Escherichia coli DNA polymerase I (Klenow fragment) uses a hydrogen-bonding fork from Arg668 to the primer terminus and incoming deoxynucleotide triphosphate to catalyze DNA replication.

Interactions between the minor groove of the DNA and DNA polymerases appear to play a major role in the catalysis and fidelity of DNA replication. In particular, Arg668 of Escherichia coli DNA polymerase I (Klenow fragment) makes a critical contact with the N-3-position of guanine at the primer terminus. We investigated the interaction between Arg668 and the ring oxygen of the incoming deoxynucleotide triphosphate (dNTP) using a combination of site-specific mutagenesis of the protein and atomic substitution of the DNA and dNTP. Hydrogen bonds from Arg668 were probed with the site-specific mutant R668A. Hydrogen bonds from the DNA were probed with oligodeoxynucleotides containing either guanine or 3-deazaguanine (3DG) at the primer terminus. Hydrogen bonds from the incoming dNTP were probed with (1 'R,3 'R,4 'R)-1-[3-hydroxy-4-(triphosphorylmethyl)cyclopent-1-yl]uracil (dcUTP), an analog of dUTP in which the ring oxygen of the deoxyribose moiety was replaced by a methylene group. We found that the pre-steady-state parameter kpol was decreased 1,600 to 2,000-fold with each of the single substitutions. When the substitutions were combined, there was no additional decrease (R668A and 3DG), a 5-fold decrease (3DG and dcUTP), and a 50-fold decrease (R668A and dcUTP) in kpol. These results are consistent with a hydrogen-bonding fork from Arg668 to the primer terminus and incoming dNTP. These interactions may play an important role in fidelity as well as catalysis of DNA replication.

The high fidelity of DNA replication synthesis is accomplished despite the similarity in energy between correctly and incorrectly paired bases (1,2). Since inter-strand hydrogen bonds cannot account for this selectivity (3,4), other mechanisms have been proposed to supply the fidelity (reviewed in Ref. 5). These include solvation (6), base stacking (7,8), steric exclusion (3,4,9), and minor groove binding (10). These mechanisms are not mutually exclusive and may all enhance selection for the correct base pairs.
X-ray crystallographic studies of polymerases that have a structure similar to that of KF Ϫ (Taq (13) Bacillus stearothermophilus (12) and T7 (14) DNA polymerases) predict a hydrogen-bonding interaction between an arginine residue (Arg 668 of KF Ϫ ) and the N-3-position of a purine or the O 2 -position of a pyrimidine at the primer terminus. Site-specific mutagenesis of Arg 668 implicated the importance of this residue in the catalysis (16,28) and fidelity of DNA replication (20). Nucleotide analog studies have indicated that the N-3-position of a purine at the primer terminus is essential to catalysis (24,26) and fidelity 2 of DNA replication.
The question that we address in this manuscript is how the interaction at the primer terminus influences the rate of incorporation of the incoming dNTP. The ternary crystal structures of Taq (13) and T7 (14) DNA polymerases bound to DNA and an incoming dNTP show that while one imino group of the critical arginine residue is in hydrogen bonding distance with the O 2 -cytosine (Taq polymerase) or N-3-adenine (T7 polymerase) with primer terminus, the other imino group may potentially form a hydrogen bond with the ring oxygen of the incoming dNTP. This interaction may be the mechanism by which disruption of the arginine-primer terminus hydrogen bond decreases the rate of polymerization. To examine this hypothesis we compared the kinetic parameters of the reactions containing wild-type Escherichia coli DNA polymerase I (Klenow fragment) with the exocyclic exonuclease activity inactivated (KF Ϫ ) or the R668A mutant, DNA containing guanine or 3-deazaguanine at the primer terminus, and dUTP or its carbocyclic analog (1ЈR,3ЈR,4ЈR)-1-[3-hydroxy-4-(triphosphorylmethyl)cyclo-* This work was supported by National Institutes of Health Grant CA 75074. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. pent-1-yl]uracil (dcUTP). In each of these nucleotide analogs, a heteroatom is replaced by a carbon (Fig. 1). Thus, if a hydrogenbonding interaction from Arg 668 to either position was important, then the analogs should exhibit reduced rates of reaction or reduced binding to the polymerase.

EXPERIMENTAL PROCEDURES
General-[ 32 P]ATP was purchased from Amersham Biosciences at 6,000 Ci/mmol. T 4 polynucleotide kinase and the Klenow fragment of E. coli DNA polymerase I with the exonuclease activity inactivated (KF Ϫ ) were obtained from United States Biochemical. The double mutant R668A and D424A KF was a gift from Catherine Joyce of Yale University. The D424A mutation knocks out the proofreading exonuclease activity. The dNTPs (ultrapure grade) were purchased from Amersham Biosciences, and the concentrations were determined by UV absorbance (29). dcUTP was synthesized as described in the supplemental literature (see supplemental material).
The oligonucleotides containing 3DG were synthesized, purified by PAGE followed by reverse-phase HPLC, and characterized by enzymatic hydrolysis with HPLC analysis (23,24). The concentrations of oligodeoxynucleotides were determined from the absorbance at 260 nm, using the method of Borer (30) in which it was assumed that the spectroscopic properties of 3DG were identical to G. The primer was 32 P-labeled with [␥-32 P]ATP and annealed with a 50% excess of the template as described previously (24).
Pre-steady-state Kinetics-Rapid reactions were initiated by the addition of 15.9 l of dNTP and MgCl 2 in water to 16.4 l of DNA-enzyme solution at 25°C with a KinTek-3 rapid quench instrument. Slow reactions were carried out by hand. The composition of the buffer during the reaction was 50 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , 5 mM DTT, 100 g/ml BSA. Typically the DNA concentration was 20 nM, and the polymerase concentration was 50 nM. The concentration of dNTPs varied from 0 to 1,000 M. The reactions were quenched by the addition of 300 mM EDTA.
Product Analysis by PAGE-The progress of the reaction was analyzed by denaturing PAGE in 20% acrylamide (19:1, acrylamide:N,NЈmethylene bisacrylamide), 7 M urea in 1ϫ TBE buffer (0.089 M Tris, 0.089 M boric acid, 0.002 M Na 2 EDTA). The size of the gel was 40 ϫ 33 ϫ 0.4 cm and was run at 2,500 V for 2-2.5 h. The radioactivity on the gel was visualized with a Bio-Rad GS 250 Molecular Imager. The progress of the reaction was quantitated by dividing the total radioactivity in the product band(s) by the radioactivity in the product and reactant bands. Multiple product bands appeared when the incorrect dNTP was added to the reaction.
Data Analysis-Data were fitted by nonlinear regression using the program Prism version 4.0 for Windows (GraphPad Software, San Diego, CA (www.graphpad.com)). The time course data were fit to Equa-tion 1, where P is the product formed, A is the total amount of DNA reacted, and k is the first order rate constant for the dNTP incorporation. The k pol and K d values were obtained by fitting k to [dNTP] according to Equation 2.

RESULTS
The interactions between Arg 668 , the minor groove of the primer terminus, and the ring oxygen of the incoming dNTP were evaluated kinetically using R668A, oligodeoxynucleotides with 3DG at the primer terminus, and dcUTP. These substitutions will be effective in examining the proposed interactions only if each modification does not drastically alter the chemistry of the enzyme or nucleotide analog except at the site of modification. Arg 668 has been shown to be important to catalysis and fidelity of DNA replication (16,20,28,31), but the overall structure of the protein appears unaffected by the substitution (28). 3DG has been used to examine minor groove interactions and is a good substrate at some sites in the DNA but not others and therefore, is a good mimic of guanine (24,32). Several nucleotide analogs in which the deoxyribose moiety was replaced with a carbocycle have been examined and they inhibit DNA synthesis by competitive inhibition or chain termination depending on the polymerase (33)(34)(35). The structure of a carbocyclic analog of dA paired with dT in an oligodeoxynucleotide duplex is similar to that of dA suggesting that dcUTP should be a good mimic of dUTP (36).
The incorporation of dUTP and dcUTP opposite dA was examined by reacting 50 nM KF Ϫ with 20 nM of the appropriate primer-template. Since [KF Ϫ ] Ͼ [DNA], the formation of product could be fit to a first order equation. The first order rate constants were plotted against the concentration of the incoming dNTP to obtain the k pol and K d values. The kinetic parameters are listed in Table I, and the relative k pol values are presented in Fig. 2.
As was observed previously, the Arg to Ala and G to 3DG substitutions support the existence of a critical hydrogen bond between Arg 668 and the minor groove of the primer terminus (24). In the present study, the G to 3DG substitution caused a 2,000-fold decrease in k pol (reaction 1 versus 2 in Table I), while with G as the template base the Arg 668 to Ala substitution induced a 1,600-fold decrease in k pol (reaction 1 versus 3). The double substitution of G to 3DG and Arg 668 to Ala 668 did not cause any additional decrease in k pol from the single modifications (reactions 2 and 3 versus 4). The observation that the effects of the two substitutions were not cumulative indicates that there is a hydrogen bond interaction between Arg 668 and the N-3-position of guanine at the primer terminus (24).
In similar experiments, we examined the relationship between Arg 668 and the deoxyribose ring oxygen of the incoming dNTP. The replacement of dUTP by dcUTP decreased the k pol ϳ1,200-fold with a 4-fold increase in K d (reaction 1 versus 5). Thus, the ring oxygen is very important to reactivity but only slightly affects the binding of the dNTP. The double substitution of Arg to Ala and dUTP to dcUTP resulted in a 50-fold decrease in k pol from either of the single substitutions (reactions 3 and 5 versus 7). This decrease in rate falls between no effect and the ϳ1,500-fold decrease in k pol caused by the single substitutions. This result suggests that there is an interaction between Arg 668 and the ring oxygen. However, the lack of this hydrogen bond does not entirely explain the decrease in k pol caused by the dUTP to dcUTP substitution.
We also examined the relationship between the ring oxygen of dUTP and the 3-position of guanine at the primer terminus. As discussed above, the decrease in rates associated with the Hydrogen-bonding Fork from Arg 668 to DNA and dNTP 33044 individual substitution of G with 3DG and dUTP with dcUTP indicate the importance of these sites (reaction 1 versus 2 and 5). The double substitution decreased the k pol only 5-fold from the single substitutions (reactions 2 and 5 versus 6). The rate reduction is insignificant with respect to the expected 2,000fold reduction that would be observed if the substitution were independent of each other. Consequently, these reactions indicate the ring oxygen of dUTP interacts with the minor groove of the primer terminus of the DNA. Since the crystal structures do not indicate a direct interaction between these two positions, a likely explanation is that they interact through the arginine fork.
Examination of the double substitutions versus the triple substitutions shows similar results and supports a mechanistic connection between the three moieties. No decrease in k pol was observed with the G to 3DG substitution to the reaction involving R668A and dcUTP (reaction 7 versus 8). The decrease in rate for the substitution of R668A for KF Ϫ with dcUTP and 3DG is only 10-fold (reaction 6 versus 8). A 100-fold decrease in rate was observed with the dUTP to dcUTP change with R668A and 3DG on the DNA (reaction 5 versus 8).
In summary, the substitutions significantly affect the k pol and only slightly the K d parameter. The results strongly support two interactions between 1) Arg 668 and the N-3-position of G at the primer terminus and 2) the deoxyribose ring oxygen of the incoming dNTP and N-3-position of G at the primer terminus. In addition, there also appears to be a connection between Arg 668 and the deoxyribose ring oxygen of the incoming dNTP. DISCUSSION One mechanism by which polymerases can replicate Watson-Crick base pairs is by geometric selection. For this to occur, the polymerases must have contact points with the DNA at positions in which the topology and chemistry is indistinguishable among the four base pairs. The minor groove is thought to be important for base pair independent contacts because the N-3position of purines and the O 2 -position of pyrimidines occupy similar steric positions and have similar chemistry, that of being hydrogen bond acceptors (10). In support of this hypothesis, x-ray crystal structures of several polymerases have shown that there are interactions between the minor groove of the DNA and the polymerase (11)(12)(13)(14)(15).
A critical contact between KF Ϫ and the minor groove of the DNA is made by Arg 668 (24,37). This contact is essential for both the catalytic efficiency (16,28) and fidelity 2 of the protein and may play a role in the transfer of the DNA between the catalytic and exonuclease sites (22). In this manuscript, we have described experiments indicating that in addition to forming a hydrogen bond with the N-3-position of guanine at the primer terminus it also forms a hydrogen bond with the deoxyribose ring oxygen of the incoming dNTP.
Ternary crystal structures of two family A polymerases, T7 and Taq DNA polymerases, support the existence of this hydrogen bonding fork. These crystal structures contain DNA and an incoming dideoxynucleotide triphosphate bound to the polymerase in the "closed conformation" and may represent a catalytically active intermediate in the reaction. The crystal structure of T7 DNA polymerase (Protein Data Bank code 1t7p) shows that one imino group of Arg 429 is 2.86 Å from the N-3position of adenine at the primer terminus while the other imino group is 2.82 Å from the deoxyribose ring oxygen of the incoming 2Ј,3Ј-dideoxyadenosine 5Ј-triphosphate (14). With Taq DNA polymerase (Protein Data Bank code 3ktaq) the respective distances are 2.58 and 3.44 Å (13). These distances are short enough to speculate that hydrogen bonds occur between these groups during catalysis.
The interaction of Arg 668 with the minor groove of the primer terminus and the incoming dNTP can aid in recognizing whether the DNA has Watson-Crick geometry. The distance from the deoxyribose ring oxygen to the minor groove hydrogen bond acceptor on the 5Ј-base is ϳ3.5 Å with correctly paired bases. However, this distance is altered on both sides of most mispairs. Thus, in addition to the relative position of the minor groove hydrogen bond acceptors, the absolute distance between the deoxyribose ring oxygen and the adjacent minor groove  2. A, representation of the reaction studied. Amino acid 668 is either Arg (shown) or Ala, the primer terminus is G (shown) or 3DG (DG), and the incoming nucleotide is dUTP (shown) or dcUTP (cU). B, relative k pol determined with 20 nM DNA and 50 nM KF Ϫ in 50 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , 5 mM DTT, 100 g/ml BSA at 25°C. The k pol is relative to that performed with KF Ϫ , unmodified DNA, and dUTP. The lower section shows the changes associated with each bar.
Hydrogen-bonding Fork from Arg 668 to DNA and dNTP 33045 hydrogen bond acceptor may be a physical property that the polymerase can check to determine whether a base pair is in the Watson-Crick geometry. Thus, these interactions may explain how the rates of insertion are decreased with mispairs at either the terminal base pair or the newly forming base pair. Two mechanisms can be envisioned of how Arg 668 would enhance the catalytic efficiency of correct base pairs but not mispairs, one in which Arg 668 plays an active role in selecting Watson-Crick base pairs and another in which the residue plays a regulatory role. Arg 668 could actively be involved in phosphodiester bond formation by aligning the incoming dNTP into correct position for reaction to occur. Its function may be to align the ␣-phosphate of the incoming dNTP with the 3Ј-hydroxyl group of the primer terminus. Alternatively, Arg 668 may lower the activation energy of phosphodiester bond formation by pulling the dNTP onto the primer terminus. Due to the relative positions of the correct versus incorrect base pairs, this process would occur only with Watson-Crick base pairs at the terminal base pair and the newly forming base pair. With a mispair at either position, the arginine fork would help to orient the incoming dNTP into an inactive configuration and thus, decrease the rate of replication.
Arg 668 may also modulate the activity of the polymerase by an indirect mechanism. The position of Arg 668 may be dependent on the positions of the primer terminus and incoming dNTP. If they are not in Watson-Crick geometry then the altered Arg 668 position may disrupt the overall geometry of the catalytic site to inhibit the rate-limiting conformational change and phosphodiester bond formation. Since Arg 668 is on the interface of the fingers and palm domain its position may be vital to the overall structure of the binding/active site.
The kinetics of E. coli DNA polymerase I can be described by seven steps illustrated in Fig. 3 (38 -40). The rate-limiting step for correct replication is the conformational change (step 3) (40). However, the rate of phosphodiester bond formation slows down to become rate-limiting for mispair formation (38). Thus, the decrease that we observed in k pol could be due to either a decrease in the rate-limiting conformational change or phosphodiester bond formation. If Arg 668 plays an active role, we would speculate that these changes would affect phosphodiester bond formation, while if it is a regulatory role we would speculate that the changes affect the conformational change.