Adeno-associated viral vector-mediated immune responses: Understanding barriers to gene delivery

Abstract

Adeno-associated viral (AAV) vectors have emerged as the leading gene delivery platform for gene therapy and vaccination. Three AAV-based gene therapy drugs, Glybera, LUXTURNA, and ZOLGENSMA were approved between 2012–2019 by the European Medicines Agency and the United States Food and Drug Administration as treatments for genetic diseases hereditary lipoprotein lipase deficiency (LPLD), inherited retinal disease (IRD), and spinal muscular atrophy (SMA), respectively. Despite these therapeutic successes, clinical trials have demonstrated that host anti-viral immune responses can prevent the long-term gene expression of AAV vector-encoded genes. Therefore, it is critical that we understand the complex relationship between AAV vectors and the host immune response. This knowledge could allow for the rational design of optimized gene transfer vectors capable of either subverting host immune responses in the context of gene therapy applications, or stimulating desirable immune responses that generate protective immunity in vaccine applications to AAV vector-encoded antigens. This review provides an overview of our current understanding of the AAV-induced immune response and discusses potential strategies by which these responses can be manipulated to improve AAV vector-mediated gene transfer.

1. Introduction

Adeno-associated virus (AAV) is a small non-enveloped virus from the Dependovirus genus of the Parvoviridae family (Balakrishnan & Jayandharan, 2014; Weitzman & Linden, 2011). Its virions are comprised of a ~4.7Kb linear, single-stranded, DNA genome encapsulated in a 25-nm icosahedral capsid (Balakrishnan & Jayandharan, 2014). AAV was first isolated in 1965 (Atchison, et al., 1965; Melnick, et al., 1965), and while it cannot autonomously replicate on its own, AAV is able to infect a wide range of host cells including both dividing and nondividing cells (Balakrishnan & Jayandharan, 2014). Following infection, the AAV genome remains latent in the host cell, integrated into the host cell DNA at adeno-associated virus integration sites (AAVS1). Co-infection with a helper virus such as adenovirus or herpes simplex virus (HSVs) is required for productive AAV replication and infection (Geoffroy & Salvetti, 2005). While AAV infection is common in humans, it is not associated with any known illness (Calcedo, et al., 2011; Erles, et al., 1999; C. Li, et al., 2012). The fact that AAV is non-pathogenic in humans yet able to transduce a wide range of cell types makes AAV an ideal candidate as a vehicle to deliver genetic content for gene therapy or vaccination (Nakai, et al., 2001; Nieto & Salvetti, 2014).

Recombinant AAV (rAAV) were first generated in the 1980s and since then have become the leading gene transfer tool for gene therapy (Weitzman & Linden, 2011). These rAAV are constructed by replacing the wild-type (WT) AAV genome with the gene of interest flanked by two inverted terminal repeats (ITRs). The rep gene, which encodes proteins involved in viral replication, and the cap gene, which encodes the structural proteins that make up the viral capsid, are replaced with the exogenous DNA of interest or “transgene” which remains flanked by the two ITRs (Grieger & Samulski, 2012; Wright, 2008) (Fig. 1). These rAAV vectors can be modulated to package a gene, or genes, of interest by molecular cloning in a relatively straightforward fashion (Y. L. Liu, et al., 2003; Robert, et al., 2017). rAAV particles containing the transgene expression cassette can be generated by transient transfection of mammalian cells (usually HEK293 cells) utilizing plasmids that express (1) the transgene(s), (2) the AAV rep and cap genes, and (3) the adenovirus helper genes required for AAV replication, respectively (Robert, et al., 2017; Wright, 2008) (Fig. 1B). Alternative ways for rAAV particle production utilizes transduction of all the required components into insect Sf9 cells by recombinant baculoviruses, or mammalian cells by recombinant HSVs (Penaud-Budloo, et al., 2018).

Figure 1. Recombinant AAV production.

(A) The wild-type (WT) AAV genome can be modified by replacing the viral replication (rep) and structural (cap) genes with exogenous DNA of interest (transgene). The transgene along with promoter and regulatory elements (transgene expression cassette) is cloned between the two inverted terminal repeats (ITRs) in the WT AAV genome to generate recombinant AAV (rAAV) genome. (B) rAAV particles are produced by co-transfecting permissive cells (usually HEK239 cells) with plasmids that contain the rAAV genome, rep and cap genes, and Adenovirus helper genes.

While the WT AAV genome integrates into the host genome after transduction, recombinant AAV transgenes have conversely been shown to remain episomal in transduced cells (Grieger & Samulski, 2012; Smith, 2008; Valdmanis, et al., 2012). These rAAV delivered transgenes are stably retained within transduced cells leading to long-term persistent gene expression, which, in theory, can last for multiple years following rAAV delivery (Nathwani, et al., 2014; Nathwani, Rosales, et al., 2011; Niemeyer, et al., 2009).

Studies show that rAAV can safely and effectively deliver genes to target cells with several preclinical and clinical trials demonstrating positive results in the treatment of genetic diseases such as hemophilia, lipoprotein lipase deficiency (LPLD), inherited retinal disease (IRD) and spinal muscular atrophy (SMA) (Gaudet, et al., 2013; Mingozzi & High, 2011; Nathwani, Tuddenham, et al., 2011). These successes led to the licensing of rAAV-based products in Europe, Glybera in 2012, and the United States, LUXTURNA in 2017 and ZOLGENSMA in 2019, as treatments for LPLD, IRD, and SMA, respectively (Buning, 2013; Hoy, 2019). Compared to other viral vectors, rAAV exhibits a superior safety profile with a relatively low risk of genotoxicity from the insertion of transgenes into the host genome (Mingozzi & High, 2011; Vandendriessche, et al., 2007).

Host immune responses during rAAV administration have been reported to limit long-term transgene expression in humans (Colella, et al., 2018; Manno, et al., 2006; Mingozzi & High, 2011, 2013). Previous exposure to WT AAV, which occurs in over 90% of the human population, can lead to the development of pre-existing immune responses that subsequently inhibit the clinical efficacy of specific serotypes of rAAV (C. Li, et al., 2012). This was highlighted in a clinical trial utilizing a rAAV delivering Factor IX (FIX) for the treatment of hemophilia B, where FIX expression was either prevented or lost around 6–8 weeks following rAAV administration, coinciding with the presence of either humoral or cellular immune responses against AAV capsid proteins (Manno, et al., 2006). As rAAV retain key features of the parent virion, including a similar if not exact viral capsid, they can stimulate host anti-viral immune responses against the rAAV capsid and/or the encoded transgene product (Basner-Tschakarjan & Mingozzi, 2014; Brantly, et al., 2009; Mingozzi, et al., 2009; Mueller, et al., 2013). This response includes pre-existing neutralizing antibodies (NAbs), priming and expansion of pre-existing antigen-specific immune cells, and/or the stimulation of naïve immune responses (Fig. 2). These responses are a major obstacle preventing long-term gene expression from rAAV and their safe use clinically (Herzog, 2015; Manno, et al., 2006; Nathwani, Rosales, et al., 2011; Nathwani, Tuddenham, et al., 2011). Protocols to evaluate the antibody, T cell, and cytokine responses to rAAV have been recently reviewed (Calcedo, et al., 2018). While these established techniques allow for evaluation of rAAV immunogenicity, strategies that focus on rational design of novel rAAVs will need to consider the molecular and cellular mechanisms that dictate these immune responses. Using coagulation factor IX as an example of transgenes, Herzog recently discussed the complexity of immune responses to rAAV-encoded transgene products (Herzog, 2019). Here we aim to summarize our current knowledge of rAAV-mediated immune responses, with a focus on the mechanisms of rAAV immunogenicity, along with our perspectives of strategic development to tune the balance between rAAV-mediated effector and regulatory immunity.

Figure 2. Immune responses limiting rAAV transduction and expression of transgenes.

(A) During rAAV delivery, pre-existing anti-capsid antibodies, previously generated from natural infection with AAV, can neutralize rAAV particles and prevent cellular transduction. (B) rAAV particles that evade antibody neutralization, interact with receptors and co-receptors on the cell surface to initiate their uptake and trafficking through the endosomal network. Endosomal escape of rAAV allows recombinant DNA to be delivered to the nucleus, where it usually persists episomally, and from which transgenes can be expressed to yield therapeutic protein products. (C) Capsids exposed to the cytoplasm after endosomal rupture can be degraded by the proteasome and processed in the endoplasmic reticulum (ER) for cross-presentation by MHC I. (D) Antigens from endogenously expressed transgene products can similarly be processed and presented by MHC I. (E) During transduction, rAAV can also trigger innate immune responses from Toll-like receptors on the cell surface (TLR2), within the endosome (TLR9), and cellular stress responses, including the unfolded protein response (UPR). (F) Innate immune responses induced by rAAV can lead to the production of proinflammatory cytokines, chemokines, and type I IFN, which recruit immune cells, such as neutrophils, macrophages, and dendritic cells (DCs), to rAAV-targeted cell locations. (G) These responses can culminate in the stimulation of humoral and cellular adaptive immune responses against rAAV capsids and encoded transgene products. (H) Transduced cells displaying rAAV antigens restricted by MHC I can activate pre-existing anti-capsid memory CD8⁺ T cells, previously generated from natural infection with AAV, or de novo rAAV antigen-specific CD8⁺ T cells, primed from naïve T cell populations. Activated CD8⁺ T cells can expand, differentiate, and facilitate cytotoxic effector responses, resulting in the loss of transduced target cells and expression from the transgenes they contain.

While immune responses to the rAAVs can be detrimental to the sustained production of rAAV encoded transgene products during therapeutic treatment of genetic diseases, immune stimulation is necessary for generating immunity to vaccine antigens. Therefore, new insights into the immunogenic properties of rAAVs have sparked interest in their potential use as vaccine vehicles. rAAV represent an excellent platform for genetic vaccination as they can efficiently deliver transgene encoded antigens to target cells and induce immune responses that can serve to adjuvant immunity against the antigen genes they encode (M. A. Liu, 2010; Nieto & Salvetti, 2014). It has been shown that rAAV vaccines are capable of stimulating both humoral and cellular responses to a variety of encoded antigens via multiple immunization routes (Du, et al., 2008; D. W. Liu, et al., 2005; Manning, et al., 1997; Xin, et al., 2002; Xin, et al., 2001). However, while their lack of pathogenicity makes rAAV a safer vaccine platform (as vector and adjuvant), they are inherently less immunogenic compared to other viral vectors (Vandendriessche, et al., 2007; Zaiss, et al., 2002). The relatively mild immunogenic profile of rAAV corresponds with a reduced ability to generate robust functional CD8⁺ T cell responses (J. Lin, et al., 2009; S. W. Lin, et al., 2007; Ploquin, et al., 2013), while antigen-specific CD8⁺ T cell responses are important for the generation of vaccine-elicited protective cellular immunity (M. A. Liu, 2010; Nieto & Salvetti, 2014).

As a result of these properties of AAV, strategies to prevent immune stimulation by rAAV are critical for the effective use of rAAV in gene therapy applications, while strategies to enhance the immunogenic properties of rAAV are important for their effective use in vaccination. Understanding how rAAV interacts with the host-immune system is critical for revealing avenues through which we can modulate these responses to generate improved genetic therapies and vaccines. In this review article, we summarize how AAV-mediated immune responses, including pre-existing neutralizing antibodies, innate and adaptive cellular immune activities interfere with AAV-mediated gene delivery, along with our perspectives of strategic designs to tailor immune responses to favor AAV-mediated therapies.

2. Pre-existing neutralizing antibodies (NAb) against AAV

In order to effectively reach and transduce target cells, rAAV must avoid neutralization by anti-capsid antibodies, previously generated from natural exposure to WT AAV. About 70% of the population has circulating NAbs directed against prevalent AAV serotypes, AAV1 and AAV2 (Calcedo, et al., 2009; Kotterman & Schaffer, 2014). Immunity to AAV is often generated in childhood by the age of 2 years (Calcedo, et al., 2011; Erles, et al., 1999; C. Li, et al., 2012). However, newborns often have maternal anti-AAV antibodies that are transferred from the mothers during pregnancy and nursing (Calcedo, et al., 2011; Erles, et al., 1999; C. Li, et al., 2012). This leaves only a small window of time after birth at around 7–11 months of age in which the seroprevalence of anti-AAV NAbs is low or undetectable (Calcedo, et al., 2011). In order to achieve effective gene transfer, several approaches have been investigated to overcome issues of rAAV neutralization by pre-existing NAbs. Increases in rAAV dose or co-treatment with an excess of empty capsids can improve transduction efficiency by overcoming vector neutralization, likely by saturating NAb binding sites (Mingozzi, et al., 2013; Mingozzi & High, 2013). However, this also increases antigen load which can contribute to the activation of immune responses, including cytotoxic T lymphocyte (CTL)-mediated clearance of transduced target cells, that prevent long-term transgene expression (Mingozzi & High, 2013; Pei, et al., 2018). Natural shielding by extracellular vesicles (EVs, referred to as exosome more recently) during rAAV production provides an envelope to compartmentalize rAAVs, and help these EV-associated rAAVs evade recognition and neutralization by the NAbs, resulting in higher transduction efficiency (Gyorgy, et al., 2014; Gyorgy, et al., 2017; Meliani, et al., 2017). EV-rAAV holds strong promise in gene therapy. The entry mechanisms are however different for enveloped rAAVs and naked rAAVs, and so our prior knowledge of AAV entry pathway and serotype-associated target specificity may not be applicable to EV-associated rAAV, Moreover, the potential cytotoxicity and immunogenicity towards the innate cells and CTLs of EV-rAAV are yet to be further defined.

Alternative approaches have focused on removing NAbs from patients by plasmapheresis prior to rAAV treatment (Chicoine, et al., 2014; Monteilhet, et al., 2011; Tse, et al., 2015). While plasmapheresis, an extracorporeal technique to remove large molecules from the plasma, has been shown to lower circulating Ab titers, it does not eliminate them completely and often requires several treatments to reduce NAb levels enough for rAAV gene transfer to be effective (Monteilhet, et al., 2011). Other strategies focus on evading NAb recognition by either chemically or genetically modifying the AAV capsid to prevent NAb binding (Jang, et al., 2011; Lee, et al., 2005; Tse, et al., 2015). Chemical modifications often involve linking polymers directly to the AAV capsid or engulfing them in polymer gel (Tse, et al., 2015). While this masks rAAV from being recognized by the NAbs, it can also interfere with receptor binding and internalization, reducing rAAV transduction efficacy (Jang, et al., 2011; Lee, et al., 2005). More promising efforts focus on genetically modifying rAAV to circumvent NAb detection, while retaining the efficacy of transduction.

To date, over 100 AAV variants have been isolated across different species and while the diversity across AAV serotypes corresponds with distinct differences in cellular tropism, the capsid protein amino acid sequences remain highly conserved (Asokan, et al., 2012; Tse, et al., 2015). Therefore, anti-AAV NAbs often exhibit cross reactivity over a wide range of AAV serotypes (G. P. Gao, et al., 2002; Mingozzi & High, 2013). Even low prevalence serotypes, AAV5 and AAV8, that have more variable capsid sequences can be detected by anti-capsid NAbs in ~40% of the human population (Kotterman & Schaffer, 2014; Mingozzi & High, 2013). This makes it unlikely that simply switching the rAAV serotype alone would circumvent NAbs. Rather more drastic alterations to the capsid structure are likely required to evade NAb detection. Multiple strategies, either rational design or directed evolution have been applied to engineer the AAV capsids to evade detection and neutralization by NAbs. PCR-based DNA shuffling of multiple AAV capsids creates novel chimeric AAVs that are able to evade NAbs (Kotterman & Schaffer, 2014). However, Abs that target conserve domains on AAV are not affected. Structural analysis has mapped to the antigenic epitopes on the AAV capsid and have identified many of the residues involved in NAb binding (Drouin & Agbandje-McKenna, 2013; Kotterman & Schaffer, 2014; Tse, et al., 2015). This allows for newer rational design approaches to be used to engineer synthetic rAAVs (Bartel, et al., 2011). Point mutations, insertions and deletions of the key antigenic epitopes successfully reduce the NAb recognitions but at the same time might affect tissue tropism as the antigenic footprint overlapped with the receptor binding footprint. Recent studies have combined structural information with directed evolution to generate synthetic variants of AAV that can evade NAbs without compromising receptor binding or transduction (Tse, et al., 2017). These results are promising for overcoming pre-existing NAbs to WT AAV. However, capsid-modified AAV will likely stimulate inflammatory immune responses that generate new NAbs and prevent therapeutic effects of their subsequent administrations. Therefore, multiple distinct capsid variant rAAVs may be needed for therapies requiring multiple doses or prime-boost vaccination. Alternatively, attenuating inflammatory responses that lead to the generation of NAbs during primary rAAV administration may allow the same rAAV to remain effective during multiple treatments. Current immunosuppressive techniques often involve non-specific drugs such as steroids that globally lower inflammation. Therefore, knowledge of how rAAV interacts with the host immune system is needed to identify targeted therapies capable of tailoring host immune responses during rAAV treatment.

3. AAV-mediated innate immunity

The innate immune response is the first line of defense against viral infection. Viral pathogens are sensed by pattern recognition receptors (PRRs) that detect pathogen-associated molecular patterns (PAMPs) present on the incoming virions or viral nucleic acids (Rogers, et al., 2011). These sensors activate signal transduction events to limit viral spread and stimulate pathogen-specific adaptive immune responses. Compared to other viral vectors, rAAV induces a relatively mild but detectable innate immune response (Zaiss, et al., 2002). Nevertheless, this rAAV-induced innate immunity has been documented as an important obstacle for both long-term transgene expression and the stimulation of a robust immune response to vector-expressed vaccine antigens (Colella, et al., 2018; Kuranda, et al., 2018; M. A. Liu, 2010; Manno, et al., 2006; Mingozzi & High, 2011, 2013; Nieto & Salvetti, 2014; Shao, et al., 2018).

PRRs are often stimulated by DNA viruses and viral vectors, including rAAV, during cellular transduction. Both AAV and rAAV mediate cell entry by binding receptors and coreceptors on the cell surface allowing for endocytosis (Bartlett, et al., 2000). It is also speculated that differences in tropism allow certain AAV serotypes to interact with receptors on professional antigen presenting cells (APCs), facilitating their engulfment by pinocytosis (Harbison, et al., 2008). While at the cell surface, viral capsid proteins can potentially stimulate Toll-like receptors (TLRs, such as TLR2 and TLR4), which may detect certain viral lipoproteins and glycoproteins (Barton, 2007; Kawai & Akira, 2007). Following endocytosis, AAV particles traffic through the endosome where capsid dissociation can expose the genome to TLR9, which recognizes CpG DNA (Kawai & Akira, 2007). TLR engagement initiates signaling through MyD88 or TRIF to activate nuclear factor-κB (NF-κB) and/or interferon regulatory factors (IRFs), which stimulate production of proinflammatory cytokines (Kawai & Akira, 2007). TLR2 has been reported as the cell surface sensing venue for AAV capsids in human liver cells (Hosel, et al., 2012), while TLR9 is indispensable for sensing intracellular AAV genetic contents in antigen presenting cells (Zhu, et al., 2007, 2009).

In addition to TLRs, rAAV can potentially stimulate soluble PRRs in the cytoplasm. Endosomal trafficking results in delivery of the AAV genome to the nucleus where transgene expression can ensue. It is generally accepted that AAV must escape the endosome and enter the cytoplasm in order to traffic to the nucleus (Ding, et al., 2005; Weitzman & Linden, 2011). However, the mechanisms by which AAV reaches the nucleus, uncoats, and delivers its DNA genome remain controversial (Ding, et al., 2005; Weitzman & Linden, 2011). Other DNA viruses, such as adenovirus, have been shown to undergo partial capsid disassembly and endosomal rupture prior to nuclear entry (Silvestry, et al., 2009). While these events are often necessary for DNA delivery, they can also trigger cytosolic immune responses. Endosomal rupture releases cellular cathepsins, that act as a danger signal inducing activation of the NLRP3 inflammasome and production of proinflammatory cytokines including IL-1β and IL-18 (Barlan, Danthi, et al., 2011; Barlan, Griffin, et al., 2011; McGuire, et al., 2011). Additionally, exposure of viral DNA to the cytoplasm can stimulate cytosolic PRRs, such as DAI and cGas, which also activate NF-κB and IRF3 mediated production of proinflammatory cytokine and IFN I production (Lam, et al., 2014; Takaoka, et al., 2007). In the mist of viral replication, RNA may be a replication intermediate that can further activate innate immune responses through cytosolic viral RNA sensors such as RIG-I, MDA5 and LGP2, further resulting in activation of NF-κB and IRF signaling pathways (Vabret & Blander, 2013). Both NF-κB and IRF signaling pathways can stimulate expression of a multitude of downstream genes that induce an anti-viral state and activate anti-viral adaptive immune responses. However, the precise mechanisms by which rAAV stimulate innate immune responses and how those responses impact adaptive immunity to vector antigens and transgenes remain largely unclear.

rAAV-mediated innate immunity is largely dependent on TLR9 signaling, which further influences adaptive immune responses thereafter. The rAAV genome containing self-complementary DNA simulates greater TLR9-dependent cytokine responses and CD8⁺ T cell responses, compared to rAAV with single-stranded DNA (Martino, et al., 2011). In contrast, removal of CpG sequences from the rAAV genome results in attenuated CD8⁺ T cell responses (Faust, et al., 2013). Furthermore, TLR9 deletion or blockade inhibits proinflammatory cytokine production, immune cell infiltration, and anti-capsid CD8⁺ T cell responses following rAAV treatment (Faust, et al., 2013; Rogers, et al., 2011; Rogers, et al., 2017; Zhu, et al., 2009). In addition, rAAV capsids have been shown to activate the host TLR2 (Hosel, et al., 2012) and unfolded protein response (UPR), with capsid variants exhibiting different levels of activation likely due to differences in cellular entry (Balakrishnan, et al., 2013). UPR activation may further enhance TLR signaling and activation of NF-κB pathways (Balakrishnan, et al., 2013; Pellegrino, et al., 2014). During the late stage after rAAV transduction, dsRNA is produced as an AAV genome-derived replication intermediate, stimulates intracellular dsRNA sensors including MDA5 and RIG-I, and promotes type I IFN production (Shao, et al., 2018). Blocking dsRNA activation pathways, including MDA5 and MAVS, can inhibit IFNβ expression from rAAV-transduced cells and increase transgene expression (Shao, et al., 2018).

The innate immune sensors triggered by rAAV components largely mediate anti-viral signaling through production of proinflammatory cytokines. These molecules in turn facilitate immune cell recruitment and activation which allows for activation and expansion of anti-transgene and/or anti-capsid adaptive immune cells, in particular CD8⁺ T cells, in vivo (Fig. 2). In the next sections, we discuss how rAAV regulates the activation, development and function of these adaptive immune cells and how they affect rAAV function and rAAV-mediated therapeutic effects.

4. AAV-mediated interactions between APCs and T cell

Differences of the AAV serotypes in cell entry, cell tropism or interaction with host innate immune factors can further dictate adaptive immune responses in hosts that received rAAV transfer. For example, different AAV capsids that may exploit alternative entry mechanisms could induce variable immune responses to rAAV (Mays, et al., 2009). The structural difference in two capsid variants, AAV8 and AAVrh32.33, influences transgene expression and activation of both anti-capsid and anti-transgene CD8⁺ T cell responses (Mays, et al., 2009). AAVrh32.33 capsids are able to stimulate more robust T cell responses, with the domain responsible mapped to hypervariable regions of the VP3 capsid protein (Mays, et al., 2013). Similarly, differences in VP3 motifs are also attributed to the greater capsid-specific T cell response induced by AAV2 compared to AAV8 vectors (Vandenberghe, et al., 2006). Alternatively, the tropism of AAV serotype in antigen presenting cells (APCs) may dictate the priming of functional T cell responses. Certain serotypes, including AAV1, have been shown to transduce dendritic cells (DCs), resulting in greater transgene immunogenicity (Lu & Song, 2009). Vaccine studies using multiple rAAV encoding HIV-1 Env pg160 with capsids from AAV serotypes 1–9 resulted in different magnitudes of HIV-specific CD8⁺ T cell response (Xin, et al., 2006). AAV2/5-gp160 induced the greatest HIV-specific immune response, which is correlated with an increased tropism for DCs (Xin, et al., 2006). Moreover, interactions between AAV and host factors may interfere rAAV uptake by the host cells and affect host immune stimulation. For example, AAV2 capsids bind complement iC3b, which can increase rAAV uptake by APCs and stimulate proinflammatory cytokine responses to further dictate adaptive immunity (Zaiss, et al., 2008).

While antigen-specific memory T cells can be re-activated upon recognition of antigen presented by any cell type, APCs mediate the activation of AAV antigen-specific T cells from both memory and naïve compartments (C. Li, et al., 2013; Mingozzi & High, 2013; Schmidt & Mescher, 2002). Therefore, APCs potentially have a large role in shaping cellular immune responses to rAAV.

APCs can take up rAAV antigens and/or protein products expressed by rAAV-encoded transgenes through transduction or phagocytosis and present them on either MHC I or MHC II (Fig. 3). Following rAAV delivery, transduced cells that express transgene products are able to subsequently process and directly present intracellular transgene peptides in the context of MHC I. Conversely exogenous proteins taken up through transduction or phagocytosis, including the structural proteins that make up the viral capsid and transgene products expressed by other cells, can be processed in the cytosol and cross-presented onto MHC I or processed within the endosome for direct presentation on MHC II. APCs presenting rAAV antigens on their surface can activate CD8⁺ T cells via MHC I or CD4⁺ T helper cells via MHC II thereby driving the T cell response to rAAV treatment. Several studies have sought to define the mechanisms contributing to the generation of anti-capsid and anti-transgene responses following rAAV treatment in order to understand how T cell immunity can be modulated to improve the clinical efficacy of rAAV.

Figure 3. APC processing and presentation of rAAV antigens.

(A) rAAV can transduce APCs via receptor mediated endocytosis, depicted in blue. Capsid fragments from rAAV can be processed in the endosome and directly presented by MHC II. Endosomal escape of rAAV allows recombinant DNA to be delivered to the nucleus, where it can be expressed to yield therapeutic protein products, and exposes rAAV capsids to the cytosol. (B) APCs can phagocytose rAAV or secreted transgene products, depicted in red. Exogenous antigens can be processed in the endosome and presented by MHC II. Endosomal escape results in exposure of exogenous antigens to the cytosol. (C) Cytoplasmic processing of rAAV antigens, depicted in black. Cytoplasmic capsids and other exogenous rAAV antigens in the cytosol can be degraded by the proteasome and processed in the endoplasmic reticulum (ER) for cross-presentation by MHC I. Similarly, endogenously expressed antigens, from rAAV encoded transgenes, can be processed and directly presented via MHC I.

TLR9 induced IFN I production and cooperation between plasmacytoid dendritic cells (pDCs) and conventional dendritic cells (cDCs) have been shown to play an important role in priming anti-capsid and anti-transgene CD8⁺ T cell responses (Brewitz, et al., 2017; Rogers, et al., 2017; Rogers, et al., 2015). The amount of IFN I produced from TLR9 activation was shown to differ between AAV serotypes but not between different encoded transgenes (Zhu, et al., 2009). Whether this contributes to differences in immunogenicity between serotypes in vivo is unknown. Furthermore, Tlr9^−/− mice generate reduced numbers of transgene-specific CD8⁺ T cells, following rAAV administration, but similar proportions of polyfunctional transgene-specific CD8⁺ T cells suggesting other pathways also contribute to the priming of anti-transgene and anti-capsid CD8⁺ T cells (Rogers, et al., 2015). While pDCs are responsible for producing IFN I in response to rAAV in vitro, Kupffer cells (KCs) seem to be required for CD8⁺ T cell recruitment to the liver where gene therapy is being targeted (Martino, et al., 2011; Zhu, et al., 2009). Therefore, multiple cell types may contribute to activating immune responses to rAAV, allowing for the priming and recruitment of CD8⁺ T cells. However, the exact mechanisms may vary among rAAV serotypes and under different tissue environments.

CD4⁺ T cell help can aid in the priming of functional CD8⁺ T cell responses. While the role of CD4⁺ T cells remains controversial for the induction of CD8⁺ T cells responses to rAAV, both anti-capsid and anti-transgene CD8⁺ T cell responses can be primed in a CD4⁺ T cell dependent manner (Mays, et al., 2009; Sarukhan, et al., 2001). A high proportion of polyfunctional effector, effector memory, and central memory CD8⁺ T cells are produced via this mechanism (J. Lin, et al., 2009). Sufficient recruitment and activation of DCs, upon primary exposure, is required to stimulate the CD4⁺ T cell dependent response(Mays, et al., 2014). In classical APC licensing, CD40L on CD4⁺ T cells interacts with CD40 on DC to facilitate DC stimulation on CD8⁺ T cells by upregulating CD80/86, inflammatory cytokines, MHC molecules, and adhesion molecules on DCs (Cella, et al., 1996; Ridge, et al., 1998; Schuurhuis, et al., 2000). However, co-administration of a CD40 agonist to Cd40l^−/− mice did not rescue the CD8⁺ T cell response, suggesting that CD40L is not the sole mechanism that CD4-dependent help prime CD8⁺ T cell responses to AAV infection (Mays, et al., 2009). It is also important to state that the structure of the capsid is capable of driving the CD4-dependent transgene-specific and capsid-specific CD8⁺ T cell response, however the immunogenicity of the transgene itself could prime CD8⁺ T cell independently of CD4⁺ T cell help (Mays, et al., 2009; Sarukhan, et al., 2001).

Therefore, many factors likely contribute to the priming of functional CD8⁺ T cell responses to rAAV including the stimulation of innate immune responses, activation by APCs and help from CD4⁺ T cells (Fig. 4). However, differences in AAV serotypes, rAAV doses, transgene immunogenicity, and routes of delivery may all impact the immune threshold needed to stimulate functional CD8⁺ T cell responses. Alternative immune responses to rAAV may occur, which could lead to tolerance and transgene-persistence.

Figure 4. Priming of T cell responses to rAAV.

Proposed model of APC priming T cell responses to rAAV: (A) Plasmocytoid dendritic cells (pDCs) trigger innate immune responses to rAAV through TLR9, which recognizes CpG sequences within recombinant DNA, resulting in production of type I IFN. (B) Type I IFN is required to license conventional dendritic cells (cDC) for antigen-presentation and T cell priming. (C) CD4⁺ T cells that are activated by MHC II restricted antigens and CD40/CD40L interactions also contribute to APC licensing. Activated CD4⁺ T cells can stimulate antigen-specific B cell responses, including antibody production. (D) Licensed APCs present rAAV antigens via MCH I to CD8⁺ T cells. TCR engagement and interactions between CD28 with MHC I presented antigens and B7 co-receptors on APCs prime antigen-specific CD8⁺ T cells. (E) Antigen-specific CD8⁺ T cells can recognize target cells displaying rAAV antigens on their surface, resulting in the cytotoxic-mediated destruction of rAAV transduced cells and loss of expression from the transgenes they contain.

5. Immunity to rAAV-encoded transgenes

In rAAV gene replacement therapies and vaccines, transgene products are often foreign antigens, which can stimulate host immune responses that lead to generation of transgene-specific antibodies and cytotoxic T lymphocytes (CTL). These humoral and cellular immune responses can neutralize extracellular transgene products, prevent transgene expression, and eliminate transgene expressing cells, that present endogenous transgene antigens on MHC I (Basner-Tschakarjan & Mingozzi, 2014; Boisgerault & Mingozzi, 2015) (Fig. 2). Transgene-specific CTL cells have been found in patients following intramuscular delivery of rAAV, in several preclinical trials, and in genetic immunization studies in mice (Flotte, et al., 2011; Manning, et al., 1997; Mays & Wilson, 2011; Mingozzi & High, 2013).

Transgene-specific CD8⁺ T cell responses have been shown to vary following treatment with different rAAV capsid pseudotypes. In vaccine studies rAAVrh32.33 stimulating a greater proportion of transgene-specific memory T cells and transgene-specific T cell that could functionally produce multiple cytokines compared to rAAV8, which correlated with their ability to expand following subsequent transgene-product exposure boosted with an adenovirus vector (J. Lin, et al., 2009). This suggests that AAV structural differences alone can account for the quality of CD8⁺ T cell response, primed by rAAV. In addition to the molecular and cellular properties of rAAV-encoded genes and interacting host immune cells, such as transgene DNA composition and APC transducing ability, route of administration and targeted tissue also contribute to the magnitude of subsequent anti-transgene immune responses (Brockstedt, et al., 1999; Mingozzi, et al., 2003). Intramuscular delivery methods are shown to generate more robust CTL responses to rAAV transgenes compared to rAAV delivered to more tolerant tissues environments such as the Liver (Mingozzi, et al., 2003; Wang, et al., 2005). These differences may help to explain why some preclinical trials have failed to generate long-term transgene expression while others find lasting transgene durability and an absence of anti-transgene responses (Nathwani, Rosales, et al., 2011; Niemeyer, et al., 2009). However interestingly, a recent study using ovalbumin (OVA) as model transgene found that concurrent muscle and liver targeting by OVA-expressing rAAV resulted in systemic OVA-specific CD8⁺ T cell exhaustion, leading to successful transduction and expression of OVA in muscle tissue (Bartolo, et al., 2019). After dual muscle-liver transduction, polyclonal OVA-specific CD8⁺ T cells in the blood and spleen expressed high levels of PD-1 and lacked IFN-γ production, despite the presence of preexisting immunity to OVA and successful induction of OVA-specific CD4⁺ T cells (Bartolo, et al., 2019). These results suggest that it may be possible to exploit the tolerogenic environment of some particular tissues, such as the liver, to induce transgene-specific CTL exhaustion systemically and help establish tolerance to the transgene in other tissues.

6. Capsid-specific CD8⁺ T cell responses

Many rAAV gene therapies fail to generate long-term transgene expression in humans despite their previous success in animal models. This loss of transgene expression was linked to capsid-specific CTL responses (Manno, et al., 2006; Mingozzi, et al., 2007; Mingozzi, et al., 2009; Nathwani, Tuddenham, et al., 2011). For example, rAAV encoding Factor IX (FIX) to treat hemophilia B (by transduction of hepatocytes) resulted in a reduction in FIX levels over time that was accompanied by a rise in liver transaminases and an expansion of capsid-specific IFN-γ⁺ CD8⁺ T cells (Manno, et al., 2006; Mingozzi, et al., 2007; Nathwani, Tuddenham, et al., 2011). Transgene expression remained stable and liver enzymes returned to normal levels following anti-inflammatory treatment by steroid prednisolone, providing further evidence of immune mediated destruction of transduced hepatocytes. Further, in vitro studies found that human hepatocytes, transduced with rAAV, display capsid peptides on MHC I and are targeted for destruction by capsid-specific CD8⁺ T cells (Pien, et al., 2009). Additionally, clinical trials transducing skeletal muscle with rAAV have also reported the loss of transgene expression, coinciding with a rise in creatine phosphokinase (CPK), indicative of muscle cell damage, which was linked to the detection of capsid-specific IFN-γ⁺ CD8⁺ T cells in peripheral blood and at the injection sites (Flotte, et al., 2011; Mingozzi, et al., 2009). Together these data suggest that the loss of transgene expression from rAAV seen in humans is due to the activation of capsid-specific CD8⁺ T cells which recognize and eliminate transduced target cells. Therefore, capsid-specific cytotoxic CD8⁺ T cell responses represent a major obstacle for achieving long-term transgene-expression in rAAV gene transfer applications.

It was initially thought that the capsid-specific CD8⁺ T cells were the result of reactivation of preexisting capsid-specific CD8⁺ memory T cells, generated during prior AAV infection(s) and subsequently stimulated by rAAV capsid antigens during therapy. However, preclinical studies using non-human primates (NHPs), which are also natural hosts for AAV, only predicted the effects of pre-existing antibody responses but not the activation of anti-capsid cellular immune responses on rAAV treatment (G. Gao, et al., 2009; Jiang, et al., 2006; Nathwani, et al., 2007; Nathwani, Rosales, et al., 2011). Although capsid-specific T cells can be detected in both humans and NHPs, capsid-specific CD4⁺ and CD8⁺ T cell frequencies, subset distribution, differentiation status and function differ substantially between human and NHP (H. Li, et al., 2011). This could be due to differences in the AAV life cycle or evolutionary differences. For example, NHP T cells express CD33-related sialic acid-binding Ig-like lectins (Siglec), a subset of inhibitory signaling molecules that regulate immune cell activation, which is lost in human T cells. Loss of these regulatory molecules could lower the activation threshold of human T cells compared to, less reactive, NHP T cells (G. Gao, et al., 2003; G. Gao, et al., 2004; Nguyen, et al., 2006; Roy, et al., 2009). These discrepancies highlight the need for appropriate animal models for evaluating rAAV as well as for further understanding of how rAAV stimulates cellular immune responses.

Capsid-specific CD8⁺ T cells can recognize and remove target cells carrying MHC I/rAAV capsid antigens (Fig. 2). Unlike transgenes products, viral capsid proteins, are not expressed from rAAV. Therefore, exogenous capsid antigens must be processed from incoming rAAV particles for target cell cross-presentation on MHC I. A putative model for cross-presentation of AAV2 capsid peptides has been proposed (Fig. 2&3) (C. Li, et al., 2013). In Brief, AAV is endocytosed, following receptor binding, in both clathrin/dynamin-mediated and clathrin-independent manners (Bartlett, et al., 2000; Nonnenmacher & Weber, 2011). AAV then traffics through the endosome where acidification leads to AAV escape into the cytosol (C. Li, et al., 2013). In the cytosol, tyrosine residues on the surface of AAV capsid can be phosphorylated, leading to subsequent ubiquitination and proteasomal degradation of the capsid (Finn, et al., 2010; C. Li, et al., 2013; Yan, et al., 2002; Zhong, et al., 2007). Peptides from the degraded capsid are transported into the endoplasmic reticulum (ER) (H. Li, et al., 2011). In the ER, processed antigens bind to MHC I and are then trafficked via the Golgi apparatus to the cell surface (C. Li, et al., 2013). These mechanisms were required for MHC I cross-presentation of capsid peptides on the surface of target cells transduced with AAV2, as confirmed by their ability to stimulate antigen-specific CD8⁺ T cells, which was prevented by pharmacological inhibitors of these processes (C. Li, et al., 2013). Defining the detailed mechanisms that lead to capsid antigen presentation following rAAV transduction may further reveal potential strategies to prevent cellular immune recognition and enable persistent transgene expression.

7. Reducing CD8⁺ T cell response to rAAV

Several strategies are being investigated to block target cell presentation of rAAV capsid antigens on MHC I. These include reducing the capsid antigen load delivered by rAAV, use of proteasome inhibitors, and mutation of surface-exposed tyrosine residues on rAAV capsids. Cells transduced with empty AAV capsids, also present capsid peptides on MHC I in vivo, although possibly by different pathways (Pei, et al., 2018) and/or at lower efficiency (C. Li, et al., 2013), compared to gene-encoding full rAAV. Several clinical trials using rAAV therapies in human patients have found a positive correlation between rAAV vector dose and capsid-specific CTL responses against transduced target cells (Flotte, et al., 2011; Mingozzi, et al., 2009; Nathwani, Tuddenham, et al., 2011). Capsid-specific CTLs can be generated and expanded from human peripheral blood mononuclear cells (PBMCs) by stimulation with target tissue cells (eg., hepatocytes) pulsed with capsid peptides or transduced with empty AAV capsids in vitro, while hepatocytes transduced with either AAV2 vectors or empty capsids could both present capsid peptides via MHC I and act as targets for destruction through cytotoxic lysis by the resultant CTLs in vitro (Pien, et al., 2009). Using this model, a capsid antigen dose dependent lysis of the targeted hepatocytes by the CTLs has been observed, ie., higher doses of capsid peptide pulsing, and AAV2 vector or empty capsid transduction in target cells all resulted in increased capsid peptide presentation and CTL killing of target cells in vitro (Finn, et al., 2010; Pien, et al., 2009). To minimize capsid-antigen load during rAAV administration, lowering the dose and/or removing empty AAV capsids that lack recombinant DNA, from vector preparations have been proposed. These approaches attempt to reduce the amount of capsids available for antigen presentation, which may reduce the CD8⁺ T cell response enough to allow long-term gene expression from transgenes. However, these approaches may also lower transduction efficiency and enhance neutralization by anti-AAV NAbs (Mingozzi, et al., 2013; Mingozzi & High, 2013). When capsid dose is lower than the neutralizing threshold by pre-existing NAbs, rAAV-mediated gene delivery and therapy would also fail. Therefore, an effective dose of rAAV capsids for gene transfer should be carefully determined within the range that is high enough to decoy NAbs but lower than the levels that can activate strong CTLs (Mingozzi & High, 2013).

Other strategies focus on inhibiting the proteasomal processing of capsid in order to prevent subsequent presentation on MHC I, with particular interest in proteasome inhibitor bortezomib, as it is currently licensed for use in humans (Colella, et al., 2018; Duan, et al., 2000; Finn, et al., 2010; C. Li, et al., 2013). The proteasome inhibitor Nacetyl-L-leucyl-L-leucyl-norleucine has been shown to prevent the degradation of internalized rAAV2 capsids in human bronchial epithelial cells (Duan, et al., 2000), while bortezomib and MG132 could reduce the presentation of AAV2 capsid peptides on human hepatocytes in vitro (Finn, et al., 2010; C. Li, et al., 2013). Furthermore, bortezomib protected transduced human hepatocytes from capsid-specific CTL killing in vitro (Finn, et al., 2010). Another benefit of proteasome inhibitors is their ability to enhance rAAV transduction by increasing nuclear translocation. This was observed in multiple human cell lines in vitro and in mice hepatocytes in vivo (Duan, et al., 2000; Finn, et al., 2010; C. Li, et al., 2013). However, there are several limitations and risks associated with the use of proteasome inhibitors. The effects of proteasome inhibitors on rAAV capsid degradation may be organ and AAV serotype specific and therefore not effective in every rAAV application (Duan, et al., 2000; Finn, et al., 2010; Monahan, et al., 2010; Nathwani, et al., 2009). Moreover, proteasome inhibitors are nonspecific and potential off-target effects from their use could limit the clinical safety of rAAV treatment. For example, bortezomib has displayed potentially serious side effects, such as neurotoxicity (Ale, et al., 2015; Ale, et al., 2014). Furthermore, proteasome inhibitors may need to be administered throughout rAAV treatment, requiring patients to be consistently monitored, which limits the utility of these treatments for gene therapy (Finn, et al., 2010; Kuhn, et al., 2009).

Alternatively, AAV capsids have been engineered by modifying surface exposed tyrosine residues, targets of phosphorylation and ubiquitination, to prevent capsid processing and presentation. AAV vectors with tyrosine modified capsids elicit reduced presentation of capsid peptides on MHC I following transduction, and require lower vector doses to achieve therapeutic levels of transgene expression (C. Li, et al., 2013; Martino, et al., 2013; Zhong, et al., 2008). This strategy was shown to reduce capsid-specific CTL-mediated destruction of transduced murine and human hepatocytes in vitro (Martino, et al., 2013). In murine models, mice that received capsid-specific CD8⁺ T cells and immunized with an rAAV2 with a tyrosine mutated capsid, rAAV2(Y-F), exhibited reduced capsid-specific CD8⁺ T cell cytotoxicity against the transduced target cells and sustainable expression of the transgene, in contrast to the mice that received the WT rAAV2 (Martino, et al., 2013). These data suggest that modifying tyrosine residues on the rAAV capsid can allow transduced target cells to avoid recognition and destruction by capsid-specific CD8⁺ T cells thereby enabling long-term transgene expression during rAAV-mediated therapy. However, the clinical efficacy of capsid modified rAAV and the precise mechanisms through which capsid-specific CTL response is induced remain to be further investigated.

Despite promising results indicating that these strategies effectively reduce capsid-specific CD8⁺ T cell responses, the kinetics of capsid degradation and presentation in vivo remain unclear. Reports related to these process are conflicting and potential variability in antigen-presentation following rAAV delivery may undermine the clinical efficacy of these strategies (Basner-Tschakarjan & Mingozzi, 2014). For example, AAV1 capsids were found to persist in human biopsies years after intramuscular injection while AAV8 capsids were only detectable in the liver of NHPs up to 6 weeks post transduction (Jiang, et al., 2006; Mueller, et al., 2013). Further, AAV8 capsids were shown to persist longer than AAV2 capsids following similar methods of systemic rAAV delivery to the liver in mice (Martino, et al., 2013). While this corroborates reports that show AAV2 and AAV8 vectors differ in their induction of CD8⁺ T cell immunity, others conversely report that AAV2 and AAV8 elicit similar kinetics of T cell activation (Basner-Tschakarjan & Mingozzi, 2014; He, et al., 2013; Manno, et al., 2006; Mingozzi, et al., 2007; Nathwani, Tuddenham, et al., 2011; Wu, et al., 2014). Therefore, further evaluation is necessary to determine whether capsid presentation differs between different AAV serotypes and whether these differences account for the variable CD8⁺ T cell responses to rAAV.

Differences in rAAV serotype, route of administration, and/or host species can impact capsid presentation processes, and affect processes of intracellular trafficking, transduction efficiency, capsid persistence, and reactivation of pre-existing capsid-specific T cell populations. Since current strategies of inhibiting capsid presentation following rAAV delivery still activate CD8⁺ T cells, antigen-presentation must still be occurring either via an alternative pathway or at levels too low to mount detectable CTL killing in vivo (Martino, et al., 2013; Rogers, et al., 2017). Any residual presentation of rAAV antigens by target cells may still lead to CTL responses that limit transgene expression (Martino, et al., 2013). Alternative antigen presentation pathways may also contribute to MHC I presentation of AAV antigens (Basner-Tschakarjan & Mingozzi, 2014; Oliveira & van Hall, 2013). Potential differences in MHC I presentation across other AAV serotypes or between species could impact the ability of activated CD8⁺ T cells to mediate lysis. Additionally, delayed antigen presentation from capsids that persist in tissue following rAAV administration could induce CD8⁺ T cell responses after the therapeutic window of co-administered drugs. Therefore, additional research is needed to fully define mechanisms of rAAV antigen presentation across variable rAAV treatment conditions.

8. T cell tolerance and exhaustion during rAAV treatment

The immune system must be highly regulated in order to rapidly respond to pathogenic infection while also preventing excessive inflammatory responses that lead to auto-immunity or chronic inflammation. Therefore, several regulatory mechanisms exist to moderate the immune response by promoting immunological tolerance to antigens. Evidence of immune tolerance to rAAV encoded transgenes has been demonstrated in certain clinical trials where patients exhibited long-term transgene expression despite generating detectable anti-capsid T cell responses (Brantly, et al., 2009; Mueller, et al., 2013; Nathwani, et al., 2014). While clinical rAAV studies currently focus on using non-specific immunosuppressive drug treatments to prevent CD8⁺ T cell elimination of target cells that successfully express the transgene, alternative approaches that specifically modulate the immune response towards tolerance are desirable, with a focus on defining the regulatory immune pathways that control the CD8⁺ T cell response to rAAV.

Studies have found that tolerance induction coincides with an expansion of antigen-specific CD4⁺CD25⁺Foxp3⁺ T regulatory (Treg) cells as well as the expression of antigen-specific T cell exhaustion markers, PD-1 and PD-L1, at inflammatory sites of rAAV delivery (Mueller, et al., 2013). Higher doses of rAAV have been reported to correlate with higher levels of transgene expression and an increase in CD4⁺CD25⁺Foxp3⁺ Treg cells (Kumar, et al., 2017; Markusic, et al., 2013; Mingozzi, et al., 2003; Mueller, et al., 2013). Kumar et al. investigated the impact of vector dose on rAAV gene delivery and stimulation of CD8⁺ T cell responses and found that delayed CD8⁺ T cell responses leading to transgene loss, occurred ~2 months after administration with low or intermediate doses of rAAV (Kumar, et al., 2017). Despite the presence of systematic antigen-specific CD8⁺ T cells that were detected 4 weeks after intermediate doses of rAAV, these cells remained unresponsive until 8 weeks post injection (p.i.) when transgene clearance began. The onset of these CD8⁺ T cell responses coincided with a reduced frequency of PD-1 expression on antigen-specific CD8⁺ T cells, from ~90% at 4 weeks (p.i.) to ~50% at 8 weeks (p.i.), and an increase in IFN-γ and TNF-α production in response to antigen re-stimulation. These studies suggest that PD-1 expression can prevent the function of antigen-specific CD8⁺ T cells, as tolerance was not observed in Pdcd1^−/− mice (Kumar, et al., 2017). However, how CD8⁺ T cells lose PD-1 and become functional months after rAAV treatment remains to be further investigated. Counter to what was observed with the low and intermediate dose groups, mice that received the high doses of rAAV exhibited persistent transgene expression and lacked CD8⁺ T cell responses (Kumar, et al., 2017). While Kumar et al. reported that blockade of PD-1/PD-L1 could not overcome tolerance in the high dose group (Kumar, et al., 2017), others showed that prolonged PD-1 expression can lead to permanent and irreversible exhaustion, and that the PD-1/PD-L1 pathway can promote differentiation and maintenance of Foxp3⁺ Treg cells (Mueller, et al., 2013; Pauken, et al., 2016; Turner & Russ, 2016). In mouse models with transient depletion of Foxp3⁺ Treg cells, or deficient in Fasl or Il10, the abundance of antigen-specific CD8⁺ T cells induced by high doses of rAAV was increased, implicating multiple pathways in the regulation of tolerance and immune suppression to rAAV. Therefore, the role of the PD-1/PD-L1 pathway in tolerance to rAAV warrants further investigation along with the evaluation of other contributing factors including Treg cells, Fas-L, and IL-10.

Treg cells can functionally suppress immune responses through a variety of mechanisms including production of immunosuppressive cytokines and cell contact-dependent suppression or killing of immune cells (Sakaguchi, et al., 2008). Treg cells induced by rAAV were shown to suppress CD4⁺CD25⁻ T cells, inhibiting their production of IL-2 in vitro and preventing antibody formation in vivo (Cao, et al., 2007; Cooper, et al., 2009). B cell tolerance to rAAV was also shown to depend on the Fas/Fas-L-mediated cell death pathway, with Fas-deficient mice exhibiting higher-titers of anti-transgene antibodies (Mingozzi, et al., 2003). Alternatively, IL-10, which is produced by Treg cells and Kupffer cells in the liver, was shown to suppress rAAV-induced CD8⁺ T cells (Hoffman, et al., 2011). High levels of IL-10 are produced from CD4⁺CD25⁺ T cells following rAAV treatment, however, due to their low frequency, the exact classification of these regulatory cells could not be determined (Breous, et al., 2009). Therefore, the role of different regulatory T cell subsets in tolerance to rAAV warrant further investigation, in particular the contribution of Foxp3⁻ type 1 regulatory T (Tr1) cells, which characteristically produce high levels of IL-10 (Roncarolo, et al., 2006). While it remains unclear how regulatory T cells are induced by rAAV, high levels of IL-4 and IL-13 have been reported following rAAV treatment (Breous, et al., 2009). IL-4 and IL-13 are associated with stimulating differentiation of naïve CD4⁺CD25⁻ T cells into CD4⁺CD25⁺ Treg cells (Skapenko, et al., 2005). However, further studies are needed to determine how IL-4 and IL-13 production is induced by rAAV and the effect these cytokines have on regulatory T cells. Further, while regulatory T cells are capable of inducing suppression through all of the mechanisms described above, contributions to tolerance from other cell types and the molecules they express cannot be ruled out. Overall, our current knowledge of these regulatory responses suggests that multiple mechanisms contribute to tolerance following rAAV delivery (Fig. 5).

Figure 5. Immune regulation of the CD8⁺ T cell response to rAAV.

Potential mechanisms of immune regulation that limit CD8⁺ effector T cell responses to rAAV and promote tolerance. (A) Tolerogenic responses from APCs, including production of IL-10, can suppress the activation and expansion of CD8⁺ T cell. (B&C) Insufficient APC licensing that occurs in the absence of CD4⁺ T cells or proinflammatory cytokines can prevent the priming of CD8⁺ T cells. (C) Stimulation of CD4⁺ T cell responses towards the differentiation and activation of regulatory T cell subsets, can suppress CD8⁺ T cells. Regulatory T cells can further regulate the levels of humoral and cellular adaptive immune responses through multiple mechanisms including cell-to-cell contact-dependent inhibition or secretion of regulatory cytokines such as IL-10. (D) CD8⁺ T cells can be stimulated to up-regulate Fas/FasL, which results in activation of cell death pathways and removal of the activated cytotoxic CD8⁺ T cells. Inadequate stimulation or prolonged antigen-exposure can cause CD8⁺ T cells to become exhausted and up-regulate immune checkpoint molecules such as PD-1. PD-1/PD-1L interactions prevent CD8⁺ T cell responsiveness. (E) A lack of responsive CD8⁺ T cells, results in persistent transgene-expression following rAAV administration.

9. Increasing “regulatory-immunogenicity” of rAAV

While the exact mechanisms responsible for generation of regulatory T cells during rAAV delivery are unknown, strategies to induce regulatory immune responses to rAAV are currently being investigated. Certain MHC class II epitopes within human IgG, termed “Tregitopes”, have been shown to induce Treg cell proliferation and suppression of Th1 and Th2 responses (De Groot, et al., 2008). These Tregitopes were used to generate rAAV, which co-express the AAV capsid protein fused to the Tregitope sequence, resulting in modulation of CD8⁺ T cell responses (Hui, et al., 2013). Tregitope expressing rAAV stimulated Treg expansion and the suppression of antigen-specific CD8⁺ T cell responses (Hui, et al., 2013). Tregitope expanded CD4⁺CD25^hiFoxp3⁺ Treg cells inhibited CD8⁺ T cell responses via a contact-dependent mechanism, requiring MHC I, that resulted in anergy of CD8⁺T cells in vitro (Hui, et al., 2013). These Treg cells likely interfere with IL-2 signaling as anergic CD8⁺ T cells can be functionally restored when they are subsequently cultured in the presence of IL-2 (Hui, et al., 2013). Alternatively, co-administration of rAAV with immunomodulatory drugs in rAAVs has been proposed as a strategy to modulate immune responses during rAAV administration. Rapamycin, encapsulated in synthetic vaccine particles (SVP), has been demonstrated with promising effects in reducing rAAV immunogenicity and enabling effective vector re-administration (Meliani, et al., 2018). Rapamycin targets the mTOR pathway and promotes expansion of functional CD25⁺ Foxp3⁺ Treg cells that can promote antigen-specific immune tolerance (Battaglia, Stabilini, Draghici, et al., 2006; Battaglia, Stabilini, Migliavacca, et al., 2006; Sauer, et al., 2008). When SVP-encapsulated rapamycin were co-administered with rAAVs, capsid-specific humoral response, T cell recall response, and CTL infiltration into the target organ were decreased, while transduction of the target cells and expression of the transgene were enhanced (Meliani, et al., 2018). These beneficial effects by SVP-encapsulated rapamycin were attenuated if CD25⁺ cells were depleted, suggesting that Treg cells may be involved in modulating rapamycin-mediated anti-AAV immunity for sustainable rAAV-mediated gene delivery (Meliani, et al., 2018). While these results provide promising evidence that support a novel approach for modulating the immune response to rAAV, further studies are needed to evaluate the clinical efficacy of Tregitope-expressing or rapamycin-assisted rAAVs, and whether they might compromise the host’s immunity against infections by natural pathogens.

10. Novel applications of rAAV in gene editing

Recent advancements in clustered regulatory interspaced short palindromic repeat (CRISPR) /CRISPR-associated protein 9 (Cas9) gene editing technology allow CRISPR components to be packaged by viral vectors, including rAAV, for targeted cellular delivery and in vivo gene editing applications (Cong, et al., 2013; Kim, et al., 2017; Ran, et al., 2015; Yin, et al., 2017). Due to previous successes in gene therapy, AAV has quickly become the leading platform for targeted gene editing by CRISPR (AAV-CRISPR). However, contrast to gene therapy, which requires lasting expression of therapeutic transgene products, gene-editing by AAV-CRISPR results in permanent alterations in the host genome. Therefore, AAV-CRISPR technology has the potential to correct or remove genetic elements beyond the limitations of conventional AAV-mediated gene therapy. AAV-CRISPR gene editing systems have been able to correct disease-causing mutations in vivo, demonstrating therapeutic success in animal models for Duchenne muscular dystrophy (Tabebordbar, et al., 2016), Huntington’s disease (Yoshiba, et al., 2019), and cervical cancer (Yoshiba, et al., 2019). As of November 2018, human clinical trials are underway using an AAV-CRISPR in vivo (EDIT-101) to correct mutations, in the CEP290 gene, causing retinal dystrophy (Maeder, et al., 2019; Mullard, 2019). However, caution will need to be taken as AAV-CRISPR gene editing may face similar drawbacks from host immune responses to those previously discussed for rAAV based gene therapies.

In addition to the rAAV-induced immune responses previously described, immune responses may be stimulated by the, bacterially derived, CRISPR components expressed from AAV-CRISPR. Pre-existing immunity to CRISPR proteins (Cas9) has been reported in humans, including the detection of antibodies against both SaCas9 and SpCas9 homologs and antigen-specific T cells to SaCas9 (Mullard, 2019). Therefore, clinical efficacy of AAV-CRISPR may be limited by immunogenic properties of the rAAV capsid and Cas9 expression. Evidence of immune responses to AAV-CRISPR was already found following intramuscular delivery of an rAAV9 based CRIPSR in mice. In these studies, antigen-specific immune responses to CRISPR Cas9 and AAV9 capsid, as well as cellular responses to CRISPR Cas9 (Chew, et al., 2016). Despite the presence of CD8⁺ T cells in Cas9-expressing tissues, no evidence of tissue damage was detected by histology (Chew, et al., 2016). However, anti-Cas9 CD8⁺ T cells have the potential to recognize and remove edited target cells expressing Cas9, which could limit therapeutic efficacy in humans. Similar strategies used to lower immunogenicity in rAAV gene therapies could be used to prevent or circumvent immune responses stimulated during AAV-CRISPR gene-editing. Additionally, AAV-CRISPR systems that limit or elicit transient expression of Cas9, such as a self-deleting AAV-CRISPR system, could potentially limit undesirable anti-Cas9 immune responses (A. Li, et al., 2019). While some of the existing clinical trials on AAV-CRISPR could potentially circumvent the aforementioned immunogenicity issues leveraging ocular immune privilege, further experiments are needed to determine how different AAV-CRISPR systems impact host immunity. This is critical for the broader application of AAV-CRISPR technology in translational settings. Results from ongoing and anticipated clinical trials should shed more light on potential immune complication of AAV-CRISPR gene-editing.

11. Conclusions

While research efforts have begun to define the mechanisms mediating rAAV immunogenicity, further studies are needed to fully characterize the immune responses stimulated by rAAV. Variable host immune responses to rAAV have been reported which correspond with differences in AAV serotype, recombinant DNA configuration, transgene immunogenicity, vector production and purification processes, rAAV dose, administration route, host species, and the host’s history of natural AAV infection. Yet, little is known about how these differences individually contribute to anti-vector immunity. Further evaluation of individual vector components is needed to identify precise vector-host interactions contributing to rAAV immunogenicity.

Current efforts mapping AAV capsid residues and motifs associated with NAb binding, APC tropism, or stimulation of CTL responses will provide critical knowledge that should allow for the rational design of rAAV that are engineered to elicit desirable immune responses specifically geared toward their application. Recent studies utilizing these data describe the generation of novel rAAVs capable of evading host immune responses and successfully expressing encoded transgenes (Martino, et al., 2013; Tse, et al., 2017). While further studies are needed to assess the clinical efficacy of these vectors, these results suggest capsid modification of rAAV is a promising strategy for overcoming host immunity. These efforts could further be utilized to design more immunogenetic rAAV capable of stimulating vaccine immunity against encoded antigens.

Current research suggests, less immunogenic rAAV platforms can be generated using rAAV that evade pre-existing immune responses, do not transduce APCs, and stimulate more dysfunctional CD8⁺ T cell responses associated with immune tolerance. Therefore, use of rAAV from AAV8, or that are modified to remove NAb binding sites and/or tyrosine residues associated with antigen processing should be considered. Recombinant DNA CpG content and conformation, as single stranded or self-complementary, should be also carefully considered. Single-stranded rAAV that are devoid of CpG may be better for evading immune detection.

Similar considerations can be made to enhance the immunogenicity of rAAV vaccines by evading pre-existing NAbs, targeting APCs for transduction, and stimulating immune responses that prime antigen-specific humoral and cellular adaptive immune responses towards rAAV-encoded antigens. Capsids from AAVrh32.33 or AAV-5, and those modified to remove NAb binding sites, or contain VP3 motifs associated with APC transduction and CD8⁺ T cell stimulation may be better for eliciting vaccine immunity. Additionally, the use of self-complementary rAAV and recombinant DNA containing CpG sequences, may be more appropriate for vaccine applications. Vaccination efforts could also include additives to adjuvant immunity to rAAV encoded antigens. Therefore, further studies should focus on assessing the effects common vaccine adjuvants have on immune responses to rAAV.

Overall, our current knowledge of the rAAV-mediated immune response leaves many questions unanswered. Yet, despite the need for further research, promising approaches exist that could modulate immune responses to rAAV and improve their clinical utility. Continued investigation of the mechanisms driving diverse immune responses to rAAV could aid endeavors to enhance rAAV efficacy in gene therapy and vaccination, as well as expand our general understanding of immune homeostasis. This knowledge could be additionally applied towards combating dysfunctional immune responses involved in cancer, autoimmunity, infections and others.

Abbreviations

AAV

Adeno-associated virus

rAAV

recombinant AAV

AdV

adenoviral vector

APC

antigen-presenting cell

CRISPR

clustered regulatory interspaced short palindromic repeat

CTL

cytotoxic T lymphocyte

DC

dendritic cell

EV

extracellular vesicle

MHC

Major histocompatibility complex

NAb

Neutralizing antibody

NHP

non-human primate

PAMP

Pathogen-associated molecular pattern

PRR

Pattern recognition receptors

TLR

Toll-like receptor

Tr1

IL-10^high Foxp3⁻ type 1 regulatory T cells

Treg

Foxp3⁺ regulatory T cells

WT

wild-type