For gene transfer to be therapeutically successful, expression must be sufficient to sustain the desired clinical effect. This may be achievable using heterologous promoter systems, but fully regulated physiological gene expression is preferable, and particularly important for genes which encode active participants in control o f cell cycle or cell differentiation. The following preliminary studies were designed to investigate some of the mechanisms that direct both myeloid and differentiation-specific expression of the
3.1.2 Defining functional transcription units
Basal promoter elements for most protein-coding genes in mammalian cells lie within the first 200bp upstream of the transcription start site. Although necessary for transcription, and often able to direct some tissue-specific or developmental specificity, these regions in isolation are rarely sufficient to support fully regulated high level expression. For this to occur, the basal promoter interacts with distally occurring elements. Enhancer elements are characterised by the ability to potentiate transcription independently o f orientation, and distance from initiation site. However, isolated enhancer elements invariably show diminished activity when introduced at ectopic integration sites in comparison to the natural site, and many integration events will fail to support desired levels of gene expression. This may in part reflect the need for multiple enhancer elements to function in a co-operative rather than hierarchical manner, but extensive evidence now indicates that different regions o f chromatin have different properties, and that these properties as well as individual regulatory elements are critical determinants of local gene expression.
Functional expression domains in chromatin have been defined by characteristic properties that distinguish them from bulk chromatin, including sensitivity to DNAasel
(Weisbrod, 1982; Reeves, 1984). In contrast to bulk chromatin which is relatively resistant to DNAasel, genes become preferentially susceptible in tissues in which expression occurs. This does not reflect transcription itself, but the potential for active transcription, and always extends over the whole transcribed region and often some distance beyond. The basis for nuclease sensitivity is not clear, but is associated with alterations of nucleosomal structure including depletion of histone H I, and histone acétylation (Lewin, 1994). Superimposed on regional sensitivity, are hypersensitive sites which are likewise are usually related to potential for gene expression in specific cells or tissues, rather than active transcription (Gross and Garrard, 1988). In some circumstances, however, events critical for induction of transcription result in the appearance of new hypersensitive sites (Beato, 1989). They may lie long distances upstream of the start of transcription, but may also be found within intronic or 3’ untranslated regions of the gene, and represent segments o f so-called ‘open chromatin’ within which the structure and/or the dynamics of nucleosome assembly are altered.
The state of gene méthylation also correlates with potential for gene transcription, rather than the active process. Most of the methyl groups added to eukaryotic DNA occur at the 5 position of cytosine residues in the dinucleotide CG (5-meC), the majority of which are methylated (Razin and Riggs, 1980; Bestor and Coxon, 1993). Genomic under-representation of CG (frequency 0.008) compared to GC (frequency 0.04) and méthylation are probably linked. In contrast to oxidative deamination at the N4 position of 5-meC which forms a T residue, deamination of un-methylated C forms uracil which is readily recognised, and excised by DNA repair mechanisms. Retention of methylated CG may therefore be a reflection of associated gene function. Most CG sites are always either methylated or un-methylated, but a few exhibit tissue-specific variability in which undermethylation is a feature of domains which contain transcribed genes, and which are significantly DNAasel sensitive (Cedar, 1988).
In some regions of DNA known as CpG-rich islands, the density of CG dinucleotides approaches the predicted value (Bird, 1986). These have an average G-C base pair content of 60% compared with the 40% average of bulk DNA, and are stably unmethylated. Although many of the genes associated with CpG islands are
constitutively expressed ‘housekeeping’ genes, some tissue-specific genes are linked to islands which are unmethylated irrespective of the state of expression of the gene. The functional significance of methylated DNA is controversial, particularly as yeasts and invertebrates such as Drosophila and the nematode C.elegans lack meC (Uriel-Shoval et al. 1982), but it has been implicated in the control of a number of cellular processes in eukaryotes, including transcription (Busslinger et al. 1983; Boyes and Bird, 1991), genomic imprinting (Swain et al. 1987), developmental regulation (Antequera et al. 1989), mutagenesis (Cooper and Youssoufian, 1988), DNA repair (Hare and Taylor, 1985), X-inactivation (Pfeifer et al. 1990), and chromatin organisation (Lewis and Bird, 1991). Abnormal méthylation of an expanded triplet repeat through which FMR-1 gene transcription is silenced in Fragile X syndrome (Sutcliffe et al. 1992), and aberrant promoter méthylation of tumour suppressor genes associated with tumorigenesis (Sakai et al. 1991) are both significant indicators of a prominent role in gene activation and mutation. Furthermore, the DNA methyl transferase gene has been shown to be essential for the normal embryonic development o f mice (Li et al. 1992), and to be associated with DNA replication foci (Leonhardt et al. 1992).
Locus control regions (LCR) are either single or clusters of sites which are necessary for proper expression o f a linked gene or genes. They are defined by gene expression in a transgenic system that is insensitive to position o f integration, and dependent on the number of copies o f the transgene. The human p-globin LCR is the paradigm, and is localised to 4 DNAasel hypersensitive sites, corresponding to transcription factor binding-sites, within a region of about 15kb (Grosveld et al. 1987; Dillon and Grosveld, 1993). Full LCR activity can be achieved by joining shorter fragments, but full level and regulated expression is dependent on linkage to the globin promoter. LCR function is tissue specific, and in non-erythroid tissues is susceptible to position effect. Domains may also be defined both physically and functionally by boundary elements, including matrix attachment sites (MARs), at which points DNA is attached to the nuclear matrix, and insulator sequences which prevent gene regulatory influences from crossing domain boundaries (Kellum and Schedl, 1991; Chung et al. 1993; Lewin, 1994). The boundary regions o f the chicken lysozyme gene, which is located within a 21kb chromatin domain o f elevated DNase 1 sensitivity flanked by two MARs, have been functionally
characterised (Bonifer et al. 1994). The entire locus is expressed at high level and independent of chromosomal position in macrophages o f transgenic mice. Position independent expression is, however, lost if one of the tissue-specific enhancer regions is deleted, although tissue-specificity is largely retained. Deletion o f boundary regions has no effect on copy number-dependency, but increases the incidence o f ectopic expression, and are therefore necessary to suppress transgene expression in inappropriate tissues. Whether LCR regions and boundary elements are necessary for the regulation of all genes is unknown.
3.1.3 Regulation of NADPH-oxidase gene expression
Gene regulation for all components o f the NADPH-oxidase is poorly understood. The most extensively investigated component is gp91^^'’^. The nucleotide sequence of the proximal promoter region, and partial characterisation of cw-elements and trans-dicimg
factors involved in regulating tissue-specific expression have been described (Skalnik et al. 1991a). One feature of particular interest is a duplicated CCAAT box between - 106bp and -124bp, of which the distal motif is recognised by the classical, and ubiquitous CCAAT-binding factor (CPI). A second protein, CCAAT displacement protein (CDP), was also found to bind to the region surrounding this motif, but in contrast was expressed primarily in cells in which the gp91^^^^ gene is transcriptionally inactive. CDP-binding activity is therefore associated with repression o f gp91^^^^ transcription, and must be disrupted for induction of myelomonocytic expression of the gp9 l/’^ox Functionally, a 450bp fragment of the proximal promoter was sufficient to target reporter transgene expression to murine monocytes and macrophages, but not to other phagocytic cells that express endogenous gp91^^°^, implying that other cis-
acting elements are necessary for fully regulated expression (Skalnik et al. 1991b). Other regions o f interest within 1.5kb o f the initiation site include an Alu repeat between -930 and -650, and several regions of simple sequence. No significant homology was identified with 5’ flanking regions o f other characterised myeloid specific genes, including cathepsin 0 , myeloperoxidase, and neutrophil elastase.
Regulation o f gene expression by cytokines has only been studied in detail for IFN-y. Specifically, IFN-y has been shown to induce levels of gp91^^^^ mRNA in vitro in
normal neutrophils, monocyte-derived macrophages and the promonocytic cell-line THP-1 (Newberger et al. 1988; Cassatella et al. 1989,1990). In the latter two cell types, increased abundance of gp91^^^^ mRNA occurred largely as a result o f an increase in the rate of transcription, although levels of immunoreactive protein were little changed. Sig nal transduction pathways that mediate the action o f IFN-y on gp91^^"^ gene expression are not clear, but seem to be specific, and independent of those that mediate changes in expression of other oxidase components (Amezega et al. 1992). Regulation of gene expression for does not seem to be influenced by IFN-y.