Inhibition of the 3' --> 5' exonuclease of human DNA polymerase epsilon by fludarabine-terminated DNA.

Incorporation of the anticancer drug fludarabine (9-β-D-arabinofuranosyl-2-fluoroadenine 5′-monophosphate; F-ara-AMP) into the 3′-end of DNA during replication causes termination of DNA strand elongation and is strongly correlated with loss of clonogenicity. Because the proofreading mechanisms that remove 3′-F-ara-AMP from DNA represent a possible means of resistance to the drug, the present study investigated the excision of incorporated F-ara-AMP from DNA by the 3′ → 5′-exonuclease activity of DNA polymerase ε from human leukemia CEM cells. Using the drug-containing and normal deoxynucleotide oligomers (21-base) annealed to M13mp18(+) DNA as the excision substrates, we demonstrated that DNA polymerase ε was unable to effectively remove F-ara-AMP from the 3′-end of the oligomer. However, 3′-terminal dAMP and subsequently other deoxynucleotides were readily excised from DNA in a distributive fashion. Kinetic evaluation demonstrated that although DNA polymerase ε has a higher affinity for F-ara-AMP-terminated DNA (Km = 7.1 pM) than for dAMP-terminated DNA of otherwise identical sequence (Km = 265 pM), excision of F-ara-AMP proceeded at a substantially slower rate (Vmax = 0.053 pmol/min/mg) than for 3′-terminal dAMP (Vmax = 1.96 pmol/min/mg). When the 3′-5′ phosphodiester bond between F-ara-AMP at the 3′-terminus and the adjacent normal deoxynucleotide was cleaved by DNA polymerase ε, the reaction products appeared to remain associated with the enzyme but without the formation of a covalent bond. No further excision of the remaining oligomers was observed after the addition of fresh DNA polymerase ε to the reaction. Furthermore, the addition of DNA polymerase α and deoxynucleoside triphosphates to the excision reaction failed to extend the oligomers. After DNA polymerase ε had been incubated with 3′-F-ara-AMP-21-mer for 10 min, the enzyme was no longer able to excise 3′-terminal dAMP from a freshly added normal 21-mer annealed to M13mp18(+) template. We conclude that the 3′ → 5′ exonuclease of human DNA polymerase ε can remove 3′-terminal F-ara-AMP from DNA with difficulty and that this excision results in a mechanism-mediated formation of “dead end complex.”

phosphate; F-ara-AMP) 1 is a major new drug in the treatment of hematologic malignancies (1)(2)(3). The action of fludarabine, mediated by its 5Ј-triphosphate metabolite F-ara-ATP, is characterized by its inhibitory effect on DNA metabolism (4,5). Previous studies revealed that several mechanisms may be involved in the drug-induced cytotoxicity. F-ara-ATP competes with dATP for incorporation into DNA by several DNA polymerases (6 -10), including DNA polymerase ⑀ (pol ⑀) (10), which possesses 3Ј 3 5Ј exonuclease activity (11,12). Because F-ara-ATP also inhibits ribonucleotide reductase (6 -8) and decreases dNTP pools (13), the incorporation of F-ara-ATP into DNA may be increased as a result of raising the ratio of F-ara-ATP to dATP in cells. The incorporated F-ara-AMP, found predominantly at the 3Ј-termini, impairs the function of DNA polymerases by effectively terminating DNA strand elongation (10). Acting at an additional process in DNA synthesis, F-ara-ATP inhibits primer RNA synthesis by DNA primase and thus may affect lagging strand initiation (14,15). Finally, the 3Ј-terminal F-ara-AMP in DNA is a poor substrate for human DNA ligase I (16). F-ara-ATP also interacts directly with DNA ligase I to inhibit the enzyme (16). In whole cells, F-ara-ATP incorporation in DNA is associated with the loss of large fragments of DNA from surviving cells (17) and is required for drug-induced DNA fragmentation in cells undergoing apoptosis (18 -20). The amount of F-ara-AMP incorporated into cellular DNA is linearly correlated with the loss of clonogenicity (10,21). It is therefore likely that this is essential for drug-induced lethality.
The 3Ј 3 5Ј exonuclease activities associated with prokaryotic and eukaryotic DNA polymerases that remove terminal mismatched nucleotides increase the fidelity of DNA replication (22)(23)(24)(25)(26)(27). Conditions that inhibit exonucleolytic proofreading decrease the fidelity of DNA polymerization (27,28), and cells lacking this activity exhibit higher mutation rates than those that express the wild type enzymes (29 -31). Because the toxicity of many anticancer and antiviral nucleotide analogues is expressed only after their incorporation into nascent DNA chains, such analogues may also be recognized as substrates for 3Ј 3 5Ј exonucleases. Thus, the ability of exonucleases to excise antimetabolites may serve as a drug resistance mechanism. As a corollary, incorporated analogues that resist excision removal are predicted to be highly cytotoxic.
Inasmuch as most of the incorporated F-ara-AMP residues are located at the 3Ј-termini of the DNA strands (10), 3Јterminal F-ara-AMP might well be excised by the 3Ј 3 5Ј exonuclease activities associated with DNA polymerases. Indeed, preliminary studies indicated limited excision of 3Ј-terminal F-ara-AMP by pol ⑀ (10). In the present study, we used an in vitro DNA excision assay to quantitatively investigate the ability of 3Ј 3 5Ј exonuclease activity of human DNA pol ⑀ to remove F-ara-AMP from the 3Ј-end of DNA. For comparison, excision of normal nucleotides was also evaluated. Our results demonstrated that DNA pol ⑀ bound to F-ara-AMP-terminated DNA with high affinity but excised the analogue from DNA at a low velocity. Once the phosphodiester bond between the 3Ј-F-ara-AMP and its adjacent nucleotide was cleaved by pol ⑀, the excision products appeared to remain associated with the enzyme, inactivating the exonuclease and preventing further exonucleolytic degradation or polymerization of the DNA products.
Preparation of 3Ј-F-ara-AMP-21-mer-The 17-mer M13 sequencing primer was labeled at its 5Ј-terminus with [␥-32 P]ATP by T4 polynucleotide kinase and annealed to the complementary site of the M13mp18(ϩ) single-stranded DNA template as described previously (10). The labeled 17-mer primer/M13mp18 template was incubated with pol ␣ and 100 M each of dCTP, dGTP, and 10 M F-ara-ATP at 37°C for 30 min in a reaction mixture containing 20 mM Tris-HCl, pH 7.5, 8 mM MgCl 2 , 0.5 mM dithiothreitol, and 10 mM NaCl. F-ara-ATP was incorporated into the primer at site 21 opposite the T site of the M13mp18 template. The reaction products were separated by electrophoresis through a 15% polyacrylamide sequencing gel. The 21-mer band containing 3Ј-F-ara-AMP-21-mer was excised from the gel and recovered from the gel slice as described previously (10). The purified 3Ј-F-ara-AMP-21-mer was annealed to M13mp18(ϩ) DNA to generate the following DNA hybrid, where F indicates the incorporated F-ara-AMP at the 3Ј-end of the 21-mer. To assure that all 32 P-labeled oligomers were hybridized to the complementary M13mp18(ϩ) DNA strands, a ratio of 1 oligomer to 10 copies of M13 DNA was used (see below). Preparation of Normal DNA Substrate for Excision by pol ⑀-A 21base oligomer with nucleotide sequence identical to the 3Ј-F-ara-AMP-21-mer except that dAMP was substituted for 3Ј-F-ara-AMP was chemically synthesized by Genosys. The normal 21-mer and the 17-mer M13 sequencing primer were separately labeled at the 5Ј-ends with [␥-32 P]ATP and purified by polyacrylamide sequencing gels as described above. The 32 P-labeled normal oligomers were annealed to the M13mp18(ϩ) DNA to produce the following DNA hybrid substrates for excision assay by pol ⑀. To assure that all 32 P-labeled oligomers were hybridized to the complementary M13mp18(ϩ) DNA strands, a ratio of 1 oligomer to 10 copies of M13 DNA was used. Under these conditions, no single-stranded 32 P-labeled oligomer was present in the mixture as evidenced by a complete conversion of all primers to high molecular weight DNA products when the mixtures were incubated with DNA pol ␣ in the presence of the four dNTPs.

21-mer
DNA Primer Excision Assay-The normal and F-ara-AMP-terminated DNA hybrids described above were used as the substrates for excision by pol ⑀. The reaction mixtures contained 20 mM Tris-HCl, pH 7.5, 8 mM MgCl 2 , 0.5 mM dithiothreitol, 10 mM NaCl, 0.2 unit of DNA polymerase ⑀, and the indicated concentration of DNA substrates. The reactions were carried out at 37°C for the indicated times up to 40 min. The reaction products were analyzed by electrophoresis through 10 or 15% polyacrylamide sequencing gels. After autoradiography, the radioactivity associated with each DNA band in the gels was quantitated by a Betascope 603 blot analyzer under conditions recommended by the manufacturer (Betagen Corporation, Waltham, MA).
DNA Polymerization Assay-The normal and F-ara-AMP-terminated DNA hybrids described above were used as the substrates for polymerization by pol ⑀ or pol ␣. The reaction mixtures contained 20 mM Tris-HCl, pH 7.5, 8 mM MgCl 2 , 0.5 mM dithiothreitol, 10 mM NaCl, 20 g/ml bovine serum albumin, 100 M each of 4 dATP, dCTP, and dGTP, the indicated concentration of DNA substrates, pol ⑀, and/or pol ␣. The reactions were carried out at 37°C for the indicated times, and analyzed by electrophoresis through a 15% polyacrylamide sequencing gel. After autoradiography, the radioactivity in the gels was quantitated by a blot analyzer as described above.
Kinetic Analysis-To determine the apparent K m and V max values for the excision of normal and F-ara-AMP-3Ј-terminated DNA by pol ⑀, various concentrations of the respective DNA constructs were incubated with pol ⑀ in 10-l excision reaction mixtures (see above) at 37°C for 10 min. Under these conditions, the reaction rate remained linear for at least 15 min. The reaction products were then analyzed and quantitated as described above. The sum of radioactivity associated with all excision products (20-mer and less) in each lane was used to calculate the reaction rate and then plotted against the respective DNA substrate concentrations. The apparent K m and V max values were calculated by the Michaelis-Menten equation with a computer-assisted program (33). To determine the K i value of 3Ј-F-ara-AMP-21-mer/ M13mp18 DNA in inhibiting the excision of normal oligomer/M13mp18 DNA by pol ⑀, various concentrations of normal 32 P-labeled DNA substrates were incubated with pol ⑀ for 15 min in the presence of different fixed concentrations of nonradioactive 3Ј-F-ara-AMP-21-mer/M13mp18 DNA. Nonradioactive normal 21-mer/M13mp18 DNA were used in parallel as controls. The reaction products were analyzed and quantitated as described above, and the K i value was calculated by the Michaelis-Menten equation with a computer-assisted program (33).
Gel Retardation Assay-The following normal or F-ara-AMP-containing oligomers were used as the substrates for binding to pol ⑀.
The oligomers (50 pg/20-l reaction mixture) were separately incubated with pol ⑀ for 10 min at 37°C and then an additional 10 min at room temperature in the presence or absence of 5-fold competing DNA (nonradioactive normal oligomer, 250 pg/20 l). The reaction products were then separated by electrophoresis through a 4.5% nondenaturing polyacrylamide gel at constant power of 25 W for 100 min. Autoradiography and Betascope analysis of the radioactivity were carried out after the gels were dried under vacuum at 60°C.

RESULTS
An in vitro DNA excision assay was used to investigate the ability of 3Ј 3 5Ј exonuclease activity associated with DNA polymerase ⑀ to remove the incorporated F-ara-AMP residues from the 3Ј-ends of DNA. A 21-base oligomer with F-ara-AMP at its 3Ј-end annealed to the complementary region of M13mp18(ϩ) DNA was used as the substrate for excision by pol ⑀. The same amount of normal 21-mer/M13mp18 DNA hybrid of identical sequence with dAMP at the 3Ј-end of the oligomer was used for comparison. As illustrated in Fig. 1 (top panel), pol ⑀ was able to excise normal nucleotides from the 3Ј-end of the normal 21-mer in a nonprocessive manner. More than 10 bands of excision products with different lengths (20-mer and less) were revealed (lanes 9 -14). The excision was time-dependent, and a substantial amount of final excision products (mononucleotides at the bottom of the gel) accumulated during a 25-40-min incubation (lanes 13 and 14). When a normal 17mer/M13mp18 DNA was the substrate for pol ⑀, the excision pattern was similar to that of the normal 21-mer/M13mp18 hybrid (data not shown). However, when the 3Ј-F-ara-AMP-21mer DNA hybrid was the substrate, pol ⑀ was much less effective in removing F-ara-AMP from the 3Ј-end of the oligomer ( Fig. 1, lanes 2-7). Less than 10% of the 3Ј-terminal F-ara-AMP was removed during a 40-min incubation. After removal of 3Ј-F-ara-AMP from site 21, no significant amount of further excision of the 20-mer product was evident. When the incubation time was prolonged to 25-40 min, a barely visible faint band appeared at the 19-mer position (lanes 6 and 7). No mononucleotide was produced even at the longest incubation time (lane 7).
The radioactivity associated with each DNA band was then quantitated as described under "Experimental Procedures." Determination of the excision velocity indicated that the rate of 3Ј-terminal dAMP removal by pol ⑀ was much greater than that of excising 3Ј-F-ara-AMP (Fig. 1, bottom panel). For example, 0.2 unit of pol ⑀ removed more than 30% of the deoxynucleotides in 5 min, whereas less than 5% of the terminal F-ara-AMP was excised during the same period.
To determine the excision kinetics of normal nucleotides and F-ara-AMP from DNA by pol ⑀, various concentrations of 3Ј-Fara-AMP-21-mer/M13mp18(ϩ) or normal 21-mer/M13mp18(ϩ) DNA were incubated with pol ⑀ for 10 min, during which period the excision rates remained linear. The reaction products were quantitated, and the apparent K m and V max values were determined as described under "Experimental Procedures." As summarized in Table I, the apparent K m of pol ⑀ for F-ara-AMP-21-mer/M13mp18 DNA was approximately 37-fold less than that for the normal 21-mer/M13mp18 DNA hybrid, indicating that pol ⑀ has a higher binding affinity to F-ara-AMP-containing DNA than to normal DNA. However, the V max of F-ara-AMP excision was substantially less than the V max of the enzyme for oligonucleotides terminated by dAMP, indicating that catalysis of excision was the rate-limiting step in removal of F-ara-AMP from the 3Ј-end of the oligomer. The kinetic parameters for a normal 17-base oligomer/M13mp18 DNA were similar to that of the normal 21-mer/M13mp18 DNA (Table I).
In addition to the difference in the reaction kinetics, the excision pattern of the 3Ј-F-ara-AMP-21-mer by pol ⑀ was different from that of the normal oligomer. When 3Ј-F-ara-AMP-21-mer/M13mp18 DNA was used as the excision substrate, a band at the 20-nucleotide position was the predominant reaction product (Fig. 1, lanes 2-7). No significant amount of further excision products were detected below the 20-nucleotide band. This indicates that after pol ⑀ removed F-ara-AMP at the FIG. 1. Excision of nucleotides from normal and F-ara-AMPterminated DNA by 3 3 5 exonuclease of pol ⑀. The 32 P-5Ј-endlabeled 3Ј-F-ara-AMP-21-mer and the normal 21-mer (162 pM each) separately annealed to M13mp18 templates were incubated with 0.2 unit pol ⑀ for the indicated times. The reaction products were analyzed by a 15% polyacrylamide sequencing gel and visualized by autoradiography as described under "Experimental Procedures." A, autoradiograph of the sequencing gel. Lane 1, 3Ј-F-ara-AMP-21-mer/M13mp18 DNA incubated without pol ⑀ and without dNTPs; lanes 2-7, 3Ј-F-ara-AMP-21-mer/M13mp18 DNA incubated with pol ⑀ and without dNTPs for 3, 5, 10, 15, 25, and 40 min, respectively; lane 8, normal 21-mer/ M13mp18 DNA incubated without pol ⑀ and without dNTPs; lanes 9 -14, normal 21-mer/M13mp18 DNA incubated with pol ⑀ and without dNTPs for 3, 5, 10, 15, 25, and 40 min, respectively. B, the radioactivity associated with the excision products in each lane was quantitated by a Betascope blot analyzer and expressed as percentage of total radioactivity input. q, normal 21-mer/M13mp18 DNA; E, F-ara-AMP-21-mer/ M13mp18 DNA. Each point represents the mean Ϯ S.E. from four experiments. 3Ј-end of the oligomer, the enzyme was not able to further excise normal nucleotides from the 20-base oligomer and, perhaps, protected the 20-mer product from further exonucleolytic degradation. As the incubation time was prolonged, only a barely visible band at the 19-mer position was revealed (Fig. 1,  lanes 6 and 7). In contrast, pol ⑀ removed nucleotides consecutively from the normal oligomer (Fig. 1A, lanes 9 -14). These observations were further confirmed by scanning the radioactivity profiles associated with each DNA band by a Betascope blot analyzer. In the 25-min reaction, at least 12 excision products, with sizes decreasing in increments of a single nucleotide, were revealed from the normal 21-mer/M13mp18 substrate (data not shown). When 3Ј-F-ara-AMP-21-mer/M13mp18 DNA was incubated with pol ⑀ for the same time, only a relatively small excision product peak corresponding to 20-mer was detected, which reflected the slow cleavage of the phosphodiester bond between 3Ј-F-ara-AMP and the adjacent nucleotide at position 20. A much smaller but identifiable 19-mer peak was also revealed. Thereafter, no peaks of smaller sizes were detected. Thus, it appears that after F-ara-AMP was excised, the 3Ј 3 5Ј exonuclease activity of pol ⑀ was greatly decreased relative to its utilization of the 20-mer/M13mp18 oligomer. Further experiments were designed to investigate the mechanism by which the 3Ј 3 5Ј exonuclease activity of pol ⑀ was inactivated after removal of F-ara-AMP from the 3Ј-end of the DNA. In sequential reaction experiments, pol ⑀ was first incubated with nonradioactive F-ara-AMP-21-mer/M13mp18 DNA for 10 min, and then the 32 P-labeled normal 21-mer/M13mp18 DNA was added to the reaction for an additional 20 min. Preincubation of pol ⑀ with 81 pM or 162 pM of the nonradioactive 3Ј-F-ara-AMP-21-mer/M13mp18 DNA resulted in a decrease in exonuclease activity by 69 and 78%, respectively, against the 32 P-labeled normal substrate (Fig. 2, lanes 3 and 4). In contrast, preincubation of pol ⑀ with the same concentrations of nonradioactive normal 21-mer/M13mp18 DNA did not significantly diminish the exonuclease activity against labeled normal substrate (Fig. 2, lanes 5 and 6).
To determine the kinetics of inhibition by 3Ј-F-ara-AMP-21mer/M13mp18 DNA, various concentrations of 32 P-labeled normal 21-mer/M13mp18 DNA substrates were incubated with pol ⑀ in the presence of different fixed concentrations of nonradioactive analogue-terminated or normal DNA hybrids. The dose effect of the two DNA hybrids on the 3Ј 3 5Ј exonuclease activity of pol ⑀ is illustrated in Fig. 3A. Only the F-ara-AMPcontaining DNA showed inhibitory activity. When the velocity of the reaction was plotted as the function of substrate ( 32 Plabeled normal 21-mer/M13mp18 DNA) concentrations in a double reciprocal plot (Fig. 3B), the lines representing reactions without and with different fixed concentrations of the inhibitor (nonradioactive 3Ј-F-ara-AMP-21-mer/M13mp18) converged at the ordinate, indicating that the nature of this inhibition was most likely competitive. In fact, computer analysis (33) of the plots revealed a competitive inhibition with a K i value of 7.5 pM. This is consistent with the high affinity of pol ⑀ for F-ara-AMP-21-mer/M13mp18 DNA as evidenced by its low K m for excision (7 pM, Table I).
Because the nonradioactive normal DNA (3-162 pM) did not inhibit the enzyme activity, the loss of excision activity observed in samples preincubated with F-ara-AMP-terminated DNA was not simply due to competition between the nonradioactive DNA (3-162 pM) and the 32 P-labeled DNA substrate (1400 pM). Rather, we postulated that when pol ⑀ was preincubated with 3Ј-F-ara-AMP-21-mer/M13mp18 DNA, the enzyme remained associated with the F-ara-AMP-terminated DNA with high affinity and was effectively trapped by the reaction products after the phosphodiester bond between 3Ј-F-ara-AMP and the adjacent nucleotide (position 20) was cleaved. This hypothesis predicts that the 3Ј-end of the oligomer (20-mer) would be blocked by pol ⑀ after cleavage of F-ara-AMP. Two different approaches were taken to test this postulate.
In the first experiment, pol ⑀ was incubated with 32 P-labeled F-ara-AMP-21-mer/M13mp18 DNA for 15 min to generate the 20-base excision product. Our hypothesis predicts that if the enzyme remained bound to DNA, the 3Ј-end of the 20-mer would not be accessible for further excision by freshly added pol ⑀. Indeed, the further addition of pol ⑀ to reactions containing F-ara-AMP-terminated DNA preincubated with pol ⑀ led to the accumulation of more 20-base excision product but did not result in significant removal of nucleotide from the 20-mer (Fig.  4, lanes 3-6). This is consistent with the conclusion that the 3Ј-end of the oligomer was protected or blocked by the enzyme. When the amount of pol ⑀ was increased to a total of 5 l (0.1 unit/l), a light band was visible at the 19-mer position (lane 6), suggesting that the binding of the preincubated pol ⑀ to the 20-mer/M13mp18 DNA was tight but still reversible. No mobility shift of the 20-mer was observed on denaturing polyacrylamide gels, indicating that pol ⑀ dissociated from the DNA under denaturation conditions (95°C, 50% formamide, and 8 M urea). Therefore, the binding between pol ⑀ and DNA did not appear to involve the formation of a covalent bond. In contrast, the addition of pol ⑀ to reactions preincubated with pol ⑀ and normal DNA led to almost complete digestion of the normal 21-mer (Fig. 4, lanes 9 and 10).
In the second approach to investigating the consequences of 3Ј-terminal F-ara-AMP excision, we reasoned that if pol ⑀ was not tightly associated to the initial excision products, the 3Ј-end of the 20-mer should be available for polymerization by DNA pol ␣ in the presence of normal dNTPs. Fig. 5 demonstrates that when F-ara-AMP-21-mer was incubated with pol ␣ and normal dNTPs after preincubation with pol ⑀ for 15 min, no polymerization products appeared (lane 5). Quantitation of the radioactivity by Betascope revealed 6.6 Ϯ 1.0%, and 6.9 Ϯ 1.5% (mean Ϯ S.D., n ϭ 12) of the total input radioactivity associated with the excision product band (20-mer) in the reaction containing pol ⑀ alone (lane 3) and the reaction containing pol ⑀ and pol ␣ in the presence of normal dNTPs (lane 5), respectively. The same amount of excision product (20-mer) in lane 3 (no pol ␣) and lane 5 (with pol ␣) indicates that after F-ara-AMP was excised from the 3Ј-end of the 21-mer by pol ⑀, the excision product (20-mer) was not extended by pol ␣. In contrast, when the normal 21-mer/M13mp18 DNA was preincubated with pol ⑀ for 15 min and then with pol ␣ in the presence of dATP, dCTP, and dGTP, both polymerization and excision products were generated (lane 6). The polymerization products appeared as a single band at the 24-mer position, due to the absence of dTTP on the reaction mixture (the M13mp18(ϩ) template directs that dTTP be incorporated at sites 25 and 26; see the DNA sequence under "Experimental Procedures"). The intensity of the DNA band at site 24 is visible but weak, probably due to the presence of exonuclease activity of pol ⑀. Quantitation of the radioactivity by Betascope analysis revealed 4320 counts asso-  ciated with this band, which represented 4.3% of the total input radioactivity. No radioactivity was detected at the same site (site 24) in lane 5, confirming that the cleaved F-ara-AMP-DNA could not be extended by pol ␣. When normal DNA was used as the substrate, high molecular weight DNA products were synthesized if all four dNTPs were added (data not shown). In the absence of pol ␣ and dNTPs, pol ⑀ generated only excision products (lane 4). Together, these results support the hypothesis that pol ⑀ remained bound to the 3Ј-end of the 20-mer DNA product after excision of F-ara-AMP.
The physical association of pol ⑀ with F-ara-AMP-DNA was revealed using a gel retardation assay. As demonstrated in Fig.  6, incubation of DNA pol ⑀ with 3Ј-F-ara-AMP-DNA oligomer (labeled with 32 P at the 5Ј-end) caused a mobility retardation of the oligomer (lane 2). Due to the high molecular mass of pol ⑀, the retarded band appeared near the well position. When a 5-fold concentration of competing DNA oligomer of identical DNA sequence (except that dAMP was at the 3Ј-end instead of F-ara-AMP) was added to the reaction, the intensity of the retarded band was not significantly reduced (lane 3), indicating that pol ⑀ bound to the F-ara-AMP-DNA oligomer with high affinity. In contrast, when normal 5Ј-32 P-DNA oligomers were incubated with pol ⑀, substantial amounts of excision products were generated. No evidence of normal oligomer retardation was discerned in the absence (lane 5) or the presence (lane 6) of competing DNA, suggesting only a loose association of pol ⑀ with the normal DNA substrate.
Because F-ara-AMP was generated when the phosphodiester bond between 3Ј-F-ara-AMP at position 21 and the adjacent nucleotide (position 20) was cleaved by pol ⑀, it is possible that the F-ara-AMP molecules might inactivate the exonuclease associated with pol ⑀. To investigate this possibility, various concentrations of free F-ara-AMP were added to the excision reactions containing pol ⑀ and normal 32 P-21-mer/M13mp18 substrate (Fig. 7). Free F-ara-AMP, at concentrations as great as 100 M, did not affect the excision activity of pol ⑀. Inhibition was observed only when the concentration of F-ara-AMP was increased to 1 mM. In contrast, pol ⑀ was inactivated by 10 -30 pM 3Ј-F-ara-AMP-21-mer/M13mp18 DNA (Figs. 2 and 3). These results are consistent with a mechanism-based inactivation of pol ⑀ by the analogue-containing DNA.
GMP, an inhibitor of the 3Ј 3 5Ј exonuclease activities (34 -36), was used to evaluate the excision of F-ara-AMP from the 3Ј-end of DNA by pol ⑀ in comparison with that of normal nucleotides. The addition of GMP to reaction mixtures inhibited the 3Ј 3 5Ј exonuclease activity of pol ⑀ to the same extent when either F-ara-AMP-terminated DNA or normal DNA was used as substrate (data not shown). This inhibition was concentration-dependent between 30 and 1000 M GMP. For example, excision of terminal F-ara-AMP and normal deoxynucleotides was inhibited by 51 and 55%, respectively, by 300 M GMP. These results indicate that the presence of F-ara-AMP at the 3Ј-end of DNA did not affect the inhibitory activity of GMP on pol ⑀, suggesting that the site in pol ⑀ for GMP action may be separate from the exonuclease catalytic site. Earlier studies (34) demonstrated that the 5Ј-monophosphate of arabinosyladenine also failed to inhibit the 3Ј 3 5Ј exonuclease of what is now recognized as pol ⑀ from rabbit bone marrow (35). Taken together, these result suggest that inhibition of pol ⑀ 3Ј 3 5Ј exonuclease by F-ara-AMP is likely to be a mechanism-based process. On the other hand, reduction of pol ⑀ 3Ј 3 5Ј exonuclease activity by relative high concentrations (1 mM) of free F-ara-AMP may be mediated by a separate mechanism, perhaps similar to that of inhibition by 5Ј-GMP. DISCUSSION The present study demonstrated that DNA polymerase ⑀ recognized and bound to F-ara-AMP-terminated DNA in preference to normal DNA. Excision of the analogue from the 3Ј-terminus of DNA, however, proceeded at a much slower rate than did removal of 3Ј-dAMP. Previous studies demonstrated that mismatched deoxynucleotides at the 3Ј-terminus of DNA caused an induced fit conformational change of T7 DNA polymerase (36,37). This change slowed the polymerization step and allowed sufficient time for the intramolecular transfer of the 3Ј-mismatched nucleotide from the polymerase site to the exonuclease site. The kinetic partitioning between exonuclease and polymerase sites favored error correction during DNA replication when a DNA polymerase encountered the 3Ј-terminal mismatched nucleotide. This was associated with a reduced rate of polymerization (37,38). A similar interpretation may be FIG. 6. Mobility retardation assay of pol ⑀/DNA complex. The 32 P-5Ј-end-labeled normal or F-ara-AMP-DNA oligomers (50 pg/20 l reaction mixture) were separately incubated with pol ⑀ for 10 min at 37°C, and then 250 pg/20 l of competing DNA (nonradioactive normal oligomer) was added to the indicated reactions and incubated for an additional 10 min at room temperature. The reaction products were separated in a nondenaturing polyacrylamide gel, and an autoradiograph was made as described under "Experimental Procedures." Lanes 1, 2, and 3, F-ara-AMP-oligomer incubated without pol ⑀, with pol ⑀, and with pol ⑀ plus competing DNA, respectively; lanes 4, 5, and 6, normal oligomer incubated without pol ⑀, with pol ⑀, and with pol ⑀ plus competing DNA, respectively. CO, enzyme-DNA oligomer complex; FD, free DNA oligomer; EX, excision products. applied to the case of an incorporated nucleotide analogue such as F-ara-AMP, which effectively inhibits the polymerization activity of pol ⑀ (10). As was seen with the mismatched terminal nucleotide-induced intramolecular transfer of T7 DNA polymerase, it is likely that when pol ⑀ encounters the 3Ј-terminal F-ara-AMP, which it cannot readily extend, a similar transfer of the 3Ј-terminal F-ara-AMP from the polymerase site to the exonuclease site might occur.
The present study showed that the 3Ј 3 5Ј exonuclease of pol ⑀ bound to 3Ј-F-ara-AMP-DNA with high affinity (K m ϭ 7 pM). This was in contrast with substrates with 3Ј-terminal dAMP, for which the 3Ј 3 5Ј exonuclease of pol ⑀ exhibited 37-fold less affinity. However, the 3Ј-terminal F-ara-AMP was a poor substrate for the exonucleolytic activity of pol ⑀; excision proceeded at a substantially slower rate than it did for 3Ј-terminal dAMP. Furthermore, the 3Ј 3 5Ј exonuclease of pol ⑀ became inactive after the 3Ј-terminal F-ara-AMP had been excised. Because it seems likely that the enzyme activity would be restored if the excision products were released, we postulate that one or both of the products remained associated with the enzyme after the cleavage of the phosphodiester bond. This hypothesis is supported by the unique excision patterns shown in Fig. 1, in which it is seen that after the phosphodiester bond between the 3Ј-terminal F-ara-AMP and the penultimate nucleotide was cleaved, pol ⑀ failed to excise the subsequent deoxynucleotides. The mobility retardation experiment (Fig. 6) provided evidence that pol ⑀ was physically associated with F-ara-AMP-DNA. We also attempted to test directly the hypothesis that F-ara-AMP remained associated with the enzyme after cleavage by using [2-3 H]F-ara-ATP to label the 3Ј-terminal F-ara-AMP on the primer. Unfortunately, these experiments were not conclusive because the specific activity of the incorporated [ 3 H]F-ara-AMP at the 3Ј-termini of DNA was not adequate to detect the small amount of F-ara-AMP excised.
Additional evidence indicates that after excision of F-ara-AMP, the 20-mer/M13mp18 DNA product remained in contact with pol ⑀. This interpretation is supported by the finding that after incubation of 3Ј-F-ara-AMP-21-mer/M13mp18 with pol ⑀, the DNA product was not a substrate for further excision upon the addition of fresh pol ⑀ (Fig. 4, lanes 3-6). Furthermore, after removal of 3Ј-F-ara-AMP, the inability of newly added pol ␣ and dNTPs to polymerize the 20-mer/M13mp18 DNA product (Fig. 5) is also consistent with the likelihood that one or both excision products remained associated with the enzyme. This association may have contributed to the sustained inhibition of both the exonuclease and polymerase functions. Because no mobility shift of the 20-mer product was observed in denaturing polyacrylamide gels, it is unlikely that pol ⑀ bound to the DNA excision product (20-mer/M13mp18 DNA) by forming a covalent bond.
Inactivation of DNA polymerization by formation of a 3Јanalogue DNA-enzyme complex was also observed by Reardon and Spector (39,40). They demonstrated that when acyclovir triphosphate was incorporated into the 3Ј-end of DNA primer by herpes simplex virus (type 1) DNA polymerase, upon binding of the next nucleotide as directed by the template, the viral enzyme and the analogue-containing DNA formed a "dead end complex." A similar mechanism was postulated for the inhibition of mammalian DNA polymerase ␣ by 2Ј,3Ј-dideoxycytidine triphosphate (41). In a different mechanism, aphidicolin inhibited both DNA polymerization and the 3Ј 3 5Ј exonuclease activity of pol ⑀ simultaneously (42)(43)(44) by sequestering the enzyme to the mismatched DNA region (43).
Incorporated arabinosyl nucleotide analogues are known to alter the configuration of duplex DNA in a specific manner (45). We speculate that although pol ⑀ readily recognizes and adapts its shape to bind to the 3Ј-terminal F-ara-AMP, cleavage of the phosphodiester bond is achieved with relative difficulty. It is likely that the conformational changes in pol ⑀ that are required to conduct this reaction are so extensive that the enzyme is unable to release the reaction products. If so, the change in enzyme conformation induced by excision of F-ara-AMP may be a key event responsible for the failure to release the excision products and for the ensuing inactivation of the exonuclease activity. This proposed mechanism of reactioninduced inactivation of pol ⑀ may provide an explanation for the inability of free F-ara-AMP to inhibit the enzyme in the presence of normal DNA (Fig. 7). Because the binding of pol ⑀ to normal DNA did not abnormally change the enzyme conformation, the presence of free F-ara-AMP should not lead to sequestering of the enzyme to its normal DNA substrate. Recent studies using genetic approaches indicated that pol ⑀, in addition to its polymerization and excision functions, may serve as a sensor involved in S phase checkpoint signaling in yeast (46). Thus, it will be important to investigate the biological consequences of the sequestering of pol ⑀ by F-ara-AMP-DNA in whole cells.
The data presented in Table I and Fig. 3 suggest that the inhibition of pol ⑀ excision of normal DNA by F-ara-AMP-DNA is probably competitive in nature. This may reflect the initial competitive binding of the two DNA species to the enzyme. It is possible that the fraction of pol ⑀ molecules that bound to F-ara-AMP-DNA might have reacted differentially to produce two possible consequences. A small portion of the enzyme molecules cleaved F-ara-AMP at a slow rate and thereby become sequestered in the dead end complex. Most of the enzyme, however, was unable to cleave the analog and eventually detached from the uncleaved DNA. This portion of pol ⑀ could have been recycled in the reaction, free to enter a second phase of competitive distribution between normal DNA and F-ara-AMP-DNA. This nonproductive binding of pol ⑀ to F-ara-AMP-DNA slowed the excision of normal DNA in a competitive manner. Nevertheless, because a small portion of the enzyme was sequestered in the dead end complex and was unable to recycle, the overall reaction is most likely a mixed-type competitive inhibition.
Although relatively few investigations of the ability of proofreading exonucleases associated with DNA polymerases to remove nucleotide analogues have been reported, a review of the field suggests that our findings with F-ara-AMP-terminated DNA may not be generalized to other arabinosyl nucleosides or to nucleoside analogues with different nonphysiological carbohydrates that result in either relative or absolute inhibition of DNA elongation. For instance, 2Ј,2Ј-difluorodeoxycytidine monophosphate, the active form of the new anticancer drug gemcitabine was resistant to excision by DNA pol ⑀ (47) when placed at either the 3Ј-terminus or in the penultimate position. With respect to antiviral drugs, the incorporated 3Ј-azidothymidine monophosphate was removed by human pol ⑀ with relative difficulty (48,49), whereas 2Ј,3Ј-didehydro-2Ј,3Јdideoxyadenosine was not a substrate for the same enzyme (48). The resistance to exonucleolytic action of 3Ј-terminal nucleotides with phosphorothioate linkages makes them substrates of choice for PCR primers when using thermostable DNA polymerases that possess 3Ј 3 5Ј exonucleases (50). In contrast, 3Ј-terminal arabinosylcytosine monophosphate was shown to be a relatively good substrate for excision by the 3Ј 3 5Ј exonuclease of E. coli Klenow fragment (51) and human DNA pol ⑀ (47). Following excision of the analogue, the appearance of excision products of intermediate length indicates that removal of subsequent nucleotides proceeded in a nonprocessive fashion. Comparative studies with DNA terminated by the mono-phosphates of arabinosyladenine and 2-fluoro-2Ј-deoxyadenosine will help determine the relative importance of the fluorine on the 2-carbon and the arabinosyl hydroxyl to the activity of pol ⑀ against F-ara-AMP.
After incubation of cells with radioactive F-ara-A, more than 94% of the incorporated drug in DNA was located at 3Ј-termini (10). This is strongly correlated with loss of viability, suggesting that such terminal incorporation of F-ara-A nucleotide is a critical mechanism of drug action (10,19,21). Further strand elongation by DNA polymerases is greatly impeded (10), and ligation of DNA strands by DNA ligase I (16) is inhibited by 3Ј-terminal F-ara-AMP in DNA, thus interfering with DNA replication and probably with DNA repair. Because efficient excision of 3Ј-terminal F-ara-AMP from DNA could constitute a resistance mechanism by which cells may circumvent these inhibitory actions, it will be important to evaluate whether cells possess other enzymes that may be more capable than pol ⑀ of removing the incorporated analogue. For instance, the amino acid sequence of the active site region of 3Ј 3 5Ј exonucleases of aphidicolin-sensitive DNA polymerases is highly conserved (30). It will be interesting to determine whether 3Ј-terminal F-ara-AMP poses a similar problem for the 3Ј 3 5Ј exonucleases of DNA polymerases other than pol ⑀. In addition, proteins with 3Ј 3 5Ј exonucleases that either lack DNA polymerase activity (52)(53)(54)(55)(56)(57)(58) or have cryptic polymerizing activities (59) have been identified. The possibility that these may be accessory proteins that provide a proofreading function for DNA polymerases such as pol ␣ that lack inherent exonuclease activity provides a compelling rationale for characterizing their activities against DNA terminated by therapeutically relevant nucleotide analogues.