Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Many antiviral nucleoside analogs possess a poor affinity for their activating (phosphorylating) enzyme [i.e., 2',3'-dideoxycytidine (ddC; zalcitabine) for deoxycytidine kinase, 2',3'-didehydro-2',3'-dideoxythymidine (d4T; stavudine) for thymidine kinase (TK), and 2',3'-dideoxyinosine (ddI; didanosine) for 5'-nucleotidase] (Furman et al., 1986; Balzarini et al., 1987a,b; Starnes and Cheng, 1987; Johnson and Fridland, 1989). Therefore, several attempts have been undertaken to design nucleoside monophosphate prodrugs that directly deliver the 5'-monophosphate derivative of nucleoside analogs (i.e., 2',3'-dideoxynucleosides) into the intact target cells. To date, at least four different approaches have been followed to achieve this goal: 1) bis pivaloyloxymethyl (POM) derivatives (Farquhar et al., 1994, 1995; Pompon et al; 1994), 2) bis(S-acyl-2-thioethyl) (SATE) derivatives (Girardet et al., 1995; Valette et al., 1996), 3) aryloxyphosphoramidate (APA) derivatives (Balzarini et al., 1996; McGuigan et al., 1996a,b), and 4) CycloSal derivatives (Meier et al., 1997a,b, 1998a,b; for an overview, see Meier, 1998). When applied to some thymidine monophosphate analogs, all four types of prodrugs have been shown to retain biological activity in TK-deficient cells, pointing to an efficient delivery of the nucleoside 5'-monophosphate inside the intact cells. However, the success and efficiency of the nucleoside kinase-bypass prodrug approach may also depend on the nature of the nucleoside analog. For example, the APA and CycloSal prodrug of d4TMP proved highly successful in retaining antiviral activity in human immunodeficiency virus (HIV)-infected TK cells, whereas the parent compound d4T had no antiviral activity in this cell line (Balzarini et al., 1996; Meier et al., 1997a). Studies with radiolabeled d4T and the APA prodrug of d4T revealed a virtual lack of phosphorylation of d4T in CEM/TK cells, whereas the prodrug of d4TMP was able to produce considerable levels of d4TMP, d4TDP, and d4TTP in CEM/TK cells (Balzarini et al., 1996). However, when the same prodrug approach was applied on AZTMP, the APA and CycloSal prodrugs of AZTMP had no marked anti-HIV activity in CEM/TK cells (McGuigan et al., 1996b; Meier et al., 1997b). Also, the parent AZT was completely devoid of antiviral activity in this cell line. These data suggest that the APA and CycloSal prodrugs of AZTMP must have been unable to eventually generate substantial levels of AZTTP to inhibit HIV replication in the CEM/TK cells. Thus, although the d4TMP prodrugs proved antivirally effective in HIV-infected CEM/TK cells, the AZTMP prodrugs again failed to retain a substantial antiviral potency in CEM/TK cells. It is unclear, however, why the TK bypass concept succeeds with APA and CycloSal prodrugs of d4TMP but not with the corresponding prodrugs of AZTMP. In an attempt to clarify this issue, CycloSal-AZTMP, radiolabeled at the 5-methyl group of AZT, has been synthesized and used to compare the metabolism of AZT and CycloSal-AZTMP in a variety of cell lines, including the TK CEM/TK cells.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Cells. CEM, Raji, and Daudi cells were obtained from American Tissue Culture Collection (Rockville, MD). Molt4/clone 8 cells were provided by N. Yamamoto (Yamaguchi University, Yamaguchi, Japan). The TK-deficient CEM/TK cells were a kind gift from Prof. S. Eriksson (currently at Uppsala University, Uppsala, Sweden) and Prof. A. Karlsson (Karolinska Institute, Stockholm, Sweden). Human primary macrophages were obtained from the blood of HIV-seronegative individuals. Mononuclear cells were separated from blood by Ficoll gradient and layered onto plastic dishes for 5 to 7 days. At the end of this period, nonadherent cells were carefully removed through repeated washings. Remaining adherent cells were >95% macrophages. Further details about this procedure are described elsewhere (Perno et al., 1996).

Radiochemicals and Compounds. The structural formulas of AZT and CycloSal-AZTMP are depicted in Fig. 1. [CH₃-³H]3'-Azido-2',3'-dideoxythymidine (AZT; specific radioactivity, 10.9 Ci/mmol) and CycloSal[CH₃-³H]AZTMP (specific radioactivity, 14 Ci/mmol) were obtained from Moravek Biochemicals Inc. (Brea, CA). Nonlabeled AZT was derived from Sigma Chemical Co. (St. Louis, MI). Unlabeled CycloSal-AZTMP was synthesized according to a previously published procedure (Meier et al., 1997b). For the synthesis of CycloSal[CH₃-³H]AZTMP, 5-[CH₃-³H]AZT was dissolved in acetonitrile, and 1.8 equivalents of 3-methylsalicylchlorophosphane was added at 0°C in the presence of 2 equivalents of diisopropylethylamine. After stirring for 30 min, 2 equivalents of anhydrous tert-butylhydroperoxide as a solution in n-decane (5-6 M) was added, and stirring was continued for 30 min at room temperature. The crude product was obtained after evaporation of the solvent. Further purification was achieved by HPLC (Whatman C18 Partisphere, 4.6 × 250 mm; flow, 1 ml/min; mobile phase, 80% 50 mM KH₂PO₄, pH 5.5, and 20% acetonitrile).

View larger version (9K): [in this window] [in a new window]   Fig. 1.   Structural formulas of AZT and 3-methyl-CycloSal-AZTMP.

Metabolism of [CH₃-³H]AZT and CycloSal[CH₃-³H]AZTMP. The metabolism of the radiolabeled compounds was studied according to previously established procedures (Balzarini et al., 1991, 1993). Briefly, wild-type CEM or TK-deficient CEM/TK cells were seeded at 2 to 4 × 10⁵ cells/ml in RPMI 1640 culture medium supplemented with 10% FCS, 2 mM L-glutamine, and 0.075% NaHCO₃. The 5-ml cell suspensions in 25-cm² culture flasks were then incubated with varying concentrations (i.e., 0.07 or 0.09, 1, 10, and 100 µM; see Table 2) of the radiolabeled compounds. At different time intervals (i.e., 2, 6, 24, 48, and 72 h; see Table 1), cells were centrifuged, washed (twice with cold RPMI 1640 medium), and precipitated with cold methanol/water (2:1). Adherent macrophages were gently scraped, counted, and then centrifuged and washed as described for the CEM cells. After centrifugation, the supernatants were subjected to HPLC analysis. A linear gradient of 0.007 M (NH₄)H₂PO₄ (pH 3.8) to 0.25 M (NH₄)H₂PO₄ plus 0.5 M KCl (pH 4.5) was used. The eluate was collected in different fractions and assayed for radioactivity in a toluene-based scintillant.

Intracellular Retention of [³H]AZT and Its Metabolites in CEM Cells after Removal of [³H]AZT and [³H]CycloSal-AZTMP from Cell Culture Supernatant. CEM cells were seeded at 4 × 10⁵ cells/ml and incubated with 0.09 µM [³H]AZT or 0.07 µM [³H]CycloSal-AZTMP (5 µCi/5 ml cell culture) for 24 h. Then, the extracellular drug was removed by centrifugation of the cells and three times careful washing of the cell pellet with warm culture medium. After further incubation of the cells during 0, 2, 4, and 8 h after removal of the radiolabeled compound, cell extracts were prepared and the radiolabeled metabolites were determined by HPLC analysis as described earlier.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Phosphorylation of [³H]AZT and [³H]CycloSal-AZTMP in CEM cells as a function of different incubation times. After the incubation of CEM cells with 0.09 µM [³H]AZT or 0.07 µM [³H]CycloSal-AZTMP, the intracellular levels of the parent compounds and their metabolites were measured at different time points during incubation (Table 1). As previously observed, the exposure of CEM cells to AZT resulted in a substantial accumulation of AZTMP at all time points measured, at concentrations that exceeded the intracellular AZTTP levels by 10- to 16-fold. The AZTMP and AZTTP levels were highest after 6-h incubation, after which they progressively decreased. At 72 h, 25% of the highest AZTMP concentration (observed at 6 h) and 40% of the highest AZTTP concentration (observed at 6 h) was still present. The appearance and disappearance of AZTDP closely followed AZTMP and AZTTP levels; the AZTDP levels were ~50% of the AZTTP levels over all time points examined.

When CycloSal-AZTMP was exposed to the CEM cell cultures, the formation of AZTMP, AZT-DP, and AZTTP was inferior to that recorded for the parent compound AZT and peaked somewhat later (~24 h) than with AZT (~6 h; Table 1). Strikingly, during the first 2 and 6 h of CycloSal-AZTMP exposure, relatively small amounts of AZTMP, AZTDP, and AZTTP were formed relative to the cultures exposed to AZT (<2% after 2 h and ~15% after 6 h). However, after 24, 48, and 72 h of incubation with the AZTMP prodrug, the AZT metabolites were half the amounts of those that were formed when parental AZT was given.

Intracellular Retention of [³H]AZT Metabolites on Removal of [³H]AZT and [³H]CycloSal-AZTMP from Culture Medium. CEM cells were incubated with 0.09 µM [³H]AZT or 0.07 µM [³H]CycloSal-AZTMP for a 24-h period, on which the compound was removed from the extracellular medium by thorough washing of the cells (Fig. 2). The decay in phosphorylated AZT metabolites was very similar for the AZT- and CycloSal-AZTMP-exposed cell cultures. At 2 h after removal of the radiolabeled compounds from the extracellular medium, the intracellular levels of phosphorylated AZT metabolites had decreased to 40% of the initial value and further progressively decreased to 15% after 8 h. When considering the intracellular decay of AZTTP, the disappearance of this metabolite was somewhat slower than the total amount of metabolites, with levels being 60, 30, and 20% of the initial values after 2, 4, and 8 h, respectively (Fig. 2). The intracellular T_1/2 of AZTTP was calculated as ~3 h, in both AZT- and CycloSal-AZTMP-exposed CEM cell cultures. The intracellular T_1/2 found in this study is in agreement with the T_1/2 of AZTTP reported by Furman et al. (1986) and Törnevik et al. (1992) in CEM and other cell lines.

View larger version (40K): [in this window] [in a new window]   Fig. 2.   Decay of intracellular [3H]AZTTP (A) and total [3H]AZTMP plus [3H]AZTDP plus [3H]AZTTP metabolites (B) as a function of time after removal of [3H]AZT or [3H]CycloSal-AZTMP from the extracellular medium. The CEM cell cultures were loaded for 24 h with 0.09 µM [3H]AZT or 0.07 µM [3H]CycloSal-AZTMP before drug removal from the culture medium.

Metabolism of [³H]AZT and [³H]CycloSal-AZTMP in TK-Deficient CEM/TK Cell Cultures. Conversion of [³H]AZT and [³H]CycloSal-AZTMP to AZTMP, AZTDP, and AZTTP was examined in the wild-type CEM cell line and its cytosolic TK-deficient counterpart CEM/TK (Fig. 3). AZT phosphorylation was dramatically reduced in the AZT-exposed TK-deficient CEM/TK cells compared with wild-type CEM cells. In the CEM/TK cells, AZTMP levels were below detection limit after 2 h and were four orders of magnitude lower than in wild-type CEM cells after 6 and 24 h. Although no AZTDP could be detected in CEM/TK cells, AZTTP was present at levels that were three to four orders of magnitude lower than in wild-type CEM cells (Fig. 3).

View larger version (36K): [in this window] [in a new window]   Fig. 3.   Metabolism of [3H]AZT (top) and [3H]CycloSal-AZTMP (bottom) in wild-type (CEM/0; left) and TK-deficient CEM/TK cell cultures (right) on 2, 6, and 24 h (top panels) or 2, 20, and 44 h (bottom panels) incubation of the cells with radiolabeled drugs at a concentration of 0.07 µM (CycloSal-AZTMP) or 0.09 µM (AZT) of the radiolabeled drugs.

Metabolism of [³H]AZT and [³H]CycloSal-AZTMP in CEM Cells as a Function of Different Input Concentrations. The formation of AZT metabolites from AZT and CycloSal-AZTMP increased progressively with higher input concentrations (Table 2), and no differences were observed in the intracellular levels of phosphorylated AZT metabolites derived from AZT- and CycloSal-AZTMP in CEM cell cultures. Although the AZTMP levels markedly increased (up to 50-fold) on increasing the initial AZT or CycloSal-AZTMP concentrations from ~0.1 to 100 µM, the AZTTP levels differed by only 5- to 7-fold within an input concentration ranging from 0.1 to 100 µM (Table 2). The markedly slower increase in AZTTP than in AZTMP levels on exposure of increasing extracellular AZT concentrations is in agreement with previously published data in other cell lines (i.e., MT-4; Balzarini et al., 1989). This phenomenon is due to the progressive inhibition of thymidylate (dTMP) kinase on intracellular accumulation of high AZTMP levels (Frick et al., 1988; Balzarini et al., 1989).

Metabolism of [³H]CycloSal-AZTMP in Different Cell Lines. Next, we investigated the metabolism of 0.07 µM [³H]CycloSal-AZTMP in human T-lymphocyte CEM, MT-4, and Molt4/C8 cells, human B-lymphoblast DAUDI cells, primary (phytohemagglutinin-stimulated and nonstimulated) human peripheral blood lymphocytes (PBLs), and primary M/M (Table 3). The metabolism of AZT and CycloSal-AZTMP in resting and activated PBLs and in M/M was included in this comparative study because the metabolic properties of the drugs in these cell lines probably mimic closer the in vivo (patient) situation than the findings in the established laboratory cell lines. AZTTP levels formed in the three T-lymphocyte cell lines after 24-h incubation ranged between 385 and 755 pmol/10⁹ cells. AZTMP levels were 5- to 13-fold higher (2002-7854 pmol/10⁹cells) than the AZTTP levels in these cell lines. The B-lymphoblast DAUDI cells generated the highest AZTMP and AZTTP levels after 24 h, amounting to 11,572 and 1,920 pmol/10⁹ cells, respectively. In sharp contrast, activated PBLs and primary M/M produced AZTMP levels of 843 to 1044 pmol/10⁹ cells but only 29 to 33 pmol/10⁹ cells AZTTP after 24 h. Thus, there was at least one order of magnitude less AZTTP produced in PBLs and M/M than in the T-lymphocyte cell lines. In resting (nonactivated) lymphocytes, AZTMP and AZTTP levels were marginal (2.6 and 0.5 pmol/10⁹ cells, respectively).

Metabolism of [³H]CycloSal-AZTMP in CEM Cells that Were Incubated in Culture Medium at pH 6.9. Because CycloSal-AZTMP is much more stable at pH values lower than 7.0 (Meier et al., 1998), metabolism of [³H]CycloSal-AZTMP was determined after 24-h drug exposure in CEM cell cultures that were incubated in culture medium for which the pH was adjusted to pH 6.9. The AZTMP, AZTDP, and AZTTP metabolites amounted to 4791, 197, and 357 pmol/10⁹ cells, and these levels were comparable to those observed in CEM cell cultures that were incubated in regular culture medium (pH 7.32).

Stability of AZTMP and d4TMP in CEM Cell Extracts. AZTMP and d4TMP have been added to freshly prepared CEM cell extracts at 500 µM. After 6 h of incubation at 37°C, 21% of AZTMP was converted to AZT, whereas only 1.8% of d4TMP was converted to d4T. The higher stability of d4TMP versus AZTMP in CEM cell extracts found in this study was in agreement with our previous investigations (Saboulard et al., 1999). In this study, a variety of AZTMP and d4TMP phosphoramidate prodrugs were exposed to a rat liver enzyme preparation. All of the AZTMP prodrugs tested showed mainly conversion to free AZT, whereas for the d4TMP prodrugs, formation of the free nucleoside was much less pronounced or even undetectable. Moreover, when AZTMP and d4TMP were directly exposed to the rat liver enzyme preparation, the percentage of hydrolysis of d4TMP was 6 and 35% after 1- and 16-h incubations, respectively, whereas for AZTMP, the percentage hydrolysis was 24 and 70% after 1- and 16-h incubations, respectively (Saboulard et al., 1999).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The CycloSal nucleotide prodrugs were originally designed to be cleaved by a chemical hydrolysis mechanism involving a successive, coupled cleavage of the phenylester and benzylester of the nucleotide prodrug phosphotriester molecule (Meier, 1998). The preferential cleavage of the phenylester bond is most likely to occur first, and its speed can be controlled by the nature of 3- and 5-substituents in the phenyl part of the prodrug molecule. The resulting 2-hydroxybenzyl phosphodiester is then spontaneously cleaved to yield the nucleotide and the remaining diol (salicyl alcohol; Fig. 4). The chemical T_1/2 of the 3-methyl-substituted CycloSal-AZTMP derivative used in this study (Fig. 1) was determined as being ~0 h in phosphate buffer, pH 7.3 (Meier et al., 1997b). It cannot be excluded, however, that inside some cell types of the body (i.e., liver cells), enzymatic degradation of the CycloSal-AZTMP prodrug to AZTMP occurs in addition to chemical hydrolysis of the molecule. If this happened, the release of AZTMP from the prodrug molecule and eventual formation of AZTTP should be the result of the complex interplay of both a chemically and an enzymatically driven process.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Proposed activation scheme of CycloSal-AZTMP.

The intracellular AZTTP (and AZTMP) formation peaked later in CEM cells when CycloSal-AZTMP was administered (24 h) than after AZT administration (6 h). Thus, it seems that TK-dependent phosphorylation of AZT to AZTMP in AZT-treated cells proceeds much faster than the release of AZTMP from CycloSal-AZTMP by the two-step prodrug hydrolysis. In addition, the CycloSal-AZTMP-derived levels of AZTTP, the antivirally active metabolite of AZT, were slightly lower than the AZT-derived AZTTP levels (~50% at 24, 48, and 72 h), an observation that is in agreement with the antiviral activity of AZT and CycloSal-AZTMP in CEM cells (EC₅₀ = 0.005-0.006 and 0.006-0.013 µM, respectively; Meier et al., 1997b, 1998a). The comparable antiviral potencies of AZT and CycloSal-AZTMP also indicate that the antiviral activity of these compounds is mainly determined by the intracellular AZTTP levels obtained at, or after, 24 h rather than the AZTTP levels generated before this time point. In fact, time-of-addition experiments in which nucleoside reverse transcriptase inhibitors are added at different time points after virus infection revealed that there is a marked lag period before the reverse transcriptase process starts during the infection process (Witvrouw et al., 1997). These observations are in agreement with our metabolic findings.

It is interesting to note that an intermediate metabolite (i.e., the benzyl phosphodiester derivative of AZTMP) cannot be detected at measurable levels in the cell extracts. This is in sharp contrast with the marked intracellular accumulation of the intermediate metabolite (i.e., alaninylphosphoramidate d4TMP diester) in the case of the APA prodrug of d4TMP (Balzarini et al., 1996). Thus, the conversion (spontaneous reaction) of the intermediate metabolite to AZTMP (and/or AZT) must occur rapidly and is clearly not rate limiting to obtain the free nucleoside or nucleotide (Fig. 4). Also, the initial intracellular appearance of AZTMP in cells exposed to CycloSal-AZTMP is markedly lower than that in AZT-treated cells. This suggests that cellular uptake of the intact prodrug molecule is slower than that of AZT and/or that the chemical or enzymatic conversion of the prodrug to AZTMP is a relatively slow process (chemical T_1/2 ~ 10 h). Although the first hypothesis cannot be completely excluded, it may be rather unlikely because the prodrug, due to its high lipophilicity, should be taken up by passive diffusion (as is also the case for AZT; Zimmerman et al., 1987), rather than by a carrier-mediated transport. However, given the partition coefficients of CycloSal-AZTMP and AZT of 26.9 and 1.09, respectively (Meier et al., 1998a), it may not be unlikely that the CycloSal-AZTMP prodrug may be retained longer in the cell membrane than AZT due to its much higher lipophilicity.

The low levels of AZTMP and AZTTP in the TK-deficient CEM/TK cells exposed to CycloSal-AZTMP correlate with the observed poor antiviral efficacy of this prodrug in CEM/TK cells (anti-HIV-2 activity of CycloSal-AZTMP in CEM/0 and CEM/TK cells: 0.013 and 15 µM, respectively; Meier et al., 1997b, 1998a). This is in sharp contrast with the CycloSal prodrug of d4TMP that retained full antiviral activity in CEM/TK cells (Meier et al., 1997b) and that was found to generate substantial intracellular amounts of d4TMP, d4TDP, and d4TTP in CEM/TK cells (data not shown). Thus, our findings demonstrate that AZTMP either is not efficiently released from the prodrug molecule or is efficiently released but rapidly degraded to AZT as soon as it is formed, rather than being further anabolized to AZTTP. If the first hypothesis is valid, it means that CycloSal-AZTMP is processed intracellularly differently than CycloSal-d4TMP. Because the chemical T_1/2 values of both molecules were found to be identical (T_1/2 ~ 10 h), this phenomenon should clearly be ascribed to an enzymatic cleavage process, which discriminates between the nature of the nucleotide in the prodrug molecule. On the other hand, the second hypothesis that AZTMP is released in the CEM/TK cells but more extensively catabolized to AZT instead of being further anabolized to AZTDP and AZTTP may be equally valid. Indeed, we found that AZTMP is much less stable than d4TMP in freshly prepared CEM cell extracts and that AZTMP is quickly converted to its free nucleoside AZT in the CEM cell extracts. If this phenomenon also occurs in the intact cell, it may be that AZTMP release from the prodrug occurs to a comparable extent in both wild-type CEM and CEM/TK cells and that in both cell lines AZTMP is rapidly converted to AZT, giving it no chance to accumulate for further conversion to the 5'-diphosphate and 5'-triphosphate metabolite. The free nucleoside is then again subject to rapid conversion to AZTMP by TK in wild-type CEM cells but not in CEM/TK cells that lack the cytosolic TK.

The fact that AZTMP markedly accumulates on exposure of wild-type CEM cells to AZT or CycloSal-AZTMP suggests that the anabolism of AZT to AZTMP by cytosolic TK occurs faster than the catabolism of AZTMP to AZT by phosphatases or 5'-nucleotidases. This is not surprising, because cellular TK has a strong affinity and high catalytic conversion rate for AZT (K_m and V_max for AZT are comparable to the values for the natural substrate thymidine; Furman et al., 1986; Balzarini et al., 1989). In contrast, thymidylate kinase has a good affinity for AZTMP (K_m = 8 versus 4 µM for dTMP) but a very low V_max value (0.3% of that of dTMP; Furman et al., 1986), resulting in a substantial retardation of further anabolism of AZTMP to its diphosphates and triphosphates. It remains to be determined which enzyme is responsible for the rapid conversion of AZTMP to its free nucleoside. It is reasonable to assume that the relatively slow release of AZTMP from the CycloSal-AZTMP prodrug molecule contributes to the rapid conversion of the majority of AZTMP molecules to AZT as soon as they are formed in the cells. Therefore, we may hypothesize that an intracellular equilibrium must be rapidly established between dephosphorylation of AZTMP to AZT and the TK-driven phosphorylation of AZT to AZTMP. Both processes must occur at a considerable speed, explaining why very low levels of AZTMP are recorded in CycloSal-AZTMP-exposed CEM/TK cells, in which the TK-driven phosphorylation is virtually absent due to the TK deficiency of these cells.

It is notable that in M/M, CycloSal-AZTMP is eventually converted to a markedly lower extent to AZTTP than in T- and B-lymphocyte cell lines. These data, however, support our view that AZTMP, as soon as it is released from the CycloSal prodrug, is quickly converted to AZT. Indeed, macrophages have TK levels that are markedly lower than lymphocytes (i.e., 5.7 pmol formed dTMP/mg protein/min in macrophages compared with 407 pmol formed dTMP/mg protein/min in lymphocyte H9 cells; Perno et al., 1988). Thus, it is conceivable that the eventual formation of AZTTP in prodrug-treated M/M is relatively low compared with lymphocytes when AZTMP released from the prodrug is mainly converted to AZT before back-conversion to AZTMP by cytosolic TK.

In conclusion, we investigated the metabolism of the CycloSal-AZTMP phosphotriester prodrug compared with that of the parent AZT in both wild-type CEM and CEM/TK cell lines. This revealed the pharmacological basis for the differential anti-HIV activity of this prodrug and its parent nucleoside AZT and for their dramatically reduced antiviral efficacy in TK-deficient CEM/TK cells. We also revealed the much higher instability of AZTMP than d4TMP in cell extracts, which may be the pharmacological basis of the failure of CycloSal-AZTMP to be active in CEM/TK cells. The observed metabolic features are in full agreement with the antiviral data obtained previously (Meier et al., 1997b, 1998a). The study revealed that any pronucleotide approach on AZTMP to efficiently form 5'-diphosphate and 5'-triphosphate metabolites directly derived from the released AZTMP is likely to fail. In this respect, AZT may not be the only nucleoside in this position. Each new nucleoside included in the prodrug approach must be investigated on a case-by-case basis. Resistance of the nucleotide 5'-monophosphates to hydrolytic cleavage in cell extracts may provide good evidence for the eventual successful kinase bypass of their corresponding prodrugs.