Enantioselectivity of human AMP, dTMP and UMP-CMP kinases

Abstract

l-Nucleoside analogues such as lamivudine are active for treating viral infections. Like d-nucleosides, the biological activity of the l-enantiomers requires their stepwise phosphorylation by cellular or viral kinases to give the triphosphate. The enantioselectivity of NMP kinases has not been thoroughly studied, unlike that of deoxyribonucleoside kinases. We have therefore investigated the capacity of l-enantiomers of some natural (d)NMP to act as substrates for the recombinant forms of human uridylate-cytidylate kinase, thymidylate kinase and adenylate kinases 1 and 2. Both cytosolic and mitochondrial adenylate kinases were strictly enantioselective, as they phosphorylated only d-(d)AMP. l-dTMP was a substrate for thymidylate kinase, but with an efficiency 150-fold less than d-dTMP. Both l-dUMP and l-(d)CMP were phosphorylated by UMP-CMP kinase although much less efficiently than their natural counterparts. The stereopreference was conserved with the 2′-azido derivatives of dUMP and dUMP while, unexpectedly, the 2′-azido-d-dCMP was a 4-fold better substrate for UMP-CMP kinase than was CMP. Docking simulations showed that the small differences in the binding of d-(d)NMP to their respective kinases could account for the differences in interactions of the l-isomers with the enzymes. This in vitro information was then used to develop the in vivo activation pathway for l-dT.

INTRODUCTION

Recent developments indicate that l-nucleosides and l-nucleoside analogues are potent anti-viral, anti-tumour and even anti-malarial agents (1–4). The l-derivative, lamivudine (β-l-2′,3′-dideoxy-3′-thiacytidine, 3TC), has been approved for treating both HIV and HBV, while emtricitabine (β-l-2′,3′-dideoxy-5-fluoro-3′-thiacytidine, FTC) is used in HIV therapy. Telbivudine (β-l-thymidine, l-dT) was recently approved for treating hepatitis B by the US Food and Drug Administration. Many l-nucleosides are currently in advanced clinical trials for the treatment of a variety of virus diseases. These include clevudine [1-(2-fluoro-5-methyl-β-l-arabinofuranosyl)uracil, l-FMAU], elvucitabine (2′,3′-didehydro-2′,3′-dideoxy-β-l-5-fluorocytidine, β-l-d4FC), valtorcitabine (val-β-l-2′-deoxycytidine, val-l-dC), pentacept (2′,3′-didehydro-2′,3′-dideoxy-3′-fluoro-β-l-cytidine, β-l-3′-Fd4C) and β-l-2′-Fd4C (2′,3′-didehydro-2′,3′-dideoxy-2′-fluoro-β-l-cytidine). Most l-enantiomers considered for treating virus diseases have similar activities to their d-counterparts, but are less sensitive to degrading enzymes and have better safety profiles (5,6). These properties are due mainly to the enantioselectivity of the enzymes that interact with these substrates in vivo (7). l-nucleosides must be phosphorylated by cellular or viral kinases before they can reach the targeted virus enzymes. The enantioselectivity of the enzymes involved in nucleoside de novo synthesis and salvage pathway is governed by no general rule. Each individual metabolic enzyme must therefore be studied. Among the four deoxyribonucleoside kinases in human cells, cytosolic thymidine kinase 1 (TK1) is strictly enantioselective while mitochondrial thymidine kinase 2 (TK2) is less specific (8–10). Deoxycytidine kinase (dCK) and deoxyguanosine kinase (dGK) are poorly enantioselective in vitro (9,11–12). The phosphorylation of nucleoside monophosphates and their analogues to their diphosphate derivatives is then carried out by NMP kinases in both de novo and salvage pathways. The NMP kinases in human cells include one dTMP kinase, one UMP-CMP kinase, six isoenzymes of adenylate kinase and several guanylate kinases (13). Only the abilities of hUMP-CMP and dTMP kinases to phosphorylate some l-deoxynucleoside monophosphate analogues has been studied to date (14–15). Finally, NDPK is strictly enantioselective (16). The final step can be carried out by phosphoglycerate kinase, which has a broad substrate specificity (17–18). Other ATP-synthesizing enzymes such as creatine kinase can also be involved in this step (19). Triphosphorylated l-derivatives can interact with viral polymerases, acting as competitive inhibitors or alternate substrates, usually leading to chain termination. Viral polymerases incorporate l-derivatives more readily than do the human ones (20–21). Nucleoside degrading enzymes, such as nucleoside deaminases and phosphorylases, tend to be strictly selective for d-nucleotides although only limited data is available (11,22). The lack of degradation of l-nucleosides has however been demonstrated in mice, where l-nucleosides were gradually excreted in the unchanged form following intraperitoneal administration (5). 5′-Nucleotidases also seem to be moderately to highly enantioselective (7,23). The ribonucleotide reductases, which catalyze the de novo conversion of ribonucleotides to 2′-deoxyribonucleotides, are also key enzymes in DNA replication and repair. In vitro, these enzymes are inhibited by 2′-azido-2′-deoxynucleoside 5′-diphosphates (24–27). However, N₃-d-dUrd is not cytotoxic due to its poor intracellular phosphorylation. The mechanism of enzyme inactivation and the enzymatic monophosphorylation of the parent nucleosides has been thoroughly studied (24–25). More recently, we have shown that ribonucleotide reductase is enantiospecific with respect to the natural configuration of the sugar moiety (26–27). This report examines the phosphorylation of the β-D and β-L stereoisomers of 2′-azido pyrimidine monophosphates by human UCK, TMPK and AKs and characterizes the enantioselectivity and the possible cross-activities of these three human kinases for natural (d)NMP (Figure 1).

Figure 1.

Structures of β-L and β-D (deoxy)ribonucleoside monophosphates (Base = adenine, thymine, uracil, cytosine).

Part of this work was presented during the XVIIth Round Table for Nucleosides, Nucleotides and Nucleic Acids in Bern, in September 2006.

MATERIALS AND METHODS

Natural d-nucleotides were purchased from Sigma chemicals (St Louis, MO, USA). MABA-dTDP was synthesized as described (28).

Synthesis of L-nucleoside 5′-monophosphates

l-dA, l-dU, l-dC and l-dT were a generous gift from Idenix Pharmaceuticals (http://www.idenix.com). l-ribonucleosides were synthesized starting from l-ribose following usual procedures. The azidonucleosides were synthesized as previously described (26). l-Nucleoside 5′-monophosphates (l-NMP) were prepared by selective 5′-phosphorylation of the corresponding l-nucleosides with POCl₃ in triethylphosphate (29). l-NMPs were purified on DEAE-Sephadex A-25 (elution: linear gradient of TEAB pH 7.6 from 10 to 300 mM) followed by RP18 chromatography (elution: water to methanol 50%). The triethylammonium counter ions were exchanged for sodium by passing the nucleotide solution through a DOWEX-AG 50WX2-400 column. Yields were 43–80%. The structures and purities of l-NMP were assessed by nuclear magnetic resonance (¹H, ¹³C, ³¹P), fast-atom-bombardment MS, UV spectroscopy, HPLC and polarimetry. In some cases, it was necessary to separate the nucleoside 5′-monophosphate from traces of nucleoside 3′-monophosphate. This was done by semi-preparative HPLC using Hypercarb®, a porous graphite carbon stationary phase (30).

The physico-chemical properties of the 2′-azido-2′-deoxy d-nucleotides were identical, except for the value, to those of the corresponding 2′-azido-2′-deoxy l-nucleotides.

N₃-l-dUMP: (c 1.1, MeOH); ¹H NMR (D₂O, 300 MHz) δ 7.96 (δ, 1H, J = 8.1 Hz, H–6), 5.93 (d, 1H, J = 5.3 Hz, H–1′), 5.87 (d, 1H, J = 8.1 Hz, H5), 4.48 (pt, 1H, J = 5.1 Hz, H3′), 4.30 (pt, 1H, J = 5.4 Hz, H2′), 4.16 (sl, 1H, H4′), 4.04–3.90 (m, 2H, H5′, H5″); ¹³C NMR (D₂O, 300 MHz) δ 166.2, 151.6, 141.6, 102.7, 86.9, 84.4, 70.3, 65.4, 63.2; ³¹P NMR (D₂O, 300 MHz) δ + 2.47; UV (H₂O) λ_max 260 nm (ɛ 9400); MS FAB + m/z 372 (M + H)⁺, 350 (M – Na + 2H)⁺. N₃-d-dUMP: − 29 (c 1.1, MeOH). N₃-l-dCMP: (c 1, H₂O); ¹H NMR (D₂O, 300 MHz) δ 8.07 (d, 1H, J = 7.6 Hz, H–6), 6.14 (d, 1H, J = 7.6 Hz, H5), 6.06 (d, 1H, J = 4.5 Hz, H–1′), 4.54 (pt, 1H, J = 5.4 Hz, H3′), 4.34 (pt, 1H, J = 5.1 Hz, H2′), 4.26–4.22 (m, 1H, H4′), 4.16–4.04 (m, 2H, H5′, H5″); ¹³C NMR (D₂O, 300 MHz) δ 166.1, 157.3, 141.3, 96.6, 87.7, 83.5, 69.8, 66.0, 62.9; ³¹P NMR (D₂O, 300 MHz) δ + 2.15; UV (H₂O) λ_max 269 nm (ɛ 10 500); MS FAB + m/z 371 (M + H)⁺, 349 (M – Na + 2H)⁺. N₃-d-dCMP: (c 1.1, H₂O).

Bacterial production and purification of His-tagged NMP kinases

Human UCK, TMPK, AK1 and AK2 were produced in Escherichia coli as recombinant proteins as previously reported (15,31, Topalis, personal communication). E. coli Rosetta (DE3)pLysS cells transformed with the appropriate expression plasmid were grown at 37°C in LB medium supplemented with 34 μg·ml⁻¹ chloramphenicol and 50 μg·ml⁻¹ kanamycin. Gene expression was induced by adding 0.5 mM IPTG when the absorbance at 600 nm reached 0.8, and cells were grown at 30°C for a further 3 h. The cells were harvested, lysed by sonication and centrifuged at 5000 r.p.m., 8°C, for 30 min. The supernatant was loaded onto a Ni-NTA column (Qiagen, Germany) equilibrated with lysis buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM imidazole, pH = 8.0). The column was washed with lysis buffer and the proteins eluted with a linear gradient of imidazole, (0–250 mM). The fractions containing the enzyme were pooled and dialysed against dialysis buffer (50 mM Tris-HCl pH = 7.4, 20 mM NaCl, 1 mM DTT, 50% glycerol). SDS-PAGE indicated that the protein was over 95% pure.

Enzymatic assays

The activities of the NMP kinases were followed by a coupled spectrophotometric assay (32). Assays were carried out at 37°C in the following reaction mixture (total volume: 140 µl): 50 mM Tris-HCl, pH = 7.4, 50 mM KCl, 5 mM MgCl₂, 1 mM ATP, 0.2 mM NADH, 1 mM phosphoenolpyruvate, 1 mM DTT, 4 U·μl⁻¹ pyruvate kinase, 4 U·μl⁻¹ lactate dehydrogenase. As the NDP kinase from Dictyostelium discoideum is strictly enantioselective, this enzyme was only added (4 U·μl⁻¹) for assays with a d-enantiomer substrate (32). The reaction was started by adding the (d)NMP or (d)NMP analogue and the decrease in absorbance at 340 nm was measured. The kinase concentrations were 4 nM to 8 µM in order to measure initial rates below 0.2 ΔA/min. For inhibition studies, the inhibitor was added after the enzyme, the reaction was then started by adding the substrate(s). The results were analysed using the KALEIDAGRAPH software. Assays were carried out in duplicate or triplicate.

Only 0.2 mM ATP was added in the reaction mixture for studies on the inhibition of hAK1 and hAK2. Inhibition was studied at three inhibitor concentrations: 0, 400 and 600 µM l-AMP for inhibiting the phosphorylation of d-AMP by hAK1, and 0, 38 and 134 µM l-AMP for inhibiting the phosphorylation of d-dAMP by hAK2. The inhibitor was added immediately after the enzyme, and adding the substrate started the reaction. Data were plotted on double reciprocal plots and the K_I values were determined by plotting the slopes against the concentrations of inhibitor.

Fluorescence assays

The dissociation constants of hTMPK for both dTMP enantiomers were measured using a fluorescence assay (28). Briefly, the binding of the fluorescent probe MABA-dTDP to hTMPK in buffer T (50 mM Tris-HCl, pH = 7.5, 5 mM MgCl₂, 50 mM KCl and 5% glycerol) resulted in a 190% increase in fluorescence intensity (excitation at 340 nm and emission at 430 nm with emission and excitation slits at 2 and 4 nm, respectively). The binding constant was determined by titration of the fluorophore with the enzyme (K_D = 6 µM). Competition titrations with d- or l-d(T/U)MP were done by mixing MABA-dTDP (6 µM) with the enzyme (6 µM), so that, as recommended, about half of the fluorophore was enzyme-bound at the start of the experiment (33). Upon addition of increasing amounts of ligand, the fluorescence probe was gradually displaced from the enzyme active site and the resulting decreases in fluorescence were monitored. The total specific signal was determined at the end of each experiment by adding excess dTDP. The data were corrected for dilution and plotted: IC₅₀ values were obtained at half-displacement. The IC₅₀ values are related to the dissociation constants for the ligand, K_D, and for the fluorophore MABA-dTDP, , by the following equation (33):

(1)

where B is the initial concentration of bound MABA-TDP, A is the total concentration of MABA-dTDP and P is the total concentration of kinase (considered to be monomers).

Modelling studies

l-nucleotides and nucleotides analogues were drawn on Molsoft (http://www.molsoft.com, 2D to 3D converter) and Smiles Translator (http://cactus.nci.nih.gov/services/translate/) to produce the PDB files. The conformation of the sugar ring was checked by visualizing the molecules within the PyMOL graphic system and, if necessary, modified to the β-L or β-D configuration, as desired (34). Docking of the l-nucleotides and nucleotide analogues was performed using ArgusLab software (35). The binding site was defined from the coordinates of the ligand in the original PDB files 2UKD for hUCK and 1E2D for hTMPK. Docking precision was set to ‘high’ and the ‘flexible ligand docking’ mode was used for each docking run. Resulting complexes were visualized with the PyMOL graphic system and the diagrams for each nucleotide analogue/NMP kinase model obtained were drawn using Chemdraw CambridgeSoft (ChemDraw, CambridgeSoft Corporation, USA, http://www.camsoft.com).

RESULTS AND DISCUSSION

Strict stereospecificity of recombinant human AMP kinases 1 and 2

The activities of the major adenylate kinases hAK1 (cytosolic) and hAK2 (in the inter-membrane space of mitochondria) were measured in vitro as a function of the d- and l-(d)AMP concentrations as well as of pyrimidine nucleoside monophosphates. Human AK1 and AK2 were both specific for the d-enantiomers. They did not phosphorylate l-(d)NMP, except for hAK2 having very little action on l-dAMP corresponding to a ‘catalytic efficiency’ about 10⁴ smaller than for the d-stereoisomer (Table 1). The rate of d-(d)AMP phosphorylation increased with the substrate concentration, as expected for Michaelis curves. The maximum rate V_max of d-AMP phosphorylation by hAK2 was not reached due to substrate inhibition at concentrations above 0.3 mM (data not shown). This inhibition is generally attributed to unproductive binding of the ligand at the same site or at a secondary site. The reaction rates for hAK2 were slightly below those for hAK1 but were almost compensated for by the K_M values, resulting in somewhat similar catalytic efficiencies as might be expected from the similarities of their active site sequences. d-dAMP was a slightly poorer substrate than d-AMP for both hAK1 and hAK2. The deoxyribonucleotides, d-dCMP and d-dUMP, were not substrates for the AKs although these enzymes phosphorylated the pyrimidine nucleotides, d-CMP and d-UMP, to a minor extent (Table 1) (15).

Table 1.

Catalytic parameters for the (d)AMP enantiomers with human AMP kinases 1 and 2

Enzyme	Substrate	KM (mM)	Relative Vmax (%)	kcat/KM (M−1 s−1)
hAK1	d-AMP	0.14 ± 0.02	100a	3 × 106 (100b)
	d-dAMP	1.5 ± 0.3	48 ± 4	1.6 × 105 (5.3)
	d-CMP	3.0 ± 0.8	5.6	9.5 × 103 (0.32)
	d-dCMP	–c	–c	–c
	d-UMP	6 ± 2	2.9	2.6 × 103 (0.08)
	d-dUMP	–c	–c	–c
	l-AMP	–c	–c	–c
	l-dAMP	–c	–c	–c
hAK2	d-AMPd	0.08 ± 0.02	16	106 (33)
	d-dAMP	0.21 ± 0.05	22	5 × 105 (17)
	d-CMP	6 ± 1	13	104 (0.33)
	d-dCMP	–c	–c	–c
	d-UMP	9 ± 2	0.32	180 (0.006)
	d-dUMP	–c	–c	–c
	l-AMP	–c	–c	–c
	l-dAMP	–c	–c	70 (0.002)

^aThe relative V_max were obtained by comparing the k_cat value of the substrates to that of d-AMP. For V_max standardization, 100 corresponds to 1240 µmol of substrate transformed/min/mg, i.e. to a k_cat = 500 s⁻¹.

^bThe relative efficiencies were obtained by comparing the k_cat/K_M (M⁻¹ s⁻¹) value of a substrate to that of d-AMP.

^cNon detectable.

^dhAK2 activity was inhibited by [AMP] > 0.3 mM, (K_I = 0.5 mM).

l-AMP was bound by both enzymes. It competitively inhibited the phosphorylation of d-AMP and d-dAMP by hAK1 and hAK2, with an estimated K_I of 200 µM for hAK1 and 18 µM for hAK2 (data not shown). d-dAMP was used because the mitochondrial AK2 was inhibited by excess d-AMP. The binding of l-nucleotides to the AMP site of hAK1 and hAK2 was thus unproductive. Additionally, l-(d)AMP was also not a substrate for hUCK or hTMPK (data not shown).

Despite this poor phosphorylation, l-dA reduced the virus load in vivo in the woodchuck model of chronic B hepatitis although less efficiently than l-dC or l-dT (6). The other isoenzymes of AK, i.e. hAK3, 4, 5 and 6 could contribute to the phosphorylation of l-(d)AMP and explain this in vivo activity (36–37).

Relaxed stereospecificity of recombinant human dTMP kinase

The recent approval of l-dT for treating hepatitis B has prompted us to study the phosphorylation of l-dTMP by hTMPK. The catalytic efficiency of the enzyme for l-dTMP was around 0.7% that d-dTMP (Table 2). The catalytic turnover number k_cat was 10 times smaller than that for d-dTMP, while the K_M was 20-fold higher, resulting in a lower k_cat/K_M (about 900 M⁻¹s⁻¹) (Figure 2A, Table 2). l-dTMP also gave rise to substrate inhibition at 1 mM and the K_I value was estimated to be 5 mM (Table 2). d-dUMP was a good substrate with an efficiency in the 10⁴ M⁻¹s⁻¹ range, as already reported (38), but the phosphorylation of l-dUMP very slow (k_cat/K_M = 60 M⁻¹ s⁻¹).

Table 2.

Catalytic parameters for the dTMP and dUMP enantiomers with human TMP kinase

Substrate	KM (mM)	kcat (s−1)	kcat/KM (M−1 s−1)	KI (mM)
d-dTMP	0.020 ± 0.005	3.0 ± 0.1	1.5 × 105 (100a)	–
d-dUMP	0.17 ± 0.01	4.8 ± 0.6	2.8 × 104 (18)	2.6 ± 0.9
l-dTMP	0.38 ± 0.07	0.34 ± 0.03	900 (0.7)	5 ± 1
l-dUMP	2.3 ± 0.2	0.13 ± 0.01	60 (0.04)	–

^aRelative efficiency expressed by comparing the k_cat/K_M (M⁻¹ s⁻¹) value for a substrate to that of dTMP.

Figure 2.

Reaction of human dTMP kinase with the natural nucleoside monophosphates dTMP and dUMP and their corresponding l-enantiomers. (A) Saturation curves of dTMP kinase with (open circle) d-dTMP, (filled circle) l-dTMP, (open triangle) d-dUMP and (filled triangle) l-dUMP. The experiments were carried out in the presence of 2 mM ATP and 2 mM Mg²⁺ using the standard coupled assay. The quantity of enzyme used in each experiment was: 0.18 µM for d-dTMP; 3 µM for l-dTMP; 0.12 µM for d-dUMP and 5 µM for l-dUMP. The K_M and k_cat values obtained by fitting to a hyperbole are shown in Table 1. The reaction rates v for d-dUMP and l-dTMP as substrates [S] were best fitted with Equation (2). v = (V • [S])/(K_M + [S] + ([S]² / K_I)) (2) (B) Fluorescence competition assays of different substrates with MABA-dTDP bound to human dTMP kinase. The fluorescence of MABA-dTDP (6 µM) was monitored at 430 nm (excitation wavelength at 340 nm, excitation slit = 2 nm, emission slit = 4 nm) in the presence of 6 µM enzyme in T buffer resulting in the binding of 40% of the fluorophore. The complex was titrated with (open circle) d-dTMP, (filled circle) l-dTMP and (open triangle) d-dUMP. The IC₅₀ values were 4.6, 82 and 260 µM respectively. The dissociation constants K_D calculated according to Equation (1) were 2 µM, 45 µM and 0.15 mM, respectively.

The relative binding constants for the l- and d-isomers of dTMP were measured using the fluorescent competition assay based on the fluorescent probe MABA-dTDP (28). The dissociation constants for d-dTMP and l-dTMP from hTMPK were 2 µM and 45 µM, respectively (Figure 2B). The K_D ratio was 22, comparable to the K_M ratio, i.e. 19 (Table 2). Using the same assay, the nucleosides, d- and l-dT, also competed with MABA-dTDP and had K_D values of 40 and 450 µM, respectively, indicating that the l-nucleoside also binds to the active site less efficiently than its d-counterpart. The affinity of the enzyme for l-dTMP (K_D = 45 µM), was greater than its affinity for d-dUMP (K_D = 150 µM), which is a good substrate of the enzyme (Figure 2B).

Activity of recombinant human UMP-CMP kinase with D- and L-nucleoside monophosphates

l-dUMP and l-dCMP were substrates of hUCK, but they were phosphorylated much more slowly than were the natural nucleotides, despite the K_M of the enzyme being slightly more favourable for the l-compounds. The K_M values were 1.3 mM for d-dUMP and 1.0 mM for d-dCMP, compared to 0.70 mM for l-dUMP and 0.73 mM for l-dCMP (Figure 3, Table 3). But the reaction rate was 20 times slower for l-dUMP than for d-dUMP and 100 times slower for l-dCMP than for its d-counterpart. l-CMP was also phosphorylated by hUCK but it was a very poor substrate for this enzyme, with a catalytic efficiency of 100 M⁻¹ s⁻¹, 10 times lower than that for l-dCMP and 10⁴ times lower than for d-CMP. This decreased efficiency is due to a much lower k_cat, as the K_M was similar to that for l-dCMP (0.75 mM). Excess of l-CMP was inhibitory, unlike the natural substrate. The l-derivate that was the best substrate for hUCK was l-dCMP, but it was far less efficient than the monophosphate form of the antiviral analogue l-3TC. l-3TCMP is indeed phosphorylated to l-3TCDP by hUCK with a catalytic efficiency k_cat/K_M of 2.8 × 10⁵ M⁻¹ s⁻¹ (15).

Figure 3.

Reaction of human UMP-CMP kinase with the enantiomers of (d)CMP and (d)UMP and their 2′-azido-derivatives. (A) Saturation curves of hUMP-CMP kinase as a function l-derivatives as substrates: (filled circle) l-dCMP, (filled triangle) l-dUMP and (cross mark) l-CMP obtained with respectively 0.4, 1 and 7.6 µM of hUCK in the reaction mixture. (B) Saturation curves of hUMP-CMP kinase with the d-enantiomers of (open circle) dCMP and (inverted triangle) N₃-d-dCMP. The quantity of enzyme used was 4 nM for dCMP and 0.8 nM for N₃-d-dCMP. The dash line represents the saturation curve with d-CMP obtained with 4 nM hUCK as shown in (15) and fitted to Equation (2). (C) Saturation curves of hUMP-CMP kinase with the d-enantiomers of (open triangle) dUMP and (filled diamond) N₃-d-dUMP. The quantity of enzyme used was 4 nM for dUMP and N₃-d-dUMP. The dash line represents the saturation curve with d-UMP obtained with 4 nM hUCK as shown in (15) and fitted to Equation (2).

Table 3.

Catalytic parameters for the (d)CMP, (d)UMP enantiomers and their 2′azido analogues with human UMP-CMP kinase

Substrate	KM (mM)	Relative Vmax (%)	kcat/KM (M−1 s−1)
d-CMP	0.020 ± 0.005	100a	6.5 × 106 (100c)
d-dCMP	1.0 ± 0.1	61 ± 4	7 × 104 (1.2)
l-CMP	0.75 ± 0.07	0.054 ± 0.001	100 (0.0015)
l-dCMP	0.73 ± 0.06	0.63 ± 0.02	1000 (0.015)
N3-d-dCMP	0.36 ± 0.03	408 ± 9	1.5 × 106 (23)
N3-l-dCMP	1.2 ± 0.1	0.36 ± 0.02	390 (0.006)
d-UMP	0.05 ± 0.01	100	2.8 × 106 (43)
d-dUMP	1.3 ± 0.3	5.1 ± 0.5	6000 (0.1)
l-dUMP	0.70 ± 0.09	0.28 ± 0.02	530 (0.008)
N3-d-dUMP	0.9 ± 0.2	89 ± 6	1.4 × 105 (2.15)
N3-l-dUMP	1.3 ± 0.1	0.07 ± 0.02	270 (0.004)
L-3TCMPb	0.15 ± 0.02	27	2.8 × 105 (4.3)

^aThe relative V_max was obtained by comparing the k_cat value of the substrates to that of d-CMP. V_max = 100 corresponds to 350 µmol of substrate transformed/min/mg, i.e. to a k_cat = 130 s⁻¹ [from (15)].

^bdata from (15) for comparison.

^cThe relative efficiencies were obtained by comparing the k_cat/K_M (M⁻¹ s⁻¹) value of the substrates to that of d-CMP.

Activity of recombinant human UMP-CMP kinase with 2′-azido-2′-deoxynucleotides

The activity of recombinant hUCK was evaluated using the nucleoside analogues, N₃-d-dCMP and N₃-d-dUMP (Figure 3). Both compounds were substrates of the enzyme and followed Michaelis–Menten kinetics. High concentrations did not give rise to substrate inhibition of hUCK, unlike the natural substrates, d-CMP and d-UMP. The K_M values were 0.36 mM for N₃-d-dCMP and 0.9 mM for N₃-d-dUMP compared to 20 μM for d-CMP and 50 μM for d-UMP (Table 3). Despite the higher K_M values, the catalytic efficiencies for phosphorylation of the 2′-azido derivatives were in the same order of magnitude (10⁵–10⁶) as those for the natural substrates of the enzyme. The reaction rate with N₃-d-dUMP was similar to that with d-UMP (Figure 3, Table 3). The reaction rate of hUCK with N₃-d-dCMP was almost 4 times faster than that with d-CMP, demonstrating that N₃-d-dCMP is currently the best-known substrate for hUCK (Figure 3B). Other enzymes hAK1, hAK2 and hTMPK did not significantly phosphorylate these substrates.

The enantioselectivity of hUCK for 2′-azido-2′-deoxynucleoside monophosphates was also evaluated. The K_M value for N₃-l-dCMP was increased (1.2 mM) while the V_max was lower (1.1 U·mg⁻¹), leading to a 4000-fold lower catalytic efficiency than for the d-enantiomer (Table 3). N₃-l-dUMP was also a poor substrate of the enzyme, with a K_M of 1.3 mM and a k_cat of 0.34 s⁻¹(Table 3). The 2′-azido group was thus only favourable for the d-analogues. Overall, in the l-series, the pyrimidine 2′-deoxyribonucleotides were better substrates than the pyrimidine ribonucleotides and the 2′-azido-2′-deoxyribonucleotides, indicating that 2′-substitutions may cause steric hindrance for l-NMP binding to hUCK.

Structural analysis of the enantioselectivity of human AMP, TMP and UMP-CMP kinases

The NMP kinases all have a highly conserved structure with a central CORE domain that contains an ATP binding loop (P-loop) and two mobile domains: an NMP binding domain and an LID domain, which provide the catalytic residues for the reaction (39). Both the NMP and LID domains are extremely mobile and undergo large ‘hinge bending’ motions (40). The same conformational changes are believed to occur in all NMP kinases when they switch from their opened to closed conformation upon substrate binding. The X-ray structure of hAK1 complexed with the bisubstrate inhibitor Ap5A was recently solved: it showed the interactions of AMP with the active site in the closed conformation (41). The structure of hTMPK complexed with various ligands has been thoroughly explored (42), but the structure of the free apoenzyme is not known. In contrast, the structure of hUCK is only known in its open conformation (43). Substrate-free hTMPK probably also exists in an opened conformation and substrate-bound hUCK probably adopts a closed conformation as does the homologous enzyme from Dictyostelium (44).

l-AMP and l-dTMP were tentatively docked in the closed conformation of hAK1 and hTMPK, and l-dCMP was docked in an enzyme model based on the Dictyostelium UCK (44). The docking of l-(d)AMP in hAK1 failed to provide a model (data not shown): this could be due to the rather specific interactions of d-AMP with the protein, which are not mediated by water molecules, thus limiting its capacity to accommodate other substrates. The docking of l-deoxypyrimidine monophosphates in hTMPK and UCK was more successful (Figures 4 and 5). The dTMP and CMP binding sites are represented as four interacting motifs: (i) the LID domain, (ii) the P-loop, (iii) the mobile part of NMP binding domain and (iv) the stable part of NMP binding domain. The binding of l-dTMP differed from that of d-dTMP, especially at the deoxyribose moiety (Figure 4). The H-bonds between the first layer residues and the thymine ring involved several water molecules, providing a ‘flexible’ NMP binding domain that readily accommodates modified substrates. The thymine ring was stacked onto Phe72 in l-dTMP binding, but it was not optimal, as the base was shifted through 15°. The interaction of the phosphate group with Mg²⁺ion and Asp15 in the P-loop, via a water molecule, was conserved, as was its interaction with Arg97 from the mobile NMP domain. However, the deoxyribose 3′OH, which interacts with the LID Gln157 via a water molecule in the d-dTMP/hTMPK complex, was modelled as being H-bonded to the NH₂ of the highly conserved Arg45 of the immobile NMP domain in the l-dTMP/hTMPK model.

Figure 4.

Scheme of the acceptor-binding site of human dTMP kinase in the closed form: (A) with d-dTMP bound to the acceptor-binding site; (B) with l-dTMP modelled in the acceptor-binding site. LID (red), P-loop (pink), immobile NMP domain (black) and mobile NMP domain (blue).

Figure 5.

Scheme of the acceptor-binding site of human UMP-CMP kinase in the closed form. (A) with d-CMP bound to the acceptor-binding site. (B) with l-dCMP modelled in acceptor-binding site. (C) N₃-d-dCMP bound to acceptor-binding site.

The major interactions responsible for the binding of d-CMP to hUCK involved the base and phosphate (Figure 5). The interaction of Asn100 with the 4-amino group of the cytidine base played an important role in the base specificity and explains why d-(d)CMP is a better substrate than d-(d)UMP (43). The phosphate does not interact with the P loop as in hTMPK but strongly connects the LID domain (Arg134 and Arg140) to the immobile NMP domain (Arg39, Arg96 and Glu36) (Figure 5A). These residues still interacted with l-dCMP phosphate group in the l-dCMP/UCK model, except for Arg134 (Figure 5B). The cytidine moiety for both d-CMP and l-dCMP was H-bonded to Val63 and Asn100 from the mobile NMP domain, resulting in similar positioning of the base for both enantiomers. The l-nucleotide did not have the major interaction of the d-CMP 2′OH with the main chain Lys61 carbonyl, contained within the NMP mobile domain. l-dCMP was a better substrate for hUCK than was l-CMP, probably due to 2′-OH steric hindrance. However, the ring oxygen in l-dCMP interacted with Arg96, anchoring the sugar to the immobile NMP domain (Figure 5B).

The model obtained with N₃-d-dCMP bound to hUCK had favourable interactions as the azido group interacts with both Asp142 in the LID domain and the α-carbonyl of Gly60 in the mobile NMP domain (Figure 5C). The 2′OH of the ribonucleotide has been shown to contribute to the LID closure of hUCK by its interaction with the carbonyl of Lys61 located in the mobile NMP domain (43). This interaction also explains the higher affinity of the enzyme for D-ribonucleotides compared to d-deoxyribonucleotides. The replacement of 2′-hydroxyl by a bulkier azido group prevents complete closure of the LID, which results in higher K_M values. However, the opening of the LID, which promotes product release, is proposed to be the rate-limiting step of the reaction catalyzed by hUCK. A recent NMR experiment on hyperthermophilic and mesophilic homologues of AK also suggested that the opening of the AMP binding and/or LID domains upon product release was the rate-limiting step influencing the catalytic turnover (45,46). The azido group might thus facilitate the LID opening or could connect the LID and mobile-NMP domains and synchronize their opening, explaining the higher rate of phosphorylation of N₃-d-dCMP compared to d-CMP. Dynamic studies should help to understand these higher rates.

CONCLUSION

We have compared the activities of the kinases on l- and d-nucleotides and shown that both hTMPK and hUCK have relaxed enantioselectivities for dNMP, while hAK 1 and 2 are strictly devoted to d-(d)NMP. The pyrimidine kinases phosphorylated only l-derivatives of the deoxy series. UCK was active with l-dCMP but far less efficiently than with l-3TCMP (15). hTMPK, hUCK, hAK1 and hAK2 all phosphorylated their respective d-(d)NMP with some minor cross-reactivity. For example, d-CMP was phosphorylated efficiently by hUCK (k_cat/K_M = 6.5 × 10⁶ M⁻¹ s⁻¹) and more slowly by hAK1 and hAK2 (k_cat/K_M about 10⁴ M⁻¹ s⁻¹). d-dUMP was a better substrate for hTMPK (k_cat/K_M = 2.8 × 10⁴ M⁻¹ s⁻¹) than for hUCK (k_cat/K_M = 6 × 10³ M⁻¹ s⁻¹). N₃-d-dCMP was the best d-series substrate for hUCK with a k_cat 4-fold higher than that for the natural substrate (d-CMP). This could be due to a tighter interaction between the LID domain and the NMP domain as explained by the structural analysis.

RNAi studies have shown that hTMPK is implicated in the activation of l-FMAU (47). There was generally a good correlation between kinetic studies and the antiviral effect, emphasizing the need for efficient cellular activation of the antiviral drug to its triphosphate form, as shown for l-3TC (15). The present study demonstrates the activity of hTMPK on l-dTMP, even if the catalytic efficiency was relatively low (k_cat/K_M = 900 M⁻¹ s⁻¹). l-dTMP had the highest relative phosphorylation efficiency of all the l-(d)NMP tested. The phosphorylation of l-dTMP correlates with studies on the intracellular metabolism of l-dT in HepG2 cells and primary cultures of human hepatocytes. The conversion of l-dTMP to l-dTDP clearly appears to be the rate-limiting step in both cell types (48). The first phosphorylation of l-dT is believed to be carried out by either hdCK or hTK2 with catalytic efficiencies of 1.6 × 10⁴ M⁻¹ s⁻¹ and 1 × 10⁷ M⁻¹ s⁻¹, respectively (9). The recent crystallization of dCK with l-3TC and troxacitabine shows how the nucleoside binding site of dCK can conserve all essential interactions with these analogues or l-dC and thus maintain productive substrate positioning for phosphoryl-transfer (49). Given the great similarity between dCK and hTK2, we can assume that they are similarly flexible for the productive positioning of l-dT, explaining the high catalytic efficiency. The binding of l-dTMP to hTMPK is not optimal and may be the rate-limiting step in the pathway. NDPK does not recognize l-(d)NDP as substrates and phosphoglycerate kinase is probably involved in the conversion of l-dTDP into l-dTTP, with a catalytic efficiency of 500 M⁻¹ s⁻¹ (18). From these in vitro data, the second and third phosphorylation steps have low catalytic efficiencies and appear rate-limiting in the activation pathway. This study demonstrated the relaxed enantioselectivity of hTMPK, which is most likely essential to the formation of l-dTTP in vivo and thus to the antiviral properties of l-dT.