Poly(ADP-ribosyl)ation is a posttranslational modification of proteins in eukaryotic cells performed by a family of NAD+ ADP-ribosyl transferases, the poly(ADP-ribose) polymerases (PARPs). NAD+ molecules as precursor are cleaved into nicotinamide and ADP-ribose moieties, and the latter are covalently attached to glutamic or aspartic acid residues of proteins, with PARP itself being the major acceptor. Poly(ADP-ribosyl)ation may alter the activity of acceptor protein, leading to inactivation due to the attachment of a highly complex, branched polymer carrying large numbers of negative charges (Fig. 1). The polymer is rapidly degraded by the enzyme poly(ADP-ribose) glycohydrolase (PARG), limiting the half-life of the polymer to approximately 1 min under conditions of DNA breakage. Poly(ADP-ribosyl)ation was originally discovered as an immediate response of cells to DNA strand-break-inducing agents, i.e., ionising radiation, alkylating agents and oxidants. Polymer formation in living cells after genotoxic stresses is mainly dependent on PARP-1 (encoded by the ADPRT gene), the first-discovered and best-investigated member of the PARP family. After binding of PARP-1 with its 2 zinc-fingers within the N-terminal DNA-binding domain to double-stranded DNA with interruptions in the sugar-phosphate backbone (i.e., single- and double-strand breaks), activity of the C-terminal catalytic centre is markedly stimulated, resulting in modification of target proteins (“acceptor” proteins). In the case of PARP-1, the polymer is attached to the centrally located automodification domain. Not only the covalent modification but also noncovalent yet tight binding of the polymer seems to modify the function of the involved protein.1 PARP-1 and PARP-2, i.e., the only other known family member activated by DNA strand breaks, are necessary for efficient base-excision-repair (BER) in living cells as the respective knockout mice and cells isolated thereof exhibit increased sensitivity to γ-irradiation and alkylating agents.2, 3 There is also in vitro evidence for a role of PARP-1 and PARP-2 in efficient base-excision repair.3, 4 Conversely, overactivation of PARP-1 by severe DNA damage leads to a depletion of cellular NAD+ (i.e., the substrate of PARPs) and ATP, as ATP is consumed in the process of NAD+ resynthesis.5 Such disturbance in cellular energy balance may lead to cell necrosis, thus revealing the Janus-like nature of PARP-1 (Janus, the double-faced Roman god of doors, beginning and changes): Mild genotoxic stress activates PARP-1 and stimulates DNA repair, whereas severe damage induces NAD+ depletion and necrotic cell death (Fig. 2). PARP-1 is also a substrate for caspase 3 in the early execution phase of apoptosis. Apoptotic stimuli lead to an initial burst of (ADP-ribose) polymer formation, but the cleavage in the bipartite nuclear localisation signal (NLS) of PARP-1 shuts down the enzyme activity and stops energy consumption. PARP-2 displays significant overlap regarding sequence homology and function with PARP-1.6, 7 Apart from PARP-2, there have been identified or predicted more than a dozen further proteins sharing homologies in their catalytic centre (PARP signature), but only a few of them have been characterised, mainly PARP-3, a cell cycle regulator located at centrosomes,8 PARP-4 or vault-PARP,9 which is a part of the vault particle probably involved in multidrug resistance, as well as PARP-5 and 6, belonging to the subfamily of Tankyrases,10 involved in telomere regulation and vesicle trafficking. In our present review, we focus on recent publications regarding the DNA damage-activated PARPs, i.e., PARP-1 and PARP-2. Evidence is emerging that the inhibition of PARP-1 and -2 activity can have beneficial effects for the organism under certain conditions, once again highlighting the double-faced function of PARPs. Treatment of tumour cells with a genotoxic chemotherapeutic agent and simultaneous inhibition of poly(ADP-ribosyl)ation enhances the (apoptotic) cell-killing effect of the agent by suppressing base-excision repair; on the other hand, the deleterious consequences of tissue damage by ischaemia-reperfusion and inflammation can be suppressed by inhibition of polymer formation, by blocking the necrotic pathway and/or by interfering with NF-κB signalling.
Poly(ADP-ribose) polymerases are involved in many aspects of regulation of cellular functions. Using NAD+ as a substrate, they catalyse the covalent transfer of ADP-ribose units onto several acceptor proteins to form a branched ADP-ribose polymer. The best characterised and first discovered member of this multiprotein family is PARP-1. Its catalytic activity is markedly stimulated upon binding to DNA strand interruptions, and the resulting polymer is thought to function in chromatin relaxation as well as in signalling the presence of damage to DNA repair complexes and in regulating enzyme activities. Moderate activation of PARP-1 facilitates the efficient repair of DNA damage arising from monofunctional alkylating agents, reactive oxygen species or ionising radiation, but severe genotoxic stress leads to rapid energy consumption and subsequently to necrotic cell death. The latter aspect of PARP-1 activity has been implicated in the pathogenesis of various clinical conditions such as shock, ischaemia-reperfusion and diabetes. Inhibition of ADP-ribose polymer formation has been shown to be effective, on the one hand, in the treatment of cancer in combination with alkylating agents by suppressing DNA repair and thus driving tumour cells into apoptosis, and on the other hand it appears to be a promising drug target for the treatment of pathologic conditions involving oxidative stress. In view of the existence of several members of the PARP family in mammalian cells, one has to be aware of possible side effects but also of a wide spectrum of potential clinical applications, which calls for the development of more specific inhibitors. © 2004 Wiley-Liss, Inc.
PARP INHIBITORS IN CHRONIC DISEASES
One of the major and most promising research areas of PARP inhibition is the treatment of cancers. But there is emerging evidence that inhibition of the cellular poly(ADP-ribosyl)ation system may also have positive effects in the therapy of other diseases, as was shown in a variety of cell culture systems as well as animal models. But one has to keep in mind that interference with a part of the cellular repair machinery may have deleterious side effects in the long run (see “Concluding Remarks”).
For the treatment of malignant tumours, 2 major options exist apart from surgery: irradiation and chemotherapy. As PARP-1 and -2 contribute to DNA BER and survival of cells under genotoxic stress, much effort has been invested to exploit the effects of PARP inhibition for rendering tumour cells more susceptible to cancer therapy based on DNA damaging agents, as it was first proposed in the early 1980s.11 Since then, a wide panel of PARP inhibitory compounds have been identified,12, 13 and this field is constantly growing. The main focus in the literature is on combination of PARP inhibitors with DNA damaging agents. The preferred cytotoxic compound is temozolomide (8-carbamoyl-3-methyl-imidazo[5,1-d]-1,2,3,5-tetrazin-4[3H]-one; TZM), a blood-brain barrier crossing alkylating agent first described in the mid-1980s as an antitumour agent14 and currently being evaluated in phase III clinical trials with patients affected by metastatic melanoma and high-grade recurrent gliomas. TZM induces methylation of guanine at the N7 or O6 position and adenine at the N3 position. These lesions are repaired by the cell via the O6-alkylguanine DNA alkyltransferase (OGAT), the BER pathway, or the mismatch repair system (MMR). If OGAT is not present in cells, O6-alkylation of guanine leads to inappropriate incorporation of thymine in the complementary DNA strand during S-phase and triggers the activity of MMR, which in turn excises the newly synthesised base, leading to futile circles of synthesis and excision and subsequently to apoptosis induction. Defects in MMR often lead to resistance of tumour cells because the BER takes over and excises the methylated base, thus eliminating the damage and allowing further undisturbed cell proliferation. It may be hypothesised that by decreasing the activity of the BER pathway due to inhibition of the enzymes PARP-1 and -2, sensitivity of tumour cells to alkylating agents may be restored. In this case, methylpurine repair would be impaired since BER would be stalled after the step of base excision and strand-break formation, and DNA single-strand breaks can accumulate over time. There would be cell cycle arrest (mainly at G2/M for tumour cells) and reduction of cell viability, probably by inducing the apoptotic pathway. Most PARP inhibitors developed to date have structural similarity to the substrate NAD+, thus competing for the active centre of the enzyme. The compounds are not cytotoxic on their own at concentrations necessary to achieve PARP inhibition.
Boulton et al.15 described that double treatment with TZM and different inhibitors of PARP activity (benzamide, 3-aminobenzamide, PD128763 and NU1025) led, in a concentration-dependent manner, to increased cytotoxicity and DNA strand-break levels compared to TZM alone and inhibited the drop in the cellular NAD+ level induced by TZM. The main results from this work have been reproduced and refined with several other PARP inhibitors and chemotherapeutic agents in different cell systems.16, 17, 18 Not only the cell killing activity of alkylating agents but also of topoisomerase-1 inhibitors can be potentiated by inhibition of the cellular poly(ADP-ribosyl)ation system. For example, 12 different human tumour cell lines were treated with TZM or the topoisomerase-1 inhibitor topotecan. These cells were of different tissue origin (colon, breast, ovary, lung) resembling the most common human tumours and also differed in their p53 status.19 The PARP inhibitors NU1025 and NU1085 potentiated growth inhibition and cytotoxicity independent of the p53 status of the cell line. Multiple treatment with TZM seems to be more effective than a bolus dose with the same overall amount of the agent, but in both cases additional inhibition of PARP potentiated the cytotoxic effects.20 More recent publications have addressed the effect of PARP inhibition and chemotherapeutic agents in tumour-bearing mice. Tumour cells were injected subcutaneously into nude mice, and the mice were systemically treated with TZM or camptothecin, a topoisomerase-1 inhibitor. An additional treatment with the PARP inhibitor CEP-6800 potentiated the tumour growth reduction seen with the cytotoxic agents and, depending on the cell line used, even led to total regression of the tumours.21 Tentori et al.22 injected lymphoma cells intracranially into mice and treated them systemically with TZM by single or repeated intraperitoneal administration. Double treatment with a PARP inhibitor at the same time as TZM by intracerebral injection of NU1025 significantly enhanced the survival of the tumour-bearing mice, with the fractionated TZM administration being most effective. A recent article by Tentori et al. describes the enhancement of survival of intracranial tumour-bearing mice by systemic administration of a novel blood-brain barrier crossing PARP inhibitor GPI 15427 in combination with TZM treatment.23 Other mechanisms for treatment of tumour cells are, for example, introduction of p53 by an adenovirus-mediated gene transfer, which reestablishes the sensitivity of MMR-deficient tumour cells to TZM itself and also in combination with 3-aminobenzamide.24 Inhibition of telomerase is thought to be another good candidate for tumour cell treatment, but inhibiting telomerase activity by introduction of a dominant negative version of telomerase surprisingly led to increased resistance of melanoma cells to TZM, associated with a reduced growth rate. Here, addition of 3-aminobenzamide restored the cytotoxic effects of TZM.25 In summary, the combined treatment of tumour cells in vitro and, more important, in vivo with chemotherapeutic compounds such as alkylating agents or topoisomerase-1 inhibitors and PARP inhibitors yielded beneficial effects such as increasing the number of DNA strand breaks, growth inhibition and cell killing and thus enhancement of the survival of tumour-bearing animals. But it also has to be noted that the components of any such combinatorial treatment have to be selected carefully to avoid deleterious effects.
Insulin-dependent diabetes mellitus (IDDM, juvenile diabetes or type-1 diabetes) originates from extensive β-cell destruction in the islets of Langerhans. Available genetic model systems are the diabetes-prone NOD mouse and the BB rat. In addition diabetes can be induced chemically by streptozotocin, a compound selectively killing β-cells through DNA alkylation and induction of oxidative stress by NO production. In patients, islet cell destruction is triggered by inflammatory processes, leading to production of cytokines, nitric oxide (NO) and reactive oxygen species (ROS) as well as increased protein glycation resulting from high blood glucose. In response to the stimulus, β-cell death can be apoptotic, but a recent article suggests that the major pathway in IDDM β-cell destruction is necrosis.26 NO (via peroxynitrite formation) as well as ROS induce DNA strand breaks and thus activate PARP-1 and -2, which in turn leads to a drop in energy levels. Depletion of cellular NAD+ and ATP subsequently triggers necrosis. Once diabetes is established, high blood glucose levels are accompanied by endothelial dysfunction and vascular alterations in several organs (retina, brain, kidneys, heart), which are the main determinants of diabetes-related mortality. In all 3 diabetes models mentioned above, treatment with PARP inhibitors like PJ34 reverses all effects despite persistent hyperglycaemia. PARP inhibitors are effective even when given after the onset of diabetes and only for a short period of time.27 Inhibitor treatment leads to restoration of endothelium-dependent vasodilatation and cellular energy levels and reduces cytokine formation and inflammatory response. Such effects have been shown in heart28 and kidney.29 In case of the endothelial alterations of the heart, compromised functions were improved even 3 weeks after discontinuing PARP-inhibitor treatment. One other aspect in diabetes is the possibility of regeneration of islet cells. It was demonstrated that PARP-1 is a component of the active transcription complex for the Reg gene and regulates its DNA binding by poly(ADP-ribosyl)ation.30 The Reg protein is important for the proliferation of β-cells and is induced by the concerted action of dexamethasone and IL6. Thus, inhibition of PARP activity in combination with dexamethasone and IL6 treatment leads to increased transcription of Reg and accelerated regeneration of pancreatic islet cells.
Inflammatory bowel disease
Crohn's disease is an inflammatory bowel disorder characterised by severe necrotic tissue damage. Inflammatory processes lead to increased formation of NO and ROS and are accompanied by a drop in cellular energy levels, neutrophil infiltration, increased intestinal permeability due to endothelial dysfunction, intestinal ulceration and weight loss. Available model systems are the chemically induced colitis by instillation of nitrobenzene sulfonic acid (NSA) derivates as well as the Il10 knockout mouse. Administration of PARP inhibitors before NSA treatment prevents pathologic changes at least partially,31, 32 but the most striking effects are seen in the IL10 model.33 Using the PARP inhibitor 3-AB, the authors were able to improve the status of the mice even after onset of colitis, normalizing the intestinal permeability and reducing the levels of pro-inflammatory mediators. Another approach is the administration of antisense oligonucleotides targeting PARP-2 mRNA. Such oligonucleotides proved therapeutically effective when applied to Il10 knockout mice with clinically manifest colitis.34 The same effects including reduction of iNOS activity were obtained with pharmacologic PARP inhibition, but the oligonucleotide approach is far more specific. As Adprt knockout mice are protected against ischaemia-reperfusion damage induced by occlusion of the superior mesenteric artery, which in turn leads to the release of the same mediators as in Crohn's disease,35 it is tempting to speculate about a pathophysiologic interaction between PARP-1 and PARP-2 in the context of inflammatory bowel disease.
In the system of chronic heart failure by ligation of the left anterior descending coronary artery, it has been shown that the nitrotyrosine content—a measure for the overall NO production—as well as PARP activity increased and the heart tissue exhibited an impaired vascular relaxation. Administering of the PARP inhibitor PJ34 after ligation improved the cardiac dysfunction and vascular relaxation but still yielded the same amount of nitrotyrosine.36 In heart transplantation in rats,37 PARP inhibition before excision of the donor heart and transferring it to the recipient improved the oxidative stress effects like DNA strand breaks, lipid peroxidation and drop of NAD+ and ATP levels. The size of the necrotic area was reduced, whereas the number of apoptotic cells was increased. So it remains unclear whether PARP inhibition is beneficial in heart transplantation. In ischaemia-reperfusion, the size of myocardial infarcts can be significantly reduced by abrogating PARP activity. Endothelial functions and myocardial contractility and relaxation as well as coronary blood flow were all improved.
Parkinson's disease is characterised by a loss of dopaminergic neurons of the substantia nigra. A well-established experimental model is treatment with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), leading to selective death of these dopaminergic cells and subsequently to clinical symptoms. NO and oxidative DNA modifications have been found in the substantia nigra from affected patients.38Adprt knockout mice proved to be protected against the deleterious effects of MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine).39
INVOLVEMENT OF PARP IN ACUTE DISEASES
PARP-1 overactivation can contribute to several pathophysiologic situations such as cell death and dysfunction or degeneration of tissues. The pathogenic role of PARP activity has been reported in many experiments over the last few years using animal models for a wide variety of common, debilitating or even lethal human disorders. In these experiments, the strategy was firstly to demonstrate the presence of PARP activation in the target tissue under conditions of induction of pathology and secondly to analyse the phenotype of Adprt1-knockout mice or of normal animals treated with PARP inhibitors. Protective effects of PARP inhibitors could be shown in various pathophysiologic situations in acute diseases as discussed below.
Ischaemia-reperfusion damage in brain, heart, liver, kidney, eye, bowel and skin, and allergen-induced asthma
It is well known that upon reperfusion of ischaemic tissue there is a massive release of ROS and NO in the affected area, and these compounds in turn lead to the generation of peroxynitrite.40 Peroxynitrite cytotoxicity occurs via multiple pathways involving the initiation of lipid peroxidation, the oxidative modification and inactivation of proteins and the extensive generation of DNA strand breaks, which lead to an overactivation of PARP.41 This overactivation of PARP appears to be crucial for the ensuing acute cell death, which at the macroscopic level is typically manifested as an infarct. An important role of PARP in ischaemia-reperfusion was demonstrated in a variety of organs such as brain, heart, liver, kidney, eye, bowel, skin and lung.
A reduction in the volume of brain infarct as well as an attenuation of the associated neurologic dysfunction has been observed in Adprt1-knockout mice42 or in wild-type animals when treated with PARP inhibitors.43 The size of myocardial infarcts could be significantly reduced by abrogating PARP activity using the inhibitor PJ34 before the onset of reperfusion after a 1 hour ischaemia.44 When administering the PARP inhibitor 5-aminoisoquinolinone (5-AIQ) to rats 5 min prior to the onset of liver ischaemia, a reduction in liver injury was observed as judged by reduced serum levels of transaminases, lactate dehydrogenase, gamma-glutamyl transferase and less staining for intercellular adhesion molecule-1 (ICAM-1) in liver sections.45 Intraperitoneal administration of PARP inhibitors after ischaemia-reperfusion injury of the kidney accelerated the recovery of normal renal function, as assessed by monitoring the levels of plasma creatinine and blood urea nitrogen during 6 days after the ischaemic insult.46 The application of 3-aminobenzamide (3-AB) as late as 12 to 18 hr after the onset of reperfusion significantly attenuated retinal ischaemia-reperfusion injury when morphology and morphometry of the inner retina of rats was analysed 7 days after reperfusion.47 Intestinal PARP overactivation upon haemorrhagic shock and resuscitation resulted in gut hyperpermeability, which developed in Adprt(+/+) but not in Adprt(−/−) mice. Adprt(−/−) mice were also protected from the rapid decrease in blood pressure after resuscitation and showed an increased survival time as well as reduced lung neutrophil sequestration.35 UV-induced acute photodamage in mouse skin could be reduced by topical application of a novel PARP inhibitor (BGP-15M) that was shown to decrease the number of single-strand DNA breaks and to downregulate IL-10 and TNFα-levels in epidermal cells.48 Finally, in a murine model of allergen-induced asthma, 3-AB prevented airway inflammation elicited by ovalbumin. Likewise, Adprt knockout mice were resistant to ovalbumin-induced inflammation.49
Septic or haemorrhagic shock
Adprt-knockout mice were reported to be resistant to septic shock induced by injection of lipopolysaccharide (LPS).50, 51 A severe deficiency in NF-κB transactivation seems to be responsible for this phenotype, as a reduction in typical LPS-induced increases in serum levels of TNFα, IL-6 and NO could be observed in PARP-1-deficient animals.52 It was shown that PARP-1 is required for specific NF-κB-dependent gene activation and can act as a co-activator for NF-κB in response to pro-inflammatory stimuli and genotoxic stress.53
In wild-type mice, LPS is captured by LPS-binding protein and then transferred to extracellular receptor complexes localised on effector cells such as monocytes and macrophages. Binding of LPS to these receptor complexes activates different signalling cascades including NF-κB (Fig. 3). NF-κB together with its co-activator PARP-1 might then upregulate the expression of pro-inflammatory mediators like cytokines (TNFα, IFN-γ) and interleukins (IL-6) in macrophages. The massive release of cytokines from macrophages in turn can activate the NF-κB/PARP-1 complex together with other transcription factors in target cells (monocytes, macrophages, epithelial and endothelial cells), leading to the expression of pro-inflammatory mediators both in effector and target cells. Thus, a positive feedback loop evolves, resulting in the persistent activation of the NF-κB/PARP-1 complex.54 As cytokines (e.g., IFN-γ) and interleukins (e.g., IL-1β) have been shown to induce iNOS-expression,55 their overexpression leads to massive production of NO in effector and target cells. NO in turn can react with superoxide released as a by-product of the mitochondrial respiratory chain to form peroxynitrite (NO·+ O → ONOO−) that gives rise to the formation of the highly reactive hydroxyl radical (· OH).56 Due to extensive DNA damage caused by the above-mentioned ROS, PARP-1 overactivation in target cells leads to NAD+/ATP depletion, which in turn causes necrotic cell death. This is the main mechanism underlying the so-called “PARP-1 suicide model.”57 PARP inhibition either by administration of various pharmacologic PARP inhibitors or by genetic knockout can reduce the effects of NF-κB-transactivation and subsequent ROS formation by preventing NAD/ATP depletion and necrotic cell death and thus confer resistance to septic shock.
To circumvent the problem that endotoxin-induced shock in mice may not be a perfect model for human sepsis, PARP inhibitors were applied in a porcine model of severe hypodynamic sepsis induced by E. coli clot implantation.59 Administration of the PARP inhibitor PJ34 immediately before intraperitoneal implantation of E. coli-laden fibrin clots to produce peritonitis and bacteraemia led to a marked increase in survival rates (83% survival after 24 hr compared to only 12% within the control group) accompanied by a significant reduction in serum levels of TNFα and prevention of poly(ADP-ribose) formation in septic animals compared to the control group.
As pointed out above, both sides of DNA damage-dependent PARP activation have recently been in the focus of therapeutic approaches. On the one hand, the activity of PARP in ischaemia-reperfusion damage leading to ROS and NO release needs to be blocked in order to prevent cellular necrosis and keep the infarct volume small. On the other hand, in cancer therapy, inhibition of PARP activity is an attractive option for suppressing repair and achieving more extensive cell killing, but this time via the apoptotic pathway. Therefore, the outcome of inhibition of PARP activity depends on the tissue background and additional pharmacologic treatments. Research is moving on very fast in this newly emerged field of therapy. But there are two caveats: First, long-lasting inhibition may have deleterious effects due to persistent suppression of BER. Therefore, interference with this part of the repair machinery may give rise to mutations in cells outside the target of chemotherapy and in a worst-case scenario may even propagate tumour formation. Second, as there are several PARP family members with unknown functions, unspecific inhibition of poly(ADP-ribosyl)ation in general may have unexpected and severe side effects. PARP-2, PARP-3 and the tankyrases are vital mediators of genomic stability by their involvement in repair, mitosis and telomere function, respectively. The development of more specific inhibitors should therefore increase the safety of such kind of therapy and open the future for a broad range of possible applications to patients in a variety of pathophysiologic conditions.
Another possible option of interfering with PARP activity would be an indirect one, i.e., by inhibition of PARG, the polymer-degrading enzyme. It is conceivable that PARG inhibitors might be able to shut down PARP activity via overaccumulation of polymer and the ensuing feedback regulation. For PARP-1, a plausible mechanism would be the loss of affinity to DNA as a result of hyperautomodification, thus abrogating its crucial activation signal (i.e., contact with DNA strand breaks). This area of research, however, is very much under debate at present and publications are scarce. A much broader and more consolidated basis of experimental data would be necessary for any reasonable estimation of the potential usefulness of PARG inhibitors as future therapeutic agents.