**3. AAV capsid serotypes**

well as acquired diseases. By inducing modification of the gene pool, gene therapy aims to permanently and non-invasively treat the disease. Among the gene modifications that the therapy allows, a gene could be added, by direct introduction of a gene copy, silenced, by administering shRNA or siRNA, or removed, by the ZFN technology. Therefore, the spectrum

Even though the idea of gene transfer has been pursued for decades using an array of diverse delivery approaches, several setbacks hampered its success for some time. In 1999, the death of Jesse Gelsinger after receiving an adenoviral-based gene therapy for the treatment of severe combined immunodeficiency disorder forced the halt on gene therapy progress [1]. Following this tragic incident, a more serious regulatory scrutiny was established and the use of alter‐ native viral and nonviral vectors was investigated. Among viral platforms for gene delivery, adeno-associated virus (AAV) emerged in 1965 and has attracted much attention since then because the virus is not pathogenic, does not induce significant immune response and/or

Adeno-associated virus was first discovered in 1965 as a contamination of rhesus monkey kidney cell cultures that were infected with adenovirus stocks [2]. Initially, the virus was called defective as it was incapable of self-replicating in the absence of a helper virus, adenovirus or herpesvirus. Later, it was classified as a member of the Parvovirus family, genus Dependovi‐

Further investigation determined that it is a small virus (approximately 20 nm) composed of an icosahedral protein capsid, which contains single-stranded DNA of 4.7 kb. The viral genome is flanked at each end by inverted terminal repeats sequences of 145 bp called ITRs. These sequences self-assemble into hairpin structures, generating a double-stranded sequence, which serves as a template for replication. The viral genome encodes for two proteins: Rep and Cap. Rep is required for single-stranded DNA replication and packaging. Cap is necessary

AAV has never been associated with a disease or pathology [3]. Furthermore, due to the homology between the Rep-binding element present on the ITR, and the rAAVS1 sequence found on human chromosome 19, the viral genome can result in integration into the human genome [4]. This last feature is important because it shows that the virus can facilitate longterm expression of the viral genome. Additionally, specific integration of AAV in a defined locus minimizes the risks of mutagenesis due to random insertions, as other vectors do.

However, several genetic modifications of AAV have been performed in order to guarantee further safety for its translation into the clinic. First, the gene required for viral replication, called Rep, and the element required for site-specific integration were eliminated from the AAV genome. Therefore, this AAV variant, called recombinant vector (rAAV), will exist in an extrachromosomal state with very low integration efficiency into the genomic DNA, reducing

of diseases that could potentially benefit from this technology is expanding.

toxicity to humans while it allows long-term transgene expression.

**2. Emergence of rAAV as a therapeutic platform**

to form the viral capsid and transduce cells efficiently.

rus.

120 Gene Therapy - Principles and Challenges

Even though serotype 2 has been more extensively used and studied, other capsids are gaining more interest. The existence of a variety of serotypes makes rAAV gene therapy more attractive as they differ in infectivity rates and tissue specificity. For instance, a biodistribution analysis of different AAV capsid serotypes carrying the same luciferase reporter gene showed a broad dissemination of the virus in the mouse following intravenous administration [6]. In an attempt to study phylogenetic relationships among serotypes 1 to 12, their capsid amino acid sequences (NCBI reference sequences: NP\_049542.1, YP\_680426.1, NP\_043941.1, NP\_044927.1, YP\_068409.1, AAB95450.1, YP\_077178.1, YP\_077180.1, AAS99264.1, AY631965.1, AY631966.1 and AX753364.1) were aligned using ClustalOmega [7] and JalView, version 2.8.2 (Figure 1). According to the degree of similarity that a residue has with the consensus residue for each column, a certain color is given. Intensive blue corresponds to more than 80% agreement, light blue to agreement between 60% and 80%, light grey to agreement between 40% and 60% and white for agreement lower or equal to 40%. Below the alignment, conservation, quality and consensus information are provided. Conservation graphic highlights alignment regions where physicochemical properties are conserved. The more intense the color, the more conserved the physicochemical property is in the column. Alignment quality indicates the likelihood of observing substitutions in a particular amino acidic position. Finally, the residue consensus provides the most common residues and their percentage for each column of the alignment. column. Alignment quality indicates the likelihood of observing substitutions in a particular amino acidic position. Finally, the residue consensus provides the most common residues and their percentage for each column of the alignment.

Figure1. Multi-sequence alignment of AAV serotypes from 1-12

using ClustalOmega and JalView software.

each column of the alignment.

Figure1. Multi-sequence alignment of AAV serotypes from 1-12

using ClustalOmega and JalView software.

column. Alignment quality indicates the likelihood of observing

consensus provides the most common residues and their percentage for

AAV Biology, Infectivity and Therapeutic Use from Bench to Clinic http://dx.doi.org/10.5772/61988 123

column. Alignment quality indicates the likelihood of observing

each column of the alignment.

122 Gene Therapy - Principles and Challenges

substitutions in a particular amino acidic position. Finally, the residue

consensus provides the most common residues and their percentage for

Figure1. Multi-sequence alignment of AAV serotypes from 1-12

using ClustalOmega and JalView software.

**Figure 1.** Multi-sequence alignment of AAV serotypes from 1-12 using ClustalOmega and JalView software.

Figure 1. Multisequence alignment of AAV serotypes from 1 to 12 using ClustalOmega and JalView software. Following the multisequence alignment, the percentage of sequence homology was deter‐ mined by performing BLAST alignments of dual AAV sequences at the time (Figure 2a). Furthermore, the phylogenetic relationships among these AAV serotypes were determined by creating a neighbor-joining tree (Figure 2b), which uses the percent identity [8].


Following the multisequence alignment, the percentage of sequence

neighbor-joining tree (Figure 2b), which uses the percent identity [8].

sequences at the time (Figure 2a). Furthermore, the phylogenetic

homology was determined by performing BLAST alignments of dual AAV

Figure 2. A, Blast alignment of dual combination of AAV serotypes to determine **Figure 2. A**, Blast alignment of dual combination of AAV serotypes to determine percentage of homology. **B**, Phyloge‐ netic tree to determine phylogenetic relationships among the studied serotypes.

percentage of homology. B, Phylogenetic tree to determine phylogenetic relationships among the studied serotypes. This tree shows that serotype AAV5 has the most divergent amino acid capsid sequence, sharing between 53% and 59% homology with the rest of the human serotypes that have been discovered so far (highlighted in orange). AAV4 also shows a considerable degree of diver‐ gence, when comparing sequences of AAV1 to 9 (between 53% and 64%). However, AAV4 shares a more recent common ancestor with serotypes 11 and 12. Furthermore, AAV1 and AAV6 share 99% homology, being the closest AAV serotypes in sequence. The most common AAV serotype, AAV2, is closer in amino acid sequence to all the AAV serotypes, especially AAV3, but greatly differs from serotypes AAV5, AAV4 and therefore, AAV11 and AAV12. Serotypes AAV8 and AAV10 are also very close between each other, sharing 93% amino acid sequence homology. These differences in sequences were observed in other studies, not only when analyzing the sequence similarities but also when studying antigenic reactivities [9]. Remarkably, the variabilities in amino acidic sequences were mainly localized in the loopedout domains that are exposed to the surface of the capsid, rather than evenly distributed along the capsid protein sequence [10]. More interesting, Gao et al. compared phylogenies from human and nonhuman primate AAV serotypes. They observed that human AAV4 and AAV5 serotypes were the most divergent, and after they emerged, the rest of the viruses were clustered in groups that included human serotypes (AAV1, AAV6, AAV2, AAV3 and AAV9), exclusive rhesus serotypes (AAV7) or a combination of both (AAV8). Considering that human AAV serotypes share a high similarity in sequences with nonhuman AAV serotypes, they are both well disseminated and are able to cross species barriers. Therefore, there is a possibility that AAV from nonhuman primates could be used for treating human diseases. This is the case of AAVrh10, a serotype isolated from rhesus macaques. This virus was found to be signifi‐ cantly more efficient in transducing neurons from different areas in a healthy dog brain as compared with AAV1 or AAV5, but to a similar extent with AAV9 [11]. More importantly, the rhesus serotype is currently being tested for safety and efficacy in the clinic for the treatment of CNS diseases, such as Battens (NCT01414985 and NCT01161576, clinicaltrials.gov) and MLD (NCT01801709, clinicaltrials.gov). Additionally, a new study is planning to test the safety of this virus for delivering human alpha 1-antitrypsin cDNA to individuals with alpha 1 antitrypsin deficiency (NCT02168686, clinicaltrials.gov); although they are not yet recruiting patients. According to the biodistribution study of AAV following tail vein injection into the mouse, rAAV9 has the broadest tissue tropism, demonstrating robust transduction of all tested tissues other than the testes [6] (Figure 3). Moreover, it is the only one so far that can reach the brain, followed by AAV8. On the other hand, AAV7 showed strong tropism for the liver and to a lesser extent for the muscle. Meanwhile, AAV6 had more preference for the heart, in comparison to liver, lung and muscle. AAV4 was found in higher viral copies in the lung, followed by the heart. The rest of the serotypes transduced the selected tissues with lower efficiencies. AAV1 and AAV2 were more prone to reach the liver. In terms of infection kinetics, AAV7 and AAV9 were the fastest in targeting the tissue and showing expression of the reporter gene, luciferase. Meanwhile, AAV3 and AAV4 were the slowest ones. Additionally, among all the serotypes, AAV2, 3, 4 and 5 showed the lowest transduction efficiency.

Figure 3. Biodistribution of AAV serotypes 1–9 in mouse. **Figure 3.** Biodistribution of AAV serotypes 1–9 in mouse.

Following the multisequence alignment, the percentage of sequence

sequences at the time (Figure 2a). Furthermore, the phylogenetic

124 Gene Therapy - Principles and Challenges

homology was determined by performing BLAST alignments of dual AAV

relationships among these AAV serotypes were determined by creating a neighbor-joining tree (Figure 2b), which uses the percent identity [8].

Figure 2. A, Blast alignment of dual combination of AAV serotypes to determine

percentage of homology. B, Phylogenetic tree to determine phylogenetic

This tree shows that serotype AAV5 has the most divergent amino acid capsid sequence, sharing between 53% and 59% homology with the rest of the human serotypes that have been discovered so far (highlighted in orange). AAV4 also shows a considerable degree of diver‐ gence, when comparing sequences of AAV1 to 9 (between 53% and 64%). However, AAV4 shares a more recent common ancestor with serotypes 11 and 12. Furthermore, AAV1 and AAV6 share 99% homology, being the closest AAV serotypes in sequence. The most common AAV serotype, AAV2, is closer in amino acid sequence to all the AAV serotypes, especially AAV3, but greatly differs from serotypes AAV5, AAV4 and therefore, AAV11 and AAV12. Serotypes AAV8 and AAV10 are also very close between each other, sharing 93% amino acid sequence homology. These differences in sequences were observed in other studies, not only when analyzing the sequence similarities but also when studying antigenic reactivities [9]. Remarkably, the variabilities in amino acidic sequences were mainly localized in the looped-

**Figure 2. A**, Blast alignment of dual combination of AAV serotypes to determine percentage of homology. **B**, Phyloge‐

relationships among the studied serotypes.

netic tree to determine phylogenetic relationships among the studied serotypes.

According to the biodistribution study of AAV following tail vein injection into the mouse, AAV9 has the broadest tissue tropism, demonstrating robust transduction of all tested tissues otherthanthetestes[6](Figure3).Moreover,itisthemostefficientinreachingthebrain,followed by AAV8. On the other hand, AAV7 showed strong tropism for the liver and to a lesser extent for the muscle. Meanwhile, AAV6 had more preference for the heart, in comparison to liver, lung and muscle. AAV4 was found in higher viral copies in the lung, followed by the heart. The rest of the serotypes transduced the selected tissues with lower efficiencies. AAV1 and AAV2 were more prone to reach the liver. In terms of infection kinetics, AAV7 and AAV9 were the fastest in targeting the tissue and showing expression of the reporter gene, luciferase. Mean‐ while, AAV3 and AAV4 were the slowest ones. Additionally, among all the serotypes, AAV2, 3, 4 and 5 showed the lowest transduction efficiency.

The first AAV primary receptor that was identified was heparin sulfate proteoglycan (HSPG). It is the receptor that AAV2 and AAV3 bind when infecting cells (Figure 4). Even though AAV6 was shown to have moderate binding affinity for heparin, it does not have the two residues R585 and R588, that participate in AAV2 binding to HSPG. On the other hand, while sequence alignment comparison between AAV1 and AAV6 capsids revealed only a six–amino acid difference, AAV1 is not able to bind heparin. Mutagenesis analysis revealed that amino acid 531 was responsible for providing the heparin binding ability to AAV6 and not to AAV1 [12]. Furthermore, AAV1 binds both α2–3 and α2–6 *N*-linked sialic acid (SIA), same as AAV6. Interestingly, AAV5 also binds α2–3 SIA, although it only shares ~40% homology with capsid serotypes AAV1 and AAV6. Crystallography studies of AAV5 showed differences in the surface loop regions, specifically smaller HI loop and larger VR-VII loop, which are located on the depression wall at the icosahedral 2-fold axis and determine receptor binding, tissue transduction efficiency and antigenic reactivity [13].

AAV4 capsid serotype follows AAV5 in terms of low sequence similarity with the rest of the serotypes and between themselves (53% sequence homology). A study, in which sialic acids were removed from cell surfaces, by neuraminidase treatment, showed that both viruses require SIA for infectivity [14]. However, when cellular glycosylation was inhibited, only treatment with *O*-linked inhibitor decreased binding of AAV4 to cultured cells. Meanwhile, treatment with *N*-linked inhibitors of glycosylation blocked AAV5 binding to the cell surface. Resialylation experiments with neuraminidase-treated red blood cells further confirmed that AAV4 binding to SIA is through α2–3 *O*-linkage, rather than through α2–6 *N* linkage, which is the interaction that AAV5 establishes for the initial infection of a cell.

Still, currently, receptors for AAV7 and AAV8 are unknown. Glycan binding analysis on microarrays revealed that AAV7 and AAV8 did not bind to any of the glycans that commonly bind serotypes AAV1–6 [15]. However, similarly to AAV2 and AAV3, AAV8 and AAV9 interact with the 37/67 kDa laminin receptor (LamR), as a secondary receptor, for efficient internalization and transduction [16]. LamR participates in interactions of extracellular laminin1 with proteases and with the cell; therefore, it is widely distributed among human tissues. Even though, AAV2, 3, 8 and 9 serotypes mediate direct tissue transduction via interaction with the LamR molecule, they significantly differ in their tissue tropism. AAV8 and 9 are able to infect a broader spectrum of tissues, even the brain, compared to serotypes AAV2 and AAV3. This result suggests that the primary receptor or the combination of both receptors is required for viral biodistribution. At the UNC gene therapy center, the Asokan laboratory was able to identify that *N*-linked glycans with terminal galactosyl residues are involved in AAV9 tissue binding and transduction [17]. The high abundance of these glycans in various animal tissues could explain the broad tropism observed after AAV9 systemic administration.

Lastly, the brain is one of the most difficult tissues that AAV can access and infect following systemic administration. The presence of a mature blood–brain barrier constitutes a physical barrier to potential harmful molecules and infectious pathogens. Therefore, most of the AAV serotypes are not able to access the brain without direct intraparenchymal administration. However, AAV9 and AAV8 (to a lesser extent), have the capability to reach the brain following intravenous administration to neonatal or adult mice [18].

Furthermore, the rhesus serotypes AAV10 and AAV11 were found to be sequence homolo‐ gous and structurally closest to the previously described serotypes AAV8 and AAV4, respec‐ tively [19].

AAV12, which was isolated from a simian adenovirus stock, showed 74% homology with AAV4 and 84% with AAVrh11. However, it does not bind SIA and appears to have strong affinity for human cancer cell lines [20].

was shown to have moderate binding affinity for heparin, it does not have the two residues R585 and R588, that participate in AAV2 binding to HSPG. On the other hand, while sequence alignment comparison between AAV1 and AAV6 capsids revealed only a six–amino acid difference, AAV1 is not able to bind heparin. Mutagenesis analysis revealed that amino acid 531 was responsible for providing the heparin binding ability to AAV6 and not to AAV1 [12]. Furthermore, AAV1 binds both α2–3 and α2–6 *N*-linked sialic acid (SIA), same as AAV6. Interestingly, AAV5 also binds α2–3 SIA, although it only shares ~40% homology with capsid serotypes AAV1 and AAV6. Crystallography studies of AAV5 showed differences in the surface loop regions, specifically smaller HI loop and larger VR-VII loop, which are located on the depression wall at the icosahedral 2-fold axis and determine receptor binding, tissue

AAV4 capsid serotype follows AAV5 in terms of low sequence similarity with the rest of the serotypes and between themselves (53% sequence homology). A study, in which sialic acids were removed from cell surfaces, by neuraminidase treatment, showed that both viruses require SIA for infectivity [14]. However, when cellular glycosylation was inhibited, only treatment with *O*-linked inhibitor decreased binding of AAV4 to cultured cells. Meanwhile, treatment with *N*-linked inhibitors of glycosylation blocked AAV5 binding to the cell surface. Resialylation experiments with neuraminidase-treated red blood cells further confirmed that AAV4 binding to SIA is through α2–3 *O*-linkage, rather than through α2–6 *N* linkage, which

Still, currently, receptors for AAV7 and AAV8 are unknown. Glycan binding analysis on microarrays revealed that AAV7 and AAV8 did not bind to any of the glycans that commonly bind serotypes AAV1–6 [15]. However, similarly to AAV2 and AAV3, AAV8 and AAV9 interact with the 37/67 kDa laminin receptor (LamR), as a secondary receptor, for efficient internalization and transduction [16]. LamR participates in interactions of extracellular laminin1 with proteases and with the cell; therefore, it is widely distributed among human tissues. Even though, AAV2, 3, 8 and 9 serotypes mediate direct tissue transduction via interaction with the LamR molecule, they significantly differ in their tissue tropism. AAV8 and 9 are able to infect a broader spectrum of tissues, even the brain, compared to serotypes AAV2 and AAV3. This result suggests that the primary receptor or the combination of both receptors is required for viral biodistribution. At the UNC gene therapy center, the Asokan laboratory was able to identify that *N*-linked glycans with terminal galactosyl residues are involved in AAV9 tissue binding and transduction [17]. The high abundance of these glycans in various animal tissues could explain the broad tropism observed after AAV9 systemic administration. Lastly, the brain is one of the most difficult tissues that AAV can access and infect following systemic administration. The presence of a mature blood–brain barrier constitutes a physical barrier to potential harmful molecules and infectious pathogens. Therefore, most of the AAV serotypes are not able to access the brain without direct intraparenchymal administration. However, AAV9 and AAV8 (to a lesser extent), have the capability to reach the brain following

Furthermore, the rhesus serotypes AAV10 and AAV11 were found to be sequence homolo‐ gous and structurally closest to the previously described serotypes AAV8 and AAV4, respec‐

transduction efficiency and antigenic reactivity [13].

126 Gene Therapy - Principles and Challenges

is the interaction that AAV5 establishes for the initial infection of a cell.

intravenous administration to neonatal or adult mice [18].

tively [19].


proteoglycan (HSPG). It is the receptor that AAV2 and AAV3 bind when infecting cells (Figure 4). Even though AAV6 was shown to have moderate binding affinity for heparin, it does not have the two residues R585 and Figure 4. Primary and secondary receptors used for AAV serotypes from 1 to 9 to infect and transduce cell types. HSPG, heparin sulfate proteoglycan; FGFR1, FGFR1 fibroblast growth factor receptor 1; PDGFRB, platelet-**Figure 4.** Primary and secondary receptors used for AAV serotypes from 1 to 9 to infect and transduce cell types. HSPG, heparin sulfate proteoglycan; FGFR1 fibroblast growth factor receptor 1; PDGFRB, platelet-derived growth fac‐ tor receptor beta; HGFR, met/hepatocyte growth factor receptor; EGFR, epidermal growth factor receptor; LamR, lami‐ nin receptor.

R588, that participate in AAV2 binding to HSPG. On the other hand, while

derived growth factor receptor beta; HGFR, met/hepatocyte growth factor receptor; EGFR, epidermal growth factor receptor; LamR, laminin receptor.

sequence alignment comparison between AAV1 and AAV6 capsids

The first AAV primary receptor that was identified was heparin sulfate

#### revealed only a six–amino acid difference, AAV1 is not able to bind heparin. Mutagenesis analysis revealed that amino acid 531 was **4. Preclinical studies of rAAV in large animal models**

responsible for providing the heparin binding ability to AAV6 and not to AAV1 [12]. Furthermore, AAV1 binds both α2–3 and α2–6 N-linked sialic In an effort to translate the rAAV gene therapy to the clinic, preclinical studies for safety, efficient rAAV dosing and capsid transduction, transgene expression and immune responses towards the new transgene and/or the rAAV capsid were performed. In Figure 5, we summa‐ rize which serotype has been evaluated for targeting a certain disease in a large animal model, such as nonhuman primate, pig, cat, dog, rabbit and sheep. However, this section will be focused on preclinical studies using large animals for the treatment of hemophilia. and/or the rAAV capsid were performed. In Figure 5, we summarize which serotype has been evaluated for a certain disease in a large animal model, such as nonhuman primate, pig, cat, dog, rabbit and sheep. However, this section will be focused on preclinical studies using large animals for the

Even though rAAV therapies for treating hemophilia in mice produced successful results, their translation into large animals, such as a dog or

is drastically affected in the presence of an immunological response

of the foreign protein. However, treatment of a large animal with the

nonhuman primate, was not straightforward [21]. The vector efficacy does not completely follow a dose–response correlation in large animals, and it

Figure 5. Diseases treated in large animal models with different AAV

towards the viral capsids. Furthermore, in both mice and dogs, there is no direct correlation between the transgene copy numbers and the expression

therapy was promising as FIX is a secreted protein and only 1%–2% normal

treatment of hemophilia. **Figure 5.** Diseases treated in large animal models with different AAV serotypes

5. rAAV targeting the liver for hemophilia treatment

factor IX levels is enough to correct the disease [21].
