Intracellular Transport of Recombinant Adeno-Associated Virus Vectors

Abstract

Recombinant adeno-associated viral vectors (rAAVs) have been widely used for gene delivery in animal models, and are currently evaluated for human gene therapy after successful clinical trials in the treatment of inherited, degenerative or acquired diseases such as Leber congenital amaurosis, Parkinson disease, or heart failure. However, limitations in vector tropism, such as limited tissue specificity and insufficient transduction efficiencies of particular tissues and cell types, still preclude therapeutic applications in certain tissues.

Wild-type AAVs are defective viruses that require the presence of a helper virus to complete their life cycle. One the one hand, this unique property makes AAV vectors one of the safest available viral vectors for gene delivery. On the other hand, it also represents a potential obstacle because rAAV vectors have to overcome several biological barriers in the absence of a helper virus to transduce successfully a cell. Consequently, a better understanding of the cellular roadblocks that limit rAAV gene delivery is crucial and, during the last 15 years, numerous studies resulted in an expanding body of knowledge of the intracellular trafficking pathways of rAAV vectors. This review describes our current understanding of the mechanisms involved in rAAV attachment to target cells, endocytosis, intracellular trafficking, capsid processing, nuclear import and genome release with an emphasis on the most recent discoveries in the field and the emerging strategies used to improve the efficiency of AAV-derived vectors.

INTRODUCTION

Gene therapy consists of the introduction of a therapeutic nucleic acid into a target cell in order to i) replace the function of a defective gene, ii) increase the expression of a down-regulated gene or iii) reduce detrimental levels of a protein, via RNA interference or antisense technology. Gene transfer is considered the last and only option to treat certain life-threatening or otherwise debilitating diseases for which no other therapeutic strategy is available such as, for instance, genetic disorders, drug-resistant cancers, heart failure or neurodegenerative diseases. Despite significant progress in the field of non-viral DNA transfer¹, in vivo delivery of unprotected nucleic acids has proven difficult, since naked DNA is prone to rapid extracellular degradation and in the cell cytoplasm. For successful transfection of a cell the DNA has to be readily transported across the plasma membrane, escape from endocytic vesicles and finally cross the nuclear membrane. The transfection reagents developed so far have achieved considerable efficiencies in in vitro transduction of cell lines but the efficiency in in vivo transduction remains limited. Animal viruses, on the other hand, have evolved over millions of years in order to optimize the delivery of their genetic material into host cells and, despite potential issues relating to their clinical safety and/or their immunogenicity, viruses are still considered the most effective and promising vectors for in vivo gene delivery. Among the variety of viruses used as vectors for gene transfer (retroviruses, herpesviruses, adenoviruses, poxviruses, etc...), the most impressive therapeutic successes so far have been obtained with retrovirus vectors in the treatment of children suffering from X-linked Severe Combined Immunodeficiency Disease (SCID)^2-4 and with adeno-associated virus-derived vectors in the treatment of Leber congenital amaurosis, an inherited eye disease leading to retinal degeneration^5-7. Adeno-associated viruses (AAV) are small non-enveloped DNA viruses first discovered as contaminants of adenovirus preparations⁸. They belong to the genus dependovirus of the sub-family parvovirinae and the family parvoviridae and require co-infection with a helper DNA virus (adenovirus, herpesvirus or papillomavirus) to complete their productive replication cycle⁹. AAVs are nonpathogenic in humans or animals, and show a low immunogenicity in comparison to other viruses¹⁰. AAVs have an icosahedral capsid (22-26 nm in diameter) containing a single-stranded DNA genome of approximately 4.7 kilobases that is flanked by two T-shaped inverted terminal repeats (ITRs). The left-hand of the viral genome contains the rep gene, encoding the nonstructural proteins rep78, 68, 52 and 40 that are required for DNA replication and packaging. The right-hand of the AAV genome contains the cap ORF, encoding the capsid proteins VP1, VP2 and VP3 and a newly identified chaperone, the so-called assembly-activating protein (AAP), that is necessary for capsid formation¹¹. The isolation of an infectious clone of AAV2^{12, 13} and the demonstration that the AAV2 backbone could be used to express foreign proteins in cultured cells^{14, 15} paved the way for using AAVs as gene delivery vectors. In contrast to other viruses such as first and second generation adenoviruses, which require several viral genes in cis for viral vector production, the ITRs are the only cis-elements of the AAV genome necessary for DNA replication and packaging, and the entire rep and cap ORFs can be replaced by any sequence of interest within a size limit of approximately 5kb¹⁶. Furthermore, any transgene flanked by AAV2 ITRs can be packaged into the capsid of any of the nine AAV serotypes (AAV1-9) currently evaluated for in vivo transduction^{17, 18}. This is of particular importance since different AAV serotypes show very different tissue tropisms^19-21 (summarized in Table 1). AAV-derived vector DNA rarely integrates in the host cell genome but can lead to sustained, long-term transgene expression in non-dividing cells²², which makes them particularly suitable for gene therapy of largely post-mitotic tissues like the brain, the retina, the liver, skeletal muscles or the heart.

Table 1.

AAV receptors and preferential tissue tropism

Virus	Glycan Receptor	Co-receptor/other	Tissue tropisma
AAV1	N-linked Sialic acid122, 123	Unknown	SMb124, CNS125, Retina126, Pancreas127
AAV2	HSPG25	FGFR126, HGFR30, LamR29, CD9 tetraspanin128	VSMC129, SM130, CNS131, Liver132, Kidney133
AAV3	HSPG18	FGFR1134, HGFR135, LamR29	Hepatocarcinoma136, SM124
AAV4	O-linked Sialic acid137	Unknown	CNS138, Retina139
AAV5	N-linked Sialic acid137, 140	PDGFR141	SM124, CNS138, Lung142, Retina126
AAV6	N-linked Sialic acid123, HSPG122	EGFR143	SM144, SM (IV)145, Heart24, Lung146
AAV7	Unknown	Unknown	SM23, Retina147, CNS148
AAV8	Unknown	LamR29	Liver23, SM23, 149, CNS148, Retina147, Pancreas150, Heart149
AAV9	N-linked galactose151	LamR29	Liver152, Heart (I.V.)152, 153, Brain (I.V.)154, SM (I.V.)155, Lungs156, Pancreas153, Kidney (I.V.)156
BAAV	Ganglioside GM137	Unknown	Unknown

^aPreferential tissue tropism following local delivery, unless otherwise indicated (I.V.=Intravenous injection)

^bAbbreviations : SM, Skeletal Muscle; CNS, Central Nervous System; VSMC, Vascular Smooth Muscle Cells

CELLULAR BARRIERS TO AAV TRANSDUCTION

Despite the great potential of rAAV for in vivo gene transfer and the availability of several serotypes and a large number of capsid variants²³, AAV-derived vectors display insufficient transduction efficiency in certain tissues and their relatively low organ specificity can be problematic for specific therapeutic applications. For instance, serotypes 1, 6 and 9, which are often referred to as “cardiotropic” due to their efficient transduction of cardiomyocytes, also transduce the liver to a significant extent following intravenous injection. Conversely, overall transduction levels are comparatively low for all serotypes in organs such as the brain, the pancreas or the kidneys following systemic injection^{20, 24}. Like all non-enveloped DNA viruses, AAVs must balance the need for a high capsid stability to prevent degradation in the extracellular environment as well as within the cell (e.g., by lysosomal proteases or the proteasome) with the capacity to readily release their genome in the nucleus of the target cell. Thus, AAV-based vectors must overcome several limiting and complex steps between the cell membrane and the nucleus, where they ultimately deliver their genome for successful transduction. First, the virion must bind to the surface of the target cell via one or more receptors/co-receptors. The second step consists in the uptake of the virus-receptor complex by endocytosis following invagination of the cell membrane. The next steps include the maturation of virus-containing vesicles into more acidic endocytic compartments, either late endosomes, recycling tubular endosomes or lysosomes, followed by retrograde transport to the trans-Golgi, medial-Golgi, cis-Golgi, ER-Golgi intermediate compartment (ERGIC, also called cis-Golgi network) and/or the endoplasmic reticulum. Those stages, which can be collectively characterized as “endosomal trafficking”, can trigger irreversible modifications of the viral capsid both due to the lower pH and through the (enzymatic) action of resident proteins of these compartments. Retrograde transport is followed by escape of the viral particle into the cytoplasm. This is most likely followed by the nuclear import of the intact viral particle through the nuclear pore complex (NPC; see below). After nuclear translocation, single-stranded DNA is released by capsid uncoating and converted into double-stranded DNA to allow transgene expression.

Although receptor expression pattern and viral attachment are obviously important in determining the tropism of AAV vectors, there is now strong evidence that post-attachment steps, such as endosomal escape or nuclear import, can drastically alter transduction efficiency. In fact, almost each AAV trafficking stage, from endocytosis to nuclear import, can constitute a rate-limiting step in a serotype- and cell type-dependent manner. While many details of AAV trafficking remain to be determined, the use of both small drug and genetic inhibitors and enhancers (Table 2) combined with fluorescence microscopy have greatly contributed to our understanding of AAV trafficking. Importantly, this increasing knowledge has also resulted in the development of novel strategies that might help in overcoming some of the limitations that AAV trafficking poses to successful transduction.

Table 2.

Inhibitors and enhancers of AAV trafficking/transduction

Virus	Drug/Protein/Mutation	Effect	Fold enh/inhiba	Possible Target
AAV2	Heparin25	Inhibitor	- 100-fold	Blocks attachment
AAV2	Bafilomycin A139, 51	Inhibitor	- 10-fold	Blocks endosomal acidification/trafficking
AAV2	Brefeldin A51	Inhibitor	- 100-fold	Disrupts the Golgi apparatus, blocks EE-LE transport and Golgi-ER transport/trafficking
AAV2	Chloroquine65	Inhibitor	- 100-fold	Blocks endosomal acidification/trafficking
AAV2	Wortmannin43	Inhibitor	- 4-fold	Blocks EE-EE and CLIC/GEEC-EE fusion/trafficking
AAV2-8	Cathepsin B-L inhibitors84	Inhibitor	- 3-fold	Blocks endosomal capsid processing
AAV2	Rac1 T17N43	Inhibitor	- 2-fold	Blocks macropinocytosis/entry
AAV2	Dynamin1 K44A39, 40	Inhibitor	- 2-3-fold	Blocks clathrin-mediated endocytosis/entry
AAV2	Rab7 siRNA67	Inhibitor	- 2-fold	Blocks late endosome-Golgi transport/trafficking
AAV2	Rab11 siRNA67	Inhibitor	- 40%	Blocks recycling endosome-Golgi transport/trafficking
AAV2	AAV HD/ANb mutant53, 75	Inhibitor	- 500-fold	Endosomal escape
AAV2	AAV BC3 mutant65, 79	Inhibitor	- 10000-fold	Viral NLS mutant, blocks nuclear import
AAV2	AAV BR3K mutant53	Inhibitor	- 500-fold	Blocks mobilization from the nucleolus
AAV2	Capsid phosphorylation111	Inhibitor	- 3-fold	Signal for capsid ubiquitination?
AAV2-5	MG13251, 97	Enhancer	+ 10-100-fold	Proteasome inhibitor/capsid protection? Trafficking?/2nd strand synthesis?
AAV2-5	LLnL97	Enhancer	+ 10-200-fold	Proteasome inhibitor, similar to MG132
AAV2-5	Doxorubicin157, 158	Enhancer	+ 10-100-fold	Topoisomerase inhibitor
AAV2	E3 Ub ligase inhibitor96	Enhancer	+ 3-fold	Blocks cellular/capsid ubiquitination?
AAV2	hydroxyurea159-161	Enhancer	+ 10-100-fold	Triggers DNA damage response/Additive effect with MG132
AAV2	Tyrphostin-2398	Enhancer	+ 20-40-fold	Blocks capsid phosphorylation/ubiquitination? similar to MG132
AAV2	Adenovirus109	Enhancer	+ 10-20-fold	Multiple/Unknown, non-additive with MG132)
AAV2	Misfolded CFTR protein mutant	Enhancer	+10-fold	ER stress/Misfolded Protein Response inducer
AAV2	Heat shock99	Enhancer	+ 10-fold	2nd strand synthesis? Hsp70 induction? ER stress? Synergistic effect with Tyrphostin23
AAV2	TC-PTP162	Enhancer	+ 16-fold	Blocks capsid phosphorylation/ubiquitination?
AAV2	PP5110	Enhancer	+ 6-fold	Blocks capsid phosphorylation/ubiquitination?
AAV2-6	AAV Tyrosine Mutants114, 115, 163, 164	Enhancer	+ 10-100-fold	Blocks capsid phosphorylation/ubiquitination? Non-additive with MG132

^aTransduction data are expressed as ratio to control infections and may represent experiments performed in various cell lines/animals/time points.

^bAbbreviations : HD/AN, VP1 double mutant (⁷⁵HD>AN) with no PLA2 activity; BC3, VP1/2 basic cluster 3 ¹⁶⁸RK>NN mutant; BR3K, highly basic VP1/2 basic cluster 3 ¹⁶⁷A>K mutant; CFTR, Cystic fibrosis transmembrane conductance regulator; TC-PTP, T cell protein tyrosine phosphatase; PP5, serine/threonine protein phosphatase 5.

AAV ATTACHMENT : RECEPTORS AND CORECEPTORS

Despite our rapidly growing knowledge about novel AAV serotypes, the vast majority of studies on AAV biology have been performed using AAV2 as a model, for reasons that are both historical (first AAV infectious clone¹²) and practical (ability to transduce efficiently common laboratory cell lines such as HeLa or HEK-293T¹⁵). Hence, AAV2 was the first serotype whose attachment receptor was investigated and identified as heparan sulfate proteoglycan (HSPG) by competition and overexpression assays^{25, 26}. This finding was followed by the identification of several putative proteinaceous co-receptors such as fibroblast growth factor receptor 1 (FGFR1)²⁶, integrin αVβ5²⁷, integrin α5β1²⁸, laminin receptor²⁹ or hepatocyte growth factor receptor (HGFR)³⁰, suggesting that AAV2 infection requires a primary glycan receptor together with a co-receptor for optimal attachment and internalization. However, independent studies have challenged the role of integrin αVβ5 in AAV2 transduction³¹ and, more recently, a high throughput siRNA screen failed to confirm a role of FGFR1 or HGFR in AAV2 transduction of human aortic endothelial cells³². Interestingly, naturally occurring AAV2 variants recovered from human subjects do not use HSPG as an attachment receptor³³, which could suggest that AAV2 ability to bind HSPG constitutes a tissue culture adaptation, positively selected via multiple passages of AAV-contaminated adenovirus preparations in HEK293 cells. AAV2 capsid mutants deficient in HSPG binding show a markedly reduced transduction of mouse liver but retain the ability to transduce cardiac muscle³⁴, indicating that AAV2 can use alternative receptors. Overall, the identification of receptors used by many AAV serotypes has reinforced the general view that AAVs use one or more proteoglycan conjugate as a primary receptor (HSPG for AAV2, AAV3 and AAV6, O-linked 2,3-sialic acid for AAV4, N-linked sialic acid for AAV1, AAV5 and AAV6, N-linked galactose for AAV9) together with a proteinaceous receptor (FGFR1, integrins and/or HGFR for AAV2, HGFR for AAV3, EGFR for AAV6, PDGFR for AAV5, LamR for AAV2, AAV3, AAV 8 and AAV9) for efficient binding and endocytosis (Table 1). Dual receptor requirement is frequently observed for other viruses³⁵ and is reminiscent of ligand-mediated activation and subsequent endocytosis of receptors such as FGFR1, which require HSPG binding³⁶. One exception to this rule would be bovine AAV (BAAV), which requires a sialylated glycolipid, GM1 ganglioside, for transduction³⁷. Most interestingly, recent data indicate that binding of AAV2 to HSPG not only allows virus attachment, but also induces a conformational change of the capsid³⁸. This could suggest that the binding of AAV to its proteoglycan primary receptor “locks” the virion in a transition state that may increase its affinity for the secondary receptor and consequently trigger endocytosis.

AAV ENDOCYTOSIS

Following attachment to surface receptors, AAV internalization occurs via endocytosis, a process that starts with the invagination of the plasma membrane domains containing virus-receptor complexes and is followed by the scission and release of the newly formed vesicle into the cell. In the case of AAV2, this process is fast and efficient, since the totality of virions bound to the cell surface can be internalized in about 30 minutes^{39, 40}. Therefore, virus internalization per se is likely not a rate-limiting step for AAV2 transduction, at least in tissue culture. Mammalian cells have developed multiple mechanisms of endocytosis aimed at targeting and sorting membrane-bound or extracellular cargos towards specific intracellular compartments. Those pathways include, but are not restricted to, clathrin-mediated endocytosis, caveolae-mediated endocytosis, macropinocytosis, phagocytosis, or several ill-characterized clathrin- and dynamin-independent pathways^{41, 42}. Early studies showed a partial (50 to 70%) inhibition of AAV2 uptake and transduction in cells overexpressing a dominant-negative mutant of dynamin, a protein involved in the scission of clathrin-coated pits and caveolae^{39, 40}, suggesting that clathrin-mediated uptake was the default pathway of AAV2 endocytosis. Later studies showed that inhibition of the Rac1 GTPase strongly decreased cellular uptake of AAV2⁴³ and that expression of a constitutively active mutant of Rac1 dramatically increased transduction⁴⁴. These observations were surprising because Rac1 is the major effector of macropinocytosis, a dynamin-independent fluid-phase endocytic pathway distinct from clathrin-mediated endocytosis⁴⁵, and there is evidence that constitutively active Rac1 inhibits clathrin-mediated endocytosis^{46, 47}. In addition, internalized AAV2 virions showed little or no co-localization with a fluorescently labeled adenovirus type 5⁴⁰, which is known to enter target cells via clathrin-mediated endocytosis⁴⁸.

Recently, our group revisited the nature of the endocytic mechanism of AAV2 endocytosis and we could not confirm a clear involvement of clathrin, dynamin or Rac1 in AAV2 transduction but instead found that viral infectious entry is dependent on membrane cholesterol, actin, Cdc42, Arf1 and GRAF1, which together define the recently characterized Clathrin-Independent Carriers/GPI-Enriched Endocytic Compartment (CLIC/GEEC) endocytic pathway⁴⁹. Similarly, a genome-wide siRNA screen has identified several components of CLIC/GEEC as strong inhibitors of AAV2 transduction in human aortic endothelial cells³² and recent experiments show that the internalization mechanism of HSPG, the primary receptor for AAV2, shares many features of CLIC/GEEC⁵⁰. In our study CLIC/GEEC endocytosis was required for efficient transduction, whereas inhibition of dynamin activity by dynasore or by dominant-negative dynamin mutants had no effect. There is currently no clear explanation to this discrepancy with previous studies. One possibility is that the use of adenoviral vectors for protein overexpression in some of the aforementioned studies^{39, 43, 44} could modify general cell homeostasis. Pre-infection with first generation adenoviral expression vectors (that retain E2, E4 and VA regions) shows a substantial cytopathic effect (our unpublished observations) and increases AAV2 transduction by at least one order of magnitude^51-53, and it has been reported that adenovirus infection enhances trafficking of AAV2 to the nucleus⁵⁴. Furthermore, Adenovirus type 5 E4 region enhances the expression of adhesion proteins on the cell surface⁵⁵, and E4orf1 partially localizes to clathrin vesicles⁵⁶ and modulates Rac1 and PI3K activity⁵⁷. Therefore, it cannot be excluded that the use of Adenoviral vectors modulates with AAV2 trafficking.

In our studies, inhibition of either dynamin or CLIC/GEEC only partially inhibited virus internalization, but simultaneous inhibition of both pathways completely blocked viral entry, which indicates that AAV2 can enter cells via at least two distinct pathways⁴⁹. These observations are not restricted to AAV2, since a similar dichotomy has been suggested for AAV5^{58, 59}, although this serotype reportedly uses a different endocytic mechanism⁶⁰. Importantly, in the case of AAV2, only CLIC/GEEC is required for successful transduction, suggesting that viral particles internalized via dynamin-dependent endocytosis are directed towards a “dead-end” compartment and seldom contribute to transduction. Thus, efficient endocytosis is necessary but not sufficient for AAV transduction. This notion is supported by the fact that NIH3T3 cells can internalize AAV2 virions as efficiently as HeLa or 293T cells, but they are not permissive to transduction due to impaired post-entry trafficking⁶¹. In vivo, experiments performed with several AAV serotypes show no strict correlation between intracellular viral DNA accumulation and transgene expression in various organs^{20, 24, 62}. Another striking example of the coexistence of infectious and non-infectious endocytic pathways within the same cell was provided by Di Pasquale and colleagues, who showed that intact AAV4, AAV5 or BAAV (but not AAV2 or AAV6) viral particles could cross monolayers of polarized cells when applied on the apical side, by a process known as transcytosis⁶³. Interestingly, drugs known to inhibit transcytosis strongly increased cell transduction, suggesting that polarized cells can redirect attached virions towards two separate endocytic pathways, one of which leads to transcytosis, and the other results in nuclear transport and gene expression. It is yet unclear whether this dichotomy reflects the use of a different receptor, but recent experiments by DiPasquale and colleagues indicate that BAAV uses a distinct receptor for transcytosis versus transduction⁶⁴. Taken together, all these studies indicate that AAV can be internalized through a variety of mechanisms, namely clathrin-mediated endocytosis, caveolae-mediated endocytosis, via the CLIC/GEEC pathway, or possibly through transcytosis (Figure 1) the nature of which will then commit the virions to either an infectious pathway (by ultimately delivering the virion to the nucleus) or to a non-infectious pathway (by targeting the virion to lysosome/proteasome system for degradation or by transporting it to the extracellular medium by transcytosis). Thus, even though endocytosis per se is not necessarily rate-limiting from a strictly quantitative point, it constitutes a crucial step in irreversibly determining the fate of incoming AAV virions.

Figure 1.

Model of Entry and Intracellular Trafficking of AAV Vectors. Following binding to a receptor/co-receptor complex, rAAV enters target cell through endocytosis via one or more of the following pathways: CLIC/GEEC endocytosis, clathrin-mediated endocytosis (CCP) or caveolar endocytosis (CAV). Virions are sorted towards the trans-Golgi network (TGN) along a retrograde transport pathway presumably involving trafficking via early endosomes (EE), followed by late endosomes (LE), perinuclear recycling endosomes (PNRE) or both. Capsid conformation changes and exposure of the PLA2 domain (spikes) allows cytoplasmic release from the Golgi apparatus or the Endoplasmic Reticulum (ER), and nuclear import via the nuclear pore complex (NPC). After nuclear import, intact capsids accumulate in the nucleolus (No) before mobilization into the nucleoplasm (NP) and genome release by partial uncoating. Some of the steps indicated are hypothetical and have not been conclusively proven yet.

POST-ENDOCYTIC TRAFFICKING

In contrast to virus internalization, several lines of evidence indicate that intracellular trafficking of AAV is a rate-limiting, slow and inefficient process. In cultured cells, only a small fraction of viral particles access the nucleus within 16-20 hours, while the majority accumulate in a perinuclear compartment⁴⁰. AAV2 transduction can be efficiently blocked by lysosomotropic drugs like bafilomycin A1, chloroquine or ammonium chloride, that buffer endosomal pH, indicating that endosome acidification is necessary for AAV2 processing^{40, 51, 65} and inhibition of transduction by brefeldin A suggests that transport steps involving the Golgi apparatus are involved⁵¹. The retrograde transport intermediates used by AAV are not yet completely defined, but at least for AAV2 and AAV5 incoming virions appear to accumulate in the Golgi apparatus before their release in the cytoplasm^{53, 59, 66}. Previous studies indicate that a fraction of AAV2 virions transit via late and/or recycling endosomes enriched in Rab7 and Rab11, respectively⁶⁷. AAV retrograde transport from an early endosomal compartment towards the trans-Golgi network is possibly similar to previously described pathways involved in trafficking of bacterial toxins^{68, 69} or other non-fusogenic DNA viruses such as SV40 or murine polyomavirus^70-72. The transport of AAVs within the endomembrane system induces profound changes in capsid conformation that are required for efficient transduction⁷³. Perhaps the most convincing evidence for the necessity of endosomal processing of AAV came from a study by Sonntag et al.⁶⁵, who showed that microinjection of intact virus in the cytoplasm - and even in the nucleus - of HeLa cells resulted in very low transduction. In the same study, the authors also observed that transduction could be blocked by cytoplasmic injection of antibodies targeted to the intact capsids or to the N-terminal part of VP1 and VP2. In conclusion, these experiments unequivocally demonstrate that i) the AAV2 capsid must undergo a conformational change in the endocytic system, ii) passage of the virions through the cytoplasm is an obligatory stage in transduction, iii) endosomal processing exposes the VP1 and VP2 N-terminal domains outside of the capsid and iv) the presence of capsids with externalized VP1 and/or VP2 N-termini in the cytoplasm is required for transduction. The N-terminal domains of VP1, which is buried inside the capsid in intact virions⁷⁴, contains a phospholipase A2 (PLA2) domain that has been shown to be necessary for endosomal escape and transduction^75-78 as well as three basic clusters (BC1, 2 and 3) that presumably act as nuclear localization signals (NLS) for the viral capsid^{65, 79}. Taken together, these observations suggest that within the endocytic system a conformational change exposes the VP1 and VP2 N-termini outside the capsid, which then allows the virus to escape the endomembrane compartment into the cytoplasm. Such process has been observed for several non-enveloped viruses, including papillomavirus⁸⁰, polyomavirus⁸¹ or adenovirus⁸². Once in the cytoplasm, the basic residue clusters of VP1 or VP2 would then mediate the nuclear import of full virions via the NPC. AAV2 virions carrying a mutation in either the PLA2 domain or the first basic cluster in VP1 accumulate in the Golgi apparatus, which suggests that i) exposure of the VP1 PLA2 domain is not necessary for endosome-to-Golgi retrograde transport and ii) that the endosomal release may occur from the Golgi apparatus or the ER-Golgi intermediate compartment (ERGIC)^{53, 83}. In vitro exposure of AAV2 to acidic pH (pH 4-5) does not induce VP1/2 N-termini externalization⁶⁵. This indicates that while endosomal acidification is necessary for AAV2 processing, endosomal acidification alone is insufficient for VP1/2 externalization. Thus, endosomal acidification could be necessary for vesicular trafficking and/or dissociation of the virus/receptor complex, but insufficient for capsid conformational modifications and endosomal escape. Recent studies indicate that endosomal cathepsins B and L are necessary for transduction by “priming” the capsids by partial proteolysis⁸⁴, but it is yet uncertain whether this interaction modulates endosomal escape or a later step. Strikingly, cathepsin digestion of AAV2 and AAV8 showed a different cleavage pattern, which could be partially responsible for the fact that AAV2 and AAV8 show different uncoating kinetics⁸⁵. AAV5 was not sensitive to cathepsin proteolysis indicating that this process is not conserved among all AAV serotypes⁸⁴.

NUCLEAR IMPORT

The mechanism allowing AAV particles to cross the nuclear membrane is still poorly understood, but converging evidence suggest that intact particles are translocated into the nucleus before DNA release. Intact virions can still be detected inside the nucleus by immunofluorescence microscopy^{40, 43, 53, 86}, nuclear microinjection of antibodies against intact capsids blocks transduction⁶⁵, and fully infectious virions can be recovered from infected cell nuclei⁸⁶. Although Kleinschmidt and colleagues suggested that AAV2 nuclear import was inefficient and rate-limiting, based on poor infectivity of virions with exposed VP1/2 N- termini following cytoplasmic injection⁶⁵, this phenomenon could reflect an incomplete processing of the capsid rather than a low rate of nuclear import. Given the complexity of capsid modifications in the endomembrane system, as discussed earlier, it is very difficult to evaluate the efficiency and kinetics of AAV nuclear import per se.

The basic clusters BC2 and BC3, shared by the N-termini of VP1 and VP2, are necessary for transduction^{65, 79} and together confer nuclear localization to a heterologous fusion protein^{65, 87}, which strongly suggests that the BC2-3 domain form a bipartite NLS that allows virus nuclear import via the NPC. AAV2 basic clusters 2 and 3 are separated by a 23-amino acids linker, which classifies them as a non-classical NLS. Therefore it is likely that AAV2 nuclear import relies on a direct interaction between the VP1/2 N-terminus and a member of the importin-β family, rather than with a classical importinα/importinβ complex.^88-90

Once the virus has gained access to the nucleus, it is readily transported to the nucleolus, in which it is maintained as an intact particle until its egress into the nucleoplasm⁸⁶. The process leading to AAV2 nucleolar localization is unknown, but it should be mentioned that AAV2 capsid has been reported to interact directly with nucleolin⁹¹ and B23/nucleophosmin⁹², two major nucleolar proteins. The exact role of nucleolar transport in AAV biology is uncertain, but the observation that nucleolin or nucleophosmin knock-down by RNA interference strongly increase AAV2 transduction⁸⁶ suggests that nucleolar transit may be dispensable or even detrimental for AAV2 uncoating and gene expression. The process of uncoating itself is not completely understood, but it is probable, as observed with autonomous parvoviruses, that DNA can be released without complete disassembly of the capsid^{93, 94}. The kinetics of DNA release appear to be cell- and serotype-dependent. In cells from cardiac origin, AAV6-packaged genomes have been reported to be released more efficiently than their AAV2-packaged counterparts, but the opposite trend is observed in HeLa cells⁹⁵. Similarly, although AAV2, 6 and 8 show similar uptake and nuclear import rates during murine liver transduction, AAV6 and 8 appear to release their genome earlier than AAV2. Surprisingly, intact AAV2 virions are completely and rapidly dissociated following incubation with a liver nuclear extract⁸⁵. Despite an apparent contradiction with the in vivo data, this observation could be explained by the phenomenon of nucleolar sequestration suggested by Johnson and colleagues, as discussed earlier; if AAV2 sequestration in the nucleolus prevents uncoating and genome release in vivo, one could expect the uncoating to be more efficient in an in vitro reaction, in which virions are directly exposed to nucleoplasmic factors. These observations also mirror those by Miao and colleagues, who described a striking discrepancy between substantial nuclear accumulation of AAV2 genomes and poor transduction in mouse liver⁶². It will be very interesting to investigate further the possible relationship between nucleolar mobilization and the serotype- and tissue-specific differences in AAV transduction.

PROTEASOME INHIBITORS AND CAPSID UBIQUITINATION

Pioneer experiments by Douar and colleagues showed that the proteasome inhibitor MG132 dramatically enhanced AAV2 transduction in cultured cells and delayed viral DNA decay⁵¹. In addition, AAV2 and AAV5 capsids can be ubiquitinated both in vivo and in vitro^{96, 97}, and transduction is increased, albeit to a lesser extent, following treatment with a E3 ubiquitin ligase inhibitor⁹⁶. This led to the hypothesis that a significant fraction of incoming virions were targeted to proteasome degradation by ubiquitin conjugation of the capsid during cytoplasmic trafficking. However, this model was later challenged by several observations: i) other proteasome inhibitors, such as LLnL, enhance transduction to the same extent as MG132 with no visible effect on viral DNA degradation and ii) the accumulation of viral DNA following MG132 treatment (about 2-3 fold 24h post-infection) cannot account for the increase in transduction observed at the same time point (50-fold). Altogether, these observations clearly indicate the dramatic increase in transduction caused by MG132 cannot be explained by an inhibition of proteasome-mediated virus degradation alone. Along those lines, it was proposed that proteasome inhibitors increase AAV trafficking to the nucleus by a yet unexplained mechanism, in a cell- and serotype-specific manner^96-98.

Interestingly, proteasome inhibitors are strong activators of the ER stress/misfolded protein response pathway, and recent studies have shown that ER stress induction could potentiate AAV2 transduction by an order of magnitude⁸³. Consistently, heat-shock treatment⁹⁹ and alteration of cell redox status¹⁰⁰, both known as potent ER stress inducers, increase AAV2 transduction to a comparable extent. Proteasome inhibitors also induce the formation of large nucleolar stress bodies enriched in proteasome, ubiquitinated cellular proteins, nucleoplasmic proteasome targets and heat-shock proteins^101-103, reminiscent of the strong nucleolar accumulation of AAV2 observed after MG132 treatment^{53, 86}.

The exact role of capsid ubiquitination in AAV transduction is still poorly understood. For instance, it is not certain whether ubiquitination is beneficial or detrimental for AAV2 transduction. In the studies by Duan et al., transduction enhancement observed after E3 ligase inhibition might result from indirect pleiotropic effects of the drug on the cell metabolism. Mutagenesis of lysines residues exposed on the surface of the AAV2 capsid, the potential targets of ubiquitination, showed no improvement in infectivity^{34, 104, 105}. Interestingly, in vitro capsid ubiquitination is much more efficient on denatured capsid proteins than on intact virions⁹⁷, which led the authors to formulate the hypothesis that ubiquitination in vivo requires a conformational change of the capsid and the exposure of normally inaccessible lysine residues. Consistent with this view, bioinformatical analysis indicates that most of the potentially ubiquitinated lysine residues localize on the N-termini of VP1 and VP2 in AAV serotypes 1 to 12 (our unpublished observations). Hence, it will be interesting to study the function of these residues in AAV trafficking and transduction.

CAPSID PHOSPHORYLATION AND TYROSINE MUTANTS

A fascinating development in our understanding of the cellular roadblocks to AAV transduction came from a series of studies on the role of tyrosine phosphorylation in viral trafficking and gene expression. Initial reports from Qing and colleagues showed that a cellular phosphoprotein, ssD-BP (single-strand D-sequence binding protein) could bind to the ITR of the AAV2 genome and block complementary strand synthesis¹⁰⁶. General inhibition of tyrosine phosphorylation by genistein or tyrphostin-23 (Tyr23), but also treatment with the DNA-damaging drug hydroxyurea or expression of adenovirus E4orf6 protein, dephosphorylated ssD-BP and showed a strong concomitant increase in AAV2 transduction^107-109. In parallel, T-cell protein tyrosine phosphatase (TC-PTP) or protein phosphatase 5 (PP5) expression dramatically increased transduction¹¹⁰.

Later studies from the same group demonstrated that phosphorylation inhibitors also had strong positive effects on virus transport to the nucleus⁹⁸. Intriguingly, Tyr23 did not show a synergistic or additional effect with MG132, which suggests that both drugs enhance transduction via a common mechanism. Simultaneously, Zhong et al. found that Tyr23 in the intact AAV2 capsids could be specifically phosphorylated in vitro by the EGFR-protein tyrosine kinase (EGFR-PTK). Most interestingly, in vitro phosphorylated AAV2 showed a 3-fold reduction in transduction, independently of second-strand DNA synthesis. A possible explanation was that exposed phosphotyrosines acted as a positive signal for capsid ubiquitination¹¹¹. Consistently, mutagenesis of highly conserved exposed tyrosine residues (Y444F, Y500F or Y730F) on AAV2 capsids enhanced transduction up to 10-fold in HeLa cells and 30-fold in mouse liver. Transduction by tyrosine mutants was not further enhanced by adenovirus co-infection, MG132 or Tyrphostin-23 treatment, which again suggests that all these distinct treatments target a common step of AAV transduction, involving capsid phosphorylation and possibly ubiquitination. Since then, single or combined tyrosine mutants of AAV2 have been successfully tested in vitro in fibroblasts and mesenchymal stem cells¹¹² and in vivo in murine hepatocytes^{113, 114}, and retina^{115, 116}. All studies so far showed a marked increase (10- to 100-fold) of transduction, successful transduction of non-permissive cells and an additive effect of multiple (up to three) tyrosine mutations. Improved transduction of mouse skeletal muscle was also obtained with tyrosine mutants of AAV6 vectors, and interestingly showed similarly enhanced long-term persistence of intracellular viral DNA¹¹⁷.

CONCLUDING REMARKS

The last decade has seen considerable progress in our understanding of AAV biology and the rapid identification of novel serotypes with different tissue tropism is opening new avenues for gene therapy applications. It is now well recognized that intracellular trafficking of AAV vectors constitutes a major limiting step in transduction. Not surprisingly, our knowledge of this highly complex, multistep process is still incomplete and future research is needed to improve our understanding of AAV capsid processing, endosomal escape, nuclear import and uncoating. This might be a challenging quest and it could prove difficult to isolate all of these steps from each other since early capsid modifications in the endosomal system could have major consequences on every downstream event, such as, for instance, the uncoating in the nucleus. The recent investigations about AAV capsid phosphorylation, which culminated in the discovery of tyrosine mutant vectors demonstrate that every discovery in the field of AAV trafficking can potentially translate into rapid and major improvements in gene therapy applications since it could allow to reduce dramatically the vector doses¹¹¹ or to transduce cell types refractory to wild-type capsid vectors¹¹². Although it may seem counter-intuitive that targeted modification of evolutionary conserved surface residues improves viral transduction, one should keep in mind that AAV vectors are defective viruses. AAVs have evolved to co-infect target cells with a helper virus, and they are likely not optimized for autonomous infection. Almost every step of AAV transduction, from endosomal processing to transgene expression, might be enhanced by adenovirus co-infection. This means that AAV vectors offer room for improvement and might be “perfectible” by genetic manipulation of the capsid. This notion is supported by a growing body of evidence showing that directed evolution of AAV capsid from random mutants or shuffling libraries can create vectors displaying much higher transduction efficiency and/or specificity for selected cell types (as reviewed by Xiao Xiao and David Schaffer in this issue). At first sight, AAV may seem an inefficient virus with recombinant AAV exhibiting a fairly high particle-to-infectivity ratio¹¹⁸, but recent studies indicate that in the presence of Adenovirus, infectivity of wild-type AAV2 can approach a physical-to-infectious ratio of one, defining it as a near-perfect virus¹¹⁹. Recombinant AAV2 is nearly 100-fold less efficient than its wild-type counterpart in the same assay, tantalizingly opening the possibility that the sequence or secondary structure of the DNA packaged into AAV capsid may also influence transduction. If this hypothesis was confirmed, determination of sequence-specific DNA-capsid interactions could have great consequences for the design of AAV gene therapy vectors, as it could influence both the production and the infectivity of recombinant viruses, as previously observed for autonomous parvoviruses^{120, 121}.