Mechanism of action of purine analogues in chronic lymphocytic leukaemia
2003; Wiley; Volume: 121; Issue: 5 Linguagem: Inglês
10.1046/j.1365-2141.2003.04336.x
ISSN1365-2141
Autores Tópico(s)Immunodeficiency and Autoimmune Disorders
ResumoFludarabine and chlorodeoxyadenosine are deamination-resistant analogues of deoxyadenosine (dAdo) that induce apoptosis in normal and malignant lymphocytes and have important therapeutic activity in chronic lymphocytic leukaemia (CLL). Both drugs are converted intracellularly to cytotoxic 5′-phosphorylated derivatives through the action of deoxycytidine kinase. Phosphorylated purine analogues are substrates for several enzymes involved in DNA synthesis and repair. In proliferating cells, cytotoxicity seems to require incorporation of purine analogue into DNA. In resting lymphocytes, killing has been attributed to inhibition of DNA repair, accumulation of unrepaired DNA breaks and NAD+/ATP depletion due to sustained activation of poly(ADP-ribose) polymerase. Although this mechanism is demonstrable in occasional patients with CLL, p53-mediated apoptosis and direct induction of mitochondrial depolarization are probably more important. Caspase activation is not required for the induction of cell killing but determines whether death occurs by apoptosis or necrosis. Purine analogue resistance in CLL is associated with P53 mutation/deletion, relative over-expression of antiapoptotic Bcl-2 family proteins and a low capacity for nucleoside phosphorylation. To overcome drug resistance, strategies are required that target these points of blockade. Chronic lymphocytic leukaemia (CLL) accounts for 25% of all leukaemias and is the most common form of lymphoid malignancy in Western countries. The disease is characterized by the accumulation of clonal mature B lymphocytes – probably akin to memory cells (Klein et al, 2001; Rosenwald et al, 2001) – in the blood, bone marrow and secondary lymphoid tissues. Among B-lymphoproliferative disorders, CLL is highly distinctive owing to its peculiar surface immunophenotype (CD5+, CD23+, weak surface immunoglobulin) and propensity for immune dysregulation. However, the disease also serves as a paradigm for other lymphoid neoplasms in which clonal expansion results from impaired apoptosis rather than increased proliferation (Caligaris-Cappio, 2000; Kipps, 2000). CLL runs a very variable clinical course. Thus, although some patients survive for decades without ever requiring any treatment, others die of drug-resistant disease within a year or two of presentation. The biological basis for this heterogeneity is currently the subject of much interest and controversy. Suffice to say that considerable attention has been focused on the adverse prognostic significance of certain chromosomal aberrations (del 11q23 and del 17p13), CD38 expression and relative lack of mutation of the IgVH gene (Damle et al, 1999; Hamblin et al, 1999; Dohner et al, 2000). CLL is currently considered to be an incurable disease, the aim of treatment being to alleviate symptoms and prevent life-threatening complications. The alkylating agent chlorambucil has, for many years, been the cornerstone of treatment. However, since the late 1980s when the clinical efficacy of purine analogues in CLL first became apparent (Keating et al, 1989), there has been a progressive shift towards the use of these newer agents as first-line therapy. This trend was undoubtedly accelerated by the recent publication of a large American study showing superiority of fludarabine over chlorambucil in terms of remission frequency (63%vs 37%), quality (complete response in 20%vs 4%) and duration (median 25 vs 14 months), albeit at the expense of higher toxicity (Rai et al, 2000). The trial has been criticized, owing to the relatively low dose of chlorambucil used (Summerfield et al, 2002), and further randomized studies are in progress involving somewhat higher doses of the drug. Despite this, purine analogues seem already to have overtaken alkylating agents – at least in concept – as the central component of modern therapy. Despite the justified optimism that combination therapy with purine analogues and novel agents such as monoclonal antibodies might herald a new era of curative treatment for CLL, drug resistance remains a major clinical problem that threatens the realisation of this goal. Clearly, overcoming such resistance is of paramount importance, but to do so requires an understanding of drug action at the cellular level. There is a large body of literature on the subject, much of it from Dr Plunkett and colleagues at the MD Anderson Cancer Center in Houston, TX, USA. It is beyond the scope of the present review to cover all of this work or, indeed, to give a detailed account of nucleoside biochemistry and pharmacology; for these subjects, the reader is referred to review articles by Plunkett and Gandhi (1997), Galmarini et al (2001) and Gandhi and Plunkett (2002). Rather than duplicate these reviews, the present article will instead focus on those aspects of purine analogue action that are of particular relevance to CLL, or which have been experimentally addressed by the author. To understand the cytotoxic action of purine analogues, it is first necessary to consider certain aspects of nucleoside metabolism (Fig 1). For more comprehensive reviews, see Carson et al (1988) and Carrera et al (1994). To summarize, all cells maintain a pool of deoxynucleotides (dNTPs) for DNA synthesis. dNTPs exist in a state of dynamic equilibrium with the corresponding non-phosphorylated deoxynucleosides. For example, dAdo is continually being phosphorylated to dATP via deoxyadenosine-5′-monophosphate (dAMP) and dADP (deoxyadenosine-5′-diphosphate). Conversely, dATP is continually being dephosphorylated to dAdo via the same intermediates. The rate-limiting step in the conversion of dAdo to dATP is dAdo→dAMP; this reaction is catalysed primarily by deoxycytidine kinase (dCK). The rate-limiting step in the conversion of dATP to dAdo is dAMP→dAdo; this reaction is catalysed mainly by 5′-nucleotidase (5′NT). In lymphoid cells, the relative activities of dCK and 5′NT strongly favour dAdo phosphorylation to dATP. However, dAdo may also be converted into deoxyinosine (dIno) by adenosine deaminase (ADA). dIno is then converted sequentially to hypoxanthine, xanthine and finally uric acid. ADA plays a critical role in limiting the intracellular accumulation of dATP, which is cytotoxic at high concentrations. The importance of this enzyme to lymphocyte survival is illustrated by the fact that individuals with a congenital deficiency of ADA have a marked reduction in lymphocyte numbers and develop the clinical syndrome of severe combined immunodeficiency (SCID). Aspects of deoxyadenosine (dAdo) metabolism that are relevant to purine analogue cytotoxicity. There are three analogues of dAdo in routine clinical use for the treatment of lymphoproliferative disorders: 2′-deoxycoformycin (dCF), 2-chloro-2′-deoxyadenosine (CdA) and 9-β-d-arabinosyl-2-fluoroadenine (fludarabine). dCF is a fermentation product of Streptomyces antibioticus and was developed as a result of being a potent and specific inhibitor of ADA (reviewed by Cheson & Grever, 1997). Treatment of patients with this compound results in increased plasma dAdo levels and intracellular dATP accumulation. In vitro, dCF-induced dATP accumulation and cytotoxicity require the addition of exogenous dAdo. In contrast to dCF, fludarabine (reviewed by Hanauske & Von Hoff, 1997) and CdA (reviewed by Beutler et al, 1997) are synthetic halogen-substituted dAdo analogues that are resistant to ADA-mediated deamination. Treatment of patients or isolated cells with either of these agents results in the intracellular accumulation of a 5′-triphosphate derivative analogous to dATP (Plunkett & Gandhi, 1997). Phosphorylated purine analogues are potent inhibitors of DNA synthesis owing to their direct inhibitory effect on ribonucleotide reductase. This enzyme maintains dNTP pools by converting ribonucleoside diphosphates into deoxyribonucleoside diphosphates, which are subsequently phosphorylated. Analogues of dATP also inhibit DNA synthesis by a second mechanism involving incorporation into DNA and stalling of DNA polymerization (Plunkett & Gandhi, 1997). Although inhibition of DNA synthesis may account for the action of purine analogues against dividing cells, it does not adequately explain the cytotoxic effect of these drugs in quiescent cells. However, in 1985, Dr Carson's group at the Scripps clinic in La Jolla, CA, USA, published a seminal paper that appeared to resolve this problem. They demonstrated the following sequence of events in normal resting lymphocytes incubated with either CdA or dCF plus dAdo: intracellular accumulation of dATP (or triphosphate derivative of CdA), accumulation of DNA breaks and simultaneous reduction in RNA synthesis, depletion of cellular NAD+, depletion of ATP, and finally cell death detected as loss of vital dye exclusion (Seto et al, 1985). This pathway is illustrated in Fig 2. A model of the purine analogue cytotoxicity in resting lymphocytes proposed by Seto et al (1985). A strong correlation had previously been noted between purine analogue cytotoxicity and intracellular accumulation of dATP (or purine analogue triphosphate) both within and between different cell types (reviewed in Carson et al, 1988). Furthermore, deoxycytidine (dCyt), which competes with dAdo and its analogues for phosphorylation by dCK, was already known to prevent cytotoxicity induced by CdA or dCF plus dAdo (Carson et al, 1977; Kefford & Fox, 1982). As expected, therefore, all of the biochemical events described by Seto et al (1985) were inhibited by dCyt. Subsequent work has shown that the therapeutic response to purine analogues among patients with chronic B-cell leukaemias correlates with the dCK:5′NT ratio (Kawasaki et al, 1993), and it is now widely accepted that phosphorylation of dAdo (or its analogues) to dATP (or its analogues) is an absolute requirement for purine analogue cytotoxicity. The induction of DNA breaks by purine analogues has been observed by other groups in both normal (Brox et al, 1984) and malignant (Begleiter et al, 1987; Ganeshaguru et al, 1987; Ho et al, 1988) resting lymphocytes. Although controversial (Jostes et al, 1989; Boerrigter, 1991), it has been suggested that quiescent lymphocytes are continually breaking and rejoining their DNA (Greer & Kaplan, 1984; Johnston, 1984). Dr Carson's group postulated that dATP and its analogues inhibit the repair of such spontaneously occurring DNA breaks, thereby resulting in their accumulation (Seto et al, 1985). In keeping with this idea, the repair of DNA breaks induced by ionizing radiation is inhibited by CdA (Kuwabara et al, 1991) and dAdo plus dCF (Begleiter et al, 1988). Furthermore, fludarabine inhibits the completion of nucleotide excision repair induced by ultraviolet radiation (Sandoval et al, 1996) or DNA alkylation (Yamauchi et al, 2001). By analogy with previous work involving agents that damage DNA directly (Skidmore et al, 1979), Dr Carson's group speculated that the NAD+ depletion that was induced by purine analogues resulted from activation of poly(ADP-ribose) polymerase (PARP). This nuclear enzyme, which is activated by DNA breaks, covalently links ADP-ribose residues to a range of proteins, including itself, and histones using NAD+ as a substrate. Although its precise role is unclear, it is thought to contribute to the recognition and signalling of single-strand DNA breaks (DeMurcia & Menissier-deMurcia, 1994; d'Amours et al, 1999). To address the possible role of the enzyme in mediating the cytotoxic action of purine analogues, Dr Carson's group conducted experiments with the PARP inhibitors nicotinamide and 3-aminobenzamide (3AB). Both compounds prevented purine-analogue-induced depletion of NAD+ and ATP, and loss of vital dye exclusion. In contrast, there was no inhibition of dATP accumulation or DNA breaks (Seto et al, 1985). Based on these observations, it was concluded that PARP activation was indeed the link between purine-analogue-induced DNA breaks and subsequent depletion of NAD+. The ATP depletion that followed was entirely consistent with the pivotal role of NAD+ in ATP generation, and it was not difficult to understand why cells that were depleted of NAD+ and ATP should cease to be viable. The riddle of how purine analogues killed resting lymphocytes had been solved. Or so it seemed. In 1993, Dr Plunkett's group formally demonstrated that fludarabine and CdA kill CLL cells by inducing apoptosis (Robertson et al, 1993). It should be noted that the classic morphological appearances of apoptosis had been observed 8 years earlier in hairy-cell leukaemia cells treated with dCF plus dAdo (Matsumoto et al, 1985). The significance of this observation was not appreciated at the time, probably because apoptosis research was very much in its infancy then. However, for the last 10 years or so, considerable effort has been put into relating purine analogue cytotoxicity to the key regulators of apoptosis: p53, caspases and the Bcl-2 family proteins. Intriguingly, in parallel with the surge of interest in apoptosis as the mode of purine-analogue-induced killing, the previously accepted mechanism of cytotoxicity (PARP activation and metabolic depletion) seems to have been forgotten about, even by the group that first described it (Genini et al, 2000a)! More about that later, but first a brief review of how the p53 pathway contributes to purine analogue cytotoxicity. p53 is a tumour suppressor protein that protects the genome from mutation by either eliminating damaged cells through the induction of apoptosis or by facilitating DNA repair through the initiation of cell cycle arrest (Lane, 1992; Levine, 1997). The TP53 gene is mutated and/or deleted in around 10–15% of patients with CLL, the usual defect being mutation of one allele and deletion of the other. We have recently shown that p53 dysfunction occurs in a further 15–20% of patients as a result of mutation in the gene encoding ATM (ataxia telangiectasia mutated): a kinase that regulates p53. Such ATM-mutant patients display impaired accumulation of p53 and its transcriptional target p21 in response to ionizing radiation, and are partially resistant to radiation-induced killing (Pettitt et al, 2001; Lin et al, 2002). In the early to mid 1990s, several independent groups showed that CLL patients with TP53 mutation/deletion not only had a short survival but also responded extremely poorly to therapy with purine analogues (El Rouby et al, 1993; Wattel et al, 1994; Dohner et al, 1995). A small number of in-vitro studies were subsequently performed examining the role of p53 in purine analogue cytotoxicity. The results were less compelling and, indeed, conflicting (Thomas et al, 1996; Johnston et al, 1997). We have partly explained this discrepancy by showing that p53-dysfunctional CLL cells (Pettitt et al, 1999a) and unstimulated splenocytes from p53-knockout mice (Pettitt et al, 1999b) display significant but only partial resistance to purine analogues in vitro. These findings suggest that purine analogues can kill resting lymphocytes by both p53-dependent and -independent mechanisms. However, further explanation is still required for the discord between the modest in-vitro resistance and the profound in-vivo resistance to purine analogues observed in CLL patients with p53 defects. It is possible that purine analogues may kill CLL cells in vivo predominantly through indirect mechanisms that are heavily dependent on p53. It is equally possible that the killing of self-renewing CLL cells in 'proliferation centres' might depend more heavily on p53 than that of quiescent 'end-stage' cells in the blood. This would explain why some patients initially respond but then progress once treatment is discontinued. Alternatively, therapeutic resistance to purine analogues may develop as a secondary phenomenon, owing to the genomic instability and propensity for clonal evolution that are a recognized feature of tumours with p53 dysfunction. Thus, treatment may select out tumour-cell subclones that have acquired resistance to purine analogues for reasons that are not directly related to p53. Irrespective of the extent to which purine analogues depend on p53 for their cytotoxicity, the drugs are potent activators of p53 in both thymocytes (Szondy, 1995) and CLL cells (Gartenhaus et al, 1996). In proliferating cells, the purine analogue gemcitibine appears to activate p53 following its incorporation into DNA by interacting with a p53/DNA-PK complex. The activated complex then interferes with further DNA synthesis by stalling DNA polymerases α and ε at the site of drug incorporation (Achanta et al, 2001). It is less clear how purine analogues activate p53 in resting lymphocytes. Although drug incorporation into DNA may be involved, the accumulation of DNA breaks may be more important. Having established that purine analogues can kill resting lymphocytes by p53-dependent and -independent mechanisms, we next set out to examine p53-independent mechanisms of cell killing. In particular, we turned our attention to the pathway proposed by Dr Carson's group in 1985 and subsequently forgotten, namely PARP-mediated metabolic depletion as a response to the accumulation of unrepaired DNA breaks. Careful consideration of this model of purine analogue cytotoxicity raised a number of concerns. First, it is inconsistent with the observation that purine analogues kill resting lymphocytes by inducing apoptosis. Thus, apoptosis is an energy-dependent process, whereas lowering of cellular ATP is associated with failure of plasma membrane function and cellular necrosis (Eguchi et al, 1997; Leist & Nicotera, 1997). Second, DNA breaks identical to those induced by purine analogues can be detected in thymocytes undergoing apoptosis in response to dexamethasone, which does not directly cause DNA damage (Hoshino et al, 1993). This raised the possibility that the DNA breaks that are detected in purine-analogue-treated lymphocytes do not initiate apoptosis but are instead a non-specific secondary event (Eastman & Barry, 1992). Third, PARP can be activated as a consequence of apoptotic DNA fragmentation (Kaufmann et al, 1993). This raises the possibility that the PARP-dependent metabolic sequelae of purine analogue exposure might also be a secondary, rather than death-inducing, event. Fourth, loss of vital dye exclusion is a late event in apoptotic cells that reflects the cell membrane disruption of secondary necrosis (Zamzami et al, 1995). It, therefore, follows that compounds that prevent loss of dye exclusion (e.g. PARP inhibitors) may simply be delaying the onset of secondary necrosis in cells that are already apoptotic. Finally, in the experiments reported by Seto et al (1985), the PARP inhibitors were used at concentrations that are now known to inhibit other enzymes that use NAD+ as a substrate (Rankin et al, 1989; Banasik et al, 1992). In the light of these various considerations, we conducted our own study of CLL cells to examine the role of PARP in purine analogue cytotoxicity (Pettitt et al, 2000). To do this, we used 3AB at a concentration known to produce selective inhibition of PARP in intact cells. We also measured cell death by a number of different methods. In keeping with its previously reported effects on normal lymphocytes, 3AB delayed purine-analogue-induced loss of propidium iodide exclusion in the great majority of cases of CLL studied, indicating that PARP activity was indeed involved in such killing. However, when cell death was examined in more detail, it became clear that, in most cases, 3AB was preserving the membrane integrity of cells that were dead by other criteria (Table I). Similar results were obtained using two other highly specific PARP inhibitors. This indicates that PARP activity was contributing to the secondary necrosis of cells that had already undergone purine-analogue-induced apoptosis, but not to the initiation of cell killing. This observation illustrates the importance of using the appropriate analytical techniques in experiments involving inhibition of cell killing, and calls into question the results of previous studies involving PARP inhibition in which cytotoxicity was detected as loss of vital dye exclusion. Importantly, we identified one case of CLL among the 30 studied in which cell death occurred by a mixture of apoptosis and necrosis, and in which all manifestations of nucleoside-induced killing were dramatically inhibited by 3AB and the other PARP inhibitors. Interestingly, this patient had a bi-allelic p53 gene defect. This indicates that PARP activation can occasionally be central to nucleoside-induced killing, and that such killing is p53 independent. The fact that this type of killing was largely necrotic was entirely consistent with the notion that it was mediated by ATP depletion. Our findings, therefore, support the idea that PARP activation can, at least under some circumstances, constitute a bona fide cytotoxic DNA damage response pathway. In 1997, it was reported that caspases were activated during the apoptotic killing of CLL cells by purine analogues (Bellosillo et al, 1997) and that purine analogue-induced DNA fragmentation could be prevented by caspase inhibition (Chandra et al, 1997). This was not, perhaps, unexpected given that these enzymes are activated during the induction of apoptosis in response to most stimuli (Thornberry & Lazebnik, 1998) and are responsible for activating endonucleases (Wyllie, 1998). The key question, so it seemed, was whether caspases contributed to the induction of apoptosis (as is the case for cell killing induced by ligation of 'death receptors' such as Fas) or whether the enzymes were merely activated as a secondary event in cells that had already undergone mitochondrial permeability transition. It should be noted that the latter is a pivotal step in the induction of apoptosis that has fatal consequences for the cell irrespective of downstream events. Thus, blocking the caspase activity of cells prior to the direct induction of mitochondrial permeability transition converts the mode of cell death from apoptosis to necrosis, but does not prevent cell death from occurring (Hirsch et al, 1997). We, therefore, conducted our own study in which CLL cells were treated with purine analogues in the presence or absence of the broad-spectrum caspase inhibitor Z-VAD.fmk (Pettitt & Cawley, 2000). In keeping with previous reports, Z-VAD.fmk produced marked inhibition of purine-analogue-induced PARP cleavage and DNA fragmentation. In contrast, there was no inhibition of cell death as detected by other criteria (Table I). This indicates that, although caspases are required for cell death to occur by apoptosis, these enzymes do not contribute to the induction of killing. Although mitochondrial disruption is a convergence point for many of the biochemical pathways that regulate apoptosis (e.g. p53 activation), in a recent study, Dr Carson's group showed that the purine analogues CdA and 2-chloro-2′-ara-fluoroadenine (clofaribine), but not fludarabine, can induce the depolarization of isolated mitochondria directly, resulting in the release of cytochrome c and apoptosis inducing factor (AIF) (Genini et al, 2000a). Although these experiments were conducted using non-phosphorylated nucleosides, it was assumed that the mitochondria possessed the enzymes required for conversion of purine analogues to their phosphorylated metabolites, and that the latter were responsible for the observed effects. This study, therefore, provides evidence for a novel mechanism through which certain phosphorylated purine analogues can induce apoptosis independently of DNA damage. The same group also showed that the 5′-triphosphate derivative of CdA can replace dATP in the cytochrome c/Apaf-1 apoptosome complex (Genini et al, 2000b). In this way, purine analogues might, therefore, facilitate the activation of downstream 'effector' caspases. However, such a mechanism is unlikely to contribute to the initiation of cell killing, as components of the apoptosome, being located in the mitochondrial intermembrane space, only become accessible to purine analogues following mitochondrial permeability transition. Furthermore, as already discussed, we have shown that blocking caspase activity does not inhibit purine-analogue-induced killing. Purine analogues have, for some time, been known to inhibit RNA synthesis (Seto et al, 1985). Furthermore, a correlation was recently demonstrated in CLL cells between fludarabine-induced inhibition of RNA synthesis and subsequent cell killing. This raises the very plausible possibility that global repression of gene transcription may contribute to the cytotoxic action of purine analogues by reducing the expression of proteins that are important for cell survival (Huang et al, 2000). In fludarabine-treated cell lines, inhibition of RNA synthesis appears to result primarily from incorporation into RNA and termination of transcription (Gandhi & Plunkett, 2002). In resting lymphocytes, however, it is unclear whether purine analogue incorporation into RNA or DNA break accumulation is more important. The latter mechanism seems highly plausible and would explain why the phenomenon is seen with dCF plus dAdo, where the toxic moiety is dATP itself (Seto et al, 1985). Figure 3 speculates on how the various cytotoxic pathways discussed above are likely to relate to one another. dATP and its analogues produce cytotoxicity through two main mechanisms: direct induction of mitochondrial permeability transition and accumulation of unrepaired DNA breaks. These DNA breaks induce mitochondrial permeability transition via p53 activation and suppression of RNA synthesis. Mitochondrial permeability transition results in cell death irrespective of downstream events. In the presence of functioning caspases, death occurs by apoptosis. Apoptotic cells are rapidly phagocytosed in vivo but, in vitro, secondary necrosis supervenes. This process is characterized by loss of cell membrane integrity and is largely due to PARP-mediated NAD+/ATP depletion. Under certain circumstances (i.e. when other killing pathways are blocked), PARP-mediated necrosis may be the primary mechanism of purine-analogue-induced killing. Current model of purine analogue cytotoxicity in chronic lymphocytic leukaemia. Established mechanisms of therapeutic resistance are shown on the right. Although CLL is predominantly an accumulation of non-cycling lymphocytes, clonal expansion could not occur without proliferation. This consideration also applies to disease relapse following remission induction. It is, therefore, relevant to consider the action of purine analogues in dividing, as well as resting, cells. Intriguingly, there are important differences. In dividing cells, incorporation of purine analogues into DNA, resulting in impaired replicative DNA synthesis and induction of apoptosis, appears to be a fundamental determinant of cytotoxicity (Huang & Plunkett, 1995). In contrast, this does not appear to be the case in resting lymphocytes, even though purine analogues are incorporated into DNA during repair synthesis (Huang et al, 2000). Exactly how apoptosis is triggered in dividing cells remains unclear. However, the S-phase-specific killing of ML-1 cells by fludarabine incorporation into DNA involves activation of the proapoptotic kinase JNK-1 (Sampath & Plunkett, 2000). As already mentioned, p53 may also contribute to cytotoxicity in dividing cells by binding to incorporated purine analogue as a p53/DNA-PK complex, by stalling DNA polymerization and by signalling apoptosis (Achanta et al, 2001). As p53 has 3′-5′ exonuclease activity in vitro, this could theoretically constitute a mechanism of purine analogue resistance in dividing cells. However, excision of incorporated purine analogues from DNA does not seem to occur to any great extent in intact cells, irrespective of their p53 status (Feng et al, 2000). Furthermore, cells with wild-type p53 are generally more sensitive, rather than resistant, to purine analogues as compared with p53-mutant cells. Purine analogues have been used in combination with other DNA damaging agents to good effect in the treatment of chronic lymphoproliferative disorders, including CLL. Fludarabine plus cyclophosphamide is a typical example of such a combination, and there is evidence of synergy between these two drugs in vitro (Bellosillo et al, 1999; Yamauchi et al, 2001). However, it is unclear whether DNA alkylation induced by cyclophosphamide enhances the action of fludarabine by initiating nucleotide excision repair, thereby promoting incorporation of the purine analogue into DNA, or whether fludarabine enhances the action of cyclophosphamide by preventing the completion of nucleotide excision repair. The latter mechanism has been demonstrated in quiescent CLL lymphocytes (Yamauchi et al, 2001), but the former may be more relevant to dividing cells in which purine analogue incorporation into DNA is required for cytoxicity. At the clinical level, retrospective data from the MD Anderson Cancer Center suggest that fludarabine plus cyclophosphamide is superior to fludarabine alone (O'Brien et al, 2001). However, a large randomized study has shown comparable response rates and greater toxicity in CLL patients treated with fludarabine plus chlorambucil as compared with fludarabine alone (Rai et al, 2000). Other randomized studies are in progress, but to date fludarabine is usually given as a single agent, at least initially. It is now recognized that cognate interactions between CLL cells and T cells play an important role in the pathogenesis of the disease. Furthermore, recent studies have shown that the killing of CLL cells by purine analogues is inhibited by ligation of CD40, an important molecule on B cells that engages with CD40L on T cells. This protective effect appears to be mediated by nuclear factor (NF)-κB, as it was blocked by a phosphorothioate κB decoy oligodeoxynucleotide (Romano et al, 1998, 2000). Another study has shown that inhibition of purine-analogue-induced apoptosis by CD40 ligation is associated with upregulation of Mcl-1 and Bcl-XL (Kitada et al, 1999). Interaction of CLL cells with the extracellular matrix protein fibronectin also confers resistance to fludarabine-induced killing via engagement of α4β1 integrin and increased levels of Bcl-XL (de la Fuente et al, 2002). In CLL, there is evidence to support several different mechanisms of resistance to purine analogues. These are shown in Fig 3. As already mentioned, therapeutic response to purine analogues correlates with the intracellular dCK:5′NT ratio (Kawasaki et al, 1993). This suggests that drug resistance may, at least in some patients, result from a low capacity for purine analogue phosphorylation. The p53 pathway is also clearly important in determining the therapeutic efficacy of purine analogues (and alkylators) in CLL. Indeed, the response rate for patients with p53 mutation/deletion is so poor (El Rouby et al, 1993; Wattel et al, 1994; Dohner et al, 1995) that it could be argued that purine analogues (and alkylators) should not be given to such patients, at least not as single-agent therapy. Given the role of ATM as an important regulator of p53, it would seem likely that CLL patients with ATM mutations might also respond poorly to therapy with purine analogues (and alkylators). However, there are to date no data to support or refute this hypothesis. The therapeutic response to purine analogues (and alkylating agents) also appears to correlate with the expression of certain Bcl-2 family proteins. This is not unexpected given the role of these proteins in regulating the opening of the mitochondrial transition pore: a key event in the induction of apoptosis by many different stimuli (Zamzami & Kroemer, 2001). In one study, failure of CLL patients to achieve complete remission was strongly associated with high levels of Mcl-1 (Kitada et al, 1998). Other work has shown a correlation between resistance to purine-analogue-induced cell killing in vitro and a low Bax:Bcl-2 ratio (Pepper et al, 1999a). It has been reported that Bcl-2 can specifically inhibit the function of p53 by relieving p53-mediated transcriptional repression (Shen & Shenk, 1994; Sabbatini et al, 1995) and/or preventing p53 nuclear import (Beham et al, 1997). However, it is becoming increasingly clear that the regulation and function of p53 depend very much on the cellular context, and that data generated in one cell type are not necessarily universally applicable. Although Bcl-2 over-expression is a consistent feature of CLL lymphocytes (Hanada et al, 1993), accumulation of p53 and its transcriptional target p21 in response to ionizing radiation is impaired only in the 25% or so of patients who have a TP53 or ATM mutation (Pettitt et al, 2001). This observation is clearly inconsistent with the notion that over-expressed Bcl-2 inhibits p53 activation or function in these cells. There are a number of potential strategies for overcoming resistance to purine analogues in CLL. Combining the drugs with other genotoxic agents has already been discussed above. Another approach is to combine fludarabine-based chemotherapy with monoclonal antibodies. As single agents, Campath-1H (humanized rat anti-CD52) and, to a lesser extent, rituximab (human–mouse chimaeric anti-CD20) have useful activity in CLL (Osterborg et al, 1997; Keating et al, 2002). Although binding of these antibodies to their corresponding antigen on the surface of the malignant cell may be directly cytotoxic, there is emerging evidence that immunotherapy may also sensitize tumour cells to the induction of apoptosis by chemotherapy. For example, rituximab has been shown to downregulate Bcl-2 and restore chemosensitivity to a cisplatin-resistant B-cell lymphoma cell line (Alas et al, 2002). At the clinical level, combinations of monoclonal antibodies and fludarabine-based chemotherapy have produced some very impressive results. Thus, fludarabine combined with cyclophosphamide and rituximab has produced response rates of over 90% (Keating et al, 2002), while fludarabine plus Campath has been shown to induce remissions in heavily pretreated patients who are resistant to each agent alone (Kennedy et al, 2002). Given the apparent importance of Bcl-2 family proteins as determinants of purine analogue sensitivity in CLL, another approach to overcoming drug resistance is to combine the agents with Bcl-2 antisense oligonucleotide. This rationale is both theoretically attractive and feasible, given that CLL cells express unusually high levels of Bcl-2 protein as a result of hypomethylation of the BCL-2 gene (Hanada et al, 1993). Indeed, Bcl-2 antisense has established in-vitro activity against CLL cells (Pepper et al, 1999b) and enhances the cytoxicity of chlorambucil (Pepper et al, 2001). A clinical trial of fludarabine in combination with Bcl-2 antisense is currently in progress. Although p53 activation is only one of several cytotoxic pathways that are employed by purine analogues, the profound resistance to therapy observed in CLL patients with TP53 mutation/deletion suggests that purine analogue resistance might be overcome in such patients if function could be restored to the mutant p53 protein. Indeed, peptides corresponding to the C-terminus of wild-type p53 have been shown to perform this function in vitro (Hupp et al, 2000), and much effort is currently underway to produce a clinically useful compound that will withstand degradation and penetrate into the nucleus. If such a peptide were to become available, combination therapy with purine-analogue-based chemotherapy might prove effective in this very difficult group of patients. CLL is still considered to be incurable. But history clearly demonstrates that incurability is not an intrinsic feature of any disease. History also demonstrates that the first step towards curative treatment in a malignancy is to identify therapeutic agents that are capable of inducing complete remissions in a substantial proportion of patients. Purine analogues have been heralded as the first class of drugs capable of doing this in CLL, although formal proof that they are more effective than alkylating agents has yet to be obtained. However, before we can consider CLL to be a 'curable' disease, the quality of remissions currently achievable with purine analogues will have to be improved upon until we have therapies that can eliminate every last self-renewing tumour cell without producing unacceptable toxicity. This is a tall order, given our still sketchy understanding about the biology of CLL. However, with each new development – the most recent being combinations of purine-analogue-based chemotherapy and monoclonal antibodies – more and more patients are apparently achieving better and better remissions, and the idea that CLL may one day be amenable to cure becomes ever more realistic. Despite this justified optimism, it is important to remember that CLL is a heterogeneous disease, and that there will always be some patients who do badly, even with state-of-the-art therapy. Improving the outlook for these difficult cases will depend on understanding the biological reasons for their non-response. Our role as clinical investigators is to achieve this understanding, apply new developments from the pharmaceutical industry in a rational way, and generate new ideas about how specific mechanisms of drug resistance may be overcome.
Referência(s)