Herpes simplex virus-based vectors

Abstract

Herpes simplex virus (HSV)-based vectors have primarily been developed for neuronal gene delivery, taking advantage of the virus' natural neurotropism. Two types of vector are available: replication defective viruses, whose cytotoxicity has been abolished by deleting viral gene products, and amplicon vectors, which are plasmids packaged into HSV particles with the aid of a helper virus. In this review I discuss how the cytotoxicity of the wild-type virus has been abolished, the progress which has been made toward defining promoter elements capable of directing long-term transgene expression form the latent viral genome and some of the potential clinical uses of these versatile vectors.

Introduction

As a highly cytotoxic human pathogen with a large and complex genome encoding at least 80 gene products, herpes simplex virus (HSV) is not at first sight the ideal gene delivery vector. However, not all natural infections with HSV lead to cell death; in neurones, the virus establishes a latent infection during which the viral genome persists indefinitely without any discernible adverse effects on the host cell. This natural neurotropism makes HSV an attractive candidate for neuronal gene delivery.

There are other features of HSV biology, which also make it attractive as a vector. The virus grows well in tissue culture and high titre stocks are easily obtained. Genetic manipulation of the viral genome is straightforward, and it can accommodate large amounts of foreign DNA. Efficient retrograde axonal transport of the virus means that neuroanatomical connections can be exploited to obtain efficient gene delivery to neurones distal from the site of inoculation. The latent HSV genome is maintained as an episome, and the potential risks of viral integration into host cell chromatin are avoided.

The major obstacles to the development of HSV as a vector have been the production of safe vector backbones and overcoming the tight transcriptional repression of the latent viral genome. In this review, I will discuss the strategies which have been used to address these issues and the potential applications of the resulting vectors.

HSV life cycle

It is a common feature of herpesviruses that they establish latent infections during which the viral genome persists in the absence of detectable infectivity. Latency allows virus to persist within the host in the presence of an immune response. Periodic reactivation and shedding of infectious virus leads to the infection of other, immunologically naïve, individuals. Herpesviruses are classified into subfamilies depending on the cellular site of latency; HSV-1 and HSV-2 are members of the Alphaherpesvirinae and establish latency in sensory neurones. HSV-1 primarily establishes latency in the trigeminal ganglia, and reactivation gives rise to orolabial cold sores, whilst HSV-2 establishes latency in the sacral ganglia and is associated with genital herpes, although there is considerable overlap. For the purpose of this review, I will confine my discussion to HSV-1.

HSV-1 is normally acquired in childhood or adolescence (Smith & Robinson 2002). Primary infection of the oral mucosa is mostly asymptomatic, although an ulcerative gingivostomatitis can occur (Whitley 2001). Progeny virus produced in the skin then enters sensory nerve terminals and the nucleocapsid is translocated to the cell body by microtubule-dependent retrograde axonal transport (Tomishima et al. 2001). This leads to the establishment of latency: using PCR-based methods, viral DNA can be detected in up to 3% of trigeminal ganglion neurones from seropositive cadaveric tissue (Cai et al. 2002). If primary infection occurs in the genital tract, or in other areas of skin, latency will be established in the relevant sensory ganglia (Whitley 2001).

In latently infected individuals, periodic reactivation of viral infection can lead to the reappearance of infectious virus at the periphery and recurrent cold sores, most commonly at the vermilion border of the lip. In the developed world, more than 40% of adults are seropositive for HSV-1 and presumably harbour latent virus, and in high-risk populations seropositivity rates can approach 100% (Smith & Robinson 2002). Only a minority of these individuals, however, suffer from recurrent cold sores, although many more may periodically shed virus asymptomatically in their saliva.

Although man is the only natural host for HSV infection, the virus is easy to grow in tissue culture and can be used to establish latency in a variety of animal models (Wagner & Bloom 1997), and this has allowed detailed molecular analysis of viral infection.

Viral entry

Entry of HSV-1 is a complex process involving several viral glycoproteins and a number of different cellular receptors (Spear & Longnecker 2003). Initial attachment to the cell is mediated by the interaction of envelope glycoproteins gB and gC with heparan sulfate on the cell surface. This brings the virus into proximity with the cell and allows gD to bind to the viral entry receptor. This receptor can also be heparan sulfate, but herpesvirus entry mediator (HVEM) and members of the nectin family bind gD as well. These molecules are widely expressed, which accounts for the broad cell tropism of HSV-1; human epithelial cells express HVEM and nectins and neurones express nectins.

Binding of gD to a cellular receptor results in fusion of the viral envelope with cellular membranes, mediated by gB and the gH-gL complex, and release of the nucleocapsid into the cytoplasm. The nucleocapsid is then transported to the nuclear pore, probably in a microtubule-dependent manner (Mabit et al. 2002), where the genome is extruded into the nucleus.

When infecting neurones, the viral nucleocapsid is transported retrogradely up the axon. The molecular basis of this transport is not well understood. It has been shown that the tegument protein US11 can interact with the microtubule-dependent motor protein kinesin (Diefenbach et al. 2002) and the ability to reconstitute retrograde and anterograde transport in the giant squid axon may help to elucidate other cellular motors which drive these processes (Bearer et al. 2000; Satpute-Krishnan et al. 2003). Axonal transport of HSV is not confined to the peripheral nervous system (PNS), and HSV has been used as a transneuronal tracer in many neural circuits (Norgren & Lehman 1998).

Productive cycle infection

The lytic pathway of HSV infection has been extensively studied in tissue culture cells and is characterized by the highly regulated sequential expression of three temporal classes of viral gene products: immediate early (IE), early (E) and late (L) (Roizman & Knipe 2001). Efficient expression of the five IE genes is dependent on VP16, a component of the viral tegument. VP16, in association with cellular factors, binds to a common sequence element found in IE gene promoters and activates transcription (Wysocka & Herr 2003). Four of the five IE genes (ICP0, ICP27, ICP4 and ICP22) are involved in the regulation of viral gene expression. The fifth, ICP47, has a role in immune evasion (Hill et al. 1995).

Successful establishment of IE gene expression leads to transcription of the E genes. E gene products include the viral DNA polymerase and other proteins which are involved in altering the intracellular milieu to favour viral replication. The L genes predominantly encode structural proteins and are, for the most part, transcribed following viral DNA replication. The productive cycle of HSV infection inevitably leads to death of the host cell.

Latency

HSV latency is characterized by the persistence of episomal viral DNA within sensory neurones in the absence of detectable infectivity. Infectivity can be reactivated by explantation of the ganglia or other stimuli specific to individual animal models (Lachmann 2003).

In the first few days after infection, it is possible to detect IE promoter-driven gene expression within sensory neurones which innervate the site of inoculation (Lachmann et al. 1999). After this time, however, all viral productive cycle promoters become transcriptionally repressed (Preston 2000) and no viral protein products have been reliably detected in latently infected neurones. There is, however, continuing transcription from the viral latency-associated promoter (LAP) which gives rise to a family of nuclear RNAs termed the latency-associated transcripts (LATs) (Stevens et al. 1987). LAT expression seems to have a role to play in the efficient establishment of latency (Thompson & Sawtell 1997). LATs can down-regulate IE gene expression and act as antiapoptotic factors, but their exact function is not presently well understood (Kent et al. 2003).

Reactivation

Reactivation of infection requires the re-establishment of productive cycle gene expression. Studies using deletion mutants in tissue culture and animal models of latency suggest that expression of ICP0 has a key role to play in derepression of the viral genome (Lachmann 2003), but the molecular events underlying reactivation are not currently well understood.

Development of HSV vectors

Different applications place different requirements on a vector system. In most cases, the vector needs to be fully replication disabled and to lack all cytotoxicity, although some ‘gene therapy’ approaches to cancer utilize mutant viruses which selectively replicate in and kill malignant, but not nonmalignant, cells. Gene therapy to correct genetic deficiencies requires long-term expression of a transgene and delivery of physiologically appropriate levels of protein to all affected cells. Cancer gene therapy often only requires transient, high level expression of a potentially toxic product specifically in malignant cells.

For these reasons, it is not possible to produce a universal vector which would be suitable for every application. Nonetheless, two parallel themes have emerged in the development of HSV-based vectors: the production of noncytotoxic, safe vectors and the definition of promoter elements capable of overcoming the transcriptional repression of the latent viral genome.

Development of safe vector backbones

Vector cytotoxicity

Attenuated mutants

For many years, it has been known that a number of mutant viruses demonstrate attenuated neurovirulence. Neurovirulence factors such as thymidine kinase (TK) (Field & Wildy 1978; Chrisp et al. 1989; Efstathiou et al. 1989) and ribonucleotide reductase (RR) (Cameron et al. 1988; Jacobson et al. 1989) are involved in nucleotide metabolism. These genes are not essential for viral replication in dividing cells but are required in fully differentiated cells, such as neurones, where the virus needs to synthesize its own DNA precursors. The avirulence of ICP34.5-null viruses (Chou et al. 1990; Bolovan et al. 1994) relates to the inability of these mutants to block the interferon (IFN) response of cells to viral infection. ICP34.5 prevents the IFN-induced accumulation of phosphorylated eIF2α which is required for inhibition of protein synthesis (Chou et al. 1995; He et al. 1997).

In vivo, viruses which are null for ICP34.5, either alone or in combination with RR, have been shown to replicate selectively in rapidly dividing malignant cells and are being developed as oncolytic agents (Post et al. 2004). A number of phase 1 clinical trials in malignant glioma have been performed (Shah et al. 2003). Although they lack neurovirulence, after inoculation these viruses can infect and express viral gene products in a variety of normal, nonmalignant cell types (Markovitz et al. 1997; Kesari et al. 1998), and they also elicit a marked inflammatory response (McMenamin et al. 1998). Therefore, although they do not spread efficiently through the brain and cause encephalitis, these attenuated viruses are toxic to the cells which they infect and are not suitable as gene delivery vectors.

Replication defective vectors

HSV encodes more than 80 viral gene products and about half of these are essential for viral replication. It is relatively easy to make a replication defective mutant by deleting one of the essential genes and providing the gene product in trans by means of a complementing cell line. Such mutant viruses are unable to replicate in noncomplementing cells, but they remain cytotoxic due to the fact that they still express many viral gene products. Even mutant viruses deleted for the essential IE3 (ICP4) gene, which can not initiate E and L gene expression, are cytotoxic (Johnson et al. 1992). This toxicity is due to the expression of the other IE gene products (Johnson et al. 1994). Therefore, the best way to minimize vector toxicity is to prevent viral gene expression altogether. Two different strategies have been used to generate such vectors.

In an attempt to establish an in vitro model of viral latency, Preston and colleagues have constructed vectors based on conditional mutations in a number of viral gene products. They first created a mutation in the virion transactivator, VP16, which prevents it from forming a protein–DNA complex and transinducing IE gene expression (Ace et al. 1989). This virus has a very high particle to plaque-forming unit (pfu) ratio and is avirulent when injected into mice. At high multiplicity of infection, however, it is still capable of undergoing full productive replication. Further mutations were introduced, a temperature-sensitive mutation in ICP4 (tsK) and disabling mutations in ICP0 (Preston et al. 1997; Preston et al. 1998). The resulting vectors can be propagated to high titres in permissive conditions [using BHK cells, which have an endogenous activity which complements the ICP0 defect (Preston et al. 1997), at 31 °C, the permissive temperature for the tsK mutation, with the addition of hydroxymethyl bisacetamide, a transcriptional activator which complements the VP16 defect (McFarlane et al. 1992)]. In nonpermissive conditions, however, infection at a multiplicity of 5 pfu per cell resulted in no detectable cytopathology and the persistence of the viral genome within cells.

The alternative approach, originally reported by DeLuca's laboratory, has been to delete all the IE genes from the viral genome and construct a complementing cell line to provide the gene products which are necessary for efficient viral growth (ICP0 and the two essential IE products ICP4 and ICP27) (Samaniego et al. 1997; Samaniego et al. 1998). The virus d109, which has been deleted for all five IE genes, could be used to infect noncomplementing cells at MOIs of up to 30 with no detectable cytotoxicity (Samaniego et al. 1998).

Vectors based on both conditional viral mutants and deletion mutants can safely be administered to experimental animals by direct stereotactic injection into the brain and are efficiently translocated to distal sites by retrograde axonal transport (Lilley et al. 2001; Scarpini et al. 2001).

Amplicon vectors

The HSV amplicon was first developed as a eukaryotic cloning vector (Spaete & Frenkel 1982). If cells are cotransfected with viral DNA and a plasmid which contains the HSV origin of replication and DNA packaging signals, the resulting progeny will be a mixture of virions which contain the viral genome and virions which have packaged concatameric plasmid sequences. Although amplicons could be used to efficiently express genes in a variety of eukaryotic cells including neurones (Kwong & Frenkel 1985; Lowenstein et al. 1994), their use as gene therapy vectors was limited by the fact that amplicon stocks tended to be of low titre and were invariably contaminated with cytotoxic helper virus.

By providing the helper virus as a set of cosmids which could not be packaged, Fraefel et al. were able to generate helper virus-free amplicon stocks (Fraefel et al. 1996), but this method relied on cotransfection and produced low-titre stocks which could not be passaged and amplified. Other groups used a strategy where replication of helper virus and amplicon were made interdependent by using mutant helper viruses which required trans-acting functions provided by the amplicon plasmid to be able to replicate efficiently (Pechan et al. 1996; Zhang et al. 1998). This allowed serial passage and the production of high-titre stocks, but helper virus was still present which, as it was not fully replication defective, produced cytotoxic effects.

Attempts to use the cre-lox system to excise the packaging sequences from the helper virus are subject to similar limitations (Logvinoff & Epstein 2001; Zaupa et al. 2003). When amplicon stocks are generated and passaged using a helper virus in which the packaging sequences are flanked by lox sites, it is possible to prevent the helper virus from being packaged by performing the final passage in a cre-expressing cell line. Although this method can generate high-titre amplicon stocks, there is still helper virus contamination.

The next step was to return to a transfection-dependent helper-free system by cloning a HSV genome in which the packaging signals had been deleted as a bacterial artificial chromosome (BAC) (Saeki et al. 1998; Stavropoulos & Strathdee 1998) (Figure 1). Cotransfection of eukaryotic cells with the BAC and an amplicon plasmid resulted in the production of supernatants containing 10⁷ amplicon particles per ml. These stocks, however, still contain small amounts of replication-competent helper virus, probably resulting from recombination events between the BAC and amplicon plasmids (Saeki et al. 1998; Saeki et al. 2001). By deleting an essential gene from the BAC and using stuffer DNA to make it too big to be packaged into a viral capsid, Saeki et al. have managed to generate stocks of amplicon in which replication-competent helper virus is undetectable (Saeki et al. 2001).

Figure 1.

The helper-free amplicon plasmid-packaging system. The two components of the amplicon system are the amplicon plasmid, which contains the herpes simplex virus (HSV) origin of DNA replication (Ori), the HSV-packaging sequence (Pac), an antibiotic resistance gene (Amp) and a promoter-transgene cassette, and a helper virus to provide viral replication functions in trans. In the most advanced systems, a helper viral genome which has had its packaging signals deleted is cloned into a bacterial artificial chromosome (BAC). When eukaryotic cells are cotransfected with amplicon and HSV-BAC DNA, the resulting viral particles can only recognize and package the amplicon plasmid, resulting in helper virus-free stocks of amplicon.

The most advanced amplicon vectors incorporate the plasmid-maintenance functions of Epstein–Barr virus. These eukaryotic cloning vectors can package more than 100 kb of DNA and are maintained indefinitely in cells in tissue culture (Wade-Martins et al. 2001).

As amplicon particles and virus particles share the same nucleocapsid and envelope, stereotactic injection of amplicons into the brain also allows efficient gene delivery to neurones (Sandler et al. 2002). Amplicon stocks contaminated with helper virus produce considerable inflammation throughout the brain after stereotactic injection (Wood et al. 1994), but this is markedly reduced when helper virus-free preparations are used (Olschowka et al. 2003).

Recombinatorial stability

One of the potential risks of using viral vectors is that recombination events will give rise to replication-competent virus. If this occurred during vector propagation, by recombination between the vector and viral sequences present in a complementing cell line, it would give rise to vector stocks which are contaminated with wild-type virus and pose a potential risk to the gene therapy subject (see discussion of amplicon vectors above). Additionally, recombination between vector and wild-type HSV DNA, either during vector propagation or in the patient, could produce a replication-competent virus capable of expressing a biologically active transgene, which could then escape into the environment. In order to avoid such events, a number of strictures have been put on the construction of vectors and complementing cell lines.

To prevent recombination between cell lines and vectors, it is necessary to ensure that all viral sequences inserted into the complementing cell line have been entirely deleted from the vector: if no sequence elements are shared between the cell line and the vector, there can be no recombination. An example of this is the gH-deficient disabled infectious single cycle (DISC) virus which has been developed as a potential vaccine (Forrester et al. 1992). Stocks of infectious virus are produced using a cell line which expresses gH under the control of a viral late promoter (the gD promoter). If this virus is used to infect noncomplementing cells, in vitro or in vivo, it undergoes a single round of replication but the progeny virions are noninfectious as they lack gH which is essential for infectivity. DISC stocks produced using the complementing cell line have been shown to contain not more than one in 10⁹ pfu of revertant, gH+, virus (Speck et al. 1996).

Vector and replication-competent viruses could infect the same cell either during vector propagation, if stocks become contaminated, or in the gene therapy subject. Given the high seroprevalence rates, it is likely that the majority of patients will harbour latent HSV within sensory ganglia and possibly elsewhere in the nervous system (Itzhaki et al. 1997; Arbusow et al. 2000). During a reactivation event, it is possible that wild-type virus could replicate in a cell harbouring a vector genome and give rise to recombinant progeny which was replication competent and carried the therapeutic transgene. The risks associated with such recombination can be minimized by ensuring that the transgene is inserted into an essential viral locus. In this way, it is possible to ensure that recombination of the transgene into any viral genome will, at the same time, remove an essential gene and inactivate the resulting virus.

Strategies for transgene expression

Obtaining transient gene expression in cells infected with replication-defective viral vectors or with amplicon vectors is straightforward, both in tissue culture and in vivo. Defining promoter elements capable of directing sustained, long-term gene expression has, however, proved much more challenging.

Transcriptional repression of the latent viral genome

One of the major characteristics of the latent viral genome is the tight transcriptional repression of viral lytic cycle promoters (Preston 2000; Lachmann 2003). This repression is not only seen in latently infected neurones but also applies to quiescent infection of tissue culture cells by fully replication defective viruses (Preston & Nicholl 1997; Samaniego et al. 1998). This strategy is advantageous to the virus as it allows it to persist within the host in the presence of an active immune response, but it is an obstacle to vector development, especially as the repression also applies to other promoter elements inserted into the viral genome; a recombinant virus containing a reporter gene under transcriptional control of a strong eukaryotic promoter will express abundant transgene product during productive infection, in tissue culture or in animal models, but in latency the promoter will be repressed and little transgene will be expressed (Lokensgard et al. 1994; Bloom et al. 1995; Ecob-Prince et al. 1995; Lachmann et al. 1996; Goins et al. 1999) (the occasional neurones expressing transgene seen in some of these studies may represent occult reactivation events).

Transgene expression during neuronal latency

The only viral promoter which is not subject to this transcriptional silencing is the LAP, although this promoter only remains active in a proportion of latently infected neurones (Ramakrishnan et al. 1994; Mehta et al. 1995). Studies attempting to utilize the LAP to drive reporter gene expression in latently infected neurones have shown that sequences downstream of the transcriptional start site are important for authentic promoter activity. Recombinant viruses in which the rabbit β-globin gene is inserted immediately downstream of the LAP TATA box (Dobson et al. 1989), or in which the first 1.5 kb of primary LAT-coding sequence is replaced with the lacZ gene (Margolis et al. 1993), display high levels of transcription from the LAP during acute infection of ganglia, but this drops during latency, and only small amounts of transcript are observed in a few cells. Similarly, a virus in which the rat β-glucuronidase gene was inserted into a 900-bp deletion downstream of the core LAP (Wolfe et al. 1992) showed a marked reduction in the number of latently infected neurones in which the LAP was transcriptionally active as compared to a wild-type virus (Deshmane et al. 1995). Attempts to define a core LAP which can be used to drive latent phase gene expression from an ectopic locus in the virus have also been made. If an 865-bp fragment containing the core LAT promoter is removed from the terminal repeats and used to drive β-gal expression from the gC locus, then, although there is transcription during the acute phase of infection, it is not possible to detect any β-gal activity in latently infected neurones (Lokensgard et al. 1994). Sawtell and Thompson, using the same recombinant viruses as Margolis et al. and Lokensgard et al., have reported better levels of latent phase β-gal expression in murine trigeminal ganglia (Sawtell & Thompson 1992). Nonetheless, as in the other studies, they saw a drop in promoter activity as the virus entered the latent phase. There is evidence that transcription from the endogenous LAP is repressed during the acute phase of infection (Margolis et al. 1992; Smith et al. 1994; Lachmann et al. 1999) (probably by ICP4), and that LATs gradually accumulate as latency is established. Thus, all these LAP constructs display kinetics which are markedly different from the ‘true’ promoter.

For authentic LAP activity, it is necessary to conserve the regions 3′ of the core promoter, and we have shown that if an insertion is made 1.5 kb downstream of the transcription start site, it is possible to maintain latent phase transcription (Lachmann et al. 1996). The resulting transcript has a long 5′ leader sequence which contains multiple stop codons making it unlikely that a ribosome scanning from the 5′-cap would ever reach the initiation codon of a transgene inserted at this locus. If the transgene is linked to an internal ribosomal entry site (IRES), allowing the ribosome direct access to the initiation codon, this problem can be overcome and such a virus is capable of directing reporter gene expression which is stable for many months and exhibits kinetics which mirror those of the viral LATs (Lachmann & Efstathiou 1997; Lachmann et al. 1999).

Attempts have also been made to utilize elements of the LAP to confer latent-phase activity on other promoter elements. Hybrid promoters combining LAP sequences with the Moloney murine leukaemia virus long-terminal repeat (MoMuLV LTR) (Lokensgard et al. 1994; Carpenter & Stevens 1996) or the human cytomegalovirus major IE (HCMV IE) promoter (Lilley et al. 2001) are able to drive long-term transgene expression in significant number of latently infected neurones. Interestingly, the downstream promoter region which is required for authentic activity of the LAP is not maintained in all these constructs, and it remains unclear exactly what sequence elements are required to overcome the transcriptional silencing of the latent genome.

By inserting promoter constructs into replication defective vectors, it is possible to investigate long-term gene expression in the CNS as well as the PNS (Figure 2). There have been many reports of ‘long-term gene expression’ in the brain using a variety of expression cassettes, but often this involves only a few cells, and studies which quantitatively address the kinetics of transgene expression are rare. We have shown that LAP-driven transgene expression peaks around 2 weeks after inoculation, at which time between 10⁵ and 10⁶β-galactosidase-positive neurones can be detected throughout the ipsilateral cortex and substantia nigra after a single striatal injection of 2 × 10⁷ pfu of vector (Scarpini et al. 2001). After this time, however, there is a gradual decrease in the number of transgene expressing neurones such that, after 6 months, in the order of 10³ positive cells remain. This loss of transgene expression appears to relate to transcriptional down-regulation rather than an immune response to the transgene, as similar results were obtained when a vector with an intact LAT region was injected and in situ hybridization for the LAT transcripts (which do not encode a protein product) was used to quantify the number of latently infected neurones. It is possible that latently infected neurones were gradually lost over time, although total viral DNA levels in the brain remained constant.

Figure 2.

Viral recombinants used to obtain long-term transgene expression in the central nervous system. A schematic of the herpes simplex virus (HSV) genome is shown with its long and short unique regions (U_L and U_S) each flanked by terminal inverted repeats (TR_L and IR_L, TR_S and IR_S). The region of the internal repeats has been expanded to show where transcription of the primary latency-associated transcript (LAT) starts. The position of the major LAT intron is indicated. Scarpini et al. have reported the use of a recombinant virus in which an internal ribosomal entry site (IRES)-linked reporter gene is inserted into a 168-bp deletion within the major LAT-coding region. Lilley et al. have reported the use of a recombinant virus in which a reporter gene under transcriptional control of the HCMV IE promoter has been inserted into an 188-bp deletion which starts 82 bp upstream of the locus used by Scarpini et al.

Lilley et al. have demonstrated that by inserting a lacZ gene driven by the HCMV IE promoter into the region downstream of the LAP transcription start site (Figure 2), they can obtain transgene expression in a wide variety of CNS neurones (Lilley et al. 2001). Once again, however, this expression appears to peak at around 1-week post inoculation and then to gradually diminish.

The kinetics of gene expression from the LAP in the PNS are therefore different from those seen in the CNS, although there may be populations of CNS neurones which will support indefinite expression of the LATs (Lachmann & Efstathiou 1997; Smith et al. 2000). Obtaining genuine long-term expression from the latent viral genome in the brain remains an outstanding challenge.

Transcription from amplicon vectors

Amplicons can be used to deliver transgenes in the same way as replication defective HSV vectors (e.g. by peripheral inoculation or direct injection into the brain). Interestingly, although they only contain viral packaging and origin sequences, they are subject to a transcriptional repression which is very similar to that seen with the latent viral genome (Ho et al. 1993; Fraefel et al. 1996) and it has not been possible to obtain long-term gene expression using standard promoter elements.

The best long-term gene expression has been observed when large, neuronal-specific promoter elements are inserted into amplicon plasmids, which can potentially accommodate inserts of almost 150 kb of DNA. A 2.7-kb fragment of the rat preproenkephalin promoter was able to direct transgene expression in relevant neuronal populations for at least 2 months, although expression levels appear to have dropped dramatically over time (Kaplitt et al. 1994). Even larger regions of the tyrosine hydroxylase promoter (6.8 and 9.0 kb) have been shown to drive stable transgene expression for at least 10 weeks in dopaminergic neurones (Jin et al. 1996; Song et al. 1997).

The large packaging capacity of amplicon vectors would potentially allow the delivery of entire genomic DNA loci rather than cDNAs. An amplicon vector containing a 135-kb genomic insert encompassing the human low-density lipoprotein receptor (LDL-R) has been used to obtain authentically regulated LDL-R expression in LDL-R-deficient cells in vitro (Wade-Martins et al. 2003), although no in vivo data have so far been reported.

Experimental applications

Many studies have been published on the use of replication-defective HSV and amplicon vectors to deliver transgenes to a wide variety of cell types both in vitro and in vivo. Here, I will concentrate on a few areas of research which illustrate the versatility of HSV-based vectors, and some of the problems associated with their use, and which could potentially lead to clinical applications in the future.

Cancer immunotherapy

The host immune response against a tumour can be stimulated by expressing immunomodulatory cytokines, such as granulocyte/macrophage-colony stimulating factor (GM-CSF), within the malignant cells (Dranoff 2002). A number of different viral vectors are being developed as cancer vaccines, but one particularly attractive candidate, as it has entered phase I/II clinical trials as a vaccine against genital herpes (Whitley & Roizman 2002), is based on the gH-deficient DISC system described above. DISC viruses can be used as gene delivery vectors, producing transient, high level gene expression before killing the transduced cell (Rees et al. 2002). DISC-GM-CSF, which expresses human GM-CSF under transcriptional control of the HCMV IE promoter (Ali et al. 2000), is being developed for immunotherapy of a variety of human cancers (Ali et al. 2002; Loudon et al. 2003; Parkinson et al. 2003) and phase I clinical trials have been conducted.

Neuroprotection in stroke

In order to take advantage of the natural neurotropism of the virus, HSV vectors have primarily been designed for gene delivery to the nervous system. Of the many potential applications of gene therapy within the brain, perhaps the most attractive are those which require short-term, unregulated expression of a transgene in a well-defined anatomical location. These approaches allow the use of simple expression cassettes and single injections of vector.

Acute ischaemic stroke fulfils these criteria. Although the primary therapeutic goal in acute stroke is reperfusion, there has also been progress in developing pharmacological strategies to protect neurones from the ischaemic insult and minimize cell death (Lees 2000). Such neuroprotection could also be provided by gene therapy. As the molecular pathways which result in neuronal death following ischaemia have been elucidated, it has been possible to identify a number of potential targets for therapeutic gene delivery (Sapolsky & Steinberg 1999; Yenari et al. 2001a); genes involved in glucose metabolism and excitotoxicity, neurotrophins, chaperonins and antiapoptotic factors.

Amplicons have been used to deliver a number of therapeutic genes in a rat model of ischaemic stroke. It has been shown that overexpression of the glucose transporter GLUT-1 improves glucose uptake (Ho et al. 1993) and reduces neuronal loss (Lawrence et al. 1996). Neuronal survival can also be enhanced by delivering the calcium-binding protein calbindin D-28K (Yenari et al. 2001b) and by inhibiting apoptotic cell death by overexpression of the heat shock protein HSP72 (Hoehn et al. 2001; Kelly et al. 2002), the proto-oncogene Bcl-2 (Lawrence et al. 1997; Yenari et al. 2003) and the free radical scavenger glutathione peroxidase (Hoehn et al. 2003). Neurotrophic factors have been used to achieve similar results in stroke (Harvey et al. 2003) and other models of neuronal toxicity (Bowers et al. 2002; Kalwy et al. 2003; Lamigeon et al. 2003).

These results are promising, but if they are to be therapeutically useful, it will be necessary to transduce large number of cells in the ischaemic territory and to define the temporal window during which transgene delivery will be successful in achieving a good functional result (Dumas & Sapolsky 2001; Yenari et al. 2003). To date, the vectors have been delivered by stereotactic injection, which is unlikely to be practicable in the acute clinical setting. A more attractive scenario would be to administer the vector intravascularly at the time of reperfusion (when disruption of the blood brain barrier might allow virus to enter the brain parenchyma), but to date, there are no experimental data concerning this approach.

Parkinson's disease

The application of neuroprotection strategies to neurodegenerative diseases is more challenging as these conditions often affect large volumes of the brain and require long-term delivery of neuroprotective agents. Parkinson's disease (PD) is an attractive target for gene therapy, as the affected neurones are found in a single anatomical region and well-characterized animal models exist (Orth & Tabrizi 2003).

The initial attempts at gene therapy in PD were based on the pharmacological approach of supplementing deficient dopamine (DA) production. An amplicon vector was used to deliver the tyrosine hydroxylase (TH) gene to the striatum of Parkinsonian rats. This resulted in increased striatal DA production and a striking and prolonged behavioural recovery (During et al. 1994). This response was all the more remarkable given that the TH gene was under the control of a viral IE promoter, which is only capable of directing significant transgene expression for a few days. Interpretation of the data was further affected by the fact that a proportion of the animals died as a result of helper virus contamination of the amplicon stocks.

Recently, the same group have developed a helper virus-free amplicon containing the genes for TH and aromatic acid decarboxylase (the enzyme responsible for the conversion of L-DOPA to DOPA) under control of a TH-neurofilament chimeric promoter (Sun et al. 2003) which has been shown to drive reporter gene expression in rat striatum for at least 6 months (Zhang et al. 2000). This refined system produced biochemical correction in the striatum of Parkinsonian rats for 4 months and behavioural correction for 5 weeks.

An alternative approach has been to utilize a neuroprotective strategy to prevent the loss of dopaminergic neurones in the substantia nigra (SN). Replication-defective HSV vectors have been used to transiently express Bcl-2 and glial cell-line-derived neurotrophic factor (GDNF) in the SN of 6-hydroxydopamine-lesioned rats (Yamada et al. 1999; Natsume et al. 2001). This resulted in up to 50% increase in the number of surviving dopaminergic neurones in the SN and a decrease in abnormal rotational behaviour.

In most cases, PD in humans is a gradual neurodegenerative process rather than being secondary to an acute toxic insult as it is in lesioned rats. Extending these results to human patients will therefore be difficult, as it will require long-term expression of the neuroprotective factors which may itself lead to unforeseen adverse effects.

Pain

Although HSV vectors can be used to deliver genes to a wide variety of cells within the nervous system and in other tissues, the peripheral sensory neurone is the most attractive target, as it exploits the biology of the virus. Peripheral inoculation of replication defective vectors leads to efficient establishment of latency in the relevant dorsal root ganglions (DRGs) (Marshall et al. 2000). LAP-driven gene expression in primary sensory neurones is lifelong (Lachmann & Efstathiou 1997; Marshall et al. 2000), and there is also evidence of long-term CMV IE promoter-driven transgene expression in a small proportion of latently infected neurones (Ecob-Prince et al. 1995; Wilson et al. 1999).

The perception of pain is complicated and involves structures at all levels of the neuraxis, including the peripheral sensory neurone and the dorsal horn of the spinal cord (Bolay & Moskowitz 2002). The mainstay of treatment for pain is the opioid drugs which are active in the dorsal horn as well as at other sites (Sawynok 2003). Systemically administered opioids have a number of undesirable side effects, and chronic administration can lead to tolerization. The use of gene therapy to deliver peptide opiate agonists directly to the sensory neurone could avoid these problems (Pohl et al. 2003).

Peripheral inoculation of a replication-defective HSV vector containing the proenkephalin gene under control of the HCMV IE promoter results in expression of proenkephalin in DRG neurones (Wilson et al. 1999). In a rat model of inflammatory pain, this vector gave an analgesic effect for up to 4 weeks which was re-established after reinoculation (Goss et al. 2001). Analgesic efficacy has also been demonstrated in models of arthritis (Braz et al. 2001), bone pain in cancer (Goss et al. 2002b) and neuropathic pain (Hao et al. 2003a; Yeomans et al. 2004).

Other promising approaches have been to deliver the neurotrophic factor GDNF (Hao et al. 2003b), which has an analgesic effect when given intrathecally, and to use an antisense strategy to reduce expression of the nociceptive neuropeptide calcitonin gene-related polypeptide (Wilson & Yeomans 2002). HSV vectors have also been used to deliver neurotrophins in models of peripheral neuropathy (Goss et al. 2002a; Chattopadhyay et al. 2003; Chattopadhyay et al. 2004).

The data from animal models are encouraging and gene therapy may be a valuable approach to the treatment of chronic pain. The next step will be to design a clinical trial to extend these findings to human patients (Mata et al. 2003).

Conclusions

Aspects of its biology make HSV an attractive candidate for neuronal gene delivery: it is efficiently translocated from the periphery to sensory neurones in the PNS and along neural pathways in the CNS; during latency, the viral genome persists indefinitely as an episomal element in the absence of viral protein expression. Noncytotoxic vector backbones have been developed, and these will potentially allow HSV to be used for gene delivery to non-neuronal cells as well. To date, stable long-term transgene expression has only been obtained in peripheral sensory neurones and further work will need to be done to determine whether it is possible to modify the LAP to remain active in significant number of neurones within the brain.

Although HSV vectors were initially conceived for gene delivery to the brain, the most promising potential applications at present are in the PNS and in oncology. The fact that various genetically modified HSVs, which are much less replication disabled than the potential vector backbones, have already been administered to patients in trials for malignant glioma (Markert et al. 2000) and as potential vaccines (Whitley & Roizman 2002), should make it easier to obtain regulatory approval for clinical trials. As with all gene therapy applications, it will be interesting to see if the positive results obtained in small rodent models can be reproduced in human patients.

Perhaps the most exciting prospect is the possibility of using amplicon vectors to deliver entire genes, complete with all potential upstream and downstream regulatory elements and introns. The most advanced amplicon vectors have enormous packaging capacities, are easily manipulated in bacterial systems, can be grown to high titres without contaminating helper virus and are capable of infecting a wide variety of different cell types in vitro and in vivo. Such vectors might turn out to be the nearest thing we have to a universal gene delivery system.