Nucleoside analogues (NA) are essential components of AML induction therapy (cytosine arabinoside), effective treatments of lymphoproliferative disorders (fludarabine, cladribine) and are also used in the treatment of some solid tumors (gemcitabine). These important compounds share some general common characteristics, namely in terms of requiring transport by specific membrane transporters, metabolism and interaction with intracellular targets. However, these compounds differ in regard to the types of transporters that most efficiently transport a given compound, and their preferential interaction with certain targets which may explain why some compounds are more effective against rapidly proliferating tumors and others on neoplasia with a more protracted evolution. In this review, we analyze the available data concerning mechanisms of action of and resistance to NA, with particular emphasis on recent advances in the characterization of nucleoside transporters and on the potential role of activating or inactivating enzymes in the induction of clinical resistance to these compounds. We performed an extensive search of published in vitro and clinical data in which the levels of expression of nucleoside-activating or inactivating enzymes have been correlated with tumor response or patient outcome. Strategies aiming to increase the intracellular concentrations of active compounds are presented.
Nucleoside analogues (NA) constitute an important class of antimetabolites used in the treatment of hematological malignancies and, more recently, in solid tumors.1 These therapeutic compounds mimic physiological nucleosides in terms of uptake and metabolism and are incorporated into newly synthesized DNA resulting in synthesis inhibition and chain termination. Some of these drugs also inhibit key enzymes involved in the generation of the purine and pyrimidine nucleotides and RNA synthesis, and directly activate the caspase cascade. All of these effects may lead to cell death.
The NA family includes various pyrimidine and purine analogues. Among the pyrimidine analogues cytosine arabinoside (ara-C, cytarabine) is extensively used in the treatment of acute leukemia, while gemcitabine has more recently demonstrated activity in pancreatic, breast and lung cancer.2,3,4 The two first purine analogues to have been described were the thiopurines 6 mercaptopurine (6-MP) and 6-thioguanine (6-TG). These compounds could more appropriately be designated as ‘nucleobases’. Other new purine analogues are 2-chlorodeoxyadenosine (2-CdA) and fludarabine.5,6,7 These latter drugs have mostly been used in the treatment of low-grade hematological malignancies.8,9
The importance of NA chemotherapeutic agents has recently increased as a result of the introduction of new compounds into clinical use and the expansion of indications into the field of solid tumors. During the same period, there has been rapid progress in the understanding of mechanisms of drug resistance to NA. In this review, we highlight the current knowledge concerning the cellular mechanisms of resistance to NA and possible strategies that may be used to overcome such resistance.
Normal nucleoside transport and phosphorylation
Because physiologic nucleosides are generally hydrophilic and do not readily permeate the plasma membrane, their cellular uptake occurs primarily via nucleoside-specific membrane transport carriers (NT). Of the seven distinct NT activities observed in human cells, only four have been defined in molecular terms. These are classified into two categories: equilibrative (ENT) or concentrative (CNT).10,11 Human equilibrative NTs (hENTs) are found in virtually all studied cell types and have broad permeant selectivity, accepting both purine and pyrimidine nucleosides as substrates. The two cloned hENTs differ in their sensitivity to inhibition by nanomolar concentrations of nitrobenzylmercaptopurine ribonucleoside (NBMPR): hENT1 has es activity (equilibrative and NBMPR-sensitive) while hENT2 possesses ei (equilibrative and NBMPR-insensitive) nucleoside transport activity.12,13,14 The human concentrative nucleoside transporters (hCNT) are inwardly-directed transporters that are capable of transporting nucleosides against a concentration gradient by utilizing the transmembrane sodium concentration gradient.15,16,17 The hCNT1 protein has greater affinity for pyrimidine nucleosides but also transports adenosine, while the hCNT2 protein transports purine nucleosides and uridine.18 In all cases, human NTs accept only dephosphorylated compounds.
Deoxyribonucleotide triphosphate (dNTPs) pools present within cells come from two sources, the de novo pathway which is specifically activated in replicating cells, and the salvage pathway, which is the main source of nucleotides in quiescent cells.19,20 The key step in the de novo pathway is the conversion of ribonucleoside diphosphates into deoxyribonucleoside diphosphates by ribonucleotide reductase (RR). Replicating hematopoietic cells are heavily dependent on the de novo pathway. Conversely in resting cells the salvage pathway, which recycles bases and nucleosides derived from DNA or RNA catabolism, is the unique provider of dNTPs. These compounds are first phosphorylated by nucleoside kinases such as deoxycytidine kinase (dCK), thymidine kinase 1 and 2, or deoxyguanosine kinase (dGK). This initial phosphorylation step constitutes the key step in the salvage pathway that concludes with the formation of triphosphate or deoxytriphosphate derivatives. There are thus two dNTP pools: one derived from the de novo pathway that is predominantly directed into replicating DNA and a second derived from salvage synthesis used for DNA repair.21,22
Mechanisms of action of nucleoside analogues and drug metabolism
All of the NA share common characteristics including active transport by membrane transporters (Table 1), activation by kinases such as dCK allowing retention of the monophosphate residues in the cell and the formation of the active triphosphates metabolites, and dephosphorylation by 5′-nucleotidase (5′-NU) (Figure 1). Moreover, most of them have the ability to ‘self-potentiate’ their own cytotoxic effects. However each of these compounds also possesses specific properties in terms of drug–target interactions which may explain their differences in activity in various diseases. In particular the cytotoxic effects of the purine analogues fludarabine and 2-CdA on quiescent cells may be explained by interaction with targets involving DNA repair rather than replication, and direct or indirect effects on mitochondria.
Ara-C (1-β-D-arabinofuranosylcytosine, cytosine arabinoside, cytarabine) is an structural analogue of deoxycytidine (dCyd) (Figure 2) used for the treatment of acute leukemias and lymphomas. Ara-C differs from dCyd by the presence of a hydroxyl group in the β-configuration at the 2′-position of the sugar moiety. Intracellular penetration of ara-C depends on plasma ara-C concentrations.23,24,25,26 Standard-dose ara-C (SD ara-C; 100–200 mg/m2) achieves steady-state plasma levels of 0.5–1 μM.27,28 At these concentrations the expression of the hENT1 protein is the rate-limiting factor in ara-C uptake. In the presence of plasma concentrations greater than 50 μM, such as those reached with high-dose ara-C (HD ara-C; 2–3 g/m2), simple inward diffusion rates exceed those of pump-mediated transport.29
Once inside the cell, ara-C is phosphorylated by dCK and pyrimidine kinases to the active 5′-triphosphate derivative ara-CTP.30,31 The catabolism of ara-C results from rapid deamination by cytidine deaminase (CDD) to the non-toxic metabolite arabinoside uridine while ara-CMP is dephosphorylated by the action of cytoplasmic 5′-nucleotidase (5′-NU).32 Ara-C cytotoxicity is believed to result from a combination of DNA polymerase inhibition and from incorporation of ara-CTP into DNA, in competition with deoxycytidine triphosphate (dCTP). This incorporation causes chain termination, resulting in a block of DNA synthesis.33,34,35 Sustained high cellular concentrations of ara-CTP relative to that of dCTP are thought to favor drug incorporation into replicating DNA, thereby initiating the leukemic cell death associated with therapeutic response.36 Low incorporation of ara-CTP into the DNA of blast cells in vitro has been shown to be predictive of an adverse outcome in AML patients receiving ara-C-based therapy.37,38
Gemcitabine (difluorodeoxycytidine, dFdC) (Figure 2) is a new pyrimidine analogue with promising activity in solid tumors.39,40 This compound is a dCyd analogue with two fluorine substituted for the two hydrogen atoms in the 2′ position of the deoxyribose sugar.41 After the initial phosphorylation into dFd-CMP by dCK,42 gemcitabine is subsequently phosphorylated by pyrimidine kinases to the active 5′-diphosphate (dFd-CDP) and triphosphate (dFd-CTP) derivatives.43 Inactivation of gemcitabine can occur by deamination into its inactive metabolite dFd-U by CDD or dephosphorylation of dFd-CMP by 5′-NU.44,45
dFd-CTP is incorporated into DNA by replication synthesis in the C sites of the growing DNA strand.46 Once incorporated into the DNA strand, an additional natural nucleoside is added, masking gemcitabine and preventing DNA repair by base pair excision. Thereafter, the DNA polymerases are unable to proceed,47 a process designated as ‘masked DNA chain termination’. The active diphosphate metabolite of gemcitabine also inhibits DNA synthesis indirectly through inhibition of RR.47 This effect blocks the de novo DNA synthesis pathway and self-potentiates gemcitabine activity by decreasing intracellular concentrations of normal dNTPs (particularly dCTP). Reduction in cellular dCTP results in increased gemcitabine nucleotide incorporation into DNA and increased formation of active gemcitabine di- and triphosphates, since dCK activity is down-regulated by high cellular dCTP levels. Other self-potentiating mechanisms of gemcitabine include a decreased elimination of gemcitabine nucleotides by direct inhibition of CDD.48 Gemcitabine is not only incorporated into DNA, but also into RNA.49
The spectrum of activity of gemcitabine in solid tumors may be due to different characteristics. Compared with ara-C, gemcitabine serves as a better transport substrate for membrane pumps, is phosphorylated more efficiently, and is eliminated more slowly, favoring a longer retention time of the active metabolite in tumor cells.48,50 Moreover, gemcitabine cytotoxicity is enhanced by a number of unique self-potentiating mechanisms that contribute to maintaining high intracellular concentrations of the active metabolites. This prolonged presence of active gemcitabine derivatives may explain part of its effect on slowly dividing tumor cells. These differences, together with greater lipophilicity, masked chain termination, RNA incorporation and RR inhibition, which are not observed with ara-C, may explain why gemcitabine is, and ara-C is not, active in solid tumors.
2-CdA (Figure 3) is an analogue of deoxyadenosine (dAdo) resistant to deamination by adenosine deaminase (ADA). 2-CdA is commonly used in the treatment of indolent lymphoid malignancies, in particular hairy-cell leukemia.51,52,53 2-CdA enters cells via NT54 and is phosphorylated to 2-CdATP by dCK, AMP kinase, and nucleoside diphosphate kinase. Mitochondrial dGK has also been identified as a 2-CdA phosphorylating enzyme.55 2-CdAMP can be dephosphorylated by 5′-NU. The net accumulation of the triphosphate derivative has been shown to depend on the relative concentrations of dCK and 5′-NU.56
2-CdA is cytotoxic both to dividing and resting cells.57,58 In dividing cells, this compound inhibits DNA synthesis by incorporation of its triphosphate metabolite into DNA.59,60 Once incorporated 2-CdATP is capable of terminating chain elongation mediated by DNA polymerases, inducing an S phase-specific apoptosis.61 2-CdA also inhibits DNA replication indirectly through its inhibitory action on RR,62 causing a subsequent reduction of the dNTPs pool required for DNA synthesis.59,63 The reduction of intracellular dATP results in self-potentiation of 2-CdA as it enhances the incorporation of triphosphate metabolites into DNA. Moreover, the decrease in dCTP could result in an increased rate of phosphorylation of 2-CdA and other NA, since dCK activity is regulated by cellular dCTP levels. An original mechanism of 2-CdA cytotoxicity is related to the inhibition of ADA and S-adenosylhomocysteine hydrolase (SAHH).64 Inhibition of ADA could lead to increased levels of Ado and dAdo, thereby causing the inactivation of SAHH and reduced homocysteine formation. This might result in altered methylation reactions and indirectly inhibit nucleotide synthesis and cellular proliferation.65,66
Different mechanisms have been proposed to explain why 2-CdA is toxic to a cell that is not undergoing replicative DNA synthesis. First, 2-CdA-induced killing in non-dividing cells occurs by inhibition of cellular DNA repair processes. Incorporation of 2-CdATP into DNA by the repair machinery terminates the nucleotide excision repair process leading to the progressive accumulation of DNA single-strand breaks which eventually initiate apoptosis by p53-dependent or p53-independent pathways.67,68 p53 activation subsequently regulates the levels of expression of p53-dependent proteins such as Bcl-2 and Bax leading to the activation of the caspase pathway.69 Second, incorporation of 2-CdA into DNA may alter gene transcription with a consequent depletion of proteins required for cell survival. It has been demonstrated that when 2-CdA metabolites were present in one or both DNA strands, the yield of full-length transcripts was reduced.70 Moreover, RNA polymerase did not bind as well to 2-CdATP-containing promoters. In addition, on binding to the substituted promoter, RNA polymerase had an altered conformation that led to enhanced proteolytic clipping by endoproteinase Glu-C.70
Alternatively, it has been shown that 2-CdA cytotoxicity may involve alterations of mitochondrial function or integrity. 2-CdA has been shown to disrupt the integrity of mitochondria in CLL cells.69 The nucleoside-induced damage leads to the release of the pro-apoptotic mitochondrial protein cytochrome c, thereby initiating the caspase proteolytic cascade. Moreover, 2-CdATP can cooperate with cytochrome c and Apaf-1 to activate caspase-3 and trigger a caspase pathway.71,72 2-CdA also interferes either directly or indirectly with mitochondrial transcription that will eventually reduce the levels of mitochondrial proteins necessary for electron transport and oxidative phosphorylation. In vitro studies have demonstrated that 2-CdATP is incorporated into mitochondrial DNA by mitochondrial DNA polymerase.62 This incorporation may decrease transcription by mitochondrial RNA polymerase. On the other hand, high intracellular 2-CdATP levels could indirectly repress transcription of mitochondrial genes by inhibiting reduction of ADP to dADP.73 This leads to high ADP concentrations that cause a net influx of adenine nucleotides into mitochondria and an increase in the intramitochondrial ATP concentration.59,74 High intramitochondrial ATP levels in turn inhibit mitochondrial RNA polymerase activity.74
PARP activation has been suggested to be involved in p53-independent mechanisms of cell killing by 2-CdA.57 In addition to being a target for proteolytic cleavage during apoptosis, PARP is thought to play a role in DNA repair by signaling genomic damage.75 PARP binds to DNA breaks, becomes activated and, using NAD as a substrate, adds ADP-ribose polymers to a range of nuclear proteins including itself, histones and DNA repair enzymes.76,77 These and other findings have led to a proposed model of 2-CdA action involving the following sequence of events: inhibition of ongoing DNA repair by 2-CdA with accumulation of DNA breaks, PARP activation with resultant consumption of NAD, and depletion of total adenine nucleotides.57,76,77,78 Given that the cell membrane requires ATP for many of its functions, it seems likely that PARP-mediated ATP depletion might contribute to the cell membrane disruption that occurs during apoptosis.79
9-β-D-arabinosyl-2-fluoroadenine (F-ara-A) (Figure 3) is an analogue of adenosine (Ado) which is resistant to deamination by adenosine deaminase (ADA).80 F-ara-A is administered as the 5′-monophosphate form (fludarabine). The phosphate group on the 5′- position of the arabinose has the sole function of making the clinical formulation water-soluble. Before entering cells, fludarabine is rapidly dephosphorylated by membrane CD73 into F-ara-A which is transported inside the cell via hNTs.81 Like other NA, it requires phosphorylation to its triphosphate form, F-ara-ATP for cytotoxic activity.82,83 The initial step in this activation process is performed by dCK.
Like 2-CdA, F-ara-A is cytotoxic both against dividing and resting cells. In dividing cells, this compound inhibits DNA synthesis and RR.82 DNA synthesis inhibition is mediated by competition of the triphosphate metabolite with deoxyadenosine triphosphate (dATP) for incorporation into the A sites of elongating DNA by DNA polymerases.84 Once incorporated it is capable of terminating chain elongation mediated by DNA polymerases α (F-ara-ATP).85 RR inhibition causes a subsequent reduction of dNTPs pools.63,86 Similarly to 2-CdA, the decrease of cellular dATP and dCTP results in a self-potentiation of the drug and other NA. Other enzyme targets of F-ara-ATP in dividing cells include DNA primase, DNA polymerase α, DNA ligase and topoisomerase II.84 Thus, the mechanism of action of F-ara-A in proliferating cells is mainly cell cycle-specific, and incorporation of F-ara-A into DNA during S phase is required for the induction of apoptosis.87
In non-dividing cells, the inhibition of cellular DNA repair processes appears to be the major mechanism of cytotoxicity of F-ara-A. Incorporation of F-ara-AMP into DNA by the repair machinery terminates the polymerization step of the nucleotide excision repair process causing irreversible damage recognized by the cells, which would signal for p53-mediated or PARP-mediated apoptosis.20 As occurs after exposure to 2-CdA, fludarabine may induce PARP activation resulting in consumption of NAD and depletion of total adenine nucleotides with subsequent cell death.88
Other mechanisms of action of F-ara-A may explain its cytotoxicity in non-cycling cells.82 First, F-ara-AMP and F-ara-ATP incorporation into RNA results in premature termination of the RNA transcript, impairing its function as a template for protein synthesis.89 F-ara-ATP also inhibits RNA synthesis by suppressing the activity of RNA polymerase II.89 Inhibition of RNA transcription correlates with the cytotoxic action of fludarabine in CLL cells.90 Secondly, F-ara-ATP is the most potent nucleotide activator of APAF-1, with the consequent activation of the caspase-9 and caspase-3 pathways.71 Moreover, F-ara-A has been reported to downregulate Bcl-2 mRNA expression, thereby favoring apoptosis.91 Finally, F-ara-A affects mitochondria but only as a late event, suggesting an indirect mechanism related to the p53-apoptotic pathway.69 The different effects of F-ara-A and 2-CdA on mitochondria integrity may be related to the fact that 2-CdATP may accumulate more rapidly than F-ara-ATP in the mitochondria because it is converted with higher efficiency by mitochondrial dGK.92 It has also been stated that 2-chloro derivatives are slightly more hydrophobic than the corresponding 2-fluoro derivatives and therefore may be able to insert more easily into the mitochondrial membrane, thereby affecting their integrity.69 In summary inhibition of DNA repair, termination of mRNA transcription and the consequent depletion of proteins required for cell survival, as well as the capacity of F-ara-ATP to activate the apoptosome pathway appear to contribute to the cytotoxicity of fludarabine in non-dividing cells.
Mechanisms of resistance to nucleoside analogues
Three general mechanisms of resistance to NA have been described in cell lines and in clinical samples. A primary mechanism of resistance to NA arise from an insufficient intracellular concentration of NA triphosphates, which may result from inefficient cellular uptake, reduced levels of activating enzymes, increased NA degradation by increased 5′-NU or CDD activity, or expansion of the dNTP pools.93 A second type of resistance mechanism may be due to the inability to achieve sufficient alterations in DNA strands or dNTP pools, either by altered interaction with DNA polymerases, by lack of inhibition of RR, or because of inadequate p53 exonuclease activity. Finally, drug resistance may be a consequence of a defective induction of apoptosis.
In vitro studies have demonstrated that NT-deficient cells are highly resistant to NA.94 Several cell lines exhibit cross-resistance to a spectrum of NA because of reduced uptake.11,54,95 In isolated leukemic blasts, sensitivity to ara-C, fludarabine and 2-CdA correlated with the cellular abundance of hENT196,97 determined by cellular binding of NBMPR. In these cells, the number of such binding sites ranged from 500 to 27600 sites/cell.11,26 As initial rates of cellular drug uptake are, in general, proportional to hENT1 abundance, this wide variation suggests that transport-deficient cells would display a drug-resistant phenotype. In AML patients, Wiley et al98 observed that low transport rates correlated with poor clinical responsiveness to ara-C therapy. Although the measurement of hENT1 expression by a quantitative flow cytometric assay with the hENT1 ligand, SAENTA fluorescein, shows promise in identifying transport-deficient leukemic cells, the clinical significance of NT deficiency to in vivo resistance is not yet well established. This is due to the difficulty of assessing drug transport in clinical samples, the contribution of transporters other than hENT1, and the technical difficulties of quantifying NTs in neoplastic clones admixed with normal cells.11
Deoxycytidine kinase is the rate-limiting enzyme that catalyzes the initial step of the 5′-phosphorylation of dCyd to dCMP and of many NA to their corresponding monophosphates.99,100,101 The dCK activity is high in quiescent cells in which it phosphorylates nucleosides necessary for DNA repair and, depending on the cell type, may increase several-fold when the cells enter the S-phase of the cell cycle.102,103,104,105
Deoxycytidine kinase activity has been shown to be decreased or absent in different cell lines resistant to NA.106,107,108,109 Transfection of the dCK gene in dCK-deficient tumor cell lines restores in vitro sensitivity to ara-C.110,111 Moreover, in vitro models have shown crossresistance between 2-CdA, gemcitabine, fludarabine and ara-C with reduced dCK activity as the underlying determinant of resistance.109,112 It is noteworthy that in vitro only significant reductions in the levels of dCK activity (10% or less) are associated with a resistant phenotype, suggesting that under baseline conditions dCK is present in excess.
In the clinical setting, it is important to note that bone marrow and lymphoid tissues exhibit the highest activities of dCK resulting in the clinical success observed when treating these target tissues with NA.113,114 However, the relevance of dCK deficiency in clinical resistance to NA is controversial (Table 2). In childhood acute lymphoblastic leukemias (ALL),115,116 AML117,118 and lymphoproliferative disorders119 some authors observed decreased expression of the dCK gene or significant decrease in the activity level of this enzyme as one of the mechanisms responsible for clinical resistance to ara-C, fludarabine or 2-CdA. Conversely, other authors reported no significant relationship between dCK activity and clinical outcome to NA in these diseases.120,121
Structural analyses of the dCK gene have show inactivating mutations and deletions as a cause of dCK deficiency in vitro. Owens and co-workers122 claimed that to acquire a drug-resistant phenotype, cells required two independent mutations that markedly disrupt the catalytic activity of the enzyme. However, these cDNA mutations rarely occur in vivo and therefore may not constitute a major mechanism of clinical NA resistance.115,116,123 Moreover, it seems that most of the aminoacid replacements did not occur in a functionally relevant region of the enzyme.124 More recently alternatively spliced forms of dCK mRNA were detected in leukemic blasts from patients with clinically resistant AML but not in leukemic blasts from patients with sensitive AML.125 These alternatively spliced forms of dCK mRNA were shown to be inactive in an in vitro dCK activity assay. These data indicate that the presence of dCK splice products may contribute to the occurrence of clinical ara-C resistance.
Structural alterations of the dCK gene complex other than mutations may play a role in drug resistance to NA in vivo.126 Downregulation of dCK activity in cytarabine-resistant cells has been shown to correlate with hypermethylation of the promoter region of the gene.127,128 Decreased levels of dCK mRNA has been shown to be induced under various conditions in vitro.127,129 Finally dCK activity may be regulated post-translationally, since this enzyme is feedback-inhibited by dNTPs (particularly dCTP) and by some phosphorylated analogues such as ara-CTP.130
Deoxyguanosine kinase (dGK) is another enzyme responsible for the phosphorylation of purine deoxynucleosides, in mammalian cells. This enzyme showed a broad substrate specificity towards natural purine deoxynucleosides, as well as towards nucleoside analogues such as ara-G and 2-CdA.55,73,92 dGK is located predominantly in the mitochondria.131 However, Petrakis et al132 described a novel amino-terminally truncated isoform that corresponds to about 14% of the total dGK mRNA population in mouse spleen. This isoform cannot translocate into the mitochondria and thus may represent a cytoplasmic enzyme.
Mitochondrial dGK is responsible for phosphorylation of purine deoxyribonucleosides and their analogues in the mitochondrial matrix providing the dNTPs necessary for mitochondrial DNA synthesis.133,134,135,136 Cytoplasmic dGK may also contribute to the supply of purine deoxynucleotides for nuclear DNA replication and repair.132 In this way, nucleoside analogues phosphorylated by dGK may exert their cytotoxic effects by interference with both mitochondrial and nuclear DNA synthesis. Whether inactivation of dGK is involved in resistance to NA is not presently known.
5′-NU dephosphorylates nucleoside monophosphates and cytotoxic mononucleotides through its hydrolysis of the ester-bond between the 5′-carbon and the phosphate group. 5′-NU activity therefore opposes that of dCK. 5′-NUs comprise a large and complex group of enzyme activities differing in cellular localization, pH sensitivity, and inhibition or stimulation by ATP (Table 3). Two broad classes of 5′-NU are implicated in drug resistance to NA: cytosolic and membrane-bound 5′-NUs. Membrane-bound or ‘ecto-5′-NU’ is also known as CD73, and is frequently expressed in acute leukemias.137,138 The physiological function of CD73 is the salvage of extracellular nucleotides into nucleosides capable of cellular entry by means of nucleoside transport proteins. In contrast, several intracellular 5′-NU activities have been identified and cloned to date.139,140,141 The ‘low-Michaelis-Menten constant’ (Km) 5′-NU is specific for pyrimidine monophosphates at low micromolar Km values and is inhibited by ATP. The other, high Km 5′-NU is purine selective at high micromolar Km values, and it is activated by ATP.142,143 In vitro, increased 5′-NU activity has been consistently associated with nucleoside drug resistance.45,58,109
In the clinic, levels of cyto-5′-NU activity have been correlated to clinical response in CLL, HCL and AML (Table 4). In CLL and HCL, Kawasaki et al119 demonstrated that pretreatment levels of high Km 5′-NU allowed prediction of 2-CdA responsiveness, since 5′-NU levels were significantly lower in responders than in non-responding patients. In AML, we have reported that patients whose blasts expressed high Km 5′-NU mRNA had shorter time-to-relapse and overall survival than patients with no expression.144 However, due to the multiplicity of 5′-NU activities, and the limited ability of classical biochemical methodology to differentiate between these activities, the particular 5′-NU enzyme(s) with the largest impact on clinical resistance to NA have not yet been convincingly identified.145
Cytidine deaminase catalyses the inactivation of cytidine and dCyd to uridine and deoxyuridine, respectively. CDD also deaminates ara-C and gemcitabine.146 Several lines of evidence have indicated a role for increased levels of CDD in the development of resistance to ara-C.147,148,149 Different studies have revealed that transfection of the human CDD gene into murine fibroblasts and hematopoietic cells confers drug resistance to ara-C and gemcitabine.44,150,151
However, the in vivo correlation between CDD activity and ara-C resistance remains controversial and the relative contribution of CDD to drug resistance has not yet been fully elucidated (Table 4). The clinical relevance of a high level of CDD activity as a major cause of ara-C resistance in AML patients has been emphasized by several investigators.118,152,153,154 Conversely, Tattersal et al117 did not find a relationship between increased CDD activity and resistance to therapy in AML patients.
Structural analyses of the CDD gene correlated polymorphism at codon 27 to substantially different deamination rates of ara-C in vitro.155 This structural aberration did not seem to represent a major cause of the observed differences of CDD activities between ara-C-sensitive and ara-C-resistant patients. Schröder and co-workers154 demonstrated a significant correlation of the amount of CDD mRNA with CDD enzyme activities in AML blasts suggesting that variations in CDD activity result from differences in gene expression. It therefore appears that CDD activity in vivo is correlated with transcriptional regulation rather than with CDD gene aberrations.
Levels of dCTP
Various cell lines and AML blasts containing high levels of dCTP have been found to be resistant to ara-C.126,156,157 dCTP levels regulate ara-C metabolism at three levels: by feedback inhibition of dCK activity with the consequent decrease in ara-C phosphorylation;117 by allosteric activation of the catabolic enzyme, CDD;126 and by competing with ara-CTP for incorporation into DNA.158,159 Moreover, Ohno and coworkers160 have shown that dCTP in sufficiently high concentrations can overcome ara-C DNA strand termination so that strand elongation continues past the ara-C incorporation site. dFd-CTP also compete with dCTP for DNA polymerase and for incorporation into DNA.46 Thus, elevated levels of dCTP will favor the survival of cells exposed to these NA.
Ribonucleotide reductase and DNA polymerases
RR consists of two non-identical proteins called M1 and M2, both of which are required for activity. Protein M1 contains the substrate and effector binding sites and is present throughout the cell cycle in proliferating cells. Protein M2 is a dimer that contains stoichiometric amounts of non-heme iron and a unique tyrosyl free radical essential for reductase activity. Its expression is specific of S phase and in quiescent cells M1 and M2 mRNA and proteins are not detectable.161 Clinically useful NA target RR, altering dNTPs concentrations.62,162 Recently, Goan et al163 reported a human KB cancer cell line resistant to gemcitabine with overexpression of the M2 subunit as the sole mechanism explaining this resistance.163
Most of the NA inhibit DNA polymerase activity by competing with dNTPs,164 thereby contributing to the inhibition of DNA synthesis.165 The degree of inhibition of DNA polymerases is not the same for all the NA. For example, DNA polymerase α has been shown to be more sensitive to inhibition by ara-CTP than DNA polymerase β.166 F-ara-ATP is not a potent inhibitor of DNA polymerases β and γ or DNA primase.84,162 In vitro, overexpression of DNA polymerases may be a significant factor in the development of drug resistance.167 Tanaka et al168 found that DNA polymerase α sensitivity to ara-CTP and F-ara-ATP was lower in blast cells obtained from ALL patients than in cells obtained from AML patients, and concluded that DNA polymerase α from ALL blast cells has a decreased affinity for NA.168
Defective cell death pathways and p53 exonuclease activity
DNA damage caused by NA induces expression of p53, leading to the induction of pro-apototic molecules such as Bax and to down-regulation of anti-apoptotic proteins, such as Bcl-2.169,170,171 These events lead to induction of apoptosis.172,173 Various investigators have shown that B-CLL and AML patients with mutations in the p53 gene displayed resistance to fludarabine, 2-CdA or ara-C and/or poorer response and shorter remission and/or overall survival174,175,176,177 (Table 5). Conversely, other authors failed to find any relationship between p53 abnormalities and resistance to NA in these diseases.178,179 Other genes regulating apoptosis, such as Bcl-2, Bcl-X/L, Mcl-1 or Bag-1, have been suggested to influence sensitivity to NA, although the results remain controversial.180,181,182,183,184,185,186,187
Alterations in p53 function have been related to in vitro resistance to NA.68,188 p53 appears to have a specific role in resistance to NA since it possesses exonuclease activity, and may therefore be able to excise illegally DNA-incorporated NA. Recently, Feng et al reported that although the 3′-5′ exonuclease of wild-type p53 (wt-p53) protein was able to bind and excise gemcitabine residues from DNA in vitro, removal of the drug molecules from cellular DNA was slow in sensitive cells containing wt-p53, and undetectable in drug-resistant mut-p53 cells. Collectively, these data imply that molecular regulators of apoptosis should be taken into consideration when analyzing factors associated with resistance to NA.
It is now widely accepted that the antileukemic effect of NA requires sufficiently high intracellular concentrations of triphosphate analogues in tumor cells.38,82,189 One approach has been to develop compounds which are insensitive to degradation. Fludarabine and cladribine are resistant to the effect of ADA. Another approach could consist in the use of specific inhibitors of degradative enzymes, such as 5′-NU. Combinations of a 5′-NU inhibitor with a NA could be expected to potentiate the cytotoxicity of the NA. However, there is still uncertainty regarding the relevant 5′-NU enzyme(s) that mediate NA resistance, and there are currently no clinically available inhibitors of 5′-NU activity. Because many potential inhibitors of 5′-NU are themselves nucleoside analogues, competition with cytotoxic NA for membrane transporter sites may be problematic. Regulation of dCK activity could also be a target for modulation. It has been shown that dCK activity is stimulated after treatment with 2-CdA, as well as with fludarabine and ara-C in normal human lymphocytes and various leukemic cell lines.190,191 The increase of dCK activity induced by NA was considered to result from post- translational modifications of dCK during inhibition of DNA synthesis.192,193
The importance of the ratio of nucleoside analogues to normal nucleosides has been recognized more recently. Reducing the intracellular pools of normal nucleotides therefore appears to be a therapeutic alternative to increasing the concentrations of cytotoxic NA triphosphates. In practice, given the feedback inhibitory effects of dNTPs on NA activation, these two goals may be attained simultaneously. Various biochemical modulation strategies, based on NA combinations, have been suggested to increase the concentration of NA triphosphate in leukemic cells.
Inhibition of ribonucleotide reductase
Large amounts of dNTPs generated by RR are directed into replicating DNA, a process that effectively excludes NA from replicating DNA synthesis.194 Furthermore intracellular dNTP pools inhibit dCK activity, thereby reducing NA activation. Thus, drugs capable of depleting or reducing normal dNTPs by inhibiting RR might be expected to potentiate cytotoxicity of different NA. Sequential incubation of K562 human leukemia cells with F-ara-A followed by ara-C enhanced ara-CTP accumulation, thereby suggesting possible therapeutic synergism between ara-C and fludarabine.86 Subsequent clinical studies have demonstrated the effectiveness of this strategy.195,196 Lymphocytes isolated from patients with CLL and incubated with ara-C after in vitro F-ara-A incubation or a therapeutic infusion of fludarabine (25 or 30 mg/m2) displayed an increase in ara-CTP accumulation.197 The same phenomenon was observed in leukemic lymphocytes and AML blasts.198,199 This increase was caused by a higher rate of anabolism rather than by a slower rate of catabolism of ara-CTP. More recently, Seymour et al200 have shown that F-ara-ATP concentrations (30–50 μM, achieved with usual clinical doses of fludarabine), enhances both intracellular ara-CTP accumulation and incorporation into cellular DNA at concentrations of ara-C which are clinically achievable. Such drug combinations have induced responses in untreated and refractory patients with acute leukemias.201,202
This modulatory effect of fludarabine can be extended to other NA such as 9-β-D-arabinofuranosylguanine (ara-G).203 In vitro studies using human leukemia cell lines and primary leukemia cells obtained from patients established that the effectiveness of ara-G is due to intracellular accumulation of ara-GTP. The rate-limiting step of ara-GTP accumulation is the initial phosphorylation performed both by dGK and dCK. In vivo, Gandhi and coworkers demonstrated in a phase I clinical trial that patients who responded to therapy with the water-soluble ara-G prodrug GW506U achieved a significantly greater concentration of ara-GTP in their circulating cells than patients who did not respond.204,205 In this setting, the combination with fludarabine could increase the intracellular accumulation of ara-GTP in target cells and improve response rates in patients.205
Other inhibitors of RR can be used for the modulation of ara-C metabolism. Iwasaki and coworkers194 showed that inhibition of RR with gemcitabine and the consequent decrease in cellular dCTP pools favored ara-C incorporation into replicating DNA. Colly et al206 and Bhalla et al207 tested the ability of the RR inhibitor hydroxyurea to enhance ara-C metabolism and cytotoxicity in drug-sensitive and -resistant cell lines. This combination significantly reduced dCTP pools, increased DNA incorporation of ara-C and enhanced drug cytotoxicity, even in dCK deficient ara-C-resistant cell lines.
Similarly, as 2-CdA inhibits the reduction of CDP and ADP, a decrease in the dCTP and dATP pools is expected when cells are exposed to this compound. The decrease of these pools would increase the ratios of 2-CdATP to dATP and ara-CTP to dCTP. In vitro incubation of AML blasts with 2-CdA followed by ara-C produced a higher rate of ara-CTP accumulation than did ara-C alone.208 Gandhi and coworkers209 studied 17 patients with refractory AML receiving combination chemotherapy with ara-C and 2-CdA. Seven of nine patients studied during therapy had an increased rate of ara-CTP formation and peak ara-CTP concentrations in leukemic blasts. This modulation strategy increased the effective dose intensity of the active metabolite of ara-C by 40% in the tumor cells. The co-administration of 2-CdA and ara-C resulted in maximum inhibition of DNA synthesis, accumulation of higher concentrations of ara-CTP, greater ratios of analogue triphosphates to normal dNTPs and tandem incorporation of 2-Cl-dATP and ara-CTP in growing DNA strands.
Although potentially interesting, the usefulness of NA combinations in the clinic remains to be determined. In particular the toxicity to normal cells due to the increase of NA accumulation appears to be significant, casting doubt on an improvement of the therapeutic index in comparison to single agent NA therapy. A potential consequence related to NA modulation therapy may be the increased risk of secondary malignancies in a group of patients whose disease already places them at a great risk of second cancers, and dose-related effects like severe neurotoxicity.210,211 Apparently, the frequency and severity of neurotoxicity may be greater with the combination therapy than with either drug used alone.
Modulation of dCK and 5′-NU activitity with bryostatin
Bryostatin, a macrocyclic lactone and protein kinase C activator, has been shown to increase the ratio of dCK/5′-NU activity and thereby increase sensitivity to fludarabine and cladribine in CLL cells, both in vitro and in vivo.212,213 Phase I studies of single agent bryostatin have been completed and this compound is now in clinical trials in combination with NA.214
Use of hematopoietic growth factors
Different in vitro studies have shown that pretreatment of AML blasts by growth factors (GF) enhances ara-C mediated cytotoxicity against leukemic cells, possibly through their proliferation-inducing effects.215,216 GF may also enhance ara-C cytotoxicity by increasing hENT1 expression. High cellular proliferation rates are associated with high levels of hENT1- mediated activity that increase the nucleoside drug uptake.217,218 Wiley et al219 determined that in vitro treatment of isolated human leukemic cells with GM-CSF, resulted in an increase in hENT1 expression. Reuter et al220 established that blasts from AML patients treated with GM-CSF displayed an enhanced drug uptake after low or SDAC doses. In this study, GM-CSF also increased intracellular ara-CTP/dCTP pool ratios and enhanced ara-C incorporation into DNA by activation of DNA polymerases. In vivo, GM-CSF priming of leukemic blasts prior to induction therapy translates into a higher antileukemic activity as indicated by a prolonged survival of treated mice with advanced AML and a high rate of complete remission in patients with AML.221,222 However, various randomized clinical trials failed to demonstrate the effectiveness of the GF-priming strategy and concluded that administration of GF during and after induction chemotherapy does not improve the clinical outcome of ara-C-treated AML patients.223,224
DNA repair induction
Another way of enhancing incorporation of NA into DNA consists of using the DNA repair pathway after combined exposure of tumor cells to NA and DNA damaging agents.225 The recognition and repair of the damaged strand caused by the DNA damaging agent allow resynthesis using the opposite strand as a template. This creates an opportunity for the insertion of the NA instead of its normal counterpart into the DNA repair patch. The incorporated NA is relatively resistant to repair excision and causes irreversible damage recognized by the cell. Alternatively inhibition of DNA repair by NA may increase the accumulation and slow the removal of DNA lesions induced by DNA damaging agents, thereby potentiating the cytotoxicity of these latter compounds. In vitro, combination of gemcitabine and fludarabine with cisplatin generated synergistic cytotoxicity.225 Van Den Neste et al226 recently reported the in vitro synergistic cytotoxicity of 2-CdA and cyclophosphamide in B-CLL cells isolated from patients pretreated with alkylating agents.
In vivo, the combination of a DNA damaging agent and a NA may prove to be particularly useful to destroy cancer cells without significant DNA replication activity such as CLL and indolent lymphomas. The combination of fludarabine with doxorubicin and prednisone is clinically active against CLL.227 Fludarabine and ara-C with or without cisplatin, have been administered to patients with CLL with cytoreductive activity.228 Results of clinical studies using mitoxantrone (which induces protein-associated DNA strand breaks), fludarabine, and dexamethasone demonstrated a high response rate, especially in follicular lymphomas, with a number of patients achieving complete remission.229,230 A combination of mitoxantrone and ara-C for the treatment of acute leukemia has been encouraging.231,232 Moreover, clinical trials evaluating NA in combination with DNA damaging agents have shown effectiveness against solid tumors with low growth fractions.233,234
New nucleoside analogues
The broadening of the spectrum of nucleoside analog activity to solid tumors has triggered renewed interest in this class of antimetabolites. Many new nucleoside drugs have been synthesized and studied over the past decade, with several that are in early phase clinical trials, and others that have become standardly prescribed agents (gemcitabine and capecitabine). We will review the most promising of these agents: troxacitabine, and the GW506U78, an ara-G prodrug. We have excluded from this discussion 5-FU prodrugs such as capecitabine since their mechanisms of action and resistance mechanisms differ substantially from those of other anticancer nucleosides.
Troxacitabine (BCH-4556; (−)-2-(S)-hydroxymethyl-4-(s)-(cytosin-1′-yl)-1,3-dioxolane; β-L-dioxalane cytidine) (Figure 4) is a stereochemically synthetic nucleoside analogue that has potent antitumor activity in preclinical models.235,236,237 It is phosphorylated in vitro by deoxycytidine kinase (dCK) with a Km similar to its natural substrate, dCyd, but unlike gemcitabine and ara-C, it is not a substrate of CDD. Troxacitabine triphosphate can be incorporated into cellular DNA,235 which causes rapid chain termination. However, neither ribonucleotide reductase nor mitochondrial DNA synthesis are inhibited by troxacitabine in cell culture. Because of its broad preclinical spectrum with cytotoxicity both against leukemic and epithelial malignancies, and promising phase I activity demonstrated in acute leukemia,238 phase II testing is now being carried out in several diseases.
In 1983, Cohen et al239reported that the deoxyguanosine derivative ara-G (9-β-D-arabinofuranosylguanine) (Figure 5) was resistant to cleavage by purine nucleoside phosphorylase (PNP) and was toxic to T-lymphocytes. The devolopment of this drug was limited by water insolubility. However, 2-amino-9-β-D-arabinosyl-6-methoxy-9H-guanine (GW506U78) is 10-fold more soluble, and is rapidly converted to ara-G by plasma adenosine deaminase activity.
In phase I clinical testing, this agent has shown particular promise in the therapy of T cell malignancies,240 where 54% of patients achieved partial or complete remissions after one to two cycles of drug treatment. Of particular interest, neurotoxicity was dose limiting, with little clinical myelosuppression.241 This raises the possibility of successful combination therapy with other active agents, including other hematologically active NA. Although no formal studies have been performed to demonstrate mechanisms of ara-G resistance, accumulation of ara-GTP in leukemic blasts correlated with the cytotoxic activity against malignant cells.241 This suggests that the early steps of uptake and metabolism of ara-G may be major determinants of cellular drug sensitivity, as for other NA.
During the past decade there has been dramatic progress in the understanding of the individual mechanisms of resistance to nucleoside analogues, including the role of cellular drug uptake, drug metabolism, interaction with cellar targets, and the resulting apoptotic cascade. However, most studies have focused on a single mechanism of resistance, thereby failing to provide the relative importance of the different known mechanisms of resistance. It is clear that malignancies are heterogeneous in terms of mechanisms of resistance, and that the clinical benefit of resistance-reversal strategies will only be demonstrated when targeted to the appropriate subgroup of patients.
The ultimate goal of research into anticancer therapy remains improving patient survival. However, in diseases such as CLL or indolent lymphoma in which several treatments are effective but not curative, a realistic goal would be the identification of tumor cell traits which could allow tailored or targetted therapy. In such patients, a predictive assay to determine a phenotype of probable resistance to certain NA would be of great value in the choice of optimal therapy, and the avoidance of ineffective but toxic treatments.
In diseases such as AML in which curative drug combinations exist, the objective is to optimize the efficacy of currently available NA and to develop newer, more effective analogues that are less susceptible to the resistance mechanisms described above. Efforts to increase intracellular levels and DNA incorporation of phosphorylated NA are very promising. NA therapy combined with agents modulating apototic responses are expected to provide additional benefit. In the same way that combination chemotherapy has provided curative treatment of certain cancers, a multifactorial approach of drug resistance should allow significant progress in the treatment of currently chemoresistant disease.
Carlos Galmarini is a recipient of the ‘Michel Clavel’ grant. The authors acknowledge the support of the Ligue Contre le Cancer du Rhône and the Alberta Cancer Foundation and National Cancer Institute of Canada.
About this article
High expression of the human equilibrative nucleoside transporter 1 gene predicts a good response to decitabine in patients with myelodysplastic syndrome
Journal of Translational Medicine (2016)