Mechanism of Anti-Human Immunodeficiency Virus Activity of β-d-6-Cyclopropylamino-2′,3′-Didehydro-2′,3′-Dideoxyguanosine

Abstract

To better understand the importance of the oxygen in the ribose ring of planar unsaturated nucleoside analogs that target human immunodeficiency virus (HIV), a 6-cyclopropyl-substituted prodrug of 2′,3′-didehydro-2′,3′-dideoxyguanosine (cyclo-d4G) was synthesized, and its cellular metabolism, antiviral activity, and pharmacokinetic behavior were studied. Cyclo-d4G had selective anti-HIV activity in primary blood mononuclear cells (PBMCs), effectively inhibiting the LAI strain of HIV-1 by 50% at 1.1 ± 0.1 μM while showing 50% inhibition of cell viability at 84.5 μM. The antiviral activity in PBMCs was not markedly affected by mutations of methionine to valine at position 184 or by thymidine-associated mutations in the viral reverse transcriptase. Mutations of leucine 74 to valine and of lysine 65 to arginine had mild to moderate resistance (as high as fivefold). Studies to delineate the mechanism of cellular metabolism and activation of cyclo-d4G showed reduced potency in inhibiting viral replication in the presence of the adenosine/adenylate deaminase inhibitor 2′-deoxycoformycin, implying that the antiviral activity is due to its metabolism to the 2′-dGTP analog d4GTP. Intracellular formation of sugar catabolites illustrates the chemical and potentially enzymatic instability of the glycosidic linkage in d4G. Further studies suggest that cyclo-d4G has a novel intracellular phosphorylation pathway. Cyclo-d4G had a lower potential to cause mitochondrial toxicity than 2′,3′-dideoxycytidine and 2′,3′-didehydro-3′-deoxythymidine in neuronal cells. Also, cyclo-d4G had advantageous synergism with many currently used anti-HIV drugs. Poor oral bioavailability observed in rhesus monkeys may be due to the labile glycosidic bond, and special formulation may be necessary for oral delivery.

Highly active antiretroviral therapy utilizing a combination of different antiviral drugs has served to limit the morbidity and mortality associated with human immunodeficiency virus (HIV) infection. A vital component of these treatment regimens are nucleoside reverse transcriptase inhibitors (NRTIs). After phosphorylation by cellular metabolic pathways, NRTIs serve to chain-terminate viral reverse transcripts due to their lack of a 3′-hydroxyl group. To date, eight NRTIs have been approved by the U.S. Food and Drug Administration (FDA) for the treatment of HIV. However, HIV's high rate of replication and the lack of proofreading by reverse transcriptase (RT) during viral replication lead to frequent mutations and the establishment of resistant virus populations (8, 15, 25, 30). Some of these mutations cause broad-spectrum resistance to all NRTIs currently used (20, 33, 35). Treatments capable of decreasing viral replication in patients harboring cross-resistant virus represent an unmet challenge (22), illustrating the importance of discovering new agents efficacious against resistant HIV.

Nucleotide analogs that have planar ribose ring conformations are well-tolerated substrates for incorporation by DNA polymerases (19). The ability of planar analogs to mimic the natural substrate in the active site of HIV-1 RT may make resistance selection less likely (38). These findings make results showing the planar carbocyclic dGTP analog carbovir triphosphate (CBVTP, the active metabolite of the anti-HIV drug abacavir [ABC]) to be a relatively ineffective mimic of dGTP surprising (26). Kinetic analyses of RT mutants selected by ABC in cell culture (37) also show that they are able to effectively increase the selectivity for dGTP over CBVTP (27). CBVTP's inability to effectively mimic dGTP and the ability of multiple mutants in HIV-1 RT to further decrease its incorporation led to the hypothesis that the ribose oxygen plays an important role in HIV-1 RT incorporation and the likelihood of HIV to develop resistance through mutagenesis. Kinetic studies further support this hypothesis and show that the planar guanosine analog containing an oxygen in the ribose ring (d4GTP) to be a superior substrate for HIV-1 RT. Unlike CBVTP, HIV-1 RT with a mutation of methionine to valine at position 184 (M184V) was unable to select against d4GMP incorporation (28).

Kinetic results showing the importance of oxygen in the ribose ring for incorporation by HIV-1 RT led to the synthesis of a 6-cyclopropylamino prodrug of 2′3′-didehydro-2′,3′-dideoxyguanosine (d4G) and initial characterization of this nucleoside analog showed it to have anti-HIV activity and advantageous characteristics in comparison to the parent nucleoside (29). Here we report the mechanism of action of the 6-cyclopropylamino prodrug of d4G (cyclo-d4G) and further explore the impact of the ribose oxygen by comparing cyclo-d4G's activity to that of the corresponding carbocyclic analog ABC.

MATERIALS AND METHODS

Chemicals.

2′,3′-Dideoxycytidine (ddC), 2′,3′-dideoxyinosine (ddI), and 2′,3′-didehydro-3′-deoxythymidine (d4T) were purchased from Sigma-Aldrich (St. Louis, Mo.). Carbovir (CBV) was a gift from Robert Vince and William B. Parker. d4G, cyclo-d4G, and 1-β-d-2,6-diaminopurine dioxolane (DAPD) were synthesized by previously reported methods (6, 29). d4G was resuspended in dimethyl sulfoxide or solutions buffered above pH 7.0 to avoid decomposition due to the highly labile glycosidic linkage (29). Sugar 5′-³H-labeled cyclo-d4G was custom synthesized by Moravek Biochemicals (Brea, CA). All other chemicals were the highest grade and were purchased from Sigma-Aldrich.

Cell culture.

Peripheral blood mononuclear (PBM), CEM-CCRF, MT-2, and Neuro 2A cells were maintained under standard cell culture conditions. CEM-CCRF, MT-2 and Neuro 2A cells were all grown in RPMI supplemented with 10% heat-inactivated fetal bovine serum, l-glutamine, and antibiotics. The nucleosides' effect on HIV-IIIB (R. C. Gallo) was assessed in MT-2 cells with 3-(4,5-dimethyl-thiazol-2-yl)-2,5-diphenyltetrazolium bromide dye (Sigma-Aldrich) as previously described (21). The antiviral effect of the compounds was evaluated in human PBM cells against a parental LAI strain and a panel of resistant viruses cloned into the Bru Pitt. (J. W. Mellors) background by site-directed mutagenesis by previously reported methods (31, 32). Curves were then constructed and used to determine the effective concentration of nucleoside necessary to reduce the cytopathic effect caused by HIV in MT-2 cells or viral replication in PBM cells by 50 and 90% (EC₅₀ and EC₉₀). The toxicity of nucleoside analogs in uninfected cells was also addressed. Cyclo-d4G and d4G were tested against herpes simplex virus types 1 and -2 and hepatitis B virus by previously reported methods (2, 5).

Effect of mycophenolic acid (MPA) and deaminase inhibitors on nucleoside antiviral activity.

The effects of the deaminase inhibitors erythro-3-(adenin-9-yl)-2-nonanol (EHNA, purchased from Sigma-Aldrich) and 2′-deoxycoformycin (dCF, a gift from William B. Parker, Southern Research Institute) and the IMP dehydrogenase (IMPDH) inhibitor MPA on the anti-HIV activity of d4G, cyclo-d4G, and other related nucleoside analogs in MT-2 cells were determined. Either 10 μM EHNA, 10 μM dCF, or 200 nM MPA was preincubated with MT-2 cells for 30 min before the addition of a nucleoside analog and infection with HIV. Parallel experiments were also done to determine the anti-HIV activity of a given nucleoside in the absence of pretreatment to facilitate accurate comparisons of activity.

Metabolism of [³H]cyclo-d4G in CEM and PBM cells.

Accumulation studies in CEM and PBM cells were initiated by the addition of 10 μM [5′-³H]cyclo-d4G (1,000 dpm/pmol) to cells at a density of 10⁶ cells/ml. At specified time points, 10 ml aliquots of cell culture were taken and centrifuged for 10 min at 350 × g at 4°C. Pelleted cells were then washed three times by resuspension in 10 ml phosphate-buffered saline, followed by centrifugation. The cells were counted, and metabolites were extracted by incubation overnight at −20°C with 1 ml of 60% methanol. The next day, further extraction was performed for 1 h on ice in 200 μl of 60% methanol. Combined extracts were dried under a gentle filtered airflow and then stored at −20°C until they were analyzed by high-performance liquid chromatography (HPLC). Cellular extracts were resuspended in 200 μl of distilled water, and aliquots were analyzed by HPLC. Reverse-phase ion-pairing HPLC analysis used a Columbus 5-μm C₁₈ column (Phenomenex, Torrance, Calif.), a buffer A solution of 5 mM tetrabutylammonium phosphate in the presence of 25 mM ammonium acetate (pH 7.0), and a multistage linear gradient from 0 to 60% buffer B (100% methanol). Radioactivity was analyzed by fraction collection, followed by beta counting.

Enzymatic studies of phosphorylation by nucleoside kinases.

Reactions were conducted to determine if d4G and cyclo-d4G were substrates for cytoplasmic thymidine kinase (TK1), 2′-deoxycytidine kinase (dCK), or mitochondrial 2′-deoxyguanosine kinase (dGK) under conditions where ≥70% of the enzymes' respective natural nucleoside substrates were converted to their monophosphate forms. All kinases were a generous gift from Staffan Erickson, Swedish University of Agricultural Sciences. The assays were performed with 0.5 mM substrate by previously reported methods for TK1, dCK, and dGK (1, 39). Products were separated with HPLC and a 5 to 600 mM linear gradient over 30 min of triethylammonium bicarbonate in distilled water, pH 8.0, and a Pharmacia HR 5/5 strong anion-exchange column. The presence of d4GMP as a product of the dGK reaction was verified by liquid chromatography-electrospray ionization mass spectrometry.

Kinetics of guanosine prodrug deamination by adenosine deaminase (ADA).

Steady-state kinetic experiments were used to determine the K_m (steady-state equilibrium binding constant) and k_cat (maximum steady-state rate) for the deamination of adenosine, ABC, and cyclo-d4G by the method of initial rates (7). Experiments with adenosine were done as previously described using 1.23 nM ADA (Boehringer Mannheim), and absorbance changes were monitored at 265 nm (36). To measure the slow rate of cyclo-d4G deamination, the enzyme concentration had to be elevated to 1.23 μM, and the conversion of cyclo-d4G to d4G at various concentrations (from 10 to 200 μM) was monitored at 295 nm using the calculated extinction coefficient of 0.00609 μM⁻¹ cm⁻¹. Kinetic constants were calculated using a nonlinear least squared curve fitting to the equation k_obsd = k_cat[S]/([S] + K_m), where k_obsd is the observed rate at a given substrate concentration ([S]).

Antiviral synergy between nucleoside analogs.

Isobolograms were constructed by measuring the fractional changes in EC₅₀ (FEC₅₀s) of two anti-HIV agents when used in combination in the MT-2 anti-HIV cell culture method described above. An isobologram is defined as the FEC₅₀ of the first agent plotted against the FEC₅₀ of the second agent when it is used in combination with the first. Each point on the isobologram was generated from triplicate wells added at a single concentration of compound 1 at less than or equal to its EC₅₀, and compound 2 was then titrated (by twofold serial dilution) to determine its EC₅₀ in the presence of compound 1. Values on the additivity line connecting FEC₅₀ values of 1 represent an additive interaction between the two compounds, while points markedly above or below the line represent antagonism or synergy, respectively. Experiments were carried out essentially as described previously (4).

Mitochondrial toxicity in Neuro 2A cells.

To estimate the potential of cyclo-d4G to cause neuronal toxicity, mouse Neuro 2A cells (American Type Culture Collection 131) were used as a model system. The concentrations necessary to inhibit cell growth by 50% (IC₅₀) were measured by a 3-(4,5-dimethyl-thiazol-2-yl)-2,5-diphenyltetrazolium bromide dye-based assay. Perturbations in cellular lactic acid were measured by a colorimetric detection method (Sigma-Aldrich). Changes in mitochondrial DNA levels at defined concentrations of ddC, d4T, and cyclo-d4G were assessed by Southern blot analysis by previously reported methods (21).

Pharmacokinetics in rhesus monkeys.

Cyclo-d4G (33.3 mg/kg of body weight) was administered either orally or intravenously to three rhesus monkeys. Serum and urine samples were taken over 24 h. Cerebral spinal fluid (CSF) was taken at 1 h. 2′,3′-Isopropylidene adenosine was added to 50 μl of collected serum at a final concentration of 20 μg/ml as an internal standard. Serum proteins were precipitated by the addition of acetonitrile, followed by vortexing and centrifugation at 9,000 × g for 3 min. The supernatant was removed, vacuum dried, and frozen overnight. Dried serum samples were reconstituted in 50 μl of 10 mM potassium phosphate buffer, pH 6.8, immediately prior to injection. Urine samples were diluted 1:50 with 10 mM potassium phosphate buffer, pH 6.8, and an internal standard was added. CSF samples were diluted 1:1 with 10 mM potassium phosphate buffer, and an internal standard was added. All samples were analyzed for cyclo-d4G by reverse-phase chromatography and with a Phenomenex Synergi Hydro column with a gradient from 10 mM potassium phosphate buffer, pH 6.8, to 50% acetonitrile. Cyclo-d4G and internal standard were monitored by UV absorbance at 260 and 285 nm. Data were modeled to a single-compartment open pharmacokinetic model by general computational techniques.

RESULTS

Anti-HIV activity of cyclo-d4G against different strains and resistant forms of HIV-1.

For comparison with previous results showing d4G and cyclo-d4G (Fig. 1) to have selective anti-HIV IIIB activity in MT-2 cells (29), their anti-HIV activity was further characterized against the LAI strain of HIV-1 in PBM cells. Both EC₅₀ and EC₉₀ values and the IC₅₀ value were determined (Table 1). As suggested by previous data with MT-2 cells, cyclo-d4G was observed to be three- to fivefold-less potent at inhibiting HIV replication in PBM cells than d4G. Both d4G and cyclo-d4G were observed to have approximately 10-fold-more-potent anti-HIV activity in PBM cells than what has previously been determined for MT-2 cells (EC₅₀s of 4.8 ± 1.6 and 8.6 ± 1.3 μM previously observed for d4G and cyclo-d4G, respectively [29]). Accompanying the slightly reduced antiviral activity, cyclo-d4G was also found to be less cytotoxic to cellular growth. To assess if d4G or cyclo-d4G had antiviral activity against other viruses, they were tested against herpes simplex virus types 1 and -2 as well as hepatitis B virus; neither showed any activity up to 30 μM (data not shown).

FIG. 1.

Structures of cyclo-d4G and d4G.

TABLE 1.

Anti-HIV activity against HIV-1 LAI in PBM cells^a

NRTI	Anti-HIVLAI in PBM (μM)
	EC50	EC90	IC50
d4G	0.34 ± 0.27	3.4 ± 1.2	54.2
Cyclo-d4G	1.1 ± 0.06	17.0 ± 3.9	84.5

^aValues represent the means ± SD of at least three independent experiments done in triplicate.

To better understand the ability of common mutations in the viral reverse transcriptase to cause resistance to cyclo-d4G, its activity was determined against a panel of site-directed mutant HIV-1 isolates harboring mutations known to cause resistance to one or more nucleoside analogs (Table 2). The EC₅₀ and EC₉₀ values against the wild-type virus were compared to those obtained with mutant virus to obtain the 50 or 90% fractional inhibition values (FI₅₀ or FI₉₀), representing fold resistance. Cyclo-d4G was found to be most susceptible to resistance caused by the K65R mutation. Viruses with amino acid changes of L74V and M184V and four mutations associated with resistance to zidovudine (AZT) and other nucleoside analogs (K67N, K70R, T215Y, and K219Q) showed mild to no detectable resistance to cyclo-d4G.

TABLE 2.

Activity of cyclo-d4G against common resistant forms of HIV-1 Bru Pitt. in PBM cells

Virus	Anti-HIV activity (μM)a		FI50b	FI90b	n
	EC50	EC90
Bru Pitt.	4.4 ± 3.7	19 ± 12			6
K65R	19 ± 10	>100	4	>5	2
L74V	9.6 ± 1.5	74 ± 19	2	4	3
M184V	2.2 ± 1.3	11 ± 3	0.5	0.6	3
4× AZTc	1.4 ± 0.9	34 ± 19	0.3	2	3

^aValues represent means ± SD for n experiments.

^bFI₅₀ = EC₅₀^Mutant/EC₅₀^{Bru Pitt.}. FI₉₀^Mutant/EC₉₀^{Bru Pitt.}.

^c4× AZT contains RT mutants K67N, K70R, T215Y, and K219Q.

Intracellular metabolism of cyclo-d4G in PBM and CEM cells.

To understand the metabolic basis for the anti-HIV activity of cyclo-d4G, cell culture experiments following radiolabeled nucleoside in CEM and PBM cells were carried out. Table 3 summarizes the results of cell incubations with 10 μM cyclo-d4G, showing both catabolic and anabolic metabolism pathways. In both PBM and CEM cells, the major metabolic pathway was the breakdown of cyclo-d4G to d4G and sugar catabolites. A small amount of the intracellular nucleoside was phosphorylated to the putative antiviral agent d4GTP. Taking into account that the intracellular volume of PBM cells is approximately fivefold less than that of CEM cells, similar intracellular concentrations of d4GMP, d4GDP, and d4GTP were observed. To further characterize the phosphorylation of cyclo-d4G, the ability of various cellular kinases to catalyze the first phosphorylation step to cyclo-d4GMP was assessed. Unlike the natural substrates 2′-deoxyguanosine and thymidine or the guanosine nucleoside analogs CBV and d4G, neither dCK, TK1, nor dGK was able to phosphorylate cyclo-d4G (Table 4).

TABLE 3.

Metabolism of cyclo-d4G in CEM and PBM cells

Cell type	Time (h)	Cellular concentration (pmol/106 cells)a:
		Sugar catabolitesb	d4G	Cyclo-d4G	Cyclo-d4G-MPd	d4G-MP	d4G-DPc	d4G-TPe
CEM	1	0.41 ± 0.11	0.54 ± 0.26	7.31 ± 1.48	0.55 ± 0.34	0.26 ± 0.06	0.31 ± 0.10	0.15c
	2	0.36 ± 0.05	0.36 ± 0.19	3.32 ± 0.81	0.15 ± 0.08	0.08 ± 0.03	0.23 ± 0.13	0.09 ± 0.04
	4	0.60 ± 0.12	1.39 ± 0.06	4.40 ± 0.38	0.27 ± 0.07	0.16 ± 0.03	0.26 ± 0.02	0.18 ± 0.01
	8	0.63 ± 0.04	2.40 ± 0.21	4.65 ± 0.95	0.29 ± 0.15	0.28 ± 0.02	0.29 ± 0.05	0.17 ± 0.02
	24	—c	1.91c	3.46 ± 0.25	0.22 ± 0.07	0.92 ± 0.17	0.36 ± 0.04	0.41 ± 0.19
PBM	1	0.07 ± 0.01	0.12 ± 0.02	0.71 ± 0.18	0.13 ± 0.04	0.06 ± 0.01	0.13 ± 0.08	0.11 ± 0.05
	2	0.09 ± 0.05	0.19 ± 0.04	0.67 ± 0.26	0.10 ± 0.04	0.04 ± 0.01	0.08 ± 0.01	0.08 ± 0.03
	4	0.10 ± 0.03	0.27 ± 0.08	1.3 ± 0.17	0.10 ± 0.01	0.12 ± 0.10	0.11 ± 0.02	0.07 ± 0.01
	8	0.16 ± 0.02	0.92 ± 0.38	1.48 ± 0.18	0.07 ± 0.01	0.08 ± 0.01	0.15 ± 0.01	0.09 ± 0.01
	24	0.63 ± 0.07	1.87 ± 0.35	0.36 ± 0.07	0.30 ± 0.06	0.12 ± 0.09	0.10 ± 0.02	0.05 ± 0.02

^aValues represent means±SD of three independent experiments.

^bThree distinct peaks for sugar metabolites separated by HPLC.

^cSamples lost or affected by poor chromatographic resolution.

^dIdentity inferred from relative retention time.

^eRetention time estimated using a ddGTP standard.

TABLE 4.

Percentage of nucleoside converted to monophosphate by cellular kinases

Nucleotide	% Monophosphate formeda
	dCK	TK1	dGK
dG	90	ND	70
dT	ND	80	ND
CBV	ND	ND	10
d4G	0	0	85
Cyclo-d4G	0	0	0

^aND, not determined.

Effect of MPA on the anti-HIV activity of nucleoside analogs.

To further characterize the metabolism of cyclo-d4G relative to other nucleoside analogs, the EC₅₀ values of DAPD, ddI, CBV, ABC, d4G, and cyclo-d4G were determined in combination with the IMPDH inhibitor MPA in MT-2 cells. While 200 nM MPA showed no antiviral activity or cellular toxicity on its own (data not shown), preincubation of MT-2 cells for 30 min before the addition of nucleoside analogs caused marked changes in the antiviral activity of some of the nucleosides tested (Table 5). While CBV, ddI, and DAPD showed three- to fivefold improvements in antiviral activity, no effect was seen with ABC or d4G. Cyclo-d4G was the only nucleoside analog to show a decrease in anti-HIV activity in response to MPA. Incubation of MPA treated cells with 10 μM guanine was observed to restore cyclo-d4G activity to that observed in untreated cells.

TABLE 5.

Effect of mycophenolic acid on NRTI anti-HIV-1 IIIB activity in MT-2 cells

NRTI	Anti-HIV-1 IIIB (μM)a		FI50c	n
	EC50	EC50 + mycophenolic acid
CBV	0.71 ± 0.41	0.090 ± 0.014	8	3
ddI	9.1 ± 5.5	1.75 ± 0.35	5	3
DAPD	9.7 ± 2.1	3.5 ± 2.2	3	3
ABC	2.4 ± 1.2	2.8 ± 1.0	1	2
d4G	4.5 ± 0.7	5.1 ± 1.3	1	2
Cyclo-d4G	8.3 ± 1.5	23 ± 8b	0.4	3

^aValues represent means±SD for n independent experiments done in triplicate.

^bActivity with the addition of 10 μM guanine in addition to mycophenolic acid was 6.8 ± 1.0 (mean±SD of three independent experiments done in triplicate).

^cFI₅₀ = EC₅₀/(EC₅₀ + mycophenolic acid). Values represent the means of n experiments done in triplicate.

Interaction of cyclo-d4G and d4G with deaminase inhibitors.

To gain further evidence that the guanosine nucleoside triphosphate analog d4GTP is the active metabolite, studies were done to determine the anti-HIV activity of cyclo-d4G in the presence of an adenosine deaminase inhibitor, EHNA, and a general adenosine/adenylate deaminase inhibitor, dCF. Results show that preincubation with EHNA had no effect on the anti-HIV activity of d4G or cyclo-d4G. While dCF had no effect on d4G activity, it prevented the anti-HIV activity of cyclo-d4G (Table 6). To further explore the deaminase responsible for cyclo-d4G activity, studies were done with calf ADA. Steady-state kinetic results showed that cyclo-d4G was a poor substrate for ADA. In agreement with previous findings, the ADA-activated guanosine prodrug DAPD (13) was found to be a 200-fold-better substrate than cyclo-d4G, while the novel nucleoside monophosphate deaminase-activated prodrug ABC (12) was not a substrate (Table 7). The combined data obtained from metabolic studies of cyclo-d4G suggest the metabolic pathway presented in Fig. 2.

TABLE 6.

Effect of deaminase inhibitors on anti-HIV-1 IIIB activity in MT-2 cells

Compound	MT-2 (IIIB)a
	EC50	IC50
d4G	6.3 ± 1.8	40 ± 7
d4G + EHNA	9.8 ± 3.2	21 ± 1
d4G + dCF	8.0 ± 2.8	39 ± 9
Cyclo-d4G	11 ± 1	87 ± 2
Cyclo-d4G + EHNA	14 ± 6	53 ± 4
Cyclo-d4G + dCF	>100	85 ± 7

^aValues represent means±SD of two independent experiments done in triplicate.

TABLE 7.

Steady-state catalytic constants for guanosine prodrug deamination by calf ADA relative to adenosine

Compound	kcat (s−1)	Km (μM)	kcat/Km (μM−1 s−1)
Adenosine	110 ± 7	17 ± 4	6.5
DAPDa	0.2 ± 0.01	11 ± 0.9	0.02
Cyclo-d4G	0.01 ± 0.001	96 ± 27	0.0001
ABC	<1 × 10−4	NDb	NDb

^aData from a previous report (13).

^bND, not determined.

FIG. 2.

Putative mechanism for the intracellular metabolism of cyclo-d4G. Solid-lined arrows represent efficient metabolic steps, while dash-lined arrows represent inefficient metabolic steps. Both anabolic and catabolic processes are shown as inferred from the experimental data.

Cytotoxicity and effect on mitochondrial function of ddC, d4T, and cyclo-d4G in a neuronal derived cell line.

Some nucleoside analog therapies cause peripheral neuropathy, thought to be mediated by mitochondrial toxicity (16). Experiments were done with Neuro 2A cells to gain a better understanding of cyclo-d4G's propensity to cause mitochondrial toxicity in neuronal cells. It was found that ddC was the most cytotoxic and caused the greatest effect on lactate production, while having no effect on mitochondrial DNA at concentrations up to 1 μM. Cyclo-d4G showed a smaller increase in lactate and less of a decrease in mitochondrial DNA content than d4T at high concentrations (up to 100 μM) (Table 8).

TABLE 8.

Effects of nucleoside analogs on Neuro 2A cells^a

Nucleoside	IC50 (μM)	Concn (μM)	Lactate (mg/106 cells)	Mitochondrial DNA (% control)
No drug			9.9 ± 0.7	100 ± 30
Cyclo-d4G	140 ± 6	1	10 ± 0.3	110 ± 4
		10	11 ± 0.4	112 ± 33
		100	12 ± 1	81 ± 12
d4T	225 ± 8	1	11 ± 1	95 ± 18
		10	13 ± 1	84 ± 15
		100	14 ± 1	37 ± 11
ddC	6.0 ± 0.2	0.01	11 ± 0.4	150 ± 40
		0.1	12 ± 0.5	108 ± 13
		1	15 ± 0.2	96 ± 8

^aValues represent means ± SD of triplicate wells.

Synergism of cyclo-d4G with other FDA-approved anti-HIV agents.

To test the ability of cyclo-d4G to inhibit HIV in combination with other anti-HIV agents, combination studies were done in the presence of nucleoside analog reverse transcriptase inhibitors (ABC, AZT, lamivudine [3TC], and d4T), a nonnucleoside reverse transcriptase inhibitor (nevirapine), and a protease inhibitor (nelfinavir). Isobolograms show that only the interaction between ABC and cyclo-d4G was less than additive. Moderately synergistic behavior was observed between nelfinavir (data not shown), nevirapine, and d4T. Data points markedly beneath the line representing additive activity showed that cyclo-d4G was synergistic with AZT and 3TC (Fig. 3A). The combination of cyclo-d4G and AZT showed an advantage in synergism over d4T and AZT, which was found to only be additive (Fig. 3B). Interestingly, cyclo-d4G was synergistic with 3TC, while the similar guanosine prodrug ABC was found to be only additive (Fig. 3C). This result was verified in a second experiment showing similar levels of synergy for cyclo-d4G or ABC with 3TC, respectively (data not shown).

FIG. 3.

Advantageous interaction of cyclo-d4G with FDA approved anti-HIV agents. (A) Isobolograms for cyclo-d4G in combination with abacavir (closed circles), d4T (closed squares), nevirapine (X), 3TC (closed inverted triangles), and AZT (open circles). (B) AZT in combination with cyclo-d4G (open circles) and d4T (closed squares). (C) 3TC in combination with cyclo-d4G (closed inverted triangles) and abacavir (closed circles). The straight line from the left top to the right bottom indicates two drugs that are additive, curves below the line indicate drugs that are synergistic, and curves above the line represent antagonism.

Pharmacokinetics in rhesus monkeys.

To address the ability of cyclo-d4G to be orally delivered to patients, pharmacokinetic studies were done with rhesus monkeys. Intravenous (i.v.) and oral dosing was done with three monkeys, and cyclo-d4G levels in serum, urine, and CSF were monitored. i.v. dosing resulted in a maximum plasma concentration of 37.1 ± 2.9 μg/ml (mean ± standard deviation [SD]; 128.9 ± 10.2 μM for samples from three monkeys) after 15 min with a half-life of 2.6 ± 0.1 h (all values are means ± SD of samples from three monkeys). Analysis of urine samples showed 13.3% of the 33.3-mg/kg dose was excreted unchanged in the urine after 8 h. Only one of the three monkeys showed detectable levels of cyclo-d4G after oral administration, achieving a maximum serum concentration of 2.22 μg/ml (7.7 μM) at 30 min. Only 1.2% of the total dose was eliminated in its original form in urine after 8 h from this animal (data not shown). Neither i.v. nor oral administration resulted in detectable levels in CSF. To determine if poor plasma stability was responsible for the small amounts of drug eliminated unchanged, the stability of cyclo-d4G was assessed in monkey and human plasma, showing a half-life of >24 h (data not shown).

DISCUSSION

Similar to previous results (29), cyclo-d4G showed two- to fourfold-less activity than d4G against the LAI strain of HIV, suggesting differences in the effectiveness of activation of these nucleoside analogs. Both d4G and cyclo-d4G showed selective anti-HIV activity in the more physiologically relevant human PBM cell system, with anti-HIV activities approximately 100-fold lower than cellular toxicity. The selectivity of cyclo-d4G in PBM cells was found to be comparable to some NRTIs currently approved by the FDA. The increased anti-HIV activity of d4G and cyclo-d4G in PBM cells relative to MT-2 cells may be due to differences in viral replication between the two cell types. The emergence of viral resistance is one of the main obstacles to sustained viral suppression during therapy. Previous transient kinetic studies have shown that d4GTP is not sensitive to the M184V mutation in reverse transcriptase (28). Consistent with the enzymatic studies, the M184V virus did not show marked resistance to cyclo-d4G in PBM cells. The only mutation that showed a >4-fold resistance was the K65R mutant virus. The resistance profile of cyclo-d4G suggests that it may possess antiviral activity against a wide variety of nucleoside resistant HIV variants.

The combined results of cellular and enzymatic metabolism experiments coupled with the anti-HIV activity of cyclo-d4G in combination with MPA and deaminase inhibitors (EHNA and dCF) led to the novel activation pathway proposed in Fig. 2. The dominant metabolite observed in both PBM and CEM cells was d4G. The formation of high levels of d4G from cyclo-d4G incubations is in contrast with low levels of CBV formed from the structurally similar guanosine prodrug ABC (12) and would suggest that cyclo-d4G, like DAPD (13), is deaminated at the nucleoside level. However, similar to that of ABC, the antiviral activity of cyclo-d4G was abolished by the general ADA and AMP deaminase inhibitor dCF but was unaffected by the ADA inhibitor EHNA. Enzymatic studies also confirmed that while cyclo-d4G was a better substrate than ABC for ADA, it was a worse substrate than DAPD or the natural substrate adenosine. Together, these results suggest that cyclo-D4G is phosphorylated before the deamination reaction takes place and that d4G is predominantly formed from the dephosphorylation of d4GMP. This may be facilitated by the inefficient phosphorylation of d4GMP to its di- and triphosphate forms.

Enzymatic studies to determine the identity of the enzyme catalyzing the first phosphorylation step were inconclusive, showing that cyclo-d4G was not a substrate for a number of nucleoside kinases. The inhibition of cyclo-d4G's anti-HIV activity by MPA may shed light on the identity of the enzyme catalyzing the first phosphorylation step. CBV, ddI, and DAPD have all been shown to be phosphorylated by IMP phosphotransferase, with IMP as a phosphate donor (13, 17, 23). This is consistent with the expectation that an IMPDH inhibitor, MPA, serving to increase IMP concentrations, would potentiate the activity of each of these nucleosides. ABC did not show increased activity in response to MPA, and it is activated by an AMP-dependent process by AMP phosphotransferase (12). d4G also did not show any change in its activity in response to MPA. While the enzyme catalyzing the first phosphorylation step of d4G is not known, its efficient phosphorylation by mitochondrial dGK may suggest that it is activated by an ATP-dependent process by this enzyme. Similar to prodrugs of CBV other than ABC (24), cyclo-d4G showed a decrease in anti-HIV activity, and presumably in metabolic activation, in response to MPA. The observation that the addition of exogenous guanine could rescue the anti-HIV activity of cyclo-d4G suggests that its phosphorylation may be dependent on a guanosine nucleotide as a phosphate donor.

The second-most-abundant metabolites observed in cell culture from the sugar-labeled cyclo-d4G were ribose catabolites. These metabolites may be formed as a result of the previously described chemical instability of the glycosidic linkage present in d4G (29, 34) or an unidentified enzyme catalyzing the depurination of d4G. The appearance of multiple radiolabeled peaks consistent with modified ribose sugar moieties suggests that multiple mechanisms of degradation are occurring. The presence of a robust catabolic pathway may limit the amount of d4GTP formed in cells. Previous results show that the degradation products of d4G are more toxic than d4G (29, 34). Therefore, the presence of robust intracellular pathway for degradation may have implications for the source of d4G and cyclo-d4G's toxicity. Indirect evidence that the IC₅₀ of cyclo-d4G is due to catabolism and not d4GTP can be found in data for anti-HIV activity in presence of the deaminase inhibitor dCF. In the presence of dCF, cyclo-d4G has no antiviral activity and presumably markedly decreased levels of d4GTP; however, the cellular IC₅₀ is unaffected (Table 6).

The importance of minimal incorporation by mitochondrial DNA polymerase gamma is illustrated by the last three drugs of the nucleoside analog class to obtain FDA approval. The active triphosphate analogs of ABC, tenofovir, and FTC have all shown limited incorporation by polymerase gamma and minimal inhibition of mitochondrial DNA synthesis in cell culture (3, 10, 16). Consistent with previous findings with a liver cell line (29), cyclo-d4G showed less of an effect than d4T on mitochondrial DNA content and was also found to cause the smallest increase in lactic acid compared to ddC and d4T. Results for ddC and d4T led to conclusions similar to those derived from more in-depth studies of neuronal toxicity (9).

Due to the current use of multidrug combinations to fight HIV infection, it is important that new compounds have favorable interactions with other agents. Combination studies showed that cyclo-d4G has synergistic activity with inhibitors from the three major classes of anti-HIV drugs approved by the FDA. In fact, the only compound that cyclo-d4G was not observed to have synergistic behavior with was ABC. Interestingly, cyclo-d4G showed some differences in synergy compared to other structurally similar nucleosides. Cyclo-d4G was found to be synergistic with AZT, while d4T, which shares the same unsaturated ribose ring, was observed to be only additive with AZT. The combination of AZT and d4T has also been shown to result in poor clinical outcomes (14). Cyclo-d4G was also found to be synergistic with 3TC, while ABC was found here to be only additive. These results may suggest that cyclo-D4G has an advantage when combined with some currently used nucleoside analogs. Differences in cyclo-d4G's synergy profile from those of other NRTIs with similar structural features may reflect its unique intracellular metabolism.

Pharmacokinetic studies illustrated that cyclo-d4G has poor oral bioavailability, only showing detectable levels in one of three rhesus monkeys after oral administration. This may in part be explained by the acid labile nucleoside being degraded in the acidic stomach environment. Cyclo-d4G's low oral bioavailability is in stark contrast to the high oral bioavailability observed with ABC (approximately 90%) (11) and may illustrate a distinct advantage to having a carbocyclic ribose ring lacking the labile glycosidic linkage. ddI has acid instability similar to that of cyclo-d4G (29), and different formulations designed to modulate the acidic environment of the stomach have helped to increase its oral bioavailability (18). Future studies testing the effects of formulation on cyclo-d4G's oral bioavailability may be warranted. An analysis of the i.v. portion of the studies shows that cyclo-d4G has a short half-life when given i.v. and that only a small amount of the drug is recovered unchanged in the urine. This could be due to chemical instability in plasma; however, studies with human and monkey plasma showed cyclo-d4G to be stable for >24 h. These results, coupled with the appearance of multiple sugar degradation products in cellular metabolism experiments, may suggest cell-mediated systemic clearance to be responsible for the small amount of cyclo-d4G excreted in the urine unchanged.

In conclusion, we have shown that cyclo-d4G is a potent and selective anti-HIV agent with favorable resistance and synergy profiles. Cellular metabolism studies suggested cyclo-d4G to be phosphorylated by a unique anabolic pathway. Poor oral bioavailability may be due to the acid labile glycosidic bond present in cyclo-d4G, and effective oral delivery may require special formulation to avoid the acidic environment of the stomach.