I. Introduction

Monumental progress in several fields including DNA replication, transcription factors and gene expression, repair, recombination, signal transduction, oncogenes and tumor suppressor genes, genome mapping and sequencing, and on the molecular basis of human disease are providing the foundation of a new era of biomedical research aimed at introducing therapeutically important genes into somatic cells of patients. The main targets of gene therapy are to repair or replace mutated genes, regulate gene expression and signal transduction, manipulate the immune system, or target malignant and other cells for destruction (reviewed by Anderson, 1992; Nowak, 1995; Boulikas, 1996a,b; Culver, 1996; Ross et al, 1996).

Two main approaches have been pursued for gene transfer to somatic cells (i) direct gene delivery using murine retroviruses, adenoviruses, adeno-associated virus, HSV, EBV, liposomes, polymers, or direct plasmid injection (gene therapy in vivo); and (ii) ex vivo gene therapy involving removal of syngeneic cells from a specific organ or tumor of an individual, genetic correction of the defect in cell culture (ADA deficiency, LDL-R for FH) or transfer of a different gene (IL-2 to tumor infiltrating lymphocytes to potentiate the cytotoxicity to tumors, cytokine genes to tumor cells from a patient for cancer immunotherapy, multidrug resistance gene transfer to render bone marrow cells resistant to certain antineoplastic drugs), followed by reimplantation of the cells. The reimplanted cells produce the therapeutic protein.

Several key factors or steps appear to be involved for the effective gene transfer to somatic cells in a patient or animal model: (i) the type of vehicle used for gene delivery (liposomes, adenoviruses, retroviruses, AAV, HSV, EBV, polymer, naked plasmid) which will determine not only the half-life in circulation, the biodistribution in tissues, and efficacy of delivery but also the route through the cell membrane and fate of the transgene in the nucleus; (ii) interaction of the gene-vehicle system with components in the serum or body fluids (plasma proteins, macrophages, immune response cells); (iii) targeting to the cell type, organ, or tumor, and binding to the cell surface; (iv) port and mode of entrance to the cell (poration through the cell membrane, receptor-mediated endocytosis), (v) release from cytoplasmic compartments (endosomes, lysosomes), (vi) transport across the nuclear envelope (nuclear import); (vii) type and potency of regulatory elements for driving the expression of the transferred gene in a particular cell type including DNA sequences that determine integration versus maintenance of a plasmid or recombinant virus/retrovirus as an extrachromosomal element; (viii) expression (transcription) of the transgene producing heterogeneous nuclear RNA (HnRNA) which is then (ix) spliced and processed in the nucleus to mature mRNA and is (x) exported to the cytoplasm to be (xi) translated into protein. Additional steps may include posttranslational modification of the protein and addition of a signal peptide (at the gene level) for secretion.

All steps can be experimentally manipulated and improvements in each one can enormously enhance the level of expression and therapeutic index of a gene therapy approach. It has been proposed that the plasmid vector is unable to translocate to the nucleus unless complexed in the cytoplasm with nuclear proteins possessing nuclear localization signals (NLSs). NLSs are short karyophilic peptides on proteins destined to function in the nucleus used for binding to specific transporter molecules in the cytoplasm, mediating their passage through the pore complexes to the nucleus (see Boulikas, 1998, this volume). NLS are present on histones, transcription factors, nuclear enzymes, and a number of other nuclear proteins; nascent chains of DNA-binding polypeptides could bind to the supercoiled plasmid in the cytoplasm mediating its translocation to the nucleus.

During delivery of foreign DNA in vivo vehicles may be attacked by macrophages, lymphocytes, or other components of the immune system and the vast majority will be cleared from blood, intracellular, or other body fluids before it is given the chance to reach the membrane of the cell target; the half-life of naked plasmids injected intravenously into animals is about 5 min (Lew et al, 1995). Cationic lipids, other than being very toxic, mediate efficient gene delivery passing through biological membranes; those lipid-DNA complexes surviving the immediate neutralization by serum proteins in the blood can reach the lung, heart and other tissues after vein or artery injection with one heart beat and transform endothelial vascular cells (reviewed by Boulikas, 1996d).

A variety of viral vectors have been developed to exploit the characteristic properties of each group to maintain persistence and viral gene expression in infected cells. Retroviral vectors and AAV integrate into target chromosomes and the transgene they carry can be inactivated from position effects from chromatin surroundings. Vectors with persistence/integration functions may not result in high levels of gene delivery in vivo.

Adenoviruses and retroviruses which are of the most frequently used vehicles for gene transfer can accommodate up to 7kb of total foreign DNA into their genome because of packaging limitations. This precludes their use for the transfer of large genomic regions. Transfer of intact yeast artificial chromosome (YAC) into transgenic mice will enable the analysis of large genes or multigenic loci such as human b-globin locus (reviewed by Peterson et al, 1997).

A small portion of plasmid molecules crossing the cell membrane will escape degradation from nucleases in the lysosomes and become released to the cytoplasm; even a smaller portion of these molecules will enter nuclei; finally, after successfully reaching the nucleus, plasmids with therapeutic genes are usually degraded by nuclear enzymes and transgene expression is permanently lost after about 2-7 days from animal tissues following successful gene delivery. During the peak of transgene expression (usually 7-48 h from injection) the transgene transcript can follow the normal fate of other nuclear transcripts when proper polyadenylation signals are provided; its processed mRNA will be exported to the cytoplasm and translated into the therapeutic protein.

The choice of the appropriate delivery system for successful somatic gene transfer demands understanding of the drawbacks and advantages of each delivery system, such as limitations in the total length of the DNA that can be introduced, including the cDNA of the therapeutically important gene and control elements. Understanding the pathophysiology of the disease and the cell targets can give clues on the way of introducing the gene (i.v., i.p., intratumoral, s.c. injection) or direct the gene therapist to designing methods to target and secrete a therapeutic protein from a tissue which is not the normal site of production of a therapeutic protein. The type of control elements required for the anticipated tissue-specific expression of the construct, the presence of viral or other origins of replication as well as of the cDNA encoding the viral replication initiator protein for an episomal replication of the transgene, sequences that prompt integration and others that insulate the gene from the chromatin surroundings at the integration site, are also important for successful gene transfer.

Cancer gene therapy and immunotherapy has been the first priority of human gene therapy protocols. New gene targets are being defined and new clinical protocols are being proposed and approved. Effective eradication of a great variety of tumors with drugs which inhibit angiogenesis has been extraordinarily successful on animal models and the method moves fast to clinical trials; transfer of anti-angiogenesis genes will be the next step. A number of anticancer genes are being tested in preclinical or clinical cancer trials including p53, RB, BRCA1, E1A, bcl-2, MDR-1, HER2, p21, p16, bax, bcl-xs, E2F, antisense IGF-I, antisense c-fos, antisense c-myc, antisense K-ras and the cytokine genes GM-CSF, IL-12, IL-2, IL-4, IL-7, IFN-g, and TNF-a. A promising approach is transfer of the herpes simplex virus thymidine kinase (HSV-tk) gene (suicide gene) and systemic treatment with the prodrug ganciclovir which is converted by HSV-tk into a toxic drug killing dividing cells. Theoretically, expression of therapeutic genes preferentially in cancer cells could be achieved by regulatory elements from tumor-specific genes such as carcinoembryonic antigen.

The first gene therapy products are expected to receive FDA approval by the year 2000; the market for gene therapy products is expected to exceed $45 billion by 2010.

This article reviews the molecular mechanisms and recent developments for the gene therapy of cancer, HIV, ADA deficiency, Parkinson's disease, lysosomal storage disease, hemophilia A and B, a1-antitrypsin deficiency, cystic fibrosis, rheumatoid arthritis, hypertension, familial hypercholesterolemia, atherosclerosis/restenosis, wound healing, and obesity including the treatment of cancer and heart diseases with angiogenesis inhibitors and gene transfer to the arterial wall. It is my intention to give a general overview rather to exhaust the field.

DIVISION ONE: GENE DELIVERY SYSTEMS AND GENE EXPRESSION

II. Gene delivery using retroviruses

A. Recombinant murine retroviruses

The recombinant Moloney murine leukemia virus (Mo-MLV or MLV) has been extensively used for gene transfer. Retroviral vectors derived from Mo-MLV promote the efficient transfer of genes into a variety of cell types from many animal species; up to 8 kb of foreign DNA can be packaged in a retroviral vector. Recombinant retroviruses have been the most frequently used and promising vehicles for the delivery of therapeutic genes in human gene therapy protocols (Appendix 1). Retroviral vectors cause no detectable harm as they enter their target cells; the retroviral nucleic acid becomes integrated into chromosomal DNA, ensuring its long-term persistence and stable transmission to all future progeny of the transduced cell.

The life cycle of the retrovirus is well understood and can be effectively manipulated to generate vectors that can be efficiently and safely packaged. An important contribution to their utility has been the development of retrovirus packaging cells, which allow the production of retroviral vectors in the absence of replication-competent virus.

Recombinant retroviruses stably integrate into the DNA of actively dividing cells, requiring host DNA synthesis for this process (Miller et al, 1990). Although this is a disadvantage for targeting cells at G0, such as the totipotent bone marrow stem cells, it is a great advantage for targeting tumor cells in an organ without affecting the normal cells in the surroundings. This approach has been used to kill gliomas in rat brain tumors by injection of murine fibroblasts stably transduced with a retroviral vector expressing the HSV-tk gene (Culver et al, 1992; see below).

B. Retrovirus packaging cell lines

The use of retroviral vectors in human gene therapy requires a packaging cell line which is incapable of producing replication-competent virus and which produces high titers of replication-deficient vector virus. The packaging cell lines have been stably transduced with viral genes and produce constantly viral proteins needed by viruses to package their genome. Wild-type virus can be produced through recombinational events between the helper virus and a retroviral vector. Methods are also available for generating cell lines which secrete a broad host range retrovirus vectors in the absence of helper virus.

Retrovirus packaging cell lines containing the gag-pol genes from spleen necrosis virus and the env gene from spleen necrosis virus or from amphotropic murine leukemia virus on a separate vector have been used; retrovirus vectors were produced from these helper cell lines without any genetic interactions between the vectors and sequences in the helper cells (Dougherty et al, 1989). An ecotropic packaging cell line and an amphotropic packaging cell line, in which the viral gag and pol genes were on one plasmid and the viral env gene were on another plasmid have been constructed; both plasmids contained deletions of the packaging sequence and the 3' LTR; when the fragmented helper virus genomes were introduced into 3T3 cells they produced titers of retrovirus which were comparable to the titers produced from packaging cells containing the helper virus genome on a single plasmid (Markowitz et al, 1990).

The pBabe retroviral vector constructs which transmit inserted genes at high titers and express them from the Mo-MLV LTR have been designed with one of four different dominantly acting selectable markers, allowing the growth of infected mammalian cells in the presence of G418, hygromycin B, bleomycin/phleomycin or puromycin, respectively. The packaging cell line, omega E, generated with separate gag/pol and ecotropic env expression constructs, was designed in conjunction with the pBabe vectors to reduce the risk of generation of wild type Mo-MLV via homologous recombination events (Morgenstern and Land, 1990).

C. Pseudotyped retroviral vectors

The traditional retroviral vector enters the target cell by binding of a viral envelope glycoprotein to a cell membrane viral receptor. Coinfection of cells with a retrovirus and VSV (vesicular stomatitis virus) produces progeny virions containing the genome of one virus encapsidated by the envelope protein of the other (pseudotypes of viruses); this led to the development of pseudotyped retroviral vectors where the Moloney murine leukemia env gene product is replaced by the VSV-G protein able to interact with other membrane-bound receptors as well as with some components of the lipid bilayer (phosphatidylserine); because of the ubiquitous distribution of these membrane components pseudotyped particles display a very broad host range (Friedmann and Yee, 1995). Use of pseudotyped vectors has been a significant advancement for retroviral gene transfer.

Pseudotypes of VSV and Mo-MLV, are released preferentially at early times after infection of MuLV-producing cells with VSV; at later times, after synthesis of M-MLV proteins has been inhibited by the VSV infection, neither Mo-MLV virions nor the VSV (Mo-MLV) pseudotypes are made. There appears to be a stringent requirement for recognition of the viral core by homologous envelope components for the production of VSV (M-MLV) pseudotypes (Witte and Baltimore, 1977).

The finding that the G protein of vesicular stomatitis virus (VSV) can serve as the exclusive envelope protein component for one specific retroviral vector that expresses VSV G protein was extended to a general transient transfection scheme for producing very high-titer VSV G-enveloped pseudotypes from any Moloney murine leukemia-based retroviral vector (Yee et al, 1994). Pseudotyping of MuLV particles with VSV-G expressed transiently in cells producing MLV Gag and Pol proteins, has yielded vector preparations with a broader host range that could be concentrated by ultracentrifugation. For example, this technology allowed for efficient concentration of vector by ultracentrifugation to titers > 10⁹ colony-forming units/ml and offers hope for potential use for gene transfer in vivo. Furthermore, these vectors could infect cells, such as hamster and fish cell lines, that are ordinarily resistant to infection with vectors containing the retroviral envelope protein (Burns et al, 1993).

A human 293-derived retroviral packaging cell line was generated by Ory et al (1996) capable of producing high titers of recombinant Mo-MLV particles that have incorporated the VSV-G protein. This new packaging cell line may be used for direct in vivo gene transfer using retroviral vectors because the retroviral/VSV-G pseudotypes generated with these cells were significantly more resistant to human complement than commonly used amphotropic vectors.

A human immunodeficiency virus type 1 (HIV-1)-based retroviral vector containing the firefly luciferase reporter gene could be pseudotyped with a broad-host-range VSV envelope glycoprotein G; higher-efficiency gene transfer into CD34⁺ cells was achieved with a VSV-G-pseudotyped HIV-1 vector than with a vector packaged in an amphotropic envelope (Akkina et al, 1996).

Because the VSV-G protein is toxic to cells when constitutively expressed, Yang et al (1995) have used steroid-inducible and tetracycline-modulated promoter systems to derive stable producer cell lines capable of substantial production of VSV-G pseudotyped MLV particles. Similarly, the toxic G protein of VSV could be induced in a cell line by the removal of tetracycline and the addition of estrogen; this cell line was transduced with a modified tTA transactivator gene engineered with the ligand-binding domain of the estrogen receptor to the carboxy terminus of the tTA transactivator; a single retroviral vector could transduce both the transactivator gene and the VSV-G protein gene controlled by the tTA-inducible promoter into mammalian cells (Iida et al, 1996). The tetracycline-inducible system was modified by fusing the ligand binding domain of the estrogen receptor to the carboxy terminus of a tetracycline-regulated transactivator to regulate VSV-G expression in a tetracycline-dependent manner that could be modulated by b-estradiol in stable packaging cell lines (Chen et al, 1996).

D. Limitations and advancements using retroviral vectors

Before the in vivo gene therapy with retroviruses becomes a successful reality a number of problems must be overcome. Despite the extensive use of retroviral vectors in gene therapy, there are still problems to be solved and there is an ultimate need for the development of new, improved retroviral vectors and packaging systems to fuel further advances in the field of human gene therapy. The principle limitation of retroviruses has been poor gene expression in vivo which has been overcome through the use of tissue-specific promoters. Use of internal ribosome entry sites from picornaviruses in retroviral vectors has provided stable expression of multiple gene enhancers (reviewed by Naviaux and Verma, 1992; Boris-Lawrie and Temin, 1993).

Little is known about the factors that influence the efficiency of retroviral infection in vivo. Many commonly used experimental animal strains, such as mice, harbor endogenous C-type proviruses, some of which are expressed and have circulating antibodies against the viral envelope glycoproteins that cross-react with the Mo-MLV; the efficiency of retrovirus-mediated transfection in vivo using a variety of mouse strains was affected by humoral immune competence and interference between endogenous MLVs and exogenous recombinant Mo-MLV (Fassati et al, 1995).

One of the drawbacks of retroviruses for their exploitation in gene therapy has been the low viral titers obtained, too low to achieve therapeutic levels of gene expression; methods for the efficient concentration from large volumes of supernatant and purification of amphotropic retrovirus particles have been developed in several laboratories. For example, Bowles et al (1996) have used concentration and further purification of virus particles by sucrose banding ultracentrifugation; animal studies have shown that viral transduction increased proportionally with titer of the retrovirus.

Transduced cells producing retrovirus are tissue-incompatible and are, therefore, expected to be attacked by the immune system; this will lead to the elimination of therapeutic cells from the body, a phenomenon markedly associated also with adenoviral gene transfer. A privileged exception are brain tumor cells expressing recombinant retrovirus which persist without immunologic rejection (Culver et al, 1992).

Sodium butyrate treatment of murine retrovirus packaging cells producing a CFTR vector increased the production of the retrovirus vector between 40- and 1,000-fold (Olsen and Sechelski, 1995).

The Cre/LoxP recombinase strategy (see below) has been used to generate retroviral vectors that have the ability to excise themselves after inserting a gene into the genome, thereby avoiding problems encountered with conventional retrovirus vectors, such as recombination with helper viruses or transcriptional repression of transduced genes (Russ et al, 1996). Retroviral vectors with the Cre/LoxP technology have also been used to deliver the GM-CSF gene to K562 cell culture (Fernex et al, 1997), for the development of retroviral suicide vectors for gene therapy using the HSV-tk gene (Bergemann et al, 1995), and for the production of a high-titer producer cell line containing a single LoxP site flanked by the viral LTRs (Vanin et al, 1997).

Because retrovirus vectors are integrated into the genome, transcriptional repression of transduced genes will often take place from position effects exerted from neighboring chromatin domains; two matrix-attached regions (MARs), one at either flank of the transgene, are proposed here to insulating the gene in the retrovirus vector from chromatin effects at the integration site by creating an independent realm of chromatin structure harboring the transgene. MAR insulators have been used and can enhance up to 2,000-fold the expression of genes in transgenic animals and plants (McKnight et al, 1992; Breyne et al, 1992; Allen et al, 1993; Brooks et al, 1994; Thompson et al, 1994; Forrester et al, 1994).

E. Targeting of retrovirus to specific cell types

A number of approaches have been directed to develop retroviral vectors that are able to target particular cell types; also efforts focus toward retroviral vectors that incorporate nonretroviral features and are tailored to desired needs for specific uses (reviewed by Vile and Russell, 1995; Gunzburg and Salmons, 1996).

Ideally, therapeutic genes should be delivered only to the relevant cell type and/or expressed in this cell type. Viral and nonviral vectors can be targeted through ligand-receptor interactions. Retroviral targeting through protease-substrate interactions has also been described; epidermal growth factor (EGF) was fused to a retroviral envelope glycoprotein via a cleavable linker comprising a factor Xa protease recognition signal. Vector particles displaying the cleavable EGF domain could bind to EGF receptors on human cells but did not transfer their genes until they were cleaved by factor Xa protease (Nilson et al, 1996).

A retroviral vector that infects human cells specifically through recognition of the low density lipoprotein receptor has been described by adding onto the ecotropic envelope protein of M-MLV a single-chain variable fragment derived from a monoclonal antibody recognizing the human LDL-R; the chimeric envelope protein was used to construct a packaging cell line producing a retroviral vector capable of transfer of the lacZ gene to human cells expressing LDL-R (Somia et al, 1995).

F. Other retroviruses

Viruses that contain RNA as their genetic material may be either negative- or positive-strand RNA viruses. The very large group of negative-strand RNA viruses includes some of the most serious and notorious pathogens subdivided into those with segmented RNA (influenza viruses, comprising eight separate segments of RNA and bunyaviruses containing three segments of single-stranded RNA, the large, L, the medium, M, and the small, S) and those with nonsegmented RNA (VSV, rabies, measles, Sendai, respiratory syncytial virus, Ebola viruses). Positive-strand RNA viruses include poliovirus.

Cloned positive-strand poliovirus cDNA is infectious but neither isolated genome nor antigenome RNA of negative-strand viruses is infectious; this is because the negative-strand viral RNA is assembled with viral nucleoprotein into an RNP complex that becomes the template for the viral RNA-dependent RNA polymerase. Helper influenza virus-dependent procedures have been developed in which an influenza virus-like RNA molecule, containing a reporter gene, was mixed with disrupted virion core proteins to reconstitute RNP complexes in vitro which were then transfected into influenza virus-transfected cells. Recombinant nucleocapsid and polymerase proteins for the unsegmented RNA viruses have also been used to produce infectious virus without help from an homologous virus using full-length cDNA clones of intracellularly transcribed antigenomes (rabies, VSV, measles, Sendai) (see Bridgen and Elliott, 1996 and the references cited therein).

Plasmids containing full-length cDNA copies of the three RNA genome molecules of Bunyamwera bunyavirus and a negative-sense copy of the GFP gene, flanked by T7 promoter and hepatitis delta virus ribozyme sequences, were used to produce infectious virus particles without helper virus; these plasmids were used to transfect HeLa cells which expressed T7 RNA polymerase and recombinant Bunyamwera bunyavirus proteins by previous transfection with the appropriate plasmids; 24 h after infection about 1 in 1,000 HeLa cells displayed fluorescence indicative of transcription and replication of the reporter RNA (Bridgen and Elliott, 1996).

III. Adenoviral gene delivery

A. Adenovirus replication, transcription, and attachment to the nuclear matrix

Before understanding the principle of adenoviral gene transfer, it is essential to comprehend the molecular events which are involved in the life cycle of the adenovirus. Adenoviruses posses a well-defined origin of replication which is stimulated by transcription factors NFI and NFIII (Hay, 1985; Pruijn et al, 1986). The transcription factor NF-I (also called CTF, CCAAT box-binding protein, or C/EBP) stimulates replication of adenovirus DNA in vitro (Pruijn et al, 1986; Jones et al, 1987; Santoro et al, 1988; Coenjaerts et al, 1991) by establishing cooperative interactions with Ad-DBP (Adenovirus DNA-binding protein) (Cleat and Hay, 1989). The transcription factor NFIII (also called Oct-1 or OTF-1), involved in the regulation of the histone H2B and immunoglobulin genes, can stimulate initiation of adenovirus DNA replication in vitro (O'Neil et al., 1988; Mul et al, 1990; Verrijzer et al, 1990; Coenjaerts et al, 1991).

The adenovirus 5 protein Ad-DBP is a single-stranded DNA binding protein product of the viral E2A absolutely required for chain elongation during Ad5 DNA replication; other than facilitating unwinding of the DNA, Ad-DBP might also protect single-stranded DNA at the replication fork from nuclease attack, increase the rate of processivity of the viral DNA polymerase, and increase binding of NFI of the core origin of Ad5 (Cleat and Hay 1989). This protein has a size of 529 amino acids, is phosphorylated and apart from its role in DNA replication is also involved in transcription, recombination, transformation, and virus assembly (see Tucker et al 1994). Crystal structure at 2.6 A resolution of Ad-DBP shows that a 17 aa C-terminal domain hooks onto a second Ad-DBP molecule thus promoting its cooperativity during DNA binding; Ad DBP was proposed to act by forming a core around which single-stranded DNA winds (Tucker et al, 1994).

Adenoviruses replicate episomally; they need to attach to the nuclear matrix of the host cell for their replication. Two adenoviral proteins have been found attached to the nuclear matrix and presumably mediating the anchorage of the adenovirus: (i) the E1a protein (11 kDa), a transcription and replication factor sufficient to immortalize primary rodent cells, which was crosslinked to matrix proteins with oxidation with o-phenanthroline/Cu²⁺ (Chatterjee and Flint, 1986) and (ii) the adenovirus terminal protein (55 kDa) which is covalently attached to the 5' end of Ad DNA and initiates DNA replication; the adenovirus terminal protein mediated adenovirus anchorage to nuclear matrix was resistant to 1M guanidine extraction (Bodnar et al, 1989; Schaack et al, 1990; Fredman and Engler, 1993).

Three types of internal matrix structures were recognized in HeLa cells infected with adenovirus 2; an amorphously dense region; granular regions representing virus capsid assembly structures; and filaments connecting these regions to one another and to the peripheral lamina (Zhonghe et al, 1987); the perinuclear matrix was also rearranged after adenovirus infection.

Electron micrographs of thin sections through nuclei of adenovirus-infected HeLa cells showed that the ³H-deoxyuridine grains were located at the periphery as well as in the interior of nuclei. Simultaneous visualization of adenovirus transcription and replication showed that the two processes occurred in adjacent, yet distinct, foci throughout the interior and periphery of nuclei presumably in association with the nuclear matrix; DNA molecules were found to be displaced from the replication foci and to become spread in the surrounding nucleoplasm serving as templates for transcription (Pombo et al, 1994).

Adenovirus infection provokes dramatic rearrangements to the nuclear matrix. A reorganization in both internal and peripheral NM was also observed in HeLa cells after infection with adenovirus 2 giving structures able to support the increased replication demands and capsid assembly of the virus (Zhonghe et al, 1987). Splicing of adenoviral HnRNA takes place on the nuclear matrix. All adenovirus 2 polyadenylated RNAs could be UV crosslinked to two host HnRNP proteins that are involved in the association of HnRNA to the matrix (Mariman et al, 1982).

Adenovirus establishes foci called replication centers within the nucleus, where adenoviral replication and transcription occur; although the rAAV genome was faintly detectable in a perinuclear distribution after successfully entering the cell, AAV was mobilized to the adenovirus replication centers when the cell was infected with adenovirus; thus AAV colocalizes with the adenovirus replication centers (Weitzman et al, 1996).

B. Adenovirus E1A and E1B proteins in apoptosis and control of the host cell cycle

Viruses have developed strategies to shut down protein synthesis in the host and subdue its protein synthesizing machinery to produce progeny virus when infecting cells. In response, many cell types commit suicide after viral infection to protect the organism from further infection. Striking back, viruses have evolved mechanisms to prevent infected cells from perishing using mechanisms that inhibit apoptosis of the host cell; adenoviruses synthesize the 19 kDa E1B protein which has a domain similar to that of the cellular protein Bcl-2, the apoptosis inhibitor (Sarnow et al, 1982; van den Heuvel et al., 1990). p53 can be complexed with adenovirus E1B (Sarnow et al, 1982; van den Heuvel et al., 1990).

Expression of the adenovirus E1A protein stimulates host DNA synthesis and induces apoptosis; on the contrary E1B 19 kDa associates with Bax protein and inhibits apoptosis (Figure 1). The E1A oncogene of adenovirus exerts its effect via p53 protein (Debbas and White, 1993; White, 1993). Indeed, expression of E1A increases the half-life of p53 resulting in accumulation of p53 molecules in adenovirus-infected cells leading to apoptosis. Although induction of host DNA synthesis by E1A provides a suitable environment for virus replication, the induction of apoptosis by the same protein impairs virus production since virus-infected cells are eliminated (see Han et al, 1996 for references). p53-deficient cells are transformed by E1A because of absence of the pathway for induction of apoptosis by p53 (Lowe et al, 1994).

E1A represses HER-2/neu transcription and functions as a tumor suppressor gene in HER-2/neu-overexpressing cancer cells. Transfer the E1A gene into cancer cells that overexpress HER-2/neu is an interesting aspect of gene therapy (see E1A in gene therapy; Yu et al, 1995; Chang et al, 1996; Ueno NT et al, 1997; Rodriguez et al, 1997; Xing et al, 1997).

The E1B oncogene products inhibit apoptosis induced by E1A expression thus preventing premature death of host cells during adenovirus infection. This gives an advantage to virus for its proliferation and E1B proteins (19 kDa and 55 kDa) are necessary for transformation of primary rodent cells by E1A. E1A alone is unable to transform primary rodent cells (White, 1993).

The E1B 19K protein of adenovirus is the putative viral homolog of the cellular Bcl-2 protein; using the yeast two-hybrid system for the identification of proteins interacting with E1B, Han and coworkers (1996) have identified Bax as one of the seven 19k-interacting clones. The 50-78 amino acid domain of Bax contains a conserved region homologous to Bcl-2 which is able to interact specifically with either Bcl-2 or E1B. In p53 mutant cells expression of Bax induced apoptosis; inhibition of apoptosis by Bcl-2 may proceed via its ability to bind the death-promoting Bax protein (Han et al, 1996). The bax gene is upregulated by p53.

Expression of p53 and of adenovirus E1A induce apoptosis (Debbas and White, 1993; Lowe and Rudley, 1993). A number of proteins when expressed at sufficient amounts block apoptosis; these include Bcl-2 and E1B 19 kDa protein of adenovirus (Debbas and White, 1993; Chiou et al, 1994). All four protein molecules act upstream of Bax which is a potent inducer of apoptosis: both the cellular Bcl-2 and the 19 kDa protein E1B of adenovirus are able to interact with Bax inhibiting its involvement in induction of apoptosis (Han et al, 1996; Figure 1). E1A acts upstream of p53 by increasing the half-life of p53 resulting in an accumulation of p53 molecules in the nucleus (Lowe and Ruley, 1993); increased levels of p53 are then believed to upregulate the bax gene.

The transcription factor E2F was originally identified as an activator of the adenovirus E2 gene and is implicated in the regulation of DNA replication (Shirodkar et al., 1992). Following infection of cells with adenovirus, the DNA binding activity of E2F increases and as a consequence transcription of the E2 gene of adenovirus increases (Kovesdi et al., 1987). These changes in E2F are mediated by E1A protein of adenovirus. RB forms specific complexes with E2F keeping E2F in a form unable to upregulate its target regulatory sequences. E2F can form specific complexes also with cyclin A during S-phase in NIH 3T3 cells (Mudryj et al., 1991). Both types of complexes, E2F-RB and E2F-cyclin A, can be dissociated by the adenovirus E1A protein (Chellappan et al., 1991; Bagchi et al., 1990; reviewed by White, 1998 this volume) but also by phosphorylation of RB at G1/S causing release of E2F and stimulation in transcription of genes required for DNA replication (myc, DHFR). These events contribute to the uncontrolled proliferation of adenovirus-transformed cells (Mudryj et al., 1990, 1991). Release of E2F from RB induced by E1A is critical for transformation of cells by E1A (for references see Hiebert et al, 1995).

C. Strategies of adenoviruses to enter the cell

In order to enter the host cell the adenovirus first attaches with a high affinity to a cell surface receptor, whose nature still remains elusive, using the head domains of the protruding viral fibers; the fibronectin-binding integrin on the cell surface then associates with the penton base protein on the adenovirus triggering endocytosis of the virus particle via coated pits and coated vesicles (Svensson and Persson, 1984; Greber et al, 1996). The third step in adenovirus entry into the host cell includes

IV. Gene delivery with Adeno-Associated Virus (AAV)

A. Replication of AAV and rAAV: the role of the inverted terminal repeats

AAVs are replication-defective parvoviruses, not associated with any human disease (nonpathogenic), requiring cotransfection with a helper virus to produce infectious virus particles; they can replicate in cell culture only in the presence of coinfection with adenovirus or herpes virus. Five serotypes of distinct AAV isolates have been recovered from human and other primates. AAV infections in humans are asymptomatic acquired with other viral infections such as adenovirus or HSV infections; 80-90% of adults are seropositive for antibodies against AAV (for references see Clark et al, 1995; Berns and Linden, 1995).

The replication of the AAV is dependent on two copies of a 145-bp inverted terminal repeat (ITR) sequence that flanks the AAV genome which is the primary cis-acting element required for productive infection and the generation of recombinant AAV (rAAV) vectors.

In the absence of helper virus, the AAV particle can penetrate cells and find its way to the cell nucleus where the linear genome is uncoated and becomes integrated at a specific site on chromosome 19q13.3; several copies of AAV may integrate in tandem arrays. Thus, the AAV establishes a latent infection; the integrated viral genome can be activated and rescued by superinfection with helper virus (either adenovirus or any type of herpes virus). Inverted repeats at the ends of the viral DNA serve for the integration appearing near the junctions with cellular DNA sequences (Bohenzky et al, 1988).

Adenovirus establishes foci called replication centers within the nucleus, where adenoviral replication and transcription occur; AAV was colocalized with the adenovirus replication centers using in situ hybridization and immunocytochemistry; AAV may, thus, utilize adenovirus and cellular proteins for its own replication; the rAAV genome was faintly detectable in a perinuclear distribution after successfully entering the cell; however, rAAV was mobilized to replication centers when the cell was subsequently infected with adenovirus (Weitzman et al, 1996).

Xiao et al (1997) have engineered the pDD-2 plasmid containing two copies of the D element, a unique sequence adjacent to the AAV nicking site, flanking a single ITR (a total of only 165 bp of AAV sequence); this modified hairpin was sufficient to sustain replication of the plasmid vector when Rep and adenovirus helper functions were supplied in trans. This plasmid has a significant prospect in gene transfer because is replicated more efficiently than infectious AAV clones; as a prelude to its replication the input circular plasmid was converted into a linear substrate by resolution of the AAV terminal repeat through a Holliday-like structure, a process most likely mediated by host factors. Linear monomer, dimer, and other higher-molecular-weight replicative intermediates were generated during the replication of pDD-2, a feature characteristic of AAV replication. The replicative intermediates of this plasmid substrate were competent for AAV DNA replication, encapsidation, infection, integration, and subsequent rescue from the chromosome when superinfected with Ad and wild-type AAV (Xiao et al, 1997). The elucidation of the important role of this 165-bp ITR sequence for AAV replication and the entire life cycle invigorates the important role of inverted repeats at the origin of replication not only of viruses but also of cellular origins of replication (Boulikas, 1996e).

B. Packaging capabilities of AAVs

AAVs posses a 4.7 kb single-stranded DNA genome. Hermonat et al (1997) have examined the maximum amount of DNA which can be inserted into the wild-type AAV genome without compromising packaging into an infectious virus particle; the maximum effective packaging capacity of AAV, examined as increments of 100 bp ligated at map unit 96 of AAV, is approximately 900 bp larger than wild type. Thus, wtAAV therapy vectors can be generated carrying a foreign gene of 900 bp or less with the advantages of wtAAV such as the ease in which high titers of infectious virus can be generated and the ability to specifically integrate in chromosome 19.

On the contrary, the payload capacity of recombinant AAV, which has been deprived of its viral genes and bears only the ITRs is in the order of 4.5-4.7 kb; this means that a cDNA up to this size can be inserted into a rAAV; for example the size of the CFTR cDNA is 4.5 kb and thus, the combined length of the promoter that drives CFTR expression and ITRs needs to be kept under 500 bp (Dong et al, 1996).

Similar results were reported by Dong et al (1996) who have estimated that the optimal size of AAV vector is between 4.1 and 4.9 kb; the packaging efficiencies were sharply reduced above 5.2 kb and below 4.1 kb; two copies of the vector were packaged into each virion when vectors of 2.2-2.5 kb were provided.

C. Integration of wtAAV but not of rAAV is site-specific

Wild-type AAV is able to undergo targeted integration on chromosome 19 after infection in 15 out of 22 clones examined (Kotin et al, 1990, 1992). Of 51 integrations examined by fluorescence in situ hybridization (FISH) 48 (94%) were to chromosome 19 after infection of IB3-1 bronchial epithelial cells with wild-type AAV (Kearns et al, 1996). Site-specific integration has been reported for other viruses including avian leukosis virus (ALV) integrating adjacent to cellular oncogenes in tumors; however, the mechanism of ALV integration involves a process of selection of cells able to form tumors by overexpression of the oncogene due to virus integration rather than exclusive integration of the ALV at unique sites of the genome (Hayward et al, 1981). RSV also appears to be integrated at a limited number of sites (Shih et al, 1988). Adenovirus integration, a more rare event compared to the majority of episomal molecules, may also occur at a number of preferred sites (Jessberger et al, 1989). A larger number of recombinase molecules than those known today may be present in mammalian cell nuclei and promote site-specific integration and recombination events.

Although the human wild-type AAV (wtAAV) is unique in its ability to target viral integration to a specific site on chromosome 19, the recombinant AAV (rAAV) vectors have lost the site-specific integration and targeting ability; furthermore, rAAVs have incapacitated ability to integrate, and can be found as episomes. When wtAAV-2 was used to infect IB3-1 bronchial epithelial cells all metaphase spreads examined by fluorescence in situ hybridization (FISH) had integrated copies and 94% of the integrations were to chromosome 19; furthermore, 36 of 56 metaphase spreads had a single copy of wtAAV integrated and 20 of 56 showed two sites within chromosome 19 (Kearns et al, 1996). On the contrary, when a recombinant AAV containing the CFTR cDNA was used to infect the same cells, examination of 67 metaphase chromosome spreads identified four integrations (only 6% of total) to different chromosomes. No integration was to chromosome 19. When these studies were repeated on the A35 epithelial cell line selected for stable CFTR expression, the episomal AAV-CFTR sequences were abundant in the low molecular weight DNA fraction (Kearns et al, 1996).

Yang et al (1997) have cloned over 40 AAV and rAAV integration junctions to determine the terminal-repeat sequences that mediate integration. These studies have shown that in both immortalized and normal diploid human cells, wt AAV targeted integration to chromosome 19 in head-to-tail tandem arrays; the majority of the junction sequences were involving incomplete copies of the AAV inverted terminal repeats (ITRs); inversions of genomic and/or viral DNA sequences at the wt integration site took place. The viral integration event was found to be mediated by terminal repeat hairpin structures and cellular recombination pathways. In contrast, rAAV provirus integrated on chromosome 2 and at the same locus in two independent cell lines, in both the flip and flop orientations; genomic rearrangements took place at the integration site of rAAV, mainly involving deletions and/or rearrangement-translocations.

Similar data were reported by Rutledge and Russell (1997): recombinant AAV vectors were found to be integrated by nonhomologous recombination as single-copy proviruses in HeLa cells and at random chromosomal locations; the recombination junctions were scattered throughout the vector terminal repeats with no apparent site specificity; the flanking HeLa DNA at integration sites was not homologous to AAV or to the site-specific integration locus of wild-type AAV. Furthermore, vector proviruses with nearly intact terminal repeats were excised from the genomic HeLa DNA and were amplified after infection of cells with wild-type AAV and adenovirus.

The integration patterns of four recombinant AAV-2 genomes in individual clonal isolates of the human nasopharyngeal carcinoma cell line (KB) were different; the difference between the recombinant AAV-2 genomes were in the combinations of the genes for resistance to tetracycline, to neomycin, to ampicillin, with the genes for AAV replication, and the AAV capsid genes. None of the KB cell clones examined had the proviral genome covalently linked to the specific-site of integration of the wt AAV on chromosome 19 (Ponnazhagan et al, 1997a,b).

D. Drawbacks of AAV in gene therapy and their remedy

Gene transfer with AAV vectors has typically been low. Difficulties in generating recombinant virions on a large scale sufficient for preclinical and clinical trials and in obtaining high-titer virus stocks after the initial transfection into producer cells is a limiting factor for the widespread usage of AAV vectors; this obstacle is expected to be overcome in the near future. The high viral titers required for preclinical and clinical studies have been achieved by a new strategy developed by Tamayose et al (1996); AAV vector particles in cell lysates could be concentrated by sulfonated cellulose column chromatography to a titer higher than 10⁸ cfu/ml or 5 x 10¹⁰ particles/ml. A method for transfecting cells at extremely high efficiency with a rAAV vector and complementation plasmid while simultaneously infecting those cells with replication competent adenovirus using adenovirus-polylysine-DNA complexes has been developed by Mamounas et al (1995).

The difficulties in developing packaging cell lines for AAV relate to low levels of rep gene expression from the AAV-p5 promoter and to the propensity of Rep proteins to suppress continued growth of immortalized cell lines; expression of AAV rep under control of the LTR of the human HIV together with the development of cell populations containing rescuable AAV recombinant genomes increased 50-fold the packaging efficiency of AAV vectors (Flotte et al, 1995).

After infection of cell cultures with recombinant AAV there is a decline in the percentage of cells expressing the transferred gene with time in culture. This decline was associated with ongoing losses of vector genomes (Malik et al, 1997). For example, transfer to cultures of K562 human erythroleukemia cells of a truncated rat nerve growth factor receptor (tNGFR) cDNA as a cell surface reporter under control of the LTR of the Moloney murine leukemia virus showed that about 30% of cells expressed tNGFR on the surface early after transduction at a multiplicity of infection (MOI) of 13 infectious units (IU), which declined to 3% over 1 month of culture. At an MOI of 130 IU, nearly all cells expressed tNGFR immediately and the proportion of cells expressing tNGFR declined to 62% over 2 months of culture (Malik et al, 1997).

Another obstacle of rAAV vectors is the low rate of integration of rAAV into the host genome (which can be improved at high MOI). The efficiency of integration was about 2% at low MOI (1.3 IU) and increased to at about 49% at an MOI of 130 (Malik et al, 1997).

E. Advantages using AAV and improvements in AAV gene delivery

AAV does not elicit an immune reaction and is a nonpathogenic virus to humans. AAV contains normally a single-stranded copy of its genome. Transduction with AAV can be enhanced in the presence of adenovirus gene products through the formation of double stranded, non-integrated AAV genomes.

AAV has been reported to have advantages over other viruses for gene transfer to hematopoietic stem cells due to their high titers and relative lack of dependence on cell cycle for target cell integration. A robust CMV/LacZ reporter gene expression in primary human CD34⁺CD2^- progenitor cells induced to undergo T-cell differentiation was obtained without toxicity or alteration in the pattern of T-cell differentiation. 70% to 80% of the cells isolated from either adult bone marrow or umbilical cord blood were efficiently transduced with AAV; however, the expression was transient without integration; this limits the potential use of AAV in gene therapy strategies for diseases such as AIDS (Gardner et al, 1997).

Gene transduction by AAV vectors in cell culture can be stimulated over 100-fold by treatment of the target cells with agents that affect DNA metabolism, such as irradiation or topoisomerase inhibitors (Russell et al, 1995); great improvements in transduction efficiency can also be achieved in vivo: previous g-irradiation increased the transduction rate in mouse liver by up to 900-fold, and the topoisomerase inhibitor etoposide increased transduction by about 20-fold after direct liver injection or after systemic delivery via tail vein injection; up to 3% of hepatocytes could be transduced after a single systemic vector injection (Koeberl et al, 1997). This is a significant advantage compared to stealth liposomes which , although concentrating in the liver, spleen and tumors can transduce Kupffer cells but not hepatocytes after systemic delivery (Martin and Boulikas, 1998, following article).

A combination of the adenovirus-5 capsid protein or the Fiber protein of adenovirus with liposomes, termed adenosomes (adenovirus protein-cationic liposome complexes) improved the efficiency of gene transfer. This complex was able to mediate efficient transfer of a AAV/CMV-LacZ construct to endothelial cells (Zhou et al, 1995).

Clark et al (1996) have developed a sensitive assay system to determine infectivity of AAV vectors based on the replication of input rAAV genomes rather than transgene expression which depends on the type of promoter which drives the foreign gene; this system uses a cell line that expresses AAV helper functions (rep and cap) upon induction by adenovirus infection.

F. Examples using AAV for gene transfer

AAV will infect a broad number of mammalian cell lines and has been used as a cloning vector to transduce the NeoR gene into mammalian tissue culture cells (Hermonat and Muzyczka 1984). Antisense AAV vectors have been used to inhibit HIV replication (Chatterjee et al, 1992), and to correct Fanconi's anemia in human hematopoietic cells (Walsh et al, 1994). AAVs transduce preferentially cells in S phase; topoisomerase inhibitors increase transduction efficiency (Russell et al, 1995).

Using AAV, the genomic copy of a normal human b-globin gene under control of the DNase l-hypersensitive site 2 (HS-2) from the locus control region was expressed in K562 human erythroleukemia cells, which normally lack the b-globin gene; following selection with G418 by virtue of the neo-resistance function which was provided in the rAAV vector, stable integration of the exogenous b-globin allele was determined (Zhou et al, 1996). Similar data were reported by Einerhand et al (1995) transferring a recombinant AAV-vector containing a human b-globin gene together with the DNase1 hypersensitive sites 4, 3 and 2 of the human b-globin locus control region as an approach for the gene therapy of b-thalassemia and sickle cell anemia. The vector replicated to high titers and could efficiently transduce hematopoietic stem cells isolated from patients. In order to treat sickle cell anemia Lubovy et al (1996) have transferred lacZ with a recombinant AAV vector and stably transduced hematopoietic stem cells purified from normal and homozygous sickle cell anemia patients.

AAV was able to promote delivery of functional levels of glial cell line-derived neurotrophic factor (GDNF), in a degenerative model of Parkinson's disease (Mandel et al, 1997). AAV has also been used for the transduction of the mouse liver in vivo with Factor IX cDNA as a prelude to treatment of hemophiliacs (Snyder et al, 1997; Herzog et al, 1997) and for the human intratracheal instillation of CFTR cDNA into neonatal New Zealand white rabbits (Rubenstein et al, 1997), and to the lungs of rhesus macaques without eliciting inflammation (Conrad et al, 1996).

AAV has also been used for the transfer of the human multidrug resistance gene (hMDR1) cDNA to NIH-3T3 cells followed by selection of successfully transfected cells based on the drug-resistant phenotype conferred by the P-glycoprotein efflux pump (see below and Lee et al, 1997, this volume); integration of MDR1 sequences into the host cell genome was demonstrated by fluorescent in situ hybridization (FISH) but also the persistence of nonintegrated AAV-MDR1 episomal plasmids (Baudard et al, 1996).

Introduction of a human globin gene into murine hematopoietic bone marrow cells ex vivo with a recombinant AAV vector followed by transplantation of these cells into lethally irradiated congenic mice sustained a long-term repopulating ability: human globin gene sequences were detected in the bone marrow and spleen in primary recipient mice for at least 6 months.

Kessler et al (1996) have shown that following a single intramuscular administration of a recombinant adeno-associated virus (rAAV) vector, carrying either the lacZ or the human erythropoietin gene into adult BALB/c mice leads to the local production of the foreign protein in the muscle for at least 32 weeks; furthermore, human erythropoietin was secreted and stimulated red blood cell production in the mouse for up to 40 weeks. This finding was extended by Fisher et al (1997) who arrived to the unexpected finding that intramuscular injection of highly purified recombinant AAV can sustain a high level of transgene expression in the absence of adenovirus after direct injection to the muscle in mice (Figure 4); this expands the potential of AAV for the treatment of inherited and acquired diseases. Using this approach no humoral or cellular immune responses were elicited after transfer of lacZ against the neoantigenic E. coli b-galactosidase. The rAAV genome was integrated at single sites as head-to-tail concatamers into nuclei of differentiated muscle fibers. Transfer of the lacZ gene using a highly purified preparation of AAV which was injected into the skeletal muscle of adult mice in the presence of E2a-deleted adenovirus to enhance transduction followed by direct visualization of the b-galactosidase by X-gal histochemistry revealed high transduction of muscle fibers by day 17 associated with inflammation (Fig 4a and b). Animals that received the same AAVlacZ in the absence of adenovirus demonstrated higher levels of transduction that persisted for 240 days (Fig 4c-h).

AAV-mediated delivery of the lacZ gene by direct injection to brain tumors which were induced from human glioma cells in nude mice showed that 30-40% of the cells along the needle track expressed b-galactosidase; subsequent delivery of the HSV-tk/IL-2 genes to these tumors with AAV and administration of GCV to the animals for 6 days resulted in a 35-fold reduction in the mean volume of tumors compared with controls by a significant contribution from the bystander effect (Okada et al, 1996).

A phase I clinical trial for CF is being conducted at Johns Hopkins Hospital using AAV (see Kearns et al, 1996, and protocols #165, 166 in Appendix 1).

V. Herpes Simplex Virus-1 (HSV-1) and miniviral vectors

HSV-1 has a capacity of inserting up to 30 kb of exogenous DNA which is a clear advantage over the adenovirus (up to 7.5 kb of exogenous DNA). High titer viral stocks can be prepared from HSV-1. HSV-1 also displays a wide range of host cells and can infect nonreplicating cells such as neuron cells in which the vectors can be maintained indefinitely in a latent state. However, infection with HSV-1 is cytotoxic to cells because of residual viral proteins produced by the virus. Strategies to circumvent this drawback led to the development of viral vectors with a very large capacity for insertion (almost as large as the size of the virus) which depend on defective helper virus for replication and packaging into infectious virions (see below). A miniviral vector can combine the advantage of cloning the gene in bacterial plasmids, the high efficiency of virus-mediated gene transfer, and the possibility to transfer large genomic DNA fragments including far upstream, downstream and intronic regulatory elements.

The HSV-1 genome is a 152 kb double-stranded DNA containing three origins of replication and encoding at least 72 unique proteins; it consists of a unique long segment replicated from oriL and two repeats flanking the unique segment each replicated from oriS. Spaete and Frenkel (1982) have constructed plasmids containing the lytic viral origin of replication, foreign DNA inserts, and the terminal packaging signal sequences; in the presence of a wild-type helper virus such an amplicon was amplified into multimeric tandemly-repeated forms of the original vector by rolling-circle replication and was packaged into infectious HSV virions (Spaete and Frenkel, 1982). However, the helper virus caused death of the infected cells due to lytic replication and this system is not amenable to gene therapy.

To circumvent this bottleneck two strategies have been developed leading to replication-defective helper HSV: (i) a temperature-sensitive system permitted production of virion stocks at 31^o C whereas infection of cells at 37^o C caused inactivation of the helper virus which was incapable of entering the lytic cycle and allowed delivery of the miniviral vector to the target cell without causing its death. (ii) In a different system, the immediately early gene IE3 was deleted from the helper virus; IE3 encodes for a protein (ICP4) essential for early and late viral gene expression and replication; the helper cell line used for packaging had a genomic insertion of the IE3 gene of HSV which was functionally expressed allowing for complementation and for lytic infection using the IE3-defective HSV virus (DeLuca and Schaffer, 1987; Geller and Freese, 1990).

Two types of viral vectors have been used for gene transfer to cancer cells: replication-incompetent vectors expressing a gene product that leads to the destruction of the tumor or replication-competent vectors that are inherently cytotoxic to the tumor cells. In order to combine the two modes of action Miyatake et al (1997) used a defective HSV vector that consisted of a defective particle, containing tandem repeats of the HSV-tk gene, and a replication-competent, non-neurovirulent HSV mutant as a helper virus. When glioma GL261 cells were infected with the tk-defective vector/helper virus the HSV-TK activity was significantly higher than that in helper virus-infected cells which contained a single copy of HSV-tk; subcutaneous injection of these cells to C57BL/6 mice inducing gliomas led to a significant decrease in tumor size after GCV treatment.

An HSV-1 vector containing a 6.8-kb fragment of the rat tyrosine hydroxylase promoter (pTHlac) supported a seven- to 20-fold increase in reporter gene expression in catecholaminergic cell lines compared to noncatecho-laminergic cell lines. Furthermore, 4 days after stereotactic injection into the midbrain of adult rats and for a duration of 6 weeks, pTHlac supported a 10-fold targeting of b-galactosidase expression to tyrosine hydroxylase-expressing neurons in the substantia nigra pars compacta compared with pHSVlac; this long term expression was significant compared to that from pHSVlac which decreased approximately 30-fold between 4 days and 6 weeks after gene transfer (Song et al, 1997); this study also shows the importance of large control regions in the order of 7 kb in sustaining cell type-correct gene expression, something feasible with HSV and liposomes but nor with recombinant retrovirus, adenovirus, or AAV.

VI. HIV vectors for gene transfer

Recent studies have succeeded in exploiting the deadly HIV-1 virus, after crippling some functions, as a gene delivery vehicle. An advantage of HIV vectors has been the broad range of tissues and cell types they can transduce, a property granted because lentiviral vectors are pseudotyped with vesicular stomatitis virus G glycoprotein. Human lentiviral (HIV)-based vectors can transduce non-dividing cells in vitro and deliver genes in vivo; expression of transgenes in the brain has been detected for more than six months. HIV vectors have been also used to introduce genes directly into liver and muscle; 3-4% of the total liver tissue was transduced by a single injection of 1-3 x 10⁷infectious units (I.U.) of recombinant HIV with no inflammation or recruitment of lymphocytes at the site of injection. Whereas expression of green fluorescent protein (GFP), used as a surrogate for therapeutic protein, was observed for more than 22 weeks in the liver and for over 8 weeks in the muscle using lentiviral vectors, little or no GFP could be detected in liver or muscle transduced with the Moloney murine leukemia virus (Mo-MLV), a prototypic retroviral vector (Kafri et al, 1997).

The development of a stable noninfectious HIV-1 packaging cell line capable of generating high-titer HIV-1 vectors is another important step towards use of HIV vectors in gene therapy (Corbeau et al, 1996). A hybrid murine leukemia virus-based vector containing U3 and R sequences from HIV-1 in place of the MLV U3 and R regions gave single transcriptional unit retroviral vectors under the control of Tat; this vector has advantages for anti-HIV gene therapy (Cannon et al, 1996).

Although replication-incompetent HIV vectors displayed a strict CD4⁺ T cell tropism for gene transfer, a feature important for AIDS therapy, it was thought to preclude HIV-based vectors for other gene transfer applications; a two-step gene transfer system, however, was developed to expand the host range of the HIV vector: in the first step, the CD4 gene was introduced into target cells using a replication-defective adenoviral vector; in the second step the CD4-transfected cells were incubated with HIV vectors which resulted in stable integration and HIV-mediated gene transfer (Miyake et al, 1996).

An HIV multiply attenuated vector in which the virulence genes env, vif, vpr, vpu, and nef were deleted was able to deliver genes in vivo into adult neurons (Zufferey et al, 1997).

HIV-mediated gene transfer was used to transfer the GFP gene under control of CMV to retinal cells by injection into the subretinal space of eyes in rats; the GFP gene was efficiently expressed in both photoreceptor cells and retinal pigment epithelium; predominant expression in photoreceptor cells was achieved using the rhodopsin promoter. The transduction efficiency was high and photoreceptor cells in >80% of the area of whole retina were expressing GFP (Miyoshi et al, 1997).

VII. Epstein-Barr virus (EBV) and baculovirus vectors

EBV is an episomaly-replicating virus in synchrony with the cell cycle. EBV infects human cells causing mononucleosis; the presence of the unique latent origin of replication (oriP) in EBV allows for episomal replication of the virus in human cells without entering the lytic cycle. The presence of oriP and of the replication initiator protein EBNA1 cDNA on a vector allows episomal replication in human cells; in addition, plasmids containing only oriP can replicate episomally into cell lines expressing EBNA-1 (Sun et al, 1994; Banerjee et al, 1995).

A hybrid HSV-1/EBV vector has been developed by Wang and Vos (1996), which combines (i) the HSV-1 lytic oriS; (ii) an HSV-1 packaging sequence which allows replication and packaging in the presence of defective helper virus carrying a deletion in the IE3 gene in the E5 cell line expressing the IE3 gene; (iii) the latent oriP of EBV and (iv) the EBNA-1 cDNA allow episomal replication of the infectious vector in the E5 cell line so that viral stocks of high titer can be made. Infection of tumor-derived fibroblast and epithelial cell lines in culture and local injection of human liver tumors in nude mice was used to demonstrate 95-99% efficiency of infection and transfer of the reporter b-galactosidase gene.

Genetically modified baculoviruses (Autographa californica nuclear polyhedrosis virus) were used to efficiently deliver genes into cultured hepatocytes of different origin; delivery into human hepatocytes with baculovirus vectors approached 100% efficiency in cell culture and expression levels were high when mammalian promoters were chosen. A number of drawbacks preclude their direct application in vivo; nevertheless gene transfer was feasible in ex vivo perfused human liver tissue (Sandig et al, 1996; Hofmann et al, 1998 this volume).

VIII. Liposomal gene delivery

Abbreviations:

DC-CHOL: 3b [N-(N',N'-dimethylaminoethane)carbamoyl]cholesterol

DDAB: dimethyldioctadecyl ammonium bromide

DMRIE: N-[1-(2,3-dimyristyloxy)propyl]-N,N-dimethyl-N-(2-hydroxyethyl) ammonium bromide

DMTAP: 1,2-dimyristoyl-3-trimethylammonium propane

DOGS: Dioctadecylamidoglycylspermine (Transfectam, Promega)

DOPE: dioleyl phosphatidylethanolamine

DOSPA: 2,3-dioleoyloxy-N-[2(sperminecarboxamido)ethyl]-N,N-dimethyl -1-propanaminium trifluoroacetate

DOTAP: N-(1-(2,3-dioleoyloxy)propyl)-N,N,N-trimethylammonium chloride

DOTMA: N-[1-(2,3-dioleyloxy) propyl]-n,n,n-trimethylammonium chloride

DPTAP: 1,2- dipalmitoyl-3-trimethylammonium propane

DSTAP: 1,2-disteroyl-3-trimethylammonium propane

Lipofectin: DOTMA:DOPE 1:1 (GIBCO BRL)

A. Immune responses and toxicity of cationic lipid-DNA complexes

Cationic lipids have been widely used for gene transfer; a number of clinical trials (34 out of 220 total RAC-approved protocols as of December 1997) use cationic lipids (see Table 4 in Martin and Boulikas, 1998, this volume, pages 203-206). Although many cell culture studies have been documented, systemic delivery of genes with cationic lipids in vivo has been very limited. All clinical protocols use subcutaneous, intradermal, intratumoral, and intracranial injection as well as intranasal, intrapleural, or aerosol administration but not i.v. delivery because of the toxicity of the cationic lipids and DOPE (see Table 4 in Martin and Boulikas, 1998, this volume, pages 203-206).

Liposomes formulated from DOPE and cationic lipids based on diacyltrimethylammonium propane (dioleoyl-, dimyristoyl-, dipalmitoyl-, disteroyl-trimethylammonium propane or DOTAP, DMTAP, DPTAP, DSTAP, respectively) or DDAB were highly toxic when incubated in vitro with phagocytic cells (macrophages and U937 cells), but not towards non-phagocytic T lymphocytes; the rank order of toxicity was DOPE/DDAB > DOPE/DOTAP > DOPE/DMTAP > DOPE/DPTAP > DOPE/DSTAP; the toxicity was determined from the effect of the cationic liposomes on the synthesis of nitric oxide (NO) and TNF-a produced by activated macrophages (Filion and Phillips, 1997).

Another factor to be considered before i.v. injections are undertaken is that negatively charged serum proteins can interact and cause inactivation of cationic liposomes (Yang and Huang, 1997). Condensing agents used for plasmid delivery including polylysine, transferrin-polylysine, a fifth-generation poly(amidoamine) (PAMAM) dendrimer, poly(ethyleneimine), and several cationic lipids (DOTAP, DC-Chol/DOPE, DOGS/DOPE, and DOTMA/DOPE) were found to activate the complement system to varying extents. Strong complement activation was seen with long-chain polylysines, the dendrimer, poly(ethyleneimine), and DOGS; complement activation was considerably reduced by modifying the surface of preformed DNA complexes with polyethyleneglycol (Plank et al, 1996).

B. Mechanism of liposome entry to cells

Cationic lipids increase the transfection efficiency by destabilizing the biological membranes including plasma, endosomal, and lysosomal membranes; indeed, incubation of isolated lysosomes with low concentrations of DOTAP caused a striking increase in free activity of b-galactosidase, and even a release of the enzyme into the medium demonstrating that lysosomal membrane is deeply destabilized by the lipid; the mechanism of destabilization was thought to involve an interaction between cationic liposomes and anionic lipids of the lysosomal membrane, allowing a fusion between the lipid bilayers; the process was less pronounced at pH 5 than at pH 7.4 and anionic amphipathic lipids were able to prevent partially this membrane destabilization (Wattiaux et al, 1997).

In contrast to DOTAP and DMRIE which were 100% charged at pH 7.4, DC-CHOL was only about 50% charged as monitored by a pH-sensitive fluorophore; this difference decreases the charge on the external surfaces of the liposomes and was proposed to promote an easier dissociation of bilayers containing DC-CHOL from the plasmid DNA and an increase in release of the DNA-lipid complex into the cytosol from the endosomes (Zuidam and Barenholz, 1997).

C. Tissue targets using cationic liposomes in vivo

Although cationic lipids have been used widely for the delivery of genes very few studies have used systemic i.v. injection of cationic liposome-plasmid complexes because of the toxicity of the lipid component and certainly in animal models, not humans. Administration by i.v. injection of two types of cationic lipids of similar structure, DOTMA and DOTAP, has shown that the transfection efficiency was determined mainly by the structure of the cationic lipid and the ratio of cationic lipid to DNA; the luciferase and GFP gene expression in different organs was transient, with a peak level between 4 and 24 hr, dropping to less than 1% of the peak level by day 4 (Song et al, 1997).

Figure 5 shows the effect of cationic lipid:DNA ratio on transfection efficiency after i.v. tail injection. Luciferase activity was detected in all organs examined with the highest level in lung. In the absence of neutral lipid both DOTMA and DOTAP promoted a linear increase in luciferase activity in the lung with increasing lipid:DNA from 12:1 to 36:1 nmol lipid: mg of DNA. DOTMA was 10 times more efficient than DOTAP (10⁶ versus 10⁷ relative luciferase units (RLU) per mg protein. Cholesterol (Chol) mixed with DOTMA (1:1 molar ratio) decreased the level of gene expression in the lung whereas cholesterol did not affect the transfection efficiency of DOTAP liposomes. Inclusion of DOPE into either DOTAP or DOTMA liposomes significantly decreased the transfection efficiency by 100-fold in the lung.

When a group of four cationic lipids with identical head group but of different fatty acyl chains were tested for their transfection efficiencies (Figure 6); these included DOTAP, DMTAP, DPTAP, and DSTAP. The C₁₄ acyl chain-lipid DMTAP had a similar transfection efficiency as DOTAP which has 18 carbon atoms in the acyl chain and one double bond (C₁₈D₉); on the contrary, the transfection efficiencies of DPTAP (C₁₆) was 10-100 fold lower and that of DSTAP (C₁₈) was 100 to 1000 fold lower.

Confocal microscopy of lung tissue after injection of 25 mg pCMV-GFP plasmid DNA complexed with DOTMA liposomes to mice (Figure 7) has shown that the type of cells that express the transgene are the endothelial cells that have typical characteristics of neighboring multiple air-sac structures (Figure 7D).

A number of different organs in vivo can be targeted after liposomal delivery of genes or oligonucleotides. Intravenous injection of cationic liposome-plasmid complexes by tail vein in mice targeted mainly the lung and to a smaller extend the liver, spleen, heart, kidney and other organs (Zhu et al, 1993). Intraperitoneal injection of a plasmid-liposome complex expressing antisense K-ras RNA in nude mice inoculated i.p. with AsPC-1 pancreatic cancer cells harboring K-ras point mutations and PCR analysis indicated that the injected DNA was delivered to various organs except brain (Aoki et al, 1995).

A number of factors for DOTAP:cholesterol/DNA complex preparation including the DNA:liposome ratio, mild sonication, heating, and extrusion were found to be crucial for improved systemic delivery; maximal gene expression was obtained when a homogeneous population of DNA:liposome complexes between 200 to 450 nm in size were used. Cryo-electron microscopy showed that the DNA was condensed on the interior of invaginated liposomes between two lipid bilayers in these formulations, a factor that was thought to be responsible for the high transfection efficiency in vivo and for the broad tissue distribution (Templeton et al, 1997).

Steps to improve for successful liposome-mediated gene delivery to somatic cells include persistence of the plasmid in blood circulation, port of entry and transport across the cell membrane, release from endosomal compartments into the cytoplasm, nuclear import by docking through the pore complexes of the nuclear envelope, expression driven by the appropriate promoter/enhancer control elements, and persistence of the plasmid in the nucleus for long periods. A number of strategies for liposomal delivery and for enhancing the efficiency of uptake by the cells and release from endosomal compartments of plasmid or oligonucleotide DNA are reviewed in the following article (Martin and Boulikas, 1998).

D. Cationic lipids in oligonucleotide transfer

Encapsulation of oligonucleotides into liposomes increased their therapeutic index, prevented degradation in cultured cells and in human serum and reduced toxicity to cells (Thierry and Dritschilo, 1992; Capaccioli et al, 1993; Morishita et al, 1993; Williams et al, 1996; Lewis et al, 1996); conjugation to a fusogenic peptide enhanced the biological activity of antisense oligonucleotides (Bongartz et al, 1994). However, most studies have been performed in cell culture, and very few in animals in vivo; there is still an important number of improvements needed before these approaches can move to the clinic.

Zelphati and Szoka (1997) have found that complexes of fluorescently labeled oligonucleotides with DOTAP liposomes entered the cell using an endocytic pathway mainly involving uncoated vesicles; oligonucleotides redistributed from punctate cytoplasmic regions into the nucleus; this process was independent of acidification of the endosomal vesicles. The nuclear uptake of oligonucleotides depended on several factors such as charge of the particle where positively charged complexes were required for enhanced nuclear uptake; DOTAP increased over 100 fold the antisense activity of a specific anti-luciferase oligonucleotide. Physicochemical studies of oligonucleotide-liposome complexes of different cationic lipid compositions indicated that either phosphatidylethanolamine or negative charges on other lipids in the cell membrane are required for efficient fusion with cationic liposome-oligonucleotide complexes to promote entry to the cell (Jaaskelainen et al, 1994).

Similar results were reported by Lappalainen et al (1997); digoxigenin-labeled oligodeoxynucleotides (ODNs) complexed with the polycationic DOSPA and the monocationic DDAB (with DOPE as a helper lipid) were uptaken by CaSki cells in culture by endocytosis. The nuclear membrane was found to pose a barrier against nuclear import of ODNs which accumulated in the perinuclear area. Although DOSPA/DOPE liposomes could deliver ODNs into the cytosol, they were unable to mediate nuclear import of ODNs; on the contrary oligonucleotide-DDAB/DOPE complexes with a net positive charge were released from vesicles into the cytoplasm; it was determined that DDAB/DOPE mediated nuclear import of the oligonucleotides.

DOPE-heme (ferric protoporphyrin IX) conjugates, inserted in cationic lipid particles with DOTAP, protected oligoribonucleotides from degradation in human serum and increased oligoribonucleotide uptake into 2.2.15 human hepatoma cells; the enhancing effect of heme was evident only at a net negative charge in the particles (Takle et al, 1997). Uptake of liposomes labeled with ¹¹¹In and composed of DC-Chol and DOPE was primarily by liver, with some accumulation in spleen and skin and very little in the lung after i.v. tail injection; preincubation of cationic liposomes with phosphorothioate oligonucleotide induced a dramatic, yet transient, accumulation of the lipid in lung which gradually redistributed to liver. The mechanism of lung uptake involved entrapment of large aggregates of oligonucleotides within pulmonary capillaries at 15 min post-injection via embolism; labeled oligonucleotide was localized primarily to phagocytic vacuoles of Kupffer cells at 24 h post-injection; nuclear uptake of oligonucleotide in vivo was not observed (Litzinger et al, 1996).

Phosphorothioate oligonucleotides were found in most tissues 48 h after i.p. administration with highest concentrations in kidney and liver; complexation of the oligonucleotide with DOTMA did not affect neither the oligonucleotide uptake nor its tissue distribution in normal mice but increased the oligonucleotide cellular uptake (4-10 times) in LOX ascites tumors (Saijo et al, 1994).

Triplex-forming ODNs were delivered to cells in culture using adenovirus-polylysine-ODN complexes designed to take advantage of the receptor mediated endocytosis of adenoviruses to transfer the ODNs to the cell nucleus; nuclear uptake peaked at 4 h and intact ODN persisted in the nucleus with a half-life of 12 h (Ebbinghaus et al, 1996).

E. Fusogenic peptides enhance gene transfer efficiency

Enveloped viruses have evolved efficient mechanisms to release their genomes from the endosomes into the cytoplasm of the host cells; specific envelope proteins of the nucleocapsid are capable of destabilizing the endosomal membrane. Therefore, inactivated viruses have been used to enhance the transfer of plasmids. Addition of adenoviral particles capable of inducing endosome lysis (Blumenthal et al, 1986), mediated by a conformational change in the adenovirus penton protein induced at the lower pH of endosomes (Seth, 1994) can increase transfection efficiency 100-1000 fold using 10⁹ adenoviral particles/ml and the transferrin receptor (Curiel et al, 1991; Cotten et al, 1992; Wagner et al, 1992b; Cristiano et al, 1993; Morishita et al, 1993; Harries et al, 1993; Curiel, 1994; reviewed by Ledley, 1995).

Use of fusogenic peptides from influenza virus hemagglutinin HA-2 enhanced greatly the efficiency of transferrin-polylysine-DNA complex uptake by cells; in this case the peptide was linked to polylysine and the complex was delivered by the transferrin receptor-mediated endocytosis (Wagner et al, 1992a; Plank et al, 1994). This peptide had the sequence: GLFEAIAGFIENGWEGMID GGGYC and was able to induce the release of the fluorescent dye calcein from liposomes prepared with egg yolk phosphatidylcholine which was higher at acidic pH; this peptide was also able to increase up to 10-fold the anti-HIV potency of antisense oligonucleotides, at a concentration of 0.1-1 mM, using CEM-SS lymphocytes in culture (Bongartz et al, 1994). This peptide changes conformation at the slightly more acidic environment of the endosome destabilizing and breaking the endosomal membrane (Murata et al, 1992; Bullough et al, 1994). Fusogenic peptides have been used by other investigators (Midoux et al, 1993; Kamata et al, 1994). It is thought that several fusogenic peptides self-assemble following their conformational change forming a transmembrane channel (Bongartz et al, 1994).

Sendai virosomes were effective for delivering AAV neuropeptide Y (NPY) cDNA constructs in vivo. Injections into brain neocortex of Sendai-virosome encapsulated rAAV construct expressing NPY increased NPY-like immunoreactivity in neurons but not glia; injections into the rat hypothalamic para-ventricular nucleus increased body weight and food intake for 21 days (Wu et al, 1996). Tomita et al (1996) have found that newborn mice can sustain expression of the insulin gene delivered by Sendai virus-liposome complexes for at least 8 weeks as assayed by reverse transcriptase PCR and radioimmunoassay, compared to 2 weeks in adult animals.

A 27 residue peptide vector, containing the fusion sequence of HIV gp41 and the nuclear localization sequence of SV40 T antigen was used to deliver oligonucleotides to cell nuclei very rapidly in cell culture (1h). The complexes formed strongly increased the stability of the oligonucleotide to nucleases, enhanced passage through the plasma membrane, and led to endosomal internalization (Morris et at, 1997).

Certain cationic lipids are endowed with a better ability to disrupt the endosomal membrane and promote release of the plasmid to the cytoplasm, a prelude for its nuclear import. Presentation of plasmid DNA to COS cell cultures using three different lipid formulations: (i) vectamidine (3-tetradecylamino-N-tert-butyl-N'-tetradecylpropionamidi ne), (ii) DOTMA:DOPE (Lipofectin), and (iii) DMRIE-Chol (1:1) resulted in complex entry via endocytosis for all three cationic lipids as revealed using transmission electron microscopy. However, the endosomal membrane in contact with complexes containing vectamidine or DMRIE-Chol, but not Lipofectin, often exhibited a disrupted morphology (El Ouahabi et al, 1997).

F. Plasmid condensation with spermine, polylysine, protamine, histones enhances the transfection efficiency

DNA can be presented to cells in culture as a complex with polycations such as polylysine, or basic proteins such as protamine, total histones or specific histone fractions (Fritz et al, 1996), cationized albumin, and others (Smull and Ludwig, 1962). These molecules increase the transfection efficiency. In addition to HMG1, also histone H1 and HMG17 were identified as transfection-enhancing proteins in cell culture (Zaitsev et al, 1997). Histone H2A significantly enhanced in vitro DNA transfection whereas other histones and anionic liposomes did not (Balicki and Beutler, 1997). Gene transfer through the asialo-glycoprotein receptor-mediated endocytosis pathway was enhanced with the histones H1, H2a, H2b, H3, and H4 which were galactosylated with the crosslinker agent, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide, conjugated to DNA and then used to transfect HepG2 cells, which display the asialoglycoprotein receptor (Chen et al, 1994). Plasmid DNA and HMG1 were efficiently co-encapsulated in liposomes by agitation and sonication, and were co-introduced into cells by hemagglutinating virus of Japan (HVJ)-mediated membrane fusion; the presence of HMG1 enhanced 3-fold the transfection efficiency (Kato et al, 1991).

The interaction of plasmid DNA with protamine sulfate followed by the addition of DOTAP cationic liposomes offered a better protection of plasmid DNA against enzymatic digestion and gave consistently higher gene expression in mice via tail vein injection compared with DOTAP/DNA complexes; 50 mg of luciferase-plasmid per mouse gave 20 ng luciferase protein per mg extracted tissue protein in the lung which was detected as early as 1 h after injection, peaked at 6 h and declined thereafter. Intraportal injection of protamine/DOTAP/DNA led to about a 100-fold decrease in gene expression in the lung as compared with i.v. injection; endothelial cells were the primary locus of lacZ transgene expression (Li and Huang, 1997). Protamine sulfate enhanced plasmid delivery into several different types of cells in vitro using the monovalent cationic liposomal formulations (DC-Chol and lipofectin); this effect was less pronounced with the multivalent cationic liposome formulation, lipofectamine (Sorgi et al, 1997).

Spermine has been found to enhance the transfection efficiency of DNA-cationic liposome complexes in cell culture and in animal studies; this biogenic polyamine at high concentrations caused liposome fusion most likely promoted by the simultaneous interaction of one molecule of spermine (four positively charged amino groups) with the polar head groups of two or more molecules of lipids. At low concentrations (0.03-0.1 mM) it promoted anchorage of the liposome-DNA complex to the surface of cells and enhanced significantly transfection efficiency (Boulikas et al, in preparation).

Because the receptor for ecotropic viruses is a transporter for basic amino acids, use of a histone as a facilitator increased the efficiency of retroviral infection (Singh and Rigby, 1996). Polybrene is the usual agent employed during retroviral infection. For supernate infections, concentrations of 5-10 mg/ml of protamine provided essentially the same infection efficiency as polybrene; protamine displayed low toxicity on a range of cell types and increased 7-fold the efficiency of retroviral infection (Cornetta and Anderson, 1989).

The polycations polybrene, protamine, DEAE-dextran, and poly-L-lysine significantly increased the efficiency of adenovirus-mediated gene transfer in cell culture; this was thought to act by neutralizing the negative charges presented by membrane glycoproteins which reduce the efficiency of adenovirus-mediated gene transfer (Arcasoy et al, 1997).

G. Targeted gene delivery

Targeting a specific cell type or animal tissue is an important goal of gene therapy. Many different approaches have been undertaken to achieve targeting. A recombinant adenovirus encoding an anti-erbB-2 intracellular single-chain antibody (sFv) displayed a genetic selectivity for the erbB-2-positive prostate carcinoma cell lines DU145 and LNCaP; delivery of this recombinant adenovirus resulted in cytotoxicity to the DU145 and LNCaP, but not PC-3, cell lines and reduced the clonogenic capacity of DU145 cells cultured alone or mixed with various ratios of irradiated human bone marrow. This finding led to a strategy for effectively reducing DU145 and erbB-2-positive primary prostate tumor contamination in bone marrow cultures (Kim et al, 1997). Delivery of an anti-erbB-2 single chain (sFv) antibody gene for previously treated ovarian and extraovarian cancer patients is in clinical trials using adenoviral gene delivery (protocol #133).

A luciferase expression vector (pRSVLuc) non-covalently linked to a humanized HER2 antibody (rhuMAbHER2) covalently modified with poly-L-lysine bridges was able to direct gene transfer to HER2 expressing cells in vitro (Foster and Kern, 1997).

A targeting gene therapy approach for hematopoietic stem/progenitor cells has been directed to cell lines expressing the c-kit receptor; plasmid DNA containing a luciferase reporter gene was condensed with polylysine covalently linked to streptavidin (which binds biotinylated ligand) and with polylysine covalently linked to adenovirus (to achieve endosomal lysis) with the final addition of biotinylated steel factor; omission of the adenovirus endosomalytic agent from the vector resulted in the loss of gene expression (Schwarzenberger et al, 1996).

Systemic administration of a c-fos antisense, regulated by mouse mammary tumor virus (MMTV) control elements in a retroviral vector, showed expression only in breast epithelium although the vector could be detected in several tissues thus supporting targeting to MMTV-regulated tissues (Arteaga and Holt, 1996).

Liposomes coated with polyethyleneglycol (PEG) can be efficiently targeted to tumor cells that express folate receptors (KB cells) via conjugation of folate to a PEG spacer of 25 nm in length; shorter PEG spacers were not efficient in mediating binding of the liposomes to KB cells (Lee and Low, 1995).

Neri and coworkers (1997) were able to target an angiogenesis-associated oncofetal fibronectin (B-FN) isoform by affinity-matured recombinant antibody fragments. B-FN is present in vessels of neoplastic tissues during angiogenesis but is absent from mature vessels and could provide a target for diagnostic imaging and therapy of cancer. Phage display libraries were screened to isolate human antibody fragments able to recognize this isoform across species; imaging of F9 murine teratocarcinomas grafted in nude mice is shown on Figures 8 and 9.

H. Targeted gene delivery with peptide-displaying phages

Development of methods to display and select collections of peptides specific for binding a target provide valuable tools to identification of peptide drugs; peptides could be selected for binding biological targets including cell surface receptor molecules, DNA, antibodies, or whole cells. The technique of peptide-displaying phages has been developed for targeted gene delivery. Selection of cell surface-binding peptides, ideally specific for each type of cell in the human body, will be used for incorporation into gene delivery vehicles to achieve the long-searched tissue specificity of the vector (reviewed by Russell, 1996).

Development of the random peptide library as a source of specific protein binding molecules (Devlin et al, 1990) and exposure of random peptides on the surface of phages (Cwirla et al, 1990) has been the catalyst for progress in this promising field. Libraries of random 8 to 12 amino acid peptides expressed on the N-terminus of the pIII protein of the fd phage or on the N-terminus of the pVIII major coat protein of the same phage have been selected that bind the extracellular domain of human IL-1 receptor; screening was against immobilized IL-1 receptor extracellular domain. Two families of peptides could act as antagonists blocking triggering of the IL-1 signaling pathway; because IL-1 levels become elevated in autoimmune and inflammatory disorders, these peptide antagonists of IL-1 receptor could provide novel drugs for these diseases (Yanofsky et al, 1996).

Phages displaying known integrin-binding peptides have been shown to bind and enter mammalian cells (Hart et al, 1994). A peptide antagonist to thrombin receptor has been identified using phage display (Doorbar and Winter, 1994). Production of cell-targeting ligands has been achieved by cell-binding peptides specific for different cell types in culture; these peptides are selected through six rounds of binding (and amplification of phage clones) to a particular cell type from random peptide-presenting phage libraries; the selected peptides are apparently recognizing specific surface receptor molecules. For example, the 20mer peptide KTLTLEAALRNAWLREVGLK has been selected for its high affinity for PEA10 mouse fibroblast cells binding 1000 more efficiently to the cells than random peptides (Barry et al, 1996).

IX. Gene delivery with polymers, peptides and other means

A. Delivery of transferrin-polylysine-DNA complexes

A number of polymers have been tested and shown to enhance significantly the transfection efficiency of plasmids but also of viruses; the enhancement in transfection results from a facilitation in the interaction of plasmids with the cell surface, transport to endosomes, release to the cytoplasm, and in some cases nuclear import (reviewed by Behr, 1994).

A closely related family of ubiquitous DNA binding proteins, called MDBP, binds with high affinity to two 14 base pair (bp) sites within the human cytomegalovirus immediate early gene 1 (CMV IE1) enhancer; these MDBP sites did not require cytosine methylation for optimal binding; mutation of one of the enhancer MDBP sites to prevent MDBP recognition modestly increased the function of a neighboring CREB binding site in a transient transfection assay (Zhang et al, 1995). Furthermore, the CMV promoter competed with the Egr1 promoter for transcription factors or co-factors which might be required for activation by WT1; WT1 was converted from an activator to a repressor by co-transfection of an excess of the parental CMV-based vector (Reddy et al, 1995).

C. Tissue-specific promoters in gene therapy

A number of studies have used tissue-specific promoters and enhancers from mammalian genes in order to attain a cell type-specific expression of the transgene. The discovery of genes which are expressed at high levels in specific tumor cell types has prompted the idea of the use of their promoter or enhancer DNA sequences to express in this particular cancer cell type therapeutically important genes (Venkatesh et al, 1990; Brady et al, 1994; Dimaio et al, 1994; Osaki et al, 1994; Pang et al, 1995).

Examples include the expression of the suicidal CD gene under control of the regulatory regions of the tumor marker gene carcinoembryonic antigen (Richards et al, 1995), the expression of HSV-tk gene, under control of a-fetoprotein enhancer and albumin promoter, into adult liver cells in transgenic animals (Su et al, 1996), the expression of b-galactosidase in tyrosine hydroxylase-expressing neurons in the substantia nigra midbrain of adult rats using the tyrosine hydroxylase promoter (Song et al, 1997), and the expression of the lacZ marker gene under control of the murine pancreatic amylase promoter in the pancreas in neonatal and adult mice (Dematteo et al, 1997). Transduction of the human LDL-R cDNA under the transcriptional control of the liver-type pyruvate kinase promoter allowed high and tissue specific expression of the gene in primary hepatocytes (Pages et al, 1996b).

Fibroblasts, infected with recombinant retroviruses and selected with G418 for the expression of the vector carrying the therapeutic gene, have been used for the ex vivo treatment in animal models; when the therapeutic gene was either under control of the viral LTR or an heterologous internal promoter, expression of the transgene from the integrated retrovirus was shut off (Scharfmann et al, 1991). The use of the dihydrofolate reductase housekeeping gene promoter which is expressed in all cell types, led to sustained expression, albeit at very low levels (Scharfmann et al, 1991); it appears that the combination of a suitable enhancer and promoter for a particular cell type and the method of introduction of the transgene is crucial for sustained expression.

Combination of the mouse muscle creatine kinase enhancer with the human cytomegalovirus promoter to drive the expression of the canine factor IX gene in ex vivo infected mouse primary myoblasts led to the expression of factor IX and its secretion in the blood of mice transplanted with these myoblasts for over 6 months; however, the levels of factor IX protein secreted into the plasma (10 ng/ml for 10⁷ injected cells) were not sufficient to be of therapeutic value (Dai et al, 1992).

Joki and coworkers (1995) have used the promoter of the early growth response gene 1 (EGR-1, also known as Zif/268, TIS-8, NFGI-A, or Krox-24) to confer selective expression of the luciferase gene in glioma cell lines exposed to ionizing radiation; a 10-fold higher activity in luciferase activity was found after irradiation of the cells which was detectable at 1-3 h after stimulation with 20 Gy (stereotactic radiosurgery during treatment of isolated brain metastases, arteriovenous malformations, meningiomas, craniopharyngiomas, and glioblastomas uses a single dose of 20-30 Gy). Transfection of the HSV-tk gene under control of the EGF-1 promoter rendered irradiated, but not nonirradiated, cells sensitive to GCV. Irradiation induces DNA repair, cell cycle arrest, and reinitiation of DNA synthesis in surviving cells; g-radiation also induces higher levels of a number of proteins including p53, AP-1, NF-kB, TNF, IL-1, and EGF-1. Therefore, use of the EGF-1 promoter can activate gene expression selectively in radiation fields and could be used to drive the expression of cytotoxic genes (HSV-tk) for the killing of cancer cells.

Peptides containing the three zinc fingers of Zif268 could efficiently repress activated transcription from promoter constructs prepared with Zif268 binding sites inserted at various positions with respect to the TATA box (Kim and Pabo, 1997); such strategies could find important applications in gene therapy leading to construction of artificial promoters able to activate or repress transcription of transferred genes. A potent hybrid CAG promoter was used to drive the HSV-tk gene and showed effective eradication of pancreatic tumors in animal xenografts (Aoki et al, 1997).

D. Molecular switch systems

The ability to regulate gene expression via exogenous stimuli will facilitate the study of gene functions in mammalian cells. Molecular switch systems have been devised (Wang et al, 1994) allowing the researcher to turn on or off individual genes; the switch used by Delort and Capecchi (1996) is composed of three elements: (i) the inducible UAS promoter, a synthetic promoter containing five GAL4 response elements, normally absent from mammalian genomes; (ii) the synthetic hybrid steroid receptor (TAXI), composed of the GAL4 DNA -binding domain, a truncated human progesterone receptor, and the acidic region from VP16 protein of HSV; the hybrid molecule activates transcription from the UAS promoter when bound to an inducer drug, and (iii) the synthetic nontoxic drug inducer RU486 which is permeable to blood-brain and placental barriers; this model allows up to 100-fold induction of a gene linked to this system and can be finely tuned to lower levels of induction (Delort and Capecchi, 1996).

Transient cotransfection of HeLa cells with the UAS-CAT and the hybrid receptor expression vector showed that the hybrid TAXI protein bound to the UAS promoter only after treatment with RU486 but not progesterone; the TAXI/UAS system was successfully used in transgenic mice to regulate the expression of a human growth hormone gene; the ex vivo approach, however, did not sustain long-term expression of the transgene. This system might allow physicians to alter the level of expression of foreign genes during somatic cell transfer in response to the clinical state of the patient (Delort and Capecchi, 1996).

Iida et al (1996) have modified the tetracycline-controlled inducible system by the addition of the ligand-binding domain of the estrogen receptor to the carboxy terminus of the tTA transactivator; a single retroviral vector could transduce both the transactivator gene and the gene of interest controlled by the tTA-inducible promoter into mammalian cells; cell lines expressing the transactivator were established where the expression of a gene (the toxic G protein of vesicular stomatitis virus) depended on the removal of tetracycline and the addition of estrogen.

A different genetic switch used consisted of the cytochrome P450 1A1 promoter driving the expression of the human apolipoprotein E (apoE) gene in transgenic mice; this switch system was induced by b-naphthoflavone; the inducer could pass transplacentally and via breast milk from an injected mother to her suckling neonatal pups, giving rise to the induction of human apoE in neonate plasma and lowering the cholesterol levels in hypercholesterolemic pups (Smith et al, 1995).

XII. DNA recombination in gene therapy

A. Mechanisms of DNA recombination

Genetic recombination, i.e., exchange of segments of DNA between two molecules of DNA, is a very frequent event. It often occurs during meiosis and also between homologous chromosomes in mitosis. Homologous recombination involving double-strand DNA breaks (DSBs), has similarities to mechanisms of repair of DSB lesions by cells. Specific recombinases have played and continue to play an important role in molecular evolution and genome shuffling; deregulation in recombination procecess is connected to chromosomal aberrations (inversions, translocations) in cancer. The double-strand-break repair model was put forward by Szostak and collaborators (1983) to explain genetic recombination in yeast. Recent studies (reviewed by Stahl, 1996) have isolated the recombination intermediate molecules predicted by the DSB repair model; in this model, a 5’-3’ exonuclease is responsible for the removal of segments of single strands starting bidirectionally from the DSB followed by invasion, repair synthesis and ligation to give the joint molecule which is then reduced to a pair of duplexes by a Holliday junction resolvase.

The development of mature lymphocytes in mammals results from a complex combination of genetically preprogrammed events and interactions with antigens. Shared in its general mechanisms by both B (bone marrow) and T (thymus) lymphocytes this developmental program involves a series of cell migration gene rearrangements, cell-to-cell contacts, as well as positive and negative selection processes; recombination mechanisms take place at the immunoglobulin and the T cell receptor genes to generate a large number of immunoglobulin genes in different lymphocyte clones. One site-specific recombination event brings together the V and the J segments of the light chain immunoglobulin genes. In the case of the heavy chain genes, one recombination event joins a V to a D segment, sequentially followed in a time frame by the joining of the recombined V-D segment to a J segment. Recent studies have shown that the mechanism of V(D)J recombination is a two-step process involving: (i) site-specific DNA cleavage at the 7mer sequence and at the first nucleotide of the coding sequence, implicating the RAG-1 and RAG-2 proteins which are necessary and sufficient for this step (van Gent et al, 1996); (ii) joining of broken ends in a mechanism similar to the repair of double strand breaks. The murine SCID locus has provided crucial information in the elucidation of the second step in V(D)J recombination: thymocytes in SCID mice are able to catalyze joining of signal ends but display an accumulation of hairpin coding ends (Zhu et al, 1996). The murine SCID locus has been mapped to the gene encoding the catalytic subunit of DNA-dependent protein kinase (DNA-PK) (Kirchgessner et al, 1995).

Group I introns from a variety of organisms contain long open reading frames (ORFs) that encode site-specific DNA endonucleases which promote integration of their DNA into cognate sites via homologous recombination. These endonucleases typically cleave intron-lacking DNA near the site of intron insertion (exon-exon junction) creating a staggered DSB which facilitates intron invasion (intron homing). This mechanism has been demonstrated in mitochondria, chloroplasts and nuclei of eukaryotic cells. I-CreI is a member of this class of molecules that promotes homing of the chloroplast 23S rRNA intron in Chlamydomonas reinhardtii ; I-CreI contains once the LAGLI-DADG motif (whereas other members of the family contain two copies of this motif separated by 90-120 amino acids); this motif is important for the endonuclease activity of the molecule. DNA cleavage by I-CreI requires Mg²⁺ or Mn²⁺ and is inhibited by monovalent cations, has an optimum for catalytic activity of 50-55^oC, is stabilized by DNA and binds to 12 nt on each target strand (Wang et al, 1997).

B. Aberrant recombinations can result in human disease

Mammals carry about 1,000,000 copies of Alu sequences and 10,000 to 100,000 copies of complete and truncated versions of the L1 class of LINEs. Such sequences promote homologous recombination causing translocations of genes and have been hold responsible for a number of human disorders. Alu sequences are found in introns. One type of mutation in the LDL receptor gene responsible for familial hypercholesterolemia has incurred by Alu-Alu recombination deleting several exons and thus producing a truncated receptor molecule with loss of function (Lehrman et al, 1987). De novo insertions of an L1 element into the factor VIII gene can cause hemophilia A in humans (Kazazian et al, 1988).

Foreign DNA transferred to host cells may be rejected (degraded), integrated at random sites by illegitimate recombination, integrated at homologous sites by legitimate recombination, or remain extrachromosomal and replicate autonomously. The homologous recombination between chromosomal DNA and transfected DNA sequences, an event termed "gene targeting," can be used to correct mutated genes in cultured cells.

It has been known for a long time that the translocation of an active gene to the neighborhood of heterochromatin (transcriptionally inert part of the genome) results in silencing of the translocated gene, a process known as "position effect variegation", first described in Drosophila (Lewis, 1950; Wilson et al, 1990).

Chromosomal translocations seem to contribute to tumorigenesis either by activating proto-oncogenes to oncogenes or by inactivating tumor suppressor genes. Mammalian chromosomes contain a number of break-susceptible or fragile sites where breakage can be induced reproducibly by experimental manipulations. Such fragile sites might lie in the neighborhood of transposable elements, hypervariable minisatellites or other DNA structural peculiarities such as Z-DNA, and other hotspots of recombination (reviewed by Haluska et al., 1987).

Nonrandom chromosome rearrangements, observed in a variety of human and animal tumors are associated with the enhanced expression or deregulation of cellular oncogenes. For example, the human c-myc oncogene becomes active following its translocation close to the enhancer sequence within the immunoglobulin heavy chain gene locus (Hayday et al., 1984). The chromosomal translocation (17;19) in acute lymphoblastic leukemia produces a chimeric transcription factor consisting of the amino-terminal portion of the helix-loop-helix proteins E12/E47 fused to the DNA binding and leucine zipper dimerization motifs of the liver-specific protein factor Hlf (Hepatic leukemia factor), normally not expressed in lymphoid cells (Hunger et al., 1992). In pre-B cell acute lymphoblastic leukemia (ALL) the t(1;19) translocation brings together two gene fragments encoding for transcription factors and results in the synthesis of a chimeric transcription factor composed of truncated E2A and Pbx1 (Kamps et al., 1991).

C. Exploitation of recombinases in gene therapy: the Cre/LoxP system

A strategy has recently been developed which facilitates culturing of human cells derived from primary tumors. This method is based on the transient expression of T antigen of SV40 which has been shown to immortalize primary cells of human and murine origin and on the use of bacteriophage recombinase Cre which catalyzes sequence-specific recombination at the LoxP sequence inducing permanent deletion of T antigen cDNA; indeed, if two LoxP sequences were provided as direct repeats the intervening sequence could be deleted during Cre recombination and lost from cells. This is an advantage for primary cell cultures because the continuous expression of SV40 large T antigen may alter the antigenicity of the cells and induce other type of mutations not associated with the original tumor; according to this strategy large T antigen can be expressed in a time-dependent way (Li et al, 1997).

Figure 14 shows the structure of the two retroviral vectors used by Li et al (1997) to facilitate culturing of primary tumor cells and Figure 15 the change in morphology as a result of Cre-Puro retrovirus-induced loss of expression of T antigen in T antigen-immortalized primary cell cultures of mouse breast cancer cells.

Reversible immortalization of primary cells was achieved by Westerman and Leboulch (1996) using retrovirus-mediated transfer of an oncogene that could be subsequently excised by site-specific Cre/LoxP recombination; the FLP/FRT recombination was not efficient in primary cells. Pure populations of cells in which the oncogene was permanently excised were obtained which reverted to their preimmortalized state. Using the Cre/LoxP recombination strategy primary cells could be cultured and expanded; the method was proposed to be applicable for facilitating gene transfer to cells unresponsive to exogenous growth factors.

A retroviral vector, containing both a neomycin resistance expression unit flanked by loxP sites and GM-CSF cDNA, was used to transduce the human hematopoietic K-562 cell line. Superinfection of K562 cell clones with a retrovirus containing a Cre recombinase expression unit and molecular analyses of 30 doubly transduced subclones showed a strict correlation between Cre expression and LoxP-flanked selectable cassette excision; excision of the selectable cassette resulted in a significant increase of GM-CSF transcription driven by the retroviral promoter (Fernex et al, 1997).

Novel retroviral vectors for gene transfer were developed by Bergemann et al (1995) by inserting two LoxP sites into a retroviral vector also containing the HSV-tk gene; Cre expression in cells infected with this vector was followed by BrdU selection for cells in which site-specific recombination took place. Furthermore, replacement of the enhancer/promoter elements in both LTRs by Lox sequences led to the development of retroviral suicide vectors for gene therapy. Vanin et al (1997) have used the Cre/LoxP recombinase system to generate high-titer retroviral producer cell lines; incorporation of LoxP sites at the flanks of a Neo^R-HSV-tk cassette in the proviral DNA allowed excision of these selectable markers through expression of Cre recombinase and the production of a high-titer producer cell line containing a single LoxP site flanked by the viral LTRs. Retransfection of this cell line with a plasmid containing a gene of interest flanked by LoxP sites and the Cre expression vector allowed insertional LoxP/LoxP recombination of the gene into the favorable preexisting site in the genome and the generation of a new line with a titer equivalent to that of the parental producer cell line. The Cre/LoxP recombinase strategy has been used to generate retroviral vectors with the ability to excise themselves after inserting a gene into the genome (Russ et al, 1996).

Bushman and Miller (1997) fused retroviral integrase enzymes to sequence-specific DNA-binding domains and investigated target site selection by the resulting proteins. A fusion protein composed of HIV integrase linked to the DNA-binding domain of l repressor was able to direct selective integration of retroviral cDNA in vitro into target DNA containing l repressor binding sites. A fusion of HIV integrase to the DNA binding domain of the zinc finger protein Zif268 also directed increased integration near Zif268 recognition sites.

Introduction of foreign DNA into cell nuclei with recombinase cDNA and appropriate sequences to promote recombination may promote nor only insertion of a therapeutic gene into a specific chromosomal site but also chromosomal rearrangements that could convert therapeutically transduced cells into malignant. There is a great deal of knowledge to be derived from these very promising strategies of gene therapy before they can be successfully applied to humans.

XIII. Fate of the transgene in the nucleus

A. How to sustain transgene expression?

A major drawback in gene therapy applications is loss in gene activity within a few days from gene transfer although all previous steps were successful. In other words, the transferred gene is transiently expressed for 1-4 days and its expression thereafter declines dramatically. This is due (i) to the degradation of the gene in the nucleus; (ii) the dilution of the plasmid during replication of the cells from its inability to replicate; (iii) its inactivation by position effects from chromatin surroundings after its integration into the chromosomal DNA; (iv) the elimination of the therapeutic cells expressing the transgene by the immune system of the organism either because of the antigenicity of the expressed protein or because of the antigenicity of viral proteins, an effect often associated with adenoviral and retroviral gene delivery.

A number of strategies are being pursued to solve these problems. Sustaining the expression of a transgene into somatic cells for, lets say, 6 months would mean than a gene therapy treatment would need to be repeated twice a year, for example to a hemophilia patient or to a patient who has undergone balloon treatment after coronary heart disease and is being treated via arterial gene transfer.

An approach to sustain expression of the transgene is via episomal replication of the plasmid carrying the transgene for long periods of time, maintaining the plasmid in high copy numbers, and in a form replicating in synchrony with the cell cycle; even better a plasmid can be replicated continuously independently of the cell cycle, an approach to find application in the transfection of nondividing cells by plasmids (which to date is a virtue of adenoviruses, AAV, and HIV-1 vectors; see Table 1).

A way to sustain expression of the transgene could be achieved via targeted integration into one or several different chromosomal locations and the insulation of the transgene from neighboring chromatin domains using special classes of DNA sequences able to act as insulators and maintain independent realms of gene activity (such as matrix-attached regions, MARs). In this case flanking of the foreign gene by two MAR sequences is expected to insulate it against position effect variegation and prevent inactivation of the gene at the chromatin level by chromatin condensation or other mechanisms propagated from the neighboring domains at the integration site (Boulikas, 1995b).

Several studies have shown that linearization of plasmids with restriction enzymes favor highly their integration into the host's genome compared with supercoiled, covalently-closed plasmid DNA. Free ends of DNA are known to promote recombination and a number of nuclear proteins including p53, poly(ADP-ribose) polymerase, ligases I and II, Ku antigen, DNA-dependent protein kinase are known to bind to free ends of DNA, whereas other molecules such as helicases and endonucleases are known to function during repair of lesions in DNA inducing the appearance of strand breaks; especially important in this aspect are members of the RAD50-57 family of proteins involved in recombination and in repair of double-strand breaks.

B. Episomal plasmids for gene transfer

Integration or replication of a foreign gene introduced as a plasmid into mammalian cells is a very rare event; plasmid DNA resides transiently in the nucleus as an episomal, extrachromosomal element for short periods of time after transfection of cells in culture (usually up to one or very few days) during which transcription can take place; after that the episomal DNA is degraded and lost permanently from the cells.

Viral origins of replication have been introduced into the same plasmid as the reporter gene and found to increase the persistence of expression. A polyoma virus-based plasmid containing the polyoma virus origin of replication and the T antigen gene, as well as the neoR gene was maintained extrachromosomally in mouse embryonic stem (ES) cells at 10-30 copies per cell for at least 74 cell generations in the presence of G418 (Gassmann et al, 1995).

Prolonged episomal persistence may be an advantage for gene therapy of nondividing cells. A limited number of studies in gene transfer have used plasmids able to replicate episomally. Most of the plasmids used contain viral origins of replication but also the gene of the replication initiator protein that after its expression in the host will interact with the origin of the plasmid to maintain a relatively high copy number of plasmids which will persist for some time. The advantage using episomal replication of plasmids is enormous in somatic human gene therapy as it can sustain expression of a transgene for a few months after a single injection of the plasmid as compared to the loss of expression after about 1-10 days (maximum at day 2) following injection of nonepisomal plasmids (Zhu et al, 1993). Thierry and coworkers (1995) have succeeded in sustaining the expression of the luciferase reporter gene in mice for up to 3 months after a single intravenous injection of a plasmid including the human papovavirus BKV early region and origin of replication, the large tumor antigen (T antigen) as the replication initiator protein, and the late viral capsid proteins in the same construct harboring the luciferase gene; this plasmid was shown to be replicated extrachromosomally for 2 weeks in the lung.

Episomal replication of a hybrid HSV-1/EBV vector was achieved when the latent oriP of EBV and the EBNA-1 cDNA, which encodes for the replication initiator protein of EBV, were included in the vector (Wang and Vos, 1996).

Expression of viral replication initiator proteins (e.g. T antigen) is oncogenic. Of special interest in human gene therapy is to determine human DNA sequences able to sustain the extrachromosomal replication of plasmids into permissive human cells for longer periods. Such DNA sequences known to act as origins of replication, although poorly understood, have been found in human, monkey, and other mammalian genomes and could be used to sustain the replication of the plasmid thus increasing its copy number in the cell and the time of its persistence (see page 122-123).

To this end, a technology has been developed in our laboratory that permits us to isolate human origins of replication (ORIs) and to include selected ORIs together with the cDNA of the replication initiator protein responsible for activating this particular ORI, in plasmids with therapeutic genes (Boulikas et al, in preparation).

C. Considerations of chromatin structure of plasmids during gene delivery

Almost all supercoiled plasmids used in gene transfer, as produced in bacteria, are under negative supercoiling. Immediately after their import into nuclei plasmids are packaged into nucleosomes that absorb and constrain part or most likely all of the negative supercoils. This is true assuming that no cuts on the DNA are introduced during its passage through the cell membrane barrier to cytoplasmic lysosomes before entering nuclei; if DNA is cut the supercoils on the plasmid will be relaxed. Nicked DNA might be repaired and ligated in nuclei by DNA ligases and be subject to the same constrains as chromosomal DNA. Use of linear plasmids is expected to stimulate recombination during repair of double strand breaks (also would increase degradation of the plasmid in the nucleus and loss of the transgene) ultimately resulting in plasmid integration at variable chromosomal loci, determined to some extend by the nature of the free ends of DNA and the short terminal sequence of the DNA at the ends as well as the type of recombinase molecules in the cell type used.

Treatment of cell cultures with sodium butyrate inducing hyperacetylation of core histones would reverse in part the relieving of the negative torsional strain by the wrapping of the plasmid around histone octamers and will provide DNA in a negatively superhelical of underwound form able to sustain transcription of the template (Schlake et al, 1994).

D. Overcoming the influences of chromosomal surroundings at plasmid integration sites

Use of two MARs each flanking the reporter gene on either side is expected to form a minidomain after integration of the foreign gene into a chromosomal site. MARs potentiate the effect of promoters and enhancers when two MAR elements are placed one upstream and the other downstream from control elements but not between them. MARs will (i) shield reporter genes from the influences of chromosomal surroundings that most often cause inactivation of foreign genes. This effect of chromatin structure on neighboring sequences is known as position effect variegation. Indeed about 85% of the chromosomal sites are transcriptionally inactive assuming that 15% of the genomic DNA is transcribed; however, even integration of a foreign gene into an active chromatin locus may not warrantee its transcriptional activation as other parameters, such as proximity of the integration site to the natural promoter and enhancer elements of the active chromatin domain, or orientation of the integrated gene with respect to the active gene in the chromosomal DNA may determine its level of expression. (ii) MARs will maintain a supercoiled DNA topology within the domain thus increasing the negative supercoiling at local promoter and enhancer sites, a prerequisite for efficient transcription (see Boulikas, 1995b).

XIV. Transfer of reporter genes

A. Transfer of the b-galactosidase (lacZ) reporter gene

Before a gene therapy preclinical study or even gene transfer to cells in culture begins it is essential to test the variables and pinpoint the conditions leading to the success of the operation using reporter gene transfer. LacZ, encoding the b-galactosidase (b-Gal) from E. coli is one of the most commonly used reporter genes. A staining procedure for this enzymatic activity can result in the generation of blue color using X-Gal as a substrate leading to the direct visualization of its activity, for example, in thin sections through animal tissues.

Transfer of the reporter b-galactosidase gene to human liver tumors in nude mice was performed by Wang and Vos (1996) using a hybrid HSV-1/EBV vector which replicates episomally when the latent oriP of EBV and the EBNA-1 cDNA were included.

Many mammalian tissues, especially intestine, kidney,

epididymis, and lung contain endogenous b-Gal, a lysosomal enzyme participating biochemically in glycolipid digestion. Weiss et al (1997) were able to detect mammalian b-Gal activity on histochemical preparations of mouse, rat and baboon lung tissue (Figure 16) and also to distinguish between the endogenous and bacterial b-Gal activity in airway epithelial cells in the transgenic Rosa-26 mouse in based upon the differences in pH optima between the mammalian and bacterial enzymes (Figure 17). Time and temperature of exposure to X-Gal could not be used to distinguish between endogenous and exogenous b-Gal activity; thus, exposure of tissue preparation to pH 8.0-8.5, which minimized detection of the endogenous activity allowed unambiguous discrimination and was the method of choice to detect reporter b-Gal activity.

A phase I study involving six patients with inoperable lung cancer and an endobronchial lesion accessible by bronchoscopy was initiated to evaluate the feasibility, tolerance, and clinical effects using adenoviral delivery of the Escherichia coli lacZ marker gene driven by the RSV promoter; biopsy specimens of the tumor and surrounding mucosa in 5 patients were tested positive for b-galactosidase expression (Tursz et al, 1996).

B. Transfer of the luciferase and green fluorescent protein (GFP) reporter genes

A synthetic polyamino derivative was used by Goldman and coworkers (1997) to transfer the luciferase and b-galactosidase reporter genes in animal models bearing stereotactically implanted human glioma cell xenografts. The luciferase reporter gene was transferred in both newborn and adult rabbit lungs using polyethylenimine (Ferrari et al, 1997).

Thierry and coworkers (1995) have succeeded in sustaining the expression of the luciferase reporter gene in mice for up to three months using episomal vectors (Table 2).

GFP has been used as a reporter molecule for gene expression because it fluoresces green after blue-light excitation. However, many attempts by Hanazono et al (1997) to isolate stable retroviral producer cell clones secreting vectors containing GFP (after transfer of the neoR gene and selection in G418) have failed because stable GFP-clones were undergoing major rearrangements or other mutations which abrogated GFP expression and prevented vector production.

Additional studies using reporter gene transfer are summarized on Table 2.

DIVISION TWO: GENE THERAPY OF CANCER

XV. Cancer immunotherapy and tumor vaccines

A. The molecular basis of cancer immunotherapy

Many human tumors are nonimmunogenic or weakly immunogenic. The immune system, evolved to rid the body of unwanted intruders, could be instructed and reinforced to eliminate cancer cells. Increasing the immunogenicity of tumors by causing local cytokine production or by increase in the expression in MHC antigen can lead to local antitumor effect (Tepper et al, 1989; Fearon et al, 1990). Indeed, immune surveillance is of the major defense mechanisms against cancer; immunosuppressed individuals are more prone to cancer and nude mice, lacking immune response, are exploited in the lab to elicit tumors after injection of tumorigenic cells, a response that, in many cases, cannot be elicited in normal mice.

Cells undergoing malignant transformation are believed to be eliminated from the body by white blood cells including natural killer cells (NK), lymphokine-activated killer cells (LAK), cytotoxic T lymphocytes (CTL), tumor-infiltrating lymphocytes (TIL), and activated macrophages; since established cancers in the human body may escape this potential defense mechanism of immunologic surveillance, cancer patients have been treated with IL-2 to stimulate their cellular immune mechanisms to kill cancer cells; lengthy and complete remissions, however, were at a low rate and complications were encountered by the toxicity caused by the systemic administration of IL-2 (Rosenberg, 1992).

Transfection of the IL-2 gene into human melanoma cells increased cellular immune response (Uchiyama et al, 1993). This and similar approaches have established the foundation of the ex vivo cancer immunotherapy by transfer of autologous (cancer patient’s) cells after transduction in vitro with cytokine genes (see below). The ultimate goal is the activation of tumour-specific T lymphocytes capable of rejecting tumour cells from patients.

B. Cancer immunotherapy with tumor infiltrating lymphocytes (TILs)

Ex vivo approaches in immunotherapy have been aimed at isolating T cells directly from tumors (known as tumor infiltrating lymphocytes or TILs), stimulate TILs to proliferate in cell culture with IL-2 followed by their reintroduction into the blood stream of advanced cancer patients (Rosenberg et al, 1988). The adoptive transfer of TILs was 50-100 times more potent than that of lymphokine-activated killer (LAK) cells isolated from the patient's tumors. Treatment of 20 patients with TILs after their expansion in vitro, plus IL-2, resulted in objective regression of metastatic melanomas in lungs, liver, bone, skin, and subcutaneous sites which lasted for several months (Rosenberg et al, 1988).

TILs were also transfected in vitro with the bacterial neomycin-resistance gene and were reintroduced into patients with advanced cancer in order to follow their persistence in blood circulation with PCR methods (Aebersold et al, 1990). Such “gene marking” clinical protocols using TILs are numbers 1, 3, 9. 13, 57, and 169 in Appendix 1, pages 159-172. Having shown safety in the NeoR-modified TIL protocol, the gene for tumor necrosis factor (TNF) was added to the vector for therapy of malignant melanoma in advanced cancer patients; the first patient began treatment in January 1991. TNF, however, is effective as an anticancer agent in mice at 400 mg/kg body weight, but in humans, TNF is toxic at 8 mg/Kg and so far of no proven therapeutic value at this low concentration (reviewed by Anderson, 1992). In a similar approach, the TNF gene was replaced by the gene of interleukin-2 (IL-2) in order to develop locally high doses of IL-2 at the tumor site by immunization with TIL cells from the patient producing systemic antitumor immunity (Rosenberg et al, 1992).

TNF-a, (also IL-1b, IFN-g, and vitamin D3) after binding to their transmembrane receptors stimulate the production of the second messager ceramide from sphingomyelin in the plasma membrane by activating sphingomyelinase; this results in a cascade of signal transduction events that result in down regulation of c-myc and induction of apoptosis, to terminal differentiation, or to RB-mediated cell cycle arrest (see apoptosis further below).

IL-2 stimulates the differentiation of precursor lymphocytes into LAK cells and further stimulates LAK cell proliferation; LAK cells are probably produced by activation of NK cells or from activated T cells by IL-2. Administration of IL-2 plus amplified LAK cells into mice models led to marked regression of disseminated cancers and leukemia. LAK cells are able to destroy tumor cells that express only weakly histocompatibility antigens. IL-2, however, has several pleiotropic effects: stimulation of B cell proliferation; activation of HLA class II antigen expression on endothelial cells, TILs, and melanoma cells; and enhanced production and release of TNF-a, and IFN-g (see Cassileth et al, 1995 and the references cited therein).

However, the use of large numbers of adoptively transferred, broadly cytotoxic LAK cells in combination with IL-2 has been effective for only small subsets of cancer patients (reviewed by Wiltrout et al, 1995).

TILs, which could potentially kill tumor cells, are found in many tumors but remain suppressed or anergic; this anergy may arise from the absence of lymphokines which provide signals for TIL cell activation and stimulation to proliferation although ligands may be bound to the variable region of the T cell receptor; indeed, nonimmunogenic tumors are rejected by syngeneic mice upon transfection by IL-2 or IL-4 genes; IL-2 lymphokine production by the tumor cells bypasses T helper function in the generation of an antitumor response rendering the tumor cells immunogenic; nontransfected tumors are not rejected by the animal and grow causing its death (Tepper et al, 1989; Fearon et al, 1990).

Ex vivo gene therapy trials using cytokine gene transfer (see below) circumvent the problem of toxicity of IL-2 administration; for non-gene transfer therapies, white blood cells drawn from patients are fractionated, cultured, stimulated with IL-2 or other cytokines, and reintroduced in much higher numbers into the blood of the patient.

C. Cancer immunotherapy with cytokine genes

The combination of immunotherapy with conventional treatments such as radio- and chemotherapy may be necessary to eradicate minimal residual disease. Advanced therapies involve the transfection of lymphocytes in culture with cytokine genes followed by selection of the successfully transfected cells with a selectable marker such as the bacterial neomycin-resistance gene (Cassileth et al, 1995). Numerous phase I clinical trials employing either syngeneic genetically modified or allogenic tumor vaccines are in progress (see immunotherapy in Appendix 1, page 159-172). The development of tumor cells transduced with cytokine genes and their exploitation as tumor vaccines in patients with cancer is a very promising field (reviewed by Jaffee and Pardoll, 1997).

Cytokine genes used for cancer immunotherapy include those of IL-2, IL-4, IL-7, IL-12, IFNs, GM-CSF, TNF-a in combination with genes encoding co-stimulatory molecules, such as B7-I. The major goal of the use of immunostimulatory cytokines is the activation of tumour-specific T lymphocytes capable of rejecting tumour cells from patients with low tumour burden or to protect patients from a recurrence of the disease. As distant metastasis is the major cause for therapeutic failures in clinical oncology, treatment of patients having a low tumor volume with immunotherapy could protect the patient from recurrence of disease. Treatment of rodent tumor models with little or no intrinsic immunogenicity with this approach resulted in regression of preexisting tumors and cure of the animals from their disease; furthermore, in some instances cured animals had retained immunological memory and resisted a second challenge with the parental tumor cells (reviewed by Gilboa, 1996; Mackensen et al, 1997).

The transduction of the tumor cells of the patient with cytokine genes ex vivo and the development of tumor vaccines depends on the establishment of primary cell culture from the solid tumor. Although malignant melanomas are easy to culture, it is difficult to establish cell lines from most other primary human tumors using convenient methods; primary tumor cultures are being used (i) for the transduction of autologous cells from the cancer patient with cytokine genes to develop cancer vaccines after intradermal implantation to the patient; (ii) for characterization of tumor-specific cytotoxic T lymphocytes in order to identify specific antigens on the human primary culture; (iii) for extensive phenotypic characterization of the tumor in cell culture. The Cre/LoxP system (see recombinases in gene therapy) has been used to facilitate the establishment of primary cell lines from human tumors (Li et al, 1997).

Human gene therapy protocols 3 and 10 (Appendix 1) use immunization of cancer patients with autologous cancer cells transduced with the gene for tumor necrosis factor (TNF).

D. Cancer immunotherapy with the IL-2 gene

Active immunization with pancreatic tumor cells genetically engineered to secrete IL-2 were shown to inhibit pancreatic tumor growth in vivo; this was shown using a poorly immunogenic subcutaneous model of murine ductal pancreatic cancer by implanting tumor Panc02 cells in C57BL/6 mice; whereas 90% of animals vaccinated with irradiated parental Panc02 and subsequently challenged with parental Panc02 cells developed tumors by 48 days only 40% of animals vaccinated with irradiated Panc02 cells engineered to secrete IL-2 and challenged with parental Panc02 cells developed tumors by 48 days (Clary et al, 1997).

According to a RAC-approved clinical protocol the gene for human interleukin-2 (IL-2) was transduced into a cell line established from the neoplastic cells of a patient with malignant melanoma; this procedure established an IL-2-secreting cell line with integration of the IL-2 gene into genomic DNA. The IL-2-secreting cells were irradiated, in a manner sufficient to inactivate 100% of the cells but insufficient to completely inhibit IL-2 synthesis, and administered to 12 patients with metastatic malignant melanoma in a Phase I toxicity study. These cells have the capacity to induce an antimelanoma response as shown in animal studies (Das Gupta et al, 1997).

A significant number of RAC-approved clinical protocols use IL-2 cDNA transfer. These include protocols 11, 16, 19, 20, 46, 48, 50, 61, 71, 102, and 135, in Appendix 1 and protocols 190, 197, 198, 200, 204, 211, 213, 215, and 219 using cationic lipids for gene transfer (Table 4 in Martin and Boulikas, 1998, this volume, pages 203-206).

E. Cancer immunotherapy with the IL-3 gene

IL-3 was found to enhance the development of cytotoxic T lymphocytes; during antitumor response, macrophages could ingest whole tumor cells, cell fragments, or heat shock proteins complexed to antigenic peptides and then process the tumor antigens for presentation; IL-3 stimulated antigen-presenting cells (APCs), which are macrophage-like, within the tumor leading to generation of cytotoxic T lymphocytes (CTLs). This constitutes a plausible pathway for enhancement in tumor rejection by IL-3 stimulation (Pulaski et al, 1996).

IL-3 signaling proceeding either via the JAK-STAT or the Ras-Raf pathways, stimulates a number of genes such as the DUB-1 encoding a deubiquitinating enzyme the overexpression of which leads to G1 arrest (Zhu et al, 1996); deubiquitination might be an additional mechanism to couple extracellular signaling to cell growth. IL-3 signaling leads to stimulation in myeloid cell proliferation.

F. Cancer immunotherapy with the IL-7 gene

Primary cell cultures from 45 patients with malignant melanoma were transfected via electroporation with the gene encoding for human interleukin-7 (IL-7) resulting in the production of biologically active IL-7 without altering the expression of HLA class I and II, ICAM-1, and of a melanoma-associated antigen. Irradiation of the transfected cells with 10,000 cGy, which arrested tumor cell growth in vitro, did not affect the ability of the cells to secrete IL-7 in the culture medium; this approach, which does not use retroviruses, could be applicable in vaccination protocols for melanoma patients (Finke et al, 1997).

Transfer of the IL-7 cDNA for cancer immunotherapy is being used in a human clinical trial (protocol 70, Appendix 1).

G. Cancer immunotherapy with the IL-12 gene

IL-12 gene therapy is one of the more novel and promising approaches in cancer therapy. IL-12 is a heterodimeric cytokine composed of two subunits, p40 and p35, that requires the simultaneous expression of both the p35 and p40 chain genes from the same cell for production of biologically active IL-12. Coordinate expression of the IL-12 p40 and p35 genes in several solid tumor models has been found to induce strong and specific antitumor immune responses. A variety of biological functions have been attributed to IL-12 including the induction of IFN-g and the promotion of predominantly Th1-type immune responses to antigens (Tahara et al, 1996).

The local secretion of IL-12 achieved by gene transduction suppressed tumor growth and promoted the acquisition of specific antitumor immunity in mice. This was shown by intradermal inoculation of mice with NIH3T3 cells transduced with expression plasmids or a retroviral vector expressing the murine IL-12 gene admixed with murine melanoma BL-6 cells; CD4⁺ and CD8⁺ T cells, as well as NK cells, were responsible for the observed antitumor effects resulting from IL-12 paracrine secretion. Transduction of tumor cells with B7.1 gene enhanced the antitumor immune response (Tahara et al, 1996).

The antitumor effect of several transgene expression plasmids encoding the cytokines IL-2, IL-4, IL-6, IL-12, IFN-g, TNF-a, and GM-CSF was tested using the gene gun-mediated DNA delivery into the epidermis overlying an established intradermal murine tumor; this study showed that IL-12 gene therapy was much more effective than treatment with any other tested cytokine gene for induction of tumor regression as determined from the increased CD8⁺ T cell-mediated cytolytic activity in the draining lymph nodes of tumor-bearing mice; treated animals were able to eradicate not only the treated but also the untreated solid tumors at distant sites; elevated systemic levels of IFN-g, were found after IL-12 gene therapy. This approach is providing a safer alternative to IL-12 protein therapy for clinical treatment of cancers (Rakhmilevich et al, 1997).

Lieu et al (1997) have evaluated three IL-12 retroviral vector designs for their level of IL-12 expression in leukemia/lymphoma cells; these retroviral vectors were based on the murine stem cell virus (MSCV) which efficiently transduces functional genes into normal hematopoietic cells. MSCVpac-mlL-12 and MIPV-mIL-12 contained an encephalomyocarditis virus internal ribosome entry site for internal translation of bicistronic mRNA transcripts, while MDCVpac-mIL-12 carried an expression cassette in the U3 region of the 3' LTR. The MSCVpac-mIL-12 vector was more efficient and directed robust expression of both p40 and p35 IL-12 genes in several murine tumor cell lines of hematopoietic origin, including a T-cell lymphoma, a B-cell lymphoma, and a plasmacytoma/myeloma.

Adenoviral delivery of the IL-12 gene was effective against breast tumors (Bramson et al, 1996) or metastatic colon carcinoma (Caruso et al, 1996) in animal models: mice bearing breast tumors, injected intratumorally with a single dose of an adenovirus expressing IL-12 showed regressions in greater than 75% of the treated tumors; this effect was accompanied with a maximum expression of IL-12 within the tumor between 24 and 72 hr post-injection which lasted for 9 days and an elevation in IFN-g within the tumor; local production of IL-12 also stimulated IFN-g production in tumor-draining lymph node cells (Bramson et al, 1996). Whereas intratumoral adenoviral transfer of the HSV-tk and the murine IL-2 genes resulted in substantial hepatic tumor regression, induced an effective systemic antitumoral immunity in the host and prolonged the median survival time of the treated animals from 22 to 35 days a recombinant adenovirus expressing the murine IL-12 gene was much more effective: intratumoral administration of the IL-12 vector alone increased significantly survival time of the animals; 25% of the treated animals lived over 70 days (Caruso et al, 1996).

The immunological host response to syngeneic murine mammary carcinoma cell line variants, genetically modified to express B7-1 or secrete GM-CSF and IL-12, was examined by Aruga et al (1997). The mammary adenocarcinoma MT-901 subline was weakly immunogenic by immunization/challenge experiments and induced tumor-specific T-cell responses in lymph nodes draining progressive subcutaneous tumors; however, tumor clones from this cell line expressing B7-1 or secreting GM-CSF exhibited reduced tumorigenicity and resulted in significantly enhanced T-cell reactivity to tumor-draining lymph node (TDLN) cells as compared to wild-type TDLN cells. In contrast, transduction with the IL-12 gene led to complete tumor growth inhibition.

An adenovirus vector, AdIL12-B7-1, encoding the two IL-12 subunits in early region 1 (E1) and the B7-1 gene in E3 of adenovirus under control of the murine CMV promoter was used to treat mice tumors derived from a transgenic mouse mammary adenocarcinoma. A single intratumoral injection with a low dose (2.5 x10⁷ pfu/mouse) mediated complete regression in 70% of treated animals whereas a similar dose of recombinant virus encoding IL-12 or B7-1 alone resulted in only a delay in tumor growth. Coinjection of two different viruses expressing either IL-12 or B7-1 induced complete tumor regression in only 30% of animals treated (Putzer et al, 1997).

Human peripheral blood lymphocytes (HuPBLs), injected s.c. in mixture with human lung tumor cells into severe combined immunodeficient (SCID) mice, engrafted and displayed antitumor cytotoxic activity; this antitumor activity was dependent upon both CD8⁺ T cells and CD56⁺ natural killer cells in the donor HuPBLs. IL-12 enhanced the human peripheral blood lymphocyte-mediated tumor suppression; this implies that transfer of the IL-12 gene has a prospect in this type of immunotherapy. This could be significant under the light of studies showing that PBLs isolated from a lung cancer patient also suppressed the growth of the patient's (autologous) tumor when coinjected s.c. with the tumor cells into SCID mice (Iwanuma et al, 1997).

Tumor cell vaccines were transduced with IL-12 or IL-2 genes and the antitumor response induced in mice bearing lung metastases of the BALB/c colon carcinoma C51 were compared by Rodolfo et al (1996). The cells used for transduction with the IL-12 or IL-2 genes were the histologically related, and antigenically cross-reacting C26 tumor cells which were irradiated and injected s.c. Vaccination with C26/IL12 cells cured 40% of mice, while vaccination with C26/IL2 cells reduced the number of metastatic nodules without affecting survival; both cell vaccination regimens showed similar antitumor CTL activation in mice. Both treatments induced antibodies directed against tumor-associated antigens, but only sera from mice treated with C26/IL12 contained antibodies that lysed tumor cells. The better therapeutic efficacy of vaccination with C26/IL12 was found to be associated, among other factors, with an early infiltration of the metastatic lungs by activated T lymphocytes (Rodolfo et al, 1996).

Transfer of the IL-12 cDNA for cancer immunotherapy is being used in human clinical trials (protocols 62, 111, 180, and 183, Appendix 1).

H. Adoptive immunotherapy with GM-CSF

1. Cell culture experiments

The human hematopoietic growth factor, granulocyte-macrophage colony-stimulating factor (GM-CSF), is important in the management and gene therapy of a variety of malignant disorders of the human hematopoietic system. Infection of COS-1 monkey kidney cells with a recombinant AAV vector containing the GM-CSF gene resulted in the release of recombinant GM-CSF protein into the supernatant; the released GM-CSF was able to sustain the active proliferation of the GM-CSF-dependent human megakaryocytic leukemia cell line, M07e, (Luo et al, 1995).

2. Animal studies

The Dunning rat R3327-MatLyLu prostate tumor model (an anaplastic androgen-dependent, nonimmunogenic tumor that metastasizes to the lymph nodes and the lung) has been used for GM-CSF therapy; IL-2- or GM-CSF-secreting human tumor cell preparations (tumor vaccines) were used for the treatment of advanced human prostate cancer in rats. All animals with subcutaneously established tumors were cured; the cancer vaccine induced immunological memory that protected the animals from subsequent tumor challenge; GM-CSF was less effective than IL-2 (Vieweg et al, 1994). Using the Dunning rat prostate carcinoma model, animals with hormone refractory prostate cancer treated with irradiated prostate cancer cells genetically engineered to secrete human GM-CSF showed longer disease-free survival compared to untreated control rats.

To further test the clinical feasibility of the prostate cancer cell vaccine, cancer cells from patients with stage T2 prostate cancer undergoing radical prostatectomy were successfully transduced with MFG-GM-CSF, achieving a significant human GM-CSF secretion in each of 10 consecutive cases (Sanda et al, 1994).

Continuous secretion of GM-CSF and activation of macrophages may contribute to the antitumor effects of a recombinant vaccinia virus expressing the gene for murine GM-CSF injected to solid melanoma tumors twice weekly for 3 weeks; this injection regimen resulted in growth inhibition of the subcutaneous tumor and enhanced the survival of the animals (Ju et al, 1997).

A recent effort has been toward potentiation of T-lymphocyte-mediated antitumor effects. T-lymphocyte response incapacitation in the murine renal cancer model could arise from an impairment of critical nuclear transcription factors. A vaccine-oriented gene therapy approach used T cells and antigen-presenting dendritic cells which were recruited through the use of antigen, chemokines and GM-CSF and further potentiated by fibroblasts expressing IL-2, IL-4, IL-7, or IL-12; the goal of this approach was to optimize MHC class I- and class II-dependent pathways for induction of T-lymphocyte-mediated responses to cancer in animal models (Wiltrout et al, 1995).

Chen et al (1996) found that adenoviral delivery of a combination of HSV-tk, mouse IL-2, and mouse GM-CSF is much more effective for the treatment of metastatic colon carcinoma in the mouse liver than HSV-tk alone or HSV-tk combined only with IL-2; a fraction of the animals developed long-term antitumor immunity and survived for more than 4 months without tumor recurrence in the three gene combination regimen; thus, local expression of GM-CSF in the hepatic tumors and prolonged IL-2 expression were necessary to generate persistent antitumor immunity.

A gene gun device was used to accelerate and introduce gold particles coated with GM-CSF cDNA plasmids into mouse and human tumor cells. Transfected and irradiated murine B16 melanoma cells produced about 100 ng/ml murine GM-CSF/million cells per 24 hr in vitro for at least 10 days. Toward development of a tumor vaccine, irradiated B16 tumor cells expressing murine GM-CSF cDNA were then injected into mice. Subsequent challenge of these mice with nonirradiated, nontransfected B16 tumor cells showed that 58% of the animals were protected from the tumor by the prior vaccine treatment compared to only 2% of control animals inoculated with irradiated B16 cells transfected with the luciferase gene (Mahvi et al, 1996).

Human tumor tissue transfected within 4 hr of surgery produced significant levels of transgenic human GM-CSF protein in vitro. Human GM-CSF was readily detectable in serum and at the injection site following subcutaneous implantation of these transfected tumor cells into nude mice (Mahvi et al, 1996).

The autocrine secretion of GM-CSF by transduced tumor cells was found to serve as an effective immune adjuvant in the host response to a weakly immunogenic murine mammary carcinoma tumor: transfer of activated lymph node cells derived from mice inoculated with GM-CSF-secreting (240 ng/million cells/24 hours) murine mammary carcinoma cells resulted in the prolonged survival of animals with macroscopic metastatic disease; this was not evident utilizing lymph node cells from mice inoculated with wild-type tumor (Aruga et al, 1997).

3. Clinical trials

Autologous cells (sensitized T cells) genetically-modified to secrete GM-CSF have been used for adoptive immunotherapy on humans. GM-CSF has been used for the treatment of advanced melanoma or renal cell cancers (Chang et al, 1996). The steps included retrieval of tumor from the patient for use as a vaccine; the tumor cell line was transduced with a retroviral/GM-CSF vector; cells were reintroduced into the patient (tumor vaccination). Removal of draining lymph nodes after 7-10 days and activation of lymph node cells with a monoclonal antibody directed against CD3 and expansion of the cell population with IL-2 gave anti-CD3⁺/IL2-activated cells which were exquisitely tumor-specific and mediated the regression of established tumors in animal models (Figure 18).

According to a phase I clinical trial cancer patients are intradermally vaccinated with lethally-irradiated tumor cells that have been transfected by particle-mediated gene transfer with gold particles coated with human GM-CSF plasmid DNA; this is based on preclinical studies showing that vaccination of mice with irradiated, GM-CSF-transfected melanoma cells provided protection from subsequent challenges with non-irradiated, non-transfected tumor cells. Human tumor immunotherapy studies in course use patients' fresh specimens of melanoma or renal carcinoma; cells are dissociated, lethally-irradiated and transfected with GM-CSF plasmid DNA-coated gold particles resulting in the subsequent production of biologically active GM-CSF protein by the patient’s cells. Patient’s cells are used intradermally as a vaccine to elicit anti-tumor immune responses. Surgical excision of the vaccination sites will assess GM-CSF production and infiltration of immune effector cells; patients are being subjected to an intradermal injection in their opposite extremity of 5 million irradiated cryopreserved tumor cells taken from the patient at the time of vaccine preparation to asses immune reactions (DTH testing); if a positive reaction is noted on day 28 the DTH site will be surgically removed (Mahvi et al, 1997).

Figure 18. A clinical protocol for adoptive immunotherapy of advanced melanoma patients. Adapted from Chang et al (1996).

A number of RAC-approved human gene therapy protocols use GM-CSF cDNA transfer. These are protocols 35, 53, 63, 113, 149, 150, 162, and 181 in Appendix 1.

I. Cancer immunotherapy with the IFN-g gene

Solid tumors in nude mice have been successfully eradicated with treatment with tumor cell lines stably transfected with an IFN gene. A number of human tumor cell lines including 293, HeLa, K562, and Eskol (a malignant immunoblastic lymphoma) were infected with a rAAV carrying a synthetic type I interferon gene and the bacterial neomycin-resistant gene and geneticin-resistant cells were selected; when injected into nude mice, 293, K562, and Eskol cells failed to form tumors for a duration of up to 3 months; on the contrary, mice receiving nontransduced cells developed tumors within 7 to 10 days; in addition, treatment of an established Eskol tumor with transduced 293 cells resulted in tumor regression (Zhang et al, 1996).

Three RAC-approved human gene therapy protocols use IFN-g cDNA transfer. These are protocols 36, 54, and

71 in Appendix 1.

J. Immunotherapy with synthetic tumor peptide vaccines

Progress in the identification of tumor-specific antigens, that is proteins expressed at high levels by a specific tumor cell type such as prostate or breast cancer, most of which are surface glycoproteins easily recognizable by the immune system, as well as the deciphering of the mechanisms for enhancing the response of cytotoxic T cell lymphocytes have advanced the potential for developing cancer vaccines.

Cancer immunotherapies based on synthetic tumor peptide vaccines have been developed. Tumor-specific CD8⁺ cytotoxic T lymphocytes (CTLs) recognize short peptide epitopes presented by MHC class I molecules that are expressed on the surface of cancer cells. Bone marrow-derived dendritic cells, grown in vitro in media containing combinations of GM-CSF + IL-4, when pulsed with synthetic tumor peptides (which are loaded on the surface of the dendritic cells) became potent antigen-presenting cells (APCs) capable of generating a protective antitumor immune response. Injection of these cells into naive mice protected the mice against a subsequent lethal tumor challenge; in addition, treatment of mice bearing C3 sarcoma or 3LL lung carcinoma tumors with the same type of cells resulted in sustained tumor regression in over 80% of the animals (Mayordomo et al, 1995).

One of the obstacles of this method has been the difficulty in obtaining sufficient numbers of APCs; dendritic APCs have been isolated from CD34⁺ hematopoietic progenitor cells drawn from cord blood and expanded in cell culture in the presence of GM-CSF and TNF-a; TNF-a inhibits the differentiation of dendritic cells into granulocytes. Human peripheral blood mononuclear cells or mouse bone marrow cells depleted of lymphocytes could also yield dendritic cells when cultured in the presence of GM-CSF + IL-4 (Mayordomo et al, 1995).

K. DNA vaccines

Vaccines may be one of the first successful applications of foreign genes into mammalian cells under control of heterologous promoters and enhancers (Felgner and Rhodes, 1991; Thompson, 1992; Gilboa and Smith, 1994). Vaccination with DNA has been shown to be a promising approach for immunization against a variety of infectious diseases (Wang et al, 1993; Michel et al, 1995; Huygen et al, 1996; Kuhober et al, 1996). The method consists in introducing the gene of a viral or bacterial antigen which is uptaken and expressed by the host’s cells to elicit an antigen-specific immune response. DNA coding for an antigen can be directly injected into muscle or skin and stimulate an immune response against the expressed antigen; the gene can either code for surface molecules, which are often used for conventional peptide vaccines, or from internal microbial proteins.

During this approach the antigens are produced intracellularly where they are correctly folded and can be presented to the immune system to stimulate cytotoxic T cells; the method is safe and simple and has shown promising results on animals (reviewed by Moelling, 1997). For example, mice injected intramuscularly with an HIV-1 envelope DNA construct developed anti-HIV envelope immune responses (Wang et al, 1993); intramuscular injection of plasmid DNA expression vectors encoding the three envelope proteins of the hepatitis B virus (HBV) induced humoral responses in C57BL/6 mice specific to several antigenic determinants of the viral envelope (Michel et al, 1995). Immunization of mice with plasmid DNA constructs encoding one of the secreted components of Mycobacterium tuberculosis, antigen 85 gene induced substantial humoral and cell-mediated immune responses (Huygen et al, 1996).

Because immunization of cancer patients with tumor antigen proteins is a very promising approach used extensively in cancer therapy (e.g. Karanikas et al, 1997) many of these approaches could be transferred to the DNA level using the gene encoding the tumor antigen.

As an extension, this method could find application using human tumor antigen genes rather than bacterial/viral antigen genes, that is genes encoding for proteins expressed in tumor but not in normal cells leading to development of tumor vaccines (Graham et al, 1996; Okamoto et al, 1997); this method mimics the infection of the cell in the host by a pathogenic virus resulting in the intracellular processing of the viral proteins and their presentation on the cell surface. Human tumor antigens are, however, weakly immunogenic compared to microbial antigens a problem connected with polymorphism in the major histocompatibility complex proteins of the host and in antigen presentation.

Development of a fusigenic viral liposome vector was made possible using the HVJ (hemagglutinating virus of Japan, a Sendai virus) renowned for its cell fusion ability; plasmid DNA containing the human tumor antigen genes MAGE-1 and MAGE-3 was mixed with HMG-1 nonhistone protein (to increase nuclear import and expression of the plasmid after transfection) and was encapsulated into anionic liposomes (phosphatidylserine, phosphatidylcholine, cholesterol) followed by the addition of inactivated HVJ; intramuscular injection into mice resulted in production of MAGE-1 and -3 IgG antibodies (Okamoto et al, 1997).

XVI. Gene therapy of cancer and candidate genes

A. Mechanisms of carcinogenesis

Whereas for inborn errors of metabolism transfer of a single gene can correct the disorder, cancer is a complex disease involving mutations in a number of proto-oncogenes and tumor suppressor genes as well as an imbalance and disarray in phosphorylation events and regulatory circuits of the cell cycle. As a result of transformation, tumor cells acquire a proliferation advantage compared with normal cells, most of which are quiescent in the adult organism; cancer cells acquire partial independence from regulatory signals from neighboring cells for restricted cell growth. A crucial step in cancer development is the nonelimination of pre-cancer cells by apoptosis (usually a subsequence of a mutation in the p53 gene); such cells acquire a number of unrepaired damage in their DNA, such as strand breaks, which induce chromosomal translocations and result in clonal expansion of this cell population.

Tumor cells are able to survive after DNA damage, and display an increase in mutation rate; cancer cell populations are heterogenous with respect to translocations, loss of heterozygosity, point mutations and transpositions in various genes. Whenever the mutated cell acquires an advantage for rapid growth over other cells in the tumor mass, escaping cell cycle checkpoints, it may replace the original population, a phenomenon known as tumor progression; this may lead to appearance of a more malignant phenotype. As a result, tumor cells are of different genotypes and clones obtained from the same solid tumor may differ in the level of malignancy.

A number of candidate genes, when become mutated or overexpressed, may lead to tumor phenotype: p53, RB, and p21 appear to be the most important. The deregulation of other genes is connected to tumor progression whereas different groups of genes are associated with tumor cell metastasis. These facts make a single gene transfer approach to tumor cell mass to inhibit its growth or change its phenotype from malignant to normal very challenging.

B. Human clinical trials

The genes used for cancer gene therapy in human clinical trials include a number of tumor suppressor genes (p53, RB, BRCA1, E1A), antisense oncogenes (antisense c-fos, c-myc, K-ras), suicide genes (HSV-tk, in combination with ganciclovir, cytosine deaminase in combination with 5-fluorocytosine) which have been very effective in eradicating solid tumors in animals. Also the cytokine genes (IL-2, IL-7, IFN-g, GM-CSF) are being used for the ex vivo treatment of cancer cells isolated from human patients and are able to elicit an immunologic regression especially on immunoresponsive malignancies (melanomas, colorectal carcinomas, renal cell carcinomas) (Culver, 1996). Future directions might be toward use of genes involved in the control of tumor progression and metastasis. Discovery of new genes which are over- or under-expressed during transformation and metastasis is a promising approach for the identification of novel gene targets in cancer gene therapy (Georgiev et al, 1998, this volume).

Diseases amenable to therapy with gene transfer in clinical trials (Appendix 1 and Table 4 in Martin and Boulikas, this volume) include cancer (melanoma, breast, lymphoma, head and neck, ovarian, colon, prostate, brain, chronic myelogenous leukemia, non-small cell lung, lung adenocarcinoma, colorectal, neuroblastoma, glioma, glioblastoma, astrocytoma, and others), AIDS, cystic fibrosis, adenosine deaminase deficiency, cardiovascular diseases (restenosis, familial hypercholesterolemia, peripheral artery disease), Gaucher disease, Hunter syndrome, chronic granulomatous disease, PNP deficiency, a1-antitrypsin deficiency, leukocyte adherence deficiency, partial ornithine transcarbamylase deficiency, Cubital Tunnel syndrome, Canavan disease and rheumatoid arthritis. Several RAC-approved protocols use gene marking rather than gene therapy . An important number of protocols in cancer use ex vivo immunotherapy (Appendix 1, pages 159-172 & 203-206).

XVII. Gene therapy strategies based on p53

A. p53 as a tumor suppressor protein

The p53 has been a fascinating subject in cancer biology since its discovery (Lane and Crawford, 1979; Linzer and Levine, 1979). Originally assigned in the constellation of oncogenes was later shown to exert suppressive effects on cell growth (Finlay et al, 1989); indeed, the mutated p53 has many characteristics of an oncogene (Will and Deppert, 1998, this volume). Mutations in the p53 gene contribute to the emergence of the malignant phenotype (Diller et al., 1990; Baker et al., 1990). Alterations in the p53 tumor suppressor gene appear to be involved, directly or indirectly, in the majority of human malignancies (Vogelstein, 1990). For example, human lung cancer cell lines and specimens showed allelic loss for chromosome regions 3p and 17p (p53 is assigned to 17p13); these specimens displayed homozygous deletions of p53, DNA rearrangements involving the p53 gene, or expression of truncated p53 transcripts suggesting abnormal splicing, initiation, and termination arising from point or other mutations (Takahashi et al, 1989; Nigro et al, 1989).

An interesting approach to unravel the molecular mechanism of action of p53 in restricting cell growth and in inducing apoptosis was the cloning of genes induced by p53 before the onset of apoptosis; this led to the identification of a group of 14 genes (out of 7,202 transcripts examined) which were markedly increased in p53-expressing cells compared with control cells many of which were predicted to encode proteins that could generate oxidative stress or respond to oxidative stress (Polyak et al, 1997). Additional studies in this line have suggested that the induction of the apoptotic pathway by p53 involves (i) transcriptional induction of redox-related genes; (ii) formation of reactive oxygen species; and (iii) the oxidative degradation of mitochondrial components (Polyak et al, 1997).

p53 can inhibit transformation of rat embryo fibroblasts mediated by adenovirus E1A plus activated ras and can also suppress focus formation mediated by myc plus activated ras (Finlay et al, 1989; Eliyahu et al, 1989). Both alleles of p53 need to be mutated or altered for transformation. Introduction of a null mutation by homologous recombination in murine embryonic stem cells gave mice which appeared normal but were susceptible to a variety of neoplasms by 6 months of age (Donehower et al, 1992).

The tumor suppressive activity of p53 seems to involve at least six independent pathways: (i) induction by p53 of the p21/Waf-1/Cip-1 gene which causes growth arrest both via inhibition of cyclin-dependent kinases and via inactivation of PCNA; PCNA is the accessory molecule to DNA polymerases a and d and its absence causes arrest of DNA synthesis at the replication fork; (ii) induction of the death-promoting bax gene by p53 as a mechanism which eliminates oncogenic virus-infected and transformed cells; (iii) by a direct interaction of p53 with origins or replication preventing firing and initiation of DNA replication; (iv) via binding of p53 to a number of important molecules involved in transcription (TATA box-binding protein or TBP, TFIIH); (v) by the role of p53 in DNA repair via its patrolling the genome for small insertion deletion mismatches or free ends of DNA; (vi) p53 is able to attract RPA, an accessory to DNA polymerases a and d as well as TFIIH and RAD51 at the damaged DNA sites; TFIIH, RAD51, and RPA have a demonstrated role in DNA repair (Figure 19). Additional properties of p53 include the induction of Gadd45 involved in the arrest of the cell cycle and induction of Mdm2 which, after exceeding a threshold value in the cell associates with p53 to restrict its regulatory functions; thus, Mdm2 acts as a feedback loop for p53 to moderate its apoptotic and cell cycle restrictive functions (Figure 20).

B. Genes regulated by wild-type p53

Protein p53 appears to be a transcription factor able to recognize specific regulatory regions in a number of genes via its central DNA-binding domain; the DNA sequence-specific binding of wt p53 is regulated by the C-terminal domain of p53 and is activated by a variety of posttranslational modifications (Hupp et al, 1992). p53 is phosphorylated and is constitutively expressed at low levels in most normal tissues (Lane and Crawford, 1979; Linzer and Levine, 1979).

The sequence specificity of p53 has been determined using random synthetic oligonucleotides followed by selection by wtp53 and cloning; these studies revealed the 10 bp motif RRRCGYYY (where R is purine and Y pyrimidine) as the binding and recognition site of wtp53 recognition (El-Deiry et al., 1992); two such 10 bp motifs are required for p53 binding separated by up to 13 bp of random sequence. Since the 10 bp motif is a palindrome, the binding site of p53 comprises 4 copies of the half binding sites GYYY oriented in opposite directions, which suggested that p53 binds either as a dimer to two cruciforms or as a tetramer with each subunit interacting with one half site. The second possibility is favored since

biophysical studies indicate that p53 exists as a tetramer in solution (Stenger et al., 1992).

Increased levels of p53 upregulate the expression of specific genes including Cip-1/Waf-1/p21 (El-Deiry et al, 1993), GADD45 (Kastan et al, 1992), cyclin G (Okamoto and Beach, 1994), and mdm2 (Perry et al, 1993; Barak et al, 1993; Momand et al, 1992) which is induced by UV damage in a p53-dependent pathway (Perry et al, 1993). Gadd45 inhibits cell cycle progression (Papathanasiou et al, 1991).

Mdm2 acts as a feedback loop for the biological functions of p53 apparently to moderate the G1/S arrest or apoptosis triggered by p53 following severe damage to DNA. Mdm2 protein associates with p53 causing p53 inactivation by preventing its sequence-specific binding to regulatory targets in DNA (Momand et al, 1992; Oliner et al, 1992). Elevated levels of Mdm2 mimic the effect of T antigen, E1B of adenovirus, E6 of HPV, which also inactivate p53 in a similar manner; overexpression of Mdm2 can block the induction of apoptosis by p53 (Chen et al, 1994).

Additional genes up-regulated by p53 include human PCNA (Shivakumar et al, 1995), mouse muscle creatine kinase MCK (Zambetti et al, 1992), EGFR (Deb et al, 1994), the potent promoter of the death pathway Bax (Miyashita and Reed, 1995), and thrombospondin-1 (Dameron et al, 1994). Other cellular regulatory regions that interact with p53 include the RGC repeats in the ribosomal gene cluster (Farmer et al, 1992; Kern et al, 1992).

The PCNA promoter is up-regulated in the presence of moderate amounts of wt p53; however, at higher levels of wt p53 the PCNA promoter is inhibited whereas tumor-derived p53 mutants activate the PCNA promoter (Shivakumar et al, 1995); it has been suggested that the moderate elevation in wt p53 seen after DNA damage induces PCNA to cope with its DNA repair activities (Shivakumar et al, 1995); this inhibition in DNA replication but stimulation in repair by p53 might be accomplished by an independent pathway involving induction of p21 (El-Deiry et al, 1993) which interacts with PCNA protein auxiliary to DNA polymerase d to inhibit the replication but not the repair functions of PCNA (Li et al, 1994).

The bax gene which induces apoptosis (Figure 21) is upregulated by p53 whereas the bcl-2 gene which inhibits apoptosis in B cells is down-regulated by p53 (Miyashita et al, 1994a,b; Miyashita and Reed, 1995). Initiated cancer cells may lead to tumor development only when a dysfunction in their apoptotic pathway takes place; some of the mechanisms leading to inactivation of the apoptotic pathway in cancer cells may result from an up-regulation in the bcl-2 gene (a Bcl-2 chimeric factor is produced in leukemias as a result of a translocation) or down-regulation of the bax gene. Gene therapy for cancer could involve restoration of the apoptotic pathway in cancer cells leading to their suicidal death (see below).

Figure 21. Involvement of Bax and Bcl-2 proteins in apoptosis. Bax is a potent inducer of apoptosis; binding of Bcl-2 to Bax (also binding of the E1B 19 kDa protein of adenovirus to Bax) prevents Bax from its apoptotic functions. From Boulikas T (1997) Gene therapy of prostate cancer: p53, suicidal genes, and other targets. Anticancer Res 17, 1471-1506. With the kind permission of Anticancer Research.

Binding sites for p53 have been found at the origin of replication of polyomavirus with an inhibitory effect on virus replication in vitro (Miller et al, 1995) and at the SV40 ORI (Bargonetti et al, 1991) as well as in putative cellular origins of replication (Kern et al, 1991).

A number of genes not containing p53 response elements may be repressed by p53 (Ginsberg et al, 1991; Mercer et al, 1991; Shiio et al, 1992; Seto et al, 1992).

C. Binding of p53 to viral oncoproteins

p53 was first detected in rodent cells transformed with SV40 in a complex with T antigen (Lane and Crawford, 1979; Linzer and Levine, 1987). Subsequent studies have shown that p53 can be complexed with adenovirus E1B (Sarnow et al, 1982; van den Heuvel et al., 1990) and the E6 oncoprotein of human papilloma virus (Werness et al., 1990). SV40 T antigen was unable to act as an initiator of SV40 DNA replication in vitro when complexed with wt murine p53 (Wang et al, 1989) thought to act by blocking the interaction of T antigen with DNA polymerase a (Gannon and Lane, 1987; Braithwaite et al, 1987).

What appears to be important in understanding the involvement of p53 in tumorigenesis is that p53 is unable to transactivate the p53-inducible reporter genes in cells that express one of these viral oncoproteins (Yew and Berk, 1992). In addition, the growth suppressive effect of p53 protein may be mediated by its association with cellular proteins (Fields and Jang, 1990; Raycroft et al., 1990). Negative elements that could be required for an efficient growth shutdown leading to the reversible G₀ state or to irreversible out-of-cycle conditions such as terminal differentiation, apoptosis, and senescence, may be affected by p53 (Bargonetti et al., 1991).

D. Transcription repression by interaction of p53 with TBP

Although p53 activates a number of promoters that contain p53-responsive elements, it represses transcription from many promoters that lack p53 binding sites; central to the promoter repression by p53 was thought to be its interaction with the TATA-box binding protein or TBP (Seto et al, 1992; Mack et al, 1993; Truant et al, 1993). This interaction may activate transcription when TBP interacts with a preformed p53-DNA complex or may repress transcription when p53 interacts with DNA-bound TBP (Deb et al, 1994). However, p53 acts as a repressor only in cells undergoing apoptosis and p53-mediated transcriptional repression is released by adenovirus E1B or cellular Bcl-2 (Shen and Shenk, 1994; Sabbatini et al, 1995).

Both wild-type and mutant p53 interact with C/EBP on the human hsp70 promoter (Agoff, 1993), with TFIIH (Xiao et al, 1994), holo-TFIID (Chen et al, 1993; Liu et al, 1993) and the TAFII40 and TAFII60 subunits of TFIID (Thut et al, 1995).

E. Inhibition of DNA replication by wild-type p53

Several lines of evidence suggested inhibition in DNA replication by wild-type p53 but not by tumor-derived mutant forms of p53. Indeed, SV40 T antigen was unable to act as an initiator of SV40 DNA replication in vitro when complexed with p53 (Wang et al, 1989); mutant p53 was unable to cause inhibition in the initiating functions of T antigen in vitro (Friedman et al, 1990). Inhibition in DNA replication in vivo by p53 (Braithwaite et al, 1987) suggested that p53 might interact with cellular DNA replication initiator proteins or other components of the replication fork. p53 also interacts with replication protein A (RPA) implicated in DNA replication and in repair; interaction of p53 inhibits the replication functions of RPA (Dutta et al, 1993) although interaction of p53 with RPA via its acidic domains stimulates BPV-1 DNA replication in vitro (Li and Botchan, 1993). Immunolocalization of p53 (also of RB and host replication proteins) at foci of viral replication in HSV-infected cells (Wilcock and Lane, 1991) provided further evidence for a direct interaction of p53 with proteins (or DNA sequences) at the replication fork.

According to a second model, p53 can cause inhibition in DNA replication by a direct interaction with origins of replication at the DNA sequence level rather than via its interaction with replication initiator proteins. The potential role of p53 as a down-regulator of DNA replication in a DNA-binding-dependent manner has been suggested from replication assays of polyoma virus in vitro (Miller et al, 1995) and from the inhibition in nuclear DNA replication by a form of p53, truncated at its C-terminus, which is constitutively active for DNA binding in transcription incompetent extracts from Xenopus eggs (Cox et al, 1995). In the experiments of Miller and coworkers (1995) wild-type p53 suppressed DNA replication in vitro when the p53 binding site (RGC)₁₆ from the ribosomal gene cluster was cloned on the late side of the polyomavirus (Py) core origin; when mutated p53-binding sites were used, the inhibition in Py replication was not observed. In addition, RPA (able to interact directly with p53) was unable to relieve the p53-mediated repression in Py replication. Furthermore, tumor-derived mutants of p53 that had lost their sequence-specific DNA-binding capacity were unable to inhibit Py replication of the construct with the wild-type oligomerized RGC sites in vitro.

F. Differences in biological functions between wild-type p53 and tumor-derived p53 mutants

Tumor-derived mutant forms of p53 have lost their DNA sequence-specific binding capacities. For example the Trp-248 and His-273 mutants of p53 have poor DNA-binding abilities and are unable to activate transcription from constructs containing p53 binding sites (Farmer et al, 1992).

Wild-type (wt) p53 tumor suppressor protein negatively regulates cell growth (Hollstein et al, 1991; Prives, 1994). Whereas the wild-type p53 acts as a tumor suppressor, several of the mutant forms display oncogenic activities (Levine, 1993; Prives and Manfredi, 1993; Deppert, 1994). Although the wt p53 has been postulated to repress growth by activating genes that repress growth (p21), many of the mutant forms have lost their DNA sequence-specific binding and transcriptional activation capacities (reviewed by Zambetti and Levine, 1993).

According to one model (see Vogelstein and Kinzler, 1992), wt p53 is a positive regulator for the transcription of genes that by themselves are negative regulators of growth control and/or invasion. Indeed, p53 upregulates the genes of p21/CIP1/WAF1 (ElDeiry et al, 1993) and GADD45 (Kastan et al, 1992) whose products interact with PCNA to inhibit its association with DNA polymerase d thus causing arrest in DNA replication (Smith et al, 1994). This feature of p53 that is central to its ability to suppress neoplastic growth is lost by mutations on p53 that result in loss of its ability to bind to DNA or to interact with other transcription protein factors.

Mutant p53 can transactivate genes that up-regulate cellular growth (Deb et al, 1992; Dittmer et al, 1993) such as PCNA (Shivakumar et al, 1995), EGFR (Deb et al, 1994), multiple drug resistance (MDR1) (Chin et al, 1992; Zastawny et al, 1993), and human HSP70 in vivo (Tsutsumi-Ishi et al, 1995). These studies support the idea for an oncogene function of the mutant p53 protein compared with the tumor suppressor function of wt p53; mutation in the p53 gene may, thus, cause gain of new functions such as transforming activation and binding to a distinct class of promoters which are not normally regulated by wt p53 (Zambetti and Levine, 1993; Tsutsumi-Ishi et al, 1995). At the same time appearance of mutations in the p53 gene result in the loss of function of the wt p53 (Zambetti and Levine, 1993).

The wild-type but not mutant p53 at low levels transactivates the human PCNA promoter in a number of different cell lines; the wild-type p53-response element from the PCNA promoter functions in either orientation when placed on a heterologous synthetic promoter; thus moderate elevation of p53 can induce PCNA, enhancing the nucleotide excision repair functions of PCNA (Shivakumar et al, 1995). Whereas low levels of wild-type p53 activate the PCNA promoter, higher concentrations of wt p53 inhibit the PCNA promoter, and tumor-derived p53 mutants activate the promoter (Shivakumar et al, 1995).

While the wtp53 is endowed with a 3'-to-5' exonuclease activity, associated with the central DNA-binding domain, and thought to function during repair, replication, and recombination, the 273^His mutant of p53 has lost the exonuclease activity (Mummenbrauer et al, 1996).

G. Involvement of p53 in repair and control of the cell cycle

p53 controls the level of expression of the p21 gene, encoding a protein that inhibits the activity of cyclin-dependent kinases (CDKs); CDK activity is essential for the phosphorylation of RB at the G1/S checkpoint of the cell cycle resulting in the release of E2F transcription factor from RB-E2F complexes and in the up-regulation by the released E2F of genes required for DNA synthesis. p21 levels are reduced considerably in tumor cells that have lost the p53 protein or contain a nonfunctional mutated form of p53 (El-Deiry et al, 1993). In addition, the p21 inhibitor of cyclin-dependent kinases associates with PCNA thus blocking its ability to activate DNA polymerase d; this could give rise to the abnormal control in DNA replication or to the loss of coordination between DNA replication and cell cycle progression seen in tumor cells. Thus the upregulation of the p21 gene by p53 acts in two different ways causing a cascade of events.

p53 is linked directly to homologous recombination processes via its interaction with the RAD51/RecA protein (Stürzbecher et al, 1996).

H. A proposal for an efficient killing of cancer cells using p53/PAX5 expression vectors

Introduction of a null p53 mutation by homologous recombination in murine embryonic stem cells gave mice which appeared normal but were susceptible to a variety of neoplasms by 6 months of age (Donehower et al, 1992). Relevant to the issue that p53 is dispensable for embryonic development are the studies of Stuart and coworkers (1995) suggesting that during early embryo development p53 is not expressed because of the suppression of its gene by Pax5; at later stages of development Pax5 inactivation allows p53 to be expressed and exert its control on cell growth (Figure 22).

A significant factor to be considered in approaches aimed at transferring the wt p53 gene to tumor cells is the impairment of the wt p53 functions by the endogenous mutant p53 expressed in tumor cells which is able to tetramerize with wt p53; optimal results will be expected if the endogenous mutant p53 gene is inhibited concurrently with overexpression of the wt p53 gene. It has been proposed (Boulikas, 1997) that effective suppression of tumor growth with p53 vectors could be achieved by the simultaneous transfer of wt p53 plus Pax5 to cancer cells; Pax5 is a well established supressor of the p53 gene; its effect is exerted via a direct interaction of Pax5 with a control element in the first exon of the p53 gene (Stuart et al, 1995). Pax5 is an homeotic protein, controlling the formation of body structures during development; Pax5 is expressed in early embryo stages to keep the levels of p53 low and allow rapid proliferation of embryonic tissues. Simultaneous transfer to solid tumors of a PAX5 and p53 genes in the same expression vector but with the wt p53 mutagenized at 2-3 nucleotides to abort the PAX5 suppressive site was proposed as a strategy to effectively suppress tumor cell proliferation (Boulikas, 1997).

I. p53 gene bombs that explode in cancer cells

Exogenous genes encoding "weapons" (suicide genes) and "triggers" have been devised whose delivery to somatic cells will affect only cancer cells. The production of mutated forms of p53 at high levels by cancer cells (normal cells do not have adequate amounts of wt p53 protein) is being exploited to pull a molecular trigger resulting in the transcriptional activation of a toxic gene and in the death of cancer cells (da Costa et al, 1996). This invention is based on the fact that (i) powerful chimeric transcription factors can be engineered consisting of a DNA-binding domain (DBD) and a transactivation domain (TAD) and (ii) prokaryotic or viral enzymes are able to convert nontoxic prodrugs into toxic derivatives (suicide genes, see HSV-tk, CD and PNP further below); the toxic derivative produced in tumor cells which are transfected can diffuse to surrounding cells causing their killing even in

the absence of transfection of these cells, a phenomenon known as "bystander effect".

Trigger genes in plasmids were made up of the DNA-binding domain of GAL4 (aa 1-147) fused in frame to a protein domain that could interact with p53; the p53-binding domain was the 84-708 aa region of SV40 T antigen or of the 305-393 aa TAD domain of p53 (which acts as a tetramer) and which is similar between wt and mutant p53 (mutations on p53 are within the DBD). The constructs included the E. coli (DeoD) gene which encodes the purine nucleoside phosphorylase (PNP) under control of the GAL4 response element (known as upstream activating sequence or UAS); the PNP gene can convert the 6-methylpurine deoxyribose (MeP-dR) prodrug into the diffusible, toxic 6-methylpurine (see page 69) and can become a powerful suicide gene under these conditions (Sorscher et al, 1994). Transfection of cells in culture with these constructs followed by treatment of the cells with MeP-dR resulted in the death of the cells (da Costa et al, 1996). The mechanism was based on the fact that most normal cells do not express the p53 gene and those who do express the wt p53 are destined for programmed cell death; therefore, cancer cells containing elevated amounts of mutant forms of p53 were amenable to this strategy.

The 55- kDa E1B protein of adenovirus binds to and inactivates the p53 gene; ONYX-015 is an E1B, 55-kDa gene-attenuated adenovirus unable to replicate and show cytopathogenicity in tumor cell lines which express a wt p53 such as in RKO and U20S carcinoma lines but can cause cytolysis in cell lines expressing a mutated form of p53; a wide range of human tumor cells, including numerous carcinoma lines with either mutant or normal p53 gene sequences (exons 5-9), were efficiently destroyed following intratumoral or intravenous administration of ONYX-015 to nude mouse-human tumor xenografts; furthermore, combination therapy with ONYX-015 plus chemotherapy (cisplatin, 5-fluorouracil) was significantly greater than with either agent alone. On the contrary, normal human cells were highly resistant to cytolysis by the adenovirus (Heise et al, 1997).

J. Transfer of the p53 gene in cell culture

Preclinical studies have shown that both viral and plasmid vectors able to mediate high efficiency delivery and expression of wild-type tumor suppressor p53 gene can cause regression in established human tumors, prevent the growth of human cancer cells in culture, or render malignant cells from human biopsies non-tumorigenic in nude mice. Inhibition in cell proliferation was observed in cell culture and in tumors after induction of p53 expression with adenovirus vectors (Bacchetti and Graham, 1993; Wills et al, 1994; Zhang et al, 1994).

Transfer of the wild-type p53 gene using a defective HSV vector into a human medulloblastoma cell line containing a mutant copy of p53 resulted in p53 expression, increased the levels of mdm2 proteins and induced cell cycle arrest of the majority of transduced cells (Rosenfeld et al, 1995).

Apoptosis can be induced in cultured NCI-H 596 human non-small cell lung cancer cells, which have a wild-type p53 gene, by EGF signaling in a p53-dependent manner; whereas treatment of these cells with EGF plus p53 sense oligonucleotides induced EGF-dependent and p53-dependent apoptosis within 8 hours, antisense p53 gene therapy suppressed the induction of apoptosis. A new nucleic acid drug was developed based on a mutated p53 antisense with a mutation at three bases immediately 5' and 3' from the CG dinucleotides which potentiated the induction of apoptosis and failed to suppress the induction of EGF-dependent apoptosis (Murayama and Horiuchi, 1997).

Infection of the androgen-independent human prostate Tsu-pr1 cell line lacking functional p53 alleles with recombinant adenovirus vectors (replication-deficient) carrying the p53 gene under control of the CMV promoter resulted in expression of p53 and induced striking morphological changes: the cells were detached from the substratum, condensed, and exhibited breakdown of the nuclear DNA into nucleosome-size fragments characteristic of apoptosis; whereas control cells were able to elicit tumors in nude mice, the AdCMV/p53-infected cells failed to form tumors (Yang et al, 1995).

K. Animal studies using p53 gene transfer

Intratracheal injection of a recombinant retrovirus containing the wt p53 gene was shown to inhibit the growth of lung tumors in mice nu/nu models inoculated intratracheally with human lung cancer H226Br cells whose p53 gene has a homozygous mutation at codon 254 (Fujiwara et al, 1994). A number of other studies have shown suppression in tumor cell growth and metastasis after delivery and expression of the wt p53 gene (Diller et al, 1990; Chen et al, 1991; Isaacs et al, 1991; Wang et al, 1993).

The safety of the adenovirus-mediated p53 transduction of the liver in normal rats and in 50%-hepatectomized animals subjected to asanguineous portal perfusion was examined by Drazan et al (1994). The gene transfer rate in whole liver and after hepatectomy ranged from 20% to 40%; liver regeneration and hepatocyte function were unaffected by overexpression of p53.

Delivery of the p53 gene to malignant human breast cancer cells in nude mice using DOTMA:DOPE 1:1 cationic liposomes (400 nmoles liposomes/35 mg DNA) resulted in regression (60% reduction in tumor cell volume) in 8 out of 15 animals treated; animals were receiving one injection every 10 days (Lesoon-Wood et al, 1995). It was thought that wild-type p53 expression (tumor cells were expressing mutant forms of p53) upregulated p21 gene to inhibit cell growth by inhibition in cyclin-dependent kinases but also via induction of apoptosis preferentially in cancer cells.

When a recombinant adenovirus encoding wild-type p53 under the control of the human CMV promoter was introduced into SK-OV-3 human ovarian carcinoma cells it increased by more than 50% the life span of nude mice injected with these cells; control animals in this highly aggressive ovarian xenograft model died between 25-45 days from injection time (Mujoo et al, 1996). Adenoviral transfer of a functional p53 gene into a radiation-resistant SCCHN cell line that harbors mutant p53 restored the G1 block and apoptosis in these cells in vitro and sensitized SCCHN-induced mouse xenografts to radiotherapy in vivo (Chang et al, 1997).

The efficacy of a replication-deficient p53 adenovirus construct was tested against three human breast cancer cell lines expressing mutant p53, MDA-MB-231, -468, and -435 and was found to be highly effective against 231 and 468 cells as well as their tumor xenografts in nude mice but not against 435 cells probably due to their low adenovirus transduction. 37% of growth inhibition of 231 cells was due to p53, while 49% was adenovirus-specific (Nielsen et al, 1997).

Cytotoxic T lymphocytes (CTLs) recognizing a murine wild-type p53 were able to discriminate between p53-overexpressing tumor cells and normal tissue and caused complete and permanent tumor eradication without damage to normal tissue after adoptive transfer into tumor-bearing p53^+/+ nude mice. CTLs, presented by the MHC class I molecule H-2Kb, were generated by immunizing p53 gene deficient (p53^-/-) C57BL/6 mice with syngeneic p53-overexpressing tumor cells (Vierboom et al, 1997).

L. Transfer of the p53 tumor suppressor gene to prostate cancer cells

Although primary prostate tumors have few mutations in the p53 gene (Voeller et al, 1994; Isaacs et al, 1994), specimens from advanced stages of the disease and metastases as well as their cell lines frequently display mutations or deletions at both alleles of the p53 gene (Chi et al, 1994; Dinjens et al, 1994). Three of five prostate cancer cell lines examined (TSUPr-1, PC3, DU145) and one out of two primary prostate cancer specimens were found to harbor mutations altering the amino acid sequence of the conserved exons 5-8 of the p53 gene; transduction of the p53-defective cell lines with the wt p53 gene using lipofectin showed reduction in tumorigenicity assayed from reduced colony formation and the cells became growth arrested (Isaacs et al, 1991).

Endocrine therapy is ineffective once the prostate cancer becomes androgen-independent; these cancers remain unresponsive to conventional chemotherapy. Androgen-independent and metastatic prostate cancers were established in athymic male mice by co-inoculation with the LNCaP human prostate cancer cell line and the MS human bone stromal cell line; these tumors became necrotic and were successfully eradicated by intratumoral injection of a recombinant p53/adenovirus; the p53 gene was driven by the CMV promoter and the SV40 poly(A) signal placed in the E1 region of Ad5 (Ko et al, 1996). It was suggested that in addition to the tumor suppressor, apoptotic, and antiangiogenesis function of p53, tumor necrosis was induced by a bystander effect or a general immune response which attracted immune cells to cause tumor cell killing (Ko et al, 1996).

M. Clinical trials using p53 gene transfer

A human clinical trial at M.D. Anderson Cancer Center uses transfer of the wild-type p53 gene, in patients suffering with non-small cell lung cancer and shown to have p53 mutations in their tumors, using local injection of an Ad5/CMV/p53 recombinant adenovirus at the site of tumor in combination with cisplatin (Roth, 1996; Roth et al, 1996; protocols #29 and 124, Appendix 1). A retroviral vector containing the wild-type p53 gene under control of a b-actin promoter was used for multiple percutaneous injections or direct thoracoscopic injections at the site of the tumor into nine patients, all in advanced stages, with non-small cell lung cancers. Patients whose conventional treatments failed were selected for a p53 mutation in the lung tumor. Reduction in tumor volume was achieved via apoptosis (assayed in posttreatment biopsies) in three patients, and arrest in tumor growth in three other patients (Roth et al, 1996).

RAC-approved clinical trials (Appendix 1) using p53 cDNA transfer are #29 (treatment of non-small cell lung cancer with p53 and antisense K-ras), #124 (intratumoral delivery of adenoviral p53 cDNA plus cisplatin), #130 (intratumoral injection of adenoviral p53 to treat head and neck squamous cell carcinoma), #131 (primary and metastatic malignant tumors of the liver), #147 (percutaneous injections of adenovirus p53 for hepatocellular carcinoma), #148 (advanced or recurrent adenocarcinoma of the prostate), #152 (intra-tumoral injections of ad5cmv-p53 to patients with recurrent squamous cell carcinoma of the head and neck), #153 (intralesional delivery of adenovirus p53 in combination with chemotherapy in breast cancer), #154 (intratumoral injection of adeno p53 to patients with advanced prostate cancer), #155 (intratumoral injection of adeno p53 to patients with advanced and metastatic bladder cancer), and #156 (adeno p53 for non-small cell lung cancer).

XVIII. p21 and p16 in cancer gene therapy

A. Molecular action of p21

p53 upregulates the p21/CIP1/WAF1 gene (simply called p21) (ElDeiry et al, 1993). Induction of the p21/Waf-1/Cip-1 gene causes growth arrest via inhibition of cyclin-dependent kinases (CDKs). CDKs are upregulated by cyclins which act as positive regulators of cell cycle progression. Cdk2, also called p33^cdk2, is the master regulator of the cell cycle at the G1/S transition point. Whereas cdk2 is expressed at constant levels throughout the cell cycle, its activation by phosphorylation is first detected a few hours before the onset of DNA synthesis; furthermore, antibodies directed against Cdk2 blocked mammalian cells from entering S phase. D1 cyclin associates with Cdk2, Cdk4, and Cdk5 to control the G1®S transition point; the genes of cyclins D1 and E are overexpressed or rearranged in malignancies and conditional overexpression of human cyclins D1 and E in Rat-1 fibroblasts causes a decrease in the length of G1 and an acceleration of the G1/S phase transition. D1 appears to be specialized in the emergence of cells from quiescence (Go®G1 transition) whereas cyclin E is more oriented toward control of the G1/S transition. Cdc2, a close relative of Cdk2 and whose pattern of phosphorylation is cell cycle-regulated, becomes associated with cyclin B to regulate the G2®M transition (see Boulikas, 1995a).

CDK activity is essential for the phosphorylation of RB at the G1/S checkpoint of the cell cycle resulting in the release of E2F transcription factor from RB-E2F complexes and in the up-regulation of genes required for DNA synthesis by the released E2F. p21 levels are reduced considerably in tumor cells that have lost the p53 protein or contain a nonfunctional mutated form of p53 (El-Deiry et al, 1993).

Induction of the p21/Waf-1/Cip-1 gene also causes growth arrest via inactivation of PCNA; indeed, the p21 inhibitor of cyclin-dependent kinases associates with PCNA, the accessory of DNA polymerases a and d , thus blocking its ability to activate these DNA polymerases; this could give rise to the abnormal control in DNA replication or to the loss of coordination between DNA replication and cell cycle progression seen in tumor cells (Li et al, 1994).

B. p21 and p16 gene transfer

Introduction of the wt p53 or of the p21 downstream mediator of p53-induced growth suppression into a mouse prostate cancer cell line, deficient in p53, led to an association of p21 with Cdk2; this interaction was sufficient to downregulate Cdk2 by 65% (Eastham et al, 1995). The p21 gene, driven by CMV promoter into an Adenovirus 5 vector, was more effective than the AD5CMV-p53 vector, (harboring the p53 gene under control of the same elements as p21), in reducing tumor volume in syngeneic male mice with established s.c. prostate tumors; tumors were induced by injection of 2 million cells in each animal. These studies suggested that p21 expression might have more potent growth suppressive effect than p53 in this tumor model and that p21 may be seriously be included in the constellation of anticancer arsenals.

Transfer of p21 is an effective tool to lead carcinoma cells with inactivated p53 into less malignant phenotypes. p53 is frequently inactivated by papilloma viruses in carcinomas of the uterine cervix. Transfer of the p21 gene to HeLa cells, a widely used uterine cervix cell line, resulted in a significant growth retardation by blockage of G1 to S transition, reduced anchorage-independent growth and attenuated telomerase activity (Yokoyama et al, 1997). Introduction of p21 with adenoviral vectors into malignant cells completely suppressed their growth in vivo and also reduced the growth of established pre-existing tumours (Yang et al, 1997).

Transfer of p21 was used to suppress neointimal formation in the balloon-injured porcine or rat carotid arteries in vivo (Yang et al, 1996; Ueno H et al, 1997a). A combination therapy in mice with simultaneous transfer of the p21 gene and of the murine MHC class I H-2Kb gene, which induces an immune response that stimulates tumor regression, was more effective than treatment with either gene alone (Ohno et al, 1997).

Malignant gliomas extensively infiltrate the surrounding normal brain and their diffuse invasion is one of the most important barriers to successful therapy; one of the most frequent abnormalities in the progression of gliomas is the inactivation of the tumor-suppressor gene p16, suggesting that loss of p16 is associated with acquisition of malignant characteristics. Restoring wild-type p16 activity into p16-null malignant glioma cells modified their phenotype. Adenoviral transfer of the p16/CDKN2 cDNA in p16-null SNB19 glioma cells significantly reduced invasion into fetal rat-brain aggregates and reduced expression of matrix metalloproteinase-2 (MMP-2), an enzyme involved in tumor-cell invasion (Chintala et al, 1997).

XIX. Gene therapies based on transfer of the retinoblastoma (RB) gene

A. RB and E2F proteins in the control of the cell cycle and apoptosis

Retinoblastoma protein is a transcription factor (Lee et al, 1987) involved in the regulation of cell cycle progression genes (reviewed by White, 1998, this volume). The role of RB on cell proliferation and tumor suppression arises (i) from its association with E2F, an association disrupted by RB phosphorylation at the G1/S checkpoint resulting in release of E2F and in the upregulation of a number of genes required for DNA replication; (ii) from the direct association of RB protein with a number of viral oncoproteins or key regulatory proteins including E1A of adenovirus (Whyte et al., 1988), SV40 large T (Ludlow et al., 1990) and the human papilloma virus E7 protein (Dyson et al., 1989). Normal cellular targets of RB, such as the transcription factor E2F (Bagchi et al., 1991; Chellapan et al., 1991) become dissociated from the RB protein in the presence of these viral proteins in the cell (E1A, T antigen, E7), leading to cell cycle progression. This constitutes a mechanism (also the interaction of viral proteins with p53, see above) viruses use to render infected cells continuously cycling.

(iii) RB is able to repress directly c-fos gene expression (Robbins et al., 1990) and has been proposed to have a similar effect on c-myc expression (Pietenpol et al., 1990). (iv) RB also suppresses cell growth by directly repressing transcription of the rRNA and tRNA genes by blocking the activity of RNA polymerase I transcription factor UBF (Cavanaugh et al, 1995; reviewed by White, 1998).

Hypophosphorylated RB, but not mutant RB, was associated with the nuclear matrix, particularly concentrated at the nuclear periphery and in nucleolar remnants, only during early G1; the peripheral matrix proteins lamin A and C bound RB in vitro. This association was thought to be important for the ability of RB to regulate cell cycle progression (Mancini et al, 1994). It is interesting that mutated p53 but not wtp53 interacts with specific types of MARs (Will et al, 1998); nuclear matrix is an essential structure for replication transcription recombination and repair processes intimately connected to mechanisms of carcinogenesis.

The tumor suppressor function of RB is believed to occur by complex formation between E2F and RB or the RB-related proteins p107 and p130, a complex that down-regulates the DNA-binding activities of E2F; the transcription activating capacity of E2F on the genes it regulates can be repressed by interaction with RB (Nevins, 1992). Cyclin A, believed to facilitate DNA replication, also associates with E2F; both types of complexes, E2F-RB and E2F-cyclin A, can be dissociated by the adenovirus E1A protein. The release of E2F by E1A results in cell cycling and this constitutes an additional mechanism of interference of adenoviruses with the proliferation of the infected cells; release of E2F from RB induced by E1A is critical for transformation of cells by E1A (for references see Hiebert et al, 1995).

The p107 protein with similarities in structure and DNA-binding properties to RB also binds cyclin A; whereas RB is complexed to E2F during G1 the p107-cyclin A complex interacted with E2F as cells entered S phase (Shirodkar et al., 1992).

E2F is a transcription factor that activates the adenovirus E2 gene and a number of cellular genes that respond to proliferation signals and that control the passage of the cell cycle through S phase such as myc and DHFR genes and contributes to the uncontrolled proliferation of adenovirus-transformed cells (Mudryj et al., 1991; see White, 1998, this volume). It has been speculated that the physiological function of RB (and also of its similar protein p107) in negatively-regulating cell growth and in acting as a tumor suppressor protein are exerted via its ability to down-regulate the activity of E2F (Shirodkar et al., 1992); this has been subsequently confirmed by numerous studies. RNA ligands that bind to E2F1 were selected from RNA libraries and were used to inhibit the induction of S phase in cultured cells (Ishizaki et al, 1996). Such molecules might find applications in cancer therapy because of the important role of E2F proteins in the regulation of cell cycling.

Retinoblastoma protein has a functional domain (the pocket) for binding to transcription factor E2F implicated in cell growth control. The same domain is responsible for the association of RB with the adenovirus E1A, the SV40 large T, and the human papilloma virus E7 proteins (Kaelin et al., 1992). Using an approach for screening lgt11 expression libraries, clones encoding for RB-binding proteins were identified; among those are RBAP-1 and 2, or retinoblastoma-associated proteins 1 and 2 (Kaelin et al., 1992) and RBP3 (Helin et al., 1992). RBAP-1 binds to the RB pocket, copurifies with E2F, contains a functional transactivation domain, and binds to E2F cognate sequences (Kaelin et al., 1992).

E2F contains a RB-binding domain in its C-terminus (Helin et al., 1992; Shan et al., 1992). RB binds directly to the activation domain of E2F1 and silences it, thereby preventing cells from entering S phase. To induce complete G1 arrest, RB requires the presence of the hbrm/BRG-1 proteins, which are components of the coactivator SWI/SNF complex. This cooperation was mediated through a physical interaction between RB and hbrm/BRG-1. RB can contact both E2F1 and hbrm at the same time, thereby targeting hbrm to E2F1 (Trouche et al, 1997).

E2F cooperates with p53 to induce apoptosis (Wu and Levine, 1994) and high levels of wild-type p53 potentiate E2F-induced apoptosis in fibroblasts (Qin et al, 1994). The physiological relevance of E2F in the apoptotic mechanism was thought to arise from the ability of E2F to act as a functional link between p53 and RB; p53 levels increase in response to high levels of E2F (DP is required for the association of E2F with RB); overexpression of both E2F-1 and DP-1 led to a rapid death of (IL-3)-dependent 32D.3 myeloid cells even in the presence of survival factors (Hiebert et al, 1995). Overexpression of exogenous E2F-1 using a tetracycline-controlled expression system in Rat-2 fibroblasts promoted S-phase entry and subsequently led to apoptosis (Shan and Lee, 1994).

B. Phosphorylation of RB: the TGF-b1, IL-1, and IL-6 connection

Work from several groups has shown that RB is un- or under-phosphorylated in G0/G1 and becomes phospho-rylated in its N-terminal domain during S and G2/M (Buchkovich et al., 1989; Chen et al., 1989; DeCaprio et al., 1989; Mihara et al., 1989). Only under-phosphorylated RB interacts with E2F (Chellappan et al., 1991). Treatment with TGF-b1 maintained RB protein in its active dephosphorylated form, thus providing a link between RB growth suppression and growth inhibition by TGF-b1.

Interleukin-6 (IL-6), known to mediate autocrine and paracrine growth of multiple myeloma (MM) cells and to inhibit tumor cell apoptosis was determined to exert this function via phosphorylation of RB protein; this finding could explain the abnormalities of RB protein and mutations of RB gene associated with up to 70% of MM patients and 80% of MM-derived cell lines. Culture of MM cells with RB antisense, but not RB sense, oligonucleotide triggered IL-6 secretion and proliferation in MM cells; phosphorylated pRB was constitutively expressed in MM cells and IL-6 shifted pRB from its dephosphorylated to its phosphorylated form (Urashima et al, 1996).

Interleukin-1 (IL-1) causes G0/G1 phase growth arrest in human melanoma cells, A375-C6 via hypophosphory-lation of RB protein. Exposure to IL-1 caused a time-dependent increase in hypo-phosphorylated RB that correlated with an accumulation of cells arrested in the G0/G1 phase; this was abrogated by the SV40 large T antigen which binds preferentially to hypo-phosphorylated RB, but not by the K1 mutant of the T antigen, which is defective in binding to RB (Muthukkumar et al, 1996).

C. Genes regulated by RB protein

RB represses a number of genes by sequestering or inactivating the positive transcription factor E2F and seems to activate some other genes by interacting with factors like Sp1 or ATF-2 (Rohde et al, 1996). RB protein is a master regulator of a complex network of gene activities defining the difference between dividing and resting or differentiated cells. Using the method of differential display Rohde et al (1996) detected a number of genes which were upregulated by ectopic expression of the RB gene in RB-deficient mammary carcinoma cells including the endothelial growth regulator endothelin-1 and the proteoglycans versican and PG40.

Introduction of the wild-type RB gene via retrovirus-mediated gene transfer has provided several RB-reconstituted retinoblastoma cell lines (Huang et al., 1988; Chen et al., 1992). These RB⁺ cell lines showed little difference in their growth rates in culture when compared to the parental or revertant RB^- cells; however, RB⁺ cells invariably lost their tumorigenicity in nude mice assays (Chen et al., 1992). RB protein down-regulates its own gene and this negative autoregulation is mediated by the transcription factor E2F; this was shown by inserting the promoter of the RB gene 5' of the bacterial CAT reporter gene followed by its transfection into RB⁺ and RB- retinoblastoma cells: RB promoter activity was signifi-cantly decreased in RB⁺ cells (Shan et al., 1994).

D. Transcription factors (TFs) that regulate the RB gene

Several mutations have been found in the promoter region of the RB gene, suggesting that inappropriate transcriptional regulation of this gene contributes to tumorigenesis. The presence of E2F recognition sites in promoters of a number of growth-related genes suggested that expression of these genes might be affected by RB. Understanding the nature and availability of TFs which regulate the RB gene in particular cell types is instructive for a successful gene therapy application involving transfer of RB.

An E2F recognition site lies within a region critical for RB gene transcription; binding of E2F-1 at this site transactivates the RB promoter; striking back, the resulting overexpression of RB suppresses E2F-1-mediated stimulation of RB promoter activity and, thus, the expression of RB is negatively autoregulated through E2F-1 (Shan et al, 1994). Up-regulation of the RB gene by E2F was shown by co-transfection of RB- osteosarcoma Saos2 cells in culture with a plasmid expressing E2F-1 under the control of the CMV immediate-early gene promoter-CAT construct: expression of E2F-1 stimulated RB promoter activity 10-fold under conditions where E2F-1 had little effect on c-jun, c-myc, and EGR-1 gene expression (Shan et al., 1994). The autoregulation of RB gene by RB may be accomplished via a direct protein-DNA complex formation, via protein-protein interaction regulating the activity of other transcription factors on the promoter of the RB gene, or both.

Two distinct DNA-binding factors, RBF-1 and ATF, play an important part in the transcription of the human RB gene. The promoter of the human RB gene and of the mouse RB1 gene (Zacksenhaus et al., 1993) contain binding sites for ATF, and a Sp1-like transcription factor (Mitchell and Tijan, 1989) where the RBF-1 (retinoblastoma binding factor 1) may bind (Sakai et al., 1991). Human RB gene is also regulated by AP-1 (Linardopoulos et al, 1993), as well by the early response transcription factor, nerve growth factor inducible A gene (NGFI-A) which is expressed in prostate cells and binds to the site GCGGGGGAG at -152 to -144 within the RB gene promoter (Day et al, 1993). The ATF site of the RB promoter is a responsive element during myogenic differentiation; RB promoter activity increased about 4-fold during differentiation and was reduced when a point mutation was designed in the ATF site (Okuyama et al, 1996).

pRB activates expression of the human transforming growth factor-b2 gene through ATF-2; the human RB gene promoter is autoregulated by RB protein via an ATF-2-like binding site at the carboxyl-terminal domain of pRB; overexpression of RB stimulates RB promoter activity through the ATF binding site in a variety of different cell types (Park et al, 1994).

The candidate oncoprotein Bcl-3, previously characterized as a member of the IkB family, activated transcription of the RB gene, whose promoter has no typical NF-kB sites, via binding to a DNA element identical to E4TF1/GABP site; Bcl-3 promoted tetramerization of E4TF1. Expression of the antisense bcl-3 RNA in myoblasts suppressed induction of RB and myogenic differentiation whereas transient expression of bcl-3 in myoblasts was shown to induce expression of the endogenous RB (Shiio et al, 1996).

Two oncogenic point mutations at the Sp1 and ATF sites of the RB gene promoter were identified in two separate hereditary RB families. The Sp1 consensus site mutation was blocking the action of RBF-1, recently identified as the human GABP/E4TF1, a transactivator from the adenovirus early-region 4 promoter. The human GABP/E4TF1 protein enhanced the core RB promoter activity, whereas it did not stimulate a mutant RBF-1 site and was proposed to be the most essential transcription factor for human RB gene activation (Clark et al, 1997).

Whereas binding of the Sp1 transcription factor is not significantly affected by methylation of the CpG dinucleotide within its binding site, 5'-GGGCGG (lower strand, 5'-CCGCCC) methylation of the outer C is inhibitory (mammalian cells also have the capacity to methylate cytosines at CpNpG sites) and in particular methylation of both cytosines ^mCp^mCpG inhibited binding by 95%; endogenous ^mCp^mCpG methylation of an Sp1 site in the CpG island promoter of the RB gene was identified by genomic sequencing in a proportion of retinoblastoma tumors which were extensively CpG methylated in the RB promoter (Clark et al, 1997).

E. RB gene transfer

Functional loss of the RB gene has been implicated in the initiation or progression of several human tumor types including cancer of the eye, bone, bladder, and prostate. The cancer suppressor activity of RB was directly demonstrated by the introduction of a normal RB gene into retinoblastoma cells that have lost the RB function (inability to be phosphorylated because of mutations at the appropriate sites) by mutation at both alleles; this led to the suppression of the neoplastic phenotype and loss of the tumorigenicity of RB cells in nude mice (Huang et al, 1988). Expression of the normal RB gene into the human prostate carcinoma cell line DU145, mediated by recombinant retrovirus integration, also resulted in loss of its tumorigenic ability in nude mice (Bookstein et al, 1990). Studies with tumor cells reconstituted with RB ex vivo and implanted into immunodeficient mice, as well as with germline transmission of a human RB transgene into tumor-prone Rb^+/- mice have demonstrated cancer suppression (see Riley et al, 1996).

DU145 cells express a shorter protein lacking 35 amino acids from exon 21 due to a 105 nucleotide in-frame deletion (Bookstein et al, 1990). The human bladder carcinoma cell line J82 contains a mutated RB protein with exactly these features (Horowitz et al, 1989); this 35 amino acid stretch is required for complexation with T antigen and E1A. However, the two cell lines have lost exon 21 of RB because of a different type of mutation: J82 cells have a point AG to GG mutation in the intron 20-splice acceptor site but the type of mutation in DU145 leading to exon 21 loss is different (Bookstein et al, 1990).

Intratumoral infection of spontaneous pituitary melanotroph tumors arising in immunocompetent Rb^+/- mice with a recombinant adenovirus carrying the RB cDNA inhibited the growth of tumors, re-established innervation by growth-regulatory dopaminergic neurons, and prolonged the life spans of treated animals (Riley et al, 1996).

Retrovirus-mediated gene transfer of RB to the breast carcinoma cell lines MDA-MB468 and BT549, both of which harbor partial RB gene deletions as well as point mutations of their p53 genes, restored its expression in cells, reduced their ability to grow in soft agar, and their tumorigenicity in nude mice, although it did not significantly altered growth rate in culture (Wang et al, 1993).

Future therapeutic approaches using the RB gene are directed toward inhibition in cell proliferation (such as to inhibit neointima formation and smooth muscle cell proliferation in arterial diseases, see Arterial injury below and Chang et al, 1995) rather that aggressive suppression and apoptosis of solid tumors; p53 is a better gene than RB for tumor eradication.

XX. Induction of apoptosis for cancer gene therapy

A. Apoptosis as an essential process

Apoptosis has become a basic tool in developing cancer research in establishing new anticancer strategies. The health of a multicellular organism depends both on the ability of the body to produce new cells but also on the ability of certain type of cells to perish, self-destruct, when they become superfluous or severely damaged. Apoptosis, or programmed cell death, is a biological process associated with pronounced morphological changes, chromatin condensation, drop in pH, and intranucleosomal DNA degradation by which a cell actively commits suicide. Virtually all tissues have apoptotic cells; salient examples in the adult are: the eye lenses which consist of apoptotic cells that replaced their cytoplasm with crystallin; intestinal wall cells which migrate to the tip of the finger-like projections over several days where they die; ineffectual T cells which mature in thymus and which would attack the body’s own tissues are eliminated by apoptosis before entering the bloodstream; skin cells migrate from the deepest layers to the surface where they commit suicide forming the outer layer of the skin. Apoptosis is an essential process during embryogenesis: mammals eliminate neuron cells as the nervous system is formed; tadpoles delete their tails by apoptosis (reviewed by Duke et al, 1996).

Virus-transformed as well as severely X-ray-damaged or UV-damaged cells are similarly eliminated from the tissue via apoptosis; if they are left they can form malignant cells. Initiated cancer cells may lead to tumor development only when a dysfunction in their apoptotic pathway takes place. Although the biochemical aspects of cell death are fraught with the problem of cause versus effect, the role of apoptosis in neoplasia and its regulation by a number of oncogenes and p53 has emerged. Apoptosis is essential for normal development and homeostasis; deregulation in the positive control of apoptosis is associated with cancer and autoimmune disease whereas deregulation in the negative control of apoptosis is associated with degenerative diseases (reviewed by White, 1993; Duke et al, 1996).

B. Molecular mechanisms for apoptosis: p53, Bax, Bcl-2, c-Myc and other proteins

Apoptosis is of special interest in gene therapy not only of cancer but of other diseases such as arterial disease. Apoptosis is a complex process involving a significant number of apoptotic and antiapoptotic mechanisms. The cytotoxic (killer) T lymphocytes of the immune system of the infected organism bind to virus-infected cells inflicting their eradication with two different type of proteins: Perforin is a transmembrane molecule transferred from the killer T cell to the membrane of the infected cells forming holes on the membrane of the target cell allowing uptake of proteases called granzymes that activate ICE-like proteases to induce apoptosis. A number of antiviral drug development strategies are based on blockage of the activity of antiapoptotic viral proteins.

Expression of a number of genes induce apoptosis; their protein products include adenovirus E1A (Debbas and White, 1993; Lowe and Rudley, 1993) and c-Myc (Hermeking and Eick, 1994; Wagner et al, 1994). A number of proteins when expressed at sufficient amounts block apoptosis; these include Bcl-2 and E1B 19 kDa protein of adenovirus (Debbas and White, 1993; Chiou et al, 1994). Exposure of cells to a variety of growth factors including IL-3, IL-6, and erythropoietin, acting as survival factors, inhibit induction of apoptosis (Johnson et al, 1993; Yonish-Rouach et al, 1993; Canman et al, 1995).

The role of p53 in these molecular processes has been discussed in previous pages in this review. The involvement of p53 in apoptosis is thought to occur via upregulation of bax and downregulation of bcl-2 genes by wt p53 but not by mutated p53 proteins; Bax protein induces apoptosis and its upregulation triggers the apoptotic mechanism in cells which display elevated levels of p53 as a result, for example, of DNA damage. Down-regulation of Bcl-2 has a similar effect on the induction of apoptosis. p53 may induce apoptosis independently of transcription, although the G1 arrest by p53 requires transcription of p53 targets (reviewed by Ko and Prives, 1996). Induction of the apoptotic pathway by p53 was proposed to involve: (i) transcriptional induction of redox-related genes; (ii) formation of reactive oxygen species; and (iii) the oxidative degradation of mitochondrial components (Polyak et al, 1997). The potential of p53 in cancer gene therapy is discussed above.

While p53 and E1A activate apoptosis, Bcl-2 and E1B 19k proteins inhibit apoptosis. All four protein molecules act upstream of Bax which is a potent inducer of apoptosis: both the cellular Bcl-2 and the 19 kDa protein E1B of adenovirus are able to interact with Bax inhibiting its involvement in induction of apoptosis (Han et al, 1996; Figure 1 on page 9). E1A acts upstream of p53 by increasing the half-life of p53 resulting in an accumulation of p53 molecules in the nucleus (Lowe and Ruley, 1993); increased levels of p53 are then believed to upregulate the bax gene (Figure 1). The survival factors IL-3 and IL-6 appear to prevent p53-dependent apoptosis (see White, 1993).

p53 induces apoptosis after exposure to UV irradiation (Ziegler et al, 1994) and hypoxia (Graeber et al, 1996); this acts as a protective mechanism for the removal of severely damaged cells from the body which could become initiated cancer cells and progress to tumors. Spontaneous or radiation-induced apoptosis mediated by p53 has been shown to act for the removal of cells from the gastrointestinal tract in mice (Merritt et al, 1994) and the skin after sunburn (Ziegler et al, 1994). Epidermal growth factor (EGF) has induced apoptosis in various cancer cell lines via a novel signal transduction pathway of EGF mediated through p53 (Murayama and Horiuchi, 1997).

c-myc expression, normally induced in proliferating hematopoietic cells by mitogens, drops dramatically by mitogen withdrawal leading to cell arrest in G1. During deregulated c-myc expression, c-myc levels were not down-regulated upon mitogen withdrawal; instead, DNA synthesis continued resulting in apoptosis but not in growth arrest. The transforming segment of c-Myc was responsible for induction of apoptosis (see White, 1993).

Pax5 is a repressor of expression of the p53 gene interacting directly with a regulatory region within exon 1 of the p53 gene. At early stages during pre-B cell development the levels of Pax5 are high and p53 is down-regulated; however, later in development Pax5 levels drop and the p53 gene is activated; this process was proposed to lead to the decision of B cells to enter apoptosis or differentiate into plasma cells (Stuart et al, 1995).

Down regulation of the Cu²⁺/Zn²⁺ superoxide dismutase (SOD1) induced oxidative stress and apoptosis (Troy et al, 1996). A great deal of oxidative damage during the procedures for ex vivo-modification of cells induces their apoptosis; transfer of the Cu²⁺/Zn²⁺ superoxide dismutase to ex vivo modified cells increased their survival after implantation (see Nakao et al, 1995). This demonstrates the importance of blocking apoptotic pathways during cell manipulation for successful ex vivo gene therapy.

Gene therapy for cancer could involve restoration of the apoptotic pathway in cancer cells leading to their suicidal death; this could be effected by overexpression of the bax gene, by suppression of the endogenous bcl-2 gene (see below), or by transfer of the wt p53 gene.

C. Role of tumor necrosis factor (TNF)

The tumor necrosis factor-a (TNF-a) is a cytokine produced by macrophages, monocytes, lymphoid cells, fibroblasts and other cell types in response to inflammation and infection. TNF-a is produced by lipopolysaccharide (LPS)-stimulated macrophages; the molecular pathways leading to TNF-a production in these specialized cells involves activation by LPS of several kinases including the extracellular-signal-regulated kinases 1 and 2 (ERK1 and ERK2), p38, Jun N-terminal kinase/stress-activated protein kinase (JNK/SAPK), as well as activation of the immediate upstream MAPK activators MAPK/ERK kinases 1 and 4 (MEK1 and MEK4) and of MEK2, MEK3, and MEK6 (Swantek et al, 1997).

TNF-a binds to two type of specific receptors, TNFR1 and TNFR2, causing their trimerization and leading to activation of a number of kinases (ceramide-activated kinase, IkB kinase, Raf-1, Jun N-terminal kinases or JNKs, p38/Mpk2). Activation of Raf-1, JNK, and p38/Mpk2 contribute to the induction of AP-1 whereas activation of IkB kinase is leading to the activation of the transcription factors NF-kB. This activation leads further to upregulation of genes and induction of other cytokines, metalloproteinases, and immunoregulatory proteins (see Liu et al, 1996 and the references cited therein).

TNF can induce apoptotic death or necrosis in some tumor cells; this effect of TNF could be mediated by activation of sphingomyelinases and phospholipases, synthesis of metabolites of arachidonic acid, generation of free radicals, changes in intracellular calcium, generation of DNA strand breaks and activation of poly(ADP-ribosyl)ation, or activation of ICE-like proteases.

TNF-a, IL-1b, IFN-g, and vitamin D3 after binding to their transmembrane receptors stimulate the production of the second messager ceramide from sphingomyelin in the plasma membrane by activating sphingomyelinase; this results in a cascade of signal transduction events that result in down regulation of c-myc and induction of apoptosis, to terminal differentiation, or to RB-mediated cell cycle arrest (Figure 23).

IL-1 signaling leads to NF-kB activation and to protection against TNF-induced apoptosis. The IL-1R-associated kinase (IRAK) is homologous to Pelle of Drosophila. Two additional proximal mediators, both associating with the IL-1R signaling complex, were required for IL-1R-induced NF-kB activation: IRAK-2, a Pelle family member, and MyD88, an adaptor molecule containing a death domain (Muzio et al, 1997).

Treatment of different cell types with TNF-a results in the activation of the MEKK1 pathway of protein kinases ultimately resulting in AP-1 transcription factor activation and in the upregulation of several cytokine genes. TNF-a-stimulation also results in the activation of NF-kB and inhibition of apoptosis (Figure 24). A TNF-responsive

Figure 23. A pathway leading to the induction of growth arrest and apoptosis by the cytokines TNF-a, IL-1b, and IFN-g. The pathway is conserved between mammalian cells and yeast. Adapted from Nickels and Broach (1996). From Boulikas T (1997) Gene therapy of prostate cancer: p53, suicidal genes, and other targets. Anticancer Res 17, 1471-1506. Reproduced with the kind permission from Anticancer Research.

serine/threonine protein kinase termed GCK-related (GCKR) most likely signals via mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) kinase kinase 1 (MEKK1) to activate the SAPK pathway (Shi and Kehrl, 1997).

D. NF-kB as anti-apoptotic molecule and TNF-a signaling

Activation of NF-kB is believed to lead to the activation of antiapoptotic genes that have not been fully identified. The antiapoptotic role of NF-kB at the molecular level and the TNF-a connection consists of the following events; signaling by TNF-a induces trimerization of its receptors, an event causing three different cascades: (i) Activation of IkB kinase and activation of NF-kB, a pathway which prevents cell death. A key step for NF-kB activation leading to the activation of the stress-activated protein kinase (SAPK, also called c-Jun N-terminal kinase or JNK) is the recruitment to the TNF receptor of TNF receptor-associated factor 2 (TRAF2). (ii) induction of apoptosis via a different pathway involving activation of sphingomyelinase in plasma membrane and generation of ceramide leading to EGFR activation and induction of apoptosis; (iii) activation of MEKK1 and JNK protein kinases which is not linked to apoptotic death but to AP-1 activation (Figure 24). The antiapoptotic function of NF-kB may involve activation of the manganese superoxide dismutase and of the zinc finger protein A20; expression of these genes is induced by TNF and each of them provides protection against apoptosis (Liu et al, 1996).

bcl-2 upregulation during progression of prostate cancer was implicated in the acquisition of the androgen-independent growth; a strong antioxidant that interferes with activation of NF-kB in prostate carcinoma cells, potentiated TNF-a-stimulated apoptosis signaling through a bcl-2-regulated mechanism; based on these studies, modulation of the NF-kB survival signaling was proposed to be used to clinical advantage in the treatment of prostate cancer patients (Herrmann et al, 1997).

Transgenic mice lacking the p65 (RelA) subunit of NF-kB displayed increased apoptosis and degeneration in the liver providing further support to an apoptotic function of NF-kB (Beg et al, 1995). The TNF-induced death of mouse primary fibroblasts expressing deregulated c-Myc was inhibited by transient overexpression of the p65 subunit of NF-kB, which increased NF-kB activity in the cells (Klefstrom et al, 1997). Rel (a protooncogene, member of the NF-kB family) is implicated in both positive and negative regulation of GM-CSF expression in a variety of cell types (Gerontakis et al, 1996).

The elucidation of IL-1, TNF, IFN and other signaling pathways would lead to the discovery of new drugs causing specific inhibition; for example, members of the IL-1 signaling cascade may provide therapeutic targets for inhibiting IL-1-induced inflammation (Muzio et al, 1997).

Figure 24. TNF-a signaling via trimerization of its receptors (TNF-aR1), is causing: (i) activation of IkB kinase and activation of NF-kB, a pathway which prevents cell death via activation of the manganese superoxide dismutase and of the zinc finger protein A20. (ii) induction of apoptosis via second message ceramide (see Figure 23) and (iii) activation of JNK leading to AP-1 activation and up-regulation of cytokine genes.

E. Interleukin-1b converting enzyme (ICE) and apoptosis

The human interleukin-1b converting enzyme (ICE) is a cysteine-rich protease that can cleave the inactive 31 kDa precursor of IL-1b to generate the active cytokine; it has similarities to the C. elegans CED-3 protein. This protease plays a central role in apoptosis; the exact role and the involvement of IL-1b have not been elucidated; it is believed that signals from IL-1b, TNF-a, vitamin D3, and interferon-g, which induce an antiproliferative response, converge on sphingomyelin of plasma membrane activating a sphingomyelinase which generates a ceramide second messenger (Figure 23); in S. cerevisiae, this leads to signal transduction via activation both of a cytoplasmic protein phosphatase 2A and a protein kinase leading to down-regulation in c-myc expression and induction of apoptosis as well as RB-mediated cell cycle arrest via EGFR activation (Nickels and Broach, 1996).

At least 10 ICE-like proteases have been identified which mediate apoptotic death after their induction by a number of stimuli (see Martin and Green, 1995); these are divided into three families: (i) the ICE/CED3 family, including ICE itself; (ii) the CPP32/Yama family; and (iii) the Ich-1/Nedd2 family; they all contain the conserved QACRG pentapeptide in which the central cysteine participates in proteolytic catalysis (see Jänicke et al, 1996). Activation of these proteases by induction of apoptosis results in the cleavage of a large number of key regulatory proteins including among others poly(ADP-ribose) polymerase or PARP (Lazebnik et al, 1994), RB (Jänicke et al, 1996), PKCd (Emoto et al, 1995), Gas2 affecting microfilament reorganization (Brancolini et al, 1995), the DNA-dependent protein kinase (Casciola-Rosen et al, 1995), and the sterol regulatory element binding proteins (SREBPs) catalyzed by CPP32 ICE-like protease (PARP et al, 1996). Since cleavage of a single protein has not been shown to cause cell death it is not clear how many substrate protein molecules need to be cleaved. In addition different apoptotic pathways may exist and may operate in different cell types.

Expression of the murine ICE cDNA in Rat-1 cells induced programmed cell death and this phenomenon could be reversed by overexpression of the bcl-2 oncogene (Miura et al, 1993). Expression of members of the family of cysteine proteases related to ICE have been shown to be necessary for programmed cell death in a number of organisms (Yuan et al, 1993). Overexpression of murine ICE or of the ICE-like proteases NEDD-2/ICH-1 and Yama/apopain induced apoptosis (Miura et al, 1993). Mice lacking ICE were resistant to apoptosis induced by Fas antibody (Kuida et al, 1995).

F. Role of poly(ADP-ribose) polymerase (PARP)

PARP is a central mediator of genome integrity and transmits signals from DNA damage to recruit locally DNA repair activities (Zardo et al, 1998; Quesada, 1998, this volume). An additional role of PARP is its involvement in apoptosis causing suppression of an apoptotic endonuclease; PARP is cleaved by an ICE-like protease during TNF-induced apoptosis (Lazebnik et al, 1994). Cleavage of PARP would abort these pathways resulting in loss of recruitment of DNA repair enzymes at damaged sites but also in loss in the inhibitory function of poly(ADP-ribose) groups on key regulatory enzymes (DNA ligase, topoisomerase). It is unlikely that PARP proteolysis by an ICE-like protease is a primary event since PARP-deficient mice show normal resistance to DNA damaging agents (Wang et al, 1995).

G. Apoptosis in autoimmune disease and ischemic heart disease

T cells are produced by bone marrow and then migrate to the thymus gland where they mature. The cytotoxic or killer T cells directed against foreign bodies are released in the bloodstream; a specific apoptotic mechanism eliminates T cells directed against specific antigens on healthy cells. However, the body allows some mildly self-reactive lymphocytes to circulate; although harmless, exposure to a microbe or food antigen can stimulate them causing an expansion in their proliferation and resulting in a mild autoimmune disease. Such mild autoimmune reactions usually disappear when the stimulating antigen is cleared away; in more severe autoimmune disease, however, these lymphocytes survive longer inducing apoptosis and self-destruction in healthy cells in various tissues.

A number of classical diseases may originate by autoimmune mechanisms including initiation of atherosclerosis by apoptotic death of the epithelial cells in the arterial wall, diabetes by destruction of the pancreatic cells, lupus erythematosus, rheumatoid arthritis, and others. The mechanism via which T lymphocytes directed against self antigens defy apoptosis is not known; the mechanism might involve overexpression of the Bcl-2 gene in these lymphocytes or down-regulation of a gene encoding for the Fas ligand that sends a death message to the lymphocyte (Weih et al, 1996; reviewed by Duke et al, 1996).

Excessive necrotic death in cells of the coronary artery wall results by oxygen and glucose deprivation after blockage of a blood vessel feeding a segment of the heart (also the brain in stroke). Destructive free radicals are then produced during inflammation of the area which can cause apoptotic or necrotic death in cells in the surroundings. Since both brain and heart cells in the adult are not regenerated, Biotech Companies (for example Genentech) are focusing in developing drugs that block free radical formation, inhibit ICE-like proteases, or inhibit apoptosis via other mechanisms.

The progressive loss of neuron cells in senile or other brain diseases such as Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, and amyotrophical lateral sclerosis may ensue by apoptosis. Etiologic factors may include excessive levels of neurotransmitters, low levels of NGF, free radical-mediated damage, and deregulation in the expression of genes encoding apoptotic regulators during aging. Deregulation in apoptosis may also have a share in the induction of osteoporosis.

XXI. Genes involved in the regulation of apoptosis as targets for gene therapy

Many of the molecular controllers of apoptosis including cytokine signaling pathways (TNF-a, IL-1b), tumor suppressor proteins (p53), viral proteins (E1A of adenovirus), cellular oncoproteins (Myc), proteins that control the cell cycle (E2F), apoptosis inducers (Bax) and antiapoptotic molecules (Bcl-2, NF-kB) could constitute potential targets for pharmacological intervention for the treatment not only of cancer but of other human disease. Although for cancer treatment it is desirable to induce apoptosis, the opposite effect, that is inhibition of apoptotic pathways is desirable in the gene therapy of heart disease and degenerative brain disease (see below).

A. Gene therapy that targets bcl-2

Bcl-2 protein is overexpressed in a variety of human leukemias because of translocation of its gene to the immunoglobulin locus; Bcl-2 is associated with the outer surface of the mitochondrion and appears to be involved in scavenging oxygen radicals. Overexpression of the bcl-2 gene in tumors is thought to be responsible for the poor response of the tumors to antineoplastic drugs and radiation therapy blocking apoptosis of the tumor cells. Bcl-2 can interact with members of the Bcl-2 family including Bax, Bcl-X-S, Bcl-X-L, and Mcl-1 but also with heterologous protein molecules including BAG-1, Raf-1, and R-Ras.

Introduction of the bcl-2 gene into human diploid breast epithelial MCF10A cells (containing the wild-type p53 gene) resulted in suppression in p21 gene expression although the level of expression of p53 was not affected; these studies suggested that Bcl-2 may inhibit the functional activity of p53 protein and might regulate the commitment of cells to commit suicide or proliferate (Upadhyay et al, 1995).

Overexpression of the bcl-2 gene in tumors is thought to be responsible for the poor response of the tumors to antineoplastic drugs and radiation therapy blocking apoptosis of the tumor cells; therefore, down-regulation of the bcl-2 gene specifically in tumor cells could induce apoptosis. Primary untreated human prostate cancers were found to express significant levels of this apoptosis-suppressing oncoprotein; this is a striking difference with normal prostate secretory epithelial tissue not expressing Bcl-2 (Raffo et al, 1995). Transfection of LNCaP human prostate cancer cells with a plasmid expressing bcl-2 rendered these cells highly resistant to a variety of apoptotic stimuli (serum starvation or treatment with phorbol ester) and induced earlier and larger tumors in nude mice. The ability of Bcl-2 to protect prostate cancer cells from apoptotic stimuli correlated with the ability of the cells to form hormone-refractory prostate tumors in nude mice (Raffo et al, 1995).

The Bcl-2 oncoprotein suppresses apoptosis and, when overexpressed in prostate cancer cells, makes these cells resistant to a variety of therapeutic agents, including hormonal ablation. Overexpression of BCL-2 is common in non-Hodgkin lymphoma leading to resistance to apoptosis and promoting tumorigenesis. Therefore, bcl-2 provides a strategic target for the development of gene knockout therapies to treat human prostate cancers (Dorai et al, 1997) and non-Hodgkin lymphomas (Webb et al, 1997).

Down-regulation of Bcl-2 can be accomplished with antisense. In patients with relapsing non-Hodgkin lymphoma, BCL-2 antisense therapy led to an improvement in symptoms; antisense oligonucleotides targeted at the open reading frame of the BCL-2 mRNA showed effectiveness against lymphoma grown in laboratory animals and has entered human clinical trials. The first study was conducted on nine patients with BCL-2-positive relapsed non-Hodgkin lymphoma using a daily subcutaneous infusion of 18-base, fully phosporothioated antisense oligonucleotide administered for 2 weeks (Webb et al, 1997). A local inflammation at the infusion site was noted. A reduction in tumour size was observed in two patients (one minor, one complete response) using computed tomography scans; in two other patients, the number of circulating lymphoma cells decreased during treatment. In four patients, serum concentrations of lactate dehydrogenase fell, and in two of these patients symptoms improved (Webb et al, 1997).

A divalent hammerhead ribozyme, constructed by recombining two catalytic RNA domains into an antisense segment of the coding region for human bcl-2 mRNA was able to rapidly degrade bcl-2 mRNA in vitro; it was then tested for its ability to eliminate bcl-2 expression from hormone-refractory prostate cancer cells. When this hammerhead ribozyme was directly transfected into cultured prostate cancer cells (LNCaP derivatives), it significantly reduced bcl-2 mRNA and protein levels within 18 hr of treatment and induced apoptosis in a low-bcl-2-expressing variant of LNCaP, but not in a high-bcl-2-expressing LNCaP line (Dorai et al, 1997).

B. Bcl-xs

Many cancers overexpress a member of the Bcl-2 family of inhibitors of apoptosis, such as Bcl-2 and Bcl-xL. Members of the Bcl-2 family were found to be essential for survival of cancer cells derived from solid tissues including breast, colon, stomach, and neuroblasts (Clarke et al, 1995). On the contrary, Bcl-xs is a dominant negative repressor of Bcl-2 and Bcl-xL; thus, Bcl-xs induces apoptosis. Transient overexpression of Bcl-xs in MCF-7 human breast cancer cells, which overexpress Bcl-xL, with a replication-deficient adenoviral vector induced apoptosis in vitro; intratumoral injection of the bcl-xs adenovirus on solid MCF-7 tumors in nude mice showed a 50% reduction in size with evident apoptotic cells at sites of injection (Ealovega et al, 1996).

An adenovirus vector expressing bcl-xs specifically and efficiently killed carcinoma cells arising from multiple organs including breast, colon, stomach, and neuroblasts even in the absence of an exogenous apoptotic signal such as x-irradiation. In contrast, normal hematopoietic progenitor cells and primitive cells capable of repopulating SCID mice were not killed by the bcl-xs adenovirus. Thus, transfer of the bcl-xs gene could be used in killing cancer cells contaminating the bone marrow of patients undergoing autologous bone marrow transplantation (Clarke et al, 1995).

C. E2F-1 and TNF-a gene transfer

E2F cooperates with p53 to induce apoptosis; high levels of wild-type p53 potentiate E2F-induced apoptosis in fibroblasts. The physiological relevance of E2F in the apoptotic mechanism is thought to arise from the ability of E2F to act as a functional link between p53 and RB; p53 levels increased in response to high levels of E2F. Targeted disruption of the E2F-1 gene yields transgenic animals with an excess of mature T cells due to a defect in lymphocyte apoptosis (Field et al, 1996).

Overexpression of the transcription factor E2F-1 could induce apoptosis in quiescent rat embryo fibroblasts in a p53-dependent manner; however, Hunt et al (1997) have shown that overexpression of the E2F-1 gene after adenoviral transfer can mediate apoptosis in the absence of wild-type p53: adenovirus-mediated transfer of the E2F-1 gene under control of the CMV promoter to human breast and ovarian carcinoma cell lines resulted in the induction of significant morphological changes in four of the five cell lines that had mutations in the p53 gene within 48 h of transduction characteristic of apoptosis.

Retroviral vector-mediated transfer of the TNF-a gene into the DNA of human tumor cells induced apoptosis in high- TNF-a-producing clones generated from a human lymphoma T-cell line (ST4); the apoptotic death of the cells was associated with a downregulation of the apoptosis-preventing gene, bcl-2, while the expression of bax and p53 genes persisted (Gillio et al, 1996).

D. E6, E7 of human papillomavirus (HPV)

E6 and E7 of HPV possess transforming ability, have been shown to interact with the cellular tumor suppressors p53 and RB (Werness et al, 1990; Dyson et al., 1989) and are believed to play a central role in HPV-induction of cervical carcinogenesis as well as in the maintenance of the malignant phenotype. Viruses have developed strategies to shut down protein synthesis in the host and subdue its protein synthesizing machinery to produce progeny virus when infecting cells. Because virus-infected cells commit suicide to protect the organism from further infection viruses have evolved mechanisms to prevent apoptosis of the host cell ensuring their propagation; E6 protein interacts with p53 to exclude p53 molecules from their apoptotic functions and to inhibit apoptosis in HPV-infected cells thus giving to HPV a proliferation advantage.

27-mer phosphorothioate oligodeoxynucleotides (oligos) targeting the ATG translational start region of HPV-16 E6 and E7 sequences showed antiproliferative effects in all HPV-16-positive cell lines tested and in primary cervical tumor explants while the endometrial and two ovarian primary tumors as well as the HPV-negative C33-A cell line and HPV-18-positive cell line HeLa were relatively insensitive to the HPV-16 oligos (Madrigal et al, 1997).

E. Prevention of apoptosis for gene therapy of heart disease and for ex vivo manipulations of therapeutic cells

As induction of apoptosis is the desired effect for the gene therapy of cancer, prevention of apoptosis by gene therapy can fight heart disease. Cardiomyocyte death results from heart ischemia proceeding via necrosis and from reperfusion which induces additional cardiomyocyte death by apoptosis; prevention of apoptosis would constitute an important target for fighting heart disease. Prevention of apoptosis should also solve a major problem in cell culture cells which are subject to oxidation damage during their manipulation for ex vivo gene transfer and most important during the step of reimplantation, encapsulation in biopolymer membranes for surgical implantation, and similar processes. Prevention of apoptosis could be effected by transfer and overexpression of the bcl-2 gene. Also prevention of oxidative damage during reimplantation of ex vivo-modified cells could be reduced by transfer and overexpression of the Cu/Zn superoxide dismutase gene (Nakao et al, 1995).

Overexpression of bcl-2 delayed onset of motor neuron disease and prolonged survival in a transgenic mouse model of familial amyotrophic lateral sclerosis (Kostic et al, 1997).

XXII. E1A and HER-2/neu (c-erbB-2) in cancer gene therapy

A. HER-2/neu

The human epidermal growth factor receptor-2 (HER2), a membrane tyrosine kinase highly expressed in many epithelial tumors, could be a target for cancer gene therapy. The HER-2/neu (also called c-erbB-2) proto-oncogene is overexpressed in many human cancer cells, including those of breast cancer and ovarian cancer correlating with lower survival rate in ovarian cancer patients; amplification or overexpression of HER-2/neu has also been observed in human lung cancer and has been correlated with poor prognosis and chemoresistance.

A reversible transformation of NIH3T3 fibroblasts by overexpression of the HER2/c-erbB2 receptor tyrosine kinase under control of a tetracycline-responsive promoter has been demonstrated in tissue culture; induction of HER2 expression resulted in cellular transformation in vitro and treatment of transformed cells with the effector anhydrotetracyline switched-off HER2 expression and induced morphological and functional changes characteristic for non-transformed cells (Baasner et al, 1996).

B. E1A-based gene therapy

E1A-based gene therapy approaches are now in clinical trials (see below); the molecular mechanism behind this approach is that the E1A protein of Adenovirus 5 represses HER-2/neu transcription and functions as a tumor suppressor gene in HER-2/neu-overexpressing cancer cells. Breast cancer cells that overexpress HER-2/neu are more resistant to chemotherapeutic agents such as paclitaxel (Taxol) and docetaxel (Taxotere) than those that do not overexpress HER-2/neu; paclitaxel sensitivity correlated with HER-2/neu expression level in a panel of mouse fibroblasts expressing different levels of HER-2/neu; downregulation of HER-2/neu expression by E1A sensitized the cells to paclitaxel. Transfer the E1A gene into two human breast cancer cell lines that overexpress HER-2/neu and E1A gene transfer sensitized these cells to the drug by repressing HER-2/neu expression (Ueno NT et al, 1997).

Increased HER-2/neu expression led to more severe ovarian malignancy and increased metastatic potential in animal models; the adenovirus 5 E1A gene repressed HER-2/neu gene expression and suppressed growth of human ovarian cancer SKOV-3 cells, which overexpress HER-2/neu, in cell culture (Yu et al, 1995). Intraperitoneal injection of SKOV-3 cells into female nu/nu mice elicited tumors and the animals died within 160 days of severe tumor symptoms; cationic liposome-mediated delivery of the E1A gene into adenocarcinomas that developed in the peritoneal cavity and on the mesentery of the mice significantly inhibited growth and dissemination of ovarian cancer cells; about 70% of the treated mice survived at least for 365 days (Yu et al, 1995).

Regulatory regions derived from the 5' flank of the human prostate-specific antigen (PSA) gene were inserted into adenovirus type 5 DNA to drive the expression of the E1A gene; infection of cells in culture with this recombinant adenovirus was able to drive the expression of the E1A gene only in cell lines which expressed PSA such as the human LNCaP cells but not in human DU145 cells which do not express PSA; the recombinant adenovirus destroyed large LNCaP tumors (1x10⁹ cells) and abolished PSA production in nu/nu mouse xenograft models after a single intratumoral injection (Rodriguez et al, 1997).

A replication-deficient adenovirus containing the E1A gene, Ad.E1A⁺, was used to transduce E1A into HER-2/neu-overexpressing and low expressing human lung cancer cell lines and shown a better therapeutic efficacy in HER-2/neu-overexpressing cells. The cell culture studies were then extended to animal studies: tumor-bearing mice established by intratracheal injection of lung cancer cells overexpressing HER-2/neu and treated by i.v. tail injections of Ad.E1A⁺ showed suppression of the intratracheal lung cancer growth. However, no significant tumor suppression effect was observed in mice bearing a low HER-2/neu-expressing cell line with the same regimen (Chang et al, 1996).

C. Clinical trials with E1A and c-Erb-B2

Liposome-mediated E1A gene transfer suppressed tumor development and prolonged survival of mice bearing human breast cancer cells overexpressing HER-2/neu. These studies resulted in the initiation of a phase I clinical trial using an E1A-liposome complex administered to patients with HER-2/neu-overexpressing breast or ovarian cancer (Protocol 205 in Table 4 of following article, pages 203-206). The principal investigators are Drs. Hortobagyi, Lopez-Berstein, and Hung at MD Anderson Cancer Center, Houston, Texas). The safety of this regimen was shown by intraperitoneal injection of E1A/liposomes in normal mice and at cumulative doses 5 to 40 times the DNA-lipid starting dose proposed for the phase I clinical trial (Xing et al, 1997). A Phase I multicenter study of intratumoral E1A gene therapy using cationic liposome gene transfer is also in course for patients with unresectable or metastatic solid tumors that overexpress HER-2 /neu (protocol 209, see page 205).

Delivery of an anti-erbB-2 single chain (sFv) antibody gene for previously treated ovarian and extraovarian cancer patients is in clinical trials using adenoviral gene delivery (protocol #133). A clinical trial for tumor vaccination with HER-2 /Neu using a B7 expressing tumor cell line prior to treatment with HSV-tk gene-modified cells is in phase I for ovarian cancer (protocol #96, page 165).

XXIII. Suicidal genes for cancer therapy (prodrug gene therapy)

A. Molecular mechanism of cell killing with HSV-tk gene and ganciclovir (GCV)

Expression of genes encoding prodrug-activating enzymes can increase the susceptibility of tumor cells to prodrugs, and may ultimately achieve a better therapeutic index than conventional chemotherapy (Table 3). Direct suppression of tumor growth by cytotoxic gene therapy is a successful gene transfer approach. This approach has promise for a variety of other applications where excess cell proliferation is detrimental and has also been used to restrict intimal hyperplasia of the arterial wall and smooth muscle cell growth to limit restenosis after artery angioplasty (see below).

Cancer cells can be induced to be conditionally sensitive to the antiviral drug ganciclovir after their transduction with the thymidine kinase (tk) gene from the herpes simplex virus (HSV); ganciclovir (GCV) is the 9-{[2-hydroxy-1-(hydroxymethyl)-ethoxy]methyl}guanine (Field et al, 1983); it is converted by HSV-tk into its monophosphate form which is then converted into its triphosphate form by cellular enzymes and is then incorporated into the DNA of replicating mammalian cells leading to inhibition in DNA replication and cell death (Moolten, 1986; Borrelli et al, 1988; Moolten and Wells, 1990). It is only viral TK, not the mammalian enzyme, that can use efficiently ganciclovir as a substrate and this drug has been synthesized to selectively inhibit herpes virus replication (Field et al, 1983); indeed, the mammalian TK has a very low affinity for this guanosine analog. The toxicity of ganciclovir is manifested only when cells undergo DNA replication and it is not harmful to normal nondividing cells. This treatment strategy has been used for hepatocellular carcinoma (Huber et al, 1991; Su et al, 1996), fibrosarcoma, glioma (Culver et al, 1992, see below), adenocarcinoma (Osaki et al, 1994), prostate cancer (Eastham et al, 1996) and many other cancers.

B. Treatment gliomas in rats with HSV-tk plus ganciclovir

Brain tumors have the privilege of escaping immunologic rejection; therefore brain tumors are inaccessible to cancer immunotherapy. Culver and cowor-

The replication and expression of hepatitis B virus (HBV) could be inhibited through antisense gene transfer and this could become a new method for clinical gene therapy against HBV; infection of the human hepatoblastoma cell line 2.2.15, which expresses HBV surface antigen and releases HBV particles, with retroviral vectors carrying an antisense preS/S or preC/C genes of HBV inhibited expression of the surface antigen (Ji and St 1997).

Phosphorothioate antisense oligos directed against c-myc and p53 in different cell lines (CAOV-3, SKOV-3, and BG-1) were shown to have both antiproliferative and stimulatory activity, as single agents and in combination; it was concluded that further in vitro studies are needed before considering clinical trials with these agents in gynecologic cancers (Janicek et al, 1995).

Transfection of antisense cDNA constructs encom-passing different regions of the c-erbB-2 gene in the lung carcinoma cell line Calu3, which overexpresses the c-erbB-2 oncogene, reduced significantly anchorage-independent growth and tumor size in nude mice (Casalini et al, 1997).

Antisense oligonucleotides against PCNA and cdc2 kinase transferred into injured arterial walls by protein-liposomes greatly reduced mRNA levels for those genes and inhibited neointima formation of the injured artery for 8 weeks; double-stranded oligonucleotides containing the consensus sequence for E2F binding sites also inhibited the growth of smooth muscle cells and prevented neointima formation (Kaneda et al, 1997). Antisense oligonucleotides to angiotensinogen1-receptor mRNA and to angiotensinogen mRNA reduced blood pressure (Tomita et al, 1995; Phillips, 1997; Phillips et al, 1997).

XXVI. Triplex gene therapy

A. Molecular mechanisms for triplex formation

Natural purine^.pyrimidine sequences in regulatory regions of genes in eukaryotic cells with a mirror symmetry can form triple-helical structures; in addition, purine-rich segments in DNA unable to form triple helices on their own can be targeted by DNA or RNA oligonucleotides able to form triplex structures with their target DNA and these unusual structures can inhibit transcription factor binding, transcription initiation, and nuclear enzymatic activities. Understanding the advantages, limitations and pitfalls for using oligonucleotides as gene bullets, development of strategies for boosting their therapeutic efficiency, their covalent linkage to DNA damaging molecules to hit a specific genomic DNA sequence, and improvements to the methods for their delivery to cells could make reality their use as tools of micro-targeting specific genomic sites and as pharmacogenomic drugs.

Formation of triple helical DNA was found to take place on AT- and GC-rich stretches. A pyrimidine third strand oligonucleotide, studied by NMR and other approaches, interacts with purine residues in the major groove of the target duplex in a parallel orientation (Moser and Dervan, 1987; Rajagapol and Feigon, 1989; de los Santos et al, 1989) whereas a purine oligonucleotide binds in an antiparallel orientation relative to the purine strand in the duplex (Cooney et al, 1988; Kohwi and Kowhi-Shigematsu, 1988; Beal and Dervan, 1991). In this case G can recognize GC pairs and A or T can recognize AT pairs. Specificity is provided from T^.AT and C⁺GC base triplets where the bases of the third polypyrimidine strand establish Hoogsteen base pairing with the purine strand of the duplex (Hoogsteen, 1959; Rajagopal and Feigor, 1989).

The H form is the structural basis for S1-nuclease hypersensitivity (Mirkin et al, 1987). A restriction fragment from a human U1 gene containing the sequence d(C-T)₁₈.d(A-G)₁₈ under supercoiling and pH less than or equal to 6.0 showed S1 hyperreactivity in the center and at one end of the (C-T)_n tract, and continuously from the center to the same end of the (A-G)n tract providing strong support for a triple-helical model (Johnston, 1988).

Homopyrimidine oligodeoxyribonucleotides with EDTA-Fe attached at a single position bound the corresponding homopyrimidine-homopurine tracts within large double-stranded DNA by triple helix formation and cleaved at that site (Moser and Dervan, 1987). Studies from the group of Claude Hélène have similarly focused on the development of artificial scissor oligonucleotides based on triplex technology (Praseuth et al, 1988; Perrouault et al, 1990). However, the feasibility of employing this exciting in vitro technology to animal studies has not yet been demonstrated.

Intramolecular or intermolecular triple helices could be recognized by specific proteins that stabilize triplex structures and might play a role in gene regulation; a protein from HeLa cell nuclear extracts was identified that binds to a 55 nucleotide-long DNA oligomer that could fold on itself to form an intramolecular triple helix of the Py Pu x Py motif (Guieysse et al, 1997).

Triplex-forming oligophosphoramidates containing thymines and cytosines or 5-methyl cytosines (5' T₄CT₄C₆T 3') bind strongly to a 16 base pair oligopurine^.oligopyrimidine sequence of HIV proviral DNA even at neutral pH and are remarkably stable compared to oligonucleotides with natural phosphodiester linkages. The phosphoramidate oligomers induced an efficient arrest of both bacteriophage and eukaryotic transcriptional machineries (SP6, T7 or Pol II) under conditions where the phosphodiester oligos had no inhibitory effect and blocked the RNA polymerases at the triplex site (Giovannangeli et al, 1996).

Oligonucleotide-directed triplex formation has been shown to inhibit binding of transcription factors to their cognate DNA sequences. A 21 bp homopurine element insert flanking a single Sp1 site in the adenovirus E4 promoter was used to study the effect of oligo targeting on transcriptional efficiency in vitro; assembly of the triple helical complex repressed basal transcription by rendering the triplex target inflexible and by blocking assembly of the promoter into initiation complexes; Sp1 was unable to cause derepression (Maher et al, 1992). Thus DNA triplexes can inhibit transcription initiation not only when directed to a TF binding site occluding its binding but also to a flanking region by other possible repression mechanisms including stiffening of the double helix (Maher et al, 1992).

B. Triplex targeting of IGF-I

Oligonucleotide-directed triple helix formation targeted toward IGF-I to inhibit its expression was studied following stable transfection of C6 rat glioblastoma cells with a plasmid from which an RNA was transcribed that coded for the third strand of a potential triple helix. A plasmid encoding the oligopurine variant of the triple helix but not the oligopyrimidine or a control sequence caused a dramatic reduction of IGF-I RNA and protein levels in cultured cells, morphological changes, and increased expression of protease nexin I and MHC class I molecules; the transfected cells displayed a reduced capacity for tumor growth when injected in nude mice (Shevelev et al, 1997).

XXVII. Gene transfer to some characteristic tissues or cell types

A. Transduction of hematopoietic stem cells (HSCs)

Hematopoietic stem cells (HSCs), which can be isolated with high speed flow-cytometric cell sorting from fetal or adult bone marrow and cytokine-mobilized peripheral blood, have extensive self renewal and multilineage repopulating potential; HSCs are being used as an hematopoietic graft to treat cancer patients undergoing high dose chemotherapy which eradicates HSCs; GM-CSF treatment of the patient can enhance mobilization of true HSCs; furthermore, HSCs can be stably transduced at high efficiency (32-75%) by co-culture with a cell line producing recombinant retroviruses containing the neomycin-resistant gene and are targets for hematopoietic cell-based gene therapy especially for the treatment of patients with multiple myeloma (Chen et al, 1995).

The efficiency of gene transfer into monkey pluripotent hematopoietic stem cells (PHSCs) is at least one order of magnitude lower than what has been achieved in mice because primate PHSCs seem to require quite different culture conditions for their maintenance and transduction than mouse PHSCs. Successful retroviral vector-mediated gene transfer into monkey PHSCs supported maintenance of the long-term repopulating ability of autologous monkey grafts and has closed the gap between gene transfer experiments in mouse models and primates opening the door to the clinical application of bone marrow gene therapy to humans (Van Beusechem and Valerio, 1996).

Retroviral vectors pseudotyped with vesicular stomatitis G glycoprotein (VSV-G) and expressing a murine cell surface protein, B7-1, were used to infect the human T-cell line Jurkat and human peripheral blood lymphocytes (PBLs); the transduction efficiency of PBLs with the pseudotyped vector reached a maximum of 16-32% at an moi of 40 (Sharma et al, 1996). Introduction of a mutant H-ras gene (along with a neomycin resistance gene) into normal human bone marrow progenitor cells with a retrovirus followed by selection in cell culture with G418 suggested that expression of mutant H12-ras resulted in enhanced proliferation of early myeloid cells at the expense of differentiation (Maher et al, 1994).

Dendritic cells (DCs) which are the most potent antigen-presenting cells (APCs) for the initiation of antigen-specific T-cell activation can be highly enriched from peripheral blood-adherent leukocytes by short-term culture in the presence of IL-4 and GM-CSF; adenoviral vectors expressing luciferase, b-galactosidase, IL-2, and IL-7 readily transduced human DCs compared to other methods (Arthur et al, 1997).

Transduction of hematopoietic stem cells with human IL-1Ra cDNA was used to alleviate symptoms of RA; the HSCs were subsequently injected into lethally irradiated mice; all of the mice survived and over 98% of the white blood cells in these mice produced biologically active human IL-1Ra type from 2-13 months after transplantation; the animals had the human IL-1Ra protein in their sera for at least 15 months (Boggs et al, 1995).

B. Gene transfer to the brain

Several molecular approaches, including gene transfer with retroviral, adenoviral and herpes simplex virus vectors, as well as antisense vectors, and antisense oligonucleotides have been shown to have in vitro and in vivo activities against brain tumor cells. These approaches are especially important for the treatment of glioblastomas which remain incurable despite an aggressive combination regimens using surgery, radiation, and chemotherapy (reviewed by Yung, 1994).

Intrathecal transplantation of polymer-encapsulated cell lines genetically engineered to release the human ciliary neurotrophic factor (CNTF) provided a means to deliver CNTF continuously behind the blood-brain barrier and bypass the immunologic rejection of allogeneic cells; for example, transduction of mouse C2C12 myoblasts with human CNTF followed by membrane encapsulation and intrathecal implantation in adult rats could partially rescue motor neurons from axotomy-induced cell death (Deglon et al, 1997).

Since adult brain cells are nonproliferative, they are refractory to retroviral infection that could deliver the tyrosine hydroxylase gene to the brain to alleviate degeneration at the nigrostriatal pathway in Parkinson disease (PD). Implantation of immortalized fibroblasts, primary fibroblasts, or myoblasts, stably transfected in culture with the TH gene (Jiao et al, 1993) or direct injection of lipofectin-plasmid DNA complexes containing the TH gene under the influence of the SV40 promoter/enhancer (Cao et al, 1995) reduced behavioral abnormalities in PD animal models. A 7 kb region encompassing the TH promoter was able to confer expression of b-galactosidase in catecholaminergic cell types in the substantia nigra pars compacta compared to other regions of the brain after HSV-1-mediated transfer to adult rat brains (Song et al, 1997).

C. Gene transfer to hepatocytes

Hepatocytes are responsible for the production of many therapeutically important proteins such as LDL-R which clears LDL from the serum and the blood clotting Factors VIII and IX which are defective in hemophiliacs. Portal vein, rather than systemic intravenous injection, has been used to transduce preferentially hepatocytes (or liver macrophages, known as Kuppfer cells). For example, the Factor IX gene was delivered to a portal vein cannulated into a splenic vein in animals previously subject to two-third hepatectomy and resulted in the expression of low levels of factor IX for up to about 5 months; 0.3-1% of hepatocytes were found to be transduced (Kay et al, 1993).

An adenovirus LDL-R cDNA, infused into the portal vein of rabbits deficient in LDL receptor, resulted in the expression of human LDL-R protein in the majority of hepatocytes that exceeded the levels found in human liver by at least 10-fold (Kozarsky et al, 1994). According to an ex vivo protocol, cultured hepatocytes from a FH patient were transduced ex vivo with the LDL receptor gene and transplanted by infusion into the portal vein of the patient (Wilson et al, 1992; Grossman et al, 1994).

Delivery of a 5.6 kb genomic clone or of a 834-bp cDNA clone encoding the kallikrein gene into the portal vein or tail vein of spontaneously hypertensive rats resulted in significant reduction of their blood pressure for about 5-6 weeks (Chao et al, 1996). Portal vein injection of the human kallistatin cDNA in an adenoviral vector into spontaneously hypertensive rats resulted in a significant reduction of blood pressure for 4 weeks; this method resulted in the transduction not only of liver but also of spleen, kidney, aorta, and lung (Chen et al, 1997).

Hepatocyte Growth Factor (HGF) is the most potent mitogen of mature hepatocytes; transfer of the human HGF gene into a recombinant retroviral cell line produced HGF in the supernatant which was correctly processed and biologically active; primary mouse and human hepatocytes could be transduced with the supernatant of transfected cells and, thus, this cell line should be useful for in vivo liver gene therapy (Pages et al, 1996a).

D. Gene transfer to the embryo

Introduction of normal genes in utero or in the early postnatal period could become a successful approach to correct genetic defects; several studies have shown that adenoviral or retroviral vector-mediated gene transfer during the ebryonic or neonatal period results in prolonged gene expression. Gene transfer (or gene disruption) has been extensively studied in preimplantation embryos giving rise to transgenic animals difficient in a specific protein (e.g. Smith et al, 1995; Fong et al, 1995; Shalaby et al, 1995).

Gene transfer to the embryo has shown the importance of the promoter, large genomic regulatory regions, cell-cell interactions and gene switch taking place during embryogenesis in maintaining transgene expression in different tissues; results obtained in embryos reflect the in vivo patterns of tissue-specific expression which could be useful to direct efforts in promoter choices for somatic gene transfer to the adult (as is the case for most gene therapy applications). Furthermore these studies provide the foundation of a new era where genetic manipulation of the embryo could permanently correct monogenic genetic disorders such as hemophilias, thalassemias and others.

The promoter of the tie gene, which encodes a receptor tyrosine kinase that is expressed in the endothelium of blood vessels, was used to drive the expression of a luciferase reporter gene construct; in cultured cells the luciferase activity was not restricted to endothelial cells. In contrast, in transgenic mice expression of the reporter b-galactosidase was restricted to endothelial cells undergoing vasculogenesis and angiogenesis; in adult transgenic mice, tie promoter activity in lung and many vessels of the kidney was as high as in the vessels of the corresponding embryonic tissues, whereas in the heart, brain and liver, tie promoter activity was downregulated and restricted to coronaries, cusps, capillaries, and arteries (Korhonen et al, 1995).

A retroviral VEGF expression vector was used to infect quail ebryo and to increase the level of VEGF during critical periods of avian limb bud growth and morphogenesis. Overexpression of VEGF in the limb bud exclusively resulted in hypervascularization as reflected by an increase in vascular density from an augmentation of the VEGF signaling mechanism in a permissive environment; vascular permeability was also dramatically increased leading to local edema (Flamme et al, 1995).

An avian leukosis virus (ALV)-based retroviral vector system was used for the efficient delivery of genes into preimplantation mouse embryos; a subset of the integrated proviruses expressed the delivered chloramphenicol acetyltransferase (CAT) reporter gene either from the constitutive viral promoter contained in the long terminal repeat or from the internal nonviral tissue-specific promoter in different sets of experiments. Thus, many of the sites that are accessible to viral DNA insertion in preimplantation embryos were thought to be incompatible with expression in older animals (Federspiel et al, 1996).

Baldwin and coworkers (1997) have found that the expression of lacZ gene under control of CMV or RSV promoter transferred to early, postgastrulation mouse embryos gave tissue-specific patterns of expression which depended on the type of promoter used. Embryos were injected into the mesoderm of the neural fold (A in Figure 30) and b-galactosidase activity was detected in the head

Table 5. Ex vivo studies on animal models

Gene target	Human disease	Method	Animal model, objective, and method	Results	Reference
ADA	SCID (severe combined immunodeficiency)	Retrovirus	Immunodeficiient mice were injected with peripheral blood lymphocytes from ADA- patients transduced with a retroviral vector for human ADA	Restoration of immune functions (presence of human immunoglobulin and antigen-specific T cells)	Ferrari et al, 1991
bcl-2	prostate cancer		bcl-2 expressing LNCaP human prostate cancer cells are rendered highly resistant to apoptotic stimuli	LNCaP-bcl-2 cells induced earlier, larger, and hormone-refractory prostate tumors in nude mice	Rafo et al, 1995
Factor IX	hemophilia B		Injection of transduced primary myoblasts into the muscle	Factor IX was being synthesized and delivered to the circulation for over 6 months	Dai et al, 1992
Factor IX	hemophilia B	retr	Transplantation of retrovirus-transduced keratinocytes	Human factor IX was detected in the bloodstream of nude mice in small quantities for one week	Gerrard et al, 1993,
Factor IX	hemophilia B	retrovirus	Mouse primary myoblasts were infected with retrovirus expressing the canine factor IX under control of mouse muscle creatine kinase and human CMV promoter; myoblasts were injected into the hindlegs of recipient mice; secreted canine factor IX was monitored in the plasma	Sustained expression of factor IX for over six months without any apparent adverse effects on the recipient mice; however, the levels of the factor IX protein secreted into the plasma (10 ng/ml for 107 injected cells) were not sufficient to be of therapeutic value; 100 times higher amounts of factor IX were needed	Dai et al, 1992; Yao et al, 1994
Factor VIII	Hemophilia A	transferrin	Transfection of fibroblasts and myoblasts with B-domain-deleted factor VIII gene followed by implantation into mice	Therapeutic levels of factor VIII in the blood of the animals for 24 hours	Zatloukal et al, 1994
Factor VIII	Hemophilia A	Retrovirus	Mouse primary fibroblasts infected with a recombinant retrovirus containing factor VIII gene deleted at the B domain	Therapeutic levels of factor VIII in blood of animals for 1 week after surgical implantation into the peritoneal cavity in SCID mice of 15 million cells in the form of neo-organs	Dwarki et al, 1995
Growth hormone (human)	none	mice electroporation	Ex vivo modified C2C12 cells with the hGH gene under control of the inducible UAS promoter and a synthetic hybrid steroid receptor (TAXI), activating transcription from the inducible promoter after treatment with the synthetic nontoxic drug inducer RU486; transplanted in mouse muscle	This model allows up to 100-fold induction of the hGH gene and can be finely tuned to lower levels of induction	Delort and Capecchi, 1996
Growth hormone (human, hGH)	general	retr	Injection of genetically engineered myoblasts into mouse muscle	hGH could be detected in serum for 3 months; myoblasts were fused into preexisting multinucleated myofibers that were vascularized and innervated	Dhawan et al, 1991
HSV-tk	glioma	Retr	To directly transfer HSV TK gene and kill transduced proliferating brain tumor cells with ganciclovir without affecting nondividing normal cells	Murine fibroblasts transduced ex vivo with HSV TK retroviral vectors caused complete regression of gliomas in rat brain after intratumor injection	Culver et al, 1992
HSV-tk	pancreatic cancer	retrovirus	BXPC3 primary human pancreatic adenocarcinoma cells were transduced with retroviral vector carrying the HSV-tk gene driven by the CEA promoter; engrafted subcutaneously into nude mice eliciting pancreatic tumors	Animals treated with 0.1 mg/Kg ganciclovir exhibited a significant reduction in tumor growth	DiMaio et al, 1994
HSV-tk	proliferative vitreoretinopathy (PVR)	retrovirus-transduced rabbit dermal fibroblasts	Traction retinal detachment results from proliferation of retinal pigment cells in the vitreous cavity of the eye; PVR may ensue after retinal surgery or trauma and can be induced in rabbit models by surgical vitrectomy.	Significant inhibition of PVR (killing of proliferating cells in the retina) was observed in rabbit PVR models after injection into the vitreous cavity of rabbit dermal fibroblasts transduced in vitro; all eyes received 0.2 mg GCV on the following day and on day 4;	Kimura et al, 1996
IL-1 -receptor antagonist protein gene	Rheumatoid arthritis (RA)	Retr	Synovial cells were surgically removed from joints of animals with experimental arthritis, cultured, transduced with the IL-1 -receptor antagonist protein gene and reimplanted into the respective donors by intraarticular injection	Improvement in RA symptoms	Bandara et al, 1993
IL-1 receptor antagonist (IL-1Ra)	Rheumatoid arthritis (RA)	Retr	Degradation of cartilage in RA is stimulated by IL-1; to inhibit IL-1; RA synovial fibroblasts transfected with the IL-1Ra gene were coimplanted with normal human cartilage in SCID mice	IL-1Ra expression protected the cartilage from chondrocyte-mediated degradation.	Otani et al, 1996; Makarov et al, 1996; Muller-Ladner et al, 1997b
IL-2; GM-CSF	prostate cancer		Dunning rat R3327-MatLyLu prostate tumor model (an anplastic androgen-dependent, nonimmunogenic tumor that metastasizes to the lymph nodes and the lung); cytokine (IL-2)-secreting human tumor cell preparations (tumor vaccines) were used for the treatment of advanced human prostate cancer in rats	All animals with subcutaneously established tumors were cured; the cancer vaccine induced immunological memory that protected the animals from subsequent tumor challenge; GM-CSF was less effective than IL-2.	Vieweg et al, 1994
LDL receptor	Familial hypercholesterolemia (FH)	Retr	Watanabe heritable hyperlipidemic rabbit (deficient in both alleles of LDL receptor gene); establish hepatocyte culture from animal liver; transduce with LDL receptor gene responsible for LDL internalization into hepatocytes to reduce blood serum cholesterol; transplant hepatocytes into the animal	30-40% decrease in serum cholesterol that persisted for 4 months	Chowdhury et al, 1991
Nerve Growth Factor (NGF)	Alzheimer's disease	rat and primate	Delivery of NGF by ex vivo-modified allogeneic cells surrounded by a semipermeable membrane and implanted intrathecally	Release of NGF by the implant which is not subject to immunologic rejection due to the membrane	Kordower et al, 1994
XPD (ERCC2)	xeroderma pigmentosum (XP)	Retr	To transduce ex vivo human keratinocytes and produce skin grafts on immunodeficient mice; use it on XP patients as reconstructive surgery to alleviate cancers in UV-exposed areas	The retroviral vector carrying the XPD gene and neoR under control of SV40 enhancer fully complemented the DNA repair deficiency of primary skin fibroblasts	Gözükara et al, 1994; Carreau et al, 1995
TH (Tyrosine hydroxylase)	Parkinson's disease (PD)	Retrovirus	Rat fibroblasts transduced with tyrosine hydroxylase (TH) produced and released L-dopa to the culture medium;	When grafted to the striatum of Fischer rats with a prior 6-hydroxydopamine lesion, primary fibroblasts containing the TH transgene survived for 10 weeks, continued to express the transgene, and reduced rotational asymmetry.	Wolff et al, 1989
TH	Parkinson disease (PD)		Overexpress TH that converts tyrosine to L-DOPA to alleviate degeneration of dopaminergic nigrostriatal neurons (DNN) in PD rat models; unilateral destruction of DNN in animals with 6-hydroxydopamine and administration of apomorphine caused PD rats to turn contralaterally (7 or more rotations/min).	Implantation of transgenic immortalized fibroblasts and myoblasts intracerebrally improved rotational behavior	Fischer et al, 1991
TH	Parkinson's disease (PD)	HSV	Infection of 6-hydroxydopamine-lesioned rats, used as a model of PD, with a defective herpes simplex virus type 1 vector expressing TH	Conversion of endogenous striatal cells into L-dopa-producing cells	During et al, 1994
TH	Parkinson's disease (PD)	Lipofectin	To alleviate the symptoms of PD in TH-deficient rats (PD animal models; perform colateral rotations at 15 rounds/min upon administration of apomorphine)	Primary muscle cells were transduced with TH cDNA under control of CMV promoter; 10 million cells were injected into brains of TH-deficient rats; this resulted in 75% decrease in the number of rotations/min for more than 6 months	Jiao et al, 1993

Table 6. In vivo somatic gene transfer strategies to animal models

Gene target or delivered	Human disease	Method	Animal model, objective, and method	Results	Reference
HSV TK and ganciclovir	hepatocellular carcinoma	AAV mice	To preferentially kill hepatocellular carcinoma cells by the suicadal gene HSV TK (driven by the a-fetal protein (AFP) enhancer and albumin promoter) with ganciclovir	Selective killing of AFP-positive cells in culture; transgenic mice were established by injection of AAV ITRs, neoR, and HSV TK genes as a linear DNA fragment; HSV TK was expressed predominantly in adult liver.	Su et al, 1996
Prostaglandin G/H synthase	acute lung injury	Cat lipid rabbit	Rabbits intravenously transfected with the PGH synthase gene	Increased plasma levels of prostacyclin and PGE2; protection of lungs in rabbits against endotoxin-induced inflammation, pulmonary edema, release of thromboxane B2, and pulmonary hypertension	Conary et al 1994
Ornithine transcarbamylase (OTC)	OTC-deficiency	Adenovirus	iv injection of recombinant adenovirus to spf-ash mice (OTC-deficient)	Correction of enzyme deficiency in OTC-deficient mice for over 1 year.	Stratford-Perricaudet et al, 1990
a1-antitrypsin	a1-antitrypsin-deficiency in lung	Adenovirus	The adenovirus major late promoter was linked to a human a1-antitrypsin gene for its transfer to lung epithelia of cotton rat respiratory pathway	Both in vitro and in vivo infections have shown production and secretion of a1-antitrypsin by the lung cells for over 1 week	Rosenfeld et al, 1991
a1-antitrypsin	a1-antitrypsin-deficiency in liver	Cat lipid mice	Protect connective tissue from the lytic action of the leukocyte neutrophil elastase; plasmid was encapsulated into negatively-charged liposomes containing phoshpatidylcholine	Small liposomes were much more effective in delivering the a1-antitrypsin gene to mouse hepatocytes in vivo.	Aliño et al, 1996
a1-antitrypsin (AT, human)	acute and chronic lung diseases.	Cat lipid	Aerosol and intravenous transfection to lungs of rabbits	Human a1AT mRNA and protein were detected for at least 7 days; immunohistochemical staining showed a1AT protein in the pulmonary endothelium following intravenous administration, in alveolar epithelial cells following aerosol administration, and in the airway epithelium by either route	Canonico et al, 1994
CFTR	Cystic fibrosis (CF)	Adenovirus	To alleviate the symptoms of CF	Expression of CFTR after intratracheal instillation into lungs of cotton rats; expression between days 2-10	Rosenfeld et al, 1992
CFTR (cystic fibrosis transmembrane conductance regulator)	Cystic fibrosis (CF)	Cat lipid	To express the normal CFTR gene in lungs of Edinburgh insertional mutant mouse (cf/cf) after delivering CFTR cDNA-liposome complexes into the airways by nebulization.	Full restoration of cAMP related chloride responses in some animals; human CFTR cDNA expression in the same tissues	Alton et al, 1993
CFTR	Cystic fibrosis (CF)	Lipofectin	To express the human CFTR gene in lungs in CFTR-deficient transgenic mice by tracheal instillation of lipofectin-plasmid	Successful transfer of the CFTR gene to epithelia and to alveoli deep in the lung leading to correction of the ion conductance defects found in the trachea of transgenic mice	Hyde et al, 1993.
CFTR	CF	Lipofectin	Transduction of airway epithelial cells in normal mice by intratracheal instillation of a plasmid carrying the CFTR gene under control of the Rous sarcoma virus promoter	Airway epithelial cells were the major target and site of expression of CFTR	Yoshimura et al, 1992
CFTR cDNA (human)	Cystic fibrosis (CF)	AAV	Intratracheal instillation into neonatal New Zealand white rabbits	Epithelial expression of the human CFTR fusion protein was detected using antisera to both the human CFTR R domain and the amino-terminal epitope at up to 6 weeks after vector inoculation, a time coinciding with the completion of the alveolar phase of lagomorph lung development	Rubenstein et al, 1997
p53	lung cancer	retrovirus	Lung tumors were elicited in nu/nu mice after intratracheal inoculation with human lung cancer H226Br cells whose p53 gene has a homozygous mutation at codon 254	Intratracheal injection of a recombinant retrovirus containing the wt p53 gene was shown to inhibit the growth of the tumor	Fujiwara et al, 1994
MHC

Gene deliv. system	Advantages	Drawbacks
Murine retroviral vectors	Very safe; may achieve high efficiency of transduction; infects only dividing cells; integrates into host DNA.	Loss in expression soon after infection; low efficiency in vivo; up to 8 kb of DNA; high titers required for in vivo gene delivery; immunogenicity.
Recombinant adenoviruses	Infect nondividing cells; rarity of recombination events between adenoviral vectors and the host chromosomes; high efficiency of transduction; adenovirus vectors efficiently escape from the endosome and enter the nucleus; episomaly-replicating virus.	Induction of immune responses that eliminates therapeutic cells; may induce unwanted infections to humans; only up to 7.5 kb of exogenous DNA can be inserted; loss of adenoviral episomes in progeny cells.
AAV	Does not stimulate inflammation or immune reaction; enters nondividing cells and does not replicate; nonpathogenic virus.	Low efficiency of gene transfer; only up to 4.1 and 4.9 kb can be incorporated; wt AAV integrates on chromosome 19 but recombinant AAV integrates at different sites (e.g. chromosome 2); integration may cause inactivation of the transgene by chromatin effects.
HSV-1	Can take up to 30 kb of exogenous DNA; high titer viral stocks; wide range; can infect nonreplicating cells	Infection with HSV-1 is cytotoxic.
Baculovirus	Specificity for hepatocytes; high efficiency of infection	Not applicable in vivo at present.
EBV	Episomaly-replicating virus	wt EBV infects human cells causing mononucleosis.
HIV-1	Transduces non-dividing cells; broad range of tissues and cell types; no inflammation; sustains expression of GFP for 8-22 weeks in muscle and liver after injection to animals	Start up technology, not broadly tested.
Hybrid HSV/EBV	High efficiency of infection (95-99% after intratumoral liver injection)	Not broadly tested.
Cationic lipids	High efficiency of transfection via membrane destabilization (cell membrane and endosomal); destabilize lysosomal membranes and promote release of plasmid in the cytoplasm.	Toxic, not suited for i.v. injection; can interact with negatively charged serum proteins in vivo causing transgene inactivation; gene expression is transient; i.v. injection targets mainly the lung
Stealth liposomes	Non toxic, escape immune surveillance and concentrate into solid tumors by extravasation.	Not taken up by tumor cells but remain in the extracellular space.
Naked plasmid DNA	Suited for intramuscular injection and DNA vaccination; easy to use; no viral antigens.	Low transfection; not widely applicable method; naked plasmid is cleared from blood rapidly.
Gene gun	Easy to use (plasmid-coated gold particles are delivered to tumor cells using helium pressure); rapid, suited for gene transfer to tumor specimens from patients for immunotherapy.	Not broadly tested.

Gene delivered	Human disease	Vector	Method/Goal	Results	Reference
CAT and luciferase	none	DOTMA: DOPE	Expression of reporter genes in mouse tissues after a single intravenous injection.	CMV promoter was most effective; expression in vascular endothelia, extravascular and parenchymal cells present in lung, spleen, and heart; lower expression in all other tissues; persisted for 9 weeks.	Zhu et al, 1993
Luciferase/CMV construct and lacZ	none	Transfectam (DOGS:DOPE)	To show effectiveness in gene transfer after intracranial injection of liposome-plasmid complexes into the newborn mouse.	Transient expression of luciferase in striatal parenchymal cells; lipospermine:DNA charge ratio of 2 or smaller was most effective in vivo.	Schwartz et al, 1995
Luciferase/CMV	none	Cationic liposomes	Distribution of luciferase gene expression among tissues and persistence of expression after systemic injection.	Use of viral origins of replication and T antigen cDNA in vector sustained expression up to 3 months primarily in the lung in mice.	Thierry et al, 1995
lacZ and HSV-tk/IL-2	glioma	AAV	Direct injection of tumors induced from human glioma cells into the brains of nude mice.	30-40% of the cells along the needle track expressed b-galactosidase; administration of GCV to the HSV-tk/IL-2 treated animals for 6 days, resulted in a 35-fold reduction in the mean volume of tumors compared with controls.	Okada et al,. 1996
lacZ and human erythropoietin	none	AAV	Single intramuscular injection into adult BALB/c mice.	Protein expression was detected in myofibers for at least 32 weeks; dose-dependent secretion of erythropoietin and corresponding increases in red blood cell production in mice persisted for up to 40 weeks.	Kessler et al, 1996
lacZ	Retinal degeneration; retinitis pigmentosa	AAV	Subretinal injection of recombinant AAV particles encoding lacZ.	Successful transduction of all layers of the neuroretina as well as the retinal pigment epithelium; the efficiency of transduction of photoreceptors was significantly higher than that achieved with an equivalent adenoviral vector.	Ali et al, 1996
lacZ	Inherited hearing disorders	AAV	To assess the feasibility of introducing genetic material directly into the peripheral auditory system; infusion into the cochlea of guinea pigs.	Thin sections of cochleae showed intense immunohistochemical reactivity in the spiral limbus, spiral ligament, spiral ganglion cells and the organ of Corti but much weaker staining in the contralateral ear.	Lalwani et al, 1996
lacZ	Parkinson’s disease	HSV-1	Stereotactic injection into the midbrain of adult rats.	A 6.8-kb fragment of the rat tyrosine hydroxylase promoter supported a 7- to 20-fold increase in reporter gene expression in catecholaminergic tyrosine hydroxylase-expressing neurons in the substantia nigra.	Song et al, 1997
lacZ	Leptomeningeal cancer	AAV	To test the feasibility of AAV-mediated gene therapy.	Successful transduction of medulloblastoma (DAOY) cells in a nude rat model of leptomeningeal disease.	Rosenfeld et al, 1997
lacZ	vascular disorders	AAV	To develop gene therapies for vascular disorders by gene transfer into isolated segments of normal and balloon-injured rat carotid arteries.	Comparable gene transfer into medial and adventitial cells, with significantly higher efficiency of transduction in injured compared with normal vessels.	Rolling et al, 1997
human placental alkaline phosphatase (AP)	lung disease	AAV	To assess the ability of AAV vectors to transduce airway cells; AP gene was delivered to one lobe of the rabbit lung using a balloon catheter.	AP staining was almost exclusively in the epithelial and smooth muscle cells in the bronchus at the region of balloon placement; staining was in ciliated cells but was also in basal cells and airway smooth muscle cells.	Halbert et al, 1997

Prodrug-activating enzyme	Prodrug	Toxic substance it is converted to
Thymidine kinase from HSV	9-{[2-hydroxy-1-(hydroxymethyl)-ethoxy]methyl}guanine or ganciclovir (GCV)	GCV monophosphate
Cytosine deaminase (CD) from E. coli	5-fluorocytosine (5FC)	5-fluorouracil (5FU)
Purine nucleoside phosphorylase (PNP) from E. coli	6-methylpurine-2’-deoxyriboside (MeP-dR)	6-methylpurine (a very toxic adenine analog)
Purine nucleoside phosphorylase (PNP) from E. coli	Arabinofuranosyl-2-fluoroadenine monophosphate (F-araAMP) commercially known as fludarabine	A very toxic adenine analog
Human deoxycytidine kinase (dCK)	Cytosine arabinoside (ara-C)	A toxic drug inducing lethal strand breaks in DNA
Nitroreductase from E. coli	5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954)	A potent dysfunctional alkylating agent which crosslinks DNA

Drug resistance gene	Confers resistance to	Reference
MDR1 (multidrug resistance)	Daunomycin, doxorubicin, taxol	Galski et al, 1989; Podda et al, 1992; Sorrentino et al, 1992 (see below)
Mutant dihydrofolate reductase	Methotrexate (MTX)	Williams et al, 1987; Corey et al, 1990; Li et al, 1994; Zhao et al, 1997
Glutathione transferase	DNA alkylating agents	reviewed by Maze et al, 1997
O⁶-methyl guanine transferase	Nitrosoureas	Allay et al, 1995
Cytidine deaminase	Cytosine arabinoside (Ara-C)	Momparler et al, 1996
Aldehyde dehydrogenase	Cyclophosphamide	reviewed by Koc et al, 1996