Mutagenesis of non-conserved active site residues improves the activity and narrows the specificity of human thymidine kinase 2

Abstract

Human thymidine kinase 2 (TK2) is critical for the nucleotide salvage pathway and phosphorylation of nucleoside analog prodrugs in vivo; however, it remains poorly studied because of difficulties expressing it heterologously. TK2 is strictly pyrimidine-specific, whereas its phylogenetic relative, the Drosophila melanogaster deoxyribonucleoside kinase (DmdNK), shows higher activity and broader specificity towards both pyrimidines and purines. These differences are counterintuitive, as only two of 29 active site residues differ in the two enzymes: F80 and M118 in DmdNK are L78 and L116 in TK2. In addition to reporting an optimized protocol for the expression and purification of TK2, we have used site-directed mutagenesis to introduce the DmdNK-like amino acids into TK2, and characterized the three resulting enzymes (L78F-TK2, L116M-TK2 and L78F/L116M-TK2). These mutations improve the K_M for thymidine, increasing the catalytic activity of L78F/L116M-TK2 4.4-fold, yet leave activity for deoxycytidine or the purine nucleosides unchanged.

The salvage of 2’-deoxyribonucleosides from intracellular or extracellular degradation of DNA is essential for providing deoxyribonucleoside triphosphates for nuclear DNA repair and mitochondrial DNA synthesis. Deoxyribonucleoside kinases (dNKs) are key enzymes of the salvage pathway, catalyzing the initial phosphoryl transfer from a suitable donor (typically ATP) to a 2’-deoxyribonucleoside, forming a 2’-deoxyribonucleoside monophosphate [1]. In the treatment of several cancers and viral diseases, this reaction is the rate-limiting step in activating nucleoside analog prodrugs to their cytotoxic, phosphorylated forms [2]. This central role of dNKs has lead to interest in investigating the determinants of their substrate specificity, to facilitate engineering novel dNKs that can phosphorylate nucleoside analogs more effectively.

Human mitochondrial thymidine kinase 2 (TK2) is a pyrimidine-specific dNK, phosphorylating thymidine (dThd) and deoxycytidine (dCyd) [3, 4]. It is constitutively expressed throughout the cell cycle [2], making it critical for the phosphorylation of pyrimidine analog prodrugs such as AZT (3’-azido-2’,3’-dideoxythymidine), FIAU (1-(2’-deoxy-2’-fluoro-β-D-arabinofuranosyl)-5-iodouracil) and AraT (1-β-D-arabinofuranosylthymine), that are routinely used in antiviral and cancer chemotherapy [3, 5]. Mutations in the TK2 gene have also been linked to mitochondrial DNA depletion syndrome [6]. Despite the clinical importance of TK2, there is little structural or functional information on this enzyme; unlike other members of the dNK family, TK2 has not been engineered for improved nucleoside analog activation or altered substrate specificity. This may reflect difficulties in expressing TK2 heterologously, as the protein tends to aggregate under standard overexpression conditions [7]. These problems have also hampered efforts to determine the structure of the enzyme.

In this study, we have developed an improved protocol for heterologous expression and purification of recombinant TK2, and used site-directed mutagenesis to explore the roles of nonconserved active site residues on the structure/function relationship of the enzyme. Aligning the sequence of TK2 with its closest phylogenetic relatives, the dNKs from Drosophila melanogaster (DmdNK), Anopheles gambiae (AgdNK) and Bombyx mori (BmdNK), reveals remarkable conservation of the active site residues, with the four enzymes only differing in two positions (Fig. 1). Unlike TK2, however, the insect enzymes are each able to phosphorylate all four native 2’-deoxyribonucleosides [8–10]. We hypothesized that the different substrate specificities of these enzymes may be due to the differences in the non-conserved amino acid positions. We attempted to engineer TK2 into an enzyme with DmdNK-like broad substrate specificity and high catalytic efficiency, by using site-directed mutagenesis to replace L78 and L116 of TK2 with the corresponding residues in DmdNK (F80 and M118). We have characterized the intermediate enzymes L78F-TK2 and L116M-TK2, as well as the double mutant, L78F/L116M-TK2. Our results show that the DmdNK-like mutations do not broaden the specificity of TK2; however, they do improve activity towards dThd, and affect the quaternary structure of the enzyme.

Fig. 1.

Structure-based multiple sequence alignment of TK2, DmdNK, AgdNK and BmdNK computed by 3DCoffee [24]. Numbering is according to TK2. Active site residues, as determined from the structure of DmdNK [13], are highlighted in gray. Residues targeted for mutagenesis (L78, L116) are underlined and bold.

Materials and methods

Materials

Enzymes were from New England Biolabs (Beverly, MA) unless otherwise indicated. Pfu Turbo DNA polymerase (Strategene, La Jolla, CA) was used for all cloning. Pyruvate kinase and lactate dehydrogenase were from Roche Biochemicals (Indianapolis, IN). Gel filtration standards were from BioRad (Hercules, CA). All other reagents were from Sigma-Aldrich (St. Louis, MO) unless noted.

Molecular Modeling

The entire TK2 sequence was submitted for First Approach Mode to the SwissModel server [11]. Residues 9-207 of TK2 were able to be modeled from the structures of DmdNK complexed with the substrates dThd (PDB code 1OT3 [12]) and dCyd (PDB code 1J90 [13]).

Cloning and mutagenesis

The genes encoding TK2 and DmdNK were isolated previously [14]. The first 19 codons of TK2 were removed by PCR, with a new initiator methionine codon introduced in place of K19 (numbering according to [4]) using primers TK2-F and TK2-R (Table 1). This truncated form of TK2 has identical kinetic properties to the full-length enzyme [3]. Primers DM-F and DM-R were used for PCR amplification of the full-length DmdNK gene. The TK2 gene was corrected for codon usage in Escherichia coli using overlap extension PCR with primers CO1-CO10 (Table 1); the rare arginine codons AGA and AGG were mutated to CGC at amino acid positions 79, 93, 99, 107, 129, 151, 155 and 157. This codon-optimized variant was used as the “wild-type”. The L78F and L116M mutations were introduced by overlap extension PCR, using primers L78F-F and L78F-R, and L116M-F and L116M-R, respectively (Table 1). All genes were subcloned via NdeI and SpeI restriction sites (introduced by the primers) into pET-14b (Novagen, Madison, WI) for overexpression. Constructs were confirmed by DNA sequencing.

Table 1.

Primers used for plasmid construction and sequencing

Name	Sequence
TK2-F	5'-CAGCCATATGTCAGTGATCTGTGTCG-3'
TK2-R	5'-GCGACTAGTTCATGGGCAATGCTT-3'
DM-F	5’-CTATCATATGGCGGAGGCAGCATCCTGTG-3’
DM-R	5’-CTCACTAGTTCATCTGGCGACCCTCTG-3’
CO1	5'-CATGCTGGACCGCCATACTCGTCCTCAGGTGTC-3'
CO2	5'-GACGAGTATGGCGGTCCAGCATGGTGAGCTGCA-3'
CO3	5'-ATGGAGCGCTCGATTCACAGCGCACGCTACATTTTTGTAGAAAAC-3'
CO4	5'-GCGCTGTGAATCGAGCGCTCCATCAACCGTACAGATG-3'
CO5	5'-AAACCTGTATCGCAGTGGGAAGATGCCAGAAGT-3'
CO6	5'-TCTTCCCACTGCGATACAGGTTTTCTACAAAAA-3'
CO7	5'-CTGGATCTTGCGCAACATGGACGTGTCTGTTGA-3'
CO8	5'-CGTCCATGTTGCGCAAGATCCAGTCAAACCATT-3'
CO9	5'-CCAGCGCTTAAAGAAGCGCTGCCGCGAAGAGGAGAAGGTCATTCC-3'
CO10	5'-GCAGCGCTTCTTTAAGCGCTGGTAACAAGTCTCAGGAT-3'
L116M-F	5'-GTAGAAAACATGTATCGCAGTGGGAAGAT-3'
L116M-R	5'-CACTGCGATACATGTTTTCTACAA-3'
L78F-F	5'-TCTTACGTTCCAGACTTATGTGCAGCTC-3'
L78F-R	5'-CTGCACATAAGTCTGGAACGTAAGAC-3'

Restriction sites are italicized. Mutations are underlined.

Expression and purification

E. coli BL21(DE3) pLysS cells (Novagen) were transformed with plasmids containing the wild-type and mutated genes. Cells were grown at 37°C in LB broth containing carbenicillin (100 μg/ml), chloramphenicol (34 μg/ml) and ethanol (3%, v/v), to A₆₀₀ ~0.6. Cultures were shifted to 22°C, isopropyl-β-D-thiogalactopyranoside was added (0.4 mM, final concentration) and cells were grown for 12 h. DmdNK was overexpressed as described previously [14]. Cells were harvested by centrifugation and the pellets stored at −20°C. For protein purification, cell pellets were resuspended in buffer A (50 mM potassium phosphate, 300 mM NaCl, 1 mM β-mercaptoethanol, 10% glycerol, pH 8.0) with protease inhibitor cocktail, rLysozyme (Novagen) and Benzonase nuclease (Novagen) added according to manufacturers’ protocols. Bacteria were lysed by sonication on ice, and the lysates clarified by centrifugation (4500g, 4°C, 30 min) before being loaded onto Talon metal affinity resin (Clontech, Mountain View, CA). After two wash steps with 5 column volumes of buffer A containing 5 mM and 15 mM imidazole, the target proteins were eluted in buffer A containing 150 mM imidazole. Amicon-Ultra centrifugal filter units (Amicon Bioseparations, Billercia, MA) were used to exchange the purified proteins into either buffer B (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) for biophysical characterization, or storage buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM β-mercaptoethanol, 10% glycerol, pH 8.0). Protein concentrations were quantified by measuring A₂₈₀ (TK2, ε = 38,690 M⁻¹.cm⁻¹; DmdNK ε = 39,880 M⁻¹.cm⁻¹, calculated according to Gill and von Hippel [15]). Aliquots were stored at −80°C.

Enzyme Assays

Nucleoside phosphorylation activity was determined using a spectrophotometric coupled-enzyme assay [10, 16]. Activity was measured at 37°C, in a 500 μl reaction mixture containing 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 2.5 mM MgCl₂, 0.18 mM NADH, 0.21 mM phosphoenolpyruvate, 1 mM ATP, 1 mM 1,4-dithio-DL-threitol, 30 units/ml pyruvate kinase and 33 units/ml lactate dehydrogenase. Substrate concentrations ranged from 1-3000 μM. Measurements were made in triplicate and corrected for background. Apparent k_cat (calculated assuming one active site per enzyme monomer) and K_M values were determined by fitting data to the Michaelis-Menten equation, using non-linear regression analysis in Origin (OriginLab, Northhampton, MA).

Biophysical characterization

Far-UV circular dichroism (CD) spectra were obtained using a J-810 spectropolarimeter (Jasco, Inc., Easton, MD). Spectra were recorded at 22°C from 260-190 nm (0.5 nm increments) using a 0.1 mm pathlength cell, 20 nm/min scan rate, 4 s response time, and 2 nm bandwidth. Proteins were analyzed at concentrations of 0.9–1.2 mg/ml. Recorded data represent the means of eight scans. Spectra were corrected for buffer absorbance and converted to mean residue ellipticity ([θ]_mrw).

Intrinsic tryptophan fluorescence emission spectra were measured using a FluoroMax-3 spectrophotometer (Horiba Jobin Yvon, Edison, NJ). Protein samples were excited at 295 nm in a 1 cm pathlength cuvette. Emission data were accumulated at 20°C, from 305–405 nm, in 1 nm increments, using 5 nm excitation and emission bandwidths. Protein concentrations were 3.2–3.6 μM. Each spectrum represents the mean of three scans, normalized for protein concentration and corrected for contribution of buffer to the signal.

Oligomeric states of the purified enzymes were determined by size-exclusion chromatography using a Superdex 200 10/300 GL column and AKTApurifier (both Amersham Biosciences). The column was equilibrated in buffer B and calibrated with the standard proteins thyroglobulin, bovine γ-globulin, chicken ovalbumin and equine myoglobin. Blue dextran and acetone were used to determine the void volume (V_o = 8.00 ml) and total exclusion volume (V_t = 20.75 ml) of the column, respectively. The partition coefficients of the standards, K_av = (V_e−V_o)/(V_t−V_o), where V_e is the elution volume, were linearly related to log molecular mass. Approximately 100 μl of each purified protein (1 mg/ml) was loaded onto the column, and eluted at a flow rate of 0.5 ml/min.

Results and discussion

Rational redesign of the TK2 active site

Our sequence alignment (Fig. 1) suggested that L78 and L116 were the only two active site residues that are not identical in TK2 and the multi-substrate insect kinases. This was confirmed when we constructed a homology model of TK2 using the structure of DmdNK [13] as a template (Fig. 2A). Overall, we identified 29 amino acid residues within the active site; 27 of these are identical in TK2 and DmdNK. Consistent with previous models [7, 13], L78 and L116 in TK2 are in structurally equivalent positions to F80 and M118 of DmdNK. It has been proposed that the broad specificity of DmdNK is due to its large, hydrophobic active site [13]. Interestingly, our model predicts that the active site of TK2 (with the two leucines replacing a phenylalanine and a methionine) is even larger and more hydrophobic. Based on these findings, we used site-directed mutagenesis to produce the L78F and L116M single-site mutants, as well as the L78F/L116M double mutant.

Fig. 2.

Comparison of TK2 and DmdNK conserved residues and structure. (A) Homology model of TK2 monomeric subunit. Active site residues are shown as sticks and colored green. L78F and L116M are yellow. (B) Overlay of TK2 homology model and DmdNK with dThd (PBD code 1OT3 [12]) or (C) dCyd (PDB code 1J90 [13]) in the binding pocket. DmdNK and TK2 residues are colored grey and yellow, respectively.

Expression and Purification

Expression and purification of TK2 in sufficient quantity and quality is critical for enzyme engineering, yet has been problematic. Previously reported protocols have suffered from low yields [3] or required complex, multi-step expression and purification [7]. We therefore sought a robust, simple and high yield protocol to prepare TK2 for in vitro characterization.

First, we optimized codon usage in the human gene by replacing nine AGG and AGA arginine codons with the preferred E. coli codon CGC. This codon-optimized variant of TK2 served as the wild-type control and as the template for site-directed mutagenesis. We next attempted to follow the protocol of Wang et al. [3], as the authors have shown that an N-terminal (His)₆-tag does not affect activity and therefore would not need to be proteolytically cleaved (in turn, obviating the need for further purification). However, this approach yielded negligible amounts of soluble enzyme. Instead, we found that using the pET system with a reduced induction temperature and ethanol added to the medium (to induce a heat shock-like response in the bacterial host [7, 17–19]) increased the yield of soluble, purified TK2 to an average of 11 mg per liter culture. This yield is similar to that reported by Barroso et al. (10 mg/L; [7]), but was obtained with a simpler protocol, requiring fewer purification steps. Our improved protocol was used to purify L78F-TK2, L116M-TK2 and L78F/L116M-TK2 to >95% purity.

Kinetics

To examine the effects of the mutations on the activity and substrate specificity of TK2, each enzyme was evaluated for its ability to phosphorylate the four natural 2’-deoxyribonucleosides (Table 2).

Table 2.

Kinetic parameters of wild-type and mutant deoxyribonucleoside kinases

	dThd			dCyd			dAdo	dGuo
	kcat (s−1)	KM (μM)	kcat/KM (M−1s−1)	kcat (s−1)	KM (μM)	kcat/KM (M−1s−1)	kcat/KM (M−1s−1)	kcat/KM (M−1s−1)
DmdNKa	17.5 ± 0.6	1.9 ± 0.3	9.1 x 106	19.3 ± 0.5	2.2 ± 0.3	9.0 x 106	8.3 x 104	1.8 x 104
TK2	1.8 ± 0.2	6.7 ± 1.1	2.7 x 105	2.0 ± 0.2	16.2 ± 1.1	1.2 x 105	n.d.	n.d.
L78F-TK2	2.5 ± 0.2	3.7 ± 0.6	6.7 x 105	1.7 ± 0.1	7.8 ± 0.6	2.2 x 105	n.d.	n.d.
L116M-TK2	1.1 ± 0.1	4.9 ± 1.2	2.3 x 105	1.0 ± 0.1	36.1 ± 4.9	2.7 x 104	n.d.	n.d.
L78F/L116M-TK2	1.3 ± 0.1	1.1 ± 0.3	1.2 x 106	1.7 ± 0.1	14.9 ± 1.5	1.1 x 105	n.d.	n.d.

Values represent the mean of three experiments. n.d. = not detectable.

^aThe kinetic parameters of DmdNK have been determined previously using a radiometric assay [10]; our results using the coupled-enzyme spectrophotometric assay are in agreement.

The catalytic efficiency (k_cat/K_M) of L78F-TK2 was improved 2.5-fold for dThd, and 1.8-fold for dCyd (Table 2). The L78F mutation introduces the favorable base stacking interaction observed in DmdNK [13] and consequently improves the Michaelis constants for dThd and dCyd. We expected that this mutation would also have equal effects on the turnover numbers for each pyrimidine, since the new aromatic side chain is angled towards the edge of the substrate that is identical in dThd and dCyd (i.e. C2 and N3). However, the mutation improves the k_cat for dThd and worsens it slightly for dCyd, suggesting that this single-site mutation causes structural rearrangements that are not strictly paralleled in DmdNK.

Replacement of L116 with methionine had little effect on the kinetics of dThd phosphorylation. In contrast, L116M-TK2 retained only 23% of dCyd phosphorylation activity relative to wild-type TK2; the k_cat and K_M each decreased approximately two-fold. Examination of the DmdNK structures with dThd and dCyd reveals that upon binding dCyd, the side chain of Q81 rotates, forming hydrogen bonds with the nucleobase, while the side chain of M118 must move 2.8 Å to avoid a steric clash with the amino group of Q81 (Q79 in TK2; Fig. 2B and C). In DmdNK, this conformational change is accommodated by a small hydrophilic residue (S123) located behind the methionine. TK2 has a lysine residue at the equivalent position (K121), which may interfere with the analogous conformational rearrangement in L116M-TK2, due to its size and charge.

The double mutant, L78F/L116M-TK2, retains wild-type activity towards dCyd, suggesting that the favorable stacking interaction introduced by the L78F mutation and the steric constraints of the L116M mutation offset one another. For dThd, the combination of the two mutations is synergistic, raising the catalytic performance almost 5-fold over wild-type TK2. L78F/L116M-TK2 has a K_M that is improved >6-fold, and now matches the binding affinity of DmdNK for dThd (Table 2). These two substitutions, which flank Q79, are apparently crucial in optimizing the active site geometry for dThd binding.

Interestingly, the activity of mutant enzymes towards purine nucleosides remained below the detection limit of the assay (k_cat/K_M < 10² M⁻¹s⁻¹). The L78F substitution generates a TK2 active site that approximates that of BmdNK (with the exception that L116 in TK2 is an isoleucine in BmdNK), yet it fails to introduce purine activity at levels comparable to the insect kinase [9]. The double mutant L78F/L116M-TK2 has all 29 active site residues identical to DmdNK, completely recreating the active site of DmdNK on an amino acid level, yet no improvement in purine activity was detected in the engineered enzyme. The combined effect of the L78F and L116M mutations was to narrow the substrate specificity of TK2, making it almost 5-fold more specific for dThd. This change in specificity results from an improved catalytic efficiency for dThd, rather than a decrease in activity towards dCyd (Fig. 3).

Fig 3.

Summary of catalytic efficiencies.

Biophysical characterization

The wild-type and mutant TK2 enzymes were characterized by CD spectroscopy, fluorescence spectroscopy and size-exclusion chromatography (Fig. 4) to probe the effects of the mutations on the secondary, tertiary and quaternary structure of the enzymes.

Fig. 4.

Biophysical characterization of wild-type and mutant TK2 enzymes. (A) Far-UV CD spectra. (B) Intrinsic tryptophan fluorescence spectra. (C) Size-exclusion chromatography profiles.

The far-UV CD spectra of all four proteins were essentially identical (Fig. 4A), indicating that the changes in specificity and activity were not brought about by changes in secondary structure. We also probed for changes in tertiary structure, by measuring intrinsic tryptophan fluorescence (Fig. 4B) to detect perturbations in the environment of the five tryptophan side chains in TK2 (W55, W74, W133, W136 and W189). Three of these (W74, W133 and W156) are located in the dimerization interface, W55 stacks with the base of the substrate in the enzyme active site and W188 is located at the end of α-helix 8. The emission spectra of TK2 and L116M-TK2 are identical. In contrast, both enzymes with the L78F mutation exhibit a decrease in fluorescence intensity of 15%, and a 1 nm blue shift in emission maxima. The small changes observed for L78F-TK2 and L78F/L116M-TK2 are most likely due to introduction of an aromatic side chain (L78F) in close proximity to the active site tryptophan (W55).

Size-exclusion chromatography was used to examine possible effects of the mutations on quaternary structure. Untagged wild-type TK2 is known to exist predominantly as a dimer (~70%) in solution, though with tetrameric, hexameric and higher-molecular weight forms also present [7, 20]. Similarly, (His)₆-tagged TK2 eluted in two peaks (Fig. 4C), corresponding to estimated molecular masses of 91.1 kDa (24% total protein) and 44.7 kDa (76% total protein). These match the expected masses of compact tetrameric and dimeric forms, respectively (monomer M_r = 27 600). In contrast, all three mutant enzymes showed a shift in the dimer-tetramer equilibrium; in each case the fraction of enzyme in the dimeric form increased to 90–97% of total protein with only a small shoulder in the elution profile indicating the presence of the homotetramer (Fig. 4C). Although the dimer-tetramer ratio has no effect on catalytic activity [20], the observed equilibrium shift suggests that the active site mutations impact enzyme topology through long-range conformational changes.

Summary

We have generated a mutant, L78F/L116M-TK2, that is 4.4-fold more efficient than wild-type TK2 at phosphorylating dThd, making it the first reported example of an improved TK2 variant. We anticipate that this more active variant will be an ideal starting point for engineering kinases with improved activity towards nucleoside analogs. Our improved protocol for protein expression and purification will also facilitate future engineering, in addition to enabling structural studies.

Surprisingly, L78F/L116M-TK2 showed no change in catalytic efficiency with dCyd, nor any detectable activity towards purine nucleosides. Rather than broadening specificity as hypothesized, mutagenesis at positions 78 and 116 of TK2 actually narrowed it. It has generally been assumed that active site residues determine the substrate specificity in dNKs [13, 21–23]. Here, we have provided direct evidence that active site residues are not the sole determinant of the narrow specificity of TK2, but rather that more distant residues must play equally important roles. This demonstrates the limited ability of rational design to predict subtle changes, including those that are mediated by long-range interactions within the structure of the protein. Other probes of protein structure/function relationships, such as homologous or non-homologous recombination, will be required to search beyond the active sites of dNKs for determinants of their activity and specificity.