HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via Google Scholar Google Scholar Articles by Russell, W. C. Search for Related Content PubMed PubMed Citation Articles by Russell, W. C. Agricola Articles by Russell, W. C.

Update on adenovirus and its vectors

Biomolecular Sciences Building, School of Biology, University of St Andrews, North Haugh, St Andrews, Fife KY16 9ST, UK¹

Author for correspondence: Willie Russell. Fax +44 1334 462595. e-mail wcr{at}st-andrews.ac.uk

	Introduction

TOP Introduction General properties of... Defence mechanisms Adenoviruses as vectors Application of adenovirus... Future developments References

 
Adenoviruses have been characterized extensively since their initial description in the early 1950s (Hillemann & Werner, 1954 ; Rowe et al., 1953 ) and there is now a panoply of observations on the properties of many of the virus gene products. Nevertheless, there is still a lack of understanding of a number of the molecular mechanisms that operate in the infected cell, particularly in respect of how the virus gene products interact with cellular components and of the nature of the responses mounted by the host in response to infection.

In this regard, it is significant that, although there were almost 4000 references to adenoviruses in the 3 years from 1997 to 1999, most of these have been concerned with the results of investigations using adenoviruses as vectors and relatively few have dealt with the basic virology and immunology of virus infection. Indeed, it is now accepted that the initial enthusiasm for utilizing adenovirus gene vectors in therapy was rather prematurely optimistic and was perhaps over-hyped. In its place, there is a realization that targetting the vector effectively is not so straightforward and that, more importantly, the efficacy of host defences has not been fully appreciated and must be adequately addressed.

This paper reviews the current knowledge in the field of adenovirus vectors as well as advances in our understanding of the properties of the adenovirus gene products. Particular emphasis has been made on developments over the last two to three years. There have been a number of reviews examining different aspects of the vector field, and the reader is referred to these (Benihoud et al., 1999 ; Hitt et al., 1997 ; Zhang, 1999 ) for more comprehensive coverage. The expectation remains that a better understanding of the total spectrum of the virus–cell and virus–host interactions will lead to the design of vectors that provide more efficient delivery along with minimal deleterious host reactions.

	General properties of adenoviruses

TOP Introduction General properties of... Defence mechanisms Adenoviruses as vectors Application of adenovirus... Future developments References

 
Adenoviruses have a characteristic morphology (Stewart et al., 1993 ), with an icosahedral capsid consisting of three major proteins, hexon (II), penton base (III) and a knobbed fibre (IV), along with a number of other minor proteins, VI, VIII, IX, IIIa and IVa2 (Fig. 1). The virus genome is a linear, double-stranded DNA with a terminal protein (TP) attached covalently to the 5' termini (Rekosh et al., 1977 ), which have inverted terminal repeats (ITRs). The virus DNA is intimately associated with the highly basic protein VII and a small peptide termed mu (Anderson et al., 1989 ). Another protein, V, is packaged with this DNA–protein complex and appears to provide a structural link to the capsid via protein VI (Matthews & Russell, 1995 ). The virus also contains a virus-encoded protease (Pr) (Weber, 1976 ; Webster et al., 1989 ), which is necessary for processing of some of the structural proteins to produce mature infectious virus.

View larger version (46K): [in this window] [in a new window]   Fig. 1. Structure of adenovirus. The locations of the capsid and cement components are reasonably well defined. In contrast, the disposition of the core components and the virus DNA is largely conjectural.

Early events in adenovirus infection
The adenovirus infectious cycle can be clearly defined into two phases. The first or ‘early’ phase covers the entry of the virus into the host cell and the passage of the virus genome to the nucleus, followed by the selective transcription and translation of the early genes. These early events modulate the functions of the cell so as to facilitate the replication of the virus DNA and the resultant transcription and translation of the late genes. This leads to the assembly in the nucleus of the structural proteins and the maturation of infectious virus. The early phase in a permissive cell can take about 6–8 h (depending on a number of extraneous factors), while the late phase is normally much more rapid, yielding virus in another 4–6 h.

The adsorption of virus to target cell receptors involves high-affinity binding to cell receptors via the knob portion of the fibre; for a review see Chroboczek et al. (1995) . The prime receptor for the human subgroup C adenoviruses was shown to be identical to that for coxsackie B virus (Bergelson et al., 1997 ) and has therefore been termed the coxsackie/adenovirus receptor (CAR). This has subsequently been shown to be a plasma membrane protein of 46 kDa belonging to the immunoglobulin superfamily and to contain extracellular, transmembrane and cytoplasmic domains (Tomko et al., 1997 ), with the extracellular domain being sufficient for attachment (Wang & Bergelson, 1999 ). A more recent study has indicated that the adenovirus CAR does not completely overlap the coxsackievirus receptor (Tomko et al., 2000 ). A comprehensive survey of representative members of all the human adenovirus species A to F (Roelvink et al., 1998 ) suggested that they all bound to CAR with the exception of members of the subgroup B, which appear to recognize a different receptor (Stevenson et al., 1995 ). In the same study, it was also noted that adenovirus serotype 41 (Ad41) (of subgroup F) has two fibres of different lengths, and only one of them binds to CAR. Since Ad41 readily infects cells of the gastrointestinal tract, it seems likely that the other fibre will adsorb to a different cell receptor, perhaps displayed on enterocytes. Some cells, such as those of haemopoietic origin, appear to be largely refractory to productive infection by human adenoviruses 2 and 5 and do not display CAR molecules on their plasma membranes (Mentel et al., 1997 ). This suggests that receptor recognition could be one of the key factors involved in cell tropism. In an attempt to modify cell tropism, fibreless adenoviruses have been constructed. Not surprisingly, these particles showed drastically reduced infectivities and were extremely unstable. The low level of infectivity that could be detected possibly operated by integrin-dependent pathways, which have been demonstrated to operate in some cell systems (Huang et al., 1996 ; see below). Recent investigations have succeeded in defining the receptor-binding motif on the three-dimensional structure of the fibre head (Kirby et al., 1999 ; Santis et al., 1999 ) as well as on the CAR (Kirby et al., 2000 ). It has also been shown that receptor recognition can be altered by switching fibre heads from other subgroups (Miyazawa et al., 1999 ). Experiments have also been carried out that have substituted or added other receptor-binding motifs (Hidaka et al., 1999 ) in the fibre knob; see below. It is interesting to note that some adenovirus serotypes seem to have additional specificities of binding, suggesting that the CAR receptor may be part of a family of receptors (Segerman et al., 2000 ). Indeed, another receptor, the histocompatibility class I molecule, also a member of the immunoglobulin superfamily, has been shown (Hong et al., 1997 ) to be available for the subgroup C viruses. Moreover, a recent observation suggests that Ad37, a member of subgroup D, appears to bind to sialoglycoprotein receptors (Arnberg et al., 2000 ), indicating that receptor specificities are wider than was at first thought.

After the initial interaction of the virus with the fibre receptor, entry of the virus proceeds via clathrin-mediated endocytosis. The critical recognition mechanism for this process is an RGD motif that is exposed on the penton base (Stewart et al., 1997 ) and interacts with cellular v integrins (Wickham et al., 1993 ). There appears to be direct binding of the virus penton base to the integrins in the presence of divalent cations (Mathias et al., 1998 ). Integrins normally react with the extracellular matrix to facilitate adhesion, differentiation and other cell–cell phenomena (Meredith et al., 1996 ). They form a large family of heterodimeric receptors and it appears that integrins v3 and v5 both support adenovirus internalization. It is noteworthy that integrin v5 is expressed on human bronchial epithelial cells, a major site of primary adenovirus infection in vivo (Mette et al., 1993 ). Integrins may also play an important part in defining tropism in some situations such as the intestinal epithelium (Croyle et al., 1998c ) while in others, such as hepatocytes, they have a minimal role (Hautala et al., 1998 ).

Interaction of the virus with the plasma membrane can induce a number of signalling pathways and there is good evidence for the activation of the phosphoinositide-3-OH kinase (PI-3K) pathway, which in turn triggers the Rho family of GTPases and the polymerization and reorganization of actin to facilitate endocytosis (Li et al., 1998 ; Rauma et al., 1999 ). As early as 20 min post-infection, activation of the Raf/mitogen-activated protein kinase (MAPK) pathway and consequential production of IL-8 have been observed (Bruder & Kovesdi, 1997 ). Since the activation of Raf/MAPK is insensitive to the addition of cycloheximide and is sensitive to prior heating of the virus inoculum at 56 °C, it seems plausible that the initial events at the cell membrane are triggered by a structural component, and this could be via the penton base, since it is heat sensitive (Russell et al., 1967 ). Triggering these pathways may act as an ‘early-warning system’ for the induction of defence mechanisms induced in the host (see below).

As noted above, progress of the virus through the endosomes and into the cell cytoplasm is normally mediated by clathrin and the coated pit pathway (Wang et al., 1998 ). Thereafter, the virus-encoded protease appears to assist in the further disruption of the virus capsid by the proteolysis of the structural protein VI (Greber et al., 1996 ), which functions as a linker between the capsid and the core components (Matthews & Russell, 1994 , 1995 ). The partially disrupted virus is then transported to the nuclear membrane and the genome is passaged through the nuclear pore and into the nucleus, where the primary transcription events are initiated. The passage through the cytoplasm to the nucleus has been postulated to be mediated by the association of the virus core (the virus DNA and the covalently attached TP, together with the basic proteins VII and V and mu peptide) with a cellular protein, p32 (Matthews & Russell, 1998b ). The p32 protein is primarily located in the mitochondria but can also be detected in the nucleus, and it has been suggested that it is a component of a cellular transport system that shuttles between the mitochondria and the nucleus and that the virus can hijack this system to gain access to the nucleus. This passage to the nucleus is relatively rapid and also involves the participation of dynein and microtubules (Leopold et al., 2000 ; Suomalainen et al., 1999 ). Virus-like particles can be detected at the nuclear membrane by electron microscopy within 1 h of infection (Dales & Chardonnet, 1973 ) and virus DNA and proteins V and VII can be detected within the nucleus between 1 and 2 h (Greber et al., 1997 ; Matthews & Russell, 1998 a ). Once inside the nucleus, the genome is targetted to the nuclear matrix (NM), where the TP forms a tight complex with the cellular CAD pyrimidine synthesis enzyme and possibly other NM components (Angeletti & Engler, 1998 ; Fredman & Engler, 1993 ). It is interesting that nuclear lamin B, which is a component of the NM, readily binds to p32 (Simos & Georgatos, 1994 ) and this may allow for the disassociation of p32 from the incoming genome.

Transcription and replication
As noted above, adenovirus transcription can be defined largely as a two-phase event, early and late, respectively occurring before and after virus DNA replication (Fig. 2). Transcription is accompanied by a complex series of splicing events, with four early ‘cassettes’ of gene transcription termed E1, E2, E3 and E4 (Fig. 2).

View larger version (15K): [in this window] [in a new window]   Fig. 2. Transcription of the adenovirus genome. The early transcripts are outlined in green, the late in blue. Arrows indicate the direction of transcription. The gene locations of the VA RNAs are denoted in brown. MLP, Major late promoter.

NF-B is a nuclear transactivator that is released by proteolysis of an associated inhibitory factor, IB, in the cytoplasm (Hay et al., 1999b ), thus leading to its migration to the nucleus and the activation of NF-B-responsive genes, among the latter being the E3 gene promoter (Deryckere & Burgert, 1996 ) (see below). Phosphorylation of IB by a kinase complex, IKK, appears to be crucial for the proteolysis of IB.

The protein p53 is a tumour suppressor that regulates the transcription of a variety of genes involved in cell cycle arrest and apoptosis. In normal cells, p53 is present in small amounts, but levels increase in response to genotoxic and other stresses. The regulation of p53 seems to mainly at the protein level, utilizing the cellular protein mdm2, which binds to p53 and acts as a ubiquitin ligase, targetting p53 to the proteosome for degradation. Another cellular protein, p19arf, also contributes to this system by binding to mdm2, blocking its ligase activity and thereby stabilizing p53 (de Stanchina et al., 1998 ; Honda & Yasuda, 1999 ; Tao & Levine, 1999 ; Weber et al., 1999 ). An additional factor in this regulation has also been uncovered by the finding that p53 can be modified by the small ubiquitin-like modifier (SUMO), leading to activation of p53 (Rodriguez et al., 1999 ).

E1A proteins interfere with the processes of cell division and with the regulation of NF-B and p53, and do this by a great variety of strategies involving both direct and indirect interaction with cellular proteins. They can also modulate transcription patterns in favour of virus transcription. A summary of the characteristics of E1A is found in Table 1 and Fig. 3. It should be pointed out that many of the properties ascribed to E1A in Table 1 are based on in vitro studies, whereas the availability of the relevant cellular components in vivo will depend on the nature of the infected cell and its metabolic state. Moreover, other virus gene products can modulate these cellular interactions significantly. For instance, the E4 gene products can co-operate with E1A in a variety of ways (Goodrum & Ornelles, 1999 ; Hall et al., 1998 ; Yun et al., 1999 ). The E1B gene product 19K also seems to function co-operatively with E1A and p53 in promoting oncogenesis and transformation (Kannabiran et al., 1999 ), mainly by ensuring that the downstream consequences of cell cycle release do not induce apoptosis.

View larger version (25K): [in this window] [in a new window]   Fig. 3. Binding map of proteins to E1A. The locations have been determined on the basis of in vitro studies using deletion and mutational analysis. The abbreviations are explained in Table 1. CR1, CR2 and CR3 are constant regions present in a wide range of adenoviruses. PLDLS and LXCXE etc. are recognition motifs. E2F-1 A.D. homology refers to sequences that can displace E2F-1 from Rb and p300.

The E2 gene products are subdivided into E2A (DBP) and E2B (pTP and Pol). These provide the machinery for replication of virus DNA (Hay et al., 1995 ) and the ensuing transcription of late genes, and this is mediated by interaction with a number of cellular factors.

The E3 genes, which are dispensable for the replication of virus in tissue culture, provide a compendium of proteins that subverts the host defence mechanisms (see below) and their properties are summarized in Fig. 4(A). One of these E3 gene products has been termed the adenovirus death protein (ADP), since it facilitates late cytolysis of the infected cell and thereby releases progeny virus more efficiently (Tollefson et al., 1996 ). The E3 gp19K is localized in the ER membrane and binds the MHC class I heavy chain and prevents transport to the cell surface, where it would be recognized by CTLs. This gene product, in addition, delays the expression of MHC I (Bennett et al., 1999 ). The E3 proteins RID& and 14.7k inhibit proapoptotic pathways (see below). A recent review on these proteins can be consulted for further information on the molecular mechanisms involved (Wold & Chinnadurai, 2000 ).

View larger version (31K): [in this window] [in a new window]   Fig. 4. Characteristics of E3 and E4 gene products. Horizontal arrows indicate the relative directions of transcription.

Adenoviruses also transcribe a set of RNAs (see Fig. 2) that are not translated, termed the VA RNAs, and these play a role in combating cellular defence mechanisms (see below).

Other characteristics of these early gene products are described below, and a cartoon depicting the effects of some of them on a few cellular pathways is provided in Fig. 5. DNA replication begins from both DNA termini and requires sequences within the ITRs as origins of replication (Hay et al., 1995 ). Thereafter, late transcription ensues, with five cassettes of transcripts (termed L1 to L5) resulting from a complex series of splicing events. These lead to the production of the virus structural components and the encapsidation and maturation of virus particles in the nucleus. A key player in the control of transcription is the major late promoter (MLP), which is attenuated during transcription of the early genes. However, it should be noted that there is a low basal level of late transcription occurring early in infection, even before the MLP comes into play. After the onset of virus DNA replication, the IVa2 and IX genes are expressed at high levels (see Fig. 2) and transcription via the MLP is fully functional by specific activation. This is accomplished via the IX and IVa2 gene products (Lutz & Kedinger, 1996 ; Lutz et al., 1997 ) and is also influenced by effective competition for the limiting transcription factors (Fessler & Young, 1998 ). The encapsidation process is governed by the presence in the virus DNA of a packaging signal at the conventional left end, which consists of a series of AT-rich sequences (Hearing et al., 1987 ). These events are accompanied by major changes in the nuclear infrastructure and the permeabilization of the nuclear membrane (Rao et al., 1996 ; Tollefson et al., 1996 ). This facilitates the egress of the virus into the cytoplasm and is followed by the disintegration of the plasma membrane and the release of virus from the cell.

View larger version (140K): [in this window] [in a new window]   Fig. 5. A cartoon (not to scale) illustrating some of the sites of action of the virus and virus gene products (in red) on a few of the cellular pathways (in yellow). A virus particle at the receptor site is in green.

	Defence mechanisms

TOP Introduction General properties of... Defence mechanisms Adenoviruses as vectors Application of adenovirus... Future developments References

Host cells have a range of strategies to combat any incursion by an intruder; these can be considered as innate and adaptive. With respect to the former, it has recently been established that some epithelial cells release 3–4 kDa antimicrobial peptides termed defensins (Ganz & Lehrer, 1998 ) and it has been shown that these compounds can provide significant protection from adenovirus infection (Gropp et al., 1999 ). Indeed, an adenovirus vector expressing a defensin has been utilized to supplement innate defences (Bals et al., 1999 ). Some tissues, on receiving the appropriate signal (perhaps via the Raf/MAPK pathway; see above), will release multiple chemokines that, in turn, recruit neutrophils and invoke an inflammatory response (Charles et al., 1999 ; Muruve et al., 1999 ). Innate defence mechanisms such as recruitment of macrophages, activation of complement and natural killer (NK) cells have been shown to play a significant role in clearing an adenovirus infection in vivo (Worgall et al., 1997a , 1999 ). The transcription factor NF-B appears to be a key regulator of the innate antiviral response (Ferreira et al., 1999 ), since it can activate the transcription of cytokines and adhesion molecules, leading to the production of a range of proinflammatory cytokines and the orchestration of other signalling pathways. It has been claimed that adenovirus infection, especially at high multiplicities, can lead to the activation of NF-B at early stages of infection (Clesham et al., 1998 ; Lieber et al., 1998 ). One mechanism for achieving this could be by the binding of E1A to the p65 subunit of NF-B, although it is also apparent that this activation can be suppressed by E1B 19K (Pahl et al., 1996 ; Schmitz et al., 1996 ). Reference has been made above to the very early induction, possibly by the interaction of the penton base in the virus with cellular integrins, of the Raf/MAPK and other pathways. These may play a role in the activation of NF-B (Ghoda et al., 1997 ) as well as in the early release of chemokines (Kuhnel et al., 2000 ; Muruve et al., 1999 ) and interferons, which are important components of the innate response to infection. However, the induction of these endogenous genes following infection appears to be quite cell dependent, with human endothelial cells displaying a range of signalling molecules at 24 h, a scenario not seen in human dermal fibroblasts or alveolar macrophages (Ramalingam et al., 1999 ).

Interferons
The interferons are subdivided into two main classes, type I (containing interferons and ) and type II (interferon ). They are cellular proteins ranging in size from 15 to 35 kDa and are released from cells very early after infection by viruses and as a result of other insults to the cell and display a fair degree of cell specificity. In the case of adenoviruses, induction seems to be by interaction with a structural component, since they can be produced by virus particles in the absence of protein synthesis (Reich et al., 1988 ). The interferons function by binding to cell receptors, thereby activating the cellular Jak/STAT pathways, which lead to STAT complexes being transferred to the nucleus and binding to interferon-response elements (ISREs) on the cellular DNA. The ISREs regulate the transcription of a range of gene products, such as a dsRNA-induced protein kinase (PRK) and a 2'–5' oligoadenylate synthetase as well as a variety of immunomodulators. These form an impressive array of weapons to combat the intracellular activities of the invading virus; for a recent review see Goodbourn et al. (2000 ). Adenoviruses are generally refractive to interferons, since they have provided themselves with a number of strategies to overcome this assault on their activities. Thus, gene products from E1A downregulate the STAT activators (Look et al., 1998 ; McDonald & Reich, 1999 ; Paulson et al., 1999 ; Leonard & Sen, 1996 , 1997 ).

It has also been claimed (Feigenblum et al., 1998 ) that adenoviruses induce an interferon-regulatory factor (IRF) at later stages of infection that plays a role in cytopathogenicity. In addition, the VA RNAs (Mathews & Shenk, 1991 ) bind to and inactivate PRK.

Apoptosis
As another means of combating virus infection, the cell can redirect its metabolism to switch on its apoptosis circuits. Cells have complex mechanisms for ensuring that their integrity is not compromised and they have devised a fall-back strategy to switch on proapoptotic proteins when specific alarm pathways are activated (see Fig. 5). Chief among these is the tumour suppressor p53, which regulates the transcription of genes involved in cell cycle arrest and apoptosis. Among the latter are members of the Bax family (Pearson et al., 2000 ), which interact with mitochondria and are involved in the induction of caspases, leading to apoptosis. There are other members of this family, such as Bcl-2, which function to inhibit apoptosis and they carry this out by binding to Btf, an important transcriptional repressor. Btf promotes cell death (Kasof et al., 1999 ) by inducing the permeabilization of mitochondrial membranes (Imazu et al., 1999 ) and releasing cytochrome c, thereby initiating the caspase cascade. Adenoviruses can subvert the operation of this pathway by utilizing virus gene products from its E1 cassette; thus, E1B 19K can inactivate Bax (Han et al., 1996 ; Ohi et al., 1999 ) and has a similar function to Bcl-2 in binding to Btf, thus counteracting the proapoptotic response of the E1A gene product in activating p53. A parallel mechanism of apoptosis is mediated by TNF, which is secreted by monocytes and lymphocytes following activation as part of the innate response. This cytokine appears to play a significant part in the elimination of adenovirus vectors (Elkon et al., 1997 ) and functions by activating cytosolic phospholipase A2 (pL A2), which permeabilizes cell membranes, releasing arachidonic acid (Wolf & Laster, 1999 ) and initiating the production of prostaglandins and leukotrienes, which also play a role in inflammation (Krajcsi et al., 1996 ). This pathway is normally modulated by Bcl-2 and its virus analogue, E1B 19K (see above), via downregulation of IB transcription, thus releasing NF-B to the nucleus (de Moissac et al., 1999 ). In this regard, it is intriguing to note that the ability of p53 to induce apoptosis requires the participation of NF-B (Ryan et al., 2000 ), implying a degree of co-operative ‘cross-talk’. It is important, however, to note that many of these effects seem to be cell specific: thus, in endothelial cells, Bcl-2 serves to protect the cells from both apoptosis and proinflammatory responses (Badrichani et al., 1999 ). Another route of TNF action is by the direct induction of caspases (Kimura & Gelmann, 2000 ). In contrast, in oligodendrocytes, apoptosis by TNF appears to be mediated by p53 and involves initiating the JNK signalling pathway (Ladiwala et al., 1999 ). TNF-induced apoptosis can be ablated by E3 gene products (Lukashok et al., 2000 ) (see Fig. 5). Other key players in apoptosis are Fas and Fas ligand interactions, and these have been shown to be the major mediators of the elimination of adenovirus vectors from the liver (Chirmule et al., 1999 ). In this case, the E3 gene products RID and RID cause Fas to be removed from the cell surface and degraded (Tollefson et al., 1998 ). A cellular protein termed FIP-3 (Li et al., 1999b ) has also been implicated in these proapoptotic events. FIP-3 appears to be a scaffolding component of the IKK complex (Ye et al., 2000 ) and blocks the release of NF-B by inhibiting the kinase activity of IKK (Fig. 5) (Li et al., 1999b ). Adenoviruses modulate these events via the E3 gene product 14.7K, which binds to FIP-3 (Li et al., 1999 b ) and effectively restores NF-B transcription and thereby cell survival. Many of these apoptotic mechanisms involve the activation of a range of proteases, such as caspases, and it has been demonstrated that inhibition of the related ICE-like proteases can boost adenovirus yields (Chiou & White, 1998 ). A recent investigation has also shown that E1A can induce apoptosis by activation of caspase-8 and is independent of the status of p53 (Putzer et al., 2000 ). Interferons can also act proapoptotically by inducing caspase-8 (Balachandran et al., 2000 ), and this can be amplified in infected cells via the dsRNA route (Tanaka et al., 1998 ). In this case, however, the inhibition of interferon induction by E1A suppresses this apoptotic response. From the above discussion it will be seen that there is a complex interplay of cellular and virus components seeking to control cell survival and promotion of virus replication and spread. Thus, in utilizing adenoviruses as vectors, it is critical to take these factors into account in devising the optimum conditions for delivery and effective expression of the transgene.

Cellular immune responses
T cells provide an effective defence via both CD8⁺ cytotoxic cells (CTLs) and CD4⁺ helper cells. CTLs function by recognizing a virus antigen in a complex with class I proteins of the MHC on the cell surface. This event releases perforin, resulting in cell lysis, thereby eliminating the infected cells even at an early stage before any virus is assembled and released. The recognition mechanism depends on a virus antigen being available to complex successfully within the ER membrane with an MHC component and then being transported to the plasma membrane. The complex formation is a function of the nature of the cell being infected, as well as the MHC status of the host. It also appears that different virus gene products can provide the target depending on their ability to interact with a particular MHC. It is significant, however, that there appears to be some cross-reactivity of human CTLs in recognizing different adenovirus subgroups (Smith et al., 1998 ).

Adenoviruses can combat this cellular strategy as described above by utilizing E3 gp19K to retain the MHC antigens in the ER and hence disrupt the recognition process (Kvist et al., 1978 ). E4 gene products have also been demonstrated to function in the inhibition of T cell cytolysis (Kaplan et al., 1999 ).

The CD4 helper cells are important in mounting a proliferative response to infection. This is mediated in a similar fashion by recognition of a virus target antigen in association with class II MHC. These helper T cells can thereby stimulate proliferation of B cells to provide immunoglobulins for the humoral response (see below). Very few attempts have been made to examine the adenovirus antigens involved in the initiation of the proliferative response. A study on the lymphoid cells from one individual suggested that either the fibre or IIIa structural polypeptides could be targets (Souberbielle & Russell, 1995 ). However, a more general investigation noted that proliferative responses to the uncommon Ad35 occurred in individuals without any serological evidence of previous Ad35 infection (Flomenberg et al., 1995 ), implying that CD4⁺ T cells recognized a conserved antigen. This suggests that this arm of the immune system may play a role in modulating infection with a wide range of serotypes.

The humoral response
The humoral response is a major component of the defence strategy of the host and depends on the ability of B cells, elaborating surface immunoglobulins, to recognize a specific epitope on a foreign antigen. This recognition initiates a massive proliferation via T helper cells and thus the release of specific immunoglobulins of various classes into plasma to interact directly with these antigens. Where these are important in the initial interaction with the host cell, virus infection can be neutralized very efficiently. Given the importance of the fibre and penton base in the recognition of the receptors (see above), it is not surprising that adenovirus-neutralizing antibodies are directed against epitopes on these capsid components (Gahery-Segard et al., 1997 , 1998 ; Willcox & Mautner, 1976 ). However, there are also antigens on the hexon that induce neutralizing antibodies, and these seem to function by aggregating virus particles and thereby inhibiting adsorption. The efficacy of the humoral response in the case of adenovirus gene therapy is very important and depends on the nature of pre-existing immunity as well as the route and target of infection (Harvey et al., 1999 ).

For humans, there are 51 different adenovirus serotypes, classified on the basis of their specific neutralizing abilities, and protection by humoral antibodies is therefore tightly restricted to a given serotype. Further subdivision into species or subgenera A to F has also been made, using a variety of criteria (Benk et al., 1999 ). Type-specific antigens have been described in the fibre that are associated with the trimeric knob and the proximal regions of the stem (Watson et al., 1988 ; and W. C. Russell, unpublished data). In the case of the hexon, the type-specific epitopes reside, not surprisingly, on the hexon surface, whereas the internal antigens are conserved, being critical in the formation of the capsid structure, and therefore have a very much wider ‘group’ specificity. Group-specific hexon antibodies have been used extensively as general adenovirus diagnostic reagents.

	Adenoviruses as vectors

TOP Introduction General properties of... Defence mechanisms Adenoviruses as vectors Application of adenovirus... Future developments References

 
Adenoviruses can infect a wide variety of cell types and tissues in both dividing and non-dividing cells. This characteristic, together with their relative ease of preparation and purification, has led to their extensive use as gene vectors.

The virus can incorporate only about 2 kb of foreign DNA without significant affects on its stability or its infectivity, and the introduction of longer sequences therefore requires the removal of some or all of the virus genes. There are a range of techniques for constructing recombinant adenoviruses, and these are described in detail elsewhere (Hitt et al., 1997 ; Tashiro et al., 1999 ; Zhang, 1999 ).

Vectors can be utilized for: (i) cancer therapy to deliver genes that will lead to tumour suppression and elimination; (ii) gene therapy, i.e. to deliver genes to tissues to augment defective genes; (iii) supplementary therapy to deliver genes, expression of which will combat disease processes.

First-generation vectors
In the first generation of vectors, the E1 and/or E3 gene cassettes were removed, allowing the introduction of up to 6·5 kb of foreign DNA, often under the control of a heterologous promoter. In the case of the E1 deletions, care was taken to ensure the retention of the ITR and the packaging sequences. Removal of the E1 region had the additional apparent advantage of impairing the transcription of the E2 genes (which are E1 dependent) and consequently the replication of virus DNA and the production of the virus capsid proteins. However, it will be evident from the description of the E1 genes given above that there is also the disadvantage of the cellular environment being much less conducive to vector transcription. The defective E1 viruses could be propagated by infection of 293 cells (Graham et al., 1977 ), which provide the E1 gene products in trans. Although many of the initial studies in vitro provided much promise, it soon became evident that the expression of the transgene in vivo was only transient and was depressed because of the overwhelming immune response, mounted mainly against the virus capsid antigens as well as the expressed transgene. One of the reasons for this was the observation that many cells harboured E1-like proteins that allowed the E2 genes to function, albeit at reduced levels. In turn, this facilitated virus DNA replication and the synthesis of the late structural antigens and the production of replication-competent adenovirus (RCA). It also became evident that, at higher m.o.i., the E1 dependence of E2 gene transcription could be ablated. Bearing in mind these problems, a number of strategies have been adopted in an attempt to minimize the production of RCA (Hehir et al., 1996 ; Gao et al., 2000 ). Furthermore, as described above, removal of the E1B products also effectively disarmed one of the mechanisms for combating proapoptotic defences. In the case of the E3-deleted vectors, there were similar sequelae as a result of the elimination of the E3 gene-mediated defences against host responses (Poller et al., 1996 ).

Second- and third-generation vectors
The next approach was to construct vectors (using suitable complementing cell lines) with some or all of the E2 genes excised (Lusky et al., 1998 ; Moorhead et al., 1999 ) and hence with the capacity to replicate virus DNA and to produce RCAs removed. Generation of RCAs could also be prevented by constructing cell lines that do not contain adenovirus sequences that overlap those in the vector (Fallaux et al., 1998 , 1999 ). Nevertheless, the host immune response was still a major impediment to achieving persistent transgene expression and was particularly evident when repeated infections were attempted. A number of studies confirmed that the infecting recombinant virus itself was sufficient to induce the immune response, perhaps not surprising in view of the early activation of signalling cascades noted above and the potent antigenicity of the capsid components.

Other, rather more sophisticated vectors (third generation) have been constructed by deleting other virus genes (Amalfitano et al., 1998 ) and the latest of these have all or nearly all of the virus genes removed. These so-called ‘gutless’ vectors (Hardy et al., 1997 ; Kumar-Singh & Chamberlain, 1996 ; Lieber et al., 1999 ; Morsy et al., 1998 ; Steinwaerder et al., 1999 ) originally retained only the ITR and packaging sequences and required helper virus and appropriate complementing cells for propagation, followed by careful purification. Nevertheless, there were problems associated with these techniques, mainly due to contaminating helper virus and vector instability. A further development, which prevented the packaging of the helper virus, involved the use of the Cre-lox helper-dependent system (Chen et al., 1996 ; Hartigan-O’Connor et al., 1999 ; Ng et al., 1999 ; Parks et al., 1996 ; Tashiro et al., 1999 ).

Other methods to simplify and improve the construction of vectors have been described (He et al., 1998 ; Mizuguchi & Kay, 1998 ). A more comprehensive review (Hitt et al., 1997 ) provides details of most of the different techniques available for construction of vectors. One factor in fabricating these vectors is the need to maintain the vector size for efficient DNA packaging (Parks & Graham, 1997 ). This has been achieved by using ‘stuffer’ DNA, although the nature of this stuffer segment has been shown to influence transgene expression (Parks et al., 1999b ). These latest vectors have increased expression dramatically in vivo (Morral et al., 1999 ; Morsy et al., 1998 ; Ji et al., 1999 ). However, it has become clear that the retention of some of the E4 genes is important in combating the T cell response (Kaplan et al., 1999 ; Lusky et al., 1999 ; Yew et al., 1999 ) and more recent vectors have been modified accordingly (Gorziglia et al., 1999 ).

An extension of this approach involves the formation of hybrid vectors with adeno-associated virus (AAV) ITRs, which facilitate transgene integration (Lieber et al., 1999 ; Recchia et al., 1999 ). A similar strategy has been developed recently by using the long terminal repeats of Maloney leukaemia virus (Zheng et al., 2000 ) and has shown promise both in vitro and in vivo in a model system in facilitating transgene persistence. Hybrids with other viruses such as Epstein–Barr virus and retroviruses have also been developed (Caplen et al., 1999 ; Tan et al., 1999 ). Adenoviruses derived from other species (avian, ovine, bovine, canine) have been investigated as vectors for human gene therapy, since they do not normally invoke endemic humoral immunity (Hofmann et al., 1999 ; Kremer et al., 2000 ; Michou et al., 1999 ; Reddy et al., 1999 ; Zakhartchouk et al., 1998 ). Animal adenovirus vectors have also been used for animal vaccination (Hammond et al., 2000 ; Rasmussen et al., 1999 ).

Strategies for ensuring effective vectors
The effectiveness of gene therapy is governed in the main by the ability of the vector to be delivered to the relevant tissue and, once there, to express the gene product in appropriate quantities. This, of course, is exactly what the virus has sought to achieve in the course of evolution, and has been demonstrably successful in that adenoviruses are very prevalent without causing excessive morbidity. This seems to have been accomplished by utilizing virus gene products to delay the early innate and immune host defences, thus ensuring that the primary infection produces large amounts of virus. Propagation to other host cells can then be accomplished before the infected host’s full immunological armoury is deployed successfully. Whether adenoviruses are able to adopt other strategies to ensure their continuance, such as persistence or latency in the presence of an immune response, has never been adequately demonstrated, although adenovirus sequences can be detected in a proportion of the lungs from healthy individuals (Eissa et al., 1994 ; Elliott et al., 1995 ) as well as patients with pulmonary disease (Keicho et al., 1999 ) (for reviews see Lukashok et al., 2000 ; Mahr & Gooding, 1999 ). However, the facts that adenovirus immunity appears to be long lasting and that antibodies to the common serotypes 2 and 5 can be detected in almost 90% of individuals do suggest that persistence may be a factor in virus survival. This latter characteristic is obviously a desirable feature of an effective vector in some applications, but to achieve such an outcome in the tissue of choice will require a much greater understanding of the total spectrum of molecular mechanisms that operate in infection and of the resulting immunity.

Minimizing immune and apoptotic responses.
In view of the importance of the immune response in relation to transgene persistence, a number of studies have been carried out to unravel the role of the different arms of the immunological repertoire. Most of them have been carried out in model rodent systems, but a few have involved human subjects.

Humoral responses can be mounted, as noted above, with a single immunization (Juillard et al., 1995 ), but this can be modified to some extent by ensuring the retention of the E3 gene cassette in the vector as well as by treatment of the host with anti-CD4 reagents (Poller et al., 1996 ). This latter protocol reduced the population of T helper cells, which are needed for the activation of B cells and for the production of neutralizing antibodies. Another study implicated factors other than the capsid antigens in modulating the humoral response. It was shown that deletion of the E4 gene cassette diminished Th2 and B cell activities and it was postulated that an E4 gene product facilitated antigen presentation and the production of IL-6 and IL-8, which are important in B cell maturation (Armentano et al., 1997 ). Not surprisingly, a number of studies demonstrated that the administration of immunosuppressive agents such as cyclosporin, cyclophosphamide (Smith et al., 1996 ), FK506 (Ilan et al., 1997 ), deoxyspergualin (Kaplan & Smith, 1997 ) and CTLa4 Ig (Jooss et al., 1998b ) enhanced the persistence of the transgene product. Induction of tolerance has also been shown to be successful in some cases, leading to significant transgene persistence (Ilan et al., 1996 , 1998 ; Lee et al., 1999b ). Another route to immunosuppression was brought into play by the co-administration of an adenovirus vector with another that had a transgene expressing soluble CD8 or CD8 fused to the extracellular regions of a TNF receptor (Peng et al., 1999 ). This procedure successfully inhibited the action of TNF and significantly reduced the humoral antibody responses to both adenovirus and the transgene product. Another, more direct approach to minimizing antibody neutralization was achieved by covalently linking polyethylene glycol (O’Riordan et al., 1999 ) or a hydrophilic polymer based on N-(2-hydroxypropyl)methacrylamide (HPMA) to the capsid components of the virus (Fisher et al., 2000 ). This latter procedure also allowed retargetting of the vector. It should be noted, however, that the antibody response to the administration of a virus vector is influenced significantly by the pre-existing antibody status and by the route of administration (Harvey et al., 1999 ). Some improvement in transgene persistence can be achieved by repeated administration with vectors of different serotypes (Parks et al., 1999a ), although this has its limitations in view of the T cell cross-reactivities described above. Nevertheless, in spite of this plethora of techniques available to minimize the humoral antibody response, there is no doubt that the inability to ablate the response effectively remains a major impediment to exploitation of vectors (Benihoud et al., 1999 ).

In terms of T cells, a number of investigations have indicated that adenovirus gene delivery can elicit a complex panoply of cellular immune responses. CD4⁺ and CD8⁺ T cells specific for the transgene product as well as the vector can all be elaborated with variations dependent on the route of administration, the target organ and other factors such as the host genotype (van Ginkel et al., 1997 ) and development status (Kass-Eisler et al., 1994 ). In addition, for delivery to the lung, innate immune mechanisms involving the migration of alveolar macrophages seem to be very important (Worgall et al., 1997b ).

Apoptosis can also play an important part in minimizing transgene expression, and this can be combated to a significant extent by using vectors that express Bcl-2 both with and without NF-B inhibitors (Bilbao et al., 1999a , b ; de Moissac et al., 1999 ; Lieber et al., 1998 ), the expression of the inhibitors encouraging greater transgene persistence in mouse livers. The role of the E3 14.7K protein in attenuating inflammation was shown neatly by constructing transgenic mice in which this gene was expressed selectively by using a human SP-C promoter. There was a significant difference in lung inflammation and prolonged transgene expression when an E1/E3-deleted vector was administered (Harrod et al., 1998 ). A systematic investigation of the target proteins for CTLs and their histocompatibility restriction was undertaken in a murine model of liver gene therapy and revealed that the levels of CTL responses to adenovirus antigens and to the transgene product were varied and very dependent on the MHC haplotype of the host. A range of adenovirus antigens were examined in this system (pTP, Pol, DBP, hexon, penton and fibre) and the structural proteins, especially hexon, appeared to be the major targets (Jooss et al., 1998a ). Another survey examined both apoptosis and antibody formation in different strains of mice as a result of infection with an adenovirus vector and also concluded that there were differing responses depending on the mouse strain (Schowalter et al., 1999 ). An interesting study with nude mice, where the immune system was ablated, also concluded that the persistence of a transgene in mouse lung depended on the nature of the vector backbone and on the host background (Kaplan et al., 1997 ). These results suggest that the efficacy of therapy with adenovirus vectors will exhibit considerable heterogeneity in human populations.

Factors that affect delivery of transgenes.
The primary cellular receptors for adenoviruses appear to be distributed so widely in cells that effective and specific delivery to target cells would normally be precluded. On the other hand, a number of tissues and cells express very little, if any, of these receptors (Leon et al., 1998 ). Thus, the apical surfaces of ciliated airway epithelia, so important in dealing with treatment of cystic fibrosis, do not appear to have CAR available (Walters et al., 1999 ; Zabner et al., 1997 ), and the same is true of some primary tumours (Li et al., 1999c ; Miller et al., 1998 ). To permit targetted gene delivery, therefore, novel strategies need to be developed and a number of vectors have been constructed in an attempt to do this. Some of these have bispecific conjugates that can ablate the normal receptor binding and introduce novel tropisms, e.g. by using growth factor receptor (Miller et al., 1998 ), CD3 (Wickham et al., 1997 ), fibroblast growth factor (Printz et al., 2000 ), heparin (Wickham et al., 1996 ) or gastric releasing peptide (Hong et al., 1999a ). In this way, inflammatory vascular endothelial cells exhibiting E selectin can be targetted by complexing an anti-selectin E MAb with an anti-FLAG MAb and then attaching this dual antibody to a vector expressing the FLAG epitope (Harari et al., 1999 ). Another, more direct approach has been to incorporate binding motifs into the C-terminal domain of the fibre protein. This procedure facilitated binding to other cells without altering the endogenous binding, but this technique showed that specific delivery could be obtained in cells where the normal CAR was not expressed (Hidaka et al., 1999 ). A similar strategy using a variety of ligands proved promising in a model system for mouse gliomas (Staba et al., 2000 ).

One problem in attempting to produce vectors with novel receptors is the need for simultaneous development of culture cells that would allow good propagation of the vector. A neat approach to dealing with this has been demonstrated by incorporating six histidine residues (a His tag) into the H1 loop of the fibre knob (Krasnykh et al., 1998 ; Michael et al., 1995 ) and then using the modified virus to infect human glioma cells successfully (which lack normal receptors), which themselves had been modified to display a single-chain antibody against the His tag (Douglas et al., 1999 ).

A similar method, using a peptide from influenza virus haemagglutinin inserted into either the fibre or the penton base, was used successfully to infect cells expressing the single-chain antibody ligand (Einfeld et al., 1999 ). This approach could, in principle, be developed to construct vectors that have lost their native tropisms through mutation of the receptor-binding site on the fibre (Bewley et al., 1999 ; Kirby et al., 1999 ; Santis et al., 1999 ) and the RGD motif on the penton base (Chiu et al., 1999 ; Mathias et al., 1998 ) and therefore have the capacity to infect cells with other specificities. It has also been shown that delivery can be inhibited by protective extracellular matrices (van Deutekom et al., 1999 ) and that there are also anatomical barriers to overcome (Fechner et al., 1999 ).

A survey of a range of human adenovirus serotypes has demonstrated that some of them exhibit different and wider host tropisms, indicating that factors other than CAR must also operate. Thus, a chimeric type 2 adenovirus with a type 17 fibre can enhance gene transfer to airway epithelia (Zabner et al., 1999 ), in contrast with type 2 on its own. Similar use could be made of the properties of adenoviruses of subgenus H; thus, Ad41 binds selectively to differentiated gut enterocytes (Croyle et al., 1998b ). Viruses of subgroup D infect primary central nervous system cells more efficiently than do subgroup C (Chillon et al., 1999 ). One other strategy to ablate the binding properties of the fibre is to use fibreless virus. In this case, infectivity is reduced drastically but entry to cells can still be achieved at low levels via the RGD motif in the penton base (Legrand et al., 1999 ; Von Seggern et al., 1999 ). There is additional evidence that virus uptake can be mediated via the penton base alone and that the interaction with integrins can lead to a different route to the nucleus (Hong et al., 1999b ).

Other, more non-specific ways of bypassing normal receptor-mediated entry are by transfection with the aid of cationic lipids and polymers and by using calcium phosphate (Alton et al., 1999 ; Campain et al., 1998 ; Croyle et al., 1998a ; Dodds et al., 1999 ; Fasbender et al., 1997 , 1998 ; Lee et al., 1999a ; Qiu et al., 1998 ). Although not strictly a vectorial procedure, the ability of adenovirus to enter cells efficiently has been exploited by condensing a plasmid with polyethylenimine and then complexing with psoralen-inactivated adenovirus (Baker et al., 1997 ; Bischof et al., 1999 ; Edgell et al., 1998 ). A variation, with a simpler technique involving the ability of the virus mu peptide to package and deliver DNA to the nucleus with the aid of liposomes, has also been developed recently (Murray et al., 2000 ). A peptide derived from adenovirus fibre has recently been shown to target to the nucleolus and may provide a vehicle for gene delivery (Zhang et al., 1999 ). A combinatorial approach, using adenovirus transduction and plasmid transfection as well as lipofection, can also lead to enhancement of expression (Dunphy et al., 1999 ).

Ensuring expression of the transgene.
Assuming that there is effective delivery of the transgene to the host cell, the next step in successful expression depends greatly on the