Inhibition of Thymidine Kinase by Pl-(Adenosine-5’)-P5-(thymidine-5’)-pentaphosphate*

Potential bisubstrate analogs, with adenosine and thymidine joined at their 5’ positions by polyphos-phoryl linkages of varying lengths (Ap,dT, where n = the number of phosphoryl groups), were examined as inhibitors of cytosolic thymidine kinase from blast cells of patients with acute myelocytic leukemia. Ki values were 1.2 PM for Ap,dT, 0.31 p~ for Ap4dT, 0.12 pM for Ap,dT, and 0.19 ~ L M for ApedT. The best inhibitor of the cytosolic enzyme, ApsdT, was somewhat less effective as an inhibitor of the mitochondrial enzyme (Ki = 0.50 p ~ ) . In addition to their inhibitory modes of binding by the cytosolic enzyme, these compounds were bound at considerably lower concentrations for Ap5dT, and PM for Ap4dT), in such a way as to protect the cytosolic enzyme from thermal inactivation at 37 “C in the absence of substrates.

Thymidine kinase, an enzyme of the pyrimidine salvage pathway, catalyzes the transfer of phosphate from ATP to thymidine in the presence of Mg2+ (1, 2). Cytosolic and mitochondrial forms of thymidine kinase have been isolated from the blast cells of acute myelocytic leukemia (3, 4) as dimers with M, = 90,000 and 70,000, respectively. ATP is bound cooperatively by the cytosolic enzyme, and both enzymes are allosterically inhibited by dTTP (5). Because of its increased activity during DNA synthesis (6), in neoplastic cells (7,8) and in some virally infected cells (9), and its role in the activation of antiviral nucleosides (10,11), cytosolic thymidine kinase has attracted interest as a possible target for the development of antimetabolites (12,13).
Adenylate kinase, catalyzing a similar reaction, is inhibited by compounds in which two molecules of adenosine are joined at their 5'-positions by linear polyphosphoric acid. That enzyme was strongly inhibited by Ap5A1 (14), but relatively weakly by Ap4A (15). Tight binding was considered to arise from the positioning of the 2 adenosine residues in a spatial relationship resembling that which they might adopt in the transition state for phosphate transfer (14). Strong inhibition by ApBA also provided the first evidence that the active site * This work was supported by National Institutes of Health Grants GM-18325 and CA-27448. 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 U.S.C. Section 1734 solely to indicate this fact.
Analytical Methods-Proton magnetic resonance spectra of ApldT and ApsdT were recorded using Varian EM 390 and Bruker WM 250 spectrometers, operating at 90 MHz and 250 MHz, respectively. 31P NMR spectra were obtained using a Bruker WM 250 spectrometer operating at 101.3 MHz, employing acquisition and delay times of 2.2 and 2.0 s, respectively. Inverse gated proton decoupling was used to eliminate nuclear Overhauser effects. Mass spectra of inhibitors were obtained on a VG micromass mass spectrometer (model ZAB 2F) using fast atom bombardment with either positive or negative argon ions. Mass spectra of the tetraammonium salt of PI-P4-di(adenosine-5'btetraphosphate established that the mass/charge for the molecular ion corresponded to that expected for the singly ionized free acid. In addition to the spectroscopic methods described above, inhibitors were analyzed for purity in TLC systems A (cellulose developed in 0.25 M NH4HC03:2-propanol (3565)) and B (polyethyleneimine cellulose developed in 0.9 M LiC1). After digestion for 1 h using phosphodiesterase (0.5 mg/ml), alkaline phosphatase (0.5 mg/ml), and MgClz (1.0 mM), product mixtures were again analyzed in TLC System A using authentic adenosine and thymidine standards (Table   I), and inorganic phosphate was determined using ammonium molybdate (19). Extinction coefficients of Ap4dT and ApsdT were determined by measuring the absorbance at 262 nm before and after enzymatic digestion, and comparison with the absorbance of equimolar mixtures of adenosine and thymidine of known concentration ( Table I).
Synthesis of Inhibitors-Tributylammonium salts of nucleotides (20) were prepared by first converting their barium or calcium salts (0.1 mmol) to pyridinium salts by treatment with Dowex 50W-X8 (pyridinium form, 30 g) in water. Filtrates were evaporated under reduced pressure to a volume of 10-20 ml, and 1 molar equivalent of tributylamine/mol of phosphorus was added, along with sufficient ethanol to produce a homogeneous solution. The tributylammonium salt was then evaporated to dryness several times, from ethanol and then from dry pyridine.
Ap3dT (201, ApddT, ApsdT, and ApsdT were synthesized by the 5731  phosphorimidazolidate method of Hoard and Ott (21). For the synthesis of ApsdT, Ap4dT, and ApsdT, tributylammonium adenosine nucleotide (ADP, ATP, or adenosine 5'-tetraphosphate, 0.1 mmol) was dissolved in a minimal volume of dry dimethylformamide and stirred for 2 h with l,l'-carbonyldiimidazole (0.5 mmol). Methanol (0.6 mmol) was then added, the solution was stirred for 30 min, and tributylammonium dTMP (0.25 mmol) was added in a minimal volume of dry dimethylformamide. For the synthesis of ApddT, this procedure was reversed and tributylammonium dTTP (0.1 mmol) was first converted to the imidazolidate, and then allowed to react with tributylammonium ATP (0.25 mmol). After standing for 48 h at room temperature, reaction mixtures were evaporated to dryness under reduced pressure, dissolved in methanokwater (1:1), adjusted to pH 10.5 with triethylamine, and stirred for 2 h (18). After evaporation to dryness under reduced pressure, the residue was redissolved in water and subjected to chromatography on DEAE-Sephadex (2.5 X 25 cm), eluting with a linear gradient of aqueous NH4HCO1 (2000 ml, 0-0.8 M). Several peaks of UV-absorbing material resulted. In each case, the peak containing the desired compound, emerging at a salt concentration of 0.45-0.65 M, was identified by its absorption maximum (262 nm), resistance to alkaline phosphatase digestion, conversion to an equimolar mixture of adenosine and thymidine when digested with phosphodiesterase and alkaline phosphatase, and phosphate content, with the results shown in Table I. Products were desalted by repeated evaporation to dryness, rechromatographed, and worked up in the same way under reduced pressure from water/ethanol mixtures. They were stable during storage for at least 3 months in vacuo over Pzos in the absence of light, or when stored frozen in aqueous solution.
Properties and spectroscopic data of the products are listed in Table I. After rechromatography on DEAE-Sephadex, ApSdT, AprdT, and ApsdT appeared to be more than 97% free of UV-absorbing impurities by TLC analysis and gave satisfactory integrated intensities when analyzed by 'H and 31P NMR spectra (Table I)   ApddT was eluted last, at salt concentrations between 1.35 and 1.55 M, and these functions were repeatedly evaporated (60°C) under reduced pressure until less than 5 g of ammonium acetate remained.
The residue was dissolved in water (400 ml) and subjected to chromatography on DEAE-Sephadex (2.5 X 25 cm) eluting with a gradient of NH4HC03 (2000 ml, from 0.1-0.7 M). Ap6dT emerged in the second of two broad peaks that partially overlapped. Fractions free of the front-running material were desalted by repeated evaporation under reduced pressure. Ap6dT now appeared to be greater than 95% pure by TLC and NMR criteria (Table 111). Enzyme Assays-Thymidine kinase was prepared as previously described (3,4) and assayed at 37 "C in a final volume of 100 pl that contained Tris-HCl(50 mM, pH 7.5), dithiothreitol (2 mM), NaF (7.5 mM), bovine serum albumin (1.2 mg/ml), creatine phosphate (13 mM), creatine kinase (22.8 units/ml), [meth~l-~HIthymidine (0.2-0.5 pCi), and varying concentrations of thymidine and ATP (3,4). MgC12 was maintained equal t o the total nucleotide concentration except as otherwise stated. Reactions were initiated by the addition of enzyme stock solution (10 pl, 1-20 X units') and quenched 5y spotting the reaction mixture (50-70 pl) on a disc of DE81 filter paper that was then immersed in 95% ethanol (5 ml/filter) for at least 5 min. The ethanol was decanted and the filters were washed for 5 min two additional times with 95% ethanol (5 ml/filter). Filters were dried, incubated (15 min) in scintillation vials with a solution of NaCl (2 M) and HCl(O.2 M), and finally scintillation fluid was added (6 ml of Scintiverse E). Reaction velocities were found to remain constant over a period of 60 min.
Rate constants for inactivation were calculated from half-times of inactivation, using the relationship k = 0.693/t1/z.

RESULTS
Effects of ApsdT, Ap,dT, ApSdT, Ap6dT, and ApA on the activity of cytosolic and mitochondrial thymidine kinases were first determined in the presence of fixed concentrations of ATP (100 p~) and thymidine (5.0 p~) near their K, values, as summarized in Table 11. ApsdT produced the strongest One unit of thymidine kinase is the amount of enzyme which converts 1 nmol of thymidine to dTMP per min under standard assay conditions. inhibition, Ap4dT and Ap6dT were nearly as effective, and Ap,dT and Ap5A were much less effective.
In view of the resemblance of these compounds to possible reaction intermediates, it was of interest to examine the stability of the strongest inhibitor, Ap,dT, in the presence of the enzyme. Rates of reactions catalyzed by both cytosolic and mitochondrial enzymes were found to be constant for at least 1 h either in the absence or presence of sufficient ApdT to inhibit the reactions by 90%, suggesting that no significant breakdown of Ap5dT had occurred during that period. To test the stability of Ap5dT in the presence of the enzymes more directly, Ap5dT (20 p~) was incubated with MgClz (100 flM) and thymidine kinase (6.7 x units) in standard assay buffer without substrates or ATP-regenerating components. After 1 h at 37 "C, the reaction was quenched by immersion in boiling water for 2 min, and the denatured protein was removed by centrifugation. Aliquots (20 pl) of the supernatant fluid were analyzed by high performance liquid chromatography using a Zorbax-SAX column (4.6 mm X 25 cm), eluting with NH4H2P04 (1.1 M, pH 3.5) and monitoring the eluate absorbance at 254 nm. Negligible hydrolysis of the inhibitor was found to have occurred; experiments with control samples, incubated without enzyme or Ap5dT, showed that 5% hydrolysis would have been detected. Fig. 1 shows the dependence of initial reaction rates on thymidine concentration (0.5-20 p~) in the presence of Ap5dT (0, 15, and 30 p~) and A T P (2.0 mM). Fig. 2 shows double reciprocal plots of initial reaction velocity as a function of changing ATP concentration in the presence of saturating concentrations of thymidine (50 mM) in the absence (circles) and presence of Ap5dT (0.475 p~, triangles; 1.9 p~, squures). Nonlinear double reciprocal plots were observed, suggesting that ATP was bound cooperatively by cytosolic thymidine kinase (5). The substrate concentration required for halfmaximal velocity was estimated from linear plots of l / v versus l/(ATP)Z (Fig. 3). Table 111 shows Ki values observed for  Ap,dT, determined from double reciprocal plots of initial reaction rates as a function of changing concentrations of each of the two substrates, in the presence of saturating concentrations of the co-substrate.
To avoid complications arising from the cooperative binding of ATP by cytosolic thymidine kinase, the affinities of ApadT, Ap4dT, ApsdT, and ApsdT were estimated by monitoring activity at substrate concentrations (ATP = 7.5 PM, thymidine = 100 nM, MgCl:, = 100 WM) far below their K,,, values, in the absence and presence of increasing concentrations of inhibitors. K i values were estimated from the reciprocal of the slope of a plot of uninhibited initial velocity divided by inhibited initial velocity as a function of inhibitor concentration (Fig. 4). Kj values determined in this way were  absence of substrates or Ap5dT, but not in their presence. Substrates have been reported previously to confer thermal stability on thymidine kinase, when present at concentrations above those that produce half-maximal velocity (4, 22). The ability of inhibitors to protect cytosolic thymidine kinase from irreversible inactivation was used to determine apparent dissociation constants for their complexes with the cytosolic enzyme in a different way. Activity remaining was monitored as a function of time in the absence and presence of various concentrations of ApbdT, and rate constants were determined from the half-time for inactivation (Fig. 5 ) . These rate constants, plotted as a function of increasing concentrations of ApsdT yielded an apparent dissociation constant of 2.5 nM for ApsdT (Fig. 5 ) . This dissociation constant was considered an upper limit, since binding was so tight that some depletion of ApsdT could not be avoided under these conditions. Protection from inactivation was unaffected by the addition of MgClz (100 @f). Under the same experimental conditions, the rate of inactivation was reduced 47% by 1.0 FM Ap3dT, 55% by 30 nM Ap4dT and 88% by 20 nM Ap6dT. Dissociation constants were estimated from these results to be 1.1 PM for ApadT, 29 nM for Ap4dT, and 2.7 nM for ApsdT.

DISCUSSION
Ap,dT was earlier shown to inhibit rat cytosolic thymidine kinase with an affinity comparable with the apparent affinity of the substrate-ATP (18). The rat cytosolic enzyme apparently differs from the human enzyme in binding ATP noncooperatively, and in being somewhat more sensitive to inhibition by Ap3dT (18). The present inhibitors were bound by the human enzyme substantially more tightly than Ap3dT, consistent with the possibility that these compounds bear a somewhat closer resemblance to the phosphoryl donor and acceptor, in a spatial relationship similar to that which they may adopt during direct phosphoryl transfer. The ordered sequential kinetics of substrate binding by human liver cytosolic ( 5 ) and mitochondrial (14) thymidine kinases accord with a mechanism involving direct phosphate transfer. In keeping with the established specificity of thymidine kinase for thymidine as an acceptor ( 5 ) , Ap4A was found in the present study to be relatively ineffective as an inhibitor (Table  11).
Somewhat surprisingly, the affinities of the most effective (ApsdT; Ki = 0.12 PM) and the least effective (ApadT; Ki = 1.2 p~) inhibitors examined in the present study differed by only one order of magnitude. This lack of discrimination by thymidine kinase stands in marked contrast to the behavior of adenylate kinase, which showed 104-fold higher affinity for Ap5A (Ki = 2.5 nM) than for Ap4A (Ki = 24 KM) (15). Since ApsA was very weakly bound by thymidine kinase (Table 11), it seems likely that both ends of these two-headed inhibitors are involved in enzyme-inhibitor interaction. The insensitivity of binding to chain length, in inhibitors of the type Ap,dT, suggests that the inhibitors may adapt in different ways to fit the stereochemical requirements of the active site, or that recognition sites on the enzyme may be somewhat flexible in their relative orientation. In this latter connection, it appears possible that the binding sites are aligned in such a way as to encourage nucleophilic displacement on phosphorus, but are sufficiently mobile to allow some latitude in the distance separating the phosphoryl donor and acceptor; thus, inhibitors with different numbers of phosphoryl groups might be accommodated with similar affinities by the active site.
The unexpected ability of these inhibitors to protect thymidine kinase from thermal inactivation, at concentrations substantially below those at which half-maximal inhibition was achieved, indicates the presence on the enzyme of an additional binding site or sites. In view of the considerations that went into their design, it is surprising that these twoheaded analogs should adhere with such high affinity to noncatalytic sites on the enzyme. These sites differ from the inhibitory site not only in their absolute affinity, but also in their relative affinity for ligands. Thus, ApsdT was bound at least as tightly as Ap,dT, Ap,dT was bound an order of magnitude less tightly, and Ap,dT protected the enzyme only at the high concentrations that also produced inhibition. Detailed information is not yet available concerning the overall structure of this enzyme, but it is worth noting that inhibition of cytosolic thymidine kinase by dTTP has been found to result in an increase in cooperativity, so that more than two molecules of ATP appear to be bound (5). Thus, each subunit of this dimeric enzyme probably contains at least one allosteric binding site for ATP and one for dTTP, in addition to binding sites for the substrates ATP and thymidine. Inhibitor binding at any two of these sites, on the same or different subunits, could presumably result in thermal stabilization of the enzyme.