Circumventing Antivector Immunity: Potential Use of Nonhuman Adenoviral Vectors

Abstract

Adenoviruses are efficient gene delivery vectors based on their ability to transduce a wide variety of cell types and drive high-level transient transgene expression. While there have been advances in modifying human adenoviral (HAdV) vectors to increase their safety profile, there are still pitfalls that need to be further addressed. Preexisting humoral and cellular immunity against common HAdV serotypes limits the efficacy of gene transfer and duration of transgene expression. As an alternative, nonhuman AdV (NHAdV) vectors can circumvent neutralizing antibodies against HAdVs in immunized mice and monkeys and in human sera, suggesting that NHAdV vectors could circumvent preexisting humoral immunity against HAdVs in a clinical setting. Consequently, there has been an increased interest in developing NHAdV vectors for gene delivery in humans. In this review, we outline the recent advances and limitations of HAdV vectors for gene therapy and describe examples of NHAdV vectors focusing on their immunogenicity, tropism, and potential as effective gene therapy vehicles.

Introduction

Adenoviruses (AdVs) are nonenveloped viruses with an icosahedral capsid containing a double-stranded linear DNA genome of 26–46 kb (Davison et al., 2003). Since their discovery in 1950s in human cell cultures (Rowe et al., 1953), AdVs have been identified in many species, ranging from fish to humans. Today AdVs form five accepted genera (Mastadenovirus, Aviadenovirus, Siadenovirus, Atadenovirus, and Ichtadenovirus) within the family Adenoviridae. Recently, one more genus was proposed, named Testadenovirus (Doszpoly et al., 2013). So far 57 different AdV serotypes (Lukashev et al., 2008; Walsh et al., 2011) have been isolated from humans (HAdVs) and are classified within the genus Mastadenovirus as 7 species, HAdV A through HAdV G (Jones et al., 2007). While classification was previously based on biological characteristics (erythrocyte agglutination, oncogenicity, and by specific antisera neutralization), phylogenetic relationships have been revised and clarified on the basis of genotype similarity (Bailey and Mautner, 1994; Davison et al., 2003). Accordingly, individual AdV serotypes frequently share characteristics such as tropism and associated host pathologies with other serotypes within their species. The tropism is in large part determined by the capsid proteins that interact with the host cell surface receptors to mediate cell entry. The capsid consists primarily of hexon and penton proteins that form pentameric structures at each of the 12 vertices, securing a trimeric fiber that projects outward. Other viral proteins include the cement proteins (IIIa, VI, VIII, and IX) and the core proteins (TP, V, VII, and mu) that associate with the AdV genome (Vellinga et al., 2005). While the basic structure of these proteins is maintained, variations in the length and composition of the amino acid sequences of these proteins across species and genera cause them to bind to a number of receptors and host factors with variable affinity. Thus, the coxsackie and AdV receptor (CAR), CD46, desmoglein-2 (DSG2), CD80, CD86, vascular cell adhesion molecule-1 (VCAM-1), heparan sulfate proteoglycans (HSPGs), major histocompatibility complex class I-α2 (MHC-I-α2), sialic acid, dipalmitoylphosphatidylcholine, lactoferrin, and others can be utilized by one or more HAdV serotypes to transduce cells (Sharma et al., 2009b; Arnberg, 2012).

Viral vectors remain to be the most promising gene transfer vehicles for gene therapy because of their higher gene-transfer efficiencies when compared with other approaches. Currently, AdV vectors have been used in 23.5% of all clinical trials as they can efficiently transduce a wide variety of dividing and quiescent cells, they very rarely integrate into the host genome minimizing the risk of insertional mutagenesis (Stephen et al., 2010), they have a large packaging capacity up to 36 kb, and they can routinely be produced at high titers. Of the 57 HAdV serotypes, HAdV-5 of species C has been the best characterized and the most widely used in clinical trials so far, mainly in the treatment of cancer but also as a vaccine (Sullivan et al., 2003; Shiver and Emini, 2004).

It has, however, been estimated that over 80% of the adult population have been naturally exposed to the common HAdV serotypes (Garnett et al., 2002). As a result, preexisting humoral and cellular immunity may preclude efficient gene transfer (Kuriyama et al., 1998), and local systemic immune responses against the vector and transgene product may result in a number of clinically undesirable side effects. Severe immune responses against AdV vectors can have serious consequences in humans, which have been exemplified by a fatal case of systemic inflammatory response syndrome that was directly attributed to the systemically administered AdV gene transfer vector (Raper et al., 2003). Moreover, upon systemic administration of recombinant HAdV-5, transduction of immune cells in the liver and spleen leads to immune responses against the AdV vector that can limit its therapeutic efficacy. Eliminating liver tropism and the epitopes involved in viral proteins recognition by neutralizing antibodies (NAbs) has been proposed to reduce these immune responses. Taking advantage of the diversity of AdV capsid protein isoforms and their variety of ligand–receptor interactions, vectors based on low seroprevalent AdVs could potentially enable targeting to different cell types as well as overcoming preexisting immunity. Naturally, nonhuman AdV (NHAdV) species have a low seroprevalence in all human populations and therefore represent a source of potential vectors that could fit this niche. Since the beginning of their development in 1990s (Mittal et al., 1995), recombinant NHAdV vectors have been derived from many different AdV species: canine AdV-2 (CAdV-2), bovine AdV-3 (BAdV-3), porcine AdV-3 (PAdV-3), ovine AdV-7 (OAdV-7), murine AdV-1 (MAdV-1), several simian (SAdVs), and fowl (FAdVs) AdVs. Here we outline the current status of HAdV vectors, summarize the development of NHAdV vectors, and discuss the advantages and shortfalls of each.

Innate and Adaptive Immune Responses Against AdVs

Innate and adaptive immune responses play a critical role in resolving AdV infections. The presence of specific antiviral antibodies significantly increases antibody constant fraction receptor (FcR)-dependent viral internalization in macrophages, triggering an innate immune response to AdVs that leads to type I interferon (IFN) production and caspase-dependent interleukin (IL)-1β maturation (Zaiss et al., 2009). Internalized AdV DNA induces maturation of pro-IL-1β in macrophages dependent on NALP3 and ASC, components of the innate cytosolic molecular complex called the inflammasome, but independent of Toll-like receptors (TLRs) and IFN regulatory factors (Muruve et al., 2008). Later studies reported that this response requires the binding of the arginine-glycine-aspartic acid (RGD) motif in the AdV penton base with macrophage β3 integrins (Di Paolo et al., 2009). Barlan and colleagues also demonstrated that the specific AdV penetration mechanism of the endosomal membranes is required for IL-1β release (Barlan et al., 2011a,b). Macrophages and conventional dendritic cells (cDCs) can detect cytosolic AdV DNA in a TLR9-independent manner (Basner-Tschakarjan et al., 2006; Zhu et al., 2007), while plasmacytoid DCs (pDCs) detect nonmethylated AdV CpG-containing DNA in the endosomes through TLR9. Also, internalized HAdV-5 and HAdV-6 immune complexes have been reported to induce DC maturation via TLR9 agonist motifs present in the AdV genome, whereas HAdV-35, -36, and -26 have a limited potency to induce DC maturation because of having a smaller number of motifs (Perreau et al., 2012). TLR9 signaling pathway is dependent on myeloid differentiation primary response 88 (MyD88) (Yamaguchi et al., 2007; Zhu et al., 2007) and induces the secretion of IL-6 and IL-12 (Yamaguchi et al., 2007). Heparin-sensitive receptors on DCs bind to the AdV shaft fiber region and trigger T cell immune responses (Cheng et al., 2007). HAdV-5-induced natural killer (NK) cell activation relies on T cell contribution in vitro (Pahl et al., 2012), and also appears to be related to the presence of type I IFN in vivo (Zhu et al., 2007, 2008). In contrast to HAdV-5, in response to HAdV-35 pDCs alone are sufficient for NK cell activation in a TLR9-dependent manner, after which NK cells enhance pDC IFN-α secretion (Pahl et al., 2012). This indicates that both pDCs and T cells may be necessary for NK cells' activation. Therefore, it is clear that different AdV serotypes can trigger immune responses in different ways; thus, it is essential to study immune responses in detail to ensure that efficient strategies to circumvent immune responses are developed.

The complement system also plays a critical role in AdV immune responses with effects mediated through the classical and alternative complement pathway. The complement system appears to be related to humoral immune responses to the vector capsid or the encoded transgene (Appledorn et al., 2008). IgM isotype neutralization of virions activates the classical complement pathway hampering receptor–ligand interactions with the target cell (Xu et al., 2013). On the other side, AdV-induced thrombocytopenia has been demonstrated to be dependent on factor B and C3 of the alternative complement pathway, and AdVs mediate induction of the nuclear factor-kappaB (NF-κB) upon binding to C3 (Appledorn et al., 2008). In addition to macrophages, DCs, and NK cells, binding of the CAR protein and αv-integrins on nonimmune cells also activates immune responses (Liu et al., 2005; Schoggins et al., 2005). In particular, CAR-mediated responses have been reported in epithelial cells (Tamanini et al., 2006) and integrin-mediated responses in kidney epithelium-derived (REC) cells, inducing expression of CXCL10 (IP-10) through NF-κB in an RGD motif-dependent manner (Liu et al., 2005).

These anti-AdV innate immune responses are linked to adaptive immune responses through direct interactions and production of cytokines (Ginsberg, 1996; Appledorn et al., 2008). When a cell is infected by AdVs, it releases type I (α and β) IFN. Plasmacytoid DCs, cDCs, and macrophages produce IFN-α upon AdV recognition. The surrounding uninfected cells recognize IFN by IFN receptors and they induce, via the Jak-STAT signaling pathway, the expression of antiviral enzymes such as MxA, 2′,5′-oligoadenylate synthase, and protein kinase R (Katze et al., 2002). Also, type I IFN induces upregulation of costimulatory molecules in DCs such as CD80, CD86, and CD40, leading to DCs' maturation (Vujanovic et al., 2009). Moreover, IFN takes part in B cells' activation for antibody production. Type II IFN (IFN-γ) is an immunoregulatory cytokine secreted by Th1-type T CD4+ cells, T CD8+ cells, and NK cells. IFN-γ induces the expression of major histocompatibility complex class I (MHC-I) on nearly all cells and major histocompatibility complex class II (MHC-II) on professional antigen presenting cells (APCs) promoting antigen presentation to helper T CD4+ cells, and it also activates macrophages (Gattoni et al., 2006).

Dendritic cells have been shown to engulf infected cells and present antigens to naive T CD8+ cells through MHC-I (Albert et al., 1998). At the same time, APCs present antigens to Th1-type T CD4+ cells through MHC-II and upon activation they produce IL-2. If the T CD8+ cell receives both antigen presence and IL-2 signal, it gets activated (Jankovic et al., 2001). Once the T CD8+ cell interacts with an infected cell, mainly recognizing conserved epitopes in hexon protein (Olive et al., 2002; Leen et al., 2004; Tang et al., 2006), it begins a cytotoxic response leading to apoptosis of the infected cell (Schumacher et al., 2004). On the other hand, Th2-type T CD4+ cells induce B cell maturation and isotype switching (Jankovic et al., 2001).

NAbs, mostly IgG isotype, interact with AdV capsid proteins (Gall et al., 1996; Sumida et al., 2005). It was reported that NAbs recognize mainly the hypervariable regions (HVRs) of the hexon protein (Roberts et al., 2006; Abe et al., 2009; Shiratsuchi et al., 2010; Bradley et al., 2012b) and to a lesser extent the fiber knob domain (Myhre et al., 2007; Bradley et al., 2012a; Yu et al., 2013) and penton base proteins (Yu et al., 2013). AdV neutralization prevents cell attachment and facilitates aggregation and phagocytosis mediated by opsonization (Wohlfart, 1988). FcRs in macrophages and NK cells recognize NAbs bound to AdV antigens, facilitating their phagocytic and cytolytic action, respectively (Gattoni et al., 2006).

In summary, AdV gene transfer vectors are subjected to antiviral immune responses (Fig. 1) once administered to the patient. This process highly limits their efficiency for gene therapy applications. Therefore, it is critical to understand the immune response to AdVs and have strategies to circumvent antiviral immune responses. These will be discussed in the following sections, focusing on the main limiting factors: NAbs against AdV proteins and CD8+ cytotoxic T lymphocytes (CTLs) against infected cells.

FIG. 1.

Innate and adaptive antiadenoviral immune responses. (1) Activation of macrophages and natural killer (NK) cells by recognition of opsonized adenovirus (AdV) through antibody-constant fraction receptor (FcR). (2) Interleukin (IL)-1β maturation upon binding of arginine-glycine-aspartic acid (RGD) motif in the AdV penton base with macrophage β3 integrins. (3) Induction of IL-6 and IL-12 secretion in plasmacytoid dendritic cell (pDC) upon AdV DNA recognition through Toll-like receptor (TLR)-9. (4) Dendritic cell (DC) maturation through internalized immune complexes (ICs). (5) Triggering of T cell immune responses upon binding of AdV fiber shaft region to heparin-sensitive receptors on DCs. (6) Interferon (IFN)-α production by activated pDCs, conventional DCs (cDCs), and macrophage and IFN-α recognition by uninfected cells leading to a cellular antiviral state. (7) Induction of CXCL10 expression by AdV penton base RGD motif binding to α_v integrins. (8) Secretion of IFN-α and β by infected cells. (9) Classical complement pathway activation by virion neutralization with IgM or recognition by IgG. (10) Thrombocytopenia related to alternative complement pathway activation on AdVs. (11) Major histocompatibility complex class II (MHC-II) induction on professional antigen presenting cells (APCs) by IFN-γ signal promoting antigen presentation to helper T CD4+ cells. (12) Induction of B cell maturation and isotype switching through AdV antigen presentation by Th2-type T CD4+ cells. (13) Activation of T CD8+ cells through AdV antigen presentation by DC via MHC-I. (14) IL-2 production by Th1-type T CD4+ cells upon AdV antigen presentation by APCs through MHC-II, which activates T CD8+ cells. BCR, B cell receptor; C3b, complement component 3b; CAR, coxsackie virus and adenovirus receptor; IFN R, interferon receptor; TCR, T cell receptor.

Implications of Antiviral Immune Responses Against HAdV Vectors and Strategies to Circumvent Them

Since the tragic death of a patient participating in a clinical gene therapy trial involving administration of a replication-incompetent HAdV-5 vector (Raper et al., 2003), many efforts have been made to avoid antiviral immune responses against AdV vectors to improve their safety and efficacy. Since the AdV vector recognition by the immune system depends on the existence of antigenic epitopes in the AdV proteins (i.e., NAbs recognize AdV proteins, and CD8+ CTLs recognize infected cells), most of the strategies to circumvent the immunogenicity of HAdV vectors are based on protein modifications.

Functionally significant HAdV-5-specific NAbs are directed primarily against the HAdV-5 hexon protein and to a lesser extent to fiber and penton base proteins (Gall et al., 1996; Sumida et al., 2005; Tian et al., 2011). Despite several groups successfully generated hexon-chimeric HAdV-5 vectors that partially evaded anti-HAdV-5 immune responses (Gall et al., 1998; Roy et al., 1998; Ostapchuk and Hearing, 2001; Wu et al., 2002; Roberts et al., 2006; Bruder et al., 2012), hexon pseudotyping appears to be a limited strategy as it has been reported the formation of many nonviable virions (Youil et al., 2002).

The hexon protein (Rux and Burnett, 2000) contains 7 HVRs (Crawford-Miksza and Schnurr, 1996). Many groups have exchanged HAdV-5 HVRs with those from low-prevalent serotypes and demonstrated that the resultant AdV vector can evade HAdV-5 immunity (Roberts et al., 2006; Abe et al., 2009; Shiratsuchi et al., 2010). Roberts et al. (2006) showed that HAdV-5 with all seven HVRs replaced with the corresponding ones from the rare serotype HAdV-48 could circumvent HAdV-5 NAbs in both mice and rhesus monkeys. Further studies (Bradley et al., 2012b) exchanged only some HVRs and demonstrated that HAdV-5-specific NAbs target multiple HVRs, suggesting that mutation or replacement of all seven HVRs would probably be necessary to evade anti-HAdV-5 immunity. However, in 2012, Coughlan et al. demonstrated that, despite HAdV-5 containing all seven HVRs from HAdV-48, it displayed a decreased hepatocyte transduction and accumulation in Kupffer cells, and it triggered a robust proinflammatory response not present with the wild-type HAdV-48 (Coughlan et al., 2012). In order to generate novel viable HVR-chimeric vectors able to evade immune responses, it would be necessary to use HVRs from other serotypes, and for that it is essential to determine the viability of HVR replacements in AdV vectors based on the structural and biochemical constraints of the different HVRs that limit their manipulation.

Regarding the immunogenicity of the fiber protein, little is known about the epitopes involved in fiber recognition by NAbs, although it is thought that antiknob NAbs represent the major population of antifiber NAbs (Myhre et al., 2007; Bradley et al., 2012a; Yu et al., 2013). It has been reported that fiber pseudotyping of AdV vectors can lead to reduced AdV-associated innate (Schoggins et al., 2005) and adaptive immune responses (Parker et al., 2009; Rogée et al., 2010). Despite the C-terminal and shaft region heterogeneity, the N-terminal region (fiber tail) that is bound to penton base protein presents high similarity among diverse AdV species (Tarassishin et al., 2000). Thus, designing chimeric AdV vectors containing all proteins from HAdV-5 but fibers from other serotypes (mainly species B and D) is a very interesting approach to retarget HAdV-5 to different tissues while circumventing antiviral preexisting immunity. Thus, several chimeric AdV vectors with tropism for different cell types such as HAdV-5/3 (Haviv et al., 2002; Kanerva et al., 2002; Volk et al., 2003; Ulasov et al., 2007), HAdV-5/11 (Havenga et al., 2001; Stone et al., 2005; Wang et al., 2011b), HAdV-5/16 (Havenga et al., 2001), and HAdV-5/35 (Havenga et al., 2001) were generated. Interestingly, HAdV-5/35 and -5/11 were assessed in vivo in murine and nonhuman primate animal models, showing reduced toxicity and limited induction of inflammatory cytokines (Ni et al., 2005). However, other studies showed that HAdV-5/35 presented tropism for CD34+ human hematopoietic stem cells in vitro (Shayakhmetov et al., 2000), monocytes, granulocytes, and blast cells of human bone marrow (Rogozhin et al., 2011), and CD4+ and CD8+ T lymphocytes in vitro (Zhang et al., 2013b). Since transduction of immune cells can lead to immune responses, extensive characterization of chimeric AdV vectors is essential.

Other retargeting strategies such as the truncation or removal of the fiber knob domain and its replacement with foreign trimerization motifs and heterologous peptides fused to the fiber shaft are also recently being assessed (fiber de-knobbing) (Coughlan et al., 2010). Despite that the removal of the fiber knob domain results in a lower number of fiber copies per virion, leading to a less efficient production of the AdV vector, it has been shown that this approach may contribute to evade NAbs (Myhre et al., 2007).

Regarding the adaptive immunity mediated by CD8+ CTLs, the response depends on viral antigenic epitopes loaded in the MHC of the infected cell. High-capacity, helper-dependent (HD) vectors (also called gutless vectors) bypass this problem (Kochanek et al., 1996) as they are devoid of all viral genes, and therefore they are less immunogenic and allow long-term transgene expression (Maione et al., 2001; Ehrhardt and Kay, 2002; Dudley et al., 2004). Furthermore, as in these vectors DNA is constituted of stuffer DNA as part of the vector backbone to maintain optimum vector size, nonmethylated viral CpG-containing DNA (recognized by DCs through TLR9) can be eliminated. However, one study showed that HD AdVs also trigger an innate immune response dependent on TLR9 (Cerullo et al., 2007). Despite the advantages they offer, HD vectors are equally neutralized by preexisting antibodies against capsid proteins and also may present transgene-encoded protein immunogenicity (Tripathy et al., 1996).

On the other hand, the activation of immune responses because of interactions between AdV capsid proteins and receptors in the target cell can limit AdV vector efficacy in gene therapy applications. HAdV-5 binds to the CAR protein on human erythrocytes in the bloodstream, limiting its delivery to the in vivo target tissue and contributing to toxicity (Seiradake et al., 2009). To avoid this interaction, Nicol et al. (2004) described two CAR-binding mutations that abolished the previously described agglutination in human and rat erythrocytes. Moreover, HAdV-5 binding to CAR activates immune responses that can limit its potential as a gene transfer vector (Schoggins et al., 2005; Tamanini et al., 2006). Other interactions such as RGD with macrophage β3 integrins (Cheng et al., 2007; Di Paolo et al., 2009) or the one between the shaft fiber region and the heparin-sensitive receptor from DCs (Cheng et al., 2007) have also been associated with immune responses and, in the case of RGD motif, through IL-1α production (Di Paolo et al., 2009). Interestingly, the mutation of the RGD motif to arginine-glycine-glutamic acid (RGE) resulted in a significant fivefold reduction in spleen uptake, diminishing the antiviral inflammatory response after intravascular administration (Bradshaw et al., 2012). These studies indicate that de-targeting from native receptors could be a useful approach to avoid immune responses against the AdV vector. Nevertheless, further research needs to be done to describe the pathways through which the AdV vectors trigger immune responses. There are more factors involved in cellular recognition that could also be implicated in promotion of immune responses. These include alternative receptors and coreceptors such as FcR (Xu et al., 2008), complement-3 receptor (CR-3) (Xu et al., 2008), HSPGs and low-density lipoprotein receptor-related protein (LRP) (Shayakhmetov et al., 2005), complement receptor-1 (CR1) (Carlisle et al., 2009), scavenging receptor-A (SR-A) (Khare et al., 2012), bridging molecules such as the coagulation factor IX, X, and VII and protein C (Parker et al., 2006), complement component C4-binding protein (C4BP) (Shayakhmetov et al., 2005), and IgM antibodies (Xu et al., 2008). Further studies will have to be done in order to describe their influence on the use of HAdV vectors for gene therapy applications.

Apart from genetic engineering of HAdV vectors, other strategies such as chemical modifications by addition of cationic polymers or lipid molecules to shield AdV vectors and prevent them from binding undesired proteins have been widely studied. Several variants of polyethylene glycol (PEG) have been used, but better shielding was found with multivalent copolymers of poly[N-(2-hydroxypropyl)methacrylamide] (pHPMA). Shielding of HAdV vectors with pHPMA showed an increased biological stability of the vector and a longer persistence in blood, decreased liver transduction, and therefore a lower immune response (Green et al., 2004). Furthermore, coated HAdV vectors have been shown to be less vulnerable to recognition by preexisting NAbs in vitro (Romanczuk et al., 1999; Wang et al., 2011a) and by Kupffer cells in the liver (Mok et al., 2005), likely because of the hampered interaction of PEGylated AdVs and Kupffer cell SR-A (Haisma et al., 2009). However, other factors could be affecting liver transduction of the coated HAdV such as the limited size of the sinusoidal endothelial cells' fenestrae (Hofherr et al., 2008). The use of PEGylated HAdV vectors resulted in decreased levels of IL-6 production in vitro (Mok et al., 2005) and of IL-6 and IL-12 in vivo (Croyle et al., 2000). Moreover, coated HAdV vectors induce lower IL-6 levels in serum because of decreased levels of spleen AdV uptake (De Geest et al., 2005). Furthermore, several groups successfully retargeted coated AdV vectors (Parker et al., 2005; Stevenson et al., 2007; Morrison et al., 2008, 2009; Wang et al., 2010). Although in some studies they presented similar levels of antitumoral efficacy (Morrison et al., 2009) or transgene expression (Parker et al., 2005; Stevenson et al., 2007; Morrison et al., 2008; Wang et al., 2010) to unmodified wild-type HAdV-5, others achieved a greater therapeutic effect with negligible side effects (Eto et al., 2010; Yao et al., 2010). Despite the limited retargeting efficiency in comparison with unmodified HAdV-5, the use of PEGylated AdV vectors achieved lower hepatic tropism and reduction in liver and spleen toxicity in vivo. Moreover, evidence highlights the potential of AdV polymer coating strategies to circumvent preexisting immunity to all viral capsid proteins (Romanczuk et al., 1999; Croyle et al., 2000; De Geest et al., 2005; Mok et al., 2005; Haisma et al., 2009; Zeng et al., 2012). Nevertheless, the reported PEG immunogenicity could hamper PEGylated AdVs efficacy (Shimizu et al., 2012).

Despite the advances made in HAdV vectors to evade anti-AdV immune responses, the current strategies are limited by the formation of nonviable virions following molecular engineering and by the limited knowledge and characterization of capsid epitopes motifs responsible for cross-reactivity. Therefore, the use of low-seroprevalent HAdV serotypes has been proposed as an alternative to hexon or fiber-chimeric AdV vectors in order to circumvent preexisting NAbs. More than 15 years ago, several studies showed that the alternate use of AdV vectors from different serotypes, even within the same AdV species, can circumvent anti-AdV humoral immunity (Kass-Eisler et al., 1996; Mastrangeli et al., 1996; Mack et al., 1997; Parks et al., 1999). A study on the presence of NAbs against different HAdV serotypes in the Belgian population reported high seroprevalence toward species A, C, and E, while low seroprevalence toward species B and D, among which, serotypes 11, 34, 35, 50 (species B), 43, and 48 (species D) showed the lowest values (Vogels et al., 2003). This indicates that these serotypes could be potentially used to evade preexisting immunity. Vectors have been developed from multiple HAdV serotypes with lower seroprevalence than HAdV-5: HAdV-11 (Holterman et al., 2004; Stone et al., 2005; Abbink et al., 2007), HAdV-35 (Gao et al., 2003; Sakurai et al., 2003, 2009; Seshidhar Reddy et al., 2003; Vogels et al., 2003; Barouch et al., 2004; Wu and Tikoo, 2004; Abbink et al., 2007; Brouwer et al., 2007; Sakurai, 2008; McVey et al., 2010; Geisbert et al., 2011), HAdV-49 (Lemckert et al., 2006; Abbink et al., 2007), HAdV-26, -48, and -50 (Abbink et al., 2007; Geisbert et al., 2011), HAdV-7 (Nan et al., 2003), and HAdV-6 (Capone et al., 2006). However, since the presence of NAbs depends on AdV exposure, which is closely related to geographical location, seroprevalence studies in different specific populations are very important. Moreover, better characterization of such serotypes in terms of tropism and specificity, biodistribution, and toxicity is yet required to design safe HAdV vectors.

Use of Nonhuman Adenoviral Vectors

The idea about NHAdVs as vectors appeared in 1990s, with the assumption that they would be more useful for vaccines and gene therapeutic approaches in human medicine than HAdV-based vectors. Design and characterization of NHAdV vectors was reviewed earlier (Bangari and Mittal, 2006), with a focus on vaccine vectors. In this review we focus on the progress in the development of NHAdV vectors (Table 1) for gene therapy.

Table 1.

Most Popular Nonhuman Adenovirus Used as a Vector

Species	Genus	Serotype
Canine adenovirus A	Mastadenovirus	CAdV-2
Bovine adenovirus B	Mastadenovirus	BAdV-3
Porcine adenovirus A	Mastadenovirus	PAdV-3
Human adenovirus E	Mastadenovirus	ChAdV-68/Pan 9/SAdV-25
		ChAdV-5/Pan 5/SAdV-22
		ChAdV-6/Pan 6/SAdV-23
		ChAdV-7/Pan 7/SAdV-24
Human adenovirus G	Mastadenovirus	SAdV-7
Murine adenovirus A	Mastadenovirus	MAdV-1
Fowl adenovirus A	Aviadenovirus	FAdV-1
Fowl adenovirus C	Aviadenovirus	FAdV-10
Fowl adenovirus D	Aviadenovirus	FAdV-9
Fowl adenovirus E	Aviadenovirus	FAdV-8
Ovine adenovirus D	Atadenovirus	OAdV-7

Different species of nonhuman adenovirus are listed, specifying the genus to which they belong and the serotypes from which the vectors have been constructed.

Canine AdV vectors

Canine AdV (genus Mastadenovirus, species CAdV A)-derived vectors are currently the best described NHAdV vectors. Up to 90% of the residues involved in CAdV-2 fiber tail-penton base interactions are conserved compared with HAdV-5 (Schoehn et al., 2008). CAdV-2 binds to CAR (Tomko et al., 1997; Soudais et al., 2000), but it does not contain an integrin-interacting RGD motif in the penton base (Chillon and Kremer, 2001) in contrast with HAdV-5 (Greber, 2002). CAdV-2 penton base appears to be ∼20% shorter than that of HAdV-5 (Schoehn et al., 2008). In addition, CAdV-2 has shorter protuberances on the top of the penton base, mainly where RGD motif is contained in HAdVs, and a shorter N-terminal region. The hypervariable loops in the hexon protein are shorter in CAdV-2 and the protruding structures containing the epitopes recognized by antihexon HAdV-5 antibodies are absent, while the fiber is more complex since the shaft region contains two bends (Schoehn et al., 2008). Also, protein IX and IIIa are shorter in CAdV-2 (Schoehn et al., 2008). Production of CAdV-2 vectors started more than 15 years ago (Klonjkowski et al., 1997) in an attempt to avoid the inhibitory effect of the preexisting humoral and cellular immunity against HAdVs when used in the clinics. Later, it was shown that CAdV-2 vectors rarely cause neither the activation and proliferation of T cells nor the maturation of DCs (Perreau and Kremer, 2006). The advantage of these vectors over HAdV vectors is that CAdV-2 mainly interacts with CAR (Soudais et al., 2000) and they present specific tropism for neurons in the central nervous system (Soudais et al., 2001) with great capacity for axonal transport in vivo. Taking into account that CAdV-2 vector development was recently reviewed (Bru et al., 2010), the following data will focus on the more recent trials performed with the mentioned vector. Until very recently, the production of CAdV-2 vectors was based on dog kidney (DK) cell lines (Kremer et al., 2000), which is not accepted by the Food and Drug Administration (FDA) and the European Medicine Agency (EMA). To circumvent this problem, CAdV-2 E1-transcomplementing cell lines, based on Madin-Darby canine kidney (MDCK) cells accepted by the FDA and EMA for the production of vaccines, have been developed (Fernandes et al., 2013a). CAdV-2 vector DNA replication in the MDCK-E1 cell line was shown to correlate with the expression levels of the transcomplementing E1A and E1B. While Cre recombinase expression in MDCK-E1-Cre cell lines (for HD CAdV-2 vectors production) can impair cell growth, controlled expression of Cre was determined not to result in negative effects on viral production (Fernandes et al., 2013b). CAdV-2-based vectors are ideally suited for treating neurological disorders, but because of vector titer limitations, large volumes are necessary for applicability in the clinics. Therefore, better purification and scalable production according to good manufacturing practice (GMP) are being currently developed (Segura et al., 2012; Fernandes et al., 2013a), and the improvement of producer cell lines has increased vector titers (Ibanes and Kremer, 2013). In a comparative study on HD HAdV, HD CAdV-2, and a VSV-G lentiviral vector, the transduction efficiency and transgene expression in human midbrain neuroprogenitor cells was assessed. HD CAdV-2 was determined to have the most promising vector profile: equal transduction efficacy, no negative effect to neuronal development, and milder induced immune response (Piersanti et al., 2013). HD CAdV-2 was also used as a vector in order to evaluate a gene therapy approach to correct the neurodegenerative lysosomal storage disorder mucopolysaccharidosis type IIIa (Lau et al., 2012). It was shown to effectively transduce neurons and provide long-term transgene expression; however, its use in vivo was limited to discrete areas of the brain, indicating a need to find a way to increase transgene expression throughout the whole brain. Furthermore, in a clinically relevant stroke model in rats, cortical CAdV-2 vector injection transduced neurons at a greater level and covered a larger region than lentiviral vectors (Ord et al., 2013), reducing infarct volume and improving neurologic recovery. For all the potential of CAdV-2 vectors, there ultimately remains the need for further understanding of the effects of CAdV-2 vectors on human cells before they are able to enter human clinical trials.

Bovine AdV vectors

Bovine AdVs (genus Mastadenovirus, species BAdV A, B, C; genus Atadenovirus, species BAdV D) were the first NHAdVs used for vectoring purposes. Mittal and colleagues substituted the nonessential E3 region of BAdV-3 (BAdV B species) with the firefly luciferase reporter gene, and luciferase expression lasted at least 6 days postinfection in human embryonic kidney (HEK)-293 cells (Mittal et al., 1995). BAdV-3 might be a better option to target a specific tissue or organ, since its cell entry is independent of CAR (Bangari et al., 2005b) and other primary receptors of HAdV-5 (Bangari et al., 2005a). BAdV-3 uses sialic acid molecules as a primary receptor to enter into host cells (Li et al., 2009) and it can efficiently transduce different cell types of various species. However, it shows limitations in transducing human cell lines (Rasmussen et al., 1999; Wu and Tikoo, 2004; Bangari et al., 2005b). To circumvent this problem, capsid proteins (fiber and pIX) were genetically modified to increase the ability of BAdV-3 to transduce nonbovine cells (Zakhartchouk et al., 2007). Studies on BAdV-3 vectors in vivo biodistribution in a mouse model showed efficient transduction of the heart, kidney, and lung in addition to liver and spleen with a longer duration and higher levels of transgene expression than HAdV-5 vectors (Sharma et al., 2009a). In a mouse model of breast cancer, the biodistribution of BAdV-3 vector was comparable to that of HAdV-5 vectors, the vector had the ability to efficiently transduce tumor and systemic tissues in the presence of HAdV-5 immunity, and it evaded sequestration by Kupffer cells (Tandon et al., 2012). Also, it has been shown that there is minimal cross-neutralization by preexisting anti-HAdV immunity in humans (Bangari et al., 2005b). These results show the suitability of BAdV-3 vectors as a delivery system for cancer gene therapy while circumventing preexisting antivector immunity. Moreover, BAdV-3 and HAdV-5 vectors have a comparable biological behavior in human, bovine, porcine, and mouse cell lines and both genomes persist at high levels in vector-permissive cell lines in the absence of immune pressure, proving BAdV-3 vectors' safety for their use as gene delivery vehicles (Sharma et al., 2011).

Porcine AdV vectors

The idea of using PAdV (genus Mastadenovirus, species Porcine A, B, C) vectors was published over 20 years ago (Tuboly et al., 1993). A few years later it was discovered that the virus can enter but is unable to replicate in dog, sheep, bovine, and human cells (Reddy et al., 1999a). Soon after, PAdV-3 (PAdV A species) vectors were constructed with the long-term objective to develop replication-competent PAdV-3 as a live gene transfer vector to induce a mucosal immune response in pigs (Reddy et al., 1999b). Recombinant PAdV-3 is an effective delivery system in pigs (Hammond and Johnson, 2005), and it is often studied together with BAdV-3 vectors, showing similar properties: lack of cross-neutralization with preexisting anti-HAdV immunity in humans (Bangari and Mittal, 2004; Bangari et al., 2005b), CAR- and integrin-independent cell entry (Bangari and Mittal, 2005; Bangari et al., 2005b), similar safety profile and biological characteristics (Sharma et al., 2011), efficient transduction of several cell types of various species with PAdV-3 having an advantage in specific targeting of the breast tissue (Bangari et al., 2005b), and the vector genomes remain as linear episomes (Sharma et al., 2009a). In contrast with BAdV-3 vectors, the genome levels of PAdV-3 vector were shown to be comparable or lower than those of HAdV-5 vectors in vivo in a mouse model, except in the case of the heart, where the levels were comparable or higher (Sharma et al., 2009a). The specificity of PAdV vectors for distinct tissues is very interesting for the design of vectors for targeted gene delivery.

Simian AdV vectors

Simian AdVs are grouped into several human (certain HAdV species contain simian or ape AdVs, based on species-demarcation criteria such as phylogenetic distance and nucleotide composition) and simian AdV species within the genus Mastadenovirus. The first gene transfer vector derived from chimpanzee AdV-68 (ChAdV-68/Pan 9/SAdV-25; species HAdV E) was able to grow in HEK293-complementing cells without being neutralized by antibodies against HAdV serotypes (Farina et al., 2001). Since neutralizing seroprevalence to ChAdV-68 in different regions of the world is significantly lower than that to HAdV-5 (Xiang et al., 2006; Ersching et al., 2010; Zhang et al., 2013a), that makes it a good candidate for application in humans. Interestingly, a single surface loop defines a major neutralization site for the ChAdV-68 hexon (Pichla-Gollon et al., 2007), which when mutated, permitted the virus to escape from neutralization by polyclonal antisera obtained from animals in vitro. This was actually the first neutralizing site identified for any AdV, which was later very important for the successful application of the vectors in the clinics. Gene delivery by ChAdV-68 is CAR dependent (Cohen et al., 2002), and unlike the E1 of HAdV-5, the flanking sequences of ChAdV-68 are nonhomologous with cell-derived E1, preventing the formation of replication-competent viruses (Xiang et al., 2002). Development of vectors derived from chimpanzee AdVs Pan 5 (ChAdV-5/SAdV-22), Pan 6 (ChAdV-6/SAdV-23), and Pan 7 (ChAdV-7/SAdV-24), members of HAdV E species, started in 2004 (Roy et al., 2004). The vectors transduced skeletal muscle with the same efficiency as HAdV-5 vectors, but without being neutralized by human sera, and the neutralization domains showed to be different in Pan 6 and Pan 7 both in vitro and in vivo.

One study investigated the importance of neutralizing determinants on capsid proteins by using chimeric vectors with different combinations of Pan 7 and Pan 6 penton base, and hexon and fiber proteins (Roy et al., 2005), reporting that both hexon and fiber have neutralization epitopes and that in vivo transduction was more affected by antihexon antibodies, with no effect of antipenton base NAbs. Other studies on seroprevalence rates of various ChAdVs were carried out, with Pan 6 and Pan 7 seeming promising as gene therapy vectors although depending on the region where the potential patients live (Jian et al., 2013). A tropism-modified derivative of Pan 7 showed several advantages over HAdV-5 for the use in gene therapy: lower degree of inactivation, diminished liver transduction caused by poor stability of FX-Pan 7 complexes, greater gene delivery efficiency than the wild-type control Pan 7 vector in vitro, and higher transduction in vivo of the Her2 (human epidermal growth factor receptor type 2)-expressing human tumor cells injected intravenously in mice but not statistically significant compared with the control (Belousova et al., 2010). The last finding indicates that there is a need to improve the vector to achieve the desired levels of target-specific transduction. Pan 6 vector demonstrated a better capacity than HAdV-5 vector in transduction of brain tumor cells and glioma cells, indicating that it is a promising vector candidate for the brain tumor therapy (Skog et al., 2007). Twenty E1-deleted vectors were generated from different AdVs isolated from chimpanzees, bonobos, and gorillas, of which 12 could be propagated, rescued, and expanded in HEK293 cells (Roy et al., 2011b). SAdV-7 (HAdV G species) was proposed for vectoring a few years earlier (Purkayastha et al., 2005b), and was subsequently constructed as an E1-deleted vector that could be complemented by HAdV-5 E1 genes in HEK293 cells (Roy et al., 2011a). The latter observation was surprising since SAdVs from HAdV G species are phylogenetically quite distant from HAdV C members.

Recently, it was shown that humans have low seroreactivity against simian-derived (SAdV-11, -16) or chimpanzee-derived (ChAdV-3, -63) AdV vectors compared with HAdV vectors (rHAdV-5, -28, -35) across multiple geographic regions (Quinn et al., 2013). Chimpanzees and macaques elicited a systemic humoral but not systemic cellular immune response to endogenous AdVs. The structure of the E3 locus, involved in modulating the host's response to infection, is smaller in HAdVs than in ape AdVs, which may impact the ability of the ape AdVs to evade host immune detection and elimination (Calcedo et al., 2009). ChAdV-68 showed the ability to transduce immature and mature human DCs, followed by secretion of IFN-α and IL-6 but not IL-12 or tumor necrosis factor (TNF)-α, and transduced immature DCs could stimulate proliferation of autologous T lymphocytes (Varnavski et al., 2003). Therefore, ChAdV-68 could be used as a vector for transduction of human DCs.

Fowl AdV vectors

Construction of FAdV (genus Aviadenovirus, species FAdV A to FAdV E) vectors started with the demonstration that recombinant FAdV-10 (FAdV C species) could be used to express an antigen to induce a protective immune response in pathogen-free chickens (Sheppard et al., 1998). After the identification of FAdV-9 (species FAdV D) regions that are nonessential for the virus (Ojkic and Nagy, 2001, 2003; Corredor and Nagy, 2010a), FAdV-9 recombinants demonstrated wild-type growth kinetics, virus titers, cytopathic effect and plaque morphology (Corredor and Nagy, 2010a), expression in avian and mammalian cells, and lack of replication in mammalian cells, indicating their applicability as gene delivery vehicles for mammalian systems (Corredor and Nagy, 2010b). FAdV-8 (species FAdV E) vectors were also constructed and showed efficacious in vivo delivery of antibody fragments against pathogenic infectious bursal disease virus (Greenall et al., 2010). FAdV-1 (species FAdV A) or CELO (Chicken Embryo Lethal Orphan) vectors have been widely used. The genomic regions that could be deleted were evaluated more than 10 years ago and a cosmid-based strategy was developed for generating CELO vectors (Michou et al., 1999; François et al., 2001). FAdV-1 has several interesting properties: it interacts with CAR and its capsid architecture allows changes in its tropism (Tan et al., 2001), it transduces mammalian cells as efficiently as HAdV-5 vector, and it has a potential for mammalian gene transfer applications, especially because it has larger DNA packaging capacity and greater physical stability (Michou et al., 1999). It was shown that FAdV-1 is able to deliver p53 transgene into various tissues in vivo restoring p53 function in human tumor cell xenografts in mice and leading therefore to the inhibition of tumor growth (Logunov et al., 2004). Moreover, FAdV-1 encoding the human IL-2 gene can successfully produce biologically active recombinant IL-2 in vitro, in ovo, and in vivo, and it is capable of increasing the median survival time of mice carrying melanoma tumors (Shmarov et al., 2002; Cherenova et al., 2004). Also, recombinant FAdV-1 encoding the HSV-1 thymidine kinase showed cytotoxicity in carcinoma cell lines and suppressed tumor growth and increase median of survival of mice with melanoma tumors (Shashkova et al., 2005), demonstrating its efficacy for the gene delivery to tumor cells in vitro and in vivo. Furthermore, one study showed that FAdV-1 successfully enhanced expression levels of secreted alkaline phosphatase reporter gene in transduced mammalian cells in vitro and in vivo (Tutykhina et al., 2008). Finally, PEGylated and retargeted with fibroblast growth factor, FAdV-1 showed increased levels of binding and internalization in a variety of human cell lines (Stevenson et al., 2006), with transgene expression being greater than the unmodified version in PC-3 human prostate cells, and being fully resistant to inhibition by human serum in vitro (Stevenson et al., 2006). All together indicates that retargeting of CELO virus to human cells via disease-specific receptors is possible, while avoiding preexisting humoral immunity.

Ovine AdV vectors

The production of OAdV (genus Mastadenovirus, species OAdV A, B; genus Atadenovirus, species OAdV D) vectors started more than 15 years ago (Xu et al., 1997) after determining the nonessential regions within the genome (Vrati et al., 1996). Surprisingly, much larger insertions than expected are tolerated, and portions of sequences can be deleted without affecting virus viability (Xu et al., 1997). The primary receptor for OAdV-7 (OAdV D species) is not CAR, as it was shown that OAdV-7 does not compete with HAdV-5 for the entry into cells (Xu and Both, 1998; Voeks et al., 2002). Recombinant OAdVs can infect a variety of nonovine cells such as rabbit, human, and murine cells but cannot replicate within them (Khatri et al., 1997; Hofmann et al., 1999; Löser et al., 2000; Xu et al., 2000; Kümin et al., 2002; Lockett and Both, 2002). They are not neutralized by polyclonal serum against HAdV-5 (Xu and Both, 1998) and they can overcome preexisting humoral immunity against HAdVs in vivo (Hofmann et al., 1999). A cosmid-based system proved useful for efficient generation of recombinant OAdVs and vectors could be rescued in an ovine fetal skin fibroblastic producer cell line (Löser et al., 2003). OAdV-7 carrying purine nucleoside phosphorylase (PNP), which converts the prodrug fludarabine into its activated form 2-fluoroadenine, was effective in vivo against prostate cancer progression upon prodrug addition (Voeks et al., 2002; Martiniello-Wilks et al., 2004). Moreover, recombinant OAdV-7 expressing ovalbumin showed efficacy in inducing antitumor response in a mouse model (Tang et al., 2012). OAdVs' biosafety profile and their application as gene delivery vectors were reviewed in the last decade (Both, 2004).

Murine AdV vectors

The production of MAdV (genus Mastadenovirus, species Murine A, B, C) mutants started with the MAdV-1 (MAdV A species) E3-deleted recombinant (Beard and Spindler, 1996; Cauthen et al., 1999) and E1-deleted recombinant (Ying et al., 1998). MAdV-1 presents mouse endothelial cell tropism (Charles et al., 1998; Kajon et al., 1998; Lenaerts et al., 2005), but it can also infect human endothelial cells (Nguyen et al., 1999) and it displays higher affinity for primary human smooth muscle cells than recombinant HAdV-5 (Lenaerts et al., 2009). MAdV-1 lacks the RGD motif in the penton base, but it contains it in the fiber knob domain (Raman et al., 2009). The RGD motif was reported to play a role in MAdV-1 infection through α_v integrins, which act as a receptor for the virus, and it was shown that cell surface heparan sulfate glycosaminoglycans are involved in MAdV-1 infection (Raman et al., 2009). Conversely, the primary attachment of MAdV-1 is independent of CAR (Lenaerts et al., 2006). Distribution of MAdV-1 to the liver is markedly lower in intravenously injected immunodeficient mice than that observed with recombinant HAdV-5 (Lenaerts et al., 2009). Moreover, MAdV-1 has been used as an oncolytic vector, proving that it is a suitable murine homolog model to test the mechanism of action of murine oncolytic AdV vectors in an immunocompetent, tumor-bearing host (Robinson et al., 2009). Furthermore, MAdV-1 has preference for ovarian carcinoma cell lines, probably because of their increased expression of the enzyme involved in HSPG biosynthesis, therefore having the potential to be used in the oncolytic treatment of ovarian cancer (Lenaerts et al., 2012).

Concluding Remarks and Future Perspectives

The main limitation of clinically applied HAdV-based vectors today is their failure in achieving high levels of efficient transduction in their target tissues. This is primarily because of the preexisting NAbs leading to the rapid clearance of circulating vector and cellular immune responses resulting in the elimination of transduced cells, greatly limiting the expression of the transgene. These immunological responses most likely stem from the high frequency of HAdV exposure in multiple human populations across the globe, which results in immunological memory against HAdV serotypes that most of today's vectors are based on. In order to circumvent this, several strategies have been developed for HAdVs, including molecular engineering of chimeric vectors or the use of low-seroprevalent HAdV vectors.

As an alternative, NHAdV vectors have been developed and are being tested in various nonhuman and human cell lines. Most of the NHAdV vectors have two important features that make them more appropriate than HAdV vectors: there is no or very low pathogenicity for their natural host, and there is no presence of preexisting NAbs, CD4+ T cells, and CD8+ CTLs against them. However, it should be noted that caution must be applied when using nonhuman primate AdVs, since their ability to infect human cells and their structural similarity could potentially result in anti-HAdV T-cell and/or NAb cross-reactivity with nonhuman primate AdV antigens. Different NHAdVs have specific tropism for different cell types, which makes them applicable in treating different diseases especially when the targeting of specific cells is required, and therefore they could extend the range of potential target tissues. Furthermore, the inability of NHAdVs to co-replicate with wt HAdVs, as well as the fact that they are replication defective in human cells, makes their vectors much safer than HAdV vectors. Many of the NHAdVs are able to tolerate longer sequences of exogenous DNA than HAdVs. For some of the NHAdVs, there are low-cost propagation systems available, for example, chicken embryos for CELO virus (Laver et al., 1971; Michou et al., 1999). Despite these advantages, there are still many steps necessary before using the NHAdV vectors in the clinics. The first challenge is to find the appropriate cell lines for efficient production. There are also safety issues compared with HAdVs such as the possible occurrance of interspecies adaptation. For example, HAdV-4 is the only HAdV member of HAdV E species that is very similar to SAdVs from that group, indicating its zoonotic origin and molecular adaptation to its new host (Purkayastha et al., 2005a; Dehghan et al., 2013). As AdVs have coevolved with their hosts (Benkö and Harrach, 2003; Davison et al., 2003), there is a possibility that humans and monkeys could be infected with the same AdVs, a concern when considering chimpanzee or simian AdV vectors for use in humans. Nevertheless, NHAdV vectors present great potential as effective gene therapy vehicles and will hopefully be part of the successful AdV vectors used in the clinics in the coming years.

Acknowledgments

We would like to thank all members of the “ADVance” international training network (Adenoviruses as Clinical Treatments; FP7 ITN, EU grant agreement ref. 290002) and in particular Andrew Baker and Eric J. Kremer for their advice and encouragement to write this review.

Author Disclosure Statement

No competing financial interests exist.