Mechanism of Ribonucleotide Incorporation by Human DNA Polymerase η*

Ribonucleotides and 2′-deoxyribonucleotides are the basic units for RNA and DNA, respectively, and the only difference is the extra 2′-OH group on the ribonucleotide sugar. Cellular rNTP concentrations are much higher than those of dNTP. When copying DNA, DNA polymerases not only select the base of the incoming dNTP to form a Watson-Crick pair with the template base but also distinguish the sugar moiety. Some DNA polymerases use a steric gate residue to prevent rNTP incorporation by creating a clash with the 2′-OH group. Y-family human DNA polymerase η (hpol η) is of interest because of its spacious active site (especially in the major groove) and tolerance of DNA lesions. Here, we show that hpol η maintains base selectivity when incorporating rNTPs opposite undamaged DNA and the DNA lesions 7,8-dihydro-8-oxo-2′-deoxyguanosine and cyclobutane pyrimidine dimer but with rates that are 103-fold lower than for inserting the corresponding dNTPs. X-ray crystal structures show that the hpol η scaffolds the incoming rNTP to pair with the template base (dG) or 7,8-dihydro-8-oxo-2′-deoxyguanosine with a significant propeller twist. As a result, the 2′-OH group avoids a clash with the steric gate, Phe-18, but the distance between primer end and Pα of the incoming rNTP increases by 1 Å, elevating the energy barrier and slowing polymerization compared with dNTP. In addition, Tyr-92 was identified as a second line of defense to maintain the position of Phe-18. This is the first crystal structure of a DNA polymerase with an incoming rNTP opposite a DNA lesion.

RNA and DNA are fundamental to life in all forms. The two nucleic acid polymers are composed of ribonucleotides and 2Ј-deoxyribonucleotides as the basic units, respectively. However, ribonucleotides have been found in DNA; they constitute a large proportion of the "DNA lesions" in the genome. In mice, they have been shown to be collectively the most frequently occurring DNA lesions, even more than abasic sites and 7,8dihydro-8-oxo-2Ј-deoxyguanosine (8-oxodG). 2 The presence of ribonucleotides in DNA increases the possibility of spontaneous hydrolysis, causing DNA to break more frequently. These ribonucleotides are mainly removed from DNA by the RNase H2 pathway (1)(2)(3)(4)(5).
DNA polymerases introduce ribonucleotides into DNA by misinserting them (6). Concentrations of cellular rNTPs are 1-6 orders of magnitude higher than those of dNTPs, depending on the cell type and the stage of the cell cycle (6 -9). To discriminate dNTP from rNTP, a steric gate residue (typically a tyrosine or phenylalanine) is generally conserved in DNA polymerases and thought to create a clash with the extra hydroxyl group of the incoming rNTP (10 -17). Although limited in extent, ribonucleotide incorporation has still been observed by a variety of DNA polymerases, from replicative ones with relatively high fidelity and small active sites (e.g. pol ⑀ and pol ␦) to error-prone X-family pol and pol ␤ and Y-family pol (6,8,(17)(18)(19)(20)(21).
Compared with other DNA polymerases, those in the Yfamily are known for high misinsertion rates and tolerance of DNA adducts in the template strand by relying on their spacious active sites (22)(23)(24). Y-family human DNA polymerase (hpol ) is of particular interest, as it is the only known DNA polymerase directly related to a human genetic disorder, mainly due to its unique role in translesion synthesis past the UVinduced DNA lesion cyclobutane pyrimidine dimer (CPD). Patients defective in hpol , the result of which is a form of xeroderma pigmentosum (XP-V), are typically highly sensitive to UV light, with increased incidence of skin and other types of cancer and (for some of the individuals) neurodegeneration (25)(26)(27)(28). hpol is also involved in translesion synthesis of other DNA lesions, e.g. 8-oxodG (29,30), abasic sites (31,32), and cisplatin or other therapeutic drug-induced DNA damage (33)(34)(35)(36)(37)(38).
In this study, we investigated the ability of hpol to incorporate ribonucleotides into DNA and crystallized hpol with incoming rNTPs opposite both undamaged DNA and an 8-ox-odG lesion. Our results demonstrate that hpol can incorporate ribonucleotides into DNA with relatively high selectivity but low efficiency, even when the template strand contains the DNA lesion 8-oxodG or CPD. The x-ray crystal structures show that the incoming rNTP at the hpol active site adopts a slightly different orientation relative to dNTP to avoid a clash with the steric gate residue, without significantly disrupting the pairing with the template dG or 8-oxodG. The latter appears to be the first crystal structure of an incoming rNTP opposite a DNA lesion within a DNA polymerase.

Experimental Procedures
Materials-Oligonucleotides were purchased from Integrated DNA Technologies (Coralville, IA) or TriLink BioTechnologies (San Diego) and purified by HPLC by the manufacturers. rNTPs and dNTPs were purchased from New England Biolabs (Ipswich, MA). These experiments were conducted with the catalytic core of hpol or the Y92A mutant (1-432 amino acids), and the former has been shown to have similar catalytic activity as the full-length protein in vitro (33). The hpol and the Y92A mutant (1-432 amino acids) were expressed and purified as reported previously (39). Polyethylene glycol monomethyl ether 2000 (from Hampton Research, Aliso Viejo, CA) was used for crystallization.
For steady-state kinetic assays, 5 M DNA substrate was incubated with 1.5 to 500 nM hpol or the Y92A mutant as well as varying concentrations of each dNTP or rNTP at 37°C for 5 min, in the same reaction buffer as extension and single base incorporation assay. Reactions were stopped by the addition of a quench buffer (90% formamide (v/v) and 10 mM EDTA). Products were separated on 18% (w/v) denaturing polyacryl-amide gels and visualized using a Typhoon system (GE Healthcare). The data for steady-state kinetic assays were fit to a (hyperbolic) Michaelis-Menten equation using Prism software (GraphPad, La Jolla, CA).
Crystallization, Diffraction, Data Collection, and Structure Determination-Each DNA substrate and hpol were mixed at a molar ratio of 1.1 to 1, before adding CaCl 2 (final concentration 3.33 mM) and excess amount of rNTP. To obtain crystals by hanging drop vapor diffusion, 1 l of each protein/DNA mixture was mixed with 1 l of reservoir solution, which contained 100 mM sodium MES (pH 6.0), 5 mM CaCl 2 , and 16 -22% (w/v) polyethylene glycol monomethyl ether 2000, followed by equilibration with 500 l of the reservoir solution at 18°C. The crystals were washed through cryoprotectant (composed of 3 volumes of reservoir solution and 1 volume of glycerol) before flash freezing in liquid nitrogen. Diffraction data were collected on the 21-ID-F or 21-ID-G beamline at the Advanced Photon Source (Life Sciences Collaborative Access Team, Argonne National Laboratory, Argonne, IL). Data were integrated and scaled using HKL2000 (40), processed with Phaser MR (41) for molecular replacement phasing (PDB code 4O3N as the search model for hpol ⅐dG⅐rCTP, and PDB code 4O3P for hpol ⅐dG⅐rCTP and hpol ⅐dG⅐rATP) (29), and refined with PHE-NIX (42). Model building was performed with ARP/ ARP classic (43,44) and COOT (45). All illustrations were generated with the program UCSF Chimera (46).

Results
hpol Inserts Ribonucleotides Opposite Undamaged Template dG or dA-To assess the ribonucleotide incorporation ability of hpol , we compared the extension of the primers in the presence of all four dNTPs or rNTPs with different incubation times. The primer was extended at 5 min and reached full-length after 30 min in the presence of rNTPs, compared with 5 min for full-length extension with dNTPs (Fig. 1A). To estimate the fidelity of ribonucleotide incorporation, hpol was incubated with annealed undamaged DNA substrate as well as individual dNTP or rNTP. Under our reaction conditions, multiple nucleotides were added to the primer in the presence of a single dNTP in an error-prone manner. However, only one or two nucleotide extensions was observed with a single rNTP. Among the four ribonucleotides, rCTP (able to Watson-Crick pair with template dG) was the most efficient one for insertion (Fig. 1B). To further measure the efficiency and fidelity of rNTP incorporation by hpol , steady-state kinetic experiments were conducted. The catalytic efficiency (k cat /K m ) for rCTP insertion opposite template dG was 770-fold less than that for dCTP and 2-5-fold less relative to those for the mismatched dNTPs. However, rates of incorporation of the other ribonucleotides (other than rCTP) were very low and could not be measured experimentally. In addition, to investigate the effect of the template base, another set of DNA substrates with dT in the template instead of dG was included in the steady-state kinetic study. The change for catalytic efficiency between rATP and dATP insertion opposite template dT was 3400-fold, slightly higher than that between rCTP and dCTP opposite dG (770fold, Table 1).
hpol Inserts Ribonucleotides Opposite 8-oxodG-We conducted primer extension assays by hpol with 8-oxodG in the template strand in the presence of dNTPs or rNTPs. The pattern for primer extension past 8-oxodG was very similar to that against undamaged DNA (Figs. 1A and 2A). The primer was partially extended to full length by 30 min with the mixture of all four rNTPs, whereas hpol elongated the primer to full length within 5 min with dNTPs. In single nucleotide insertion assays, only one rC was added to the primer within 5 min, although in the presence of the other individual rNTPs significantly elongated primers were only observed after 30 min of incubation (Fig. 2, A and B).
In steady-state kinetic assays, hpol incorporated dCTP opposite template 8-oxodG 2300-fold more efficiently than rCTP and 43,000-fold more than rATP. The difference between the catalytic efficiencies for rCTP and rATP incorporation was about 19, compared with 5 for dCTP and dATP, indicating that during nucleotide incorporation hpol has similar base selectivity regardless of the sugar ( Table 2).
hpol Incorporates Ribonucleotides Opposite CPD-Given that hpol is capable of bypassing CPD by inserting the correct nucleotide dA (26,28), we investigated the incorporation of ribonucleotides against this lesion. hpol extended the primer past CPD in a manner similar to that observed opposite undamaged DNA or 8-oxodG. With only rATP, more than 50% of the primers were elongated by two nucleotides after 5 min, much faster than with the other single rNTPs (Fig. 3, A and B). Quantitatively, rATP insertion was 1400-fold less efficient than dATP opposite CPD (Table 3).
hpol Scaffolds the Incoming rCTP and Template dG to Form a Watson-Crick Base Pair with a Propeller Twist-To understand the mechanism of ribonucleotide insertion by hpol , we    Fig. 4A. The incoming rCTP formed a Watson-Crick base pair with the template dG, but a significant propeller twist was observed. The dihedral angle between the two base planes was 27°. In addition, the base pair was slightly shifted toward the major groove, compared with the dG:dCTP pair at the hpol active site. Phe-18 was identified as the steric gate residue, and the distance between either 2Ј-OH or 3Ј-OH of rCTP and their closest atoms of the Phe-18 side chain was 3.2 Å (Fig. 5). In comparison with the hpol ⅐dG:dCTP structure (PDB code 4O3N) (29), the position of the phenyl ring of Phe-18 was almost identical. In addition, the distance between the 3Ј-OH of the primer end (two conformations) to P␣ of the incoming rCTP was 4.5 or 4.3 Å, about 1 Å further than that in   the hpol ⅐dG:dCTP structure (3.5 or 3.3 Å), providing an explanation for the 10 3 -fold lower catalytic efficiency of the polymerization reaction (Fig. 5).
hpol Accommodates 8-oxodG in Two Conformations Opposite the Incoming rCTP-We also crystallized hpol with an incoming rCTP opposite 8-oxodG (Table 4). The final Fourier (2F o Ϫ F c ) sum electron density map is shown in Fig. 4B. Noticeably, the electron density of 8-oxodG did not fit either a pure anti conformation as seen at the active site of hpol ⅐(8-oxodG): dCTP or a pure syn conformation as in hpol ⅐(8-oxodG):dATP (29). Instead, the electron density was indicative of two alternative conformations of 8-oxodG. After automated refinement using the program Phenix, the occupancy of 8-oxodG in the anti conformation refined to 50%, with the base in the syn conformation contributing the other half. In contrast to the dual conformation of 8-oxodG, the incoming rCTP fit the electron density very well, consistent with a single conformation (Fig.  4B). When 8-oxodG is in an anti conformation, the active site closely resembles that of the hpol ⅐dG:rCTP complex. Incoming rCTP and 8-oxodG are engaged in a Watson-Crick base pair with formation of a significant propeller twist. The dihedral angles of the bases of the incoming rCTP and 8-oxodG were 29°( anti conformation) and 25°(syn conformation) (Fig. 6). In the syn conformation, 8-oxodG maintained three H-bonds with rCTP as follows: O6 of 8-oxodG H-bonded with N4(H) of rCTP, with a distance of 2.8 Å, and N7(H) of 8-oxodG donated in a bifurcated H-bond to O 2 and N3 of the incoming rCTP, with distances of 3.1 and 3.2 Å, respectively. The O8 atom of 8-oxodG in the syn conformation was positioned at 3.6 Å from O2 of rCTP and therefore outside the range for interacting.
With either conformation adopted by 8-oxodG, the 8-ox-odG:rCTP pair was shifted toward the major groove compared with the structure of hpol ⅐(8-oxodG):dCTP (PDB code 4O3P) (29). The closest distance between 2Ј-OH of rCTP and Phe-18  3.3 Å, and the distance between 3Ј-OH and the steric gate residue was 3.4 Å. The 3Ј-OH of the primer end (two conformations) was 4.3 or 4.1 Å away from the P␣ of rCTP, whereas the corresponding distance was 3.4 or 3.2 Å in the hpol ⅐(8-oxodG):dCTP complex, respectively, similar to the relative positions for rCTP and dCTP opposite template dG at the hpol active site (Fig. 6).

8-oxodG Adopts a syn Conformation in the Pair with the
Incoming rATP at the hpol Active Site-Given that 8-oxodG adopts a syn-conformation in the structure of hpol ⅐(8-ox-odG):dATP (PDB code 4O3O) (29), we also examined the orientation of rATP opposite 8-oxodG at the active site of hpol . The final Fourier (2F o Ϫ F c ) sum electron density map around the base pair at the active site is shown in Fig. 4C. 8-oxodG indeed adopted a syn conformation, but instead of a planar Hoogsteen base pair, a propeller twist of 32°was observed for 8-oxodG:rATP. As a result, the distance between O6 of the lesion and N6 of rATP was 3.7 Å, too long for formation of an H-bond. Because of the propeller twist, 2Ј-OH of rATP was 3.1 Å away from the closet atom of Phe-18, whereas the closest distance between 3Ј-OH and the side chain of the steric gate residue was 3.2 Å. In the hpol ⅐(8-oxodG):rATP complex, 8-oxodG moved closer to the backbone of the template strand, whereas rATP shifted further into the major groove, compared with the hpol ⅐(8-oxodG):dATP complex. The distance between 3Ј-OH of the primer end to P␣ of rATP was 4.4 Å. By comparison, the corresponding distance in the hpol ⅐(8-oxodG):dATP complex was 3.3 or 3.1 Å (Fig. 7). Tyr-92 Acts as a Second Line of Defense to Stabilize the Steric Gate Residue Phe-18 at the Active Site of hpol -Superimpositions of the crystal structure with incoming rNTP and the corresponding structure with incoming dNTP revealed that the side chain of Phe-18 adopts an almost identical position in both cases. Upon closer investigation, a second guard residue, Tyr-92, was identified. The side chain of Tyr-92 is capable of forming ainteraction with the phenyl ring of Phe-18, with a distance of about 3.9 Å (Fig. 8). This stacking interaction appears to stabilize the side chain of Phe-18, leaving it in position and preventing the incoming nucleotide from sliding into the minor groove.
To address this hypothesis, we used site-directed mutagenesis to introduce the mutation Y92A in hpol and examined its activity in steady-state kinetic assays opposite template dG. The catalytic efficiency (k cat /K m ) of the mutant Y92A for dCTP insertion (1.6 M Ϫ1 min Ϫ1 ) was 67-fold higher than for rCTP (0.024 M Ϫ1 min Ϫ1 ), in comparison with the 770-fold difference for wild-type hpol ( Table 1), indicating that the introduction of the Y92A mutation reduced the sugar discrimination ability of hpol .

Discussion
Cellular rNTP concentrations are much higher than dNTP concentrations, thus presenting DNA polymerases with a challenge to discriminate against the former (6 -9). How DNA polymerases regulate rNTP incorporation has been studied for at least 20 years (6, 8, 16 -21). In general, a steric gate residue has been considered to constitute a physical barrier to prevent ribonucleotide incorporation by individual DNA polymerases (10 -17). The results of site-directed mutagenesis experiments illustrate the importance of the steric gate effect. For example, the Y39A mutation in hpol causes the enzyme to almost totally lose its ability to discriminate between the ribose and deoxyribose sugars (17). Also, the Sulfolobus solfataricus DNA polymerase Dpo4 Y12A mutant is capable of incorporating ribonucleotides into primers, and alanine, in place of the steric gate residue Tyr-12, allows space for the 2Ј-OH of the incoming ribonucleotide, as seen in the x-ray crystal structure (16). A recently published study of yeast pol shows that mutation of the steric gate residue Phe-35 to an alanine leads to increased ability of ribonucleotide incorporation (47).
2Ј-OH of the Incoming Ribonucleotide Leads to a Propeller Twist between the Paired Bases at the Active Site of hpol and Slows the Reaction-We crystallized wild-type hpol with primer-template DNA duplexes containing either dG or 8-ox-odG opposite the incoming rNTP. Interestingly, the extra 2Ј-OH of the incoming rNTP does not directly point into the phenyl ring of the steric gate residue. Instead, to avoid close contact, in the structure of rCTP opposite template dG, rCTP undergoes a shift relative to the position of dCTP in the corresponding crystal structure. However, the sugar pucker of the incoming nucleotide is of the C3Ј-endo type in both cases. Because of the shift of rCTP, a significant propeller twist occurs between cytosine of the incoming nucleotide and guanine of the template (Fig. 5). Similar situations were observed in the crystal structures of the hpol ⅐(8-oxodG):rCTP and hpol ⅐(8-ox-odG):rATP complexes (Figs. 6 and 7). The dihedral angle between the planes of the two bases in the three structures varies between 25 and 32°, compared with 3°in the hpol ⅐G: dCTP complex (PDB code 403N) (29). These particular structures accommodated by hpol are a consequence of its unique, unusually spacious active site that is relatively open on the major groove side. For example, a similar shift by the ribose sugar and the resulting significant propeller twist of the base pair at the active site were not observed in the structure of the wild-type hpol ␤⅐DNA:rCTP complex (PDB code 3RH4) (48). Noticeably, in each of the three structures in our study, the distance between the 3Ј-OH of the primer end to the P␣ of the incoming ribonucleotide is increased by 1 Å compared with the corresponding complexes with incoming dNTPs. As a result, the energy barrier for the nucleotidyl transfer reaction is elevated, and the overall reaction rate is reduced by about 3 orders of magnitude.
hpol Maintains Base Discrimination When the Incoming Nucleotide Is a Ribonucleotide-Our results show that hpol is still capable of discriminating bases among incoming rNTPs. The Watson-Crick base pair between an incoming rCTP and template dG is maintained, albeit with a propeller twist between bases because of different positioning of sugars of rNTPs relative to dNTPs (Fig. 5). Even with a DNA lesion (8-ox-odG or CPD) on the template strand, hpol still displays relatively high selectivity for the incoming nucleotide as a result of a generally conserved base pairing preference. Thus, it is noteworthy that the catalytic efficiency for rCTP incorporation opposite 8-oxodG is 19-fold higher than that of rATP insertion. By comparison, the change between the catalytic efficiencies of the correct insertion of dCTP opposite the lesion and erroneous insertion of dATP is only 5-fold. In addition, the orientations of 8-oxodG in the template strand vary somewhat in the crystal structures of the hpol ⅐(8-oxoG):rCTP and hpol ⅐(8-oxoG)⅐rATP ternary complexes (Figs. 6 and 7). These results indicate that the incoming rNTP may influence the position and conformation of the template lesion to some extent and therefore affect base selectivity.
Second Line of Defense That Stabilizes the Steric Gate Residue Has Been Observed in Other DNA Polymerases-In our study, we identified Tyr-92 as the second "guard" stabilizing the steric gate residue Phe-18 byinteraction, i.e. loss of Tyr-92 makes the polymerase an order of magnitude less discriminating. This second line of defense is also seen in some other DNA polymerases, in comparing their structures. For example, hpol Tyr-102 acts as the second guard, stacking with the steric gate residue Tyr-39 (17). In hpol , Tyr-112 constitutes the steric gate by forming a stacking interaction with Tyr-174 (49). Similarly, in DNA polymerase Dpo4, a pol homolog from S. solfataricus, the steric gate residue Tyr-12 is accompanied by Tyr-81 that serves as the second line of defense (50). Conversely, the replicative B-family bacteriophage RB69 uses a single DNA polymerase steric gate residue, Tyr-416, packed against an ␣-helix (19).
Biological Relevance of Ribonucleotide Incorporation by hpol -Considering that the cellular concentration of rNTPs is 10 1 -10 6 -fold higher that that of dNTPs (6 -9) and the catalytic efficiencies for ribonucleotide insertion are only about 10 3 -fold lower than those for deoxynucleotides, it is highly possible that hpol inserts a considerable amount of ribonucleotides into DNA as part of the various biological processes in which it is involved.
As a member of the DNA polymerase Y-family, one of the most important roles of hpol is the bypass of lesions during DNA replication. The base selectivity of hpol mainly depends on the type of the template base/lesion and the flanking DNA sequence (e.g. results in Tables 1-3). However, in this study, we also investigated the ability of hpol to discriminate between ribose and deoxyribose sugars of the incoming nucleotides, especially opposite the DNA lesions 8-oxodG and CPD in the template strand. Judging from our study, the active site of hpol can simultaneously accommodate a lesion on the template strand and a ribo-sugar of the incoming nucleotide, while still maintaining significant base selectivity.
Author Contributions-Y. S. designed and conducted the experiments, crystallized the proteins, and solved the structures. Y. S., F. P. G., and M. E. conceived the studies, analyzed the results, and wrote the paper.