Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
T. Skovgaard, NNF Center for Protein Research, University of Copenhagen, Blegdamsvej 3B, 2200 Copenhagen N, Denmark Fax: +45 35325001 Tel: +45 35325052 E-mail: email@example.com
The first step for the intracellular retention of several anticancer or antiviral nucleoside analogues is the addition of a phosphate group catalysed by a deoxyribonucleoside kinase such as thymidine kinase 1 (TK1). Recently, human TK1 (HuTK1) has been crystallized and characterized using different ligands. To improve our understanding of TK1 substrate specificity, we performed a detailed, mutation-based comparative structure–function study of the active sites of two thymidine kinases: HuTK1 and Caenorhabditis elegans TK1 (CeTK1). Specifically, mutations were introduced into the hydrophobic pocket surrounding the substrate base. In CeTK1, some of these mutations led to increased activity with deoxycytidine and deoxyguanosine, two unusual substrates for TK1-like kinases. In HuTK1, mutation of T163 to S resulted in a kinase with a 140-fold lower Km for the antiviral nucleoside analogue 3’-azido-3′-deoxythymidine (AZT) compared with the natural substrate thymidine. The crystal structure of the T163S-mutated HuTK1 reveals a less ordered conformation of the ligand thymidine triphosphate compared with the wild-type structure but the cause of the changed specificity towards AZT is not obvious. Based on its highly increased AZT activity relative to thymidine activity this TK1 mutant could be suitable for suicide gene therapy.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Mammals have four different deoxyribonucleoside kinases (dNKs) with overlapping substrate specificities. Two are cytosolic – thymidine kinase 1 (TK1) and deoxycytidine kinase (dCK) – and two are mitochondrial – thymidine kinase 2 (TK2) and deoxyguanosine kinase (dGK). All four are encoded by nuclear genes. Based on sequence and structure analyses, the evolutionary origin of TK1 has been found to be different from that of TK2, dCK and dGK and consequently the kinases can be divided into TK1-like and non-TK1-like groups [1–3]. In addition to the mammalian TK2, dCK and dGK, the non-TK1-like group includes insect dNKs and herpes simplex virus type 1 thymidine kinase. Vaccinia virus thymidine kinase belongs to the TK1-like group [1,4]. TK1 is a key enzyme in the salvage of thymidine for DNA synthesis. It catalyses the first of three phosphoryl transfers from ATP to thymidine (dThd), following which thymidine monophosphate (dTMP) is quickly phosphorylated to thymidine diphosphate (dTDP) and thymidine triphosphate (dTTP) by monophosphate and diphosphate kinases. The negative charge imparted by the first phosphate traps thymidine (and its analogues) in the cells, and given the passive transport of thymidine across cell membranes, TK1, in effect, serves as a thymidine pump.
The end product of the thymidine salvage pathway, dTTP, is a feedback inhibitor of TK1, but it is also an allosteric effector of ribonucleotide reductase where it shifts the enzyme’s specificity from pyrimidine to purine nucleotide reduction . Hence, an imbalance in the dTTP pool can cause an imbalance in both the purine and pyrimidine pools and thus a disturbance in DNA metabolism and repair.
Apart from phosphorylation of naturally occurring deoxyribonucleosides, dNKs are also important for the activation of anticancer and antiviral nucleoside analogue pro-drugs such as 1-β-d-arabinofuranosylcytosine (cytarabine, AraC) and 3′-azido-3′-deoxythymidine (zidovudine, AZT). TK1, in particular, is known for its activation of AZT, the cancer-diagnostic positron emission tomography imaging agent 3′-[(18)F]fluoro-3′-deoxythymidine  and the anticancer experimental radiosensitizer 5-iodo-2′-deoxyuridine (idoxuridine) . Cellular uptake of nucleoside analogues is generally rate-limited by the first phosphorylation step catalysed by dNKs, and the toxic triphosphorylated nucleoside analogues can cause DNA-chain termination, inhibit DNA polymerase and induce apoptosis [8,9]. Here, AZT is an exception as the human thymidine monophosphate kinase has very limited capacity to phophorylate AZT monophosphate, thereby creating a bottleneck in the activation of this analogue .
Similarly to insects that have only one multisubstrate dNK, Caenorhabditis elegans also has only one. However, the C. elegans thymidine kinase 1 (CeTK1) belongs to the TK1-like family and has no homology to the multisubstrate insect dNKs or to the other non-TK1-like kinases. Like human TK1 (HuTK1), CeTK1 has dThd as its preferred substrate, but it also displays a low capacity to phosphorylate deoxyguanosine (dGuo) .
With the exception of the N- and C-terminal regions, CeTK1 is similar in sequence to HuTK1, with 43% amino acid identity and 63% similarity . Despite their similarity, CeTK1 is a dimer, also in the presence of ATP, as determined by gel filtration at concentrations of 0.25 μm or lower. In contrast, HuTK1, at the same concentrations, occurs as a tetramer when eluted in the presence of ATP and a dimer in the absence of ATP. At enzyme concentrations higher than 8 μm, HuTK1 is a tetramer irrespective of the presence of ATP. The HuTK1 tetramer has a low Km with dThd (0.5 μm) and the dimer has a high Km (15–17 μm) but both forms have the same kcat [11–13].
The structure of HuTK1 has been shown to differ significantly from other human dNKs. It has been crystallized with the feedback inhibitor dTTP [3,14] and with the bisubstrate inhibitor P1-(5′-adenosyl)P4-(5′-(2′-deoxy-thymidyl)) tetraphosphate (TP4A) . The base and sugar moieties of dTTP is buried in the interior of the protein between the two domains, while the phosphates extend into the p-loop of the α/β-domain. In the non-TK1 like dNKs, the thymine base is stabilized by different side-chain interactions, and the variability of the side chains lead to different substrate specificities in the non-TK1-like dNKs [16,17]. In HuTK1 the thymine base is stabilized by main-chain interactions along the edge of the ring, explaining the narrow substrate range of TK1-like enzymes.
The multisubstrate dNK found in Drosophila melanogaster (DmdNK) has been the subject of intensive mutation studies. Random mutagenesis has led to a mutant kinase with increased specificity towards the antiviral nucleoside analogue AZT [18,19], and site-directed mutagenesis has revealed details about the catalytic function and substrate specificity of DmdNK [20,21]. Here we present the first active-site structure–function study of TK1-type enzymes.
A detailed comparison of the active site of the two TK1 enzymes presented in this study will provide an increased understanding of the substrate specificity of TK1 that could provide a platform for designing new and more efficient nucleoside analogues. It also opens up possibilities for creating new mutated HuTK1 enzymes with increased nucleoside analogue specificity that could be used in suicide gene therapy. The advantage of using a mutated human kinase over foreign kinases, such as herpes simplex virus type 1 thymidine kinase or DmdNK, is avoidance of the antigenic response to foreign protein expression.
Results and Discussion
Comparison of the active sites of CeTK1 and HuTK1 revealed that the majority of amino acids surrounding the deoxyribonucleoside are identical in both enzymes (Fig. 1). One exception was T163 in HuTK1 which is S182 in CeTK1. Thymidine kinase from vaccinia virus also have an S in this position, and is known to accommodate substrates that cannot be phosphorylated by HuTK1 . Considering this, and the fact that CeTK1 can phosphorylate dGuo, the significance of T163/S182 and some of the nearby amino acids was investigated in order to elucidate their effect on substrate specificity.
In HuTK1, T163 is part of an otherwise hydrophobic pocket around the 5-methyl group of thymine. Apart from T163, the 5-methyl group is surrounded by L124, Y181 and M28. The latter is also in close proximity to the 5′-oxygen of the ribose moiety where it stabilizes the position of the oxygen, forcing it into the proximity of E98. The E98 carboxy group is believed to act as the base that abstracts a proton from the deoxyribose 5′-OH, leading the 5′-oxygen to act as a nucleophile in the phosphor transfer reaction [3,14].
In this study we attempted to direct the substrate specificity from dThd towards other deoxyribonucleosides and nucleoside analogues by mutating selected amino acids near the substrate base in HuTK1 and CeTK1. The amino acids T163, M28 and L124 in HuTK1 and the corresponding amino acids S182, M42 and L143 in CeTK1 were mutated by site-directed mutagenesis. In previous studies either 40 or 41 C-terminal amino acids were removed in order to facilitate crystallization of HuTK1 [3,14]. Removal of 41 amino acids (HuTK1-CΔ41, named HuTK1-CΔ40 in ) has also been shown to increase the stability of the kinase compared with the nontruncated enzyme . However, a similar C-terminal truncation in CeTK1 has been proven to destabilize this enzyme  and therefore all HuTK1 mutants were made on HuTK1-CΔ41 and all CeTK1 mutants were made on the wild-type CeTK1.
The mutated TK1 genes were expressed in Escherichia coli as glutathione S-transferase (GST)-tagged enzymes and were purified by GST affinity chromatography. The purified enzymes were tested for activity with 100 μm of the natural substrates dThd, dGuo, deoxycytidine (dCyd) and deoxyadenosine (dAdo) and the two nucleoside analogues AZT and AraC. Kinetic parameters were determined for all enzyme/substrate combinations unless the activity was < 3000-fold of the activity with dThd. None of the kinases displayed activity with AraC but a few showed low activity with dGuo and dCyd (Table 1). One promising candidate, HuTK1-T163S, which showed improved specificity towards the nucleoside analogue AZT, compared with dThd, was crystallized in order to understand the structural basis of the AZT specificity. Crystallization with dThd and AZT was also attempted but yielded no diffracting crystals. However, for visualization, dThd and AZT were docked into the HuTK1-T163S structure (Fig. 2).
Table 1. Kinetic parameters. Kinetic parameters are based on initial velocity measurements and are presented with standard deviation from the fitting. kcat is determined using the theoretical molecular weight of one monomer. ND, not detectable.
a From a previous publication . b The catalytic efficiency, kcat/Km, was estimated from the slope of a v0 versus [S] plot at [S] ≪ Km. c Parameters are based on a single measurement.
1.4 ± 0.6
26600 ± 3080
9.5 ± 1.1
6.8 × 106
0.52 ± 0.1
10000 ± 324
3.5 ± 0.11
6.7 × 106
349 ± 29
4180 ± 155
1.5 ± 0.05
4.2 × 103
254 ± 42
675 ± 31
0.24 ± 0.01
317 ± 54
4032 ± 236
1.4 ± 0.08
4.5 × 103
204 ± 41
955 ± 50
0.34 ± 0.02
1.7 × 103
36 ± 3.9
5210 ± 175
1.8 ± 0.06
5.1 × 104
33 ± 3.7
4060 ± 142
1.4 ± 0.05
4.4 × 104
59 ± 14
7720 ± 655
2.7 ± 0.23
4.6 × 104
0.43 ± 0.1
3740 ± 291
1.3 ± 0.10
3.1 × 106
2.3 ± 0.2
12700 ± 256
6.4 ± 0.13
2.8 × 106
12 ± 1.8
18600 ± 495
9.3 ± 0.25
7.8 × 105
565 ± 94
5590 ± 410
2.8 ± 0.21
5.0 × 103
3.6 ± 0.8
0.34 ± 0.02
1.7 × 10−4 ± 8.2 × 10−6
2318 ± 460
1610 ± 209
0.81 ± 0.10
336 ± 76
9780 ± 723
4.9 ± 0.36
1.5 × 104
1.3 ± 0.2
1930 ± 67.6
0.97 ± 0.03
7.5 × 105
18 ± 2.9
3470 ± 152
1.5 ± 0.08
9.8 × 104
74 ± 13
1470 ± 111
0.74 ± 0.06
1.0 × 104
406 ± 51
1160 ± 49
0.58 ± 0.02
1.4 × 103
Mutants of HuTK1-M28 and CeTK1-M42
M28 in HuTK1, corresponding to M42 in CeTK1, is positioned at the side of the thymine ring with the tip of the amino-acid side chain above the 2′- and 3′-carbons of the deoxyribose and the sulfur pointing towards the six-carbon atom of thymine at a distance of 3.8 Å. The mutation M28I results in an increased space above the deoxyribose but only a minor increase in space near the thymine base. The short branch of I extends towards the β-phosphate of dTTP in the structure and may influence the transfer of phosphate between ATP and dThd (Fig. 1). In both HuTK1 and CeTK1 the mutation from M to I had a negative impact on enzyme activity with dThd as the substrate, affecting both the Km and the kcat. The Km was increased by 250-fold for both mutated enzymes, relative to the wild-type enzyme, and the kcat was reduced by six- and two-fold for mutated HuTK1 and CeTK1, respectively, compared with the wild-type enzymes (Table 1). A lowered catalytic efficiency (kcat/Km) with dThd is not unexpected for HuTK1-M28I and CeTK1-M42I because this particular methionine is conserved in all TK1-like proteins and is believed to constrain the 5′-hydroxyl of thymidine into position near the catalytic glutamate (E98 in HuTK1 and E117 in CeTK1) . A spatial increase in this region could cause the 5′-hydroxyl group of ribose to move away from the putative catalytic residue. With M28I the effect on the Km was high, whereas the effect on the kcat was low.
The catalytic efficiency with AZT was reduced even more for the isoleucine mutants compared with the two wild-type enzymes. This is probably because isoleucine is positioned just above the 3′-azido group in AZT, as illustrated in the docking of AZT into the HuTK1-T163S structure (Fig. 2). The impact on the kcat was a 15-fold decrease for HuTK1 and an 11-fold decrease for CeTK1, and for the Km the impact was an increase of about 500-fold for HuTK1 and an increase of approximately 200-fold for CeTK1. Overall, the effect on catalytic efficiency with AZT is a reduction of approximately 2000–7000-fold for both of the mutated enzymes.
HuTK1-M28I had no activity with dCyd but surprisingly CeTK1-M42I was found to be active with this substrate, even though the activity was fairly low. The kinetic parameters showed a very low kcat but also a Km that was 160- and 640-fold lower than those with dThd and AZT substrates, respectively.
The mutation of M28 to A in HuTK1 results in a larger open pocket above the ribose, which can explain the decreased enzymatic activity with both dThd and AZT because the 5′-hydroxyl of the substrate sugar is not constrained into position near the catalytic glutamate (E98). The effect of the M28A mutation is a large increase in the Km and a small decrease of the kcat, similar to the effect observed for M28I.
Mutants of HuTK1-L124 and CeTK1-L143
In HuTK1, the residue L124 (which corresponds to L143 in CeTK1) is positioned near M28 at a distance 3.7 Å from the six-carbon atom of thymine. It is located on the α /β-domain side above the thymine ring plane, 3.8 Å from the ring oxygen in the sugar moiety. Mutation of L124 to A in HuTK1 results in an opening of the space above the thymine and the deoxyribose of dTTP (Fig. 1). The mutation led to a five-fold decrease in kcat and a 26-fold increase in Km with dThd as the substrate, resulting in a 130-fold reduction in catalytic efficiency compared with the wild-type enzyme. The kcat for AZT was only slightly affected by the mutation, but the Km was increased by more than 60-fold; hence, overall, the catalytic efficiency with AZT is lower in the L124A mutant than in the wild-type enzyme.
Likewise, L143 in CeTK1 was mutated to A. This gave a slightly larger space around the thymine ring compared with HuTK1-L124A as a result of the presence of S182 in CeTK1 (T163 in HuTK1) (Fig. 1). With dThd as the substrate, the presence of A instead of L resulted in a 190-fold decrease in the kcat/Km because of a 150-fold increase in the Km and a 1.3-fold decrease in the kcat. The CeTK1-L143A mutant displayed a decrease of over 1500-fold in catalytic efficiency with AZT as the substrate compared with the wild-type enzyme. Also here, the effect apparently is on the Km because complete saturation could not be reached. The kcat/Km was therefore estimated from the linear slope of initial velocity versus substrate concentration at low concentrations of substrate because the slope approached kcat/Km when [S] << Km. Why the L to A mutation had such a drastic impact on AZT phosphorylation is unclear because the azido group of AZT is located at the opposite side of the active site from amino acid L143 (see the corresponding position of L124 in HuTK1 compared with the azido group in Fig. 2). It is likely that the change from the large and bulky L to the small A created a space that was too large for proper positioning of the substrate and/or induced minor changes in other amino acids in the active site.
Furthermore, changes may have occurred around positions 4 and 5 of the substrate base as CeTK1-L143A gained the ability to phosphorylate dCyd, which differs from dThd in those positions. The kcat/Km with dCyd as substrate was very low compared to that with the substrates dThd and AZT, however. None of the HuTK1 mutants were able to phosphorylate dCyd: this could be a result of the increased space near the 5-methyl of thymine that is occupied by an S in CeTK1 (and by the slightly larger T in HuTK1), in combination with the space created by the L143A or M42I mutations. The increased space in both sections of the active site leaves room for a slight shift in substrate position that could provide space for the amino group of cytosine. Such a shift could also explain the poor activity with dCyd because it would probably affect the position of the 5′-oxygen that receives the phosphoryl group upon catalysis. M42 and L143 (M28 and L124 in HuTK1) are both in position to keep the 5′-oxygen in close proximity to the catalytic glutamate (E98 in HuTK1) but mutation of those residues to shorter amino acids will leave the oxygen more flexible. In order to fit dCyd into the active site, the 5′-oxygen will probably be moved further away from the catalytic glutamate, resulting in poor catalysis, which is reflected by the low catalytic efficiency with dCyd for CeTK1-M42I and CeTK1-L143A.
Mutants of CeTK1-S182
In the modelled structure of CeTK1, S182 is positioned on the side of the base ring plane, pointing Cβ towards the 5-methyl group of thymine at a distance of 3.5 Å. In HuTK1, the amino acid in the corresponding location is the slightly larger T163 (Fig. 1). As this is the only amino acid around the thymidine-binding site that differs between the two enzymes, S182 in CeTK1 was mutated to T in order to mimic the HuTK1 active site. Mutating S to T reduced the space around the 5-methyl group of thymine. In order to elucidate further the effect of increased space in this area, S was also mutated to the smaller residue A.
The S182T mutation resulted in a small decrease in the Km for the substrate dThd, from 2.3 μm in wild-type CeTK1 to 1.3 μm for the S182T mutant. However, the kcat for the S182T mutant was decreased further (by about seven-fold), resulting in a catalytic efficiency that was three- to four-fold lower than that of the wild-type CeTK1.
With AZT as substrate the catalytic efficiency was also reduced in the S182T mutant compared with the wild-type enzyme. In this case, the eight-fold decrease in kcat/Km was mainly caused by a six-fold decrease in kcat.
The S182T mutant is able to phosphorylate dGuo with higher catalytic efficiency than the wild-type CeTK1. Owing to the limited solubility of dGuo and to high Km values of both enzymes, it was not possible to obtain saturating dGuo concentrations. Therefore, kcat/Km was estimated as a single kinetic parameter from the linear slope at low concentrations of dGuo (Table 1).
S182 was also mutated to A, and this yielded an enzyme that had even lower activity with dThd than S182T. The catalytic efficiency of the S182A mutant was reduced by nearly 300-fold compared with the wild-type enzyme, but, unlike S182T, this reduction was caused not only by a nine-fold decrease in kcat but also by a 30-fold increase in Km. Taken together, the S182A mutation appears to have a large impact on Km, whereas the S182T mutation mainly affects the kcat for the substrate dThd. For AZT the S182T mutation mainly affects the kcat, whereas the S182A mutation negatively affects both kinetic parameters for AZT with a 30-fold increase in Km and a 16-fold decrease in kcat.
CeTK1-S182A also phosphorylates dGuo but with a catalytic efficiency lower than for the wild-type CeTK1 and for the S182T mutant. Hence, the additional space created by introducing the small amino acid A did not improve the activity with dGuo. However, a reduction of space resulting from the introduction of T instead of S gave rise to a small increase in catalytic efficiency with dGuo, making S182T the best dGuo phosphorylating TK1 presented here.
The loss of activity of the S182A mutant with all substrates is most likely to be caused by the loss of a hydrogen bond between S182 and the main-chain nitrogen of M42 (Fig. 1). The loss of this hydrogen bond may destabilize the binding pocket and lead to poorer binding, as reflected by the large impact on all kinetic parameters for the S182A mutant.
Activity parameters and crystal structure of HuTK1-T163S
Mutation of T163 to S in HuTK1 mimics the active site of CeTK1 because the amino acid composition around the phosphate acceptor in CeTK1 differs only from HuTK1 by having an S in this position (S182). T163S displayed a decrease in activity with the substrate dThd compared with wild-type HuTK1, reflected by a four-fold decrease in kcat and a large, 40-fold increase in Km. With AZT the T163S-kinase displayed a very low Km (of only 0.43 μm), and the kcat was reduced by only two- to three-fold compared with the wild-type enzyme, and the overall catalytic efficiency for AZT was thus decreased by only two-fold. Most importantly, what distinguishes this mutant from all the other kinases presented is the AZT/dThd catalytic efficiency ratio. As the nucleoside analogue competes with dThd inside cells, this ratio is of major importance to the efficiency of the thymidine kinase as a suicide enzyme when used in conjunction with AZT. T163S was the only enzyme that displayed a higher kcat/Km with the nucleoside analogue AZT than with dThd. For the wild-type kinase the catalytic efficiency was nearly the same for AZT and dThd but for T163S the AZT/dThd ratio was 70. The increased activity was a result of the 140-fold lower Km with the substrate AZT compared with the substrate dThd.
It was attempted to crystallize the C-terminally truncated HuTK1-T163S in complex with the feedback inhibitor dTTP and also with dThd and the nucleoside analogue AZT, but only crystals in the presence of dTTP were obtained. The structure of HuTK1-T163S/dTTP was solved with X-ray crystallography to 2.2 Å resolution, with R and Rfree being 16.7% and 21.8%, respectively (Table 2) [Protein Data Bank identification (PDB ID): 2WVJ). Like the previously crystallized HuTK1 (PDB ID: 1XBT) the T163S crystals contain two tetramers in the asymmetrical unit but the unit cell dimensions are different as a result of a shift in the position of the two tetramers relative to one another. Owing to flexible parts, none of the subunits display electron density before amino acid 18 and after amino acid 192. Density is also absent for the loop between amino acids 64 and 73. These are the same regions lacking electron density in the 1XBT structure of HuTK1.
Table 2. Data collection and refinement statistics. Coordinates and structure factors have been deposited in the Protein Data Bank with accession number 2WVJ.
a Data were collected at 100 K. Values in parentheses in column 2 are for the outermost shell.
Source for data collectiona
Space group and unit cell
Content of the asymmetric unit
Solvent content (%)
No. of observed reflections
No. of unique reflections
Bond length (Å)
Bond angle (°)
Mean B value (Å2)
Superposition of the T163S mutant and the wild-type HuTK1 structures revealed rmsd values of 0.21 Å2 and 0.26 Å2 over 642 Cα for the tetramers consisting of polypeptide chains A–D and E–H, respectively. Hence, the overall structural fold is not affected by the T163S mutation. The position of the ligand, dTTP, is different in the presented structure, however. In the wild-type structure published by Welin et al. the three phosphates form an arch with a Mg2+ ion positioned beside it (PDB ID: 1XBT) . In another HuTK1 wild-type structure (PDB ID: 1W4R) published shortly thereafter, two different conformations of dTTP were present with an occupation ratio of 60/40 in all binding sites . The dominant binding mode in that structure is identical to the dTTP binding described in the paper by Welin et al. .
This structure of HuTK1-T163S also displays two different conformations of dTTP. One is the conformation reported by Welin et al. and the other has the β-phosphate in place of the magnesium ion and a water molecule in place of the β-phosphate (Fig. 3A). In the present T163S structure, the occupation ratio is 75/25. In three of the sites the dominant mode is the same as in the structure reported by Welin et al. , and a magnesium ion has been assigned to this conformation. In the remaining five subunits the situation is reversed but no magnesium ion could be assigned for this mode. In the three subunits containing magnesium in the active site, the position of R60 is stable and the amino acid has hydrogen bonds to the 5′-oxygen and to an oxygen atom in the β-phosphate. In the remaining five subunits the same R60 is very flexible and has no hydrogen bonds to dTTP (Fig. 3A).
By superposing the wild-type and T163S structures, it is evident that the replacement of T163 with S does not have any major impact on the position of the amino acid in the active site compared with the wild-type protein but only leaves a little more room at the 5-methyl group of the thymine ring and some small changes in amino acids around the phosphates (Fig. 3B). Only one amino acid, R60, appears to have slightly different orientations in the two structures, but that is probably caused by a more disordered binding of dTTP in the T163S-mutant structure.
The primary reason for the improved AZT/dThd ratio in catalytic efficiency for T163S is the 40-fold increase in Km for dThd. There is no strong evidence from the structure as to what causes this increased Km because dTTP binds very similarly in the wild type and the T163S-mutant structures. It is, however, possible that the change from T to the slightly smaller S can affect the positioning of dThd, even if it is not visible in the structure containing dTTP. The positioning of dTTP is stabilized by a network of hydrogen bonds around the phosphates, which are not present upon binding of dThd. This gives a higher degree of freedom for dThd binding and hence it is affected more easily by small amino acid changes in the hydrophobic pocket around the base. However, the dockings of dThd and AZT into the HuTK1-T163S structure revealed that the position of the thymine ring of the two substrates superimpose (Fig. 2) and they are also in good agreement with the thymine ring of dTTP in the crystal structure. The main differences are to be found in the side chains near the sugar moiety, where the flexible R60 moves away to provide space for the bulky azido group in AZT. The distance from the azido group to R60 is 3.0 Å and it forms hydrogen bonds to both the carboxyl oxygens in D58. Minor differences are also observed for D58, L124 and the catalytic E98. These differences in amino-acid orientations around the sugar moiety can easily explain the difference in affinity for the two substrates, but how they are affected by the T163S mutation situated near the thymine ring is not entirely clear.
With limited dGuo phosphorylation for CeTK1 and no measurable phorphorylation of dCyd and dAdo, C. elegans is not able to salvage all four deoxyribonucleosides and, similarly to fungi, must rely mainly on the de novo pathway for deoxyribonucleotide synthesis. TK1 may be particularly important in the embryonal growth phase to provide a fast supply of dTTP, which is known to be a key regulator of the ribonucleotide reductase .
It is noteworthy that HuTK1-T163S, which has the same amino acids in the active site as the wild-type CeTK1, shows no dGuo activity. This indicates that dGuo phosphorylation in CeTK1 is caused by interactions with amino acids that do not directly surround the substrate base. Probably, P153 in CeTK1 (G134 in HuTK1) changes the hydrophobic pocket surrounding the thymine base by a twisting of F152 (F133 in HuTK1), which is involved in sandwiching the thymine ring.
This difference in the surroundings of the binding pocket could also explain why CeTK1 and HuTK1-T163S behave differently with respect to dThd and AZT affinity.
In conclusion, the majority of the mutations introduced around the hydrophobic pocket surrounding the thymine base have a negative impact on enzyme activity with both dThd and AZT, but some also give activity with dCyd or dGuo.
Mutation of M28/42 to the more bulky I in HuTK1/CeTK1 has a strong, negative impact on AZT phosphorylation, which can be explained by its proximity to the 3′-azido group. In CeTK1 the mutations M42I or L143A result in activity with dCyd, but the same mutations in HuTK1 (M28I and L124A) do not give rise to dCyd activity. dCyd and dThd differ around the fourth and fifth positions of the base, which is near S182 in CeTK1 and T163 in HuTK1 (Fig. 1). It is likely that dCyd can only bind to the two CeTK1 mutants because the amino acid near the fifth position of the base is the smaller S182. In the CeTK1 mutants M42I and L143A, this leaves room for a shift in substrate position, providing space for the amino group of dCyd.
The wild-type CeTK1 displays activity with dGuo, but HuTK1-T163S, which has the same amino acids in the active site as the wild-type CeTK1, shows no dGuo activity. The same is found for CeTK1-S182T, which has the same amino acids in the active site as the wild-type HuTK1. This indicates that amino acids not in direct contact with the substrate are responsible for dGuo phosphorylation. It is likely that the explanation is to be found in P153 in CeTK1 (G134 in HuTK1) and that this residue may affect the surroundings of the substrate base by a twisting of an F, making stacking interactions with the base.
HuTK1 has nearly the same catalytic efficiency with dThd and AZT substrates but the mutation T163S results in a drastic decrease in the Km for AZT compared with dThd and consequently in a 70-fold increase in the AZT/dThd ratio of catalytic efficiency. This ratio is partly caused by a decreased activity with dThd in T163S. The structure of the T163S mutant shows a dTTP that is bound in a less-ordered manner than in the wild-type structure, but because of the extensive network of hydrogen bonds between dTTP and the active-site residues, the minor T to S change has no visible effect on dTTP binding. dThd is smaller and binds with more degrees of freedom and is therefore more likely be influenced by the T163 to S mutation. No major differences are observed around the thymine ring when docking dThd and AZT into the structure of T163S, however, but the dockings do show differences in the amino-acid side chains around the sugar moiety, which could explain the difference in catalytic efficiency between dThd and AZT.
Materials and methods
Mutagenesis and sequence verification
Mutations in HuTK1 and CeTK1 were performed on cDNA inserted in pGEX-2T expression plasmids. Construction of the expression plasmids pGEX-2T-CeTK1 and pGEX-2T-HuTK1-CΔ41 has been described previously [3,11,27]. Mutations were introduced by site-directed mutagenesis using forward and reverse primers (designated F and R, respectively). For all CeTK1 mutants, the following primers were used with the template pGEX-2T-CeTK1: S182T-F, 5′-gc gga tcg caa gca aac ttc aca ttc cgc agc-3′; S182T-R, 5′-gct gcg gaa tgt gaa gtt tgc ttg cga tcc gc-3′; S182A-F, 5′-gc gga tcg caa gca aac ttc gca ttc cgc agc-3′; S182A-R, 5′-gct gcg gaa tgc gaa gtt tgc ttg cga tcc gc-3′; L143A-F, 5′-gct gca gcc aat gga aca ttc gag aga aag ccg ttc c-3′; L143A-R, 5′-g gaa cgg ctt tct ctc gaa tgt tcc att ggc tgc agc-3′; M42I-F, 5′-g ggg cca att ttc agt ggc aaa acc acc g-3′; and M42I-R, 5′-c ggt ggt ttt gcc act gaa aat tgg ccc c-3′. Mutated nucleotides are represented in bold and underlined. The template pGEX-2T-HuTK1-CΔ41, encoding a 41-residue C-terminal truncated version of HuTK1, was used for all HuTK1 mutants in combination with the following primers: T163S-F, 5′-gg gaa gcc gcc tat agc aag agg ctc gg-3′; T163S-R, 5′-cc gag cct ctt gct ata ggc ggc ttc cc-3′; T163A-F, 5′-gg gaa gcc gcc tat gcc aag agg ctc gg-3′; T163A-R, 5′-cc gag cct ctt ggc ata ggc ggc ttc cc-3′; L124A-F, 5′-cc gta att gtg gct gca gcg gat ggg acc ttc c-3′; L124A-R, 5′-g gaa ggt ccc atc cgc tgc agc cac aat tac gg-3′; M28I-F, 5′-g gtg att ctc ggg ccg atc ttc tca gga aaa agc-3′; M28I-R, 5′-gct ttt tcc tga gaa gat cgg ccc gag aat cac c-3′; M28A-F, 5′-g gtg att ctc ggg ccg gcg ttc tca gga aaa agc-3′; and M28A-R, 5′-gct ttt tcc tga gaa cgc cgg ccc gag aat cac c-3′.
The mutated plasmids were transformed into XL10-Gold Supercompetent Cells and isolated, and the mutations were verified by sequencing at MWG Biotech (Martinsried, Germany).
Protein expression and purification
After sequence verification the plasmids were transformed into the expression host E. coli BL21, and following induction with isopropyl thio-β-d-galactoside (IPTG) the bacteria were harvested when the thymidine kinase activity per mg of total protein was at its highest. After expression, the cells were disrupted in a French press, and the proteins were purified by GST affinity chromatography with thrombin cleavage, as previously described . Purification quality was determined by SDS/PAGE, Experion Automated Gel electrophoresis and the Bradford method .
HuTK1-CΔ41-T163S was purified on a large scale for crystallography and was subjected to additional purification steps. After GSH chromatography the protein was desalted on a Sephadex G25 column with elution buffer (20 mm Tris, pH 7.5, 5 mm MgCl2, 5 mm NaF, 10% glycerol, 0.5 mm Chaps, 50 mmε-amino caproic acid and 2 mm dithiothreitol) and purified on a carboxymethyl sepharose cation-exchange column. The column was equilibrated with 10 mm potassium phosphate, pH 6.0, 5 mm MgCl2, 10% glycerol, 0.1% Triton X-100 and 2 mm dithiothreitol, and the protein was eluted with the same buffer, in which 10 mm potassium phosphate, pH 6.0, was substituted with 50 mm potassium phosphate, pH 8.0.
Enzyme activity assays
Kinase activities were determined with tritium-labelled substrates (Moravek, Brea, CA, USA) in the DE-81 filter paper assay . Initial velocities were determined by three or four time samples at 10 different substrate concentrations, and unless stated otherwise the assays were repeated independently between two and five times. The standard assay conditions were: 50 mm Tris/HCl, pH 7.5, 10 mm dithiothreitol, 2.5 mm MgCl2, 0.5 mm Chaps, 3 mm NaF, 0.5 mg·mL−1 of BSA and 2.5 mm ATP. The enzyme concentrations are listed in Table S1. The assay detection limit varied with the amount of radioactive substrate used. In the dCyd assay with CeTK1-M42I the dCyd was 3.6 Ci per mmol and the assay detection limit was lower than 0.05 pmol·min−1 (the kcat in this assay, 1.7 × 10−4 s−1, corresponds to 0.44 pmol·min−1). In the dGuo assay with wild-type CeTK1, the dGuo was 0.18 Ci per mmol and the assay detection limit was 1 pmol·min−1 (the data points in this assay ranged from 1.2 pmol·min−1 at the lowest substrate concentration to 214 pmol·min−1 at the highest).
The results were analyzed using graphpad prism by nonlinear regression to the Michaelis–Menten equation (v = Vmax·[S]/(Km + [S]) or the Hill equation (v = Vmax·[S]n/(K0.5n + [S]n). In all cases the Michaelis–Menten equation resulted in the best fit. For enzymes with very high Km values, complete saturation could not be obtained and the kcat/Km was estimated from the linear slope of initial velocity versus substrate concentration at low concentrations of substrate. When [S] ≪ Km the slope approaches kcat/Km.
Crystals of HuTK1-CΔ41-T163S were obtained at 15 °C by hanging drop vapour diffusion after 3–5 days. The drops contained 2 μL of protein solution (10 mg·mL−1 in 50 mm, pH 8.0, 2 mm dithiothreitol, 10% glycerol, 0.1% Triton X-100, 5 mm MgCl2 and 5 mm dTTP) and 2 μL of precipitant (0.1 m sodium citrate (pH 5.6), 20% 2-propanol and 17% polyethylene glycol 4000). Before flash-freezing in liquid nitrogen, the crystals were soaked in cryoprotectant consisting of precipitant containing 20% glycerol. Attempts were also made to crystallize HuTK1-CΔ41-T163S with dThd and AZT, but no crystals were obtained.
The data set for HuTK1-CΔ41-T163S in complex with dTTP was collected at a temperature of 100 K at beamline ID23-1 at ESRF (Grenoble, France). Data were indexed, scaled and merged using MOSFLM  and SCALA .
Structure determination and refinement
The structure was solved by molecular replacement with the program Molrep  using HuTK1-CΔ41 as a search model (PDB: 1XBT) The model with waters was adjusted manually to the 2Fo − Fc electron density in Coot  and refined with Refmac5 . Data collection and refinement statistics are shown in Table 2. Coordinates and structure factors have been deposited in the PDB with accession number 2WVJ.
Docking of dThd and AZT into the HuTK1-CΔ41-T163S structure was performed with Molsoft v. 3.7. Prior to docking, AZT was subjected to a full conformation search and the azido group was found to have three-fold symmetry, with the lowest energy conformation identical to that identified previously by crystallography . dTTP was removed from chain D in the structure of HuTK1- CΔ41-T163S (PDB ID: 2wvj) followed by full flexible docking of dThd and AZT. The conformation of AZT in the best scored docking pose is essentially identical to the lowest energy conformation and the conformation reported previously . The best scored docking poses were subsequently optimized using a combined Monte Carlo and minimization procedure (using the MMFF94 force field), keeping the ligand and the surrounding protein residues (in a 6-Å radius from the starting position) flexible. All backbone coordinates were held fixed.
We thank Nils Egil Mikkelsen and Martin Welin for technical assistance with solving the crystal structure and Thomas Frimurer for assisting with the docking. This work was supported by grants from the Danish Research Council [274-05-0081].