Negative Cooperative Binding of Thymidine, Ordered Substrate Binding, and Product Release of Human Mitochondrial Thymidine Kinase 2 Explain Its Complex Kinetic Properties and Physiological Functions

Abstract

Mitochondrial thymidine kinase 2 (TK2) catalyzes the phosphorylation of thymidine (dT) and deoxycytidine (dC) and is essential for mitochondrial function in post-mitotic tissues. The phosphorylation of dT shows negative cooperativity, but the phosphorylation of dC follows classical Michaelis–Menten kinetics. The enzyme is feedback-inhibited by its end products deoxythymidine triphosphate (dTTP) and deoxycytidine triphosphate (dCTP). In order to better understand the reaction mechanism and the negative cooperative behavior, we conducted isothermal titration calorimetry (ITC) and intrinsic tryptophan fluorescence (ITF) quenching studies with purified recombinant human TK2. Cooperative binding was observed with dT but not dC by the ITC analysis in accordance with earlier enzyme kinetic studies. The phosphate donor adenosine triphosphate (ATP) did not bind to either dTTP-bound or dTTP-free enzymes but bound tightly to the dT– or dC–TK2 complexes with large differences in enthalpy and entropy changes, strongly suggesting an ordered binding of the substrates and different conformational states of the ATP and dT– and dC–TK2 ternary complexes. dTTP binding was endothermic; however, dCTP could not be shown to interact with the enzyme. ITF quenching studies also revealed tight binding of dT, dC, deoxythymidine monophosphate, deoxycytidine monophosphate, and dTTP but not adenosine 5′-diphosphate or ATP. These results strongly indicate an ordered sequential binding of the substrates and ordered release of the products as well as different conformational states of the active site of TK2. These results help to explain the different kinetics observed with dT and dC as substrates, which have important implications for TK2 regulation in vivo.

Introduction

Thymidine kinase 2 (TK2) phosphorylates thymidine (dT) and deoxycytidine (dC) to their respective monophosphate and a complete deficiency of this enzyme is lethal.¹⁻³ TK2 showed different kinetic behaviors with its natural substrates, that is, the phosphorylation of dT showed negative cooperativity with biphasic substrate saturation curves, but dC phosphorylation followed hyperbolic Michaelis–Menten kinetics.^4,5 Active native or recombinant TK2 is present in the form of monomers, dimers, or oligomers.^4,6−8 Furthermore, the enzyme is feedback-inhibited by deoxythymidine triphosphate (dTTP) and deoxycytidine triphosphate (dCTP).⁷ Recombinant TK2 contains enzyme-bound dTTP, dCTP, and to lesser extent deoxyadenosine triphosphate (dATP).⁹ However, in a later study, only dTTP was found to be present in an equimolar ratio in highly purified recombinant TK2.¹⁰

The TK2 protein is expressed at low level in all tissue types, and in cultured cells, the TK2 protein concentration was upregulated in stationery phase cells, whereas in rapidly dividing cells, the levels of TK2 were significantly lower.¹¹ TK2 plays an important role in the synthesis of a DNA precursor for mitochondrial DNA (mtDNA) replication and in repair of nuclear DNA.^12,13 Deficiency of TK2 activity, due to point mutations, deletions, or insertions in the TK2 gene, results in devastating mitochondrial diseases with predominant mtDNA depletions and in some cases with additional multiple mtDNA deletions.¹⁴⁻¹⁶ Initial studies showed that the residual TK2 activity correlated with the severity of the disease and also the time of onset; however, with more cases reported, this correlation could not be verified.¹⁶⁻¹⁸ However, alteration in the TK2 substrate specificity and loss of negative cooperativity have been observed in mutant TK2 enzymes identified in patients with severe mtDNA depletion syndrome (MDS).^7,15,19 Tissues that have high energy demand are generally most affected by mtDNA depletion or deletions, and therefore, TK2 deficiency predominantly led to skeletal muscle myopathy, but multiple organ involvement has also been reported.^16,20,21

Many pyrimidine nucleoside analogues, such as zidovudine (AZT) and cytarabine (araC) used in antiviral and anticancer therapy, are either substrates or inhibitors for TK2 and often cause mitochondrial side effects. Inhibition or downregulation of TK2 may play an important role in the observed mitochondrial toxicities.^22,23 The defect in TK2 overall activity and/or alteration in kinetics can lead to mitochondrial dCTP and dTTP pool imbalance and mtDNA depletion.¹⁸

In order to better understand the complex kinetic behavior of TK2, we conducted isothermal titration calorimetry (ITC) and intrinsic tryptophan fluorescence (ITF) studies with purified recombinant TK2 to elucidate the basis for its kinetic properties with different substrates and feedback inhibitors.

Materials and Methods

Materials

dT, dC, adenosine 5′-(gamma-thio) triphosphate (ATPγS), and 3′-fluoro-2′,3′-dideoxythymidine (FLT) were purchased from Sigma-Aldrich. Stock solutions of dTTP and dCTP (100 mM) were from Promega. 3′-Azido-2′,3′-dideoxythymidine (AZT) was from Carbosynth.

Enzyme Preparation

Recombinant human TK2, with an N-terminal truncation of 55 amino acids, was expressed in Escherichia coli and purified by metal affinity chromatography on a Ni²⁺-Sepharose column. Purified TK2 contained an enzyme-bound dTTP at 1:1 ratio. In the case where the dTTP-free enzyme was needed, the enzyme-bound dTTP was removed by incubation with dT and ATP and confirmed by high-performance liquid chromatography analysis essentially as described in the literature¹⁰ and dialyzed against 10 mM Tris/HCl, pH 7.9, and 1 mM MgCl₂ overnight at 4 °C with two buffer changes before being used in the subsequent experiment. Protein concentrations were determined by the absorbance at 280 nm and calculated using the TK2 subunit molecular mass.

ITC Studies

ITC experiments were carried out by using a VP-ITC microcalorimeter (MicroCal, GE Healthcare) at 25 °C. The TK2 protein (40 μM, dTTP bound or free) dialyzed in buffer containing 10 mM Tris/HCl, pH 7.9, and 1 mM MgCl₂ was placed in the sample cell (1.4 mL) and titrated with different concentrations of ligands. All ligand solutions were prepared in the final dialysis buffer. The ligand concentration in the injection syringe was 1–3 mM. A typical experiment consisted of a first control injection of 2 μL, followed by 24 injections, each of 8 μL and 16 s duration, at a 500 s interval.

Control experiment was done by titrating the ligand into the dialysis buffer or titrating the buffer into the enzyme. Raw data were collected, corrected for ligand heat dilution, and integrated using the Microcal Origin 7 software supplied with the instrument. A single-site binding model or two-site binding model was used to fit the data by nonlinear regression analysis, yielding the binding constants (K_B), enthalpy changes (ΔH), and entropy changes (ΔS). The best-fit model was chosen for each ligand. The Gibbs free energy (ΔG) is calculated by the equation ΔG = ΔH – TΔS. The equilibrium dissociation constant (K_d) is calculated from the K_B values (K_d = 1/K_B).

ITF Quenching Studies

The ITF studies were conducted by using a Varian Cary Eclipse fluorescence spectrophotometer (Varian) at 21 °C in a buffer containing 10 mM Hepes/KOH, pH 7.6, 0.2 M NaCl, 2 mM MgCl₂, and 2 μM TK2 (dTTP bound or free). Excitation was carried out at 295 nm to avoid any influence by the fluorescence of tyrosine residues. The fluorescence emission spectrum was recorded between 305 and 400 nm with a slit of 5 nm. The effects of ligand were measured by titrating known concentrations (0.05–4000 μM) of each ligand. In the absence of a quencher, the TK2 fluorescence spectrum shows a maximum at 330 nm. The changes in fluorescence intensity (ΔF = F₀ – F) were measured as a function of the added ligand concentration. Both background emission and ligand dilution effect were corrected. Data were fitted using the Stern–Volmer equation F₀/F = 1 + K_s^*[Q], where [Q] is the concentration of the quencher and K_s is the association constant, and the modified Stern–Volmer equation F₀/ΔF = 1/(f_aK_a[Q]) + 1/f_a, where [Q] is the concentration of the quencher, K_a is the quenching constant of the accessible fraction, and f_a is the fraction of the initial fluorescence that is accessible to the quencher.²⁴

Equilibrium binding constants were calculated by using nonlinear regression analysis of the fluorescence data using the Prism 5 GraphPad software with an in-built equation for binding the saturation model, for example, one-site specific binding ΔF = ΔF_max^*[S]/(K_d + [S]), where [S] is the ligand concentration and K_d is the equilibrium binding constant. Data are presented as the mean ± SD of four to six independent measurements.

Results

Binding of dT but Not dC Exhibited Negative Cooperativity

dT and dC were titrated to the dTTP-bound enzymes, and in both cases, binding occurred with negative ΔH and ΔS values, which resulted in negative ΔG values (Table 1), indicating that the binding of dT and dC to TK2 is an energetically favorable event and that dT and dC are able to compete out the enzyme-bound dTTP because both dT and dC could bind to the dTTP-bound enzyme (Figure 1). These results are in agreement with earlier studies demonstrating that incubation with dT or dC releases enzyme-bound dTTP.^9,10

Table 1. Association Constants and Thermal Dynamic Parameters of Ligand Binding to TK2^a.

	ΔH (kcal/mol)	ΔG (kcal/mol)	ΔS (cal/mol/deg)	KB (μM–1)	Kd (μM)
dTTPb	237.2 ± 0.2	–6.0 ± 0.1	816	0.028	35.7
dTc	–72.7 ± 0.2	–7.1 ± 0.2	–220	1.8, 16.3	0.55, 6.13
dCc	–23.4 ± 0.7	–2.1 ± 0.02	–7.9	0.076	13.2
ATPγS to E–dT complex	–19.4 ± 0.8	–8.2 ± 0.1	–37.6	1.07	0.93
ATPγS to E–dC complex	–7.5 ± 0.5	–0.18 ± 0.01	–24.9	0.05	20.0

^aK_B binding constant; K_d, dissociation constant; ΔH, enthalpy changes; ΔS, entropy changes; ΔG, Gibbs free energy.

^bdTTP-free enzyme.

^cdTTP-bound enzyme.

Figure 1.

ITC titration curves and binding isotherms of dT (A) and dC (B) titrated to the dTTP-bound enzyme.

However, the binding of dT and dC to the dTTP-bound enzyme yielded two distinct titration curves; the titration curves for dT binding (Figure 1A) did not show the commonly observed hyperbolic shape as was observed in the case of dC binding (Figure 1B). Instead, it gave a sigmoidal trace of the heat data, with a more complex behavior where the heat release increased with the first few injection and then reached a maximum and subsequently decreased (Figure 1A). The dT binding isotherm was fitted using the two-site binding model, which yielded two binding constants, one with high affinity (K_d1 = 0.55 μM) and the other with lower affinity (K_d2 = 6.13 μM) (Table 1). Interestingly, the K_d values for dT and dC are in the same range as the K_M values determined using radiochemical ³H-Thd phosphorylation assays.^7,25

ATP Does Not Bind to Either the dTTP-Free or dTTP-Bound Enzyme but Binds to the Enzyme–Nucleoside Complexes

To study the binding capacity of ATP, the phosphate donor of the TK2-catalyzed reaction, a nonhydrolyzable ATP analogue, that is, ATPγS, was used. ATPγS was titrated with either dTTP-bound or dTTP-free TK2. As shown in Figure 2A, titration with ATPγS showed no heat response, indicating that there is no specific interaction of ATPγS to either the dTTP-bound or dTTP-free enzyme. However, when ATPγS was titrated to the enzyme–dT (E–dT) or enzyme–dC (E–dC) complexes, a typical binding isotherm was observed, which is exothermic with large negative enthalpy changes (Figure 2B,C). The binding constants for ATPγS to E–dT and E–dC were 1.1 × 10⁶ and 5.0 × 10⁵ M^–1 and the dissociation constants were 0.9 and 20 μM, respectively (Table 1).

Figure 2.

Representative ITC titration curves and binding isotherms of ATPγS titrated to free enzyme (A) or E–dT (B) and E–dC (C) complex.

Because ATP did not bind to the free enzyme but to the E–dT or E–dC complexes, we propose that the binding of dT or dC induced a conformational change of the enzyme, which allowed formation of an ATP–nucleoside–enzyme ternary complex. To test if this binding order is common among deoxynucleoside kinases, we titrated ATPγS to human recombinant dC kinase (dCK) in the presence and absence of dC and found that ATPγS does not bind to dCK alone but bind to the dCK–dC complex, similar to what was observed here with TK2 (Supporting Information, Figure S1).

Binding of Feedback Inhibitors

Both dTTP and dCTP are feedback inhibitors of TK2 with K_i values in the micromolar range.⁷ Purified recombinant TK2 has been shown to contain enzyme-bound dTTP, dCTP, and to a lesser extent dATP 9, and in a more recent study, only dTTP was bound to the purified TK2 enzyme.¹⁰ The enzyme-bound deoxyribonucleotide triphosphates (dNTPs) could be removed by incubation with dT or other nucleosides with or without ATP.^9,10 In order to study the binding affinity of feedback inhibitors, the dTTP-free TK2 was used. The binding of dTTP exhibited an endothermic heat response (Figure 3A) with a positive ΔH value; however, the large positive ΔS value upon dTTP binding resulted in a negative ΔG value (Table 1), which makes the binding of dTTP thermodynamically feasible and also indicates profound conformational changes upon dTTP binding. The binding constant is 2.8 × 10⁴ M^–1 and the dissociation constant is 36 μM, which is higher than the K_i values (2.5 μM) determined in enzyme kinetic experiments (Table 1).⁷ Titration of dCTP, however, showed no heat changes in the isothermogram, which most likely represents a typical unspecific interaction toward the macromolecule in an ITC experiment (Figure 3B). These results are in accordance with our earlier study, which showed that only dTTP was bound to the purified TK2.¹⁰

Figure 3.

Representative ITC titration curves and binding isotherms of dTTP (A) and dCTP (B) titrated to the dTTP-free enzyme.

ITF Spectrometric Studies

Human mitochondrial TK2 possesses five tryptophan residues⁵ that can be used to monitor the fluorescence quantum changes because of conformational changes induced by ligand binding. The recombinant TK2 used in this study is an N-terminal truncated form containing four tryptophans,⁵ and all of them are conserved in the TK2 enzyme family. On the basis of a TK2 structure model,¹⁵ three of them are located at the dimer interface and one is located at one of the two α-helices that form the lid region. These suggest that ITF could be a useful technique to study the subunit interaction and/or conformational changes in the TK2 structure upon substrate binding. The TK2 fluorescence emission spectrum was recorded between 305 and 400 nm with a λ_max of 330 nm. To study the effects of ligand binding-induced conformational changes, the natural substrates dT, dC, and ATP and their products deoxythymidine monophosphate (dTMP), deoxycytidine monophosphate (dCMP), and adenosine 5′-diphosphate (ADP), the feedback inhibitor dTTP as well as the nucleoside analogues AZT and FLT were titrated to the dTTP-bound or dTTP-free TK2 in a concentration range of 0.5–4000 μM. In addition, ATPγS was used to study the ATP effects on dT and dC binding to the enzyme. Addition of dT, dC, dTMP, dCMP, and dTTP as well as AZT and FLT caused quenching of the fluorescence emission but not the addition of ATP or ADP (Supporting Information, Figure S2). A slight λ_max shift (from 330 to 328 nm) was observed when dTMP or dCMP was added to the enzyme.

The Stern–Volmer plot, that is, a F₀/F versus dT concentration plot, yielded a downward curvature at low dT concentration (0.05–40 μM), which is a characteristic feature of two fluorophore populations, one of which is not accessible to the quencher, whereas the tryptophan residues close to the protein surface are quenched. At higher dT concentration (40–4000 μM), the Stern–Volmer plot is linear (Figure 4A and Supporting Information, Figure S2). Fitting the fluorescence data with the modified Stern–Volmer equation, that is, F₀/ΔF versus 1/dT concentrations, a fraction of the initial fluorescence that was accessible to dT (f_a = 25%) could be identified, with a quenching constant of 0.16 μM^–1 (Figure 4B and Table 2). The other ligands dTMP, dC, dCMP, AZT and FLT, and dTTP gave similar quenching profiles as dT (Supporting Information, Figure S2) and a fractional initial accessible fluorescence of approximately 20–40% (Table 2). The λ_max value and f_a data (Table 2) indicate that only one of the four tryptophan residues is involved in the initial quenching of the TK2 fluorescence emission induced by ligand binding at a low concentration range. The results also demonstrate that TK2 undergoes conformational changes upon ligand binding. The equilibrium binding constant of each ligand was calculated by the nonlinear regression fitting of the fluorescence changes (ΔF) in the presence of ligands using the best-fit binding model (Figure 4C and Table 3).

Figure 4.

Quenching of ITF by dT. (A) F₀/F vs [dT] plot; (B) F₀/ΔF vs 1/[dT] plot; and (C) ΔF vs [dT] plot with dTTP-bound TK2.

Table 2. Stern–Volmer Constants Obtained by Fitting of the Fluorescence Quenching Data to the Modified Stern–Volmer Equation^a.

ligand	Ka (μM–1)	fa
dT	0.16 ± 0.01	0.25 ± 0.01
dC	0.14 ± 0.01	0.29 ± 0.05
AZT	0.32 ± 0.03	0.38 ± 0.04
FLT	0.14 ± 0.02	0.24 ± 0.04
dTMP	0.79 ± 0.04	0.21 ± 0.02
dCMP	0.42 ± 0.03	0.21 ± 0.02
dT/ATPγS	0.13 ± 0.02	0.24 ± 0.01
dC/ATPγS	0.05 ± 0.01	0.31 ± 0.03
dTTPb	0.67 ± 0.39	0.22 ± 0.01
dTb	0.91 ± 0.10	0.17 ± 0.02
dCb	0.13 ± 0.04	0.40 ± 0.10

^aTitration was done at 21 °C with 2 μM TK2 in the reaction buffer and various concentrations of the ligands. K_a, quenching constant (μM^–1) of the accessible fraction, and f_a, fractional accessibility.

^bWith dTTP-free TK2.

Table 3. Equilibrium Binding Constants Derived from ITF Quenching of TK2^a.

ligand	Kd (μM)	ΔFmax
dT	3.26 ± 1.01	21.95 ± 1.25
dC	3.35 ± 0.81	19.86 ± 0.94
AZT	6.94 ± 2.78	12.77 ± 1.34
FLT	6.17 ± 1.35	17.81 ± 0.99
dTMP	1.53 ± 0.28	16.74 ± 0.53
dCMP	4.42 ± 0.99	19.74 ± 1.18
dT/ATPγSb	2.86 ± 0.59	25.88 ± 0.95
dC/ATPγSb	13.74 ± 3.72	15.10 ± 1.13
dTTPc	3.64 ± 1.02	57.6 ± 2.9
dTc	13.3 ± 1.2	69.7 ± 3.1
dCc	5.64 ± 1.20	67.5 ± 2.9

^aThe equilibrium binding constants (K_d) were calculated by nonlinear regression analysis of changes in ITF in the absence and presence of the quencher (F₀–F) by using the GraphPad in-built binding equations for the one-site specific binding model. Data are presented as mean ± SD.

^bIn the presence of 20 μM ATPγS.

^cWith dTTP-free TK2.

Discussion

Conformational Changes upon Substrate Binding and Negative Cooperativity

Cooperativity is ascribed to the conformational changes in the macromolecular structure induced by ligand binding. ITC is an ideal method to study cooperativity because it can register even very small changes in enthalpy and entropy and thus free energy changes that reflect both local changes, for example, ionic, hydrogen bonds, and van der Waals interactions, and global dynamic motions of a protein molecule interacting with a ligand.²⁶

TK2 exhibits negative cooperativity with dT and the dT analogue AZT but not with the alternate substrate dC as shown in earlier enzyme kinetic studies,^4,5,18 but the mechanism of this negative cooperativity is not known. TK2 belongs to the deoxyguanosine kinase (dGK), dCK, and Drosophila melanogaster–deoxynucleoside kinase (Dm–dNK) enzyme family, whose three-dimensional (3D) structures have revealed one nucleoside binding site (phosphate acceptor) and one ATP binding site (phosphate donor) per subunit. The two structural elements for the binding of a phosphate donor, for example, the lid region and the p-loop, undergo large conformational changes upon substrate binding and during catalysis.²⁷⁻²⁹ Although the TK2 structure has not been solved, a TK2 structure model, built based on sequence homology to and structure of the Dm–dNK,^15,28 predicts one nucleoside (phosphate acceptor) binding site and one phosphate donor binding site per subunit. Dm–dNK, dGK, and dCK are dimers both in crystal structures and in the native state.²⁷⁻²⁹ TK2, however, has been shown to occur as monomers, dimers, and tetramers in solution,^6,7,9,30 which support the observed cooperative behavior of multimeric enzymes. In this study, we used two biophysical methods to determine the interaction of TK2 with its substrates. The ITC studies showed a distinct difference in isothermal grams of dT and dC binding to the dTTP-bound enzyme, for example, a hyperbolic curve for dC binding and a sigmoidal curve for dT binding, indicating that only the binding of dT shows negative cooperativity. Similar binding isothermal grams have been observed for the binding of MgATP or MgADP to the Archaeoglobus fulgidus–GlnK2 protein and the binding of cyclic AMP to hyperpolarization-activated cyclic nucleotide gated channels. These binding isothermal results suggested the presence of more than one ligand binding site per molecule with different binding affinities for the same ligand that is negative cooperativity.^31,32 Fitting the dT binding data yielded two binding constants, suggesting that there are two binding sites with different affinities, one with high affinity and the other with low affinity. Because one TK2 subunit contains only one dT binding site, it is most likely that binding of dT to one subunit induced conformational changes on the adjacent subunit and thus affects the binding affinity of the next incoming dT, giving cooperativity. The large changes in enthalpy and entropy values suggest that binding of dT results in both local and global conformational changes of the TK2 structure. Binding of dC, however, showed smaller changes in both enthalpy and entropy values and thus less effect on enzyme conformation. These results suggest that dC binds to a limited set of conformational states of TK2 as compared to dT.

Intrinsic fluorescence spectrometry is a sensitive method to study the protein–ligand interaction, where changes in accessibility of tryptophan residues can reveal substrate binding-induced conformational changes.²⁴ Fluorescence quenching caused by the addition of dT and other tested substrates/products showed downward curvatures at low substrate concentrations but linear curves at high concentrations in Stern–Volmer plots. This indicates that at a low substrate concentration, apparently only one of the four tryptophan residues is accessible, and furthermore, the λ_max (330 nm) value suggests that Trp-188 located on the α-helix that formed the lid region is involved in the observed fluorescence quenching.²⁸ These results also demonstrated local as well as global conformational changes of TK2 upon ligand binding.

We also observed a significant difference in the fluorescence intensity (ΔF_max) of the dTTP-free TK2 as compared with that of the dTTP-bound enzyme (Table 3), indicating that there is a large structural (conformational) difference between the dTTP-free and dTTP-bound TK2. We also found a large difference in the equilibrium binding constant for dTTP obtained with ITF (3.64 ± 1.0 μM), which is in the same range as the K_i values (∼2.5 μM) determined by enzyme kinetic assays, but with ITC, the K_d value was much higher (36 μM). The reason for this discrepancy between the two methods is not known, but one explanation could be that a large fraction of the dTTP-free enzyme is partly denatured during the ITC measurements because the dTTP-free enzyme is not as stable as the dTTP-bound TK2.¹⁰ Furthermore, the duration of the ITC measurement is much longer (∼4 h) than the ITF experiment (<1 min). A much higher protein concentration is also used in the ITC experiment, which may lead to increased denaturation of the TK2 protein.

Comparison of Equilibrium Binding Constants with K_M and/or S_1/2 Values

Earlier studies using the radiochemical method to study the steady-state reaction rate of TK2 by measuring the formation of radiolabeled products over a period of time under the conditions that one of the substrates is in excess, the K_M or S_1/2 values were deduced by nonlinear regression fitting of data to the Michaelis–Menten or Hill equation and represent the substrate concentrations at which the reaction rate was 1/2 of the V_max. Therefore K_M or S_1/2 values are not true substrate binding constants. In this study, we measured individual substrates or product binding constants by using two different biophysical methods. The results from the ITC studies showed that the binding constant of dT is 65 times higher than that of dTTP, whereas the binding constant of dC is 12 times higher than that of dTTP. Therefore, both dT and dC are able to displace the enzyme-bound dTTP. The binding constant of dT, which is 5 times higher than that of dC, explained why dT is more efficient in removing the enzyme-bound dTTP, as shown in earlier study.¹⁰ With the dTTP-bound enzyme, the binding affinity of dT is 3 times higher than that of dC, and thus, dT is the preferred substrate (Table 1). Moreover, the binding constants for dT and dC with the dTTP-bound enzyme are in the same range as the K_M values reported earlier.^5,7 The differences in dT and dC binding constants can be explained by the larger enthalpy and entropy changes upon dT binding compared to dC binding, which also suggest that dT binding induced large conformational changes/rearrangements, whereas dC binding induced minor and possibly different changes in the active site conformations (Table 3).

It was demonstrated in this study that ATP does not bind to TK2 in the absence of nucleosides (the phosphate acceptors) using two different methods. However, the presence of ATP did influence the binding affinity of the phosphate acceptor as shown by the ITF method because significantly different binding constants for dC in the absence and presence of ATPγS were observed (Table 3). Furthermore, the binding constants for dT and dC, determined using the ITF method in the presence of ATPγS with the dTTP-bound enzyme, were similar to the K_M values reported earlier using the radiochemical activity assay.^5,7 These results strongly suggest that formation of a catalytically competent ternary complex is the rate-limiting step in TK2 catalysis.

ITC studies using nonhydrolyzable ATP analogue showed a larger change in the enthalpy and entropy values upon binding of ATPγS to the E–dT complex than to the E–dC complex, suggesting that binding of ATPγS to the E–dT complex induced large conformational changes in E–dT than to the E–dC complexes. The differences in ΔG values in ATPγS binding to E–dT as compared to the E–dC complex may explain the higher catalytic efficiency with dT as compared to dC.

Similar to TK2, the crystal structure of TK1 contained the feedback inhibitor dTTP,³³ and in the crystal structures of human dGK and dCK, the feedback inhibitor, that is, dATP, was also present.^27,34 These results suggested that under physiological conditions, these kinases are occupied by their feedback inhibitors, most likely because of the fact that the in vivo dNTP concentrations are normally much higher than deoxynucleoside concentrations.³⁵ Therefore, the overall activities of these nucleoside kinases toward each substrate apparently depend on the concentration of and the ability of the nucleoside substrates to compete with the bound feedback inhibitors, leading to active enzymes. This process probably determines the initial catalytic rates of the deoxynucleoside kinases in vivo.

TK2 Reaction Mechanism—Ordered Sequential Binding of the Substrates and Ordered Product Release

Both ITC and ITF studies provided evidence that substrate binding followed an ordered sequential pathway because ATP did not bind to either dTTP-free or dTTP-bound TK2 but only to the E–dT or E–dC complexes. Therefore, the binding order must proceed with dT or dC binding first to the enzyme and compete out the enzyme-bound dTTP, which is followed by ATP binding to the E–dT or E–dC complexes (Figure 5). ITF studies also showed that both dTMP and dCMP bind tightly to the enzyme, whereas ADP does not, indicating that ADP is leaving the enzyme prior to the release of dTMP or dCMP (Figure 5). Such an ordered reaction mechanism probably also applies to the TK2 family of enzymes because this is also the binding order of human dCK in the ITC experiment (Supporting Information, Figure S1). Earlier studies using the ITC method showed that the Herpes simplex virus type 1 TK, with some structural similarities to the dCK, dGK, and TK2 family of enzymes, has the same binding order, that is, ATP binds only to the E–dT complex.³⁶

Figure 5.

Schematic representation of the TK2 reaction mechanism—ordered sequential binding of the substrates and order release of products. dT binds to either the dTTP-bound or dTTP-free enzyme prior the binding of ATP; phosphorylation occurred after binding of both substrates and products (ADP and dTMP) formed; ADP released before the release of dTMP.

The kinetic mechanism proposed earlier for human dCK and Mycoplasma dAK (deoxyadenosine kinase) based on steady-state kinetics suggested that binding of ATP occurs before the binding of the phosphate acceptors.^37,38 Structural studies by Hazra et al. showed that human dCK adopts two conformational states: an open conformational state for substrate binding and product release and a closed conformational state for catalysis.³⁹ From a 3D structure point of view, it seems more plausible that the nucleosides bind to the enzyme before ATP because the nucleoside binding site would be blocked if ATP binds first.

Physiological Implications

Alterations in TK2 activity or kinetic behavior have been observed in mutant TK2 identified in human patients with MDS. The H121N mutant TK2 has nearly 100% efficiency with dT but only ∼25% with dC, and it has lost the negative cooperativity with dT, which resulted in an alteration in the ratio of dT and dC phosphorylation at physiological relevant concentrations, which apparently led to MDS.^7,14 Additional TK2 mutations found in MDS patients, for example, T77M and R161K¹⁹ and T230A and R225W,¹⁵ also showed altered ratios of dT and dC phosphorylation in addition to the drastically reduced total TK2 activity. Furthermore, alterations of intracellular dT concentration in the case of dT phosphorylase deficiency cause imbalanced dTMP and dCMP production by TK2 because of the difference in kinetics and competition between dT and dC.⁶ The resulting high dTTP and low dCTP concentrations are the likely cause of mtDNA depletion and deletions.⁴⁰ Mouse model studies have shown that a sustained systemic increase in the dC concentration could restore dCTP levels in the mitochondria of mice with defect dT phosphorylase. Furthermore, administration of dC and dT or dTMP and dCMP delayed the disease onset, reduced the severity of the phenotypic manifestation, and prolonged the survival of the TK2-deficient mice.⁴¹⁻⁴³ These results indicate that coadministration of deoxynucleosides could be used to reduce the mitochondrial toxicity caused by antiviral and anticancer therapy.

Thus, detailed knowledge of the structure and function of the TK2 enzyme will most likely increase our capacity to develop personalized treatment options for the severe mitochondrial disorders in the future.

Glossary

Abbreviations

TK2

thymidine kinase 2

dT

thymidine

dC

deoxycytidine

dU

deoxyuridine

ATPγS

adenosine 5′-(gamma-thio) triphosphate

AZT

3′-azido-2′,3′-dideoxythymidine

FLT

3′-fluoro-2′,3′-dideoxythymidine

ITC

isothermal titration calorimetry

ITF

intrinsic tryptophan fluorescence

mtDNA

mitochondrial DNA

MDS

mtDNA depletion syndrome.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.8b01376.

• Representative ITC titration curves and binding isotherms of human recombinant dCK; quenching of ITF upon ligand binding with dTTP-bound and dTTP-free TK2; and F₀/F versus ligand concentration plots for dT, dC, dTMP, dCMP, AZT, FLT, ATP, ADP with dTTP-bound TK2, as well as dTTP, dT, and dC with dTTP-free TK2 (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.