Cellular Phosphorylation of Anti-HIV Nucleosides ROLE OF NUCLEOSIDE DIPHOSPHATE KINASE*

analogs are used in antiviral ther- apy and particularly against AIDS. Delivered to the cell as nucleosides, they are phosphorylated into their ac- tive triphospho derivative form by cellular kinases from the host. The last step in this series of phosphorylations is performed by nucleoside diphosphate (NDP) kinase, an enzyme that can use both purine or pyrimidine and oxy- or deoxynucleotides as substrates.

The replicative cycle of human immunodeficiency viruses (HIV), 1 involves the action of a reverse transcriptase that is an important target for chemotherapeutic intervention. Indeed, nucleoside analogs substituted on the 3Ј-OH of the ribose such as azidothymidine (AZT) and dideoxyinosine (ddI) are powerful inhibitors of this enzyme; they act as chain terminators inhibiting virus replication specifically, due to the fact that cellular DNA polymerase binds to these analogs with a low affinity as compared to viral reverse transcriptase (1,2). However, to be active, nucleoside analogs need to be phosphorylated into triphosphonucleotides by cellular kinases since HIV does not carry genes for enzymes that metabolize purine and pyrimidine nucleotides. The reactions leading to mono-and diphosphates of the nucleosides are catalyzed by base-specific enzymes, i.e. the phosphorylation of purines and pyrimidines in the cell is catalyzed by distinct nucleoside kinases and nucleoside monophosphate kinase. In contrast, the step leading from the nucleoside diphosphate to the triphosphate is catalyzed by a single enzyme, nucleoside diphosphate (NDP) kinase, independent of the nature of the base and of the sugar (EC 2.7.4.6) (3).
The main function of NDP kinase in the cell is to phosphorylate the non-adenine nucleoside diphosphates into triphosphates. The reaction has a ping-pong mechanism, with a phosphohistidine intermediate according to the following reactions (Reactions 1 and 2).
ATP is believed to be the main phosphate donor in the cell. Renewed interest in this enzyme resulted recently from its cloning from several species including the prokaryote Myxococcus xanthus (4), the primitive eukaryote Dictyostelium discoideum (5), and higher eukaryotes including mammals. Two highly homologous NDP kinases, NDPK-A and NDPK-B, have been isolated in human erythrocytes and sequenced (6), and these proteins were identified to the products of the genes nm23-H1 and nm23-H2, respectively (6, 7). nm23-H1 has been shown to be involved in tumor metastasis (8,9). All NDP kinases are made of identical 17-kDa subunits. Eukaryotic NDP kinases are hexamers, whereas some bacterial enzymes are tetramers. The high resolution structure of the NDP kinases from Dictyostelium (10,11), M. xanthus (12), Drosophila (13), and human (14,15) show that the subunit fold and active site of NDP kinases are highly conserved throughout evolution. This fold is original for a phosphotransferase, showing no similarities with the usual nucleotide binding fold of nucleotidebinding proteins. High resolution data are also available for Dictyostelium and Myxococcus NDP kinase complexed with ADP, a purine nucleotide (15,16), for Dictyostelium complexed with TDP, a pyrimidine deoxynucleotide (17), and for human NDPK-B complexed with GDP (14). These data, along with the study of several mutant proteins modified in active site residues by in vitro mutagenesis (18), provide a comprehensive description of the active site.
Nucleoside analogs are thought to be phosphorylated by the same enzymes as the natural nucleotides. For example, thymidine kinase and thymidylate kinase catalyze the first and second steps in the phosphorylation of AZT. However, AZT-MP is a poor substrate for thymidylate kinase and accumulates in the cell (19), which may be responsible for a major part of its cytotoxic effects (20). In contrast to the numerous studies performed on AZT phosphorylation to AZT-MP and AZT-DP by thymidine kinase and thymidylate kinase, no study is available on the last step in the phosphorylation cascade, i.e. the phosphorylation of AZT-DP in AZT-TP. This may be due to the lack of specificity of NDP kinase toward the nucleobase of natural nucleotides, which has led to the general assumption that this enzyme would also easily phosphorylate diphosphates of nucleoside analogs and in particular AZT-DP and ddADP. However, the cellular concentration of AZT-TP is even lower than that of AZT-DP, unlike ATP which is much more abundant than ADP (19). This suggested to us that AZT-DP may be a poor substrate for NDP kinase and that the reaction catalyzed may be a second limiting step in the phophorylation pathway.
In this paper we have investigated the ability of antiviral diphospho-and triphosphonucleotides to be used as substrates by human NDP kinase. The results are discussed in the context of the crystal structure of NDP kinase and in particular of the role played by the 3Ј-OH of the ribose moiety in substrate binding and in catalysis.

Purification of Recombinant Human NDPK-B-Human
NDP kinase-B was expressed in Escherichia coli as described (21) and purified according to (14) with the following modifications. Cells were resuspended in 50 ml of Tris-HCl buffer (pH 8.4) containing 5 mM MgCl 2 , 1 mM dithiothreitol, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 10 mM benzamidine (Buffer A). They were lysed in a French press, the lysate was spun at 20,000 rpm for 30 min and the supernatant was loaded on a DEAE-Sephacel column equilibrated with buffer A. Under these conditions, the endogenous E. coli NDP kinase bound to the resin and was separated from recombinant NDPK-B which was recovered in the flow-through fractions. The latter were loaded on a Blue-Sepharose column (5 ml) equilibrated in buffer A at pH 7.4. The column was washed with 2 M NaCl and NDPK-B was eluted with a linear gradient of NaCl (2 M to 5 M). The high salt concentration, which was necessary for elution from the column, was immediately lowered by dialysis of the fractions against 50 mM Tris-HCl buffer (pH7.4) containing 5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 0.1 mg/ml bovine serum albumin. NDPK-B was purified to homogeneity as judged by SDSpolyacrylamide gel electrophoresis. It was stored at Ϫ20°C in the same buffer containing 20% glycerol.
To synthesize phosphoderivatives of AZT, the free 5Ј-OH of AZT was phosphorylated by condensation with ␤-cyanoethyl dihydrogen phosphate (22) in the presence of DCC in anhydrous pyridine to give the phosphodiester, followed by treatment with 0.4 LiOH for 1 h. AZT-DP and AZT-TP were obtained one-pot from AZT-MP via the phosphoroimidazolate prepared from the phosphomonoester and 1,1Ј-carbonyldiimidazole (23). The di-and triphosphate were isolated by chromatography on a DEAE-Sephadex A-25 column (HCO 3 Ϫ form) eluted with a linear gradient of triethylamonium hydrogen carbonate buffer (pH 7-8; 0.05-0.5 M). ddADP was enzymatically synthesized from ddATP in presence of 3-fold excess fructose 6-phosphate and phosphofructokinase in 50 mM Tris-HCl (pH 8), 5 mM MgCl 2 for 3 h at 20°C. It was purified by reversed-phase chromatography on a C-18 column eluted with acetonitrile-water (0 -25%).
Kinetic Measurements-When the ability of NDP kinase to use the analog as a phosphate donor was studied, we measured the formation of [ 14 C]ATP from [ 14 C]ADP (0.1 mM), at various concentrations of nucleoside triphosphate. When the analog was tested as a phosphate acceptor, 1.0 mM [␥-32 P]GTP was used as a phosphate donor and the amount of [␥-32 P]NTP formed was measured. [␥-32 P]ATP was not used in this study because of high background. It should be noted that the GDP formed during the reaction competes with the analog diphosphate studied, leading to nonlinear kinetics. In order to avoid this difficulty, rephosphorylation of GDP was achieved by adding pyruvate kinase (0.05 mg/ml at 600 units/mg) and phosphoenolpyruvate (1 mM) along with 50 mM KCl in the assay mix. We checked that the analog nucleoside diphosphates were not substrates for pyruvate kinase.
The assays were started by adding 3 l of enzyme to a reaction mixture (10 l) containing 50 mM Tris-HCl (pH 7.4), 5 mM MgCl 2 , and the substrates at 37°C. The amount of NDP kinase added per assay varied from 10 pg for natural nucleotides to 5 ng with poor analogs. When the nucleotide triphosphates were assayed, the reaction was stopped by adding 3-l aliquots to a 2-l stop solution consisting of 0.7 M formic acid and 10 mM each of ADP and ATP. When assaying nucleotides diphosphates with [␥-32 P]GTP, the reaction was stopped by placing a 3-l aliquot of the reaction mixture at 85°C for 2 min. After cooling on ice, 2 l of a 10 mM solution of cold nucleotide was added. The nucleotides were separated on TLC plates with UV indicator (Macherey-Nagel, Germany) which were developed with 400 mM NH 4 HCO 3 or 1 M formic acid and 1.5 M LiCl when [␥-32 P]GTP or [ 14 C]ATP were used, respectively. The products formed were quantified with the WIN-IQ program (Molecular Dynamics) using a PhosphorImager screen. Linear readings of the radioactivity were obtained in a range covering 5 orders of magnitude in nucleotide concentration. Kinetic parameters were calculated by nonlinear fitting using Kaleidagraph software.  We have investigated the ability of NDP kinase to use the diphosphate and triphosphate forms of AZT, ddA and ddT, as phosphate acceptor and phosphate donor, respectively. Since the two human isozymes of NDP kinase, NDPK-A and NDPK-B, do not differ in their enzymatic properties (6), we have used only the isozyme NDPK-B encoded by the gene Nm23-H2 (9), to perform the experiments reported in this paper. Preliminary experiments using NDPK-A gave similar results (data not shown). Fig. 1 shows a typical kinetic experiment. The rate of product accumulation was constant for at least 6 min, allowing determination of initial velocities (Fig. 1, inset). It should be noted that the K m and V max values derived from these experiments are apparent kinetic parameters measured by varying the concentration of one substrate only. Due to competition between the nucleoside di-and triphosphates, inhibition by excess of substrate makes a more complete study difficult. However, for an enzyme with a ping-pong mechanism, the ratio of the apparent k cat /K m is equal to the true value of k cat /K m ; therefore, it is a useful parameter when comparing the natural substrates to the analogs. Tables I and II, AZT nucleotides are very poor substrates for the NDP kinase reaction. When used in the diphospho-form as an acceptor, the apparent k cat is 0.17% of that of TDP (Table I), while it is 0.05% of that of TTP when used in the triphospho-form as the phosphate donor (Table II). The ratio k cat /K m is high with natural nucleotides, actually close to the value predicted for diffusion-controlled reactions. It drops by several orders of magnitude for all analogs with a modified 3Ј-OH position on the ribose moiety. This is true, for instance, for analogs in which the 3Ј-OH is missing, such as 3Ј-dATP (which yields only 0.4% of the activity with ATP) or the dideoxy analogs (Tables I and II). Very low k cat are measured when ddTTP or ddATP is used as donor (0.01% and 0.04% of TTP and ATP, respectively), or when ddADP is used as the acceptor (0.4% of ADP). These results point to the importance of the 3Ј-OH group as opposed to the 2Ј-OH. It is interesting to note that similar results were obtained with 3Ј-dATP and AZT-TP, suggesting that steric hindrance by the bulky azido group in AZT nucleotides is not the reason for their poor performance as substrates of NDP kinase. In contrast, preliminary measurements showed that arabino-ATP (where the sugar moiety is the epimer of ribose in the 2Ј position) is a good substrate for NDP kinase (data not shown).

As shown in
We also performed experiments where the analogs were tested as competitors in the reaction of phosphorylation of [ 14 C]ADP by TTP. AZT-TP and ddTTP were both inhibitors (data not shown), with I 50 values approximately equal to their apparent K m (see Tables I and II), indicating that a lack of binding to the enzyme is not the reason of the poor activity described above. Under the conditions used, no transfer of ␥-phosphate from either analog to ADP could be detected.
The x-ray structures of several NDP kinases in complex with nucleotides explain the lack of specificity of the enzyme for the nucleo-base. Unlike most nucleotide-binding proteins, NDP kinase does not form specific hydrogen bonds with the base (Fig.  2). In contrast, there is extensive bonding to the 3Ј-OH of the sugar, which accepts hydrogen bonds from the Lys-16 and Asn-119 side chains (numbers correspond to the Dictyostelium NDP kinase sequence). The role of these amino acids has been confirmed by site-directed mutagenesis (18). Moreover, the 3Ј-OH donates a hydrogen bond to one of the ␤-phosphate oxygens (16,17). This internal bond maintains the nucleotide in a folded conformation, which is probably needed to position  Phosphorylation of Anti-HIV Nucleotides by NDP Kinase 7889 the ␥-phosphate correctly for in-line attack by the N␦ nitrogen of the catalytic histidine. Its presence also suggests that the 3Ј-OH plays a role in catalysis by donating its proton to the leaving group and helping release of the nucleoside diphosphate product. Our data on the study of nucleoside analogs support this suggestion. CONCLUSION We have shown that the di-and triphosphate forms of AZT, ddA and ddT, are poor substrates for NDP kinase and that the absence of a 3Ј-OH on the sugar is largely responsible for their lack of activity. These results are in agreement with previous studies showing some in vivo accumulation of the AZT-DP (19) and dideoxynucleotides in MT-4 cells (24). Although they suggest the possibility that these and other nucleoside analogs lacking a 3Ј OH group such the acyclic nucleosides, may not be phosphorylated by NDP kinase in vivo, it should be kept in mind that the turn over of NDP kinases is unusually high (more than 1000 s Ϫ1 ), and therefore that even poor substrates may be phosphorylated in the cell. Our results may help understanding the pharmacokinetics of nucleoside analogs. They may provide a rational basis for the drug design of new active molecules, with the hope that analogs more efficiently phosphorylated by NDP kinase can be used at a lower dose and elicit less toxic and secondary effects.