1592U89, a Novel Carbocyclic Nucleoside Analog with Potent, Selective Anti-Human Immunodeﬁciency Virus Activity SUSAN M

1592U89, a Novel Carbocyclic Nucleoside Analog with Potent, Selective Anti-Human Immunodeﬁciency Virus Activity SUSAN M. DALUGE,1* STEVEN S. GOOD,1 MICHAEL B. FALETTO,1† WAYNE H. MILLER,1 MARTY H. ST. CLAIR,1 LAWRENCE R. BOONE,1 MARGARET TISDALE,2 NIGEL R. PARRY,2 JOHN E. REARDON,1 RONNA E. DORNSIFE,1 1 1 DEVRON R. AVERETT, ‡ AND THOMAS A. KRENITSKY § Glaxo Wellcome Inc., Research Triangle Park, North Carolina 27709,1 and Glaxo Wellcome plc, Stevenage, Hertfordshire SG1 2NY, United Kingdom2

Received 23 August 1996/Returned for modiﬁcation 5 November 1996/Accepted 26 February 1997

-1592U89, (؊)-(1S,4R)-4-[2-amino-6-(cyclopropylamino)-9H-purin-9-yl]-2-cyclopentene-1-methanol, is a car bocyclic nucleoside with a unique biological proﬁle giving potent, selective anti-human immunodeﬁciency virus (HIV) activity. 1592U89 was selected after evaluation of a wide variety of analogs containing a cyclopentene substitution for the 2؅-deoxyriboside of natural deoxynucleosides, optimizing in vitro anti-HIV potency, oral -bioavailability, and central nervous system (CNS) penetration. 1592U89 was equivalent in potency to 3؅-azido 3؅-deoxythymidine (AZT) in human peripheral blood lymphocyte (PBL) cultures against clinical isolates of

HIV type 1 (HIV-1) from antiretroviral drug-naive patients (average 50% inhibitory concentration [IC50], 0.26 ␮M for 1592U89 and 0.23 ␮M for AZT). 1592U89 showed minimal cross-resistance (approximately twofold) with AZT and other approved HIV reverse transcriptase (RT) inhibitors. 1592U89 was synergistic in combination with AZT, the nonnucleoside RT inhibitor nevirapine, and the protease inhibitor 141W94 in MT4 cells against HIV-1 (IIIB). 1592U89 was anabolized intracellularly to its 5؅-monophosphate in CD4؉ CEM cells and in PBLs, but the di- and triphosphates of 1592U89 were not detected. The only triphosphate found in cells -incubated with 1592U89 was that of the guanine analog (؊)-carbovir (CBV). However, the in vivo pharmaco kinetic, distribution, and toxicological proﬁles of 1592U89 were distinct from and improved over those of CBV, probably because CBV itself was not appreciably formed from 1592U89 in cells or animals (<2%). The

,␣ 5؅-triphosphate of CBV was a potent, selective inhibitor of HIV-1 RT, with Ki values for DNA polymerases ␤, ␥, and ␧ which were 90-, 2,900-, 1,200-, and 1,900-fold greater, respectively, than for RT (Ki, 21 nM). 1592U89 was relatively nontoxic to human bone marrow progenitors erythroid burst-forming unit and granulocyte-

macrophage CFU (IC50s, 110 ␮M) and human leukemic and liver tumor cell lines. 1592U89 had excellent oral bioavailability (105% in the rat) and penetrated the CNS (rat brain and monkey cerebrospinal ﬂuid) as well as AZT. Having demonstrated an excellent preclinical proﬁle, 1592U89 has progressed to clinical evaluation in HIV-infected patients.

Nucleoside human immunodeficiency virus (HIV) reverse significant anti-HIV activity in the class of carbocyclic nucleo- transcriptase (RT) inhibitors continue to be the cornerstone of sides with the analog of 2Ј,3Ј-didehydro-2Ј,3Ј-ddA. The anti- AIDS therapy. Side effects limit the use of 3Ј-azido-3Ј-deoxy- HIV activity of the racemic guanine member of this 2Ј,3Ј- thymidine (AZT), 2Ј,3Ј-dideoxyinosine (ddI), 2Ј,3Ј-dideoxycy- unsaturated carbocyclic series, carbovir, was described by tidine (ddC), and 2Ј,3Ј-dideoxy-2Ј,3Ј-didehydrothymidine (d4T) Vince et al. (64). We synthesized the (Ϫ) enantiomer of car- (15, 36, 41, 49, 54, 67). In addition, resistance to single agents bovir (CBV) and its triphosphate (CBV-TP), studied the phos- has emerged in the clinic (25, 28, 65). For these reasons we phorylation of CBV by cellular enzymes (37), and verified the have continued to search for nucleoside analogs with greater selectivity of CBV-TP for HIV RT over cellular DNA poly- efficacy, less toxicity, and improved resistance profiles for use merases (13, 33). Although we confirmed its potent in vitro in rationally designed combination regimens. anti-HIV activity, we found CBV to be deficient in several Carbocyclic nucleosides, lacking the labile glycosidic linkage respects as an anti-HIV therapy. As with other guanines, in between heterocycle and sugar, could offer an attractive in vivo vivo toxicities seen with CBV in animals could, in part, be stability advantage over the 2Ј,3Ј-dideoxynucleosides. How- attributed to low aqueous solubility. Oral absorption of CBV ever, the carbocyclic versions of dideoxyadenosine (ddA), ddI, was found to be limited (26 and 23% oral bioavailability in the ddC, dideoxyguanosine, and dideoxyribosylthymine have dem- rat and monkey, respectively) (7), a finding consistent with onstrated minimal anti-HIV activity (38). We first detected published studies in the rat (2, 22). In addition, brain penetration of CBV was inefficient relative to that of AZT (see below). We continued the synthesis and evaluation of a wide variety of * Corresponding author. Mailing address: Medicinal Chemistry Di- carbocyclic nucleosides containing the cyclopentenyl sugar vision, Glaxo Wellcome Inc., 5 Moore Dr., Research Triangle Park, mimic in search of a compound with a satisfactory preclinical NC 27709. Phone: (919) 483-2095. Fax: (919) 483-6053. E-mail: susan profile. [email protected]. We now report that 1592U89, (Ϫ)-(1S,4R)-4-[2-amino-6- † Present address: Pfizer Central Research, Groton, CT 06340. ‡ Present address: ICN Pharmaceuticals, Inc., Costa Mesa, CA (cyclopropylamino)-9H-purin-9-yl]-2-cyclopentene-1-metha- 92626. nol (Fig. 1), has properties, including a unique mechanism of § Present address: Krenitsky Pharmaceuticals Inc., Durham, NC activation, which allow it to correct for the deficiencies of CBV 27707. and prodrugs of CBV. In this and the following two papers (14,

1082 VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1083

plaque assay (29). Previously described assays were used to measure activity against hepatitis B virus (HBV) (9, 24), human cytomegalovirus (HCMV) (16), and feline immunodeficiency virus (FIV) (48). Combinations of 1592U89 with other anti-HIV agents. The anti-HIV-1 activity of 1592U89 in combination with other inhibitors was assayed by an HIV cytopathic assay in MT4 cells where cytopathic effect was quantified using either the vital dye 3-[4,5-dimethylthiazol-yl]-2,5-diphenyltetrazolium bromide (MTT) or the DNA stain propidium iodide. The 50% inhibitory concentrations (IC50s) were calculated from the percent cytoprotection for each compound alone and for each compound in the presence of multiple concentrations of the other FIG. 1. Structure of 1592U89 (compound 1). compound. Combination IC50s were converted to fractional inhibition concentrations (FIC50) by dividing the concentration of each compound in the combination by its IC50 when acting alone. Isobolograms were generated by plotting the FIC50s of one compound versus the corresponding values for the second 58), we have summarized our understanding of the manner in member of each combination. which the 6-cyclopropylamino modification in 1592U89 results Urinary recovery of CBV in rats given oral 1592U89, CBV, and analogs. in its unique biological profile. Compounds were dissolved in water with the addition of small amounts (equimolar or less) of hydrochloric acid or sodium hydroxide, as required, and adminis- (This work was presented in part at the 34th Interscience tered to male CD rats (Charles River Laboratories, Wilmington, Mass.) weighing Conference on Antimicrobial Agents and Chemotherapy, Or- 285 to 390 g (n ϭ 3 per compound) by intragastric intubation. Doses were lando, Fla., 4 to 7 October 1994 [3, 7, 13, 19] and at the Eighth equivalent, on a molar basis, to 10 mg of CBV per kg of body weight. Urine was International Conference on Antiviral Research, Santa Fe, collected for 24 h and analyzed for CBV content by HPLC using a C18 Microsorb MV column (250 by 4.6 mm [inside diameter]; Rainin Instrument Co., Woburn, N.Mex., 23 to 28 April 1995 [6, 18]. Clinical evaluation of Mass.) eluted isocratically at 1 ml/min with a mobile phase of 6% methanol and 1592U89 in HIV-infected patients has begun.) 6% acetonitrile buffered to pH 7.0 with 25 mM ammonium phosphate containing 0.3% TEA. The retention time for CBV was about 10 min. The column was purged with 60% acetonitrile in the phosphate-TEA buffer for 4 min and re- MATERIALS AND METHODS equilibrated to initial conditions for 10 min between each analysis. The UV Materials. 1592U89 (compound 1), CBV (compound 2), and related chiral absorbance of the column eluate was monitored at 260 nm, and CBV concen- carbocyclic nucleosides in Table 1 were synthesized at Glaxo Wellcome Inc. from trations were determined by comparison to a calibration curve prepared with the corresponding 6-chloropurine (compound 4) (5, 5a, 5b). All carbocyclic normal urine. nucleosides used were enantiomerically pure and have the same absolute con- Pharmacokinetics and disposition of 1592U89, CBV, and analogs in the rat. figuration as that of CBV. AZT, ddI, ddC, d4T, (Ϫ)-(2ЈR,5ЈS)-1-[2-(hydroxy- Compounds were dissolved in normal saline with the addition of small amounts methyl)-1,3-oxathiolan-5-yl]cytosine (3TC), nevirapine, and (3S)-tetrahydro-3- (equimolar or less) of hydrochloric acid or sodium hydroxide, as required, and furyl N[(1S,2R)-3-(4-amino-N-isobutylbenzenesulfonamido)-1-benzyl-2-hy- administered to male CD rats (230 to 360 g) by bolus injection into the tail vein droxypropl]carbamate (141W94) were synthesized at Glaxo Wellcome Inc. or by intragastric intubation (n ϭ 3 per compound per dose route). Serial blood [8-3H]dATP and [8-3H]dGTP used for DNA polymerase assays were from Du- samples were obtained from each rat via jugular vein cannulas according to the pont-New England Nuclear. [8-3H]1592U89 and [8-3H]CBV were radiolabelled method described by Upton (62). Plasmas were separated by centrifugation. by Moravek Biochemicals, Inc. (Brea, Calif.), and used at a specific activity of Ultrafiltrates of each plasma were prepared by centrifugation at 4,000 ϫ g for 20 1 to 2 Ci/mmol. The higher-specific-activity [5Ј-3H]1592U89 (14 Ci/mmol) was min with Centrifree micropartition devices (Amicon, Inc., Beverly, Mass.). The synthesized at Glaxo Wellcome Inc. and used where indicated instead of [8-3H] ultrafiltrates were stored frozen until analyzed by reversed-phase HPLC using a 3 3 1592U89. AZT (20 Ci/mmol), [2,8- H]dATP (19 Ci/mmol), and [methyl- H] 4.6- by 250-mm Adsorbosphere C18 column (Alltech Associates, Inc., Deerfield, thymidine (65 Ci/mmol) were purchased from Moravek Biochemicals, Inc. The Ill.). Samples of each ultrafiltrate (100 ␮l) were chromatographed at a flow rate purity of the radiolabelled compounds was determined prior to use and was of 1 ml/min by using a 30-min linear gradient from 50 mM ammonium acetate typically Ն98% for each compound. Authentic monophosphate (MP), diphos- containing 0.1% TEA, adjusted to pH 5.5 (buffer A), to a 1:1 (vol/vol) mixture of phate (DP), and TP standards of 1592U89 and of CBV and the MP of (1S,4R)- buffer A and 80% acetonitrile. The column was reequilibrated with buffer A for [2,6-diamino-9H-purin-9-yl]-2-cyclopentene-1-methanol (aminoCBV; com- 10 min prior to each analysis. 1592U89, CBV, and analogs 6, 7, 8, and 9 were pound 3) were synthesized at Glaxo Wellcome nc. by methods described else- eluted at retention times of about 23, 20, 29, 22, 28, and 27 min, respectively. The where (37). Sodium metaperiodate, D-(Ϫ)-ribose, ATP, dATP, CTP, dCTP, UV absorbance of the column eluate was monitored at 260 and 285 nm, and GTP, dGTP, UTP, and dTTP were purchased from Sigma Chemical Co. (St. concentrations of analytes were determined by comparison of their peak areas to Louis, Mo.). Human interleukin-2, diluted to 10%, was from Cellular Products those of calibration standards prepared with rat plasma. The area under the Inc., Buffalo, N.Y. High-pressure liquid chromatography (HPLC)-grade aceto- plasma drug concentration curve (AUC) was calculated from the first to the last nitrile (UltimAR grade), methanol (UltimAR grade), acetic acid, and ammo- time point with quantifiable levels by using the trapezoid rule. Plasma clearance nium hydroxide were from Mallinckrodt (Paris, Ky.). HPLC-grade o-phosphoric was calculated from the ratio of dose to AUC after intravenous dosing. Oral acid (85%) and triethylamine (TEA) were from Fisher Scientific (Fair Lawn, bioavailability was determined from the ratio of the mean AUC after oral dosing N.J.). HPLC-grade water used for all solutions was from a Hydro water purifi- to that after intravenous dosing. Plasma elimination-phase half-life was calcu- cation system (Picotech Water Systems, Research Triangle Park, N.C.). All other lated by using the elimination rate constant obtained from linear regression chemicals were reagent grade or better. Ultrapure 2Ј-deoxynucleoside 5Ј-triphos- analysis of the semilog plot of the terminal phase of the plasma drug concen- phates (dNTPs) and DNA polymerase I Klenow fragment were purchased from tration-time curve after intravenous dosing. Pharmacia Biotech Inc. (Piscataway, N.J.). Levels of 1592U89, AZT, and CBV in plasma and brains of rats. Male CD rats Cell culture. Peripheral blood lymphocytes (PBLs) were isolated from the (230 to 440 g) were dosed with CBV (12 mg/kg), 1592U89 (10 mg/kg), or AZT whole blood of donors negative for HIV type 1 (HIV-1) antibodies by density (10 mg/kg). Compounds were dissolved in normal saline to a final concentration gradient centrifugation on lymphocyte separation medium (Organon Teknika, of about 10 mg/ml by using small amounts of hydrochloric acid (1592U89 and Durham, N.C.). Cells were treated with 5 ␮g of phytohemagglutinin (PHA) AZT) or sodium hydroxide (CBV). Whole brains and plasma samples from the (Amersham Life Science, Arlington Heights, Ill.) per ml for 48 h. The cells were same animal were obtained 0.5, 1.0, and 2.0 h after intraperitoneal dosing (three then grown in an atmosphere of humidified 5% CO2 at 37°C in RPMI 1640 per time point per compound) and stored frozen. Brains were weighed and medium supplemented with L-glutamine (Gibco/BRL, Life Technologies, Grand homogenized with a series 4710 Ultrasonic Homogenizer (Cole-Parmer Instru- Island, N.Y.), 20% heat-inactivated fetal bovine serum (Hyclone Laboratories, ment Co., Chicago, Ill.) in 1 volume (1 ml/g) of HPLC-grade water. After Logan, Utah), interleukin-2, and 50 ␮g of gentamicin (Gibco/BRL) per ml. homogenization, 1 volume (1 ml/g of brain) of 10% trichloroacetic acid (TCA) The human T-lymphoblastoid CD4ϩ CEM cells (CEM-T4) used for antiviral, was added. The homogenates were centrifuged to yield supernatants that were cytotoxicity, and anabolism studies were obtained from the National Institutes of analyzed by reversed-phase HPLC. Plasmas were prepared for HPLC analysis by Health AIDS Research and Reference Reagent Program (Rockville, Md.). The the addition of 0.5 volume of 10% TCA followed by the removal of precipitates HT4LacZ-1 line of CD4ϩ HeLa cells was kindly provided by J.-F. Nicolas by centrifugation. Supernatants (100 ␮l) were chromatographed on a Partisphere (Pasteur Institute, Paris, France) (50). The human leukemic T-lymphoblastoid C18 column (4.6 by 250 mm; Whatman International Ltd., Maidstone, United CEM (CCL-119), T-lymphoblastoid Molt-4 (CRL-1582), IM-9 B-lymphoblastoid Kingdom) at a flow rate of 1 ml/min. 1592U89 was analyzed in plasma and brain (CCL-159), and U-937 monocytoid (CRL-1593) cell lines were obtained from the extracts by using a 30-min linear gradient from 0 to 40% acetonitrile in buffer A. American Type Culture Collection. Details of cell culture were described pre- The same conditions were used to quantitate CBV in plasma extracts. The UV viously (8). absorbance of the eluant was monitored at 285 nm for 1592U89 and 260 nm for Antiviral assays. Clinical isolates of HIV-1 were obtained from antiretroviral CBV and AZT. Retention times for 1592U89 and CBV were about 22 and 18 drug-naive patients and were amplified by a single passage in cultured PBLs. min, respectively. CBV in brain extracts was eluted free of endogenous interfer- Assays used for determination of activity against HIV were previously described ences at about 23 min by using a 30-min linear gradient from 0 to 60% methanol (1, 29, 56). Activity against drug-resistant HIV was assessed by the CD4ϩ HeLa in 50 mM ammonium acetate adjusted to pH 5.5. AZT in plasma and brain 1084 DALUGE ET AL. ANTIMICROB.AGENTS CHEMOTHER. extracts was eluted at about 23 min by using a 30-min linear gradient from 50 mM by strong anion-exchange HPLC. Intracellular levels of dGTP and dTTP were ammonium acetate, pH 5.5 (buffer B), to 20% acetonitrile and 30% methanol in determined by the DNA polymerase assay of Sherman and Fyfe (53). buffer B. A 10-min reequilibration to initial conditions was used between each Dried samples were reconstituted in deionized water and analyzed for radio- sample analysis. Plasma and brain drug concentrations were determined with labelled TPs by HPLC using the abbreviated 30-min gradient described by calibration curves prepared with compounds added either to plasma or to brains Faletto et al. (14). The peaks of radioactivity were identified by comparison of from untreated rats. Brain drug concentrations were converted from nanomoles their retention times to those of authentic standards of CBV-TP and AZT-TP. per gram to micromolar concentrations, assuming a specific volume of 1.0 ml/g, The amount of phosphorylated compound present per 106 cells was calculated and were corrected for levels of drug expected from contamination of brain with the specific activities of each nucleoside analog. The concentrations were preparations with blood (2.0% Ϯ 0.2% of plasma drug levels, determined with converted to micromolar concentrations by dividing by 0.2 ␮l of intracellular 14C-labelled inulin as an impermeable marker). water per 106 PBLs (8, 44). Levels of 1592U89 in plasma and CSF of monkeys. Cynomolgus monkeys (five DNA polymerase assays. Polymerases ␣, ␤, ␥, and ε were purified by a mod- per sex per dose group) were given oral 1592U89 succinate (25, 70, or 210 mg/kg, ification of published procedures (40) and characterized as described elsewhere twice daily, suspended in 0.1% methylcellulose) for 28 days as part of an oral (33). P. A. Furman and P. Ray (our laboratories) kindly supplied RT from the toxicity study. On day 25, concurrent samples of plasma and cerebrospinal fluid IIIB(HXB2) strain of HIV-1, produced in Escherichia coli and purified by chro- ε (CSF) were obtained 1 h after dosing and stored frozen. Plasmas were prepared matography. RT concentrations were determined spectrophotometrically ( 279 ϭ for HPLC analysis with the addition of 0.5 volume of 10% TCA, and precipitates 334,000 MϪ1 cmϪ1). DNA polymerase assays were performed as described were removed by centrifugation. Supernatants (100 ␮l) were chromatographed previously (8). Kinetic constants were determined at 37°C as described elsewhere on a Whatman Partisphere C18 column (4.6 by 250 mm). A mobile phase (37, 46). consisting of 0.1% TEA, 25 mM ammonium phosphate, 15% acetonitrile, and 5% methanol (pH 7.2) at a flow rate of 1 ml/min was used. 1592U89 eluted from the column at about 27 min. The UV absorbance of the column effluent was RESULTS monitored at 285 nm. CSF samples were similarly analyzed, without sample preparation. 1592U89 was eluted at about 10 min by using a 30-min linear Selection of 1592U89 based on anti-HIV-1 (IIIB) potency gradient from 0 to 40% acetonitrile in 50 mM ammonium acetate containing combined with low biotransformation to CBV. Inhibition of 0.1% TEA (pH 5.5). Concentrations of 1592U89 were determined by comparing the cytopathogenicity of HIV-1 (IIIB) in MT4 cells was used to peak areas of unknowns to those of calibration standards (prepared in normal monkey plasma for plasma samples or as aqueous standards for CSF samples). identify the most active carbocyclic nucleoside analogs for fur- Independently prepared quality control samples (in normal monkey plasma and ther evaluations. 1592U89 (compound 1) and a number of CSF) validated the calibration curves. other 6-modified purine analogs had potency comparable to Cytotoxicity of 1592U89, AZT, ddI, and ddC in human leukemic cell lines. Cell that of CBV (compound 2) (Table 1). In this assay, 1592U89 lines used for cytotoxicity assays were maintained in exponential growth in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum and 1 and CBV were approximately fourfold more active than ddI ␮g of gentamicin per ml at 37°C in a 5% CO2 humidified atmosphere. Cells (4 ϫ and somewhat less active than ddC and 3TC. AZT was the 104 cells/ml in 96-well plates) were incubated in the presence of various concentrations of 1592U89 or other compounds for 4 days (45). Growth inhibition was determined with propidium iodide (1). The effects of combinations of drugs on cytotoxicity were assessed by the method of Prichard and Shipman (43). TABLE 1. In vitro anti-HIV-1 (IIIB) activities of 1592U89 Human bone marrow progenitor cell growth assays. Human bone marrow and related compounds in MT4 cells samples were obtained from healthy donors following informed consent through Duke University Medical Center. Assays for inhibition of human bone marrow progenitor colony formation were performed as previously described (11). Inhi- bition data for 1592U89 were determined in 10 separate experiments with 10 different donors, data for CBV were determined in 8 separate experiments with 8 different donors, and data for AZT were determined in 12 separate experiments with 12 different donors. Data on the inhibition of progenitors by ddI and ddC in this assay system have been previously reported (10, 11, 63). Effects of 1592U89 and CBV on intracellular levels of dNTPs. Intracellular Mean IC50 Ϯ SD a levels of endogenous dNTPs were determined in extracts of cells treated with 10 Compound (␮m) (no. of ␮M 1592U89 or CBV by the periodate procedure of Garrett and Santi to remove R X determinations) ribonucleoside TPs (17). The amount of each dNTP present per 106 cells was quantitated by HPLC using calibration curves of authentic dNTPs. Concentra- 1 (1592U89) NH2 4.0 Ϯ 1.6 (21) tions were converted to micromolar concentrations by using a factor of 0.8 ␮lof 2 (CBV) NH 4.6 Ϯ 1.4 (21) intracellular water per 106 CEM cells (8, 44). 2 ؉ ϩ Anabolism of 1592U89 in CD4 CEM cells. CD4 CEM cells were exposed to 3 (aminoCBV) NH2 10 Ϯ 6 (5) 10 ␮M[3H]1592U89 for 48 h before HPLC analysis of MPs, DPs, and TPs as described by Faletto et al. (14). 4NH2 7.6 Ϯ 4.4 (7) ؉ ϩ Determination of CBV-TP half-life in CD4 CEM cells. CD4 CEM cells 5NH 3.0 Ϯ 0.1 (3) were exposed to 10 ␮M[3H]1592U89 or [3H]CBV for 48 h. The treated cells 2 were collected by centrifugation, rinsed twice with ice-cold phosphate-buffered 6NH2 9.0 Ϯ 3.7 (8) saline, and resuspended in fresh medium without compounds. Cells were harvested at various times following the washout of drug. Cell extracts were pre- 7NH2 6.1 Ϯ 1.9 (3) pared and analyzed by strong anion-exchange HPLC as described elsewhere

(14). The rate constant (k) of the exponential decay of CBV-TP was determined 8NH2 6.6 Ϯ 3.5 (7) from the slope of the curve obtained by regression of the natural logarithm of the Ϯ CBV-TP concentration on time. The half-life of decay was determined as 9NH2 7.3 1.6 (5) Ϫ0.693/k. 10 NH2 52 Ϯ 5 (3) Effect of HIV-1 infection on the anabolism of 1592U89. CD4ϩ CEM cells (1.2 ϫ 106 cells/ml) in culture medium were infected with 1 ml of HIV-1 (IIIB) 11 (adenine) H 12 Ϯ 3 (3) (2 ϫ 104 PFU) or with 1 ml of medium (uninfected control) and incubated for Ͼ b 1 h. Each culture was then brought up to a volume of 30 ml with medium and 12 H 200 (3) 3 transferred to T-75 flasks, and [ H]1592U89 was added to a final concentration 13 H Ͼ200 (3)b of 10 ␮M. The flasks were incubated at 37°C. Three 5-ml aliquots were taken from each flask at 24 and 48 h postinfection. The cells were pelleted, extracted, and analyzed for [3H]CBV-TP levels as described above. Reference compounds Anabolism of 1592U89 and AZT in PBLs. PBLs isolated from three healthy AZT 0.040 Ϯ 0.005 (51) donors were plated into six-well plates at a concentration of 2 ϫ 106 cells/ml of 3TC 2.1 Ϯ 0.6 (7) medium. The cultures of unstimulated PBLs were treated immediately with ddI 17 Ϯ 2 (28) either [3H]1592U89 or [3H]AZT to a final concentration of 0.5 ␮M. Cell number ddC 1.6 Ϯ 0.9 (3) was determined with a hemocytometer. Cells were harvested by centrifugation 6 and 24 h after the addition of the 3H-labelled compounds. The cells were washed a HIV-1 (IIIB)-mediated cytopathic effect assayed with a propidium iodide once with phosphate-buffered saline, and nucleotide extracts were prepared as stain for DNA after incubation for 5 days in the presence of compounds. described elsewhere (14). The dried samples were stored at Ϫ80°C until analyzed b 0% inhibition at 200 ␮M. VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1085

TABLE 2. Urinary recovery of CBV in rats given oral 1592U89 and modifications in structure such as the replacement of the cy- analogs and relative substrate efficiencies for adenosine deaminase clopropylamino group of 1592U89 by the isosteric isopro- Adenosine pylamino group of compound 10 resulted in significant reduc- deaminase activity tions in anti-HIV potency. % of dose (% of adenosine)b 1592U89 and analogs in Table 1 have the absolute configu- Purine recovered Compound D ϩ 6-substituent in urine Relative ration of natural -nucleosides. As with ( )-carbovir (37), the as CBVa substrate Relative (ϩ)-enantiomer of 1592U89 was inactive at 200 ␮M (not efficiency velocity shown), apparently due to lack of phosphorylation (14). (Vmax/Km) Our initial interest in the 2Ј,3Ј-unsaturated carbocyclic 1 (1592U89) 1.1 Ϯ 0.4 0.00003 nucleosides began with the finding of the activity of the adenine analog compound 11. Other 2-unsubstituted purine ana- 2 (CBV) 13 Ϯ 5 logs were inactive. The potency of compound 11 in this assay was equivalent to that of ddI and one-third that of CBV (Table 3 (aminoCBV) 19 Ϯ 2 0.016 0.08 1), despite the fact that it was an efficient substrate of adeno- 419Ϯ3 0.002c 0.23c sine deaminase (Table 2) and the resulting 6-oxopurine compound 12 was inactive. This carbocyclic series contrasts with 525Ϯ3 0.016 0.23 the corresponding 2Ј,3Ј-dideoxyribosides in which the 6-oxopurine (ddI) is active because of the conversion of its MP to 6 0.5 Ϯ 0.2 Ͻ0.0007 phosphates of ddA. Modifications of compound 11 were ex- 7 Not detected Ͻ0.00004 plored in an effort to stabilize it to adenosine deaminase. Unfortunately, 6-modified analogs of compound 11 such as the 6 8 8.5 Ϯ 1.0 Ͻ0.00004 N -cyclopropylamino derivative compound 13 were not detect- ably inhibitory to the cytopathic effects of HIV-1 at the highest 9 Not detected Ͻ0.00004 concentration tested (200 ␮M). Vince et al. (64) have since 11 0.42 1.7 reported that compound 11 has 1/10 the selectivity of CBV (no IC s given) in a similar assay in MT2 cells. a 50 Values are means Ϯ standard deviations. The 6-modified 2-aminopurine analogs with potency against b The assay for adenosine deaminase from calf intestine was performed as previously described (55). Relative velocity, relative maximal velocity (if Ͼ0.01) HIV-1 (IIIB) in MT4 cells similar to that of CBV were eval- or relative velocity at 100 ␮M (if Ͻ0.001). uated in rats to assess the extent of their conversion to CBV c Racemic compound used. after oral dosing. CBV levels in the urine were a good measure of metabolic conversion to CBV because renal excretion of unchanged compound is the primary route of elimination of most active nucleoside RT inhibitor evaluated, being approx- parenterally administered CBV (66). As expected, the better imately 100-fold more inhibitory than 1592U89 and CBV. substrates for adenosine deaminase, the 6-amino (compound The 6-amino (compound 3; aminoCBV), 6-chloro (com- 3), 6-chloro (compound 4), and 6-methoxy (compound 5) an- pound 4), and 6-methoxy (compound 5) 2-aminopurine nucle- alogs, yielded greater amounts of CBV in the urine. Although oside analogs were relatively efficient substrates for mamma- the in vitro hydrolysis of compounds 3, 4, and 5 by adenosine lian adenosine deaminase (Table 2). Thus, the activity of these deaminase was inefficient relative to that of adenosine (Table analogs was likely the result of enzymatic hydrolysis to the 2), the ubiquity and abundance of this enzyme, particularly in guanine (CBV) in cells. Consistent with this proposal were the intestine (20), resulted in extensive in vivo conversion to anabolism studies with compound 3 in which the in vitro anti- CBV. The 6-methoxy compound, compound 5, was an espe- HIV activity was reversed by EHNA [erythro-9-(2-hydroxy-3- cially efficient prodrug of CBV, giving a twofold enhancement nonyl)adenine] (14), a potent inhibitor of adenosine deami- in the percentage of the dose recovered as CBV in the urine. nase. Surprisingly, 1592U89 and a few other 6-OR-, 6-SR-, and Other 6-O-substituted analogs such as compound 8 were also 6-NR1R2-substituted-2-aminopurines, e.g., compounds 6 to 9, dealkylated efficiently to CBV in vivo, although they were not retained anti-HIV activity (Table 1) but were not substrates of detectable substrates for adenosine deaminase in vitro. We adenosine deaminase (Table 2). It is interesting that small carried out further in vivo characterization of 1592U89 and

TABLE 3. Pharmacokinetic parameters for 1592U89, CBV, and related compounds in rats at doses equimolar with 10 mg of CBV per kga

␮ ⅐ AUC ( M h) after: Absolute oral Plasma clearance Compoundb bioavailability C c (␮M) Half-life (h) Oral Intravenous max (liters/h/kg) administration administration (%) 1 (1592U89) 23 Ϯ 522Ϯ2 105 17 Ϯ 8 0.56 Ϯ 0.09 1.8 Ϯ 0.4 6 5.5 Ϯ 1.7 (7.7 Ϯ 1.4)d 14 Ϯ 2 (6.4 Ϯ 1.3)d 65e 7.5 Ϯ 5.4 (4.6 Ϯ 2.7)d 0.18 Ϯ 0.04 (0.61 Ϯ 0.07)d 2.9 Ϯ 0.4 2 (CBV) 3.8 Ϯ 2.1 15 Ϯ 1 25 1.4 Ϯ 0.5 0.52 Ϯ 0.01 2.7 Ϯ 1.5 711Ϯ2 0.21 Ϯ 0.02 3.7 Ϯ 0.7 8 8.8 Ϯ 1.0 0.22 Ϯ 0.06 4.6 Ϯ 0.5 9 6.1 Ϯ 0.2 0.14 Ϯ 0.02 6.6 Ϯ 0.2

a Values are means Ϯ standard deviations. b See Table 1, column 3, for purine 6-substituent. c Maximum concentration in serum after oral administration. d Values in parentheses are for 1592U89 formed in vivo by N demethylation of compound 6. e Calculated based on the sum of the AUC data for compound 6 and 1592U89. 1086 DALUGE ET AL. ANTIMICROB.AGENTS CHEMOTHER.

TABLE 4. Penetration of CBV, AZT, and 1592U89 into TABLE 5. CSF and plasma 1592U89 concentrations in monkeys the brain of rats after intraperitoneal injection and CSF/plasma drug concentration ratios 1 h after administration on day 25 of a 30-day Time Concn (␮M)a in: Brain/plasma Compound oral toxicology study postdose drug concn dosed a (h) Plasma Brain ratio ␮ b Dosea Concn ( M) in: CSF/plasma drug c 1592U89 0.5 15 Ϯ 2 0.98 Ϯ 0.35 0.07 Ϯ 0.03 (mg/kg/dose) CSF Plasma concn ratio 1.0 8.9 Ϯ 0.4 0.60 Ϯ 0.10 0.07 Ϯ 0.01 2.0 2.4 Ϯ 0.2 0.19 Ϯ 0.04 0.08 Ϯ 0.02 25 1.2 Ϯ 1.3 6.5 Ϯ 5.6 0.19 Ϯ 0.08 70 2.9 Ϯ 2.2 17 Ϯ 11 0.17 Ϯ 0.05 CBV 0.5 13 Ϯ 2 0.31 Ϯ 0.19 0.02 Ϯ 0.02 210 5.2 Ϯ 4.5 36 Ϯ 31 0.16 Ϯ 0.05 1.0 5.1 Ϯ 1.3 0.07 Ϯ 0.04b 0.01 Ϯ 0.01b a Cynomolgus monkeys received two daily doses of 1592U89 succinate about Ϯ Ͻ c 2.0 2.4 1.4 0.06 NQ 8 h apart. b Values are the means Ϯ standard deviations (n ϭ 10). AZT 0.5 9.0 Ϯ 1.3 0.49 Ϯ 0.28 0.05 Ϯ 0.03 c Values are the means Ϯ standard deviations of ratios for individual animals 1.0 3.9 Ϯ 0.8 0.39 Ϯ 0.39 0.10 Ϯ 0.09 (n ϭ 9). 2.0 0.33 Ϯ 0.19 Ͻ0.14d NQ

a Means Ϯ standard deviations of individual determinations (n ϭ 3, except where indicated). —, not determined. NQ, not quantifiable. plasma and was also undetectable 2 h after dosing (data not b CBV was not detectable in one brain. c CBV was not quantifiable (Ͻ0.13 nmol/g) in one brain and not detectable in shown). the other two brains. In cynomolgus monkeys given 1592U89 succinate (25, 70, or d AZT was not quantifiable (Ͻ0.14 nmol/g) in all three brains. 210 mg/kg twice daily) in a 28-day oral toxicology study, the mean CSF/plasma 1592U89 concentration ratio 1 h after dosing was 0.17 Ϯ 0.06 (Table 5). Similar ratios, averaging 0.21, have been reported in rhesus monkeys dosed with AZT (4). Lipophilicity (log P at pH 7.4) of 1592U89 and analogs. other active compounds (compounds 6 to 9) that were not The 1-octanol–0.1 M sodium phosphate (pH 7.4) partition coeffi- significantly converted to CBV in the rat. 6 Pharmacokinetic parameters for 1592U89 and analogs com- cients (P) of 1592U89, the N -methyl derivative (compound 6) CBV, and AZT were determined by the shake flask method at pared to those of CBV in the rat. The disposition and phar- Midwest Research Institute (Kansas City, Mo.). Log P values macokinetics of the active analogs that gave low or undetect- were averages of triplicate determinations: 1592U89, 1.22 Ϯ able urine CBV levels in the rat were compared to those of 0.03; compound 6, 1.54 Ϯ 0.03; CBV, Ϫ0.62 Ϯ 0.02; and AZT, CBV (Table 3). The superior pharmacokinetic profile of 0.09 Ϯ 0.02. The result for AZT is in good agreement with the 1592U89 was evident. 1592U89 had a sixfold greater AUC value of 0.10 reported with pH 7.0 phosphate buffer (68). after oral administration compared to that of CBV and was Activity of 1592U89 against clinical HIV-1 isolates in vitro. unique among the many analogs studied in having excellent Clinical isolates of HIV-1 from antiretroviral drug-naive pa- oral bioavailability. The plasma 1592U89 half-life was similar tients were assayed in PHA-activated human PBL cultures for to that of CBV, and the clearance was somewhat slower. A sensitivity to 1592U89 and other anti-HIV nucleosides as de- potential metabolite, the nucleobase 2-amino-6-(cyclopro- scribed in Materials and Methods. The results are shown in pylamino)-9H-purine, was not detected in rats, mice, and mon- Table 6. In this assay system the average IC50 of 1592U89 was keys (19) and would not be predicted to be formed in humans approximately equivalent to that of AZT, half that of ddI and because 1592U89, like CBV (32), is not a substrate for human ninefold that of ddC. The relative ranking of 1592U89, AZT, purine nucleoside phosphorylase (14). The 6-azetidino (com- ddI, and ddC with clinical isolates was different from that seen pound 7), 6-butyloxy (compound 8), and 6-thioallyl (compound with HIV-1 (IIIB) in MT4 cells (Table 1), where AZT was 9) analogs had poor pharmacokinetic properties, including more potent than 1592U89 and ddI by 100- and 400-fold, shorter half-lives in plasma and higher clearance values. 6 respectively. We are continuing to monitor the sensitivity of 1592U89 was chosen for further evaluation over the N -methyl clinical isolates of HIV-1 to 1592U89 as clinical studies derivative (compound 6), based on a somewhat greater anti- progress. HIV potency and more favorable pharmacokinetics. 1592U89 was also tested against HIV-1 (IIIB) in PBLs (IC50 1592U89 levels in rat brain and monkey CSF. The penetra- of 3.7 ␮M, as determined by inhibition of RT production tion of 1592U89 into rat brain was similar to that of AZT ϩ [Table 6]), CD4 CEM cells (IC50 of 3.8 ␮M, as determined by (Table 4). Brain 1592U89 levels were 7 to 8% of the concur- inhibition of p24 production [not shown]), and CD4ϩ HeLa rent concentrations in plasma between 0.5 and 2 h after a 10-mg/kg intraperitoneal dose. Concentrations of 1592U89 in the brain at or above clinical isolate IC50s (see below) were maintained for more than 1 h postdose. The concentration of TABLE 6. Sensitivity of HIV-1 (clinical isolates and strain IIIB) AZT in the brain at 2 h postdose was below the level of to 1592U89, AZT, ddI, and ddC in PHA-stimulated quantification in all three animals, whereas the level of human PBL cultures a 1592U89 consistently exceeded that of AZT and was 0.19 ␮M IC50 (␮M) for : at 2 h. In contrast, CBV levels in the brain were only 1% of Compound Clinical isolates Strain IIIB levels in plasma at 1 h and were undetectable in two animals (no. of isolates) (no. of assays) and unquantifiable in the third at 2 h postdose, despite a level in plasma equivalent to that of 1592U89 (2.4 ␮M) at the 2-h 1592U89 0.26 Ϯ 0.18 (8) 3.7 Ϯ 2.6 (4) time point. Consistent with its greater lipophilicity, levels of the AZT 0.23 Ϯ 0.24 (54) 0.09 Ϯ 0.03 (8) N6-methyl analog (compound 6) in the brain were approxi- ddI 0.49 Ϯ 0.37 (56) 16 Ϯ 10 (3) ddC 0.03 Ϯ 0.04 (29) 0.23 (1) mately 20% of concurrent levels in plasma, but compound 6 was eliminated from the brain as rapidly as it was from a Assays were run in triplicate. Values are means Ϯ standard deviations. VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1087

TABLE 7. Susceptibility of drug-resistant strains of HIV to 1592U89 determined in CD4ϩ HeLa cells

Known resistance(s) Mean 1592U89 IC50 Ϯ SD (␮M) Fold Virus clone a (fold IC50 increase ) (no. of determinations) change HIV-1 HXB2 None (wild type) 5.8 Ϯ 3.1 (15) 1.0 HIV-1 RTMN (41L/215Y) AZT (60–70) 11 Ϯ 4 (2) 1.9 HIV-1 RTMC (67N/70R/215F/219Q) AZT (120) 13 Ϯ 4 (5) 2.1 HIV-1 74V ddI (5–10), ddC (16) 21 Ϯ 8 (4) 3.6 HIV-1 181C Nonnucleoside RT inhibitors (Ͼ100) 9.6 Ϯ 0.02 (2) 1.7 HIV-1 184V 3TC (Ͼ500), ddI and ddC (2–10) 13 Ϯ 6 (4) 2.2 HIV-2 LAV2 Many nonnucleoside RT inhibitors (Ͼ1,000) 7.5 Ϯ 1.4 (2) 1.3

a Compared with that for the wild type.

cells (IC50 of 5.8 ␮M, as determined by plaque reduction [Ta- 1592U89 against these HCMV isolates (IC50s, 0.29 to 2.2 ␮M). ble 7]). The HIV-1 (strain IIIB) potency ranking of 1592U89, Activity was not detected against herpes simplex virus type 1 or AZT, and ddI was similar in MT4 cells and PBLs (Tables 1 and 2, varicella-zoster virus, or inﬂuenza A virus at concentrations 6), as well as in CD4ϩ CEM cells (not shown) and in the CD4ϩ of 1592U89 up to 100 ␮M. HeLa cell assay (59), indicating that host cell and assay end In vitro cytotoxicity of 1592U89. Growth inhibition studies point differences do not provide an explanation for the differ- were performed using 1592U89 alone and in combination with ence in ranking between strain IIIB and clinical isolates for AZT, ddI, and ddC. These studies were designed to determine these nucleoside analogs. the growth inhibitory potency of 1592U89 and to allow analysis Anti-HIV activity of 1592U89 in combination with other of the nature of the interaction between 1592U89 and AZT,

anti-HIV compounds. Assessments of the anti-HIV activity of ddI, and ddC. IC50s of 1592U89 are shown in Table 8. The 1592U89 in combination with the approved nucleoside RT IC50s were in excess of 100 ␮M for the IM-9 and CEM cell inhibitors (AZT, ddI, ddC, 3TC, and d4T), a representative lines. In Molt-4 cells, the estimated IC50 was 20 ␮M. The IC50s nonnucleoside RT inhibitor (nevirapine), and a protease in- obtained in this set of experiments for AZT, ddI, and ddC are hibitor (141W94) were conducted in MT4 assays with HIV-1 also reported in Table 8. Although a range of values was (IIIB) using MTT or propidium iodide. Similar results were obtained for each compound, 1592U89 was generally a less obtained with the two detection systems for AZT, ddI, and potent inhibitor of cell growth in the human leukemic cell lines ddC (results shown for MTT analyses). The other compounds than ddC and a more potent inhibitor than ddI. The inhibitory were assayed by only one method. FIC plots (Fig. 2) show that potency of AZT overlapped that of 1592U89. none of the combinations were antagonistic since data points The effect of combination of 1592U89 with AZT, ddI, or in general were on or below a line connecting unity on the two ddC was examined in each of the human leukemic cell lines. axes. Furthermore, data points falling well below this line indicate that 1592U89 was clearly synergistic with AZT, nevirapine, and 141W94. Additivity and/or some synergy was observed in combinations of 1592U89 with ddI, ddC, 3TC, and d4T. 1592U89 inhibition of replication of drug-resistant strains of HIV in vitro. The susceptibility of drug-resistant strains of HIV-1 to 1592U89 was determined in CD4ϩ HeLa cells with cloned drug-resistant variants of HIV-1 (IIIB) and with HIV-2 (Table 7). The IC50 of 1592U89 increased only two- to fourfold for the AZT-resistant clones RTMN and RTMC compared to the wild type. Similar increases were also seen in the IC50sof 1592U89 for clones containing mutations conferring resistance to 3TC, ddC, and a variety of nonnucleoside RT inhibitors. Activity against other retroviruses and other virus groups. The sensitivity of HIV-2 to 1592U89 was similar to that of HIV-1 (IIIB): the 1592U89 IC50 was 4.1 ␮M for HIV-2 (Zy) in MT4 cells (not shown) and 7.5 ␮M for LAV-2 in HeLa CD4ϩ cells (Table 7). FIV (Petaluma strain) was also inhibited by 1592U89. The IC50 in Crandell feline kidney cells as determined by a plaque reduction assay was 0.40 Ϯ 0.20 ␮M(nϭ3) for 1592U89 compared to 0.48 Ϯ 0.46 ␮M(nϭ10) for AZT. Human HBV was inhibited by 1592U89 in two different assay systems. Measurement of intracellular DNA forms in HepG2 (2.2.15) cells gave an IC50 of 7 ␮M, and measurement of extracellular Dane particle production gave an IC50 of 4.7 Ϯ 2 ␮M(nϭ17). Inhibition by 1592U89 of a panel of HCMV isolates was evaluated by a plaque reduction assay. Lab-adapted strain ␮ AD169 was the most sensitive, with a 1592U89 IC50 of 32 M, FIG. 2. Isobologram analyses of 1592U89 combinations with ddI, d4T, ddC, whereas each of ﬁve clinical isolates was two- to threefold less 3TC, AZT, nevirapine (Nev), and 141W94 in HIV-1 (IIIB)-infected MT4 cells. sensitive. Ganciclovir was 50- to 100-fold more active than Cytopathic effect was measured with MTT (——) or propidium iodide (– – –). 1088 DALUGE ET AL. ANTIMICROB.AGENTS CHEMOTHER.

TABLE 8. In vitro cytotoxicity of 1592U89, AZT, ddI, and ddC

Mean IC50 Ϯ SE (␮M) of: Human cell type 1592U89 AZT ddI ddC Leukemic cell linesa IM-9 110 Ϯ 40 400 Ϯ 10 600 Ϯ 60 110 Ϯ 10 CEM 160 Ϯ 20 40 Ϯ 20 700 Ϯ 500 9.0 Ϯ 1.0 CD4ϩ CEM 140 Ϯ 30 800 Ϯ 200 1600 Ϯ 200 10 Ϯ 1 Molt-4 20 Ϯ 10 20 Ϯ 10 1500 Ϯ 300 7.5 Ϯ 0.4 U-937 310 Ϯ 90 60 Ϯ 20 500 Ϯ 200 25 Ϯ 5 Normal bone marrow progenitorsb BFU-E 110 Ϯ 5 (10) 0.67 Ϯ 0.15 (12) 34 Ϯ 4 (6)c 0.70 Ϯ 0.1 (10)d CFU-GM 110 Ϯ 10 (10) 4.5 Ϯ 2.9 (12) Ͼ400 (6)c 1.6 Ϯ 0.3 (10)d Liver tumor cell lines (HBV producing) 2.2.15 130 Ϯ 10 110 Ϯ 4 Ͼ200 170 Ϯ 10 HB611 Ͼ200 49 Ϯ 4 200 Ͼ200

a Six experiments for 1592U89; two experiments for AZT, ddI, and ddC. b The number of different marrow samples from different donors is given in parentheses following the IC50. c Previously reported (11). d Previously reported (63).

The Macsynergy program was used to analyze the data (43). in these cells. The metabolite levels found after a 48-h expo- Additive interactions were observed. sure are shown in Table 9. Additional analyses and compari- 1592U89 was not appreciably toxic to the HBV-producing sons to CBV for the purpose of elucidating anabolic pathways human liver tumor cell lines used to assay for anti-HBV activity are contained in the following paper (14). At the nucleoside (Table 8). level, 1592U89 was the dominant species. Low levels of CBV The clinically observed hematopoietic toxicity of some and aminoCBV (compound 3) were detected but represented nucleosides, including AZT, has been correlated with the abil- Ͻ2% of the intracellular level of 1592U89. 1592U89-MP was ity to inhibit colony formation of human bone marrow progen- present, but 1592U89-DP and 1592U89-TP were not detected itors in vitro (10). Our results show that 1592U89 is much less (Ͻ0.0025 ␮M). The only radiolabelled DPs and TPs detected toxic than AZT (Table 8). The average IC50s of 1592U89 for were those of CBV. Consistent with observations in the rat, the granulocyte-macrophage CFU (CFU-GM) and erythroid nucleobase 2-amino-6-(cyclopropylamino)-9H-purine was not burst-forming unit (BFU-E) progenitors (both 110 ␮M) were detected. well above concentrations of 1592U89 which provide anti-HIV Dose response of CBV-TP formation. The levels of CBV-TP activity. We found that CBV was somewhat more toxic than produced following a 48-h incubation of CD4ϩ CEM cells with 1592U89 to CFU-GMs (IC50,50Ϯ8␮M) and approximately various concentrations of 1592U89 and CBV are shown in Fig. equivalently toxic to BFU-Es (IC50, 100 Ϯ 13 ␮M). Previously 3. CBV-TP levels of 1592U89 (0.004 to 9.4 ␮M) and from CBV published studies have established the low in vitro toxicity of (0.008 to 9.1 ␮M) increased in proportion to the extracellular CBV to marrow progenitors (12, 27). concentration between 0.1 and 100 ␮M. The levels of CBV-TP Human CD4ϩ CEM cells in log phase were incubated with produced from 1592U89 were not significantly different than 1592U89 over a 48-h time course to assess the effect of 1592U89 those produced from equimolar CBV, except at the 0.1 ␮M on normal DNA synthesis. The incorporation of [3H]thymidine into TCA-insoluble material was used as a quantitative measure of DNA synthesis. 1592U89 at concentrations of 1 and 10 ␮ TABLE 9. Levels of intracellular anabolites in CD4ϩ CEM cells M had no effect on the incorporation of thymidine into DNA 3 at the time points examined (4, 24, and 48 h). With 100 ␮M incubated for 48 h with 10 ␮M [8- H]1592U89 1592U89, a small (32%) decrease in thymidine incorporation Mean intracellular concn Type and compound Ϯ SE (␮M) (no. of sam- was noted at 4 h. This decrease was transient, as cells treated a with 100 ␮M 1592U89 incorporated [3H]thymidine at control ples analyzed) levels at both the 24- and 48-h time points. Nucleosides Since the in vitro growth of Molt-4 human leukemic cells was 1592U89 (compound 1) ...... 16 Ϯ 1 (7) the most sensitive of the lines tested to inhibition by 1592U89 CBV (compound 2)...... 0.38 Ϯ 0.02 (4) (Table 8), we chose this line for further evaluation of possible AminoCBV (compound 3) ...... 0.19 Ϯ 0.01 (4) effects on mitochondrial DNA synthesis as described elsewhere (8, 33). The ratio of mitochondrial DNA to genomic DNA in MPs Ϯ ␮ 1592U89-MP...... 0.13 0.04 (4) Molt-4 cells incubated with 0.10 to 100 M 1592U89 for up to CBV-MP ...... 0.016 Ϯ 0.007 (4) 8 days did not change significantly from the control. In ad- AminoCBV-MP...... NQ dition, mitochondrial DNA did not decrease in 1592U89- exposed cells compared to untreated same-day controls. In DPs and TPs contrast, ddC in this assay had an IC50 for depletion of mito- CBV-DP...... 0.071 Ϯ 0.010 (7) chondrial DNA of 0.002 ␮M after 5 days of exposure and was CBV-TP...... 0.28 Ϯ 0.02 (7) toxic to cells in longer exposures (33). 1592U89-DP ...... ND Intracellular metabolites from 1592U89. CD4ϩ CEM cells 1592U89-TP...... ND were incubated with 10 ␮M 1592U89, a concentration chosen a NQ, not quantifiable due to contaminating peak; ND, not detectable to be approximately twice the IC50 (3.8 ␮M) for HIV-1 (IIIB) (Ͻ0.0025 ␮M). VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1089

FIG. 3. Intracellular levels of CBV-TP in uninfected CD4ϩ CEM cells continuously exposed to a range of concentrations of 1592U89 (F) or CBV (ᮀ) for 48 h. Numbers in parentheses indicate the number of replicates at each concentration. For 1592U89, triplicate determinations at 0.1, 1.0, and 10 ␮M used [8-3H]1592U89 and triplicate determinations at 1.0, 10, and 100 ␮M used [5Ј-3H]1592U89. Values are means Ϯ standard deviations. concentration (P ϭ 0.04 by Student’s t test). Others have also infected CD4ϩ CEM cells were 0.79 Ϯ 0.05 and 0.76 Ϯ 0.03 reported nonsaturation of CBV anabolism over a 0.1 to 100 ␮M, respectively. The overall higher levels of CBV-TP pro- ␮M concentration range (39). duced from 1592U89 observed in this experiment compared Intracellular half-life of CBV-TP. CD4ϩ CEM cells incu- to those shown in Table 9 and Fig. 4 may be reﬂective of the bated with 10 ␮M[3H]1592U89 or [3H]CBV for 48 h accumu- different conditions required for growth and maintainence of lated CBV-TP to levels of 0.28 and 0.38 ␮M, respectively. The the CD4ϩ CEM cells in the relatively short anabolism exper- elimination kinetics of CBV-TP produced from either 1592U89 iments compared to the more lengthy experiments to deter- or CBV were essentially identical (Fig. 4). Disappearance of mine antiviral activity. intracellular CBV-TP was monophasic, with a half-life of 3.3 h, Effects of 1592U89 on intracellular levels of dNTPs. The in good agreement with a reported half-life of 2.5 h for CBV- intracellular concentrations of dCTP, dTTP, dATP, and dGTP TP from CBV (39). in uninfected CD4ϩ CEM cells were 15, 42, 59, and 23 ␮M, Effect of HIV infection on 1592U89 anabolism. Infection of respectively. These levels are in good agreement with reported CD4ϩ CEM cells with HIV-1 had no effect on the anabolism of values in this cell line (8). Exposures to 10 ␮M 1592U89 for 24 1592U89 to CBV-TP at 24 h. The levels of CBV-TP produced and 48 h did not signiﬁcantly affect the dNTP pools compared from 10 ␮M[5Ј-3H]1592U89 in uninfected and HIV-1 (IIIB)- to the levels in untreated cells (P Ͼ 0.45 and P Ͼ 0.08 at 24 and

FIG. 4. Elimination kinetics of CBV-TP from uninfected CD4ϩ CEM cells exposed to 10 ␮M [8-3H]1592U89 (F)or[3H]CBV (ᮀ). Values are means Ϯ standard deviations of triplicate determinations. 1090 DALUGE ET AL. ANTIMICROB.AGENTS CHEMOTHER.

TABLE 10. CBV-TP levels generated from [5Ј-3H]1592U89 RT, confirming previous reports of the selectivity of CBV-TP compared to AZT-TP from [3H]AZT and corresponding (40). competing dNTP pools in unstimulated normal human PBL cultures DISCUSSION Compound-TP Competing Ratio (compound- Sample (h) (␮M)a dNTP (␮M)b TP/dNTP) We conclude that 1592U89 is a potent and selective inhibitor of HIV replication in vitro. 1592U89 was strongly synergistic 1592U89 (6) 0.013 Ϯ 0.001 0.25 Ϯ 0.04c 0.05 with AZT, the nonnucleoside nevirapine, and the protease 1592U89 (24) 0.016 Ϯ 0.001 0.17 Ϯ 0.02 0.09 inhibitor 141W94 against HIV-1 (IIIB) in MT4 cells. Combina- tions of 1592U89 with 3TC, ddI, ddC, or d4T were additive to Ϯ c Ϯ AZT (6) 0.033 0.003 0.46 0.03 0.07 synergistic. The attractive cross-resistance profile of 1592U89 AZT (24) 0.043 Ϯ 0.005 0.22 Ϯ 0.01 0.20 also encourages combination in the clinic with a variety of a Concentrations of CBV-TP (from 0.5 ␮M 1592U89) or AZT-TP (from 0.5 ␮M other agents. HIV-2, the distantly related lentivirus FIV, and AZT) are the means Ϯ standard errors of a total of six samples from three donors. the hepadnavirus HBV were also inhibited by this nucleoside b Concentrations of dGTP (for 1592U89 incubations) or dTTP (for AZT incubations) are the means Ϯ standard errors. analog. Interestingly, HCMV also exhibited some sensitivity to c Five samples from three donors. 1592U89, although several other members of the herpesvirus group did not. While there was essentially no difference between the levels 48 h, respectively). All groups, including controls, had lower of activity of 1592U89, ddI, and AZT against clinical isolates of dNTP pools at 48 h, which may be related to a difference in the HIV-1 in PBLs, an exaggerated difference in relative ranking growth state or cell density at that time compared to the 24-h of these compounds was apparent with strain IIIB. 1592U89 time point. In the untreated controls at 48 h, concentrations of and ddI were significantly less active than AZT against HIV-1 dCTP, dTTP, dATP, and dGTP were 7.0, 25, 36, and 13 ␮M, (IIIB) in a variety of cell lines and assay types. While HIV-1 respectively. (IIIB) is a standard strain for drug discovery and preclinical Anabolism of 1592U89 compared to AZT in human PBLs. evaluation, these differences underscore the predictive limits TP levels, as well as the levels of their corresponding compet- of any one in vitro anti-HIV assay system. Assays with clinical ing endogenous dNTPs, dGTP and dTTP, were measured in isolates in PBL culture may be the more relevant system, normal, unstimulated human PBLs after incubation with although it is often not possible to predict the relative clinical 1592U89 and AZT (Table 10). The concentration of 1592U89 efficacy of experimental drugs based on in vitro antiviral data. and AZT used in these studies was 0.5 ␮M, well within the 1592U89, which was converted to a guanine NTP analog in range of expected in vivo exposure in humans and close to the cells, should complement the approved nucleoside RT inhibitors containing other nucleobases: thymine (AZT and d4T), average IC50s for clinical isolates in PBL cultures (Table 6). The ratio of CBV-TP to dGTP was similar to the ratio found cytosine (ddC and 3TC), and adenine (from ddI). In the sugar portion, CBV-TP resembles the TPs generated from D-2Ј,3Ј- for AZT-TP to dTTP and to the corresponding Ki/Km ratios for HIV-1 RT (0.083 for CBV-TP and 0.052 for AZT-TP). The dideoxynucleoside inhibitors (ddC, ddI, and d4T) both in ab- levels of CBV-TP generated from 1592U89 in the clinically solute configuration and in the absence of 3Ј substitution. The relevant human PBL thus appear sufficient to compete with in vitro resistance profile of 1592U89 resembles that of D-2Ј,3Ј- endogenous dGTP for incorporation by HIV RT into DNA. dideoxynucleosides in that relatively slow selection is observed for mutants that differ little in their sensitivity to drug. Specif- HIV-1 RT and human DNA polymerases. Ki values for both CBV-TP and 1592U89-TP were determined with RT from ically, the single mutations seen with in vitro passaging exper- clone HXB2 of strain HIV-1 (IIIB) by using defined-sequence iments in the presence of 1592U89 result in viruses with three- as well as wild-type template primers (Table 11). CBV-TP was to fourfold reduced sensitivity (58, 60). This profile of low-level a potent competitive inhibitor with respect to dGTP for RT- resistance contrasts with that of RT inhibitors that differ sig- catalyzed DNA synthesis. With a calf thymus DNA template nificantly in structure from the endogenous dNTP substrates primer, the K value for CBV-TP was 0.021 ␮M, 1/10 the K (nonnucleosides and L-nucleosides such as 3TC) and rapidly i m select for mutants with profoundly lowered (100-fold or more) for dGTP (0.26 ␮M). The RT Ki/Km ratio with calf thymus template primer for CBV-TP (0.083) compares favorably with sensitivity to drug. Significantly, mutations which correlate with those for other ddNTPs assayed under identical conditions: substantial resistance to AZT are quite distinct from those AZT-TP (0.052 [8]) ddC-TP (0.67 [8]), ddATP (0.064), and which reduce sensitivity to 1592U89 and other D-2Ј,3Ј-dideoxy- Ϫ1 3TC-TP (0.52). The kcat value of 0.040 s for incorporation of CBV-TP into a defined-sequence template primer (r44:d22- mer) is similar to the values for dGTP and AZT-TP of 0.047 TABLE 11. Inhibition of HIV-1 RT by CBV-TP and 1592U89-TP Ϫ1 and 0.025 s , respectively (37, 47). Mean Template Sub- K In contrast, 1592U89-TP did not significantly inhibit RT- Inhibitor K Ϯ SE m K /K primer strate i (␮M) i m catalyzed incorporation of dGMP into a defined-sequence (␮M) template primer but did inhibit dAMP incorporation. Thus, r44:d22 dGTP CBV-TPa 0.54 Ϯ 0.08 0.15 3.6 1592U89-TP appeared to be accepted by RT as a dATP ana- dGTP 1592U89-TP Ͼ500 0.15 Ͼ3,300 log. The Ki/Km value of 16 for 1592U89-TP indicated that it is a relatively inefficient inhibitor of RT. r44:d24 dATP 1592U89-TP 2.1 Ϯ 0.4 0.13 16 The Ki values of CBV-TP for mammalian DNA polymerases ␣, ␤, ␥, and ε (Table 12) were determined with the same Poly(rC)- dGTP CBV-TPa 0.04 Ϯ 0.01 12 0.0033 activated calf thymus DNA template primer under identical oligo(dG) dGTP 1592U89-TP Ͼ500 12 Ͼ42 conditions of pH, ionic strength, and divalent metal ion concentration as in the RT assay with this template primer. The Calf thymus dGTP CBV-TP 0.021 Ϯ 0.003 0.26 0.083 ε DNA Ki values for the DNA polymerases ␣, ␤, ␥, and were 90-, 2,900-, 1,200-, and 1,900-fold higher, respectively, than for HIV a Data from reference 37. VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1091

TABLE 12. Ki values for CBV-TP and 1592U89-TP against human DNA polymerases

DNA poly- Mean K Ϯ SE Mean K Ϯ SE Selectivity Substrate Inhibitor i m K /K merase (␮M) (␮M) i m for RT ␣ dGTP CBV-TP 6.9 Ϯ 0.9 0.90 Ϯ 0.05 7.7 90 dATP 1592U89-TP 260 Ϯ 60 1.3 Ϯ 0.1 200 ␤ dGTP CBV-TP 340 Ϯ 30 1.4 Ϯ 0.1 240 2,900 dATP 1592U89-TP 420 Ϯ 40 1.2 Ϯ 0.1 350 ␥ dGTP CBV-TP 14 Ϯ 2 0.14 Ϯ 0.01 100 1,200 dATP 1592U89-TP 250 Ϯ 40 0.16 Ϯ 0.02 1,600 ε dGTP CBV-TP 410 Ϯ 80 2.5 Ϯ 0.3 160 1,900 dATP 1592U89-TP 850 Ϯ 90 3.0 Ϯ 0.2 280

nucleoside analogs (28). In addition, three 1592U89-selected Some of the differences between CBV and 1592U89 are the mutations (L74V, M184V, and K65R) (58) have been reported result of differences in physical properties. CBV is relatively to revert AZT resistance (35, 56, 59), indicating that 1592U89 hydrophilic (log P, Ϫ0.62) but has limited water solubility at may be able to suppress or slow the development of resistance pH 7 (ca. 3 mM at 25°C [61]) and precipitates readily from to AZT. Thus, there is a strong rationale for combination of aqueous solutions. CBV crosses cell membranes slowly, utiliz- 1592U89 with AZT. ing a saturable nucleobase carrier to permeate CD4ϩ CEM ϩ 1592U89 was anabolized in infected and uninfected CD4 cells (30, 31, 31a). In contrast, 1592U89 is a weak base (pKa CEM cells and in uninfected PBLs to CBV-TP, the only TP 5.01) which, although not protonated at pH 7, is very water observed. In CD4ϩ CEM cells, an almost linear correlation soluble (Ͼ80 mM at 25°C). The 6-cyclopropylamino substitu- was observed between concentration of 1592U89 in the me- tion of 1592U89, by aromatizing the purine and removing the dium and intracellular CBV-TP level over a wide range of crystal lattice forces provided by H-bonding interactions of exposures (0.1 to 100 ␮M). Similar concentrations of CBV-TP the guanine lactam (42), resulted in the highly desirable were achieved in CEM cells from equivalent exposures to improvement in aqueous solubility. Fortunately, the guanine 1592U89 and CBV (Fig. 3), consistent with their similar anti- remains masked until intracellular phosphorylation (which viral potencies and suggesting that the anabolic pathway from improves solubility) occurs. The cyclopropylamino modifica- 1592U89 to CBV-TP is relatively efficient, although indirect. tion of 1592U89 also resulted in a large increase in lipophilicity The elimination half-life of CBV-TP formed from 1592U89 in (log P, 1.22) compared to CBV. 1592U89 is significantly more CEM cells was 3.3 h, which is equivalent to that reported for lipophilic than AZT (log P, 0.09), the most lipophilic of the AZT in CEM cells (21). CBV-TP was a selective inhibitor of clinically approved nucleoside RT inhibitors. Like AZT, HIV-RT over mammalian DNA polymerases ␣, ␤, ␥, and ε 1592U89 enters human erythrocytes and CD4ϩ CEM cells by when assayed under identical conditions with the same calf rapid nonfacilitated diffusion (31, 31a). The ability of 1592U89 thymus template primer. At exposures about twice the average to efficiently cross membranes may contribute to the high oral IC50 for a clinical isolate (0.5 ␮M), the ratio of CBV-TP to bioavailability observed in all species studied to date. In addi- dGTP (0.05 to 0.09) in unstimulated PBLs was similar to the tion to results reported for the rat (Table 3), excellent absolute Ki/Km ratio for CBV-TP inhibition of RT (0.083). Studies are oral bioavailability of 1592U89 has been seen after adminis- in progress to elucidate the effects of PBL stimulation on the tration of the succinate salt (14 mg/kg) to mice (92%) and novel activation pathway of 1592U89 in human lymphocytes. monkeys (77%) (19). 1592U89 penetrated rat brain and mon- In applying our preliminary selection criteria, we avoided key CSF as well as AZT, the only approved anti-HIV therapy substrates of adenosine deaminase such as aminoCBV, reason- with proven clinical benefits in treatment of CNS manifesta- ing that enhanced oral delivery of CBV would not compensate tions of the disease. CNS penetration by 1592U89, as well as for other deficiencies of CBV, such as (i) low solubility com- efficient penetration of other compartments harboring HIV, bined with high rates of renal elimination, resulting in precip- should be facilitated by its higher lipophilicity combined with itation in tubules and kidney damage in multiple species (mar- excellent water solubility. mosets and rodents) at relatively low doses (61); (ii) multifocal Modified aglycones are of concern because of possible in- myocardial necroses at relatively low doses (intravenous) in corporation into DNA. N6-Cycloalkylaminopurines are cyto- rodents (61); (iii) poor central nervous system (CNS) penetra- toxic (57). Structurally related 2Ј,3Ј-dideoxynucleoside analogs tion. In seeking active analogs that were not substrates for of 1592U89 have an unacceptable profile of in vivo toxicities adenosine deaminase and thus were not prodrugs of CBV, we (26). These ribosides produced substantial levels of agylcones succeeded in finding a compound that corrects for the prob- in animals. However, the nucleobase of 1592U89, 2-amino-6- lems seen with CBV without loss of its promising anti-HIV cyclopropylamino-[9H]-purine, was not observed in cells or in profile. We also discovered an activation pathway of possibly animals treated with 1592U89. This is consistent with the sta- general relevance to the delivery of NTPs. The distinctive an- bility predicted for carbocyclic nucleosides. Thus, the carbocy- tiviral profile of 1592U89 is attributable to two new enzymes: clic nature of 1592U89 is likely an important and necessary an adenosine phosphotransferase that phosphorylates 1592U89 contributor to its favorable overall profile, allowing modifica- to its MP and a previously unrecognized cytosolic enzyme activity tion of the nucleobase in order to alter the properties of the that converts 1592U89-MP to CBV-MP (14). Other 6-modified parent guanine without the risk of generating possibly toxic analogs (e.g., compounds 7 to 9) were also active without gen- nucleobases in vivo. erating CBV in vivo. Although these analogs were not suffi- 1592U89 was relatively nontoxic in a variety of human leu- ciently stable in vivo, the potential utility of this novel pathway kemic cell lines of T, B, and monocyte lineage and in HBV- for the delivery of NTPs should be investigated more broadly. producing human liver tumor cell lines. 1592U89 did not in- 1092 DALUGE ET AL. ANTIMICROB.AGENTS CHEMOTHER. hibit mitochondrial DNA synthesis in Molt-4 cells at the highest 7. Daluge, S. M., S. S. Good, M. T. Martin, S. R. Tibbels, W. H. Miller, D. R. concentration tested (100 ␮M). Additionally, 1592U89 did not Averett, M. H. St. Clair, and K. M. Ayers. 1994. 1592U89 Succinate—a novel carbocyclic nucleoside analogue with potent, selective anti-HIV activity, alter intracellular dNTP pools or affect thymidine incorporation abstr. I6, p. 7. In Abstracts of the 34th Interscience Conference on Antimi- into DNA during exposures to therapeutically relevant 1592U89 crobial Agents and Chemotherapy. American Society for Microbiology, concentrations (10 ␮M). In vitro evaluation of human bone mar- Washington, D.C. row progenitors indicated a relatively low potential for inhibition 8. Daluge, S. M., D. J. M. Purifoy, P. M. Savina, M. H. St. Clair, N. R. Parry, I. K. Dev, P. Novak, K. M. Ayers, J. E. Reardon, G. B. Roberts, J. A. Fyfe, of hematopoiesis (IC50sof110␮M for BFU-Es and CFU-GMs). M. R. Blum, D. R. Averett, R. E. Dornsife, B. A. Domin, R. Ferone, D. A. Thirty- and 90-day oral toxicity studies with 1592U89 succinate in Lewis, and T. A. Krenitsky. 1994. 5-Chloro-2Ј,3Ј-dideoxy-3Ј-fluorouridine mice (110, 330, and 1,000 mg/kg/day) and monkeys (50, 140, and (935U83), a selective anti-human immunodeficiency virus agent with an 420 mg/kg/day) identified the “no observable toxic effect” levels as improved metabolic and toxicological profile. Antimicrob. Agents Chemo- ther. 38:1590–1603. 110 mg/kg/day (mice) and 140 mg/kg/day (monkeys). Maximal 9. Davis, M. G., and R. W. Jansen. 1994. Inhibition of hepatitis B virus in tissue concentrations in plasma achieved in animals receiving the no- culture by alpha interferon. Antimicrob. Agents Chemother. 38:2921–2924. effect doses were approximately 100-fold above the average clin- 10. Dornsife, R. E., and D. R. Averett. 1996. In vitro potency of inhibition by ical isolate IC s. Importantly, toxicities observed with other nu- antiviral drugs of hematopoietic progenitor colony formation correlates with 50 exposure at hemotoxic levels in human immunodeficiency virus-positive hu- cleoside RT inhibitors—peripheral neuropathy and renal, mans. Antimicrob. Agents Chemother. 40:514–519. cardiac, and hematopoietic toxicities—were not observed in these 11. Dornsife, R. E., M. H. St. Clair, A. T. Huang, T. J. Panella, G. W. Koszalka, studies, even at the highest doses (3, 18). C. L. Burns, and D. R. Averett. 1991. Anti-human immunodeficiency virus Recently reported pharmacokinetics (18, 23, 34) and prelimi- synergism by zidovudine (3Ј-azidothymidine) and didanosine (dideoxyinosine) contrasts with their additive inhibition of normal human marrow nary viral load and CD4 cell count improvements (51, 52) in progenitor cells. Antimicrob. Agents Chemother. 35:322–328. HIV-infected patients support continued clinical development of 12. Du, D. L., D. A. Volpe, C. K. Grieshaber, and M. J. Murphy, Jr. 1992. In vitro 1592U89. Viral load reductions and CD4 cell count increases toxicity of 3Ј-azido-3Ј-deoxythymidine, carbovir and 2Ј,3Ј-didehydro-2Ј,3Ј- seen with the lowest dose of 1592U89 tested (200 mg, three times dideoxythymidine to human and murine haematopoietic progenitor cells. Br. J. Haematol. 80:437–445. a day) were of greater magnitude and duration than those seen in 13. Faletto, M. B., W. H. Miller, E. P. Garvey, J. E. Reardon, and S. S. Good. studies with AZT and other nucleosides in comparable patient 1994. Unique intracellular activation of a new anti-HIV agent (1S,4R)-4- populations. In this paper we have summarized our preclinical [2-amino-6-(cyclopropylamino)-9H-purin-9-yl]-2-cyclopentene-1-methanol understanding of 1592U89. Further details are presented in the (1592U89) in the human T-lymphoblastoid cell line CEM-T4, abstr. I84, p. 92. following two papers (14, 58). This research provides encourage- In Abstracts of the 34th Interscience Conference on Antimicrobial Agents and Chemotherapy. American Society for Microbiology, Washington, D.C. ment that significant improvements over the approved nucleoside 14. Faletto, M. B., W. H. Miller, E. P. Garvey, M. H. St. Clair, S. M. Daluge, and RT inhibitors can be made and that appropriate nucleoside ther- S. S. Good. 1997. Unique intracellular activation of the potent anti-human apies have untapped potential in the treatment of HIV disease. immunodeficiency virus agent 1592U89. Antimicrob. Agents Chemother. 41: 1099–1107. ACKNOWLEDGMENTS 15. Fischl, M. A., D. D. Richman, N. Hansen, A. C. Collier, et al. 1990. The safety and efficacy of zidovudine (AZT) in the treatment of subjects with mildly We thank M. T. Martin and B. R. Sickles for synthesis of interme- symptomatic human immunodeficiency virus type I (HIV) infection. Ann. diates to 1592U89 and analogs; S. D. Gabriel and J. A. Hill for syn- Intern. Med. 112:727–737. 3 16. Gadler, H. 1983. Nucleic acid hybridization for measurement of effects of thesis of the [5Ј- H]1592U89; D. Minick for pKa and lipophilicity measurements; T. Spector, A. Resetar, and J. Yale for adenosine antiviral compounds on human cytolomegalovirus DNA replication. Anti- microb. Agents Chemother. 24:370–374. deaminase evaluations; E. H. Dark and B. Gallagher for cytotoxicity 17. Garrett, C., and D. V. Santi. 1979. A rapid and sensitive high pressure liquid testing with human leukemic cell lines; A. Tanner and C. Hopson for chromatography assay for deoxyribonucleoside triphosphates in cell extracts. technical assistance with human bone marrow progenitor cytotoxicity Anal. Biochem. 99:268–273. testing; D. A. Lewis and J. L. Martin for mitochondrial DNA assays; 18. Good, S. S., S. M. Daluge, S. V. Ching, K. M. Ayers, W. B. Mahony, M. B. K. K. Biron, S. Stanat, and R. J. Harvey for HCMV assays; J. W. T. Faletto, B. A. Domin, B. S. Owens, R. E. Dornsife, J. A. McDowell, S. W. Selway and E. Littler for testing with other herpesviruses; R. W. LaFon, and W. T. Symonds. 1995. 1592U89 succinate—preclinical toxico- Jansen, L. C. Johnson, and L. Condreay for HBV assays; R. Hazen and logical and disposition studies and preliminary clinical pharmacokinetics. I. Najera for some of the HIV assays; D. W. Selleseth and M. N. Ellis Antiviral Res. 26:A229. 19. Good, S. S., B. S. Owens, M. B. Faletto, W. B. Mahony, and B. A. Domin. for FIV assays; and K. M. Ayers, R. L. Miller, W. B. Mahony, B. A. 1994. Disposition in monkeys and mice of (1S,4R)-4-[2-amino-6-(cyclopro- Domin, S. W. LaFon, and G. W. Koszalka for helpful discussions. We pylamino)-9H-purin-9-yl]-2-cyclopentene-1-methanol (1592U89) succinate, especially thank J. L. Rideout for support and L. W. Frick and D. a potent inhibitor of HIV, abstr. I86, p. 92. In Abstracts of the 34th Inter- Porter for their critical reviews of the manuscript. science Conference on Antimicrobial Agents and Chemotherapy. American Society for Microbiology, Washington, D.C. REFERENCES 20. Ho, D. H. W., C. Pincus, C. J. Carter, R. S. Benjamin, E. J. Freireich, and 1. Averett, D. R. 1989. Anti-HIV compound assessment by two novel high G. P. Bodey. 1980. Distribution and inhibition of adenosine deaminase in capacity assays. J. Virol. Methods 23:263–276. tissues of man, rat, and mouse. Cancer Treatment Rep. 64:629–633. 2. Brouwer, K. R., R. L. St. Claire, J. Lagarde, J. E. Patanella, J. S. Walsh, and 21. Ho, H.-T., and M. J. M. Hitchcock. 1989. Cellular pharmacology of 2Ј,3Ј- G. T. Miwa. 1990. The pharmacokinetics of (Ϫ)-carbovir in rats: evidence for dideoxy-2Ј,3Ј-didehydrothymidine, a nucleoside analog active against human nonlinear elimination. Drug Metab. Dispos. 18:1078–1083. immunodeficiency virus. Antimicrob. Agents Chemother. 33:884–849. 3. Ching, S. V., K. M. Ayers, R. E. Dornsife, G. L. Grebe, and J. L. Howard. 22. Huang, S.-H., R. P. Remmel, and C. L. Zimmerman. 1991. The bioavailabil- 1994. Nonclinical toxicology and in vitro toxicity studies with the novel anti- ity and nonlinear clearance of (Ϫ)-carbovir in the rat. Pharm. Res. 8:739–743. HIV agent (1S,4R)-4-[2-amino-6-(cyclopropylamino)-9H-purin-9-yl]-2-cyclo- 23. Hughes, W., J. McDowell, L. Adams, P. Flynn, S. Hetherington, M. Kline, J. pentene-1-methanol (1592U89) succinate, abstr. I88, p. 92. In Abstracts of Shenep, R. Yogev, and S. LaFon. 1996. Evaluation of the novel nucleoside the 34th Interscience Conference on Antimicrobial Agents and Chemother- 1592U89 in a phase I safety and pharmacokinetics (PK) study in HIV- apy. American Society for Microbiology, Washington, D.C. infected infants and children, abstr. 195, p. 115. In Abstracts of the 3rd 4. Collins, J. M., R. W. Klecker, Jr., J. A. Kelley, J. S. Roth, C. L. McCully, Conference on Retroviruses and Opportunistic Infections. Infectious Dis- F. M. Balis, and D. J. Poplack. 1988. Pyrimidine dideoxyribo-nucleosides: eases Society of America for the foundation for Retrovirology and Human selectivity of penetration into cerebrospinal fluid. J. Pharmacol. Exp. Ther. Health, Alexandria, Va. 245:466–470. 24. Jansen, R. W., L. C. Johnson, and D. R. Averett. 1993. High-capacity in vitro 5. Daluge, S. M. January 1990. European patent 349242. assessment of anti-hepatitis B virus compound selectivity by a virion-specific 5a.Daluge, S. M. June 1991. European patent 434450. polymerase chain reaction assay. Antimicrob. Agents Chemother. 37:441–447. 5b.Daluge, S. M. Unpublished data. 25. Kavlick, M. F., T. Shirasaka, E. Kojima, J. M. Pluda, F. Hui, Jr., R. 6. Daluge, S. M., S. S. Good, M. B. Faletto, M. T. Martin, W. H. Miller, D. R. Yarchoan, and H. Mitsuya. 1995. Genotypic and phenotypic characterization Averett, M. H. St. Clair, L. R. Boone, M. Tisdale, and J. E. Reardon. 1995. of HIV-1 isolated from patients receiving (Ϫ)-2Ј,3Ј-dideoxy-3Ј-thiacytidine. 1592U89 succinate—a potent, selective anti-HIV carbocyclic nucleoside. An- Antiviral Res. 28:133–146. tiviral Res. 26:A228. 26. Koszalka, G. W., C. L. Burns, M. St. Clair, L. W. Frick, C. U. Lambe, M. T. VOL. 41, 1997 ANTI-HIV AGENT 1592U89 1093

Paff, D. J. Nelson, B. Korba, K. Ayers, and T. A. Krenitsky. 1995. Inhibition 3Ј-deoxythymidine. J. Virol. 65:308–312. of HIV- and HBV-virus replication by 2,6-disubstituted purine 2Ј,3Ј-dide- 49. Richman, D. D., M. A. Fischl, M. H. Grieco, M. S. Gottlieb, P. A. Volberding, oxynucleosides. Antiviral Res. 26:A245. O. L. Laskin, J. M. Leedom, J. E. Groopman, D. Mildvan, M. S. Hirsh, G. G. 27. Kurtzberg, J., and S. G. Carter. 1990. Differential toxicity of carbovir and Jackson, D. T. Durack, D. Phil, S. Nusinoff-Lehrman, and the AZT Collab- AZT to human bone marrow hematopoietic progenitor cells in vitro. Exp. orative Working Group. 1987. The toxicity of azidothymidine (AZT) in the Hematol. 18:1094–1096. treatment of patients with AIDS and AIDS-related complex. N. Engl. 28. Larder, B. A. 1995. Viral resistance and the selection of antiretroviral com- J. Med. 317:192–197. binations. J. Acquired Immunodefic. Syndr. Hum. Retrovirol. 10(Suppl. 1): 50. Rocancourt, D., C. Bonnerot, H. Jouin, M. Emmerman, and J.-F. Nicolas. S28–S33. 1990. Activation of a ␤-galactosidase recombinant provirus: application to 29. Larder, B. A., B. Chesebro, and D. D. Richman. 1990. Susceptibilities of titration of human immunodeficiency virus (HIV) and HIV-infected cells. zidovudine-susceptible and -resistant human immunodeficiency virus isolates J. Virol. 64:2660–2668. to antiviral agents determined by using a quantitative plaque reduction assay. 51. Saag, M., D. Lancaster, A. Sonnerbord, J. Mulder, R. Torres, R. Schooley, Antimicrob. Agents Chemother. 34:436–441. R. Harrigan, D. Kelleher, and W. Symonds. 1996. A phase I/II study of a novel 30. Mahony, W. B., B. A. Domin, S. M. Daluge, W. H. Miller, and T. P. Zim- nucleoside reverse transcriptase inhibitor, 1592U89 monotherapy vs. merman. 1992. Enantiomeric selectivity of carbovir transport. J. Biol. Chem. 1592U89 ϩ zidovudine (ZDV) or placebo in HIV infected patients with CD4 267:19792–19797. counts 200–500/mm3, abstr. 195, p. 89. In Abstracts of the 3rd Conference on 31. Mahony, W. B., B. A. Domin, K. L. Prus, and T. P. Zimmerman. 1995. Retroviruses and Opportunistic Infections. Infectious Diseases Society of Amer- Membrane permeation characteristics of the structurally related anti-HIV ica for the foundation for Retrovirology and Human Health, Alexandria, Va. agents 1592U89 and (Ϫ)-carbovir in human erythrocytes and human T- 52. Saag, M., D. Lancaster, A. Sonnerbord, R. Torres, M. Thompson, J. Mulder, ϩ lymphoblastoid CD4 CEM cells. Proc. Am. Assoc. Cancer Res. 36:2211. R. Schooley, B. Gazzard, R. D’Aquila, M. Santin, M. Gurgui, R. Harrigan, 31a.Mahony, W. B., B. A. Domin, K. L. Prus, and T. P. Zimmerman. Unpub- and S. LaFon. 1996. Preliminary data on the safety and antiviral effect of lished data. 1592U89, alone and in combination with zidovudine (ZDV) in HIV-infected 32. Marr, C. L. P., and C. R. Penn. 1992. Stability of carbovir, a potent inhibitor patients with CD4ϩ counts 200–500/mm3, abstr. Th.B.294, p. 225. In Ab- of HIV, to phosphorolysis by human purine nucleoside phosphorylase. An- stracts of the XIth International Conference on AIDS, Vancouver, Canada. tiviral Chem. Chemother. 3:121–124. 53. Sherman, P. A., and J. A. Fyfe. 1989. Enzymatic assay for deoxyribonucleo- 33. Martin, J. L., C. E. Brown, N. Matthews-Davis, and J. E. Reardon. 1994. side triphosphates using synthetic oligonucleotides as template primers. Effects of antiviral nucleoside analogs on human DNA polymerases and Anal. Biochem. 180:222–226. mitochondrial DNA synthesis. Antimicrob. Agents Chemother. 38:1230–1238. 54. Skowron, G. 1995. Biological effects and safety of stavudine: overview of 34. McDowell, J. A., W. T. Symonds, P. N. Kumar, D. E. Sweet, M. R. Blum, and phase I and II clinical trials. J. Infect. Dis. 171(Suppl. 2):S113–S117. S. W. LaFon. 1995. Initial phase-I study of anti-HIV agent 1592U89 in a 55. Spector, T., T. E. Jones, and L. M. Beacham III. 1983. Conversion of single-dose escalation design including food effect and dosage form evalua- 2,6-diamino-9-(2-hydroxyethoxymethyl)purine to acyclovir as catalyzed by tion, abstr. I109, p. 224. In Abstracts of the 35th Interscience Conference on adenosine deaminase. Biochem. Pharmacol. 32:2505–2509. Antimicrobial Agents and Chemotherapy. American Society for Microbiol- 56. St. Clair, M. H., J. L. Martin, G. Tudor-Williams, M. C. Bach, C. L. Vavro, ogy, Washington, D.C. D. M. King, P. Kellam, S. D. Kemp, and B. A. Larder. 1991. Resistance to 35. Mellors, J. W., H. Bazmi, C. K. Chu, and R. F. Schinazi. 1996. K65R ddI and sensitivity to AZT induced by a mutation in HIV-1 reverse tran- Ϫ ␤ D mutation in HIV-1 reverse transcriptase causes resistance to ( )- - -dioxo- scriptase. Science 253:1557–1559. lane-guanosine and reverses AZT resistance, abstr. 7. In 5th Workshop on 57. Thedford, R., E. O. Leyimu, D. L. Thornton, and R. Mehta. 1989. Cytotox- 1 HIV Drug Resistance. Antiviral Ther. (Suppl.1). icity of N6-alkylated adenine and adenosine analogs to mouse hepatoma 36. Merigan, T. C., and G. Skowron. 1990. Safety and tolerance of dideoxycyti- cells. Exp. Cell Biol. 57:53–59. dine as a single agent. Am. J. Med. 88(Suppl. 5B):11S–15S. 58. Tisdale, M., T. Alnadaf, and D. Cousens. 1997. Combination of mutations in 37. Miller, W. M., S. M. Daluge, E. P. Garvey, S. Hopkins, J. E. Reardon, F. L. human immunodeficiency virus type 1 reverse transcriptase required for Boyd, and R. L. Miller. 1992. Phosphorylation of carbovir enantiomers by resistance to the carbocyclic nucleoside 1592U89. Antimicrob. Agents Che- cellular enzymes determines the stereoselectivity of antiviral activity. J. Biol. mother. 41:1094–1098. Chem. 267:21220–21224. 59. Tisdale, M., S. D. Kemp, N. R. Parry, and B. A. Larder. 1993. Rapid in vitro 38. Mitsuya, H., M. Matsukura, and S. Broder. 1987. Rapid in vitro systems for selection of human immunodeficiency virus type 1 resistant to 3Ј-thiacytidine assessing activity of agents against HTLV-III/LAV, p. 309–321. In S. Broder inhibitors due to a mutation in the YMDD region of reverse transcriptase. (ed.), AIDS: modern concepts and therapeutic challenges. Marcel Dekker, 90: New York, N.Y. Proc. Natl. Acad. Sci. USA 5653–5656. Tisdale, M., N. R. Parry, D. Cousens, M. H. St. Clair, and L. R. Boone. 39. Parker, W. B., S. C. Shaddix, B. J. Bowdon, L. M. Rose, R. Vince, W. M. 60. 1994. Shannon, and L. L. Bennett, Jr. 1993. Metabolism of carbovir, a potent Anti-HIV activity of (1S,4R)-4-[2-amino-6-(cyclopropylamino)-9H-purin-9- inhibitor of human immunodeficiency virus type 1, and its effects on cellular yl]-2-cyclopentene-1-methanol (1592U89), abstr. I82, p. 92. In Abstracts of metabolism. Antimicrob. Agents Chemother. 37:1004–1009. the 34th Interscience Conference on Antimicrobial Agents and Chemother- 40. Parker, W. B., E. L. White, S. C. Shaddix, L. J. Ross, R. W. Buckheit, Jr., apy. American Society for Microbiology, Washington, D.C. J. M. Germany, J. A. Secrist III, R. Vince, and W. M. Shannon. 1991. 61. Trennery, P. (Medicines Safety Evaluation, Glaxo Wellcome plc, Ware, Mechanism of inhibition of human immunodeficiency virus type 1 reverse United Kingdom). Personal communication. transcriptase and human DNA polymerases ␣, ␤, and ␥ by the 5Ј-triphos- 62. Upton, R. A. 1975. Simple and reliable method for serial sampling of blood phates of carbovir, 3Ј-azido-3Ј-deoxythymidine, 2Ј,3Ј-dideoxyguanosine, and from rats. J. Pharm. Sci. 64:112–114. 3Ј-deoxythymidine. J. Biol. Chem. 266:1754–1762. 63. Van Draanen, N. A., M. Tisdale, N. R. Parry, R. Jansen, R. E. Dornsife, J. V. 41. Petersen, E. A., C. H. Ramirez-Ronda, W. D. Hardy, R. Schwarz, H. S. Sacks, Tuttle, D. R. Averett, and G. W. Koszalka. 1994. Influence of stereochemistry S. Follansbee, D. M. Peterson, A. Cross, R. E. Anderson, and L. M. Dunkle. on antiviral activities and resistance profiles of dideoxycytidine nucleosides. 1995. Dose-related activity of stavudine in patients infected with human Antimicrob. Agents Chemother. 38:868–871. immunodeficiency virus. J. Infect. Dis. 171(Suppl. 2):S131–S139. 64. Vince, R., M. Hua, J. Brownell, S. M. Daluge, F. C. Lee, W. M. Shannon, 42. Pfleiderer, W. 1963. The solubility of heterocyclic compounds, p. 177–182. In G. C. Lavelle, J. Qualls, O. S. Weislow, R. Kiser, P. G. Canonico, R. H. A. R. Katritzky (ed.), Physical methods in heterocyclic chemistry, vol. 1. Schultz, V. L. Narayanan, J. G. Mayo, R. H. Shoemaker, and M. R. Boyd. Academic Press, New York, N.Y. 1988. Potent and selective activity of a new carbocyclic nucleoside analog 43. Prichard, M. N., and C. Shipman, Jr. 1990. A three-dimensional model to (carbovir: NSC 614846) against human immunodeficiency virus in vitro. analyze drug-drug interactions (“Macsynergy™”). Antiviral Res. 14:181–206. Biochem. Biophys. Res. Commun. 156:1046–1053. 44. Prus, K. L. (Glaxo Wellcome Inc., Research Triangle Park, N.C.). Personal 65. Wainberg, M. A., H. Salomon, Z. Gu, J. S. Montaner, T. P. Cooley, R. communication. McCaffrey, J. Ruedy, H. M. Hirst, N. Cammack, J. Cameron, et al. 1995. 45. Prus, K. L., D. R. Averett, and T. P. Zimmerman. 1990. Transport and Development of HIV-1 resistance to (Ϫ)-2Ј-deoxy-3Ј-thiacytidine in patients metabolism of 9-␤-D-arabinofuranosylguanine in a human T-lymphoblastoid with AIDS or advanced AIDS-related complex. AIDS (Philadelphia) 9:351–357. cell line: nitrobenzylthioinosine-sensitive and -insensitive influx. Cancer Res. 66. Walsh, J. S., J. E. Patanella, S. E. Unger, K. R. Brouwer, and G. T. Miwa. 50:1817–1821. 1990. The metabolism and excretion of carbovir, a carbocyclic nucleoside, in 46. Reardon, J. E. 1992. Human immunodeficiency virus reverse transcriptase: the rat. Drug Metab. Dispos. 18:1084–1091. steady-state and pre-steady-state kinetics of nucleotide incorporation. Bio- 67. Yarchoan, R., P. Brouwers, J. M. Pluda, K. M. Wyvill, R. V. Thomas, N. chemistry 31:4473–4479. Hartman, H. Mitsuya, D. G. Johns, and S. Broder. 1990. Long-term toxicity/ 47. Reardon, J. E., E. S. Furfine, and N. Cheng. 1991. Human immunodeficiency activity profile of 2Ј,3Ј-dideoxyinosine in AIDS or AIDS-related complex. virus reverse transcriptase substrate and inhibitor kinetics with thymidine Lancet 336:526–529. 5Ј-triphosphate and 3Ј-azido-3Ј-deoxythymidine 5Ј-triphosphate. J. Biol. 68. Zimmerman, T. P., W. B. Mahony, and K. L. Prus. 1987. 3Ј-Azido-3Ј- Chem. 266:14128–14134. deoxythymidine—an unusual analog that permeates the membrane of hu- 48. Remington, K. M., B. Cheseboro, K. Wehrly, N. C. Pedersen, and T. W. man erythrocytes and lymphocytes by nonfacilitated diffusion. J. Biol. Chem. North. 1991. Mutants of feline immunodeficiency virus resistant to 3Ј-azido- 262:5748–5754.