Purine nucleoside phosphorylase. 2. Catalytic mechanism.

X-ray crystallography, molecular modeling, and site-directed mutagenesis were used to delineate the catalytic mechanism of purine nucleoside phosphorylase (PNP). PNP catalyzes the reversible phosphorolysis of purine nucleosides to the corresponding purine base and ribose 1-phosphate using a substrate-assisted catalytic mechanism. The proposed transition state (TS) features an oxocarbenium ion that is stabilized by the cosubstrate phosphate dianion which itself functions as part of a catalytic triad (Glu89-His86-PO4=). Participation of phosphate in the TS accounts for the poor hydrolytic activity of PNP and is likely to be the mechanistic feature that differentiates phosphorylases from glycosidases. The proposed PNP TS also entails a hydrogen bond between N7 and a highly conserved Asn. Hydrogen bond donation to N7 in the TS stabilizes the negative charge that accumulates on the purine ring during glycosidic bond cleavage. Kinetic studies using N7-modified analogs provided additional support for the hydrogen bond. Crystallographic studies of 13 human PNP-ligand complexes indicated that PNP uses a ligand-induced conformational change to position Asn243 and other key residues in the active site for catalysis. These studies also indicated that purine nucleosides bind to PNP with a nonstandard glycosidic torsion angle (+anticlinal) and an uncommon sugar pucker (C4'-endo). Single point energy calculations predicted the binding conformation to enhance phosphorolysis through ligand strain. Structural data also suggested that purine binding precedes ribose 1-phosphate binding in the synthetic direction whereas the order of substrate binding was less clear for phosphorolysis. Conservation of the catalytically important residues across nucleoside phosphorylases with specificity for 6-oxopurine nucleosides provided further support for the proposed catalytic mechanism.

Purine nucleoside phosphorylase. 3. Reversal of purine base specificity by site-directed mutagenesis.

Human purine nucleoside phosphorylase (PNP) is highly specific for 6-oxopurine nucleosides with a catalytic efficiency (kcat/KM) for inosine 350000-fold greater than for adenosine. Crystallographic studies identified Asn243 and Glu201 as the residues largely responsible for the substrate specificity. Results from mutagenesis studies demonstrated that the side chains for both residues were also essential for efficient catalysis [Erion, M. D., et al. (1997a) Biochemistry 36, 11725-11734]. Additional mechanistic studies predicted that Asn243 participated in catalysis by stabilizing the transition state structure through hydrogen bond donation to N7 of the purine base [Erion, M. D., et al. (1997b) Biochemistry 36, 11735-11748]. In an effort to alter the substrate specificity of human PNP, mutants of Asn243 and Glu201 were designed to reverse hydrogen bond donor and acceptor interactions with the purine base. Replacement of Asn243 with Asp, but not with other amino acids, led to a 5000-fold increase in kcat for adenosine and a 4300-fold increase in overall catalytic efficiency. Furthermore, the Asn243Asp mutant showed a 2.4-fold preference for adenosine relative to inosine and a 800000-fold change in substrate specificity (kcat/KM) relative to wild-type PNP. The double mutant, Asn243Asp::Glu201Gln, exhibited a 190-fold increase in catalytic efficiency with adenosine relative to wild-type PNP, a 480-fold preference for adenosine relative to inosine, and a 1.7 x 10(8)-fold change in preference for adenosine over inosine relative to wild-type PNP. The Asn243Asp mutant was also shown to synthesize 2,6-diaminopurine riboside with a catalytic efficiency (1.4 x 10(6) M-1 s-1) on the same order of magnitude as wild-type PNP with its natural substrates hypoxanthine and guanine. The Asn243Asp mutants represent examples in which protein engineering significantly altered substrate specificity while maintaining high catalytic efficiency.

1alpha,25-Dihydroxyvitamin D3 modulates CD38 expression on human lymphocytes.

B cells from chronically stimulated tonsils displayed high initial mean CD38 levels that declined during in vitro culture, despite ligation of CD40 and/or the Ag receptor in the presence of IL-4. Exposure to 1alpha,25-dihydroxyvitamin D3 (1,25(OH)2D3) restored the initial CD38 expression on these stimulated cells and up-regulated the Ag on stimulated CD38-/low cells. 1,25(OH)2D3 enhanced CD38 expression by four- to sixfold on CD8- and CD8+ peripheral blood T cells following PHA activation. The EC50 values for induction were 2 to 3 nM. Although all-trans-retinoic acid also induced CD38 expression on stimulated B and T cells, it was less effective than 1,25(OH)2D3. B cell CD38 expression was augmented less by 1,25(OH)2D3 than by IFN-alpha and IFN-gamma. However, T cell CD38 expression was induced more strongly by 1,25(OH)2D3 than by IFN-alpha, and was unaffected by IFN-gamma. The CD38 density on activated 1,25(OH)2D3-treated CD38-/low B and peripheral T cells was proportional to cell size, indicating that hormonal induction depended upon entry into the cell cycle. While IFNs induced CD38 rapidly in stimulated T and B lymphocytes, 1,25(OH)2D3 exerted its effects only after initial 1- to 3-day delays, suggesting a requirement for nuclear 1,25(OH)2D3 receptor up-regulation. In HL-60 cells, which constitutively express the nuclear receptors, 1,25(OH)2D3 rapidly induced CD38 Ag and ectoenzyme activity. The CD38 density on freshly isolated unfractionated tonsillar B lymphocytes and on the activated 1,25(OH)2D3-treated cultured cells was nearly identical for cells within the same size range, indicating that in vitro hormonal exposure reconstituted in vivo CD38 expression levels.

The synthesis and biological evaluation of 4-p-nitrobenzylthio-v-triazolo[4,5-d]pyridazine and imidazo[4,5-d]pyridazine ribosides as potential nucleoside transport inhibitors.

The synthesis of S4-substituted nucleosides possessing the imidazo- and v-triazolo[4,5-d]pyridazine ring systems was undertaken and the compounds prepared were evaluated as inhibitors of nucleoside transport into human erythrocytes. 1-(2,3,5-Tri-O-acetyl-beta-D-ribofuranosyl)-v-triazolo[4,5-d]pyridazine- 4 (5H)-thione and 1-(2,3,5-tri-O-acetyl-beta-D-ribofuranosyl)imidazo[4,5-d]pyridazine-4 (5H)-thione were each synthesized by two different routes and served as precursors for the title analogues. The nitrobenzylmercaptopurine riboside (NBMPR) analogues, 4-(p-nitrobenzylthio)-1-(beta-D-ribofuranosyl)imidazo[4,5-d]pyrida zine and 4-(p-nitrobenzylthio)-1-(beta-D-ribofuranosyl)-v-triazolo[4,5-d]py ridazine, inhibited the transport of adenosine, but were approximately 4- and 28-fold less active, respectively, than NBMPR and nitrobenzylthioformycin, known potent and specific inhibitors of carrier-mediated transport.

Adenosine deaminase inhibitors. Synthesis and biological evaluation of putative metabolites of (+)-erythro-9-(2S-hydroxy-3R-nonyl)adenine.

The synthesis and biological evaluation of three chain-hydroxylated (+)-erythro-9-(2S-hydroxy-3R-nonyl)adenine [(+)-EHNA] derivatives are reported. Hydroxy groups at positions 9', 8', and 8',9' (12, 25, and 16) were introduced by either epoxidation or hydroboration of a terminal olefinic intermediate. Affinities for calf intestinal adenosine deaminase (ADA) were determined from the steady-state inhibition of adenosine deamination. Ki values of 0.82, 3.8, 6.4, and 15.8 nM were estimated for (+)-EHNA, 9'-hydroxy-(+)-EHNA (12), 8'-hydroxy-(+)-EHNA (25), and 8',9'-dihydroxy-(+)-EHNA (16), respectively, by assuming a single class of binding sites. However, the data for all inhibitors conformed more closely to the kinetics of a heterogeneous system with different affinities for two or more binding sites. The fairly high potencies of 12 and 25 suggest that other substitutions at the terminal position of the nonyl chain could yield useful ADA inhibitors.

Adenosine deaminase inhibitors. Synthesis and biological evaluation of 4-amino-1-(2(S)-hydroxy-3(R)-nonyl)-1H-imidazo[4,5-c]pyridine (3-deaza-(+)-EHNA) and certain C1' derivatives.

The synthesis of the title compound (15) and its 1'-fluoro (14) and 1'-hydroxy (12) derivatives is described. Key intermediate 10 was obtained by two routes through condensation of (2R,3R)-3-amino-1,2-O-isopropylidene-1,2-nonanediol (3) with either 2,4-dichloro- or 4-chloro-3-nitropyridine. When assayed as adenosine deaminase inhibitors, 15 was found to be almost twice as active as its racemate. While hydroxylation at the 1'-position resulted in an 80-fold decrease in activity, the 1'-fluoro derivative proved to have activity comparable to that of 3-deaza-(+)-EHNA.

Synthesis and biological evaluation of N4-substituted imidazo- and v-triazolo[4,5-d]pyridazine nucleosides.

The chemical synthesis of certain N4-substituted imidazo[4,5-d]pyridazine and v-triazolo[4,5-d]-pyridazine nucleosides is described. In both series, the 4-chloro analogues, i.e., 4-chloro-1-(2,3,5-tri-O-acetyl-beta-D-ribofuranosyl)imidazo[4,5-d]pyr idazine (5a) and 4-chloro-1-(2,3,5-tri-O-acetyl-beta-D-ribofuranosyl)-v-triazolo[4,5- d]pyridazine (5b), were used as synthons to the target nucleosides. Nucleoside 5b was far more reactive toward nucleophilic displacements than 5a. Attempted deprotection of 5b was always accompanied with displacement of the 4-chloro substituent, whereas 5a was conveniently deacetylated without loss of the chloro group. Biological evaluation of the title nucleosides included antitumor studies and substrate/inhibition studies with certain purine-metabolizing enzymes. The corresponding adenosine analogues, i.e., 2-aza-3-deazaadenosine (6a) and 2,8-diaza-3-deazaadenosine (6b), were very slowly reacting substrates and weak inhibitors of bovine adenosine deaminase, whereas the inosine analogues were highly resistant to human purine nucleoside phosphorylase. The 4-benzylamino derivatives were weak inhibitors of adenosine transport into human erythrocytes. The inosine, adenosine, and selected N4-substituted analogues exhibited no in vitro toxicity toward murine L1210 leukemia and B16 melanoma cells.

Adenosine deaminase inhibitors. Synthesis and biological evaluation of C1' and nor-C1' derivatives of (+)-erythro-9-(2(S)-hydroxy-3(R)-nonyl)adenine.

The synthesis of various chiral derivatives of (+)-erythro-9-(2-hydroxy-3-nonyl)adenine, (+)-EHNA, from (2S,3R)-3-amino-1,2-O-isopropylidene-1,2-nonanediol by condensation with 5-amino-4,6-dichloropyrimidine is described. The compounds synthesized were C1'- and nor-C1'-(+)-EHNA derivatives. When tested with calf spleen ADA, C1'-OH- and nor-C1'-(+)-EHNA had comparable inhibitory activity that was 1 order of magnitude lower than that of (+)-EHNA. Potency was reduced further in nor-C1' derivatives.

Inhibition of nucleoside transport by nitrobenzylthioformycin analogs.

The formycin analogs of nitrobenzylthioinosine and nitrobenzylthioguanosine were synthesized and evaluated as nucleoside transport inhibitors. These analogs have a potential therapeutic advantage over their parent compounds in that their C-nucleosidic linkages prevent them from being degraded to the immunosuppressive agents, 6-mercaptopurine and 6-thioguanine. 7-[(4-Nitrobenzyl)-thio]-3-(beta-D-ribofuranosyl)pyrazolo[4,3- d]pyrimidine (NBTF) and 5-amino-7-[(4-nitrobenzyl)thio]-3-(beta-D- ribofuranosyl)pyrazolo[4,3-d]pyrimidine (NBTGF) were inhibitors of nucleoside transport in human erythrocytes and HL-60 leukemia cells. The IC50 value for nitrobenzylthioinosine, NBTF and NBTGF with 10% erythrocyte suspensions were 18, 18 and 40 nM respectively. Specific binding studies with [3H]NBTF yielded a Kd of 3.4 nM with erythrocytes, approximately 10-fold higher than values reported for nitrobenzylthioinosine. NBTF and nitrobenzylthioinosine bound to HL-60 cells with Kd values of 8.1 and 0.81 nM respectively. The octanol/water partition coefficients of nitrobenzylthioinosine, NBTF and NBTGF were 3.5, 3.2, and 2.8 respectively. NBTF could be expected to be equipotent with nitrobenzylthioinosine in whole blood where inhibitor concentrations of 10(-7) to 10(-6) M are required in order to saturate erythrocytic binding sites; hence, it may exhibit the advantages inherent in a C-nucleoside.

Three-dimensional structure of human erythrocytic purine nucleoside phosphorylase at 3.2 A resolution.

The three-dimensional structure of human erythrocytic purine nucleoside phosphorylase has been determined at 3.2 A resolution using x-ray diffraction data. Intensity data were measured using radiation from the Synchrotron Radiation Source, Daresbury, England, and oscillation film techniques. Phases were determined by using multiple isomorphous replacement methods with four heavy-atom derivatives and were improved using solvent flattening techniques. Purine nucleoside phosphorylase exists in the crystal as a trimer in which subunits are related by a crystallographic 3-fold axis. Each subunit contains an eight-stranded mixed beta-sheet and a five-stranded mixed beta-sheet which join to form a distorted beta-barrel structure. This core beta-structure is flanked by seven alpha-helices in a manner that generates a novel folding pattern. The active site, which was characterized from binding of the substrate analogs 8-iodoguanine and 5'-iodoformycin B, is located near the subunit-subunit boundary within the trimer and involves seven different segments from one subunit and an additional short segment from an adjacent subunit. In the crystal, the phosphate-binding site is probably occupied by a sulfate ion. The specificity of purine nucleoside phosphorylase for guanine, hypoxanthine, and their analogs can be explained on the basis of the arrangement of hydrogen bond donors and acceptors in the active site.

Influx of 5'-deoxy-5'-methylthioadenosine into HL-60 human leukemia cells and erythrocytes.

The influx of 5'-deoxy-5'-methylthioadenosine (MeSAdo) into human HL-60 leukemia cells and erythrocytes was characterized in order to determine whether it is facilitated by the nonspecific nucleoside carrier system or by a separate transporter, as suggested by other reports. Initial velocities were measured at room temperature by means of inhibitor-stop and oil-stop assays. MeSAdo influx was strongly inhibited by Ado, dAdo, and nucleoside transport inhibitors including nitrobenzylthioinosine and dipyridamole. Ade was inhibitory only at concentrations in excess of 1 mM. Loss of nucleoside transport capacity during differentiation of HL-60 cells was accompanied by a corresponding decrease in MeSAdo influx rates. These results indicate that MeSAdo influx was mediated by the nonspecific nucleoside transport system. The kinetic data were consistent with a single saturable carrier and yielded Km values of 74 and 184 microM and Vmax values of 424 and 48 pmols/10(6) cells/min with HL-60 cells and erythrocytes, respectively, after correction for a substantial passive diffusion component, which accounted for over 50% of the influx of 1 mM MeSAdo. The passive diffusion of MeSAdo in the presence of a transport inhibitor was not rate-limiting for the salvage of 50 microM MeSAdo to methionine when HL-60 cells were cultured in methionine-deficient medium. The large contribution of passive diffusion to the influx of MeSAdo is consistent with its unusually high octanol/water partition ratio (5.7-fold greater than that of Ado).

Inhibition of human purine nucleoside phosphorylase by acyclic nucleosides and nucleotides.

In an effort to develop more potent inhibitors of human purine nucleoside phosphorylase (PNP) as immunosuppressive and cancer chemotherapeutic agents, the affinity of the erythrocytic enzyme for 30 acyclic nucleosides, nucleotides and related compounds was determined. Among the acyclonucleosides, 2'-nordeoxyguanosine [2'NDG, 9-(1,3-dihydroxy-2-propoxymethyl)guanine] had a 3-fold greater affinity than acyclovir, and 8-amino-2'NDG was the best inhibitor with Ki = 2.6 X 10(-7) M. The ether moiety of the acyclovir and 2'NDG side-chains was not important for binding. Phosphorylated 2'NDG analogs appeared to act as multisubstrate analogs with optimal binding at low (1 mM) phosphate concentration. The 2'NDG mono- and triphosphates had higher affinities than those reported for the phosphorylated acyclovir derivatives but the diphosphate had a similar Ki value of 9 X 10(-9) M. Poor affinity, independent of phosphate concentration, was found for 9-(2-phosphonoethyl)guanine. The 3'-phosphate derivative of 8-(3-hydroxypropyl)-9-methylguanine inhibited with a Ki = 2 X 10(-5) M in 1 mM phosphate. The chemical syntheses of new analogs are described.

Design of purine nucleoside phosphorylase inhibitors.

Purine nucleoside phosphorylase inhibitors hold promise as specific immunosuppressive, anti-T cell leukemic, and antiuricopoietic agents. The best inhibitors available that are biologically active have Ki values from 10(-6) to 10(-7) M and fall into two categories: noncleavable nucleosides preferably iodinated at the C-5' position and C-8-substituted guanine or acycloguanosines. More potent inhibition is shown by phosphorylated acyclonucleosides that function as multisubstrate analogs, but these compounds are excluded from cells. The X-ray analysis of the human erythrocytic enzyme is beginning to reveal the nature of the active site and to explain the structure-activity relationships that have been established with analog substrates and inhibitors.

Changes in nucleoside transport of HL-60 human promyelocytic cells during N,N-dimethylformamide induced differentiation.

The rate of nucleoside transport decreased profoundly in human promyelocytic leukemia HL-60 cells after myeloid differentiation was induced by 5-6 days of exposure to 0.8% N,N-dimethylformamide (DMF). The facilitated diffusion of 100 microM radiolabeled adenosine and 2'-deoxyadenosine, measured by rapid transport assays, decreased 10- to 20-fold. The transport of 2 microM coformycin or 2'-deoxycoformycin, which is mediated by the same mechanism and was monitored by the adenosine deaminase titration assay, decreased 29-fold. The reduction in nucleoside transport capacity after DMF treatment was confirmed by a 19-fold decrease in the number of specific binding sites per cell (from 24-30 X 10(4) to 1.2-1.7 X 10(4)) for [3H]-6-p-nitrobenzylthioinosine, a nucleoside transport inhibitor. The binding affinity of 6-p-nitrobenzylthioinosine was not altered significantly and nucleoside transport remained sensitive to the transport inhibitors, 6-p-nitrobenzylthioinosine, dipyridamole, and dilazep after DMF-induced maturation. Time-dependence studies showed that the rate of 100 microM deoxyadenosine transport was unchanged for the first 24 h of exposure to DMF but fell to about 36% of control rates at 24-26 h and then gradually decreased further to about 4-5% of control rates after 5-6 days. In contrast, transport rates of the purine bases were reduced only 2- to 3-fold in HL-60 cells after 5 days of DMF treatment. The rates of adenosine and deoxyadenosine transport were unchanged or reduced by no more than 2-fold after 5-6 days of exposure to 0.8% DMF in the following human tumor cell lines that are not inducible with DMF: ARH-77 (multiple myeloma), KG-1 (acute myelogenous), and K-562 (chronic myelogenous). Thus, changes in nucleoside transport may serve as an early, membrane-associated marker of differentiation of the HL-60 cell line.

Inhibitors of purine nucleoside phosphorylase: effects of 9-deazapurine ribonucleosides and synthesis of 5'-deoxy-5'-iodo-9-deazainosine.

9-Deazapurine ribonucleosides constitute a new class of noncleavable purine nucleoside phosphorylase inhibitors that have at least 30-fold greater affinity for the enzyme than the corresponding C-nucleosides of the formycin B series. 9-Deazaguanosine, 9-deazainosine, and 5'-deoxy-5'-iodo-9-deazainosine competitively inhibited human erythrocytic purine nucleoside phosphorylase with Ki values of 29, 20, and 1.8 X 10(-7) M. The last compound is the most potent nucleoside inhibitor of the enzyme presently available and its synthesis is described. In contrast, 7,9-dideaza-7-thiainosine is a very weak inhibitor of the enzyme. When tested as an inhibitor of 2'-deoxyguanosine phosphorolysis in intact human erythrocytes and MOLT-3 human T-cell lymphoblastic leukemia cells, 5'-deoxy-5'-iodo-9-deazainosine was equipotent with 8-aminoguanosine (which is a precursor for 8-aminoguanine, Ki = 2 X 10(-7) M). Similarly, 5'-deoxy-5'-iodo-9-deazainosine and 8-aminoguanosine both potentiated the growth inhibition of human T-lymphocytic MOLT-3 cells by 2'-deoxyguanosine, reducing the 50% inhibitory concentration from approximately 2 X 10(-5) to approximately 2 X 10(-6) M.

5-Iodoribose 1-phosphate, an analog of ribose 1-phosphate. Enzymatic synthesis and kinetic studies with enzymes of purine, pyrimidine, and sugar phosphate metabolism.

The 5'-deoxy-5'-iodo-substituted analogs of adenosine and inosine are cytotoxic to tumor cells that have high activities of 5'-methylthioadenosine phosphorylase and purine nucleoside phosphorylase, respectively (Savarese, T.M., Chu, S-H., Chu, M.Y., and Parks, R. E., Jr. (1984) Biochem. Pharmacol. 34, 361-367). 5-Iodoribose 1-phosphate (5-IRib-1-P), the common intracellular metabolite of these 5'-iodonucleosides, has been synthesized enzymatically from 5'-deoxy-5'-iodoadenosine via adenosine deaminase from Aspergillus oryzae and human erythrocytic purine nucleoside phosphorylase. The purification and chemical properties of 5-IRib-1-P are described. The analog sugar phosphate inhibited purine nucleoside phosphorylase from human erythrocytes, phosphoglucomutase from rabbit muscle, and 5'-methylthioadenosine phosphorylase from Sarcoma 180 cells with Ki values of 26, 100, and 9 microM, respectively. Enzymes that react with 5-phosphoribosyl 1-pyrophosphate (P-Rib-PP), P-Rib-PP amidotransferase, hypoxanthine-guanine phosphoribosyltransferase, adenine phosphoribosyltransferase, and orotate phosphoribosyltransferase-orotidylate decarboxylase from extracts of Sarcoma 180 cells, were inhibited with Ki values of 49, 465, 307, and 275 microM, respectively. 5-IRib-1-P had no effect on P-Rib-PP synthetase. Since the Ki values of the analog sugar phosphate for 5'-methylthioadenosine phosphorylase and P-Rib-PP amidotransferase are much lower than the Km values of the natural substrates, Pi or P-Rib-PP which are reported to be present at nonsaturating concentrations under physiological conditions, these enzymes could be significantly inhibited by 5-IRib-1-P in intact cells.

Crystallization and preliminary x-ray investigation of purine-nucleoside phosphorylase from Escherichia coli.

Crystals of purine-nucleoside phosphorylase from Escherichia coli have been grown from solutions of ammonium sulfate. The crystals are hexagonal with space group P6(1)22 or P6(5)22; the axes are alpha = 106.5 A and c = 241.3 A. The crystals are moderately stable to x-rays and diffract beyond 3.0-A resolution. It appears that the molecule, which is a hexamer, utilizes the 2-fold symmetry of the space group, resulting in three subunits/asymmetric unit.

Differential incorporation of 2'-deoxyadenosine into human peripheral lymphocytes.

Human peripheral lymphocytes incubated with 2'-deoxycoformycin and 2'-deoxyadenosine (dAdo) reached a plateau in dATP accumulation after 4 hr that lasted for up to 24 hr. Total dATP accumulation did not exceed 15% of the control ATP concentration in the lymphocytes. In contrast, the human CCRF-CEM T lymphoblastic cell line and human erythrocytes showed a nearly linear pattern of dATP formation throughout the incubation period. By 6 hr the dATP concentration in the CCRF-CEM cells exceeded the control ATP concentration. A comparison of dATP accumulation in purified peripheral T and B lymphocytes indicated differences between these cells that favor greater dATP formation in the B lymphocytes. Incorporation studies with several adenosine analogs demonstrated that arabinosyladenine, 2-F-arabinosyladenine, tubercidin, formycin A, and 9-(2'-deoxy-2'-fluoro-beta-D-ribofuranosyl)adenine form corresponding amounts of analog triphosphate in the T and B cell-enriched lymphocytes. 9-(2'-Deoxy-2'-fluoro-beta-D-arabinofuranosyl)adenine (2'-F-araA) was the only compound to show an incorporation pattern similar to that observed with dAdo by forming analog triphosphate only in the B cell-enriched lymphocyte population. Nucleoside kinase measurements showed no significant differences in dAdo, adenosine, or 2'-deoxycytidine kinase activities between the T and B lymphocytes. The inability of the T cells to incorporate dAdo or the analog 2'-F-araA into their nucleotide pools may indicate the existence of a highly specific catabolic enzyme(s).