through self-catalytic inteins. 85However, dual AAV strategies are often ineffective in practice due to low trans-splicing efficiency, and the need for two AAVs to transduce the same cell.As an example, inefficient co-delivery might have negatively impacted a Zinc Finger editing trial with AAV which required three individual AAVs per cell. 86n the future, AAV gene therapy may also be applied to more common liver diseases such as non-alcoholic steatohepatitis, viral hepatitis, cirrhosis, 87 and hepatocellular carcinoma (HCC). 88Taking the specific case of cirrhosis, there has already been substantial work in mouse and rat models of fibrotic liver disease.These include carbon tetrachloride (CCl 4 ) injury, [87][88][89][90][91][92][93] diethylnitrosamine (DEN) injection, 94 bile duct ligation, [95][96][97] and dietary models. 95,96,98While these models do not fully represent all aspects of human cirrhosis, they can provide valuable proof-of-concept for therapeutic strategies.There is a wealth of data showing that AAV delivery of microRNAs, [99][100][101] transcription factors, 90,93,102 and secreted proteins 87,97 to hepatocytes can prevent fibrosis development or progression.However, the more clinically relevant targets for treating pre-existing cirrhosis generally center on activated hepatic stellate cells (HSC).Some of the most promising strategies to target HSCs include inhibition of extracellular matrix deposition, arresting HSC proliferation, and driving HSC apoptosis. 103Perhaps the most exciting of all is the prospect of forced trans-differentiation of HSC back into hepatocytes with AAV6-mediated delivery of transcription factors. 102It should be noted that while highly efficient at delivery to hepatocytes, HSCs are poorly targeted by most AAV serotypes, even in rodent models.The feasibility of delivery to human HSC for gene therapy remains an open question.Far more basic and preclinical work is needed to identify the best molecular targets for these common liver diseases.It also must be done in a manner that does not impart risk of cancer or further liver injury.But ultimately, a one-time treatment would be favored over lifelong immunosuppression and the risks of graft rejection following liver transplantation.Even for hepatocyte-directed vectors, the efficacy and safety of AAV gene therapy can be significantly impacted by the health, architecture, and immune status of the affected liver.Diseases with high rates of hepatocyte turnover are vulnerable to loss of episomal AAV genomes over time, and may have a higher underlying cancer risk.Likewise, disease-associated liver dysfunction may exacerbate the risk of AAV-mediated liver injury, as has been suggested in the deaths of X-linked myotubular myopathy patients in the ASPIRO trial. 104Animals studies indicate that fibrosis can impact liver transduction.Cirrhotic livers caused by cholestatic liver disease blunted AAV transduction compared to disease-free mice. 105Similar results in rats showed that cirrhotic livers, induced by carbon tetrachloride (CCl 4 ) administration, reduced liver AAV transduction when injected intra-portal route, but cirrhosis increased hepatocyte transduction when injected intra-arterial route. 106More work is needed in this area, but it illustrates the importance of considering overall liver health in the risk/benefit equation for liverdirected gene therapy.

| PRE-EXISTING IMMUNITY TO AAV CAPSID
Even though it does not cause pathology in healthy humans, many of us have been previously infected by wild type AAV.For this reason, there is a high frequency of pre-existing immunity to AAV in the population, which may include both anti-capsid antibodies as well as memory T-cells.Prevalence of anti-AAV antibodies in humans varies by serotype.A global study estimated seroprevalence in people with hemophilia A to range CHUECOS and LAGOR from a high of 58.8% for AAV2 to a low of 34.8% for AAV5. 107However, these ranges have been reported to be higher among racial minorities in the US with >90% seroprevalence against AAV3B among black healthy donors. 108Due to conserved epitopes and the fundamentally similar structure of the AAV capsid, there is a great deal of cross-reactivity of antibodies for different serotypes.Even so, some serotypes have a much greater prevalence in general (i.e., AAV2 and AAV8) versus others (AAV5), which has in some cases been an important factor deciding their use in the clinic.Antibodies to the AAV capsid may be classified as "binding" if they recognize the capsid, or "neutralizing" if they bind in a way that prevents transduction. 109Neutralizing antibodies are a major impediment to AAV gene therapy because they prevent the virus from initially entering the patient's hepatocytes rendering the therapy ineffective.For this reason, a high neutralizing antibody titer is generally an exclusion criteria for clinical trials, often eliminating half of potential patients for these already rare disorders.Even in cases where neutralizing antibodies cannot be readily identified, some patients may still have pre-existing immunity in the form of T-cells.It has also been reported that approximately half of healthy adults have detectable levels of AAV capsid-specific CD8+ or CD4+ T-cells. 110rescreening for T-cell reactivity by enzyme-linked immunosorbent spot (ELISPOT) and other approaches prior to gene therapy is difficult, and is generally not performed in favor of simpler and more quantitative neutralizing or total antibody assays.Although anti-AAV capsid T-cells do not block the initial transduction event, they can mediate elimination of the transduced hepatocytes following therapy, having a profound effect on efficacy. 111It should also be noted that even in naive subjects, administration of AAV gene therapy effectively immunizes them against most AAV serotypes.In some cases, it has been demonstrated that redosing is possible using divergent serotypes 112 or with immunosuppression 113 in animal models, but this is the exception rather than the rule.

| STRATEGIES TO MANAGE IMMUNE RESPONSE IN AAV GENE THERAPY
Patients with rare liver metabolic disorders do not have the luxury of avoiding natural AAV infection and the immunity it creates.Accordingly, there is a compelling need to find approaches that remove or eliminate neutralizing antibodies so these patients can receive a life changing gene therapy with AAV.There have been many attempts to mutagenize the AAV capsid to prevent recognition by neutralizing antibodies. 114While there are many successful examples in animal models versus specific antibodies, each individual's antibody repertoire is unique.In practical terms, it is not realistically possible to mutate enough exposed epitopes on the AAV surface while still preserving transduction efficiency.Therefore, most clinical strategies are instead focused on removal of the antibodies from the patients' serum.Several approaches exist including infusion with decoy empty AAV capsids, 115 plasmapheresis, 116 hemapheresis, 117 or immunodepletion or immunoabsorption of patient serum. 118,119Although conceptually sound, these approaches will be difficult to implement at scale in humans and present significant logistical hurdles.A recent and exciting approach involves in vivo depletion of total IgG using infusion of IgG degrading enzymes such as IdeS 120 and IdeZ. 121Both of these enzymes have been demonstrated to enable efficient transduction in animal models with pre-existing immunity to the AAV capsid.In addition, there is clinical precedent for using IdeS to treat graft rejection and heparin-induced thrombocytopenia. 122lthough there has been great progress, it remains to be determined which of these approaches will be effective and clinically translatable.
Beyond achieving delivery to hepatocytes, it is critical that the adaptive immune response to the AAV capsid be carefully managed.This may also apply to anti-transgene immunity.However, it is currently difficult to know the extent of the problem of anti-transgene immunity because many patients in gene therapy trials already have immunity to transgene resulting from coagulation factor infusions or enzyme replacement therapy.In terms of cellular immune responses to the AAV capsid, there are multiple steps that can be targeted.Preventing the initial presentation of capsid peptides by professional antigen presenting cells is a worthwhile place to begin that is a new focus for the field.Immune cell-targeted nanoparticles containing immunosuppressive drugs 123 are being incorporated into the design of new clinical trials.Broad immunosuppression with prednisone has now become a standard expectation for AAV clinical trials and for dosing of approved products.In many cases prednisone may be given at initial dosing, and then as needed when transaminase elevations are detected.In many patients this is sufficient to preserve adequate transgene expression for Factor IX. 62 It remains to be determined how effective prednisone will be for AAV gene therapy of metabolic liver diseases which require transduction of a much larger fraction of hepatocytes.Much work remains to better understand the cellular immune responses to AAV capsid and transgene, so that more effective and targeted immunosuppression can be employed.Nonetheless, the recent approval of Hemgenix 64 bodes very well of AAV gene therapy of other severe liver disorders.EXPRESSION The ultimate goal of gene therapy is to achieve lifelong correction of disease.Although there have been significant successes for AAV gene therapy over the past several years, it is unlikely we have reached that milestone.Even with appropriate use of immunosuppression, it remains to be determined whether FIX expression and protection from bleeding events will persist decades after gene therapy.Persistence of expression is an even bigger question for gene therapies involving non-secreted proteins for metabolic diseases, where a much larger fraction of hepatocytes must be targeted.5][126] This creates additional challenges for gene therapy of pediatric patients with rapidly growing livers, diseases with high hepatocyte turnover, and patients with immune responses to transduced cells.In these situations, loss of AAV episomes with cell division is of critical concern.A potential solution could be generating AAV vectors containing sequences that function as an origin of replication for mammalian cells.Hagedorn et al. have provided proof of concept for this approach, by including a scaffold/ matrix attachment region (S/MAR) into the AAV genome, which allowed them to maintain transgene expression >50 population doublings in HeLa cells. 127argeted transgene insertion is another popular approach which can be accomplished with promoterless AAV cassettes, leveraging regulatory elements of highly expressed genes in the liver such as albumin (ALB) 128 or apolipoprotein A1 (APOA1). 74Targeting highly expressed loci with Zinc Fingers or CRISPR/Cas9 can dramatically increase the ratio of homology directed repair (HDR) events. 74,80,129,130This formed the foundation for three clinical trials by Sangamo therapeutics for hemophilia B, Mucopolysaccharoidosis I and II (MPS I and II). 86In these trials, AAV transgene cassettes were co-delivered with two other AAV vectors encoding Zinc Finger Nucleases (ZFN) for integration into intron 1 of the ALB gene.Although the therapy was deemed safe, clinically meaningful expression of the transgenes was not achieved.Liver biopsies confirmed the presence of an albumin-transgene fusion transcripts in one MPS I and two MPS II patients. 86These were a major milestone as the first in vivo genome editing trials for the liver.It is likely that inefficient co-transduction with three AAV vectors, as well as homologous recombination, resulted in an insufficient percentage of correctly targeted hepatocytes.Nonetheless, this is a promising strategy which would ensure the transgene is passed on to daughter cells upon hepatocyte division or liver growth.
There has been great interest in AAV vectors as a delivery system for genome editing reagents.AAV vectors have been applied extensively for somatic genome editing to model and treat liver diseases in mice 49,[131][132][133][134] and non-human primates. 135Although highly efficient at disrupting genes in the liver, some limitations of this approach have become apparent.Generating double strand breaks with genome editing nucleases such as CRISPR/Cas9 leads to a high frequency of AAV vector genome insertions through non-homologous end joining (NHEJ) repair.These insertions are mediated primarily by the recombinogenic ITR sequences, and may include ITR integrations, truncated or whole AAV genomes, and even concatamerized AAV genomes. 136These NHEJ insertions compete with correctly repaired alleles through HDR, and can also insert permanent copies of the Cas9 transgene.AAV delivery of Cas9 results in sustained expression of this bacterially-derived nuclease, increasing the likelihood of offtarget editing and unwanted anti-Cas9 immune responses.Li et al. showed that in mice with pre-existing immunity to Cas9, liver-directed genome editing with AAV-CRISPR was efficient.However, edited cells were eliminated by a CD8+ T-cell response between 6 and 12 weeks after vector administration. 137Similar effects have been observed in the muscle of three dog models of Duchenne muscular dystrophy (DMD), a species which acquire anti-Cas9 immunity through environmental exposure to bacteria. 138iven that many humans have pre-existing immunity to Cas9, 139,140 likely similar to that for the AAV capsid, it is important to develop alternative strategies for transient delivery.Lipid nanoparticle (LNP) delivery of CRISPR/ Cas9 or base editors has proven particularly efficient in the liver.Disruption of transthyretin (TTR) in cases of TTR amyloidosis has shown impressive efficacy and safety. 141Likewise, efforts to disrupt PCSK9 through LNP delivery of a base editor have strong preclinical data in non-human primates, supporting the recent initiation of a Phase I trial in New Zealand and the UK (NCT05398029).Nonetheless, precise repair of genes is far more complicated.For this purpose, AAV still has tremendous promise as a delivery vehicle for donor templates encoding therapeutic transgenes.The ability to shuttle donor templates directly to the nucleus in non-dividing cells is a major strength of AAV vectors for genome editing.In the years to come, we can expect to see more hybrid gene editing platforms which combine the strengths of LNP and AAV technologies to treat liver disease.

| SAFETY
In May 2020, the first death of a AAV gene therapy trial participant occurred in the ASPIRO trial for X-linked myotubular myopathy.This tragedy sent shockwaves through the gene therapy field, the first sign of AAV treatment-related hepatoxicity following an otherwise nearly perfect clinical safety record.To date, a total of four patients in the trial have died due to complications from liver failure, following AAV8 delivery of an MTM1 transgene. 142Most recently two children who were treated with the FDA-approved AAV9 vector for Spinal Muscular Atrophy onasemnogene abeparvovec have also died. 143In addition to liver injury, high doses of AAV can activate the complement system, resulting in potentially life-threatening thrombotic microangiopathy (TMA) which is what occurred with these two children. 144TMA have occurred in patients in gene therapy trials for DMD, 145 and methylmalonic acidemia (MMA).It appears that the risk of TMA probably increases with ascending dose.Doses used for these neuromuscular gene therapies were high (1-4 Â 10 14 genome copies/kg), and although doses used in the MMA trial were lower (5 Â 10 13 genome copies/kg), therapy was done in a younger age group (6 months to 2 years old), making comparisons difficult.Although a rare adverse event, further work is needed to understand the mechanisms underlying AAV activation of the complement system, and how to predict and mitigate TMA.
rAAV vectors have proven to be remarkably safe for liver-directed gene therapy.However, there is a risk of tumorigenesis through unintended off-target integration of AAV vectors.Although the vast majority of AAV genomes are episomal, rare integrations can occur, and are becoming easier to detect with advances in sequencing technology.The first hint that recombinant AAV could be tumorigenic in the liver came from a study by Donsante et al., 146 who found a high incidence of hepatocellular carcinoma in neonatal mice treated with AAV vectors.Integrations expanded in the tumors were found to occur primarily in the Rian locus, which expresses several microRNAs and small nucleolar RNAs. 147This prompted multiple independent investigations into the risk for AAV-mediated genotoxicity and tumorigenesis.Most studies found no increased risk in adult mice, 148,149 or did not attribute the risk specifically to the vector. 150eminal work by Chandler et al. found that AAV has a propensity to integrate into highly expressed genes in the liver. 151Integrations into the Rian locus were associated with hepatocellular carcinoma, an effect that was dependent on age of delivery, dosage, and promoter strength.The current consensus in the field is that rAAV generally do not substantially increase cancer risk in adult mice.A long-term 10 year study of AAV gene therapy for hemophilia in dogs recently identified clonal expansion of transduced hepatocytes following AAV8 and AAV9 delivery. 152 large number of these integration events occurred near genes involved in cell growth, and correlated with later increases in Factor VIII levels in a couple animals.However, none of the dogs developed tumors.
In 2015, Nault et al. found clonal integration in humans of wild type AAV2 in 11 out of 193 human hepatocellular carcinoma biopsies. 153Many integrations were found in or near cancer driver genes (CCNA2, TERT, CCNE1, TNFSF10, KMT2B).While it is certainly possible that AAV integration was the tumorigenic event in these rare cases, the correct interpretation of these findings is debated. 154,155It is also possible that these AAV integrations are passenger events rather than drivers.The fact that as much as 90% of the human population is seropositive for AAV2, would argue against a major causative role for wild type AAV infection in hepatocellular carcinoma. 156Wild type AAV infection of the liver was also recently implicated in an outbreak of acute hepatitis of unknown etiology in children beginning in April 2022, suggesting AAV may be a primary pathogen in this disease.Many experts disagree with this possibility, suggesting instead that AAV genomes are simply a more permanent biomarker of adenoviral infection. 145rAAV used in AAV gene therapy differs greatly from the wild type AAV in these previous reports.AAV gene therapy vectors express no viral genes, including Rep, which mediates integration into host genome during the latent state.In the context of human gene therapy, we have incomplete knowledge about the integration profile of rAAV vectors due to difficulty in accessing tissue.From studies performed, AAV integration appears to be a low frequency event, and for the most part broadly distributed across the genome, [157][158][159] although there are clearly preferences for actively transcribed genes. 160,161Nonetheless, the risk of insertional mutagenesis with AAV vectors will be closely watched in the coming years.

| DISCUSSION
The AAV gene therapy field is rapidly evolving and reaching maturity.Powered by a wealth of data in preclinical animal models, there are a wide variety of liver-directed gene therapies for metabolic diseases currently being tested in the clinic.This includes the recent approval of an AAV5 vector for liver-specific expression of Factor IX to treat hemophilia B, with similar therapies in the pipeline for hemophilia A and others.The stage is set for more AAV gene therapies of inherited metabolic diseases in the liver.Initially these will focus on disorders with a low threshold of correction, but will undoubtedly be expanded to situations where most hepatocytes require restoration of the defective protein or gene product.Beyond this, it is also conceivable that more common liver diseases such as nonalcoholic steatohepatitis, viral hepatitis, and cirrhosis could also be treated with AAV vectors.Advances in genome editing technology will be combined with AAV delivery to make more permanent alterations to the patient's own DNA, allowing for more precisely controlled gene expression.Understanding and managing the immune response to AAV capsid and transgene is a key focus of improving efficacy and safety across all of these genetic therapies.Further work is needed on the basic biology of AAV as a virus as well as a gene delivery system.Discoveries in this area will produce more potent and safer gene therapy vectors.The future is very bright for AAV, which will undoubtedly deliver life changing therapies for patients with inherited metabolic liver disorders in the years to come.
Figure generated with Biorender software.

T A B L E 3
Liver AAV clinical trials, as of April 2023.