Divergent anti-human immunodeficiency virus activity and anabolic phosphorylation of 2',3'-dideoxynucleoside analogs in resting and activated human cells.

The mechanism of divergent anti-human immunodeficiency virus type 1 (HIV-1) activity of various 2',3'-dideoxynucleoside analogs (ddNs) in peripheral blood mononuclear cells (PBM) was studied. We demonstrate that the in vitro anti-HIV-1 activity of various ddNs varies profoundly and that the divergent antiviral activity is related to the extent of anabolic phosphorylation of each ddN and its counterpart 2'-deoxynucleoside (dN). We also show that certain ddNs cause a reduction of their counterpart dNTP formation in PBM in the following order: 2',3'-dideoxycytidine (ddC) >> 2',3'-didehydro-2',3'-dideoxythymidine (d4T), 3'-thia-2',3'-dideoxycytidine (3TC), 2',3'-dideoxyinosine (ddI), 2',3'-dideoxyguanosine (ddG) > 3'-azido-2',3'-dideoxythymidine (AZT) > 2'-beta-fluoro-2',3'-dideoxyadenosine (F-ara-ddA). Based on the phosphorylation profiles, anti-HIV-1 ddNs can be classified into two groups: (i) cell-activation-dependent ddNs such as AZT and d4T that are preferentially phosphorylated, yield higher ratios of ddNTP/dNTP, and exert more potent anti-HIV-1 activity in activated cells than in resting cells; and (ii) cell-activation-independent ddNs including ddI (and 2',3'-dideoxyadenosine), F-ara-ddA, ddG, ddC, and 3TC that produce higher ratios of ddNTP/dNTP and exert more potent anti-HIV-1 activity in resting cells. These data should provide a basis for the elucidation of the mechanism of the divergent antiretroviral activity of ddNs.

nondividing (unstimulated) cells (3)(4)(5). It has recently been demonstrated that HIV-1 proviral DNA synthesis can be completed in resting peripheral blood mononuclear cells (R-PBM), although this synthesis occurs extremely slowly and ineficiently (6). Bukrinsky and his co-workers (7) have also demonstrated that a large proportion of the HIV-1 genome in asymptomatic individuals exists as full-length, extrachromosomal DNA, which retains the ability to be integrated upon activation (stimulation for division) of the host cell. In addition, proviral DNA can be detected in a relatively large percentage of PBM from HIV-1-infected individuals, the majority of which are resting cells, and the proviral DNAlevel seems to correlate with the progression of HIV-1 diseases (8). Thus, HIV-1 infection of resting cells appears to play an important role in both the viral replication cycle and the pathogenesis of acquired immunodeficiency syndrome (AIDS).
A group of nucleosides, 2',3'-dideoxynucleoside analogs (ddNs), has been shown to be active against HIV-1 replication both in vitro and in vivo (9)(10)(11). ddNs are successively phosphorylated in the cytoplasm of a target cell to produce ddN-5'triphosphates (TP) and become analogs of the 2'-deoxynucleoside (dN)-TP, the natural substrates for cellular DNA polymerases and the viral reverse transcriptase. It is thought that the ratio of 2',3'-dideoxynucleoside-5'-triphosphate to 2'deoxynucleoside-5'-triphosphate (ddNTP/dNTP) is one of the most important determinants of ddN antiviral activity against HIV-1(9). We have recently reported that when compared with 2',3'-dideoxyinosine (ddI or didanosine), 3'-azido-2',3'-dideoxythymidine (AZT or zidovudine) achieved higher ratios of AZlTP/dTTP intracellularly and exerted much greater activity against HIV-1 replication in phytohemagglutinin-stimulated PBM (PHA-PBM) (12). In contrast, ddI produced higher ratios of ddATP (the putative active form of dd1) to dATP relative to AZTTP/d'M'P in R-PBM (12). These results suggest that the anti-HIV-1 activity of AZT is superior in activated cells, whereas ddI is more effective in resting cells. These data appear to be in agreement with a recent clinical observation that a simultaneous regimen ofAZT and ddI had more profound and persistent anti-HIV-1 activity in patients with advanced HIV-1 infection than an alternating regimen of the two drugs (13). In the present study, we demonstrate that the in vitro activity of various ddNs against HIV-1 differs substantially in PBM and that such divergent antiviral activity is related to the differential anabolic phosphorylation of ddNs and dNs. We propose that ddNs can be classified into two groups: (i) cell activation-dependent ddNs and (ii) cell activation-independent ddNs.
Determination of Anti-HN-1 Activity of ddNs in PHA-PBM-The HIV-1 strain used was isolated from a patient with advanced HIV-1 infection (HIV-lESRlolpm) prior to antiviral therapy, as described previously (14). Briefly, PBM obtained from HIV-1 seronegative donors were stimulated with PHA (10 pg/ml) (PHA-PBM) for 72 h in recombinant interleukin 2-containing RPMI 1640-based culture media, seeded in 24-well tissue culture plates at a density of 1 x lo6 celldml, and exposed to 2500 TCID, of HIV-1 preparation in the presence or absence of various concentrations of ddNs. The antiviral potency of ddNs was defined as the drug concentration that inhibited HIV-1 p24 Gag protein production by 50% (IC5,,). All experiments were performed in quadruplicate.
Metabolism Studies-Freshly isolated, resting, nondividing PBM (R-PBM) and activated, dividing PBM (PHA-PBM) were incubated with 10 p~ L3H]ddN (1 Ci/mmol) for 12 h. Cells were then harvested, centrifuged, and washed twice with cold phosphate-buffered saline. After centrifugation, the cell pellets were subjected to extraction of nucleosides and nucleotides with 60% methanol. The extracts were then heated for 2 min at 95 "C and analyzed by HPLC using an ion-exchange column (Partisil 10-SAX; Whatman Inc., Hillsboro, NJ) as previously described (12, 15). Quantification of dNTP Pools in PBM Exposed to ddN-An enzymatic assay using a DNA polymerase, previously described by Sherman and Fyfe (16), was used with modifications (12) to quantify dNTP in PBM. In the modified assay, a dNTP, present in excess, was radiolabeled. The amount of radioactivity incorporated into DNA, catalyzed by the Sequenase enzyme (2.0 version, United States Biochemical Corp., Cleveland, OH), was proportional to the dNTP quantified. The reaction mixture contained 0.05 units of Sequenase enzyme, 50 m~ Tris-HC1, pH 7.5, 10 m~ MgCl,, 5 IIIM dithiothreitol, 0.25 p~ template-primer (the sequences of the oligodeoxynucleotides have been described elsewhere (16)), 2.5 p~ PHIdATP (15 Ci/mmol, for dCTP, d'ITP, and dGTP determinations) or 2.5 p~ l3H1dTTP (15 Ci/mmol, for dATP determination), to which known amounts of dNTP standard samples or 5 p l of extract were added. The total assay volume was 50 4. Polymerase reactions were carried out at 26 "C for 20 min, followed by spotting 40 pl of reaction mixture onto a Whatman DE81 filter. The filters were dried, extensively washed with 5% Na,HPO,, briefly rinsed with distilled water and with 95% ethanol, and dried. The radioactivity on the filters was then counted in a liquid scintillation counter. The radioactivity (disintegrationdminute) was plotted as a function of the quantity of dNTPs. The dNTP quantities in test samples were determined from these standard curves. In determining dNTP quantity in samples extracted from ddN-treated cells, an extract (5 pl) from -10' cells cultured in the absence of ddN was mixed with an extract from ddN-exposed cells in various ratios, and the mixtures were subjected to the polymerase assay. From thus obtained linear curves for correction, the reduction by ddNTP was determined and the dNTP quantities were adjusted to represent the values otherwise obtained in the absence of ddNTP. Nucleoside Kinase Assays-Nucleoside kinase assays were performed as previously described (12). One unit of TK and dCK activities is defined as 1 nmol of dTMP or dCMP formed per hour. The observed enzymatic activities were normalized to the unit of enzymatic Rad).
activitylmg of protein using the Bio-Rad protein assay reagent (Bio- ddNs Are Differentially Phosphorylated in Resting and Activated PBM-In an attempt to elucidate the mechanism responsible for the substantially different anti-HIV-1 potencies of ddNs in PHA-PBM, we studied the anabolic phosphorylation of ddNs in R-PBM and PHA-PBM. After the cells were exposed to 10 p~ [3H]ddN for 12 h, cellular ddN nucleotide metabolites were extracted with 60% methanol and quantitated using the anion-exchange HPLC method (12). Based on anabolic phosphorylation patterns and the cell activation state, the phosphorylation profiles were categorized into three groups: (i) anabolic phosphorylation of thymidine analogs such as AZT and d4T, which were most affected by PHA stimulation (cell activation) and were efficiently phosphorylated in PHA-PBM. The triphosphate levels of AZT and d4T were greater by 208-and 25-30fold, respectively, in PHA-PBM than those in R-PBM (Table I and Fig. 2). It is noteworthy that both AZT and d4T underwent little triphosphorylation in resting cells. (ii) Anabolic phosphorylation of purine analogs including ddI, F-ara-ddA, and ddG, which did not show a response to PHA stimulation and, with the exception of F-ara-ddA, were poorly phosphorylated in both resting and activated cells (Table I and Fig. 2). (iii) Anabolic phosphorylation of cytidine analogs such as ddC and 3TC, which were more efficiently phosphorylated in both resting and activated cells than the purine analogs, although their phosphorylation was not significantly potentiated by PHA-stimulation (Table I and Fig. 2).

Anti-HN-1 Activities of ddNs Differ Substantially in Vitro in
Inhibition of Cellular dNTP Formation by ddN-Since the intracellular dNTP pools correlate with the antiretroviral activity of ddN, we asked whether the level of dNTP pools was affected by ddN treatment per se. As shown in Table 11, when PBM were exposed to AZT for 72 h, substantial alterations of the dNTP pools were observed. First, the dTTP pool was decreased by an average of 40% in R-PBM, although variable results were observed in PHA-PBM (three cases of increase versus one case of decrease). Second, AZT inhibited dGTP formation by 21 and 14% for resting and activated cells, respectively. Finally, AZT stimulated dCTP formation in both R-PBM and PHA-PBM by 38 and 25%, respectively. Exposure to d4T decreased the dTTP pool of R-PBM and PHA-PBM slightly to moderately, whereas it failed to stimulate dCTP formation. Ex- posure to ddC resulted in a profound depletion of the cellular dCTP pool, especially in R-PBM. This depletion was concentration dependent with more potent inhibition of dCTP formation at higher ddC concentrations in R-PBM (data not shown). Exposure to 3TC caused only a slight inhibition of dCTP formation in both R-PBM and PHA-PBM. Reductions of the dATP and dGTP pools in R-PBM by ddI and ddG treatment, respectively, were more profound than those in PHA-PBM (40 versus 20% and 44 versus 17% for ddI and ddG treatment, respectively, Table 11). Exposure to F-ara-ddA did not cause an appreciable decrease of dATP pools in either R-PBM or PHA-PBM.
Ratios of ddNTP to dNTP in Resting and Activated PBM-One of the most critical factors that determine the antiviral potency of ddN against H N is the ratio of intracellular ddNTP to its competing (or natural) dNTP (9). We, therefore, evaluated the ratio of each ddNTP and its counterpart dNTP in PBM. It was found that d4T, like AZT, was more efficiently phosphorylated in PHA-PBM than in R-PBM, with higher ratios of d4TTP to d?TP seen in PHA-PBM (Table I). The level of d4lTP increased 25-fold upon PHA stimulation (0.02-0.49 pmoV106 cells). At the same time, the level of dTTP increased 11-fold, from 0.34 to 3.80 pmoV10' cells. As a result, the ratio of d4TTP/ dTTP was -2-fold higher in PHA-PBM compared with that in R-PBM. The level of AZTTP in PHA-PBM was found to be 208-fold higher than in R-PBM, resulting in an AZ' I"l' P/dTTP value that was increased -15-fold after cell activation.
Unlike AZT and d4T, the cytidine analogs, ddC and 3TC, as well as the purine analogs, ddI, F-ara-ddA, and ddG, produced higher ddNTP/dNTP ratios in R-PBM than in PHA-PBM (Table   I). The ddC triphosphate increased by 1.7-fold upon PHA activation, whereas the dCTP pool underwent a 7.4-fold increase. These disproportionate changes resulted in a 4.4-fold higher ddCTP/dCTP ratio in R-PBM (Table I). The triphosphate form of 3TC was increased by 3.8-fold upon PHA stimulation, whereas dCTP was increased by ?"fold, yielding an -2-fold higher ratio of 3TCTP/dCTP in R-PBM compared with PHA-PBM. As for ddI and F-ara-ddA, the triphosphate levels in R-PBM and PHA-PBM were similar, while levels of dATP increased by 19-and 8.5-fold upon PHA activation, resulting in a decrease in the ddNTP/dNTP values by 15.8-and 8.5-fold, re-Effects of dideoxynucleosides on dNTP pool sizes of resting and PHA-activated PBM   TABLE I1 R-PBM and PHA-PBM were exposed to 10 ~MAZT, d4T, ddC, 3TC, ddI, ddA, F-ara-ddA, or ddG at 37 "C. Cellular nucleotides were extracted and 2'-deoxyribonucleoside-5'-triphosphate levels were determined as described under "Experimental Procedures." Data represent means 2 1 S.D. of three experiments, each with triplicate determinations. Values in parentheses represent % reduction in dNTP pool sizes in the cells exposed to each for AZT, ddC, and ddI were 1.02,1.15,0.15, and 0.14 pmoY1O'resting PBM and 6.18,4.24, 1.92, and 1.49  spectively. For ddG, the triphosphate level decreased slightly after PHA-stimulation, whereas the level of dGTP increased 8-fold. Consequently, the ratio of ddGTP/dGTP turned out to be 12.5-fold higher in R-PBM than in PHA-PBM. It should be noted that the absolute concentration of both 3TC and F-ara-ddA triphosphates in R-PBM were substantially higher than those of other ddN analogs tested.

Effects of Pyrimidine ddNs on Thymidine Kinase and Deoxycytidine Kinase Activities in R-PBM and PHA-PBM-The S-
phase-specific cytoplasmic thymidine kinase type 1 (TK1) and the constitutively expressed mitochondrial thymidine kinase type 2 (TK2) play major roles in the phosphorylation of thymidine, while the cytoplasmic deoxycytidine kinase (dCK) and mitochondrial deoxypyrimidine kinase (dPydK) play a major role in the phosphorylation of deoxycytidine (17-21). We therefore asked whether or not 1 m~ concentration of the four pyrimidine ddNs studied here could affect the phosphorylation of [l4C1dT or [l4C1dC catalyzed by these kinases in protein extracts from R-PBM and PHA-PBM using phosphocreatine as the phosphate donor (Table 111). The addition of AZT to PHA-PBM protein extract decreased the amount of phosphorylated dT (which represents TK activity) by 88% (2.82-0.35 nmol/h/mg protein), although no decrease was detected using R-PBM protein extract. This result suggests that AZT in PHA-PBM is phosphorylated by TK1, whose activity is enhanced upon PHA activation. AZT also caused a reduction in the phosphorylation of dC in stimulated cells (Table III), suggesting that AZT could compete with ['*C]dC with respect to the activity of cytoplasmic dCK. Unlike AZT, d4T did not cause a significant decrease in the phosphorylation of dT in PHA-PBM (Table 111). This appears to agree with a report that d4T is a poor substrate for both TK1 and TK2 (17). It was noted, however, that dCK activity in the protein extract from PHA-PBM was moderately blocked by d4T.
In the presence of ddC, no significant effect on the phosphorylation of [14C]dT and [l4C1dC was detected. This agrees with a previous report that ddC is a relatively poor substrate for TK1, TK2 (191,dCK,and dPydK (20). In contrast, 3TC caused   (Table 111). This appears to support previous data showing that 3TC has a higher affinity for dCK as compared with ddC (21), although further study regarding the substrate specificity of dCK toward 3TC is required.

No Significant Metabolic Interaction Occurs When AZT and ddI Are Used
Simultaneously-A recent report shows that a simultaneous regimen ofAZT plus ddI has more significant and persistent antiviral activity in patients with AIDS or ARC than the sequential regimen of AZT and ddI (13). We then asked whether the enhanced antiviral activity observed in patients was due to a metabolic interaction ofAZT and ddI. It was found that when PBM were treated with 10 1.1~ r3H]AZT in the presence of 10 p~ ddI, the formation of AZTTP was not affected either in R-PBM or PHA-PBM (Table IV). The triphosphorylation profiles of AZT assessed by HPLC in the presence and absence of ddI were virtually identical. Concentrations of intracellular dTTP upon exposure to AZT plus ddI were also TAB^ N Effects of combinations ofAZT and ddI on their triphosphorylation in PBMfrom two donors R-PBM and PHA-PBM were exposed to 10 p~ ISH]MT or [3WlddI in the presence and absence of 10 p~ ddI or MT, respectively. Cellular nucleotides were extracted and determined as described under "Experimental Procedures." comparable upon exposure to AZT alone as examined in either of the PBM populations. As a result, the ratios of AZTl'P/d'ITP in R-PBM and PHA-PBM cultured with AZT and ddf were similar to those in the cells cultured with AZT alone. When the cells were exposed to i3HlddI in the presence of AZT, the formation of ddATP was comparable to that in the cells cultured with [3H]ddI alone in both R-PBM and PHA-PBM, and the ratios of d d A T P /~T P were also comparable in both cell populations. These results suggest that the enhanced antiviral effect seen in patients receiving the simultaneous regimen ofAZT plus ddI was not due to the influence of either agent on ddN metabolism, but due to the complementary antiviral activity of AZT in dividing, activated cells and ddI in resting, nondividing cells.

DISCUSSION
The present study has demonstrated that anabolic phosphorylation of ddNs differs substantially depending on the activation state of the target cells. The phosphorylation of ddNs in response to cell activation appears to be primarily determined by the increase in the appropriate kinase and therefore related to the base moiety of ddNs, whereas sugar modification affects the efficiency of anabolic t~phosphorylation. Moreover, we demonstrated that the formation of intracellular dNTP is substantially affected by exposure to ddNs, with the degree of inhibition varying as a Eunction of the target cell activation state. The anti-HIV-1 activities of ddN, therefore, must vary disproportionately in resting and activated cells. It is worth noting that HIV-1 infects not only activated, but also resting CD4+ cells (3-5). Thus, for antiretroviral therapy with nucleoside derivatives, it is important to define the metabolism of each derivative in regard to the triphosphorylation pattern in terms of the target cell activation state.
Our results suggest that ddNs can be classified into two subfamilies. First, cell activation-dependent ddNs, such as AZT and d4T, that are preferentially phosphorylated and exert more potent anti-HIV-1 activity in activated cells than in resting cells. In this regard, Perno et ai. (22) have reported that the anti-HIV-1 activities of AZT and several thymidine analogs in monocyteslmacrophages are potentiated by granulocyte macrophage colony-stimulating factor, suggesting that the phenomenon is not seen only with PHA activation. The second subfamily, cell activation-independent ddNs, including dd1 (and ddA), F-ara-ddA, ddG, ddC, and 3TC, that are preferentially phosphorylated and exert more potent anti-HIV activity in resting cells than in activated cells. Regarding the anti-HIV-1 activity of cell activation-independent ddNs in R-PBM, we have recently demonstrated that ddI can, on a molar basis, exert more potent anti-HIV activity than AZT in vitro.2 Furthermore, our results show that there are two different subtypes within the subfamily of cell activation-independent ddNs: cytidine analogs (ddC and 3TC) and purine analogs fddI, F-ara-ddA, and ddG). Cytidine analogs were phospho~lated quite efficiently in resting ceils ( Table I and Fig. 21, and their triphospho~lation was further enhanced upon cell activation. In contrast, tnphosphorylation of purine analogs occurred independent of cell activation ( Table I and Fig. 21, presumably because the enzymes responsible for their phosphorylation, such as cytosolic 5'-nucleotidase, are not responsive to cell activation (12). In addition, cellular dCTP formation was found to be differentially inhibited by cytidine anaiogs. Therefore, the cytidine analog triphosphate to dCTP ratio varies among different analogs. Indeed, ddC caused a more substantial inhibition of dCTP in R-PBM than in PHA-PBM, especially at higher concentrations (data not shown), resulting in a higher ratio of ddCTP/dCTP in R-PBM. In the case of 3TC, 3TCTPldCTP values in R-PBM were just 2-fold higher than in PHA-PBM, presumably because 3TC was less inhibitory of dCTP formation than ddC.
The reason for the changes of dTTP, dCTP, and dGTP pools upon exposure to AZT remains unclear. At a high ~ncentration ( 1 mM), AZT substanti~ly inhibited TK activity in PHA-PBM, but this inhibition was not distinct in R-PBM. In this regard, AZT is a relatively poor substrate for TK2, which is constitutively expressed in both resting and activated cells, as compared with TK1, which is greatly enhanced in activated cells (17). Thus, the kinase activity detected in R-PBM may represent mostly TK2 activity, and, therefore, it is possible that the inhibition of dTTP formation in R-PBM (Table 11) may involve a rnechanismb) other than TK inhibition. In contrast, TK1 activity was profoundly suppressed by AZT (Table III), which may also account for the depletion of the dTTP pool in PHA-PBM and the inhibition of thymidylate kinase activity by AZT-MP described by Furman et al. (23). Moreover, in PHA-PBM, the AZT exposure caused an increase in the dCTP pool (Table II), whereas AZT inhibited the dCK activity (Table 111). One possible explanation is that the AZT depletion of dTTP stimulated the activity of ribonucleotide reductase, an enzyme catalyzing the rate-limiting step of de novo dNTP synthesis. TTP is a feedback inhibitor of pyrimidine ribonucleotide reduction and an activator of guanine ribonucleotide reduction (24).
Thus, the observed dGTP pool decrease by U T exposure could be due to an indirect effect of dTTP depletion and a direct inhibitory effect ofAZT on deoxycytidine kinase. Cytosolic dCK exhibits a broad substrate specificity and can phosphorylate dG T. Shirasaka and H. Mitsuya, unpublished data. to dGMP (25) as well as F-ara-ddA to F-ara-ddATP (26).
In terms of the magnitude of triphosphorylation, 3TC presented a different picture than the other ddNs in this study. 3TC was very efficiently converted to its triphosphate, yielding 3TCTPJdCTP values of 8.59 and 4.58, in R-PBM and PHA-PBM, respectively (Table I). This is apparently in agreement with a report that 3TC is a good substrate for cytoplasmic dCK (21) and contrasts with its close ddN analog, ddC, which is a poor substrate for both cytoplasmic and mitochondrial dCKs (20). In fact, a moderate reduction in the phosphorylation of [l4C1dC by dCKs was seen in the presence of 3TC in both R-PBM and PHA-PBM, although no reduction was seen with ddC (Table 111). Nevertheless, the anti-HIV-1 potency of 3TC is comparable to that of ddC when examined in PHA-PBM system (27). One possible explanation is the higher K, of 3TCTP for HIV-1 reverse transcriptase compared to that of ddCTP (28).
The triphosphorylation profile of F-ara-ddA (Table I) was particularly intriguing. F-ara-ddA has been reported to be as potent against HIV-1 as its analogs ddA and ddI when evaluated in the cytopathic effect inhibition assay employing CD4+ ATH8 target cells (29,30). F-ara-ddA, however, was substantially less potent than ddA or ddI in the PHA-PBM system using the inhibition of p24 Gag protein production as an end point (Fig. 1). In this regard, considering the observation that the F-ara-ddATPIdATP ratio is quite high in R-PBM, F-ara-ddA could exert a favorable antiviral activity in vivo where the majority of target cells for HIV-1 infection are thought to be resting cells (3)(4)(5). Indeed, as examined in severe combined immunodeficiency mice reconstituted with human peripheral blood leukocytes (hu-PBL-SCID mice), F-ara-ddA has exhibited a potent anti-HIV activity that appeared to be superior to that of AZT (31). The current data on F-ara-ddA suggest further investigation of the agent as a potential anti-HIV-1 drug.
We previously reasoned that the favorable antiviral activity seen in patients receiving the simultaneous regimen of AZT plus ddI as compared with those receiving the alternating regimen of AZT plus ddI (13) was due to the complementary effect ofAZT and ddI (12). In the present study, we found that neither AZT nor ddI significantly influenced the metabolism of the other dideoxynucleoside (Table IV). These data further support the possibility that AZT and ddI complement each other by exerting antiviral effects in activated and resting cells, respectively. The present results also suggest that the potential for effective combination chemotherapy might be enhanced if drugs from each of the two categories, the cell activation-dependent ddNs and the cell activation-independent ddNs, are combined. It should be noted, however, that the pharmacokinetics of each drug and the emergence of drug-resistant HIV-1 variants also have to be considered as critical factors in designing effective combination antiretroviral chemotherapy.