Binding sites for NF-κB have been identified in the promoter regions of the genes for E-selectin, VCAM-1 and ICAM-1, while a binding site for AP-1 has been localized on the promoter region of the ICAM-1 gene. Point mutations which decrease NF-κB binding to κB elements result in diminished cytokine-induced E-selectin expression on cultured endothelial cells, suggesting that NF-κB plays an important role in cytokine induction of the E-selectin gene (Essani et al., 1996). Two closely spaced functional κB elements have also been identified in the MAdCAM-1 promoter (Takeuchi & Baichwal, 1995). The increased ICAM-1 expression on endothelial cells exposed to hydrogen peroxide (Lo et al., 1993) appears to be mediated through AP-1, and independent of κB elements (Roebuck et al., 1995). Therefore, hydrogen peroxide and cytokines appear to activate ICAM-1 gene transcription in endothelial cells through distinct cis-regulatory elements within the ICAM-1 promoter.
It was recently demonstrated that inhibitors of the proteasomal degradation pathway for IκB lead to decreased nuclear accumulation of NF-κB and the subsequent abrogation of TNF-α induced cell-surface expression of E-selectin, VCAM-1, and ICAM-1 in endothelial cells (Read et al., 1995). This response has important functional consequences because proteasome inhibitors also block both the adherence and emigration of leukocytes in human endothelial cell monolayers.
Recent evidence indicates that the transcription factors AP-1 and NF-κB are major targets for some of the commonly used anti-inflammatory drugs including glucocorticoids, aspirin, salicylates, gold salts and D-penicillamine. Glucocorticoids activate the glucocorticoid receptor (GR) in the cytosol, which then form GR-GR homodimers that bind a specific DNA sequence termed the glucocorticoid response element (GRE). Increased transcription of genes bearing a GRE in their promoter region follows. In addition to positive regulation of gene expression, glucocorticoids inhibit the expression of a wide variety of genes involved in the inflammatory process.
An improved understanding of the anti-inflammatory mechanism of glucocorticoids came from the observation that ligand activated GR inhibits AP-1 and NF-κB mediated transcription (reviewed in Cato & Wade, 1996). For example, glucocorticoids inhibit the NF-κB mediated expression of adhesion molecules ICAM-1 (Tailor et al., 1997), VCAM-1 (Tessier et al., 1996), and E-selectin (Brostjan et al., 1997). Glucocorticoids also inhibit the AP-1 expression of collagenases I and IV (Handel, 1997). Several mechanisms for the mutual antagonism between GR and NF-κB, and between GR and AP-1, have been proposed. For example, a direct protein-protein interaction between the GR and NF-κB has been proposed that will prevent the binding of GR and NF-κB to their respective DNA response elements (Ray & Prefontaine, 1994) (Figure 3). Similarly, binding of GR to Jun and Fos has been thought to cause mutual inhibition of GR and AP-1 DNA binding (Sakurai et al., 1997). Glucocorticoids also increase transcription of the gene for IκB, thereby increasing the formation of this protein which binds to activated NF-κB in the nucleus. The IκB protein probably induces the dissociation of NF-κB from κB sites on target genes and causes NF-κB to move to the cytoplasm (Scheinman et al., 1995).
Figure 3. Mechanisms used by the glucocorticoid receptor to inhibit transactivation by transcription factors. (1) After the glucocorticoid (GC) activates its receptor (GR), a direct protein-protein interaction between GR and transcription factors (TF) will prevent the binding of TF to the respective DNA response elements. (2) Binding of GR to the promoter region of the IκB gene results in transactivation; IκB binding NF-κB will prevent binding or displace NF-κB from κB sites. (3) Binding of GR to glucocorticoid responsive elements modulates TF-induced transactivation.
Download figure to PowerPoint
Recent evidence shows that in human inflammatory bowel disease (Ardite et al., 1998) and in asthma (Adcock, 1996), cessation of the inflammatory activity in response to steroid treatment is associated with disappearance of NF-κB from nuclear extracts of intestinal or bronchial mucosa, and that failure to abrogate NF-κB activation results in persistence of the inflammatory process. In patients with steroid-resistant asthma, there appears to be an exaggerated activation of AP-1 that binds to and therefore sequesters activated GR inside the nucleus; this would reduce the availability of GR to inhibit NF-κB, which is normally activated in such patients (Adcock et al., 1995).
Gold salts significantly inhibit AP-1 DNA binding in nuclear extracts at concentrations of 5 mM, which is within the range achieved in the serum of rheumatoid arthritis patients under this treatment. Gold salts have also been shown to inhibit IL-1 induced expression of NF-κB and AP-1 dependent transfected reporter genes (Williams et al., 1992) and to inhibit DNA binding activity of NF-κB in vitro (Yang et al., 1995). Consistent with these effects on pro-inflammatory transcription factors is the observation that gold salts inhibit expression of ICAM-1 and VCAM-1 in endothelial cells (Koike et al., 1994). D-penicillamine inhibits AP-1 DNA binding in nuclear extracts in the presence of free radicals, presumably by forming disulphide bonds with the cysteine residues in the DNA binding domains of Jun and Fos (Handel et al., 1996). Aspirin and sodium salicylate also inhibit activation of NF-κB (Kopp & Ghosh, 1994). It has been shown that salicylate inhibits activation of NF-κB by preventing phosphorylation and subsequent degradation of IκB, and this results in blockade of the TNF-induced increase in mRNA levels of ICAM-1, VCAM-1 and E-selectin, and a dose-dependent inhibition of TNF-induced surface expression of these adhesion molecules (Pierce et al., 1996). Indomethacin, a nonsalicylate cyclo-oxygenase inhibitor, has no effect on surface expression of adhesion molecules, suggesting that the effects of salicylate are not due to inhibition of cyclo-oxygenase (Pierce et al., 1996).
Proteasome inhibitors have been tested in experimental models of inflammation with promising results. In a rat model of experimental colitis induced by peptidoglycan/polysaccharide, proteasome inhibition using MG-341 significantly suppressed the upregulation of VCAM-1 and iNOS in the colon and this was associated with a reduction of colonic inflammation (Conner et al., 1997).
Systemic inhibition of NF-κB activation in humans for prolonged periods carries some risk since there is the potential for severe immunosuppression and enhanced cytokine-induced cytotoxicity (Beg & Baltimore, 1996; Wang et al., 1996). Removal of the p65 gene is lethal to the mouse embryo (Beg et al., 1995) and p50 knockout mice breed normally but have increased susceptibility to infections (Sha et al., 1995). However, because both NF-κB and AP-1 are inducible transcription factors that act in response to environmental stimuli, it may be possible to titrate the dose of an NF-κB directed inhibitor within a dynamic range to achieve a therapeutic, and subtoxic, response.
The potential side effects and nonspecific actions of the currently used transcription factor inhibitors have led to a search for drugs that rely on an alternative strategy for regulation of protein synthetic pathways that are specifically related to the inflammatory process. One such approach that has already yielded significant results is the use of antisense oligodeoxynucleotides (ODNs) (Agrawal, 1996).
Antisense ODNs are single stranded DNA sequences complimentary to a specific messenger RNA (mRNA). In theory, antisense ODNs, through base pairing of complementary bases, specifically bind to a mRNA thereby blocking the expression of the gene product (Sharma & Narayanan, 1995) (Figure 4). The mechanism of antisense ODN inhibition may involve multiple modalities, including the induction of RNAase H activity through the formation of an DNA:RNA hybrid, steric hindrance that interferes with translation, and inhibition of mRNA processing and transport from the nucleus (Sharma & Narayanan, 1995). Although ODNs can undeniably hit their intended targets, when an antisense molecule elicits a biological response, it can be difficult to determine whether the response was elicited because the reagent interacted specifically with its target RNA, or because some non-antisense reaction – involving other nucleic acids or proteins – took place. In order to distinguish between antisense and non-antisense effects, ODNs with an altered (scrambled) nucleotide sequence of the antisense are employed. The ability of ODNs to act as sequence-specific inhibitors of gene expression depends on their entry into the cytoplasm and/or nucleus. Since ODNs are negatively charged, they cannot passively diffuse through cellular membranes. Instead, they appear to enter cells via the processes of adsorptive and fluid-phase endocytosis (Yakubov et al., 1989). The identity of cell membrane proteins that bind ODNs is unclear, however a number of heparin binding proteins such as fibroblast growth factors and vascular endothelial growth factors have been implicated in this process (Guvakova et al., 1995). Of particular interest is the observation that ODNs may enter cells via an interaction with the heparin binding protein CD11b/CD18 (Benimetsaya et al., 1997). This suggests that ODNs may be preferentially taken up by activated leukocytes and thereby exert a more profound influence on leukocyte function. Upon entering the cell, in order to be effective, antisense ODNs must be resistant to nucleases, and have sequence-specific effects (Agrawal, 1996). Although questions of ODN specificity remain, there is growing evidence that antisense molecules can be useful pharmacological tools when applied carefully.
Figure 4. Target sites for decoy and antisense ODNs. Antisense ODNs are single stranded DNA sequences that specifically bind to mRNA, thereby blocking expression of the gene product. Transcription factor (TF) decoys are double stranded ODNs that compete for protein binding with the authentic binding elements, thereby interfering with gene regulation.
Download figure to PowerPoint
Antisense molecules have been produced to block the production of specific subunits of NF-κB. In murine embryonic stem cells, antisense ODN raised against p65 elicit a significant reduction in p65 mRNA, and have profound effects on cell adhesion properties (Sokoloski et al., 1993). Treatment of neutrophils with antisense phosphorothioate ODN to the p65 subunit of NF-κB results in a reduction in the expression of p65 and effectively abolishes the upregulation of CD11b normally elicited by either formyl-met-leu-phe and tissue plasminogen activator, indicating that antisense oligomers to p65 can interfere with neutrophil adhesion molecules (Narayanan et al., 1993). Antisense oligonucleotides to p65/p50 also reduce the expression of CD11b/CD18 on stimulated monocytic HL60 cells (Sokoloski et al., 1993), and the expression of E-selectin, ICAM-1, and VCAM-1 on stimulated human umbilical vein endothelial cells (Lee et al., 1995). A likely consequence of these actions of p65 ODNs is a profound diminution of the inflammatory response in mice with TNBS-induced colitis, as well as in IL-10 deficient mice with colitis. In both models of colitis, administration of p65 antisense was more effective in treating the inflammatory disease than glucocorticoids (Neurath et al., 1996).
ISIS 2302, an antisense phosphorothioate ODN to human ICAM-1, appears to selectively inhibit cytokine-induced ICAM-1 expression in a variety of human cells in vitro and in vivo (Bennett et al., 1994; Yacyshyn et al., 1998). Furthermore, a recent pilot study in patients with Crohn's disease showed that administration of this drug reduced ICAM-1 expression in intestinal mucosa and resulted in a significant decrease in corticosteroid usage relative to placebo treated patients (Yacyshyn et al., 1998). A murine analogue, ISIS 3082, has been shown to be active in multiple models of inflammation, including dextran-sulphate induced colitis (Bennett et al., 1997), allograft rejection (Stepkowski et al., 1994), and endotoxin-induced neutrophil recruitment in the lung (Kumasaka et al., 1996). Two studies have shown that treatment of stimulated human umbilical vein endothelial cells with antisense ODNs directed against ICAM-1, E-selectin, or VCAM-1 results in selective inhibition of protein expression and a corresponding reduction of monocyte (or HL-60 cell) adhesion to the cultured endothelial cells (Bennett et al., 1994; Lee et al., 1995).
Another strategy used to inhibit components of the inflammatory response is decoy ODNs that specifically interfere with regulatory proteins (Morishita et al., 1998). Inhibition of sequence-specific DNA-binding proteins can be achieved with double-stranded ODNs containing the specific binding elements. The transcription factor decoy (TFD) competes for protein binding with the authentic binding elements and consequently interfere with gene regulation (Figure 4). The mechanism of TFD action is distinct from the antisense approach in that the production of the target protein is unaffected and only its capacity to bind to the regulatory DNA element is affected. The TFD strategy is particularly attractive for several reasons: (1) the potential drug targets (transcription factors) are plentiful and readily identifiable, (2) synthesis of the sequence-specific decoy is relatively simple and can be targeted to specific tissues, (3) knowledge of the exact molecular structure of the target transcription factor is unnecessary, and (4) decoy ODNs may be more effective than antisense ODNs in suppressing an inflammatory reaction by virtue of their capacity to inhibit transcription of the multiple genes activated by a given transcription factor. Like antisense ODNs, the same critical parameters for optimal function exist, i.e., TFDs must be nuclease resistant, be taken up by cells, and have sequence-specific effects. A major concern regarding the use of TFDs is nonspecific effects, particularly those of phosphorothioate-substituted ODNs. Non-sequence-specific inhibition may occur from blockade of cell surface receptor activity or interference with other proteins (Gibson, 1996). Furthermore, TFDs containing guanine cytosine dinucleotides may result in immune activation (Khaled et al., 1996). To address these concerns, careful controlled experiments must be performed to eliminate the potential nonspecific effects of TFDs-mediated therapy. Furthermore, scrambled ODNs with several mutations in the consensus sequence should be used as controls.
Based on evidence showing NF-κB activation and increased adhesion molecule expression in tissues exposed to ischaemia-reperfusion, the value of treatment with decoy ODN against NF-κB has been evaluated in this model of acute inflammation (Morishita et al., 1997). In rats, intracoronary administration of NF-κB decoy ODNs before or after coronary artery occlusion markedly reduces the size of the myocardial infarct measured at 24 h after reperfusion. The selectivity of the NF-κB decoy ODN effect was confirmed by the finding that the reduction in infarct size was not observed in rats treated with antisense ODN directed against the iNOS gene. The specificity of the NF-κB decoy in inhibiting cytokine and adhesion molecule expression was also confirmed by in vitro experiments using human and rat coronary artery endothelial cells that were transfected with the NF-κB decoy; the decoy ODN inhibited the expression of ICAM-1, VCAM-1 and E-selectin (Morishita et al., 1997). In another study of rats subjected to myocardial ischaemia-reperfusion, the efficacy of a decoy ODN against NF-κB was confirmed by an enhanced recovery of left ventricular function and coronary flow in rats treated with the NF-κB decoy ODN, compared to groups of rats receiving a scrambled decoy or placebo. The protection against ischaemic damage afforded by the NF-κB decoy ODN was associated with less neutrophil adherence to endothelial cells and a lower tissue level of IL-8 (Sawa et al., 1997). The findings of these two studies suggest that in vivo administration of decoy ODNs against NF-κB may be an effective therapeutic strategy for treatment of myocardial ischaemia.
In a recent study, the two approaches to modulate gene expression were compared, i.e., the ability of an antisense that binds to the mRNA for the ReIA subunit of NF-κB to inhibit cytokine production by TNF-stimulated splenocytes was compared to the responses observed in splenocytes receiving a decoy with double-stranded ODNs that bind the NF-κB protein. TNF-α expression was reduced by both treatments, as were the levels of IL-2. However, the antisense effects did not last beyond 24 h, whereas the decoy ODN was shown to inhibit cytokine production even at 72 h after the initial TNF-stimulation (Khaled et al., 1998).
Limitations to the application of decoy or antisense ODN strategies to modulate transcription factor activity are similar to those discussed above in relation to drugs that block NF-κB activation, with potential inhibition of normal physiological responses as the major concern. Therefore, the application of decoy ODN strategies as gene therapy may be limited to the treatment of those acute inflammatory conditions (e.g., ischaemia-reperfusion) in which activation of transcription factors plays a pivotal role.