Safety of new therapeutic procedures must be weighed carefully against conventional therapies and against the likely benefits to be gained by the patient. This is particularly true of gene transfer. As a potentially curative procedure for an immunodeficiency such as CGD, gene transfer and re-engrafrment of autologous stem cells can be measured against bone marrow transplantation, which from a fully matched sibling donor is probably associated with a 20-30% average chance of mortality directly related to the degree o f immunosuppression required for engraftment (Morgan, 1992). The necessity for immuno-suppressive conditioning prior to engraftment o f genetically modified cells is partially dependent on the efficiency of gene transfer, but would significantly add to the risks associated with the procedure. Several animal studies have now shown that long-term chimaerism can be established without bone marrow ablation, and would imply that efficiency of gene transfer is the limiting factor (Stewart et al. 1993; Wu and Keating, 1993; Bienzle et al. 1994).
8.3.1 Safety of gene delivery systems
The safety o f gene transfer relates to the effects o f expression o f the inserted gene on the host cell, and to unrelated effects arising from expression of vector proteins, or from disruption o f normal cellular function. The first o f these is gene-specific, and may require regulated differentiation and tissue-specific expression. This is less of a problem
for haematopoietic cells which can be isolated and manipulated ex vivo, and to a certain extent obviates the need for specific cell-targeting strategies. Expression of vector protein in target cells has become a particular problem for adenovirus-based delivery systems, and results in toxicity at high copy number, and induction of powerful immune responses. This is not a problem for gene transfer by retroviral or AAV-based vectors which do not encode any viral proteins.
Production o f recombinant virus free from contamination by replication-competent wild-type or helper virus is of paramount importance for clinical application. Three monkeys infected with a recombinant retroviral supernatant heavily contaminated with helper virus developed high grade T cell lymphomas from which helper virus sequences were recovered (Donahue et al. 1992). The frequency of helper contamination can be minimised by construction o f more sophisticated cell lines in which homology between vector and helper sequences are minimised, and in which multiple recombination events are necessary for production of replication competent virus. In the absence of helper contamination, the risks associated with these vectors are associated with disruption of normal cellular function secondary to integration into the host cell genome. Site- preferential integration mediated by AAV Rep proteins may be advantageous, but is not currently applicable, and may not support high level expression of heterologous genetic material.
For essentially random integration events, insertional mutagenesis occurs when a vector integration event leads to disruption or abnormal regulation of a gene. The chance of insertion and inactivation of both copies of a gene involved in regulation of cell-cycle, for example a tumour suppressor gene has been estimated to be 1 in 10^^. Tumour formation by non-acute transforming viruses, including the murine leukaemia viruses is thought to arise by recombination with endogenous viruses, and activation of cellular genes secondary to proviral integration events. Transformation probably involves activation o f multiple proto-oncogenes by promoter insertion (Cloyd et al. 1980; Cuypers et al. 1984). Mitogenic stimuli transduced by binding o f envelope protein to cell-surface receptors may also be important (Li and Baltimore, 1991). While transformation secondary to integration of the recombinant provirus presents a finite
risk in primate cells, little homology is shared between human endogenous retroviral sequences (HERV) and MLV-based vectors, so that activation of a replication competent virus by this mechanism is unlikely.
Transfer o f defective endogenous murine retroviral sequences originating from the packaging cell line, the most abundant of which are the virus-like 30S RNAs (VL-30s), and intracistemal A-type particles (lAP), has been reported to occur at low levels, but has not been associated with pathogenicity (Hatzoglou et al. 1990). VL30 sequences can be packaged into type C virions, and horizontally transmitted to other cells. However, protein products o f these RNAs have not been reported. Proviruses o f murine lAPs are particularly abundant, and are expressed in mature B cells, B cell-related tumours, and embryos (Kuff and Lueders, 1988). They most probably represent retrotransposons recently descended from retroviruses, and are not known to encode infectious virus. For AAV-based vectors, risks of promoter insertion are again finite, and partially determined by the nature o f the promoter used to direct expression of the transferred gene. Similarly, replacement of retroviral LTR regulatory sequences in the U3 region by natural sequences may restrict the risk o f a detrimental integration event to cells in which the promoter is active.
8.3.2 Non-viral methods for gene transfer
Non-viral methods for gene transfer were first developed in the early 1970’s, and have distinct safety advantage over virus-based systems because no viral genetic material is co-transferred. They include chemical techniques, mechanical techniques, membrane fusion and receptor mediated endocytosis. The most commonly used laboratory methods for transfer o f DNA to recipient cells are based on co-precipitation and endocytosis of DNA with calcium phosphate, and in resistant cases, electroporation. However, the efficiency o f stable gene transfer is generally very low, and these techniques are not useful in a clinical setting.
Two techniques that may have clinical application are based on liposome-mediated membrane fusion (Alton et al. 1993), and on receptor-mediated endocytosis of conjugated DNA (Michael and Curiel, 1994). The main disadvantages of liposome-
mediated gene transfer are low efficiency, and low frequency o f stable long term expression. To overcome some of these problems, molecular conjugate vectors have been developed in which DNA is delivered via a receptor-mediated endocytosis pathway which normally internalises specific macromolecules. These vectors posess two distinct functional domains, a DNA-binding polycation such as polylysine, which is also necessary to condense the DNA, and a ligand domain for a specific cell surface receptor. This system has several potential advantages, including specific targeting determined by the ligand domain, and lack o f constraint on DNA size. Unfortunately, efficiency o f gene transfer in vitro has in general been low and appears to reflect lysosomal degradation of DNA following internalisation.
Adenoviral entry to cells is similar in many ways, except that acidification of the endosome results in disruption, allowing virions to proceed to the nucleus to complete their life cycle. Endosomal disruption is independent of viral gene expression, and is a direct function o f viral capsid protein. Conjugation of adenovirus capsid via hexon protein to the DNA-polylysine complex utilises the endosomal disruption properties of the capsid, and enhances gene transfer by the molecular conjugate (Gotten et al. 1992). Ternary complexes in which selective ablation of adenovirus-mediated receptor binding to the cell is associated with retention of endosomal disruption properties o f the capsid, and a distinct alternative ligand domain, may permit efficient gene transfer by molecular conjugates to specifically targeted cells.
Problems associated with immunogenicity and toxicity of the adenoviral capsid currently limit the application of this technology to cell culture. However, the principle of ligand mediated DNA delivery is attractive, and may in future become the gene transfer method o f choice.