Gene Ther Mol Biol Vol 9, 153-168, 2005

Key words: cancer gene therapy, viral vectors, adenovirus, targeting, oncolytic viruses, protein transduction domains

Abbreviations: central polypurine tract, (cPPT); conditionally replicative adenoviruses, (CRAds); coxsackie- adenovirus receptor, (CAR); cyclo-oxygenase-2, (COX-2); early growth response 1, (EGR-1); glycoprotein from vesicular stomatitis virus, (VSV-G); herpes simplex virus type 1, (HSV-1); human telomerase reverse transcriptase, (hTERT); inverted terminal repeats, (ITRs); long terminal repeats, (LTR); posttranscriptional regulatory elements, (PRE); Protein transduction domains, (PTDs); self-inactivating type vectors, (SIN);

Cancer is a world wide health problem. In 2002, approximately 11 million people were estimated to be diagnosed with and almost 7 million people to die from cancer (Ferlay et al, 2004). The number of cancer patients has increased partly due to achievements in medical sciences, including reduction of deaths from infectious and cardiovascular causes, but also improved diagnostic methods for more sensitive detection. Further, many factors in the Western lifestyle may predispose to carcinogenesis.

Even though cancer can be considered a genetic disease, because it is caused by mutations and epigenetic changes in tumor suppressor and oncogenes, it's rarely caused by a defect of a single gene and the development of malignant tumors usually involves interaction between the environment and heredity (Hemminki and Hemminki, 2004; Tamura et al, 2004). It has been hypothesized that three to seven genetic changes are required for carcinogenesis (Vogelstein and Kinzler, 1998). They typically cause increased and irregular proliferation activity, lack of apoptosis, avoidance of immune defences, activation of telomerase and ability to metastasise and form vascularisation into tumor (Hanahan and Weinberg, 2000). When accumulated, these aberrant features result in cells which can proliferate unrestrictedly, form neoplastic lesions, invade local tissues and eventually establish distant metastasis (Zhang et al, 1995; Rieger, 2004). Even though the knowledge of molecular mechanisms, diagnostic methods and treatments of cancer have improved during the last decades, most cancer types still imply poor prognosis and high mortality, especially when metastatic. In fact, metastatic solid tumors can be cured only very rarely. Thus, more efficient approaches and novel tools are needed for treatment of cancer.

Although the theoretical basis for gene therapy is rather simple and thus attractive, practical experience has demonstrated some obstacles to overcome. Nevertheless, perhaps the main finding thus far has been the generally very good safety data in clinical trials. Despite receiving high publicity, severe toxicity has been reported only rarely (Hacein-Bey-Abina et al, 2003; Raper et al, 2003). In particular, when compared to other experimental treatments (such as novel chemotherapeutics or eg. bone marrow transplantation) studied for life threatening illnesses; gene therapy has been well tolerated. While safety data has been promising, efficacy has been more variable. In a nutshell, trials have demonstrated that the key to safe and effective therapy is efficient and specific gene transfer, which is a sum of multiple factors. This review will focus on cytotoxic approaches, while immunomodulatory approaches are discussed elsewhere.

In order to deliver genes into target tissue, gene transfer vectors have to first negotiate the host immune system. It has been reported that 55% of adult humans may have some level of pre-existing circulating neutralizing antibodies against adenovirus serotype 5, which is one of the most used gene transfer vectors (Chirmule et al, 1999). In contrast, retroviruses, another group of viral vectors, rarely elicit neutralizing antibodies but they can be rapidly degraded by the complement system (Takeuchi et al, 1994). In addition, the route of viral vector administration plays an important role in gene delivery. It has been shown in several preclinical studies that regardless of the administration route adenoviral vectors almost invariably evoke neutralizing antibodies (Setoguchi et al, 1994; Van Ginkel et al, 1995; Smith et al, 1996; Gahery-Segard et al, 1997; Hemminki et al, 2002a). However, the proteins of the viral capsid have been reported to be differentially recognized depending on the route of administration (Gahery-Segard et al, 1997). In contrast, in humans adenoviral vectors can cause variable, administration route dependent humoral immune response (Harvey et al, 1999). Heretofore, viral vectors have been often administered directly into tumor tissue (intratumorally) to achieve therapeutically relevant gene transfer rates. In fact, the majority of published clinical studies have been carried out using intratumoral administration. However, the intratumoral route is conceivable only if the tumor mass is local and accessible. Some tumors are located primarily within specific body cavities which enables intracavitary administration. This has used in clinical trials for e.g. ovarian cancer (intraperitonenal) (Alvarez et al, 2000; Buller et al, 2002) and for malignant mesothelioma (intrapleural) (Sterman et al, 1998). Nevertheless, for treatment of most types of metastatic cancer systemic administration would be useful. Intravenous and intra-arterial administration have been used in clinical studies for e.g. metastatic osteosarcoma (Benjamin et al, 2001) and hepatic metastases for colorectal cancer (Reid et al, 2002), respectively.

The second challenge is to reach target cells and deliver the therapeutic genes into them. Because viral vector uptake usually requires binding to cellular receptors, one limiting factor is the expression level of viral receptors on target cells. For example, the expression of coxsackie-adenovirus receptor (CAR), which mediates adenoviral attachment to target cells, has been shown to be variable and often very low in tumor cells (Li et al, 1999a, b; Asaoka et al, 2000; Bauerschmitz et al, 2002a). Consequently, this receptor deficiency makes cancer cells rather refractory to adenoviral mediated gene transfer (Kanerva et al, 2002a). Further, viral receptors are often widely expressed in normal cells making healthy tissue susceptible to gene transfer (Kanerva et al, 2002b). This could cause side effects and decrease the total dose delivered to target cells.

The final step after viral uptake in cells is transgene expression. In order to get an adequate therapeutic response, transgene expression has to be at certain level for a sufficient time. When using non-integrating vectors such as adenoviruses, the transgene is maintained extrachromosomally in the nucleus, resulting in transient expression (Somia and Verma, 2000). While this is sufficient for cell killing, long term expression may be advantageous when attempting to correct genetic or acquired deficiencies responsible for disease phenotypes. Even though there are various viral vectors which integrate their payload into the host cell genome (including retroviral vectors) use of these vectors does not necessarily ensure successful long term gene expression (Kay et al, 2001). Because integration to host cell genome has random features, the therapeutic gene may integrate into an inactive part of the genome resulting in silenced transgene expression (Chen and Townes, 2000). Integrated transgenes may also inactivate host genes crucial for normal cellular function or activate harmful genes such as proto-oncogenes (Shiramizu et al, 1994; Bushman and Miller, 1997; Hacein-Bey-Abina et al, 2003) In addition, the host immune system may recognize transgene -encoded proteins as foreign and induce a response, which is likely to suppress the transgene expression (Tripathy et al, 1996).

Several approaches have been devised for overcoming some of the obstacles that have been identified in completed trials. One advance has been the characterisation of various different viruses and their utilization and further modification into safe and efficient gene transfer vehicles (Kootstra and Verma, 2003). Furthermore, various targeting techniques have been employed for modification of viral vectors to recognise cancer cells and thus cause reduced transduction of healthy, non-target tissues (Peng and Russell, 1999; Nettelbeck et al, 2000). Thus, more efficient and accurate gene transfer can be achieved with targeting. In addition, the natural replication capability of various viruses including adenovirus, herpes simplex virus, alpha virus, Newcastle disease virus, measles virus and vesicular stomatitis virus has been exploited in cancer gene therapy (Alemany et al, 2000; Lundstrom, 2001; Russell, 2002; Post et al, 2003; Csatary et al, 2004). These oncolytic viruses can destroy tumor cells via replication which can be limited to target cells by genetic modification. Instead of viral vector modification, also transgenes can be modified so that the resulting therapeutic protein can spread to surrounding cells and thus help compensate for initially low gene transfer efficiency. Several so called translocatory proteins are currently known and their features have been characterized and evaluated for cancer gene therapy purposes (Leifert and Whitton, 2003).

Viruses need to transfer their genomes efficiently into host cells in order to replicate. Thus, viruses are gene transfer machines optimized by evolution. In order to use viruses as safe gene transfer vehicles, it can be advantageous to modify virulence genes and/or genes responsible for viral replication. While increasing safety of the viral vector, partial genome deletions also enable the insertion of foreign genetic material including transgenes.

Human adenoviruses are a family of viruses (over 50 serotypes) that most commonly cause rather benign respiratory or gastrointestinal illness (Volpers and Kochanek, 2004). Adenoviruses are nonenveloped, double stranded DNA viruses, whose genome is surrounded by an icosohedral protein capsid comprising of three major proteins, hexon, penton base and knobbed fiber (Russell, 2000). The linear virus genome is about 36 bp in size and consists of immediate early (E1A), early (E1-E4), intermediate and late genes (L1-L5) (Figure 1). Transcription of these genes can be divided into early and late phase, respectively, occurring before or after DNA replication (Kootstra and Verma, 2003).

In order to get inside the host cell, adenoviruses first attached to their primary cellular receptor, which is the CAR for most serotypes (Bergelson et al, 1997) This is followed by interaction with cellular integrins, which results in internalization of the virus via receptor-mediated endocytosis (Wickham et al, 1993) (Figure 2). In the endosomes, the viral genome is released from the viral capsid and thereafter transported into the nucleus. The adenoviral replication cycle is initiated by transcription of E1A gene followed by transcription of other early genes (Volpers and Kochanek, 2004). Early gene products interfere with the host antiviral defence mechanism, alter the cell cycle and modulate cellular metabolism in favour of viral replication (Russell, 2000). The linear DNA is flanked by inverted terminal repeats (ITRs), which contain sequences required for DNA replication (Hay et al, 1995), mediated by E2 and E4 gene products. Next, intermediate genes are expressed at high levels followed by the expression of late genes driven by major late promoter (Russell, 2000; Kay et al, 2001). Late genes encode for structural viral proteins that assemble together with viral genomes in the nucleus followed by cell lysis and release of newly synthesized virions (Volpers and Kochanek, 2004).

Since their first description in the 1950s, adenoviruses have been increasingly studied and they have become one of the most used gene transfer tools in human gene therapy. From a gene therapy standpoint, adenoviruses have numerous advantages: 1) reasonable characterization and understanding of their biology, 2) relatively low pathogenicity in humans, 3) capability to infect both dividing and quiescent cells, 4) capacity to accommodate relatively large transgenes, 5) low risk for insertional mutagenesis due to inability to integrate into host cell genome and 6) relatively easy manipulation and high-titer production (Danthinne and Imperiale, 2000).

Figure 1. Adenoviral genome. The genome contains early (E1-4), intermediate (IX and IVa2) and late (L1-5) genes flanked by left and right inverted terminal repeats (LITR and RITR, respectively) MLP: major late promoter, Y packaging signal.

Figure 2. Adenoviral replication cycle. Viruses first attach to the coxsackie- adenovirus receptor (CAR) followed by interaction with cellular integrins resulting in internalization of the virus via receptor-mediated endocytosis. In the endosomes, the viral genome is released from the viral capsid and thereafter transported into the nucleus for DNA replication. Structural viral proteins assemble together with viral genomes in the nucleus followed by cell lysis and release of newly synthesized virions.

The most frequently used adenoviral vectors are based on serotype 5. By deleting partially the viral genome adenoviruses can be converted into viral gene transfer vehicles. First generation adenoviral vectors are made by deleting the E3 and E1 regions, the latter responsible for initiation of viral replication. These deletions enable inserting of ~8 kbp of foreign DNA (Danthinne and Imperiale, 2000). To increase the safety and transgene capacity, additional deletions have been engineered into second generation vectors (E1-4 regions deleted) (Armentano et al, 1995; Gorziglia et al, 1996) and so called gutless vectors (all viral genes deleted) (Kochanek et al, 1996). Although deletion of various viral genes can decrease immunogenicity and toxicity and prolonge persistence of transgene expression (Engelhardt et al, 1994; Kochanek et al, 1996; Wang et al, 1997), viruses must nevertheless be packaged into virions and capsid proteins continue to elicit an immune response, which might hinder repeated administration (Somia and Verma, 2000). One possibility to circumvent the immune system would be the use of vectors based on different human adenovirus serotypes (Mack et al, 1997; Barouch et al, 2004) or animal adenoviruses (Moffatt et al, 2000; Rasmussen et al, 1999) for readministration. A disadvantageous feature of adenoviral vectors is their propensity to accumulate into liver and cause hepatotoxicity. However, this problem can be partially circumvented by targeting the viral vectors to cancer cells.

Retroviruses are lipid-enveloped, single stranded RNA viruses, which can be divided into oncoretro-, lenti- and spumaviruses (Fields et al, 1996). Enveloped viral particles contain the viral genome which consists of two copies of 8-12 kilobase-sized RNA strands surrounded by the nucleocapsid (Kootstra and Verma, 2003). The genome is flanked by long terminal repeats (LTR) and contains three essential genes: gag, which encodes viral structural protein, pol encodes reverse transcriptase and integrase and env encodes viral envelope glycoprotein, which mediates virus entry (Figure 3) In the lentiviral genome, there are additional accessory genes: for example HIV-1 has vif, vpr, vpu, tat, rev and nef genes that encode proteins necessary for efficient viral replication and persistence of infection in the natural target cells of this virus (Kootstra and Verma, 2003) .

Early in the retroviral replication cycle (Figure 4), the virus binds to its receptor, which is followed by membrane fusion and release of the RNA genome from the viral capsid (Fields et al, 1996). In the cytosol, the viral genome is copied into double-stranded DNA by the viral reverse transcriptase (Jolly, 1994). The viral DNA is then translocated to the nucleus (retroviruses with passive migration and lentiviruses with active transport), where it becomes integrated into host cell genome by its own integrase enzyme to yield a provirus. Cellular machinery is then utilized to make viral RNA, using the provirus as a template. The viral RNA also serves as mRNA, which is translated into viral proteins (Kootstra and Verma, 2003). For viral particle formation, translated viral proteins or their precursors assemble together with two viral RNA strands followed by budding from the plasma membrane. During the budding process virus attains the lipid-coated envelope with incorporated env-glycoproteins from the host cell membrane (Jolly, 1994).

Figure 3. Genome structure of oncoretrovirus and HIV-1 lentivirus. Both genomes contain gag, pol and env genes flanked by long terminal repeats, LTRs. The HIV-1 genome contains six additional genes encoding vif, vpr, vpu, tat, rev and nef. Y Packaging signal.

Figure 4. Retroviral replication cycle. Retrovirus binds to its receptor followed by membrane fusion and release of the RNA genome from the viral capsid. In the cytosol, the viral genome is copied into double-stranded DNA by viral reverse transcriptase. The viral DNA is then translocated to the nucleus where it integrates into the host cell genome. Cellular machinery is then utilized to make new viral genomes and proteins. Translated viral proteins assemble together with two viral RNA strands followed by budding from the plasma membrane. During the budding process virus attains a lipid-coated envelope (with incorporated env—glycoproteins) from the host cell membrane.

Retroviruses were the first viral vectors used in clinical studies (Blaese et al, 1995). Several features have lead to their wide use. First, their genome is rather simple, making the genetic modification required for vector production relatively straightforward. Second, they are able to integrate into the host cell genome, enabling long-term expression of transgene in target cells. However, integration does not necessarily ensure stable transgene expression, but it may increase the risk for insertional mutagenesis (Shiramizu et al, 1994; Bushman and Miller, 1997; Kay et al, 2001; Hacein-Bey-Abina et al, 2003). Third, retroviral vectors do not elicit a strong immune response, which reduces their cytotoxicity and allows readministration. On the other hand, retroviral vectors are susceptible to rapid degradation by the complement (Takeuchi et al, 1994). The main limiting factor for most retroviral vectors (including the frequently used Moloney murine leukaemia virus based vectors) is their inability to transduce non-dividing cells (Barquinero et al, 2004). In contrast, vectors based on lentiviruses are capable of transduction of both dividing and quiescent cells (Delenda, 2004) Another limiting factor for retroviral vectors is inefficient production in high titers (Romano et al, 1999) However, by replacing the region responsible for initiation of transcription (U3 —region) from the 5' LTR with a CMV promoter, higher vector titers were obtained (Finer et al, 1994) In addition, modification of viral glycoproteins has generated more stable viral particles allowing their concentration for higher titers (Burns et al, 1993). Nevertheless, retroviral titers still lag behind adenoviral titers - adenoviruses can be routinely produced in titers 2-3 orders of magnitude higher than the current best retroviral-/lentiviral titers.

Due to the relatively small size of the retroviral genome, their transgene capacity can be up to 8 kbp (McCormick, 2001). When turning retroviruses into gene transfer vectors, the viral genes are completely replaced by the desired transgene and in many cases also an internal promoter. The viral proteins required for functionality of the viral vector are produced from separate packaging constructs, which minimize the probability for generation of replication competent viruses and thus increases the safety of these vectors (Romano et al, 1999). Instead of using packaging cell lines, transient transfection can be used to deliver the required constructs into producer cells to obtain efficient virus production (Pear et al, 1993; Soneoka et al, 1995). To further increase the safety of retroviral vectors, self-inactivating type vectors (SIN) have been developed (Yu et al, 1986). These vectors contain a deletion on the 3' LTR, which inactivates the enhancer and promoter. When the vector genome is reversely transcribed, this deletion is transferred to the 5'LTR abolishing transcriptional activity of the integrated provirus. In the context of HIV-1 based lentiviral vectors, safety aspects have been considered even more carefully. Consequently, all HIV-1 accessory genes have been abolished and virus production components are divided into 3-4 separate parts (Zufferey et al, 1997) and self-inactivating deletions have been introduced into vector backbones (Zufferey et al, 1998).

Various approaches have also been developed to enhance the transduction rates of retroviral vectors. To broaden host cell tropism, retroviral particles have been pseudotyped. For instance, the env-glycoprotein has been replaced by other viral proteins such as glycoprotein from vesicular stomatitis virus (VSV-G), which has also been shown to stabilize the vector particles (Yee et al, 1994). Furthermore, oncoretroviral vectors have been retargeted by fusing polypeptides into envelope glycoproteins (Peng et al, 1999, 2001). Furthermore, transduction efficiency and transgene expression with lentivirus vectors has been enhanced by incorporating central polypurine tract (cPPT) and posttranscriptional regulatory elements (PRE) into vector constructs. The cPPT has been reported to act by increasing nuclear transport of the viral preintegration complex (Follenzi et al, 2000; Zennou et al, 2000) and has also been suggested to facilitate nuclear import of viral RNA species, thereby increasing lentiviral transduction efficiency (Van Maele et al, 2003). Also, PREs of human or woodchuck hepatitis viral origin have been shown to stabilize viral vector RNA improving transgene expression (Patzel and Sczakiel, 1997; Zufferey et al, 1999).

Although use of retroviral vectors is mainly focused on inherited genetic disease where stable, long-term transgene expression is required, also several clinical studies have been reported for cancer diseases. Due to the inability of retroviral vectors to transduce non-proliferating cells (e.g. neurons), retrovirally mediated suicide gene therapy has studied for treatment of malignant brain (Culver et al, 1994) or lung tumors (Roth et al, 1996). In preclinical studies, lentiviral vectors have been evaluated for several cancer types including ovarian (Indraccolo et al, 2002), prostate (Bastide et al, 2003; Zheng et al, 2003) and bladder cancer (Kikuchi et al, 2004). In addition, cancer gene therapy approach based on lentiviral vector targeting to tumor endothelium has been introduced (De Palma et al, 2003). However, safety concerns have prevented clinical cancer trials with lentiviruses thus far.

Adverse side effects caused by unspesific gene transfer to non-target organs can be avoided by targeting viral vectors and/or transgenes to cancer cells (Glasgow et al, 2004). Additionally, such manoeuvres allow enhancement of gene transfer rates in tumor tissue resulting in enhanced therapeutic outcome. Especially with regard to adenoviral gene transfer, a big obstacle is the variable and often low expression level of CAR in tumor cells, making these cells rather refractory to adenoviral gene transfer. Further, it would be also important to minimize the adverse side effects by detargeting the vectors from the liver. Targeting strategies can be based on transductional or transcriptional approaches.

Transductional targeting is based on altered viral tropism via modification of viral proteins mediating receptor binding. In adenoviral vectors, re-targeting moieties allowing CAR- independent delivery can be linked physically to fiber knob or introduced genetically by incorporating necessary changes to the viral genome (Figure 5). Simultaneously, the binding to the primary viral receptor can be blocked. Alternative receptors are typically expressed at high levels in cancer cells but to a lesser extent in normal cells, which improves the tumor cell specificity of the gene transfer. Heretofore, several reports indicate successful adenoviral targeting in vitro and in animal models to a considerable number of alternative cellular receptors including a _v integrins, CD3, CD40, adenovirus serotype 3 receptor, prostate specific membrane antigen and epidermal growth factor receptor (Wickham et al, 1996, 1997; Tillman et al, 1999; Hemminki et al, 2001b; Kanerva et al, 2002a; Kraaij et al, 2005). Further, some studies have shown that adenoviral vector mediated liver toxicity can be reduced by targeting the vectors to cancer cells (Einfeld et al, 2001; Printz et al, 2000). However, liver uptake is mostly a non-receptor mediated process resulting in virus clearance by Kuppfer cells (Schiedner et al, 2003).

In addition to re-routing viral vectors to alternative receptors, expression of therapeutic genes can be limited into tumor tissue. The expression of transcriptionally

Figure 5. Transductional targeting of adenoviruses. Via transductional targeting, adenoviral vector can re-routed to alternative cellular receptors instead of its primary coxsackie-adenovirus receptor (CAR) Targeting can be based on genetically modified knobs or bi-specific ligands. The latter bind both to the viral knob and an alternative receptor.

One possibility to circumvent initially low gene transfer efficiency is to facilitate spreading of the therapeutic element inside the tumor tissue. One approach is based on oncolytic adenoviruses, which replicate in tumor cells killing the host cell and spreading to the neighboring cells and eventually throughout the tumor. Another possibility is to exploit protein transduction domains (PTD), which can spread the fused therapeutic protein from one cell to another.

Utilization of conditionally replicative adenoviruses (CRAds) is based on their theoretical ability to spread throughout the tumor as long as tumor cells persist, by virtue of viral replication and concomitant cell lysis and viral progeny dissemination (Alemany et al, 2000; Post et al, 2003).

To minimize adverse side effects and increase the safety of these anti-cancer agents, replication can be limited to tumor tissue by genetically modifying the CRAd genome. Replication can be limited either with partial deletions in the E1 region or by using tissue specific promoters to drive genes responsible for viral replication (Alemany et al, 2000). Partial viral genome deletions allow virus to replicate selectively in cells with defective p53/p14ARF (Bischoff et al, 1996) or Rb-p16 pathway (Fueyo et al, 2000) that are hallmarks of many cancer cells. Replication can also be limited by using tissue specific promoters such as human telomerase reverse transcriptase (hTERT) (Irving et al, 2004), a -fetoprotein (Hallenbeck et al, 1999) or tyrosinase (Nettelbeck et al, 2002) to drive E1A expression.

To increase the specificity and antitumoral activity of oncolytic agents, several targeting approaches have been developed. CRAds have been successfully targeted to alternative cellular receptors, e.g. the adenovirus serotype 3 receptor and integrins (Suzuki et al, 2001; Bauerschmitz et al, 2002b; Kanerva et al, 2003, 2005). In addition, the antitumoral activity of CRAds is affected by the presence or absence of native E3 region viral genes, frequently deleted in earlier generation vectors. It has been reported that retaining the adenovirus E3 region (Suzuki et al, 2002) and overexpression of the adenovirus death protein (ADP) (Yun et al, 2004) can increase the oncolytic activity of CRAds. In contrast, deletion of the apoptose inhibitor E1B-19 kDa protein might enhance oncolysis (Liu et al, 2004).

The most widely studied and first clinically tested CRAd is ONYX-015 (dl1520), in which the E1B-55 kDa gene is mutated (Bischoff et al, 1996). This protein is responsible for p53 binding and inactivation for effective replication and thus it was hypothesized that ONYX-015 could only replicate in cells with a deficient p53-p14ARF pathway, thus including most cancer cells. In initial preclinical studies, ONYX-015 was reported to replicate selectively in p53-pathway-deficient tumor cells (Bischoff et al, 1996) and in addition, it was suggested not to replicate in normal epithelial and endothelial cells (Heise et al, 1997). However, However, later studies questioned the correlation between viral replication and p53-status in tumor cells and suggested that the mechanism of selectivity was more complex than initially suggested (Goodrum and Ornelles, 1998; Rothmann et al, 1998).

ONYX-015 has been tested extensively in humans; over 10 clinical trials (phase I-II) have enrolled more than 300 patients with head and neck cancer (Ganly et al, 2000; Nemunaitis et al, 2000, 2001b), metastatic colorectal cancer (Reid et al, 2002), pancreatic cancer (Mulvihill et al, 2001) and ovarian cancer (Vasey et al, 2002). ONYX-015 has also studied in patients with lung metastasis (Nemunaitis et al, 2001a). The results from these studies show that although ONYX-015 has been a very safe, well tolerated vector, administerable by various routes (intratumoral, intraperitoneal, intra-arterial and intravenous), the majority of treated tumors did not respond to the therapy and significant antitumoral effects were rare (reviewed in Hemminki and Alvarez, 2002; Kirn, 2001). Of note, most of these patients had advanced disease already refractory to all available treatments. Further, most patients were enrolled in Phase I trials, which by definition do not attempt to assess efficacy as a primary endpoint.

A PSA -selective CRAd (CV706) has been studied in humans (DeWeese et al, 2001). The results were in parallel with results obtained from ONYX-trials: the virus was well tolerated and evidence of its replication was obtained and although PSA responses were frequently seen, the achieved antitumoral effect was relatively modest.

Taken together, these results suggest that early generation CRAds are safe anticancer agents, but their ability to completely destroy tumors as a single agent is limited. However, improved antitumoral activity has been obtained by combining CRAds with e.g. conventional therapy and suicide gene therapy. Unfortunately, no Phase III trials have yet been completed and therefore conclusive evidence with regard to efficacy is still lacking. Nevertheless, recent news with regard to ONYX-015 provides some optimism that Phase III trials may soon be initiated (Pollack, 2005).

There are several so called translocatory proteins which are reported to be secreted from cells via a non-classical Golgi-independent route and to move intercellularly in a receptor- and transport —independent manner (Schwarze and Dowdy, 2000) The best known translocatory proteins are herpes simplex virus type 1 (HSV-1) tegument protein VP22 (Elliott and O'Hare, 1997), HIV-1 tat (Frankel and Pabo, 1988; Green and Loewenstein, 1988) and Drosophila antennapediae (Joliot et al, 1991). These intercellularly trafficking proteins share several features: all of them appear to localize in the nucleus and each of them has a highly basic region (Leifert and Whitton, 2003). Although the exact mechanism of intercellular spreading is still unknown, domains responsible for protein transduction have been identified. These small PTDs contain several basic amino acid residues, which have been suggested to mediate cellular binding and penetration (Gratton et al, 2003; Lundberg et al, 2003).

The most studied PTD is VP22, a 38 kDa sized major protein of HSV-1 tegument encoded by the UL49 gene (Elliott and Meredith, 1992). Several reports suggest that VP22 is able to retain its trafficking capacity even when fused to other proteins. VP22 has been reported to spread intercellularly in vitro when fused to GFP (Elliott and O'Hare, 1997, 1999; Aints et al, 1999; Wybranietz et al, 1999; Brewis et al, 2000) and enhance the antitumoral effect in vivo when fused to HSV-1 thymidine kinase and tumor suppressors p53 and p27 (Dilber et al, 1999; Wills et al, 2001; Zavaglia et al, 2003). On the other hand, however, there are also reports suggesting that VP22 mediated intercellular trafficking might be an artifact, related to the fixation process (Fang et al, 1998; Lundberg et al, 2003), thus making the exploitation of protein transduction approach for gene therapy purposes rather debatable. According to recent findings by Lundberg and co-workers, positively charged PTDs e.g. VP22 only mediate cell surface adherence via electrostatic interactions and observed translocation across the cell membrane is due to the fixation procedure (Lundberg et al, 2003). However, PTDs have been successfully employed to enhance both lenti- and adenoviral transduction in vitro and in vivo (Gratton et al, 2003; Kretz et al, 2003). In addition, PTDs have been demonstrated to improve the viral uptake and replication of tumor-specific oncolytic adenoviruses in vitro (Kuhnel et al, 2004).

Although several gene therapy approaches have been demonstrated to destroy tumor cells, clinical trials suggest that the ability of early generation agents to completely eradicate advanced tumor masses is limited. A main reason for the not unexpected discrepancy between humans and preclinical models is the complex nature of advanced solid tumors; a large tumor mass will have areas of necrosis, hypoxia, variable stromal contents and other infiltrating normal tissue, variable vasculature, etc. There is also preclinical evidence that intratumoral complexity can compromise the efficacy of gene therapeutic agents (Pipiya et al, 2005; Sauthoff et al, 2003). Thus, for improved antitumoral effect, the following strategies can be identified (with some references as examples):

1. Transductional targeting of agents to receptors expressed preferentially on tumor cells in comparison to normal cells (Hemminki et al, 2001a, 2002b)
2. Transcriptional targeting for tumor specific activity (Barker et al, 2003a)
3. Combination of transductional and transcriptional targeting (Barker et al, 2003b)
4. Develop viruses that can be tracked in vivo to obtain maximum correlative data (Kanerva et al, 2005)
5. Combine gene therapeutics with more conventional approaches (Raki et al, 2005)
6. Arm vectors and CRAds with transgenes that can help overcome intratumoral barriers (Conrad et al, 2005)
7. Improve the intratumoral dissemination of agents (Ahmed et al, 2003; Liu et al, 2004)
8. Develop agents that allow multiple administration without inactivation by the immune system (Mastrangeli et al, 1996; Mok et al, 2005).

Conventional therapies have been efficiently exploited in cancer gene therapy and for metastatic disease, chemotherapy, radiation therapy and hormonal therapies remain front fine options. However, in most cases, metastatic solid tumors remain incurable and thus new therapeutic options are needed (DeVita et al, 2001). In this regard, there are some exciting recent results which demonstrate the utility of gene therapy in advanced cancer patients. In a Phase II randomized trial, patients with malignant glioma were resected and half received adenovirus coding for HSV-TK into the resection cavity margins, followed by i.v. ganciclovir. The survival of the patients was doubled in comparison to randomized controls (Immonen et al, 2004). In another clinical trial, patients with advanced high-grade glioma were treated with oncolytic Newcastle disease virus. This treatment resulted in survival rates of 5-9 years whereas typical prognosis for this aggressive disease ranges averagely from six months to a year. In addition, these patients were regularly treated with virotherapy for several years without interruption (Csatary et al, 2004).

Introduction of suicide, tumor suppressor or immunotherapeutic genes to the patients receiving chemotherapy has been studied in ovarian and lung cancer and multiple myeloma (Hasenburg et al, 2001; Schuler et al, 2001; Trudel et al, 2001). However, in these studies, only a minor additional benefit was evident with combination therapy and the lack of randomization obscures the magnitude of the benefit. Furthermore, transfer of MDR-1 gene into hematopoietic stem cells to reduce the toxic effects of cancer chemotherapy has also been evaluated (Hesdorffer et al, 1998). Combining radiotherapy with therapeutic genes encoding e.g. thymidine kinase and p53 has been studied in prostate and lung cancer (Swisher et al, 2003; Teh et al, 2004). Furthermore, utilizing chemotherapy and/or radiotherapy in conjunction with CRAds has been suggested to augment antitumor activity in pre-clinical studies with colon cancer (Rogulski et al, 2000) and malignant glioma (Lamfers et al, 2002) and in clinical studies with head and neck cancer (Khuri et al, 2000; Young, 2005) and metastatic gastrointestinal carcinoma (Reid et al, 2002).

In pre-clinical studies, immunotherapy combined with suicide genes has been demonstrated to induce antitumoral immunity and enhance the tumor regression in lung, colon and metastatic breast cancer (Jones et al, 2000; Majumdar et al, 2000; Park et al, 2003).

Another approach to enhance the antineoplastic activity of oncolytic adenoviruses is to utilize their ability to multiply and amplify inserted therapeutic genes during replication. Several pre-clinical and clinical studies suggest that combining therapeutic genes, such as p53, cytosine deaminase and thymidine kinase with replicative viruses might improve their oncolytic potency and increase the anti-tumoral effect (Wildner et al, 1999; Freytag et al, 2002, 2003; van Beusechem et al, 2002; Fuerer and Iggo, 2004).

In summary, several clinical trials in cancer patients have shown that gene therapy is relatively safe and the heretofore utilized gene transfer vectors are well tolerated. Results from clinical trials have also been promising in demonstrating preliminary efficacy, especially when combined with conventional treatments. Also, the first randomized and therefore unequivocal evidence of gene therapy giving clinical benefit was recently published (Immonen et al, 2004). In addition, several strategies to improve the therapeutic outcome are constantly being developed and evaluated in pre-clinical studies in order to create more powerful tools for gene therapy purposes.

This work was supported by the Finnish Cultural Foundation, Paulo foundation, HUCH Research Funds (EVO), Academy of Finland, Emil Aaltonen Foundation, Finnish Cancer Society, University of Helsinki, Sigrid Juselius Foundation, Sohlberg Foundation, Biocentrum Helsinki, Instrumentarium Research Fund.