Phosphorylation of Pyrimidine l-Deoxynucleoside Analog Diphosphates

Anticancer and antivirald- and l-nucleoside analogs are phosphorylated stepwise in the cells to the pharmacologically active triphosphate metabolites. We recently reported that in the last step,l-deoxynucleoside analog diphosphates are phosphorylated by 3-phosphoglycerate kinase (PGK). To explain the preference of PGK forl- over d-deoxynucleoside analog diphosphates, the kinetics of their phosphorylation were compared with the dephosphorylation of the respective triphosphates using recombinant human PGK. The results attributed favorable phosphorylation ofl-deoxynucleoside analog diphosphates by PGK to differences in k cat, which were consequences of varied orientations of the sugar and diphosphates in the catalytic site of PGK. The amino acids involved in the catalytic reaction of PGK (including Glu344, Lys220, and Asn337) were therefore mutated. The impact of mutations on the phosphorylation of l- and d-deoxynucleoside analog diphosphates was different from those on dephosphorylation of the respective triphosphates. This suggested that the interactions of the nucleoside analogs with amino acids during the transition state are different in the phosphorylation and dephosphorylation reactions. Thus, reversible action of the enzyme may not involve the same configuration of the active site. Furthermore, the amino acid determinants of the action of PGK for l-deoxynucleotides were not the same as for the d-deoxynucleotides. This study also suggests the potential impact of nucleoside analog diphosphates and triphosphates on the multiple cellular functions of PGK, which may contribute to the action of the analogs.

Nucleoside analogs are an important class of anticancer and antiviral agents. Among these the L-deoxynucleoside analogs are emerging as a new class of compounds (1,2). Some examples are L-SddC 1 (lamivudine), which is in clinical use for the treatment of HIV and hepatitis B virus (3)(4)(5); L-FMAU (6) and 2Ј,3Ј-dideoxy-2Ј,3Ј-didehydro-␤-L-5-fluorodeoxycytidine (7)(8)(9), which are in phase II clinical trials as anti-hepatitis B virus agents; and ␤-L(Ϫ)-dioxolanecytidine (10 -12), which is in a phase II clinical trial as an anticancer agent. Among D-dideoxynucleoside analogs, 2Ј,3Ј-didehydro-2Ј,3Ј-dideoxythymidine and ddC are some examples of anti-HIV drugs (13,14), and D-deoxynucleoside analogs like ␤-D-arabinofuranosylcytosine and gemcitabine are anticancer drugs (15). The last step in the phosphorylation of L-deoxynucleoside analog diphosphates to the respective triphosphates, which are the pharmacologically active metabolites, remained largely unexplored. Through comparison of phosphorylation of several clinically relevant D-deoxynucleoside, D-dideoxynucleoside, and L-deoxynucleoside analog diphosphates by nucleoside-metabolizing enzymes purified from the human hepatic carcinoma cell line HepG2, we recently reported that L-deoxynucleoside analog diphosphates could be phosphorylated by 3-phosphoglycerate kinase (PGK); D-deoxynucleoside analog diphosphates were likely to be phosphorylated by nucleoside diphosphate kinase; and D-dideoxynucleoside analog diphosphates were excellent substrates for creatine kinase (16). Moreover, L-deoxynucleoside analog diphosphates were better substrates for PGK than the corresponding D-deoxynucleoside and D-dideoxynucleoside analog diphosphates (at 200 M). Other nucleoside-metabolizing enzymes have not exhibited a similar preference for Ldeoxynucleoside analog diphosphates, and the property is unique to PGK.
Human PGK (ϳ46 kDa) is a glycolytic enzyme that catalyzes the conversion of 1,3-biphosphoglycerate to 3-phosphoglycerate and during the process generates one molecule of ATP (17). The reaction is reversible. In addition to its participation in the glycolytic cycle, cytoplasmic PGK is known to stimulate viral mRNA synthesis (18). It is also expressed in the nuclei where it modulates DNA synthesis and repair (19 -21). Since PGK in nuclei retains its ability to bind to its natural substrates, it has been proposed that its activity in nuclei may be regulated by the energy state of the cell (21). Extracellular PGK was recently shown to have a thiol-reductase activity (22). Reduction of plasmin by PGK results in proteolysis of plasmin to angiostatin fragments, which are inhibitors of angiogenesis (23)(24)(25). ATP and 3-phosphoglycerate could inhibit reduction of plasmin, presumably through inducing a conformational change that was not favorable to the reduction of plasmin (25). This suggested that the level of nucleoside diphosphates or triphosphates (natural or otherwise) might have a regulatory impact on the multiple cellular functions of PGK.
The sequences of mammalian PGKs are conserved over 96%, and there is also a high level of tertiary structure homology (26). The crystal structures of horse muscle PGK in * This work was supported by Grant AI-38204 from NFAID, National Institutes of Health and Grant CA-63477 from NCI, National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: L-SddC, ␤-L-2Ј,3Ј-dideoxy-3Ј-thiacytidine; L-FMAU, 2Ј-fluoro-5-methyl-␤-L-arabinofuranosyluracil; L-ddC, ␤-L-2Ј,3Ј-dideoxycytidine; HIV, human immunodeficiency virus; PGK, 3-phosphoglycerate kinase; DP, diphosphate; TP, triphosphate; HPLC, high pressure liquid chromatography; AMP-PNP, adenosine 5Ј-(␤,␥-imino)triphosphate. binary complex with 3-phosphoglycerate (27) and pig muscle PGK in binary and ternary complexes with 3-phosphoglycerate and nonhydrolyzable ATP analog AMP-PNP have been resolved (28 -30). All of the amino acids in the active site are conserved among horse muscle PGK, pig muscle PGK, and human PGK. In this study the amino acids in the catalytic site of human PGK have been extrapolated from the model proposed by May et al. (29), which was based on the crystal structure of pig muscle PGK in ternary complex with AMP-PNP and 3-phosphoglycerate. The modified catalytic reaction is shown in Fig. 1.
PGK comprises of a single subunit with two domains; the N-terminal domain has a basic patch region (rich in arginines and histidines) for binding to 1,3-biphosphoglycerate (or 3-phosphoglycerate), and the C-terminal domain contains the nucleotide-binding site. The nucleobase binds to a hydrophobic groove on the surface of the C terminus of the enzyme. The sugar is stacked on the pyrollidine ring of Pro 339 , and both hydroxyl groups on the ribose hydrogen bond with the carboxylate of Glu 344 . The oxygen on the ␣-phosphate interacts with Lys 220 , and the oxygen on the ␤-phosphate (or the bridge oxygen between ␤and ␥-phosphates) hydrogen bonds to Asn 337 . Asn 337 also forms a hydrogen bond with the amino group of Lys 220 and helps to stabilize its ion pair interaction with ␣-phosphate of the nucleotide (28 -30). Fig. 1 depicts the catalytic reaction of PGK with ATP as the substrate and 3-phosphoglycerate as the phosphate acceptor. Enzyme-substrate(s) interactions induce a conformational change in PGK resulting in the reduction of distance between the N-and C-terminal domains for associative phosphate transfer. The ␥-phosphate is then in the proximity of Arg 39 , and additional interaction of bridge oxygen between ␤and ␥-phosphate with Asn 337 creates a positive charge on the phosphorus atom of ␥-phosphate, facilitating nucleophilic attack by 3-phosphoglycerate. This disturbs the coordination of the metal ion and its linkage with ␣-, ␤-, and ␥-phosphates and results in interaction of metal ion with Asp 375 and ␣and ␤-phosphates. This leads to the collapse of the transition state to form metal-ADP and 1,3-biphosphoglycerate complexes bound to PGK, which then reverses the conformational change and releases the products (27)(28)(29). As with any reversible enzyme it is generally assumed that the phosphorylation of ADP to ATP using 1,3-biphosphoglycerate as a phosphate donor would also go through a similar transitional state (ED* ϭ EP*).
To understand the preference of PGK for L-deoxynucleoside analog diphosphates over their D-deoxynucleoside analog counterparts, the kinetics of their phosphorylation were compared with the kinetics of dephosphosphorylation of the respective triphosphates using recombinant human PGK. The preference observed was further explained through kinetic analysis of phosphorylation and dephosphorylation of nucleoside analog diphosphates and triphosphates by single amino acid mutants of PGK. These included mutations of PGK at Glu 344 , Lys 220 , and Asn 337 , all of which are involved in hydrogen bonding with the sugar and phosphate side chain of nucleoside analog diphosphates or triphosphates during the catalytic reaction.

EXPERIMENTAL PROCEDURES
Nucleoside Analog Diphosphates and Triphosphates-Nucleoside analogs L-FMAUDP, D-FMAUDP, L-SddCDP, ␤-L(Ϫ)-dioxolanecytidine DP, L-ddCDP, ␤-D-2Ј,3Ј-dideoxycytidine DP, L-FMAUTP, D-FMAUTP, and L-SddCTP were synthesized according to the following procedure (which is a modification of the protocol published by Ruth and Cheng (31)). Nucleoside analogs were dissolved in trimethylphosphate (10 l/mg of nucleoside) for 10 min at Ϫ10°C. Phosphooxychloride (10 equivalents of trimethylphosphate) was added at the same temperature to generate nucleoside analog monophosphates. Ten parts of the mix-ture of 2 mmol of phosphoric acid (in N,N-dimethylformamide) and 6 mmol of tributylamine were added to the solution containing nucleoside analog monophosphates. This resulted in the stepwise synthesis of nucleoside analog diphosphates and triphosphates. The reactions were stopped after 1 h by neutralization of the mixture with NaOH to obtain the maximum yields of both diphosphates and triphosphates (with time the nucleoside analog diphosphates are also converted to triphosphates). Nucleoside analog diphosphates and triphosphates were purified using DEAE Sephadex A-25 (Amersham Biosciences) and eluted with step gradients between 0 and 300 mM KCl. The purity of the respective nucleoside analog diphosphates and triphosphates was confirmed by HPLC (Shimadzu, Braintree, MA) in a binary gradient of water and 0.3 M potassium phosphate buffer using an anion exchange column (Partisil-SAX, Whatman, Inc., Clifton, NJ). All of the other nucleoside analog diphosphates and triphosphates included in this study were purchased from Amersham Biosciences.
Cloning PGK and Single Amino Acid Mutants of PGK-Total RNA was extracted from HepG2 cells (human hepatoma) using TRIzol and reverse transcribed using Superscript II RNase H reverse transcriptase (Invitrogen) according to the manufacturer's instructions. An aliquot of cDNA was amplified using DNA polymerase and specific primers for human PGK, including 5Ј-GGA ATT CCA TAT GTC GCT TTC TAA CAA GCT GAC G, and 5Ј-CGC GGA TCC CTA AAT ATT GCT GAG AGC ATC CAC. The PCR product was digested with NdeI and BamHI restriction enzymes, ligated with NdeI-BamHI-digested pET28a bacterial expression vector (the sequence of PGK was confirmed by DNA sequencing), and transfected into Escherichia coli strain BL21 (DE3). The resulting PGK was an N-terminal histidine fusion protein, which was purified using nickel-nitrilotriacetic acid-agarose (Qiagen) column according to the manufacturer's protocol. Single amino acid mutations of the wild-type PGK-pET28a plasmid were carried out using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The specific primers (and their complimentary oligomers) used were 40 -50-mer, and the positions for the beginning and end of primers and the base pairs around the site of mutation (shown in bold type) are listed: (i) E344A, 5Ј-1006 GTA TTT GCG TGG GAA 1049 ; (ii) K220A, 5Ј-635 GCA GAC GCG ATC CAG 681 ; (iii) K220R, 5Ј-635 GCA CAG CGT ATC CAG 681 ; (iv) N337D, 5Ј-987 GTG TGG GAC GGT CCT 1035 ; and (v) N337Q, 5Ј-987 GTG TGG CAG GGT CCT 1035 . The sequences were confirmed by DNA sequencing. The mutant enzymes were purified using the same method as for the wild-type enzyme.
Phosphorylation and Dephosphorylation of Nucleoside Analogs-Phosphorylation of nucleoside analog diphosphates by PGK and the mutant enzymes was evaluated in a buffer containing 50 mM Tris acetate (pH 7.5), 5 mM MgCl 2 , 1 mM NaF, 1 mM dithiothreitol, 10 mM sodium phosphate, 4 mM NAD ϩ , and 4 mM DL-glyceraldehyde-3-phosphate. 1,3-Biphosphoglycerate, the phosphate donor for the reaction, was generated 10 min prior to the inclusion of PGK or mutant enzymes, by the addition of 0.5 units/0.1 ml of glyceraldehyde-3phosphate dehydrogenase (16). Dephosphorylation of nucleoside analog triphosphate by PGK or the mutant enzymes was evaluated in a buffer containing 50 mM Tris acetate (pH 7.5), 5 mM MgCl 2 , 1 mM NaF, 1 mM dithiothreitol, and 4 mM 3-phosphoglycerate. All of the samples were incubated at 37°C. The reactions were stopped on ice, and the samples were deproteinized by trichloroacetic acid precipitation. The samples were then extracted in a mixture of trioctylamine and 1,1,3-trichlorotrifluoroethane in a ratio of 45:55. The diphosphate and triphosphate forms of nucleoside analogs were analyzed by HPLC as described above. The K m and k cat values were calculated from Lineweaver-Burk double reciprocal plots. All of the values shown in Tables I-VI are the means and standard deviations from at least three independent experiments.

Phosphorylation of Nucleoside Analog Diphosphates by PGK-
The K m and k cat values of nucleoside analog diphosphates for recombinant human PGK are shown in Table I. This included diphosphates of L-deoxynucleoside analogs and D-ribonucleoside, D-deoxynucleoside, and D-dideoxynucleoside analogs. As shown in Table I, the nucleobase did not have a major impact on binding; the K m values of all of the purine and pyrimidine analog diphosphates were within a 3-fold range. The k cat for purines, represented by ADP, which is the natural substrate of the enzyme, were higher than those of pyrimidines, as represented by TDP and CDP. In addition, within the same base, the k cat for D-ribonucleoside was greater than D-deoxynucleoside analog diphosphates. Phosphorylation of dCDP and ddCDP was below the experimental levels of detection. Comparison of D-FMAUDP with L-FMAUDP and comparison of ddCDP with L-ddCDP and L-SddCDP showed that favorable phosphorylation of L-deoxynucleoside analog diphosphates could be attributed to their higher k cat values. This implied that both the base and the sugar moieties of nucleosides had an impact on the transitional state of enzyme in binding nucleoside analog diphosphates, thereby affecting the efficiency of phosphate transfer.
Dephosphorylation of Nucleoside Analog Triphosphates by PGK-The effects of ␥-phosphate of nucleoside analog triphosphates on K m and the rate of dephosphorylation were evaluated, and the results are shown in Table II. The nucleobase had no significant impact on the K m value, however, the K m values for nucleoside analog triphosphates were 5-8-fold lower than those of the respective diphosphates. The k cat values for dephosphorylation of D-ribonucleoside and L-deoxynucleoside analog triphosphates examined were 100 -1000-fold lower than that for the phosphorylation of the respective diphosphates. The k cat values for all of the nucleoside analog triphosphates were, however, similar, indicating that the base configuration of sugar and the presence or absence of 2Ј-and 3Ј-hydroxyl groups had no impact on the rate of phosphate transfer. Interestingly, the k cat value for dephosphorylation of dCTP and ddCTP was better than the k cat value for phosphorylation of the respective diphosphates. The ratios of the efficiency of dephosphorylation of nucleoside analog triphosphates and the efficiency of phosphorylation of the respective diphosphates (derived from Table I) are also shown in Table II. Phosphorylation of the diphosphates of ribonucleosides and L-deoxynucleoside analogs were favored at least a 100-fold over the dephosphorylation of the respective triphosphates. This implied that phosphorylation of Ldeoxynucleoside analog diphosphates would not be limited by dephosphorylation of the triphosphates generated during the reaction. Phosphorylation of D-FMAUDP and TDP were favored only by 2-10-fold over the dephosphorylation of the respective triphosphates. These results supported the observation that favorable phosphorylation of L-deoxynucleoside analog diphosphates as compared with the corresponding D-nucleoside analogs were due to differences in the k cat value.
Phosphorylation of Nucleoside Analog Diphosphates by E344A-The 2Ј,3Ј-hydroxyl groups, which hydrogen bond to Glu 344 , had an impact on the phosphorylation of nucleoside analog diphosphates (Table I). To evaluate the role of Glu 344 in the orientation of nucleoside analog diphosphates in the catalytic site, glutamic acid was replaced with alanine. The K m and k cat values for E344A are shown in Table III. The values in parentheses show the changes in K m and k cat in comparison with the wild-type PGK. As expected, the k cat value for Dribonucleoside analogs ADP (over 10-fold) and CDP (at least 1000-fold) decreased significantly. The k cat value for the Ddeoxynucleoside analogs TDP and D-FMAUDP increased slightly, similar to that of L-dideoxynucleoside analog, L-Sd-dCDP, whereas L-FMAUDP remained unaffected. There were no discernible patterns in the changes associated with the K m values. These results indicated that the hydrogen bonds between ribonucleoside diphosphates and the carboxyl group of Glu 344 are important for the optimal orientation of the substrate in the catalytic site for efficient phosphate transfer. In the analogs lacking either or both 2Ј-and 3Ј-hydroxyl groups, substitution with alanine at position 344 had no significant effect on the rate of phosphorylation.
Phosphorylation of Nucleoside Analog Diphosphates by K220A, K220R, N337D, and N337Q-Lys 220 and Asn 337 stabilize the phosphate side chain by forming hydrogen bonds with oxygen on ␣-phosphate and ␤-phosphate, respectively. These amino acids are therefore directly involved in the interactions with diphosphates or triphosphates of nucleotides in transition state. Lys 220 was replaced with alanine (to abolish hydrogen bonding) and with arginine (to evaluate the steric impact), and Asn 337 was replaced with aspartate (to abolish hydrogen-bonding) and glutamine (to study the steric impact). The K m and k cat values are shown in Table IV, and the values in parentheses represent the changes with respect to the wild-type PGK. The rate of phosphorylation of diphosphates of cytidine analogs by both Lys 220 and Asn 337 mutants decreased to less than 0.04 min Ϫ1 , which is the limit for detection by HPLC. A similar decrease was observed with TDP and D-FMAUDP for all enzymes except K220R.
Mutation to K220A increased the K m of ADP over 5-fold; it, however, did not affect the K m of dADP and L-FMAUDP. The k cat value for all three analogs decreased significantly (ranging from 100-to 3000-fold). Replacement of Lys 220 with arginine, which can form hydrogen bonds with oxygen on ␣-phosphate, resulted in a slight recovery of activity, without affecting the K m . The K m value of ADP for K220R also improved over 2-fold as compared with K220A. Replacement of Asn 337 with aspartate resulted in a 3-fold decrease in K m of ADP and L-FMAUDP, whereas decreases in k cat ranged from 100-to 750-fold as compared with the wild-type enzyme. Substitution of Asn 337 with glutamine resulted in a slight recovery of activity as compared with N337D. Since the rates of phosphorylation of most of the pyrimidine analogs could not be determined, changes in K m and k cat values observed with ADP, dADP, and L-FMAUDP could not be categorized. These results, however, indicate that hydrogen bonds stabilized by Lys 220 and Asn 337 are essential for the phosphorylation of D-and L-nucleoside analog diphosphates. Replacement of Lys 220 and Asn 337 with arginine and glutamine, respectively, decreased the k cat value significantly, indicating that these amino acids are not sterically accommodated in the catalytic cleft. Varied impacts of mutation on nucleoside analog diphosphates could be attributed to different conformations of the analogs in the catalytic site.
Dephosphorylation of Nucleoside Analog Triphosphates by K220A, K220R, N337D, and N337Q-As shown in Tables V and VI, mutation of Lys 220 and Asn 337 did not affect the binding of nucleoside analog triphosphates, with the exception of ddCTP. Interestingly, the k cat values for all of the D-nucleoside analog triphosphates (except ddCTP) increased over 5-10-fold with K220A, K220R, and N337D mutation and by 2-4-fold with N337Q mutation. The K m value of ddCTP improved with mutation of Lys 220 , and the k cat value improved with mutation of Asn 337 . This indicated that interactions of D-ribonucleoside and D-deoxynucleoside analog triphosphates with amino acids Lys 220 and Asn 337 are different from those of D-dideoxy nucleoside analog triphosphates. These results showed that in contrast to the forward reactions, mutation of Lys 220 and Asn 337 had a stimulatory effect on the rate of dephosphorylation of the triphosphates, indicating that the interactions of D-nucleoside analogs with these amino acids are different in the forward and reverse reactions. Mutation to K220A, N337D, or N337Q, however, had a detrimental effect on the k cat value for L-SddCTP, whereas dephosphorylation of L-FMAUTP remained unaffected. This showed that for L-deoxynucleoside analogs, hydrogen bond formation with Lys 220 and Asn 337 is essential, and the orientation of L-deoxynucleoside analogs in the catalytic cleft was probably different from those of D-nucleoside analogs. However, the impact of mutating Lys 220 and Asn 337 on the phosphorylation of L-deoxynucleoside analog diphosphates was different from those on the dephosphorylation reactions.
DISCUSSION D-Deoxynucleoside analogs, which are in the natural configuration, as well as the L-deoxynucleoside analogs, are an important class of antiviral and anticancer agents. These analogs are phosphorylated stepwise to the respective triphosphate metabolites in the cells. The last step in the phosphorylation of L-deoxynucleoside analog diphosphates is inefficient, and in the cells, most of these analogs accumulate as diphosphate metabolites (7,(32)(33)(34)(35). We recently showed that in contrast to the general assumptions that nucleoside diphosphate kinases were the enzymes responsible for phosphorylation of all nucleoside analog diphosphates (36 -38), L-deoxynucleoside analog diphosphates were phosphorylated by PGK.
Since L-deoxynucleoside analog diphosphates were better substrates for PGK than the corresponding D-nucleoside ana-logs, we tried to explain the differences in their phosphorylation based on the kinetics of phosphorylation and dephosphorylation of these analogs by PGK. Such information is also important for predicting the effects of clinically relevant nucleoside analogs on PGK and therefore on glycolysis and possibly angiogenesis and DNA repair processes. Nucleoside analogs were categorized into D-ribonucleoside, deoxynucleoside, dideoxynucleoside, and L-deoxynucleoside analogs. These included: ADP, the natural substrate of PGK, and its deoxy-form dADP; thymidine analogs, TDP, D-FMAUDP, and L-FMAUDP; and cytidine analogs CDP, dCDP, ddCDP, and L-SddCDP. The reverse reactions used the triphosphates of all of these analogs. Single amino acid mutations of PGK in its catalytic site were based on the mechanism shown in Fig. 1. The thiol-reductase activities of PGK and its mutants were monitored to assess the global structure of the wild-type and mutant enzymes. All of the enzymes retained their thiol-reductase activity, which is necessary for the generation of angiostatin fragments from plasmin, and were not susceptible to proteolysis by plasmin (data not shown).
Kinetics of phosphorylation of nucleoside analog diphosphates by recombinant human PGK showed that the nucleobase did not have a major impact on the K m of purine and pyrimidine analog diphosphates; however, it was apparent that the k cat value varied significantly (with purines faring better than the pyrimidines). The hydroxyl groups on the sugar moiety also had an impact on phosphorylation, with the k cat decreasing in the following order: ribonucleoside, deoxynucleoside, and dideoxynucleoside analog diphosphates. L-Deoxynucleoside analogs, despite the absence of 2Јand/or 3Ј-hydroxyl groups were better phosphorylated than the corresponding D-nucleoside analogs. Since the catalytic action of PGK is reversible, it was possible that the reverse reaction in the dephosphorylation of nucleoside analog triphosphates could be a rate-limiting factor. The kinetics of dephosphorylation of D-and L-nucleoside analog triphosphates were evaluated. Although the structure itself had no impact on the K m value, the presence of ␥-phosphate in nucleoside analog triphosphates resulted in a 5-8-fold decrease in K m as compared with the respective diphosphates. In contrast to the forward reactions, the configuration of the sugar and the presence or absence of hydroxyl groups on the sugar moiety did not impact the k cat . For ribonucleoside and Ldeoxynucleoside analogs, the efficiency of the forward reaction in the phosphorylation of diphosphates was favored at least 100-fold over dephosphorylation of the respective triphosphates. Interestingly, for 2Ј-deoxycytidine and ␤-2Ј,3Јdideoxycytidine, the k cat value for dephosphorylation of their triphosphates was better than the V max value for phosphorylation of the respective diphosphates. This accounts for the   favorable phosphorylation of L-deoxynucleoside analog diphosphates (as compared with the corresponding Dnucleoside analogs) by PGK. Although L-deoxynucleoside analog diphosphates are efficiently phosphorylated by PGK, accumulation of the triphosphate form may limit its own synthesis.
Favorable phosphorylation of L-deoxynucleoside analog diphosphates over the D-nucleoside counterparts was solely attributed to differences in k cat . This implied that both the base and the sugar induced changes in the orientation of the diphosphates in the catalytic site, thereby affecting the rate of phosphate transfer. To study varied configurations of L-and D-nucleoside analog diphosphates in the catalytic site, amino acids Glu 344 , Lys 220 , and Asn 337 were mutated.
The 2Ј-and 3Ј-hydroxyl groups on the sugar moiety form hydrogen bonds with the carboxylate of Glu 344 . To assess the role of such hydrogen bonding on the k cat values for nucleoside analog diphosphates, the amino acid was substituted with Ala. As expected, E344A had a detrimental effect on the k cat values for ribonucleoside analogs, ADP and CDP, without significantly affecting the K m values or the rates of phosphorylation of the other nucleoside analog diphosphates. This implied that the varied conformations of the deoxynucleoside, dideoxynucleoside, and L-deoxynucleoside analog diphosphates in  a k cat lower than 0.04 min Ϫ1 , which is the limit for detection by HPLC.

TABLE VI Dephosphorylation of nucleoside analog triphosphates by N337D and N337Q
The K m and k cat values of the nucleoside analog triphosphates for the mutant enzymes were evaluated using 3-phosphoglycerate as a phosphate acceptor. The values in parentheses indicate the k cat of mutant enzymes as compared with the k cat for dephosphorylation of nucleoside analog triphosphates by the wild-type PGK.
a k cat lower than 0.04 min Ϫ1 , which is the limit for detection by HPLC.
FIG . 1. Catalytic activity of human PGK. The figure is a schematic representation of the catalytic activity of human PGK, a modified version of the model published by May et al. (29) for pig muscle PGK. ED represents a complex between NDP and PGK; EP represents a complex between NTP and PGK; the reaction goes through a transition complex: ED* or EP*. It is generally assumed that ED* ϭ EP*. Only the amino acids interacting with the sugar and phosphates of nucleotides are shown in this figure. the catalytic site were independent of their interactions with Glu 344 . Mutation of Glu 344 did not affect the reversible reaction in the dephosphorylation of nucleoside analog triphosphates (data not shown). This supported the observation that hydroxyl groups on the sugar moiety had no impact on the dephosphorylation of nucleoside analog triphosphates.
Lys 220 and Asn 337 interact with the phosphate side chain of nucleotides during the catalytic reaction. To study the importance of hydrogen bonding, Lys 220 and Asn 337 were substituted with Ala and Asp, respectively. The steric impact of increasing the length of amino acid side chain by K220R and N337Q mutations was also evaluated. The results showed that hydrogen bonding with Lys 220 and Asn 337 were essential for phosphorylation of all of the nucleoside analog diphosphates. Phosphorylation by K220R and N337Q was slightly better, although the increased side chain length conferred by Arg and Gln seemed to cause a steric hindrance. The varied impact of mutations on nucleoside analog diphosphates as substrates indicated that the orientation of the diphosphates in the catalytic site were probably different.
The effects of mutation of Asn 337 and Lys 220 on dephosphorylation of nucleoside analog triphosphates were also evaluated. In contrast to the forward reactions, the k cat value for dephosphorylation of all D-nucleoside analog triphosphates increased upon mutation of Lys 220 and Asn 337 without affecting the binding. ddCTP was an exception; Lys 220 mutation improved its K m value, whereas Asn 337 mutation improved the k cat value, which implied that the interactions of D-ribonucleoside and D-deoxynucleoside analog triphosphates with Lys 220 and Asn 337 mutants were different from those of D-dideoxynucleoside analog triphosphates. Increases in the k cat value upon K220R and N337Q mutation suggested that the increase in the length of the side chain allowed for a more favorable hydrogen bond formation between Arg or Gln and nucleoside analog triphosphates. Since both Ala and Asp are not capable of hydrogen bonding with the oxygen of ␣-phosphate and the bridge oxygen between ␤and ␥-phosphate, respectively, it is proposed that these oxygen molecules interact with Ala and Asp through hydrogen bonds via ordered water molecules. For the L-deoxynucleoside analog L-SddCTP, mutation to K220A, N337D, and N337Q had a detrimental effect on its dephosphorylation, whereas mutation to K220R caused no change. Dephosphorylation of L-FMAUTP by the mutants remained unaffected. The unique kinetics of dephosphorylation suggested that the orientation of L-deoxynucleoside analog triphosphates were different from those of the respective diphosphates in the catalytic site. In addition, interactions of L-deoxynucleoside analog triphosphates with amino acids in the catalytic cleft are different from those of D-nucleoside analog triphosphates. The impacts of mutations of Lys 220 and Asn 337 on nucleoside analog diphosphates and triphosphates are different. It is therefore concluded that the interactions between amino acids and the nucleoside analogs in transition states during the forward and reverse reactions are different; reversible action of the enzyme may not involve the same configuration of the active site. This is also supported by the observation that unlike the reactions involving phosphorylation of nucleoside analog diphosphates, the dephosphosphorylation of the respective triphosphates were independent of the structure of nucleobase, configuration of the sugar, or the presence or absence of hydroxyl groups on the sugar moiety. These results are better explained in the reaction scheme in Fig. 1. Contrary to the general assumptions that ED* ϭ EP*, this study showed that amino acid interactions during ED* are different from those in EP*. The conformational change induced by nucleoside diphosphate and its triphosphate are different despite both being substrates for PGK. The general assumption that the transitional state of enzyme is likely to be the same for the forward and reverse reactions of a reversible enzymatic process is not true for all enzymes.
In conclusion, favorable phosphorylation of pyrimidine Ldeoxynucleoside analog diphosphates as compared with the corresponding D-deoxynucleoside analogs by PGK is attributed to differences in the k cat value. These differences are consequences of different orientations of sugar and diphosphate of Land D-nucleoside analogs during the transitional state of PGK. The varied interactions of nucleoside analog diphosphates and triphosphates with amino acids in the catalytic cleft indicated that the configuration of the transition state for the forward and reverse reactions are different, which is a property unique to PGK. The intracellular role of PGK in the phosphorylation of pyrimidine L-deoxynucleoside analogs is currently under investigation. Although anticancer and antiviral D-deoxy and Ddideoxy nucleoside analogs like ␤-D-arabinofuranosylcytosine, gemcitabine, Azt, ddC, 2Ј,3Ј-didehydro-2Ј,3Ј-dideoxythymidine, etc., are unlikely to be phosphorylated by PGK, their accumulation in the cells as triphosphate metabolites may have an inhibitory effect on the enzyme. The impact of L-and D-deoxynucleoside analogs and their phosphorylated metabolites in the regulation of the multiple cellular functions of PGK will be addressed in the future.