Substrate Specificity of Mitochondrial 2’-Deoxyguanosine Kinase EFFICIENT PHOSPHORYLATION OF 2-CHLORODEOXYADENOSINE*

Mitochondrial deoxyguanosine kinase (dGK) (EC 2.7.1.1 13) was purified to apparent homogeneity from bovine brain. The molecular mass of the native protein was 56 kDa, as judged by gel filtration, and one single band of 28 kDa was seen in sodium dodecyl sulfate-gel electrophoresis. 2’-Deoxyguanosine (dGuo) (K,,,, 7.6 PM), 2’-deoxyinosine, and 2‘-deoxyadenosine (K,,,, 60 PM) were substrates for the enzyme as well as several dGuo analogs containing a lipophilic substituent at C-2‘. were substrates for the enzyme, whereas no 3”modified dGuo were effective. (CdA) to be an efficient substrate for (Km,

2-Chloro-2'-deoxyadenosine (CdA) is a deoxyadenosine analog that has marked therapeutic efficacy in the treatment of leukemia (2, 3). It is generally believed that dCK is the sole CdA-phosphorylating enzyme and that dCK activity is a prerequisite for CdA cell toxicity (4). Human brain tissue lacks dCK, but it has been shown previously that extracts of brain, in spite of this, contain significant CdA phosphorylating activity. These observations suggest the presence of a second CdA-phosphorylating enzyme (5). In the present study we purified dGK from bovine brain to apparent homogeneity and identified dGK as the CdA-phosphorylating enzyme in the brain. We also explored dGK activity in extracts of isolated mitochondria and cytosol fractions from CEM cells. CdA phosphorylating activity, insensitive to inhibition by dCyd, was detected in the mitochondria. This finding is in agreement with our studies on the purified enzyme and shows that CdA is efficiently phosphorylated by dGK in mitochondria. [32P]ATP (5,000 mCi/mmol) was from Amersham Corp. DEAE-Sepharose (fast flow) and CH Sepharose 4B (CNBr activated) were obtained from Pharmacia LKB Biotechnology Inc. Isobutyryl-4,4'-dimethoxytrityl-2'-deoxyguanosine was obtained from Sigma, p-nitrophenylphosphorodichloridate from Aldrich, hydroxylapatite from Clarkson Chemical Co., and unlabeled nucleotides and nucleosides were from Sigma. All reagents were of the highest purity available. All [8-3H]CdA. The reaction was performed as described previously (5). The assay was linear up to 45 min with a detection was performed with 100 pM [Y-~'P]ATP (10 mCi/ml), 50 mM Tris-limit of 1 pmol/min/mg of protein. The phosphoryl transfer assay HCI, pH 7.6, 10 mM MgC12, 100 mM KCl, 10 mM DTT, 0.5 mg/ml bovine serum albumin, 500 ng of dGK, and 100 pM nucleoside in a total volume of 50 pl as described previously (7). Lactate dehydrogenase activity was measured spectrophotometrically at room temperature by determination of NADH oxidation during the initial linear decrease in absorbance at 340 nm upon addition of 10 pg of protein extract to a cuvette containing 60 PM NADH and 300 p~ pyruvate in 50 mM Tris-HC1, pH 7.9.
Purification-The temperature throughout the purification was 0-4 "C. Bovine calf brain (1 kg) was homogenized in 4 liters of 0.3 M sucrose, 2 mM EDTA, 1.0 mM PMSF, 0.1 mM benzamidine, 0.5% Triton X-100, and 2 mM DTT in 20 mM potassium phosphate buffer, pH 7.4. The supernatant was collected after centrifugation for 40 min at 8,000 X g. The supernatant was first filtered through glass wool to remove fat, then 7% (v/v) streptomycin sulfate was added dropwise with stirring for 15 min. The precipitate was removed by centrifugation at 6,000 X g for 10 min. The supernatant was diluted with icecold water and applied to a DEAE column, equilibrated with buffer A (5 mM MgC12, 1 mM DTT, and 15% glycerol in 25 mM Tris-HC1, pH 7.6). A linear gradient of 0-0.3 M KC1 in buffer A was used to elute the enzyme. dGK activity was assayed in the various fractions, and those containing activity were pooled. The dGK pool was diluted 4-fold with ice-cold water and applied to a hydroxylapatite column, equilibrated with buffer B (5 mM MgC12, 15% glycerol, 1 mM DTT, and 1 mM EDTA in 10 mM potassium phosphate buffer, pH 7.6). After washing with 2 column volumes of buffer B, protein was eluted with a linear gradient from 0.01 to 0.15 M potassium phosphate in buffer B. The fractions containing dGK activity were pooled and saved for further use. 60 ml of enzyme from the hydroxylapatite chromatography was dialyzed twice against 1.5 liter of buffer C (1 mM EDTA, 1 mM DTT, 16% glycerol in 50 mM sodium acetate, pH 5.5) and applied to a 4-ml dGuo-Sepharose column equilibrated with buffer C. The column was washed with buffer C until no protein was detected in the eluate. Elution involved three steps: (i) 40 ml of 0.2 M KC1 in buffer C; (ii) 20 ml of 0.5 M KC1 in buffer C; and (iii) 20 ml of 1 mM dGuo in buffer C, containing 0.2 M KC1. From each step, 2ml fractions were collected and were measured for protein content and assayed for dGK activity. To concentrate the enzyme, the pooled fractions from step iii were adjusted to pH 7.7 with 0.3 M NaOH and applied to a 0.5-ml hydroxylapatite column and subsequently eluted with 0.2 M potassium phosphate in buffer B. The purified enzyme gave one single band on SDS-gel electrophoresis (PhastSystem, Pharmacia). The buffer of the protein was changed to 50 mM Tris-HC1, pH 7.5, by Centricon centrifugation (Amicon) for enzyme activity measurements.

Gel Filtration Chromatography
Gel filtration chromatography was performed using fast protein liquid chromatography on a Superose TM12 HR10/30 (10 X 300 mm) column with a single path monitor UV-1 detector (Pharmacia). About 45 pg of purified enzyme was injected to the column which was equilibrated and eluted with 5 mM MgC12, 0.1 M NaCl, 1 mM DTT, and 0.1% Triton X-100 in 50 mM Tris-HC1, pH 7.6, buffer. The flow rate was 0.4 ml/min, and 0.2-ml fractions were collected and assayed for dGK activity. The elution of standard proteins, bovine serum albumin, and ovalbumin was followed by absorbance at 280 nm.

Preparation of Cell and Tissue Extracts
Human tissues were obtained from the Departments of Plastic Surgery and Neurosurgery at the Karolinska Hospital (Stockholm). Bovine brain was from Scanfood (Uppsala). Tissues were homogenized in 0.3 M sucrose, 2 mM EDTA, 1 mM PMSF, 0.1 mM benzamidine, 2 mM DTT, 0.5% Triton X-100, and 20% glycerol in 50 mM Tris-HC1, pH 7.6. After sonication for 10 min, the supernatant was collected by centrifugation for 20 min, 14,500 X g at 4 "C. CEMwtand dCK-deficient CEMddC-50 cells were grown and harvested as described (5). Peripheral blood lymphocytes were separated by Ficoll-Isopaque centrifugation (Pharmacia) and treated as described earlier (8). Protein was extracted from frozen cell pellets (30-60 X lo6 cells/ ml) in 50 mM Tris-HC1, pH 7.6, 2 mM DTT, 5 mM benzamidine, 0.5 mM PMSF, 20% glycerol, and 0.5% Nonidet P-40 by freezing and thawing three times. The mixture was centrifuged for 5 min, 6,000 X g at 4 "C, and the supernatant was used for enzyme assays. Protein was determined by the Bio-Rad protein assay.

Preparation of CEM Cytosolic and Mitochondrial Protein Extracts
Mitochondria were isolated from 200 X IO6 exponentially grown CEMwt cells, using the protocol described by Tapper et al. (9), with the following alterations. The supernatant (13 ml) from the centrifugation (1,500 X g for 15 min) of the nuclei-free cell homogenate was recovered as the fraction containing cytosolic proteins and was upon addition of 150 pl of 0.5 M benzamidine, 150 pl of 50 mM PMSF, 750 p1 of 1.0 M Tris-HC1, pH 7.6, 75 p1 of Nonidet P-40, and 30 p1 of 1.0 M DTT frozen in 7 X 2-ml aliquots, stored at -70 "C until analyzed. The mitochondria were resuspended in 7 ml of 30 mM Tris-HC1, pH 8.0, 0.5 mM EDTA, pH 8.1, and 0.25 M sucrose and were, after collection from the subsequent sucrose gradient, isolated by centrifugation at 1,500 X g for 15 min. Hereafter the mitochondrial pellet was resuspended in 1.5 ml of 10 mM MgCl,, 1 mM EDTA, pH 8.1, 7.5% glycerol, 1 mM DTT, 350 mM NaC1,0.5 mM PMSF, 0.5% Triton X-100, and 10 mM Tris-HC1, pH 8.0. The mitochondria were then disrupted in a small glass homogenizer, and, after centrifugation at 130,000 X g for 60 min, the supernatant, containing extracted mitochondrial proteins, was kept at -70 "C in 6 X 250-pl aliquots until analyzed.

RESULTS
dGK Activity in Extracts of Different Tissues and Cells-Extracts of bovine brain, human brain, human spleen, human skin, CEMwt cells, CEMdCK-cells (lo), and human resting and mitogen-stimulated lymphocytes were examined for dGK activity (Table I). Spleen, CEMwt, cells and lymphocytes contained marked dGuo phosphorylating activity which was inhibited by excess dCyd. However, in human and bovine brain, high dGK activity was found which was insensitive to dCyd inhibition. Therefore this activity was most likely derived from dGK but not dCK. Based on these data, bovine brain was chosen for dGK purification.
Purification and Identification of dGK Activity in Bovine Brain-In the DEAE chromatography of the bovine brain extract, dGuo phosphorylating activity was eluted at 0.15 M KC1. This activity coeluted with CdA phosphorylation but was separated from thymidine and dCyd phosphorylating activity (Fig. 1). The pooled DEAE fractions of the dGuo phosphorylating activity were subsequently purified by hydroxylapatite chromatography and dGuo-Sepharose column chromatography as the final step. After the purification a single band of 28 kDa was detected on SDS-gel electrophoresis (Fig. 2). The yield from 1 kg of brain was 300 pg of protein, with a specific activity of 4.5 nmol/min/mg of protein.
Gel filtration of the pure protein gave an estimated size of the native protein of 56 kDa (Fig. 3).
Substrate Specificity of Pure Bovine dGK-In the final step of the preparation of dGK the pure protein was dissolved in 0.2 M potassium phosphate buffer. Activity was only found  with dGuo and dIno but not with dAdo, when this preparation was assayed directly (yielding a final potassium concentration of 40 mM in the assay). After changing the buffer to 50 mM Tris-HC1, pH 7.5, through Centricon centrifugation, the total activity increased, and dAdo could also be phosphorylated by the enzyme. For all further studies of the enzyme, it was kept in 50 mM Tris-HC1, pH 7.5.
The phosphoryl transferase reaction catalyzed by dGK was examined with several nucleoside analogs as substrates. The structures of the analogs phosphorylated by dGK are shown in Scheme 1 and Table 11. The rate of phosphorylation of the nucleoside analogs was compared with dGuo as the substrate. The capacity of purified dCK to use the same compounds as the substrate were also investigated (Table 111). No phosphorylation was found with 3"modified purine 2"deoxynucleoside analogs.
Nucleoside Kinase Activity in Cytosolic and Mitochondrial Extracts- Table V shows the phosphorylation of CdA (with and without excess of dCyd) and ara-G in mitochondrial and cytosolic extracts from CEMwt cells. Significant CdA phosphorylation was found in extracts of both cellular compart-

TABLE I11
Purine nucleoside analogs as substrate of the pure bovine dGK and recombinant dCK measured by [y-32PlATP phosphoryl transferase assay The value of substrate phosphorylation is given in relation to dGuo phosphorylation by dGK or dCK.  ments, but only that of the cytosol was strongly inhibited by excess of dCyd. The specific activity of the ara-G phosphorylation was 22-fold higher in the mitochondrial extract. Fig. 4 shows the autoradiogram of a phosphoryl transferase assay using dGuo, dIno, dAdo, and CdA as substrates, with purified bovine dGK and mitochondrial extracts from CEMwt cells as the protein source. Pure dGK converted the substrates to four distinct products (5'-monophosphate derivatives) with different retention times, whereas the crude mitochondrial extract seemed to contain adenosine deaminase activity converting dAdo to dIno. The derivatives of CdA phosphorylation by purified dGK or mitochondrial extracts were checked carefully to have the same retention time in the chromatography system used (data not shown).

DISCUSSION
2'-Deoxyguanosine is known to be phosphorylated by two different 2'-deoxyribonucleoside kinases, dCK and dGK (11,12). Assays on cell and tissue extracts, using 2'-dGuo as a substrate, will detect the activity of both enzymes. Inhibition of dCK by excess of dCyd will give a more specific and representative assay for dGK activity when using 2'-dGuo as substrate in a crude extract. Using that approach, we could detect non-dCK-derived dGK activity in spleen, skin, resting and mitogen-stimulated lymphocytes, and in wild type and dCK-negative CEM cells (Table I). Interestingly, a 10-fold higher dGK activity, which was insensitive to dCyd inhibition, was detected in human and bovine brain tissue. This is in agreement with the data of Spasokoukotskaja2 who could not detect any dCK enzyme in brain using Western blots. We purified dGK from bovine brain because of its high dGK activity and the lack of dCK activity. Our purification procedure yielded pure dGK, as judged by SDS-gel electrophoresis. The native dGK appeared to be a dimer with a molecular mass of 58 kDa, which is in accordance with earlier studies of bovine liver dGK (1). This is the first report on the purification and characterization of dGK from brain tissue. dGK has previously been reported to phosphorylate dGuo and dIno (1). In earlier studies, dAdo was reported to be a substrate for dGK, but later this finding was explained by deamination of dAdo to dIno with subsequent phosphorylation of dIno to dIMP (11,13). In this study we clearly show that dAdo is a substrate for purified dGK (Tables I11 and IV). To rule out any contaminating deaminase activity, the product of our enzyme assay was identified by high performance liquid chromatography to be dAMP, not dIMP (data not shown), which is also visualized by the autoradiogram of the * T. Spasokoukotskaja, personal communication. The values in parentheses whoe the fold difference in enzyme activity between the cytosolic and mitochondrial extracts. Presented values are the means of three enzyme activity measurements using the cytosolic and mitochondrial fractions from one cell fractionation.
Lactate dehydrogenase activity is given as units/mg or units/lO' cells, where units represent the decrease of absorbance/min at 340 nm when assayed in room temperature through reduction of pyruvate with concomitant oxidation of NADH. The shown values should not be considered as an absolute determination of lactate dehydrogenase activity in the extracts but rather as a marker for the content of cytosolic proteins in the two fractions. phosphoryl transferase assay (Fig. 4). In contrast with the pure enzyme, the mitochondrial fraction showed deamination of dAdo. dGK has also been purified previously from pig skin (13), neonatal mouse skin (14), and human placenta (15). In all preparations of dGK so far reported the specific activity of this enzyme is quite low. If this is because of the physiological state of the enzyme or a result of the purification procedure is unclear.
Using the purified dGK, we studied the phosphorylation of several antiviral and cytotoxic purine nucleoside analogs (Tables 11-V). Substitutions at the 3' position of purine 2'deoxynucleosides, acyclic purine nucleoside derivatives, and 6-substitutions at the purine moiety other than oxygen or amino are not accepted by dGK. However, dGK accepted the purine bases guanine, hypoxanthine, and adenine when linked to 2'-deoxyribose, several 2"substituted purine 2'-deoxyribosides, and purine arabinosides. Interestingly, the 2-substituted dAdo analog, CdA, is an equally efficient substrate as dAdo for dGK phosphorylation (Table IV). This is a novel finding that may have far reaching implications for the application of this compound.
The only enzyme that is known to activate CdA so far is dCK, as shown by a markedly reduced CdA toxicity against dCK-deficient mutant cell lines and by blocking the cell toxicity of CdA by the addition of excess of dCyd (4). However, the present study shows clearly that a second enzyme, dGK, is able to phosphorylate CdA. In cells that lack dCK, CdA will thus be activated at a level corresponding to the cellular content of dGK. A lower K,,, for CdA phosphorylation by dCK (7) indicates that CdA will be phosphorylated more efficiently by this enzyme if both dCK and dGK are present in the cell. Since we showed that brain tissue contains no dCK activity but high dGK activity (Table I), this tissue may be of high value for CdA toxicity and antitumor studies with regard to diverse types of brain tumors or other tumors containing a high content of mitochondria. Thus, our findings provide a rational basis to consider brain tumors as a potential target for CdA chemotherapy.
Ara-G and ara-Hx were good substrates for dGK, whereas ara-A was less efficiently phosphorylated. Ara-G, an antileukemic compound that is also reported to act as an inhibitor of mitochondrial purine nucleoside phosphorylase (16), is selectively active in T'cell malignancies with low toxicity to other cells in the bone marrow (17). Accumulation of high ara-GTP levels in T cell malignancies compared with non-T cell leukemias has been reported by Shewach and Mitchell (18). They did not find any difference in ara-CTP accumulation in their study and suggested biochemical properties of T lymphoblasts to accumulate ara-GTP. T cells, but not B cells, also accumulate dGTP under purine nucleoside phosphorylase deficiency conditions (19). Since dCK is expressed in both T and B lymphocytes as well as in all lymphoid cells (5,20), our findings may suggest that dGK could be a candidate enzyme to explain the differences of ara-G toxicity in these cells. The high efficiency of dGK for phosphorylation of ara-Hx could also account for drug-related toxicity in cells in which this metabolite is found. Ara-Hx is formed by deamination of ara-A and is also a metabolite of other biologically active purine arabinosides such as ara-DMAP and ara-M (21). Phosphorylation of ara-Hx forms ara-IMP, which can be converted to ara-AMP again. Ara-DMAP and ara-M show low toxicity in cell cultures but are neurotoxic in animals (21). Phosphorylation of ara-Hx by dCK has been suggested as a mechanism of ara-DMAP toxicity (22). However, our data strongly suggest dGK as the responsible enzyme to mediate the neurotoxicity of ara-Hx.
The mitochondrial location of dGK gives rise to some interesting questions. If nucleoside analogs must be activated by dGK to exert their effects on viral or nuclear DNA replication, they need to be transported as their phosphorylated derivatives from the mitochondria to the cytosol compartment. It is so far unclear if such a transport system exists. As a matter of fact, an accumulation of the corresponding 5'triphosphate derivative of such nucleoside analogs may occur in the mitochondrial compartment. Starnes and Cheng (23) have shown that the mitochondrial DNA polymerase y has a broader substrate acceptance for modified nucleotide analogs than the nuclear DNA polymerases. This may lead to a toxicity of the test compound which is mediated through inhibition of mitochondrial DNA replication (24). Thus, in this regard the possibility of phosphorylation of nucleoside analogs by mitochondrial dGK should be kept in mind as a potential cause of side effects of certain drugs.
In conclusion we believe that dGK-catalyzed phosphorylation may play an important role in the evaluation and rational development of new antiviral and cytotoxic nucleoside analogs. The expression of dGK in normal and malignant cells and tissues should be explored more in terms of targeted chemotherapy.