Interaction between DNA Polymerase λ and Anticancer Nucleoside Analogs*

The anticancer activity of cytarabine (AraC) and gemcitabine (dFdC) is thought to result from chain termination after incorporation into DNA. To investigate their incorporation into DNA at atomic level resolution, we present crystal structures of human DNA polymerase λ (Pol λ) bound to gapped DNA and containing either AraC or dFdC paired opposite template dG. These structures reveal that AraC and dFdC can bind within the nascent base pair binding pocket of Pol λ. Although the conformation of the ribose of AraCTP is similar to that of normal dCTP, the conformation of dFdCTP is significantly different. Consistent with these structures, Pol λ efficiently incorporates AraCTP but not dFdCTP. The data are consistent with the possibility that Pol λ could modulate the cytotoxic effect of AraC.

Nucleoside analogs are an important class of compounds that are clinically used for anticancer and antiviral treatments (1). For example, 3TC, 3 AZT, and D4T have been used for antiviral treatments (1). AraC (Fig. 1) is used to treat leukemia (2), and dFdC ( Fig. 1) is used to treat various types of cancer, such as non-small cell lung cancer (3), breast cancer (4), and pancreatic cancer (5). The main mechanism of action of nucleoside analogs is to terminate the DNA elongation process after their incorporation into viral or cellular DNA (6 -8). For example, due to the lack of a 3Ј-hydroxyl group on their sugar moieties, the incorporation of the antiviral compounds 3TC, AZT, and D4T terminates human immunodeficiency virus DNA replication immediately (9).
Although both AraC and gemcitabine have a 3Ј-hydroxyl group on their sugar moieties, the incorporation of either compound has been shown to terminate the DNA elongation processes and result in DNA fragmentation (7,10). Studies have shown that the arabinose sugar moiety in AraC and the difluoro group on the 2Ј-position of the sugar moiety of dFdC alter the DNA structure, inhibiting DNA polymerases (11,12). The cytotoxicity of AraC or dFdC is proportional to the amount of the analog incorporated into DNA (6 -8). Therefore, the cellular enzymes that are involved in activating these nucleoside analogs and the DNA polymerases that introduce them into DNA play a critical role in determining their cytotoxic activity.
Human DNA polymerase (Pol ) is an exonuclease-deficient, 575-amino acid DNA polymerase that belongs to the X-family (13). It contains a N-terminal BRCT domain that is important for protein partnerships, a proline/serine-rich region that has been suggested to be a target for post-translational modification, and a C-terminal 39-kDa polymerase domain (14). The 39-kDa domain, which is structurally well characterized (15,16) contains the fingers, palm, and thumb subdomains typical of DNA polymerases (17). It also contains an 8-kDa subdomain with dRP lyase activity (14) that contributes to base excision repair (BER) in vitro. Studies in Pol ␤-deficient mouse embryonic fibroblast extracts indicate that Pol also performs BER in vivo (18,19). Moreover, in a confocal microscopy study, Pol relocates to sites of oxidative DNA damage, and embryonic fibroblasts derived from the Pol knock-out mice are more sensitive than wild type cells to H 2 O 2 treatments (18). These studies suggested that Pol plays a protective role against oxidative DNA damage. A recent report further indicates that expression of Pol increases in order to repair smoking-induced damage in respiratory epithelium (20). In addition to BER, Pol is also strongly implicated in filling gaps in DNA during non-homologous end joining of double strand breaks (15,16,21).
It is therefore possible that Pol could incorporate therapeutic nucleoside analogs into DNA during BER and nonhomologous end joining, which may have an impact on the cellular sensitivity to these compounds following DNA damage. Therefore, in this study, we investigate the ability of Pol to incorporate AraCTP and dFdCTP, and we provide crystal structures of Pol in a complex with a physiologically relevant, one-nucleotide gapped DNA substrate with each analog present.

Purification of Pol Proteins and Primer Extension
Assays-The expression and purification processes of Pol protein were as previously described (14). The Pol Y505A mutant protein expression vector was generated using the QuikChange site-directed mutagenesis kit (Stratagene) with a forward primer, 5Ј-TGCCT-GTGCCCTGCTCGCTTTCACC-GGCTCTGCAC-3Ј, and a reverse primer, 5Ј-GTGCAGAGCCGGT-GAAAGCGAGCAGGGCACAG-GCA-3Ј, on a pET-22b vector (Merck) that harbors 39-kDa fragment of wild type Pol . The sequence was confirmed by sequencing. The Y505A mutant protein was expressed and purified using the same procedures as for the wild type Pol (14). To evaluate the polymerase activity of Pol , 1.25 nM Pol was incubated with 100 nM double-stranded DNA oligonucleotide containing a single nucleotide gap in the presence of 50 M dNTP, 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml bovine serum albumin. For steady state kinetic analysis, the concentrations of the nucleoside triphosphate metabolites were ranged from 0.01 to 50 M at a fixed enzyme concentration. For each reaction, less than 25% of the substrates were converted to products by the enzyme, even at the lowest substrate concentration. All of the kinetic data were determined based on the initial reaction rates. The reaction mixtures were analyzed with a 12% denaturing polyacrylamide DNA sequencing gel, and the results were quantified with a phosphor imager (Fuji FLA5100).
Cell Lines and Cell Sensitivity Assay-The SV40-transformed Pol ϩ/ϩ and Pol Ϫ/Ϫ MEF cells were a generous gift from Dr. Samuel Wilson (NIEHS, National Institutes of Health) (19). The cells were grown at 37°C in a 10% CO 2 incubator in Dulbecco's modified Eagle's medium and 10% fetal bovine serum. For cell sensitivity studies, 4,000 cells/well were seeded in a 24-well plate 24 h before adding different concentrations of AraC. The cells were incubated for an additional 72 h in the humidified incubator. At the end of incubation period, 5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (100 l) was added for an additional 4-h incubation, and the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide/formazan generated by live cells was dissolved in 200 l of DMSO. The absorbance was determined at 560 nm using a microtiter plate reader (Bio-Rad Benchmark Plus). The cell survival data were presented as percentage of the absorbance of treated/untreated cells.
Crystallization-Crystals of Pol in complex with oligonucleotides T11 (5Ј-CGGCGGTACTG), PG (5Ј-CAGTAC), and DT (5Ј-GCCG) were grown using the hanging drop method. AraCTP crystals grew in 50 mM Hepes, pH 5.5, 15% 2-propanol, 200 mM sodium citrate, and 10 mM magnesium chloride. dFdCTP crystals grew in 50 mM Hepes, pH 8.0, 20% 2-propanol, 200 mM sodium citrate, and 10 mM magnesium     (22). Refinement was carried out using CNS (23) and phenix (24), and model building was carried out with O (25) and Coot (26). The quality of the model was assessed using Molprobity (27), and it was found to have good stereochemistry (see Table 2). Competition between dFdCTP and dCTP for Pol Incorporation-The same DNA oligonucleotides and reaction conditions as described above were used for competition study except for the concentrations of dCTP and dFdCTP. Under each dCTP concentration (0.1, 0.2, 0.5, 1, and 2 M), increasing concentrations of dFdCTP (0, 0.5, 1, 5, and 20 M) were added to the reactions. All of the kinetic data were determined based on the initial rate of the reactions under different combinations of dFdCTP and dCTP. The reaction mixtures were analyzed with a 12% denaturing polyacrylamide DNA sequencing gel, and the results were quantified with a phosphor imager (Fuji FLA5100).

RESULTS AND DISCUSSION
Incorporation of AraCTP and dFdCTP by Human Pol -The 39-kDa form of Pol was used in this study to allow direct comparisons of incorporation with crystal structures. To examine if AraCTP and dFdCTP are substrates, Pol was incubated with a DNA oligonucleotide substrate containing a onenucleotide gap in the presence of dCTP, AraCTP, or dFdCTP. As shown in Fig. 2, Pol inserted AraCTP into DNA, albeit slightly less efficiently than it incorporated dCTP. In contrast, dFdCTP was poorly inserted by Pol . Steady state kinetic analysis was then performed ( Table 1). The K m value of Pol for dCTP was 0.04 M, and the k cat was 7.0 min Ϫ1 . The incorporation efficiency (k cat /K m ) for Pol on dCTP is 175. Compared with dCTP, Pol exhibited a higher K m and a reduced k cat (Table 1) when incorporating AraCTP, resulting in an 18-fold decrease in catalytic efficiency relative to dCTP. Incorporation of dFdCTP was even less efficient and was ϳ145-fold reduced as compared with dCTP.
Crystallization of Pol with Gapped DNA and Incoming Nucleoside Analogs-Several structures of Pol in complex with different DNA substrates and incoming nucleotides have resulted in a very detailed characterization of the polymerization reaction catalyzed by this enzyme (13, 28 -33). Because AraCTP and dFdCTP are relatively similar substrates and yet Pol has a very different ability to incorporate them as shown in Fig. 2, we decided to co-crystallize Pol with either AraCTP or dFdCTP as incoming nucleotides to investigate the structural reasons for this difference. We therefore solved the co-crystal structures of Pol bound to a double-stranded DNA containing a single nucleotide gap and either AraCTP (2.03 Å) or dFdCTP (2.00 Å; see "Materials and Methods" and Table 2) as an incoming nucleotide. As shown in Fig. 3, a-d, both analogs can be accommodated in the dNTP binding pocket of the enzyme without major structural rearrangements. AraCMP was incorporated by the polymerase and is therefore observed as part of the growing chain. In contrast, very little incorporation of dFdCMP was observed, even after soaking the crystal in the presence of Mn 2ϩ ions (see "Materials and Methods"). Both analogs appear to adopt a similar conformation. The main differences between these analog complexes and the reference structure with dCTP (PDB entry 2PFP) (28) are in the sugar moiety of the analogs. The 2Ј-OH of AraCTP is within hydrogen bonding distance of the backbone amide nitrogen of Ala 510 . The presence of this atom does not appear to impose significant constraints in either the conformation of the ribose of the incoming nucleotide or the surrounding active site residues. On the other hand, the two fluorine atoms in C2Ј of dFdCMP appear to be incompatible with the normal conformation of the ribose of the incoming nucleotide during catalysis. The dFdCMP ribose is shifted away from its usual position by about 0.9 Å, and it is apparent that if not for this shift, the fluorine atoms would clash with residues of the active site, in particular with the backbone carbonyl oxygen of Tyr 505 and the delta nitrogen of Asn 513 , which are 2.53 and 2.36 Å away, respectively, from one of the fluorine atoms in the structure. As a result of this shift, the ␣-phosphate of dFdCMP is slightly away from its normal cata-lytic position. Because the nucleophilic attack by the 3Ј-oxygen on the ␣-phosphate that is responsible for phosphoryl transfer is very sensitive to the geometry of the reactive groups, this difference can explain the observed reduction in catalytic efficiency with dFdCTP. Interestingly, discrimination against dFdCTP appears to be mostly based on the same structural features that allow Pol to discriminate against ribonucleotides. Previous studies suggested that amino acid Tyr 505 plays an important role in polymerization and substrate selection (34 -36). The impact of Tyr 505 on the incorporation efficiency of AraCTP and dFdCTP was also examined using a Y505A mutant Pol . As shown in Fig. 3e and Table 3, the Y505A mutant has significantly reduced polymerase activity as compared with the wild type enzyme because its efficiency in incorporating dCTP is reduced more than 200-fold. Similar nucleotide selection preference was observed for the Y505A mutant as that of the wild type enzyme, which is dC Ͼ AraC Ͼ dFdC. As shown in the crystal structure, the side chain of Tyr 505 does not appear to play a significant role in discriminating against dFdCTP, which is supported by the biochemical results. However, the Tyr 505 side chain appears to play a general role in substrate selection. Although the structural basis for this role is not clear at present, it is possible that during the dNTP-induced conformational change, the side chain adopts a conformation not observed in the crystal structures that is critical for sugar discrimination. It is also possible that substitution of the tyrosine side chain might lead to local structural perturbations that affect sugar discrimination.
Competition Assay between dFdCTP and dCTP-To obtain further information on the low efficiency of incorporation of dFdCMP, a competition assay was performed between dCTP and dFdCTP. Increasing concentrations of dFdCTP were added to the reactions containing a constant amount of dCTP. Fitting the results (Fig. 4) to a Michaelis-Menten equation revealed a mixed non-competitive mechanism of inhibition, where both the K m and k cat values changed with increasing concentration of dFdCTP ( Table 4). The observed increase of K m and reduction of k cat in response to increasing concentrations of dFdCTP suggest that dFdCTP affects both the substrate binding and enzyme turnover. This is consistent with our structural obser- FIGURE 3. Structure of Pol in complex with AraCTP and dFdCTP. A, the complex in which AraCTP is bound as an incoming nucleotide shows minimal distortion in the active site, and AraCMP has been fully incorporated into the primer chain. The AraCMP residue is overlaid with a normal dCTP residue (31). A simulated annealing F o Ϫ F c electron density map contoured at 3 is shown (blue). B, structure of Pol in complex with dFdCTP. Although dFdCTP can be bound in the Pol active site, it can only be accommodated in a conformation that differs significantly from that of a natural dCTP (transparent). A simulated annealing F o Ϫ F c electron density map contoured at 3 is shown (blue). C, two fluorine atoms have been modeled on the C2Ј of a natural dCTP (31). The two fluorine atoms (Van der Waals surface in magenta) would clash with the backbone of active site residues and the side chain of Asn 513 (green). D, an overlay view of AraCTP and dFdCTP within the active center of Pol . The primer strand of the AraC structure is yellow. The primer strand for the dFdCTP structure is brown. The three catalytic aspartates are shown in magenta (AraC) and green (dFdCTP). E, incorporation of AraCTP and dFdCTP by Y505A Pol mutant protein. The wild type or Y505A mutant Pol was incubated with a double-stranded DNA oligonucleotide in the presence of different nucleoside triphosphates at 37°C for 3 min under the conditions described under "Materials and Methods."  (19) were used to compare their sensitivity to AraC. As shown in Fig.  5, the Pol -deficient cells showed reduced sensitivity to AraC as compared with the wild type cells. This result is consistent with the biochemical and structural analysis and suggests that Pol plays a role in sensitizing cells to AraC in vivo.
In this study, we have reported the first co-crystal structure of a human DNA polymerase in complex with a nucleotide analog used for cancer therapy. These structures provide novel insight into the structural characteristics of these analogs that determine their efficient incorporation into DNA as well as the structural basis of sugar selectivity in Pol . Given the roles of Pol in DNA base excision repair and non-homologous end joining repair (21,37), it is possible that the cellular sensitivity to AraC may partly depend on Pol . In particular, Pol may have a more profound impact on cellular sensitivity to AraC when used in combination with other DNA-damaging agents, such as mitoxantrone, which introduces DNA double strand breaks, and/or hydroxyurea or fludarabine, which affect cellular nucleotide pools.