Gamma-phosphate-substituted 2'-deoxynucleoside 5'-triphosphates as substrates for DNA polymerases.

Several 2′-deoxythymidine 5′-triphosphate and 3′-azido-2′,3′-dideoxythymidine 5′-triphosphate analogs containing a hydrophobic phosphonate group instead of the γ-phosphate were synthesized and evaluated as substrates for human immunodeficiency virus (HIV) and avian myeloblastosis virus reverse transcriptases, human placental DNA polymerases α and β, and calf thymus terminal deoxynucleotidyl transferase. They were efficiently incorporated into the DNA chain by the retroviral enzymes but were not utilized by the mammalian ones. Also, some γ-ester and γ-amide derivatives of dTTP and 3′-azido-2′,3′-dideoxythymidine 5′-triphosphate (AZTTP) were synthesized and studied. They proved to be substrates for both the retroviral and mammalian enzymes under study. The Km values for incorporation of the dTTP derivatives into the DNA chain were close to those for dTTP and AZTTP. The Km for the AZTTP derivatives were one order of magnitude greater than those for dTTP and AZTTP. The results obtained indicate that HIV and avian myeloblastosis virus reverse transcriptases have no sterical obstacles for binding the triphosphate fragment bearing a bulky substituent at the γ-position. Modification of the γ-phosphate in AZTTP increased the selectivity of HIV reverse transcriptase inhibition versus DNA polymerase α. γ-Methylphosphonate and γ-phenylphosphonate were dephosphorylated in human serum much less rapidly than AZTTP. Besides, they were shown to be markedly more hydrophobic than AZTTP. Thus, replacement of the γ-phosphate in AZTTP with γ-phosphonate markedly alters its substrate properties toward some cellular DNA polymerases and blood dephosphorylating enzymes but does not change its substrate activity with respect to HIV reverse transcriptase.

The ␥-phosphate seems to play an important role in binding of dNTP to the enzyme ϩ template-primer complex. Indeed, 2Ј-deoxynucleoside 5Ј-diphosphates display a hundredfold lower affinity to Escherichia coli DNA polymerase (1) and human placental DNA polymerase ␣ (2, 3) as compared with dNTP. X-ray analysis of rat liver DNA polymerase ␤ complexed with a template, terminated primer, and 2Ј,3Ј-dideoxycytidine 5Ј-triphosphate revealed that the ␥-phosphate appears to in-teract with residues Asp-190, Asp-192, Arg-149, and Gly-189 of the enzyme (4). In the molecule of HIV reverse transcriptase, Asp-110, Asp-185, and Asp-186 are likely to interact with the ␥-phosphate of dNTP (5). Residue Asp-882 of E. coli DNA polymerase I is a catalytically important residue (6). According to the sequence, it is equivalent to Asp-190 of DNA polymerase ␤ (5).
The process of substrate binding to E. coli DNA polymerase I has been earlier studied (7). It has been assumed that the preformed Mg 2ϩ -dATP complex binds to the enzyme as a ␤,␥bidentate. A chemical mechanism has been proposed for substrate binding to DNA polymerases (8). At the first step, dNTP binds to the enzyme and interacts with the enzyme-bound divalent cation only via its ␥-phosphate. This complex has been detected by NMR spectroscopy (9). At the second verification step, only if a proper Watson-Crick base pair can form between the substrate and the corresponding residue of the template, the ␤-phosphate of dNTP also coordinates to the enzyme-bound metal cation and thus ensures the distance between the primer 3Ј terminus and ␣-phosphate of dNTP appropriate for nucleophilic substitution at the phosphorus atom. It is noteworthy that the regions surrounding the polymerase active site correspond to motifs that are conserved in DNA polymerases (10,11). Therefore, it seems likely that different DNA polymerases bind dNTPs by similar mechanisms.
Here we report the synthesis, substrate properties, and stability in human blood serum of ␥-methyl and ␥-phenylphosphonate diphosphates of 2Ј-deoxythymidine Ia and Ib and 3Јazido-2Ј,3Ј-dideoxythymidine IIa and IIb and compare their properties with those of ␥-phosphate-substituted nucleotides Ic, Id, IIc, and IId (Structure 1).

MATERIALS AND METHODS
Methylphosphonic and phenylphosphonic dichlorides and phenyl phosphate were from Aldrich. 2Ј-Deoxynucleoside 5Ј-diphosphates were synthesized according to Ref. 16.
For DNA polymerase assays, the samples of I and II were purified by HPLC on a Daltosil C-18 column (150 ϫ 4 mm, 4 m); elution was with a linear gradient of 0 -25% methanol in 0.05 M KH 2 PO 4 (pH 6.0) with UV detection at 270 nm. The flow rate was 0.5 ml/min.
Thin-layer chromatography was carried out on Kieselgel UV-254 (Merck) plates using the following system (v/v): dioxane-NH 4 OH-water (6:1:4). For column chromatography, Dowex 50W4 (H ϩ , 200 -400 mesh, Sigma), DEAE-Toyopearl 650 M (Toyosoda), and LiChroprep RP-18 siliconized Silica gel (40 -63 m, Merck) were used. 1 H NMR spectra were registered in a Varian AMX400 spectrometer (400 MHz) with tert-BuOH as the inner standard. 31 P NMR spectra were recorded at 162 MHz using 85% H 3 PO 4 as the external standard. In the 1 H NMR spectra, the proton signals for the compounds Ic, Id, IIc, and IId coincided with those for compounds Ia, Ib, IIa, and IIb. Fast atom bombardment mass spectra were recorded on a Kratos MS 50TC spectrometer. The reaction yield was estimated by measuring the optical density of the resulting aqueous solutions of I and II (extinction coefficient, 9800).

Enzymes and DNA
HIV reverse transcriptase was isolated according to Ref. 17. DNA polymerases ␣ and ␤ were isolated from human placenta as described in Refs. 18 and 19. AMV reverse transcriptase and terminal deoxynucleotidyl transferase were from Omutninsk Chemicals (Omutninsk, Russia) and Amersham Corp., respectively.

Primer Extension Assays
For the template-dependent DNA polymerases, the assay mixture (volume 6 l) contained 0.01 M template-primer (Scheme 1), compound under study or dTTP, enzyme (2 activity units of reverse transcriptases or 1 unit of DNA polymerases ␣ and ␤), and the corresponding buffer. The reaction was carried out for 20 min at 37°C and terminated by adding 3 l of deionized formamide containing 0.5 mM EDTA and 2% bromphenol blue and xylene cyanol. The reaction products were separated by electrophoresis in 20% polyacrylamide gel, and the gels obtained were radioautographed.
Kinetic measurements (22) were performed within the linear region of the product formation versus time curve. The reaction time was 2 min.
DNA synthesis inhibition assays were carried out as described in Ref. 23

Hydrolysis of Compounds IIa, IIb, IIc, IId, and AZTTP in Human Blood Serum
The assay mixture containing 2.5 l of 10 mM solution of IIa-d or AZTTP and 47.5 l of 100% fetal blood serum was incubated at 37°C, mixed with 50 l of water and 230 l of methanol, and cooled for 30 min at Ϫ20°C. The samples were centrifuged for 10 min at 12,000 rpm, and the supernatants were concentrated to 100 l and analyzed by HPLC. The extent of hydrolysis was assessed by measuring the amount of the starting compound.

RESULTS
Compounds Ia, Ib, IIa, and IIb were synthesized at high yields using the standard method of phosphate activation by 1,2,4-triazole (24). It includes interaction of methyl-or phenylphosphonic bis- (1,2,4-triazolides) with 5Ј-diphosphates of 2Ј-deoxythymidine and 3Ј-azido-2Ј,3Ј-dideoxythymidine. Compounds Ic and IIc were prepared as described in Ref. 25. Compounds Id and IId were prepared by one-pot synthesis using the procedure described in Ref. 16. The corresponding derivatives of ATP, dATP, UTP, and GTP have earlier been synthesized by reacting aniline with trimetaphosphate prepared in situ from nucleoside triphosphates (26). The 31 P NMR data are presented in Table I.
Compounds I and II were evaluated as substrates for HIV and AMV reverse transcriptases, human placental DNA polymerases ␣ and ␤, and calf thymus terminal deoxynucleotidyl transferase. The results of HIV reverse transcriptase assays are shown in Figs. 1 and 2.
Compounds Ib-d were proved to be substrates for HIV reverse transcriptase at 2 M; they were incorporated into the growing DNA chains one (Fig. 1, lanes 3, 7, and 11), two (lanes 4 and 5, 8 and 9, and 12 and 13), and up to one hundred (lanes 6, 10, and 14) times. Similar results were obtained for Ia (data not shown). The presence of octadecanucleotide bands on lanes 4, 8, and 12 is due to the error prone properties of the enzyme.
All modified nucleoside 5Ј-triphosphates under study were utilized by AMV reverse transcriptase (data not shown). Table  II lists the kinetic parameters for incorporation of I and II into the DNA chain by AMV reverse transcriptase. Clearly, K m for I are close to the values for dTTP and AZTTP, whereas K m (II) are 1 order of magnitude greater than K m for dTTP and AZTTP. The V max values are close for compounds studied.
It can be seen in Fig. 3 that phenylphosphonate Ib was a very poor substrate for DNA polymerase ␣ (Fig. 3, lanes 3 and  4), whereas Ic (Fig. 3, lanes 5 and 6) and Id (Fig. 3, lanes 7 and  8) were efficiently utilized by this enzyme. Methylphosphonate Ia (Fig. 3, lanes 9 and 10) was a slightly better substrate than its phenyl counterpart Ib. DNA polymerase ␤ did not recognize Ib as a substrate even at 80 M (Fig. 4, lanes 6 -9) and utilized Ic only to a small extent (Fig. 4, lanes 13-19). In the control assays (Fig. 4, lanes 2-5), only natural dNTPs were used. At the same time, the enzyme incorporated Id and, less efficiently, Ia into the DNA chain (Fig. 5).
It is evident from Fig. 6 that Ic (lanes 5 and 6) and Id (lanes 7 and 8) are incorporated into the primer by terminal deoxynucleotidyl transferase to yield long oligonucleotide products; Ib (lanes 3 and 4) and Ia (lanes 9 and 10) also serve as substrates for the enzyme.
It was found that 3Ј-azido-2Ј,3Ј-dideoxythymidine derivatives II are not substrates for DNA polymerases ␣ and ␤ and terminal deoxynucleotidyl transferase (data not shown). This finding is of no surprise, because AZTTP is a very poor substrate for these enzymes. However, both AZTTP and IIa-d at rather high concentrations inhibit DNA synthesis catalyzed by DNA polymerase ␣ and terminal deoxynucleotidyl transferase (Table III). Compounds IIb-d were as efficient as AZTTP in  inhibiting HIV reverse transcriptase, but their inhibitory effect on DNA polymerase ␣ was 5-6-fold smaller than that of AZTTP.
We studied the stability of ␥-phosphate-modified AZTTP derivatives in human blood serum. In all assays we did not observe formation of AZT 5Ј-diphosphate; only AZT 5Ј-monophosphate and AZT were formed (data not shown). It is evident from Table IV that for ␥-methylphosphonate IIa and ␥-phenylphosphonate IIb the half-lives are 8 and 11 times, respectively, greater than that for AZTTP. The dephosphorylation rates for IIc and IId were intermediate between the rates for IIa, IIb, and AZTTP.
It should be noted that compounds Ia, Ib, IIa, and IIb are surprisingly highly hydrophobic. The retention time for Ib and IIb on a reversed-phase HPLC column is even higher than that for AZT 5Ј-monophosphate (Table IV). TLC on Silica gel plates also revealed high hydrophobicity of Ia, Ib, IIa, and IIb.

DISCUSSION
The literature data outlined in the introduction indicate that replacement of the ␥-phosphate residue with a carboxyl main-  tains the substrate properties of these dNTP analogs toward several DNA polymerases (12,13). Consequently, for these enzymes one negative charge in the ␥-phosphate is sufficient to ensure formation of a catalytically competent complex. Modification of the triphosphate residue in sugar-substituted dNTPs slightly reduces their affinity to reverse transcriptases but increases their selectivity toward these enzymes (23,27).
The data obtained in this work may be summarized and analyzed as follows.
Replacement of the ␥-Phosphate with Methylphosphonate or Phenylphosphonate-Replacement of the ␥-phosphate with methylphosphonate or phenylphosphonate has a minor effect on the substrate properties of dTTP and AZTTP toward retroviral reverse transcriptases. Compounds II and AZTTP inhibited HIV reverse transcriptase-catalyzed DNA synthesis by 50% at close concentrations, but the K m values for the ␥-phosphate-modified AZTTP derivatives differed 10 -20-fold from that for AZTTP. This contradiction may be ascribed to the fact that the K m were measured in a standing start single-substrate incorporation assay, whereas the 50% inhibition values were determined in a system containing all four dNTPs. High substrate efficiency of ␥-phosphate-modified analogs of dTTP and AZTTP suggests that the dNTP-binding site of HIV and AMV reverse transcriptases has no sterical obstacles for binding triphosphate fragments bearing bulky substituents at the ␥-position. We have earlier studied the substrate properties of dTTP derivatives bearing methyl (23), phenyl, and decyl (28) residues at the P ␣ atom. It has been shown that modification of the ␣-phosphate markedly decreases the substrate efficiency of dTTP. The data obtained in this work indicate that modification of the ␥-phosphate has a minor effect on the substrate properties of dTTP toward reverse transcriptases and some DNA polymerases. These results are easily explainable, because the ␣-phosphate is the reactive group of dNTP, whereas the ␥-phosphate is not involved directly in the reaction, although it is assumed to play a role in dNTP binding to the enzyme.
Replacement of the ␥-Phosphate with Phenylphosphonate-Replacement of the ␥-phosphate with phenylphosphonate totally inactivates dTTP as substrate for DNA polymerases ␣ and ␤ and terminal deoxynucleotidyltransferase. However, the ␥-ester and ␥-amide derivatives (Ic and Id, respectively) are utilized by these enzymes. This is inconsistent with the data described in Refs. 2 and 3, which indicate that different 2Јdeoxynucleoside ␥-alkylamidotriphosphates are not substrates for DNA polymerase ␣.
It is unclear why the ␥-aryland ␥-alkylphosphonates differ in substrate activity from dTTP and its ␥-amide and ␥-ester derivatives. Conversion of 2Ј-deoxythymidine 5Ј-phosphate to 2Ј,5Ј-dideoxythymidine 5Ј-phosphonate increases pK a 2 from 6.5 to 7.0 (29). The angles O-P-O and C-P-O in 2Ј-aminoethylphosphate (30,31) and 2Ј-aminoethylphosphonate (32) differ by 5°. The O-P bond in the first compound is longer than the C-P bond in the second one by 0.26 Å. The phosphonate and phosphate groups differ in the distribution of electronic density around the phosphorus atom. However, these data are not sufficient to explain the differences in the biochemical properties of ␥-phosphates and ␥-phosphonates.
Modification of the ␥-Phosphate in AZTTP-Modification of the ␥-phosphate in AZTTP increases the selectivity of HIV reverse transcriptase inhibition versus DNA polymerase ␣. Thus, modification of the triphosphate residue may be a useful step in designing selective inhibitors of HIV reverse transcriptase. Furthermore, ␥-methylphosphonate IIa and ␥-phenylphosphonate IIb are dephosphorylated in human serum much less rapidly than AZTTP. Besides, IIa and IIb are markedly more hydrophobic than AZTTP. Their TLC mobilities on silica gel plates in dioxane-NH 4 OH-water (R f 0.40 and 0.45) are even higher than those for 2Ј-deoxythymidine 5Ј-monophosphate and AZT 5Ј-monophosphate (R f 0.38 and 0.42). For Ib and IIb the retention times in reversed-phase HPLC were 14 and 18 min, whereas for dTTP and AZTTP they were 6.8 and 7 min; for AZT 5Ј-monophosphate and AZT, they were 11 and 24 min. Thus, replacement of the ␥-phosphate in AZTTP with ␥-phosphonate markedly alters its substrate properties toward some cellular DNA polymerases and blood dephosphorylating enzymes but does not change its substrate activity with respect to HIV reverse transcriptase. It seems likely that modification of the triphosphate residue in nucleotide analogs will yield potent reverse transcriptase inhibitors stable in blood and hydrophobic enough to penetrate into the cell. Also, modified dNTPs of the proposed structure may be useful as tools for studying the cellular processes, in which nucleoside 5Ј-triphosphates serve as donors of ␥-phosphates or nucleotide residues.