Mutational Studies of Human DNA Polymerase α LYSINE 950 IN THE THIRD MOST CONSERVED REGION OF α-LIKE DNA POLYMERASES IS INVOLVED IN BINDING THE DEOXYNUCLEOSIDE TRIPHOSPHATE

The function of a lysine residue, Lys950, of human DNA polymerase α located in the third most conserved region and conserved in all of the α-like polymerases was analyzed by site-directed mutagenesis. Lys950 was mutagenized to Arg, Ala, or Asn. The mutant enzymes were expressed in insect cells infected with recombinant baculoviruses and purified to near homogeneity. The mutant enzymes had specific activities ranging from 8 to 22% of the wild type. All three Lys950 mutants utilized Mn2+ as metal activator more effectively than the wild type enzyme and showed an increase in Kmvalues for deoxynucleoside triphosphate but not kcat values in reactions with either Mg2+ or Mn2+ as the metal activator. Although mutation of the Lys950 residue caused an increase in Km values for deoxynucleoside triphosphates, mutations of Lys950 to Arg, Ala, or Asn did not alter the mutant enzymes' misinsertion efficiency in reactions with Mg2+ as a metal activator as compared with that of the wild type, suggesting that the base of the incoming deoxynucleoside triphosphate is not the structural feature interacting with the Lys950 side chain. In reaction with Mn2+ as a metal activator, all three Lys950 mutants had an improved fidelity for deoxynucleotide misinsertion compared to wild type. Inhibition studies of the three Lys950 mutant derivatives with an inhibitor, structural analogs of deoxynucleoside triphosphate, and pyrophosphate suggest that the deoxyribose sugar and β-,γ-phosphate groups are not the structural feature recognized by the Lys950 side chain. Comparison of the mutant enzymes to the wild type enzyme for their affinities for dCTPαS versus deoxynucleoside triphosphate suggests that this highly conserved Lys950 is involved in interacting either directly or indirectly with the oxygen moiety of the α-phosphate of the incoming deoxynucleoside triphosphate.

The function of a lysine residue, Lys 950 , of human DNA polymerase ␣ located in the third most conserved region and conserved in all of the ␣-like polymerases was analyzed by site-directed mutagenesis. Lys 950 was mutagenized to Arg, Ala, or Asn. The mutant enzymes were expressed in insect cells infected with recombinant baculoviruses and purified to near homogeneity. The mutant enzymes had specific activities ranging from 8 to 22% of the wild type. All three Lys 950 mutants utilized Mn 2؉ as metal activator more effectively than the wild type enzyme and showed an increase in K m values for deoxynucleoside triphosphate but not k cat values in reactions with either Mg 2؉ or Mn 2؉ as the metal activator. Although mutation of the Lys 950 residue caused an increase in K m values for deoxynucleoside triphosphates, mutations of Lys 950 to Arg, Ala, or Asn did not alter the mutant enzymes' misinsertion efficiency in reactions with Mg 2؉ as a metal activator as compared with that of the wild type, suggesting that the base of the incoming deoxynucleoside triphosphate is not the structural feature interacting with the Lys 950 side chain. In reaction with Mn 2؉ as a metal activator, all three Lys 950 mutants had an improved fidelity for deoxynucleotide misinsertion compared to wild type. Inhibition studies of the three Lys 950 mutant derivatives with an inhibitor, structural analogs of deoxynucleoside triphosphate, and pyrophosphate suggest that the deoxyribose sugar and ␤-,␥-phosphate groups are not the structural feature recognized by the Lys 950 side chain. Comparison of the mutant enzymes to the wild type enzyme for their affinities for dCTP␣S versus deoxynucleoside triphosphate suggests that this highly conserved Lys 950 is involved in interacting either directly or indirectly with the oxygen moiety of the ␣-phosphate of the incoming deoxynucleoside triphosphate.
Compilation and alignment of the protein sequences of DNA polymerases deduced from the nucleotide sequence data have classified DNA polymerases into three families, family A, B, and C, according to their similarities to Escherichia coli polymerase I, II, and III, respectively (1,2). The ␣-like DNA polymerases including the E. coli polII belongs to family B with polymerase ␣ as the prototype (3). The lack of an intrinsic proofreading nuclease in DNA polymerase ␣ makes this enzyme an ideal model to study the structure-function relationship of the active site and serves as a model for all of the ␣-like DNA polymerases, particularly in identifying residues in the active site that are responsible for DNA synthetic fidelity (4,5). We have overproduced functionally active recombinant human DNA polymerase ␣ catalytic subunit in insect cells infected with a recombinant baculovirus (6) and established a one-step immunoaffinity purification protocol to purify the enzyme to near homogeneity (7). By site-directed mutagenesis followed by physical and steady-state kinetic studies, several highly invariant residues in the catalytic site of human DNA polymerase ␣ were analyzed (4,5,8,9). We have established that three residues, -Asp-Thr-Asp-, in the most conserved region of the ␣-like DNA polymerases (region I, -YGDTDS-) are involved in metal activator binding. The two aspartate residues, Asp 1002 and Asp 1004 , like the aspartate residues in the active site of Klenow and HIV reverse transcriptase forming the "catalytic triad," directly participate in chelating the metal ion (8 -12). Mutations of Asp 1002 to Asn and Thr 1003 to Ser also yielded mutant polymerases as metal ion-induced anti-mutators (8). Thus, residues Asp 1002 , Thr 1003 , and Asp 1004 in the most conserved region also play critical roles in the observed metal-induced infidelity in DNA synthesis of cellular DNA polymerases.
Five highly conserved residues in the second most conserved region (region II) were also analyzed. Mutation of a conserved glycine, Gly 841 , appears to affect both catalysis and substrate dNTP binding, suggesting that this glycine residue is essential for the maintenance of the overall active site structure. Mutations of Tyr 865 altered the affinity of the mutant enzyme to bind the incoming dNTP. Analysis of the Tyr 865 mutant enzyme for its DNA synthetic fidelity has shown that the phenyl ring side chain of Tyr 865 directly interacts with the nucleoside base moiety of the incoming dNTP and plays a critical role in nucleotide misinsertion fidelity of DNA synthesis (5). Mutation analyses of the second serine residues in the conserved region, -SLYPSI-, have revealed that the hydroxyl side chain of this serine residue, Ser 867 , directly interacts with the 3Ј-OH terminus of the primer and plays an essential role in mispaired primer extension fidelity of DNA synthesis (4).
In this report, we continue to investigate the contributions of those highly conserved amino acid residues in the catalytic site of ␣-like DNA polymerases for either substrate binding or for catalysis. We analyzed an invariant lysine residue in the third most conserved region (region III) that is conserved from human DNA polymerases ␣, ␦, and ⑀ to E. coli polymerase II.

Methods
Site-directed Mutagenesis-The cloning and in vitro mutagenesis strategy was as described (9). Briefly, a 1.44-kilobase SalI-BamHI fragment of the human polymerase ␣ cDNA containing the conserved regions (13) was cloned into the SalI-BamHI site of M13mp19 for site-directed mutagenesis in vitro (9). The mutations were verified by restriction enzyme analysis and DNA sequencing.
Construction of Recombinant Baculovirus-The strategy for construction of transfer vector was performed as described in Ref. 9. Briefly, the SalI-BamHI fragments containing the site-directed mutation were isolated and used to replace the SalI-BamHI fragment of the plasmid pBR(XbaI) HDP␣ which contains the full-length coding sequence of human DNA polymerase ␣ (9). A 4.6-kilobase XbaI-XbaI fragment of pBR(XbaI)HDP␣ containing the full-length mutated human polymerase ␣ cDNA was then isolated and constructed into the XbaI site of pVL1393, to generate the transfer vector pVL1393/SDM containing the site-directed mutation in the polymerase ␣ cDNA.
Recombinant baculoviruses were generated by co-transfection of Sf9 cells with the transfer plasmids and linear baculovirus DNAs as described (9). Five micrograms of transfer plasmid DNA were co-transfected with 1 g of linear Ac␤Gal viral DNA, and the recombinant viruses were selected using standard baculovirus techniques.
Expression and Purification of Mutants-The amplification, infection, and harvest of recombinant baculoviruses expressing the mutant polymerase ␣ were performed as described (9). The recombinant DNA polymerase ␣ proteins were purified to near homogeneity from Sf9 insect cell lysates by the one-step immunopurification protocol with monoclonal antibody SJK237-71 cross-linked to Sepharose 4B as described (7).
DNA Polymerase Assay and Kinetic Analysis-The standard assay for DNA polymerase ␣ activity using optimally gapped calf thymus DNA as primer-template was as described (14). One unit of polymerase activity is defined as the amount of DNA polymerase that incorporates 1 nmol of labeled dNTP into acid-insoluble DNA at 37°C in 1 h with 10 mM MgCl 2 . All kinetic assays and metal ion optimum curves were performed with optimally gapped calf thymus DNA which was extensively dialyzed to remove the Mg 2ϩ ion as described (9). The metal optimum curves, k cat and K m(dNTP) , were determined using optimally gapped calf thymus DNA as the primer-template substrate. K m for primer-template was determined by varying the concentrations of oligo(dT) 12 /poly(dA) as described (6). Initial rate kinetic data were analyzed by using the computer program CRICKET GRAPH.
Processivity-Processivity assays for wild type and mutant enzyme were performed by using the DNA trap method described in Ref. 15. Oligo(dT) 12  Misinsertion Fidelity-The fidelity of misinsertion by wild type and mutant polymerase ␣ was measured by the gel electrophoresis assay described in Refs. 5 and 16 -18. The standing start primer-template used to measure correct or incorrect insertion was: 5Ј-TGA CCA TGT AAC AGA GAG-3Ј (18-mer) and 3Ј-ACT GGT ACA TTG TCT CTC ATT CTC TCT CTC TTC TCT-5Ј (36-mer), where the bold and underlined nucleotide on the template indicates the position of insertion of either the correct dTMP or the incorrect dCMP. The oligonucleotides were gel-purified before use. The primer was 5Ј-end-labeled with 32 P and annealed 1:2 to the template. The primer-template (0.5 pmol) was mixed with 0.3 unit of the enzyme in a 10-l reaction buffer containing 20 mM Tris-HCl, pH 8.0, 1 mM ␤-mercaptoethanol, 200 g/ml acetylated bovine serum albumin, and either with 10 mM MgCl 2 or with 0.75 mM MnCl 2 as metal activator. The reaction was initiated by the addition of either dTTP (correct) or dCTP (incorrect) to fill the I 1 position in the primer for 2 min at 37°C and terminated on a dry ice bath followed by the addition of an equal volume of 95% formamide sequencing gel loading dye. Reaction products were boiled for 5 min and separated on a 13% polyacrylamide sequencing gel. After electrophoresis, gels were analyzed by PhosphorImager (Molecular Dynamics). I 0 designates the primer band, and I 1 designates the site of correct or incorrect insertion. Velocity was measured as I 1 /(I 0 ϩI 1 ), and K m and V max values were deduced from Lineweaver-Burk plots.
Single-stranded DNA Inhibition-Wild type or mutant enzyme (1.4 pmol) was preincubated with varying concentrations of (dT) 174 for 5 min at 0°C in a 10-l incubation and then assayed under standard conditions with optimally gapped DNA as primer-template.
Inhibition Assay-Enzymatic assays with (40 to 400 pmol) of wild type or mutant polymerase ␣ enzymes were carried out in the presence and absence of inhibitors in a reaction mixture containing 20 mM HEPES (pH 8.0), 2 mM ␤-mercaptoethanol, 200 g/ml bovine serum albumin, 10 mM MgCl 2 , 50 M dNTP, and 800 g/ml gapped calf thymus DNA at 37°C for 10 min. The concentration of inhibitors that produced 50% inhibition (IC 50 ) are mean values from two or three independent assays.

Site-directed Mutation of a Highly Conserved Lysine Residue
A lysine residue located in the third most conserved regions of the ␣-like DNA polymerases is invariantly present in all three major mammalian cellular DNA polymerases ␣, ␦, and ⑀, in yeast POLI, -II, and -III, in several DNA virus polymerases, in E. coli polII, in polymerases of bacteriophage T4, PRD1, and 29, and in polymerase-like proteins such as PGKL1 and mitochondria S1 (Fig. 1). In this study, we used human DNA polymerase ␣ as the model for all of the three cellular polymerases to analyze the functional role of this highly conserved lysine residue. By site-directed mutation, we changed Lys 950 to Arg, thereby replacing the side chain with a larger positively charged side chain, Lys 950 to Ala, thereby completely abolishing the positively charged side chain, and Lys 950 to Asn, thereby replacing the ⑀-NH 3 group of the positively charged lysine side chain with a polar amide group. The three mutant polymerases were produced in recombinant baculovirus-infected insect cells and purified with the one-step immunopurification protocol to near homogeneity in high yield (7). Analysis of these three mutant proteins produced from recombinant baculovirus-infected insect cells showed that all three had a predominant protein of 180 kDa like the wild type enzyme with minor species of 165-140 kDa (data not shown). Furthermore, these three mutant proteins like the mutant proteins reported by us before had no detectable global structural alterations (4,5,9). The specific activities of each mutant DNA polymerase ␣ were measured by using optimally gapped calf thymus DNA as primer-template and with either Mg 2ϩ or Mn 2ϩ as metal activator. In reactions with Mg 2ϩ as the metal activator, mutant K950A had 22% of the specific activity of the wild type enzyme, while mutants K950R and K950N had 7.6% and 8% of the wild type specific activity, respectively. In reactions with Mn 2ϩ as metal ion, mutant K950R had 42% of the wild type specific activity, mutant K950A had 70% of the wild type specific activity, while mutant K950N had specific activity identical with, if not slightly higher than, the wild type enzyme (Table I).

Kinetic Parameters of the Lys 950 Mutants
Kinetic parameters of these three mutants in reactions with either Mg 2ϩ or Mn 2ϩ as the metal activator were measured and compared to the wild type (Table I). We found that mutations of Lys 950 to Arg, Ala, or Asn had a profound effect on the K m values of dNTPs in reactions with Mg 2ϩ as the metal activator. Mutants K950R, K950A, and K950N showed 76-, 25-, and 42-fold increases of K m for dNTP, respectively, as compared to the wild type. These mutants, however, only showed moderate 2-to 5-fold decreases in their k cat values in reactions with Mg 2ϩ . Moreover, in reactions with Mg 2ϩ , all three Lys 950 mutant derivatives showed striking decreases in their processivity to the nearly distributive mode of DNA synthesis. All of the three Lys 950 mutant enzymes also showed moderate decreases in their K m values for primer terminus compared to the wild type enzyme in reaction with Mg 2ϩ . These results suggest that in reactions with Mg 2ϩ as the metal activator, the observed lower specific activities of the Lys 950 mutant derivatives as compared with the wild type are due to the combination of increases in the K m for dNTP and decreases in k cat and processivity.
In reactions with Mn 2ϩ as the metal activator, mutants K950R, K950A, and K950N had 8-, 4-, and 6-fold increases in K m for dNTPs and moderate 3-, 3.5-, and 4.5-fold increases in their k cat , as compared with the wild type, respectively. In reactions with Mn 2ϩ , these three mutant enzymes also had comparable DNA synthetic processivity and K m values for primer terminus as the wild type enzyme (Table I). These kinetic parameters thus render the mutant enzymes with specific activities comparable with the wild type enzyme in reac-tions with Mn 2ϩ .
Since mutation of Lys 950 to Arg or Ala but not Asn increased the mutant's affinity (decreased the K m ) for primer terminus in reactions with Mg 2ϩ as the metal activator, we also investigated the effect on template interaction when the positively charged side chain of this lysine residue is replaced by either a larger size charged side chain or is completely abolished. Inhibition by single-stranded DNA was compared, and no apparent difference was found between wild type and all three mutant enzymes (data not shown). This result indicates that this lysine residue is not involved in template interaction.
Results of these kinetic parameter studies suggest that mutation of this highly conserved Lys 950 residue primarily affects the affinity for the incoming dNTP substrate and does not affect catalysis.

Effect of Metal Activator
The observed differences in these mutant enzymes' kinetic parameters in reactions with Mg 2ϩ from that of Mn 2ϩ led us to investigate these mutant enzymes' preference of metal activator. The optimal concentrations of each metal activator for the wild type and the three mutant enzymes were compared and are shown in Fig. 2. Like what we observed before (9), the wild type enzyme prefers to utilize Mg 2ϩ as the metal activator with an optimal concentration at 10 mM and utilizes Mn 2ϩ as metal activator poorly with an optimal concentration at approximately 0.5 mM. In contrast, all three Lys 950 mutant enzymes were able to utilize Mn 2ϩ as metal activator more effectively than the wild type enzyme (Fig. 2). Mutant enzymes K950R and K950N had similar optimal concentrations for Mg 2ϩ and   (Fig. 2C), whereas mutant enzyme K950R had 70% of the specific activity in reaction with Mn 2ϩ as compared with reaction with Mg 2ϩ , and mutant enzyme K950N had a higher specific activity in reaction with Mn 2ϩ than with Mg 2ϩ (Table  I and Fig. 2, B and D).

Misinsertion Fidelity
Given the ability of these three Lys 950 mutant enzymes to utilize Mn 2ϩ as the metal activator for catalysis like that of D1002N and T1003S in region I (8,9), we tested the misinsertion fidelity of these three Lys 950 mutants in reactions with either Mg 2ϩ or Mn 2ϩ as the metal activator. Using a standing start primer-template (see "Experimental Procedures"), we tested the incorporation of correct dTTP versus incorrect dCTP in reactions with Mg 2ϩ as the metal activator. We found that the misinsertion efficiency of the three Lys 950 mutants was identical with that of the wild type. In contrast, in reactions with Mn 2ϩ , our results show that mutations of Lys 950 had improved misinsertion fidelity over the wild type enzyme like that of the D1002N and T1003S mutant enzymes (8) (Table II). Mutant enzymes K950R and K950A had 6-and 9.6-fold improved misinsertion fidelity compared with the wild type, while mutant enzyme K950N showed a Ͼ1900-fold improved misinsertion fidelity over the wild type. Thus, the three Lys 950 mutant enzymes are metal-induced anti-mutators and the side chain of Lys 950 , like that of Asp 1002 and Thr 1003 (8), might have a role in the Mn 2ϩ -induced infidelity during DNA synthesis by DNA polymerase ␣.

Lys 950 Side Chain Has a Role in the Active Site and Is Involved in Interacting with dNTPs
Our finding of mutations of Lys 950 to Arg, Ala, or Asn affecting apparent K m values for dNTP in reactions with either Mg 2ϩ or Mn 2ϩ as the metal activator suggests that the side chain of Lys 950 may have a role in active site interacting with the incoming dNTP substrate. We, thus, used an inhibitor and several dNTP structural analogs to verify this notion.
We first tested the effect of an inhibitor, aphidicolin, on these three Lys 950 mutant enzymes. Aphidicolin is a general inhibitor for all three major cellular ␣-like DNA polymerases, ␣, ␦, and ⑀ (3). Aphidicolin acts as a competitive inhibitor of pyrimidine deoxynucleoside triphosphate, but, structurally, aphidicolin is not an analog of dNTPs. We have previously proposed a model of how aphidicolin forms hydrogen bonds with the purine base of the nucleotide in template in the active site of ␣-like DNA polymerases (5). To test if Lys 950 plays a role in the active site in interacting with metal activator(s) or the dNTP-metal activator complex, we tested the inhibitory effect of aphidicolin on the three mutant derivatives of Lys 950 and compared their 50% inhibition point to that of the wild type reactions with Mg 2ϩ as metal activator. The three Lys 950 mutant enzymes, K950R, K950A, and K950N, were 10, 14, and 33 times more sensitive to aphidicolin inhibition than the wild type enzyme, respectively (Table III). These results suggest that Lys 950 functions in the active site and aphidicolin affects the interaction between the Lys 950 side chain and the dNTP substrate.
We next compared the three Lys 950 mutant derivatives to wild type enzyme for their 50% inhibition points by an analog of dGTP, BuPdGTP, and its ␣,␤-methylene derivative, BuPdGMPCH 2 PP (Table III). All three Lys 950 mutant derivatives showed higher sensitivity to both of these compounds than the wild type enzyme. Mutant enzyme K950R had about 3 times higher sensitivity to both BuPdGTP and BuPdGMPCH 2 PP. Mutant enzyme K950A and mutant enzyme K950N both showed much higher sensitivity to the inhibition by BuPdGTP and BuPdGMPCH 2 PP than did the K950R (Table III). These indicate that the Lys 950 side chain indeed has a role in the active site and is involved in interacting with the incoming dNTP.
Results of the kinetic and inhibitor studies suggest that the positively charged side chain of Lys 950 has a function in the active site and is involved in interacting with the dNTP-metal activator complex.

Structure Feature of dNTP Recognized by Lys 950
We next investigated what structural feature of dNTP interacts with the Lys 950 side chain. The finding that all three Lys 950 mutant derivatives have misinsertion fidelity efficiency identical with the wild type enzyme in reactions with Mg 2ϩ as the metal activator (Table II)  type enzyme (Table IV). All three Lys 950 mutant enzymes showed a higher resistance to araCTP inhibition than the wild type enzyme did. In contrast, all three Lys 950 mutant enzymes showed higher sensitivity to ddCTP inhibition than the wild type enzyme. Mutant enzyme K950A with the entire positively charged side chain abolished had a 64-fold increase in its sensitivity to ddCTP and was 7-fold more resistant to araCTP than the wild type enzyme (Table IV). Thus, alteration of the furanose ring conformation of a dNTP has an effect on the interaction between a dNTP and the Lys 950 side chain. Since alterations of the deoxyribose sugar also affects the orientation of the triphosphate group of dNTP, we, therefore, analyzed whether the ␣-, ␤-, or ␥-phosphate group was the structural feature of a dNTP recognized by the Lys 950 side chain.
Does the Lys 950 Side Chain Recognize the ␤or ␥-Phosphate Group of the Incoming dNTP Substrate?-We tested whether the Lys 950 side chain has a role in properly positioning the triphosphate moiety of the incoming dNTP for metal activator chelation. Interaction between the phosphate groups of the dNTP substrate and the positively charged side chain of Lys 950 could also facilitate a nucleophilic attack of the incoming primer 3Ј-hydroxyl group for deoxynucleotidyl transfer. We, thus, investigated the effect of an altered phosphate group of dNTP on the reactivities of the three Lys 950 mutant enzymes and compared them to the wild type. In our inhibition studies of BuPdGTP and BuPdGMPCH 2 PP (Table III), the patterns and the extent of inhibition for the three Lys 950 mutant enzymes compared with the wild type enzyme were similar. Since the structural difference between these two compounds is the substitution of the oxygen group on ␤-phosphate of BuPdGTP with methylene (-CH 2 -), the result suggested that modification of the ␤-phosphate group does not have any effect on the interaction between the Lys 950 side chain and the incoming dNTP substrate. To verify this observation, we tested the effect of pyrophosphate (PP i ) and two analogs of PP i , carbonyldiphosphonate and phosphonoacetic acid, for their 50% inhibition point (IC 50 ). The three mutant enzymes were inhibited by pyrophosphate to an extent similar to the wild type enzyme (Table IV). Inhibition patterns of the three mutant enzymes and the wild type enzyme by the two pyrophosphate analogs were variable (Table IV). However, like the inhibition studies with araCTP and ddCTP, K950A, with the Lys 950 side chain abolished, showed the highest sensitivity to carbonyldiphosphonate and phosphonoacetic acid, as well as to BuPdGTP and BuPdGMPCH 2 PP. This suggests that the side chain of Lys 950 might be involved in interacting with the triphosphate moiety of the incoming dNTP substrate. Since the three mutant enzymes showed similar if not identical sensitivity to pyrophosphate, BuPdGTP, and its ␤-phosphate analog BuPdGMPCH 2 PP as the wild type enzyme, the ␤and ␥-phosphates of the incoming dNTP are not likely to be the structural feature directly interacting with the Lys 950 side chain.
Does the Lys 950 Side Chain Interact with the ␣-Phosphate Group of the dNTP?-The finding that these three Lys 950 mutant enzymes could utilize Mn 2ϩ more effectively than the wild type enzyme, had altered sensitivity to araCTP and ddCTP, and were insensitive to pyrophosphate inhibition led us to investigate whether the ␣-phosphate group of the incoming dNTP is the structural feature recognized by the Lys 950 side chain. We compared the affinity difference of the three Lys 950 mutant enzymes for an ␣-phosphate analog, dCTP␣S, and for dNTP (Table V) in reactions with either Mg 2ϩ or Mn 2ϩ as the metal activator. The dCTP␣S used in this study is the S p diastereomer which has been documented to be an active substrate for E. coli polymerase I (19).
In reactions with Mg 2ϩ , wild type enzyme had 22-fold higher affinity for dTTP than dCTP␣S (Table V) (Table V). Mutant enzymes K950A and K950N had 6 and 24 times lower affinity (higher K m values) for dCTP␣S than the wild type enzyme. These results indicate that substitution of the oxygen moiety of the ␣-phosphate by sulfur profoundly affects the affinity of the Lys 950 mutant enzymes for the ␣-phosphate analog. This suggests that the Lys 950 side chain interacts with the ␣-phosphate group of the incoming dNTP substrate.  . In contrast to the reactions with Mg 2ϩ as the metal activator, in reactions with Mn 2ϩ as metal activator, all three mutant enzymes as well as the wild type enzyme had lower affinity (higher K m values) for dCTP␣S than the wild type enzyme. Furthermore, the three mutant enzymes and the wild type enzyme had comparable or equal affinity ratio of dCTP␣S to Table V). We then compared the three mutant enzymes to the wild type enzyme for their catalysis rate (k cat ) in utilizing dCTP␣S versus dNTP as substrate. In reactions with Mg 2ϩ , the wild type enzyme had 2.6-fold higher k cat in utilizing normal dNTP versus dCTP␣S as substrate with a ratio of k cat S /k cat O of 0.38. Mutant enzyme K950R did not show a significant difference in its k cat value when either normal dNTP or dCTP␣S was used as substrate. Mutant enzyme K950A like the wild type enzyme had an approximately 2-fold higher k cat when dNTP was used as substrate than when dCTP␣S was used as substrate (k cat S ϭ 0.68 and k cat O ϭ 0.27). Mutant enzyme K950N, interestingly, had higher k cat when dCTP␣S was used as substrate versus dNTP (Table V).
When Mn 2ϩ was used as the metal activator, the wild type enzyme had identical k cat values in utilizing either dCTP or dCTP␣S as a substrate. Mutant enzymes K950R and K950N both had comparable k cat values in using dCTP␣S or normal dNTP as a substrate. Mutant enzyme K950A showed a 2-fold lower k cat in using dCTP␣S as the substrate than with dCTP as the substrate, like that observed in the Mg 2ϩ -catalyzed reaction.
By comparing the affinity (K m ) and catalysis (k cat ) of these mutant enzymes to the wild type enzyme for utilizing dNTP and dCTP␣S as substrate, it is apparent that substitution of the -PϭO by -PϭS in the ␣-phosphate group of dCTP profoundly affects the affinity of the enzyme's binding to the incoming dNTP substrate, but does not significantly affect the rate of catalysis.
These results together with the findings that the three Lys 950 mutant enzymes are able to utilize Mn 2ϩ as metal activator strongly suggest that the positively charged side chain of Lys 950 either directly or indirectly participates in interactions with the oxygen moiety of the ␣-phosphate group of the incoming dNTPs, either to position the dNTP to interact with the metal activator or to facilitate the nucleophilic attack by the 3ЈOH group of the incoming primer. DISCUSSION We have used the recombinant human DNA polymerase ␣ as the prototypic model for the three principal cellular DNA polymerases ␣, ␦, and ⑀ to elucidate the functional roles of several highly invariant amino acid residues in the active site. We altered several invariant residues by site-directed mutagenesis based on the rationale that we described (20). We generated a panel of mutants that did not have any detectable gross alteration of the protein structure (4,5,9,20). Thus, it is reasonable to assume that the structural alterations resulting from each mutation are confined to the position of the mutated side chains. By steady-state kinetic analysis of the mutant enzymes, we have defined the functions of several residues in the two most conserved regions (regions I and II) of the active site (4,5,8,9). In this report, we have extended our studies of the active site by investigating the function of a highly invariant lysine residue in the third most conserved region (Fig. 1).
Previous study has documented that the interaction of DNA polymerase ␣ with its substrates obeys a rigidly ordered sequential terreactant mechanism, with template as the first substrate, followed by primer as the second substrate and . b k cat O is the catalytic rate when dCTP, dATP, dTTP and dGTP were substrates in the DNA synthesis reaction under steady-state conditions. c k cat S is the catalytic rate when dCTP␣S, dATP, dTTP, and dGTP were substrates in the DNA synthesis reaction under steady-state conditions. dNTP as the third. Specification of which of the four dNTPs has kinetically significant binding is determined by the base sequence of the template (21). A similar ordered sequential terreactant mechanism was also proposed by Dahlberg and Benkovic (22) for the Klenow fragment of E. coli polymerase I. Given the universal ordered sequential mechanism for both eukaryotic DNA polymerase ␣ and prokaryotic E. coli polymerase I, and depending on the base sequence of the template, DNA polymerases have different modes of interaction with the incoming dNTP. Studies of Klenow fragment have shown a rate difference in using dTTP versus dGTP (23). In this study, we did not compare the difference of either the wild type or the mutant enzymes for their affinity (K m(dNTP) ) in binding each different incoming dNTP or the difference in binding purine versus pyrimidine deoxyribose triphosphate. It is possible that the side chain of this highly invariant Lys of ␣-like DNA polymerases like that of E. coli polymerase I has a different mode of interaction with each different incoming dNTP (23). Here, we assume in the enzyme where the protein contacts the dNTP, the Lys 950 side chain has the same interaction with all four of the dNTPs in all circumstance. We also only evaluated the side chain function of Lys 950 by kinetic analysis in the state of catalytically competent ternary complex at the point of phosphodiester bond formation and pyrophosphate release.

Mutations of Lys 950 Side Chain Affect dNTP Affinity and Metal
Activator Utilization-In this study we have observed the following. (i) Mutation of Lys 950 to Arg, Ala, and Asn has an effect on the mutant enzyme's K m for dNTP in reactions with either Mg 2ϩ or Mn 2ϩ as metal activator, but has only a moderate effect on their k cat . This implies that the side chain of Lys 950 is mainly involved in interacting with the dNTP substrate and not in catalysis. In addition, inhibition studies with an active site inhibitor and analogs have further verified that the side chain of Lys 950 has a role in the active site interacting with a dNTP substrate. (ii) Mutations of this highly conserved lysine residue allows the mutant enzymes to utilize Mn 2ϩ as metal activator more efficiently than the wild type enzyme (Fig. 2) as observed in mutations of Asp 1002 and Thr 1003 in region I of human DNA polymerase ␣ (9). However, a noteworthy difference between the Lys 950 mutant enzymes and the Asp 1002 and Thr 1003 mutant enzymes is that mutations of Asp 1002 and Thr 1003 have a profound effect on the enzyme's catalysis k cat , whereas mutations of Lys 950 mainly affect the mutant enzymes' K m(dNTP) values and have only a mild effect on their catalysis (k cat ) ( Table I).
Metal Activator Effect-Despite the fact that mutations of Lys 950 affect the K m(dNTP) , none of the Lys 950 mutant derivatives displayed significant differences in their misinsertion efficiency as compared to wild type enzyme when Mg 2ϩ was used as the metal activator (Table II). This suggests that in a polymerase ␣-Mg 2ϩ complex, the side chain of Lys 950 does not interact with the nucleotide base directly and is not responsible for pairing the incoming dNTP to the appropriate nucleotide base in the template. In reactions with Mn 2ϩ , all three Lys 950 mutants showed an improved misinsertion fidelity (Table II) like that observed in mutants D1002N and T1003S of human polymerase ␣ region I (8).
Studies of E. coli polymerase I have proposed that there is an indirect interaction between metal activator and the deoxyribose of dNTP (24). It has been proposed that E. coli polymerase I-Mg 2ϩ complex selectively prefers the C2Ј-endo conformation of the deoxyriboside of dNTPs, while the polymerase I-Mn 2ϩ complex is less selective for this conformation. Thus, in the Mn 2ϩ -catalyzed reaction, the deoxyribose freely equilibrates between the C2Ј-and C3Ј-endo conformations of the deoxyriboside. It is possible that the polymerase ␣-Mn 2ϩ complex like that of E. coli polymerase I-Mn 2ϩ complex, is less selective for C2Ј-endo conformation of deoxyriboside and allows the sugar ring to freely equilibrate between C2Ј-and C3Ј-endo conformations. This might enhance the stringency for the polymerase ␣-Mn 2ϩ complex in its specification for a correct dNTP over an incorrect dNTP resulting in improved misinsertion fidelity.
Based on structural data together with mutagenesis studies of Klenow fragment, Joyce and Steitz and co-workers (12,25) have proposed a possible mechanism for the polymerase reaction in which two metal ions are involved in mediating catalysis. In the active site of a DNA polymerase, several carboxylate side chains, such as the Asp 705 and Asp 882 of the Klenow and Asp 1002 and Asp 1004 of human polymerase ␣, function to anchor two divalent metal ions (Mg 2ϩ ) for catalysis. One Mg 2ϩ promotes the deprotonation of the 3Ј-hydroxyl of the primer, while the second Mg 2ϩ facilitates the formation of the pentacovalent transition state at the ␣-phosphate of the dNTP and the loss of pyrophosphate. Since mutations of Lys 950 affect both the metal activator utilization and affinity for dNTP, the side chain of Lys 950 therefore might interact either directly or indirectly with the second metal ion chelated ␣-phosphate of the incoming dNTP.
Lys 950 Side Chain Interacts with Oxygen Moiety of the ␣-Phosphate of the dNTP Substrate-The abilities of the three Lys 950 mutant enzymes to utilize Mn 2ϩ as metal activator more efficiently than the wild type enzyme suggest that the positively charged lysine side chain may have an influence on the configuration of the negatively charged phosphate groups of the incoming dNTP. Comparison of the inhibitory effect of pyrophosphate and the inhibitory effect of BuPdGTP versus BuPdGMPCH 2 PP on the wild type enzyme and the three mutant enzymes have shown that the ␤and ␥-phosphates are not likely to be in direct contact with the lysine side chain (Tables  III and IV). Analogs of dNTP containing alterations in the ribose moiety had notable effects on the reactivity of the three Lys 950 mutant enzymes as compared to the wild type enzyme (Table IV). Mutant enzyme K950A which has the positive charge side chain abolished always displays higher sensitivity or resistance to the dNTP analogs regardless of whether the analog has deletion of the 3Ј-OH group as in ddCTP or has a twisted deoxyribose ring as in araCTP (26).
Since alteration of the deoxyribose in either araCTP or ddCTP could also affect the orientation of the oxygen group of ␣-phosphate, we also tested the interactions between the Lys 950 side chain and the oxygen group of ␣-phosphate with an analog, dCTP␣S, in reactions utilizing either Mg 2ϩ or Mn 2ϩ as metal activator. We compared each enzyme's affinities (K m ) and catalysis (k cat ) for utilizing dCTP␣S versus dNTP as substrate in reactions with either Mg 2ϩ or Mn 2ϩ (Table V). We also compared each enzyme's affinity ratio for dCTP␣S versus dNTP (K m S /K m O ) with the wild type enzyme. In reactions with Mg 2ϩ , mutation of the Lys 950 side chain by either replacing it with a larger charged side chain (K950R) or abolishing the charged side chain (K950A) had a significant effect on the mutant enzyme's affinity to dCTP␣S versus dNTP substrate. In contrast, replacing the ⑀-amino side chain of Lys 950 to Asn (K950N) appears not to affect the mutant's affinity ratio for dCTP␣S versus dNTP. In Mn 2ϩ -catalyzed reactions, all three mutants showed comparable affinity ratio, K m S /K m O , as the wild type enzyme. We reason that this difference in metal effect might be due to the polymerase ␣-Mn 2ϩ complex having less selective preference for C2Ј-endo deoxyribose. The presence of the C3Јendo form of dNTP could affect the affinity between the dNTP and the side chain of Lys 950 resulting in an observed mild effect on the affinity of the mutant enzymes for dCTP␣S.
In sum, results presented in this study strongly suggest that the structural feature of the incoming dNTP recognized by the positively charged Lys 950 side chain is the oxygen moiety of the ␣-phosphate group.
A Proposed Model of How Active Site Residues Collaborate for Polymerase Catalysis-There is no structural data for any member of the ␣-like DNA polymerases (family B). However, in light of the primary sequence conservation of these three highly invariant regions among all of the ␣-like DNA polymerases and the similarities between the predicted secondary structure of these three regions in ␣-like DNA polymerases with the crystal structures of Klenow fragment and HIV-1 reverse transcriptase (10,11), it is reasonable to assume that these three highly conserved regions in the ␣-like polymerases are components of the active site. The Lys 950 described in this report is located in a predicted ␣-helix. This helix might be positioned in the active site like the O-helix of Klenow fragment (10). In the active site of Klenow fragment, binding of the incoming dNTP is not identical under all circumstances for each dNTP (23). Studies of site-directed mutations of a residue Arg 754 in the O-helix of Klenow fragment suggested that Arg 754 contacts the ␤or ␥-phosphate of the incoming dNTP substrate. When the incoming dNTP is dGTP, in addition to Arg 754 , two residues in the O-helix, Lys 758 and Phe 762 , also participate in the binding (23). Without a physical structure of the ␣-like polymerases, we cannot assume that Lys 950 of polymerase ␣ is equivalent to the Lys 758 of the Klenow fragment. However, it is interesting that both family A and family B polymerases have a highly conserved lysine residue located in an ␣-helix, and both appear to be involved in interacting with the incoming dNTP substrate.
Our mutational studies (4,5,8,9) have identified the functions of side chains of several amino acid residues localized in the active site of the ␣-like polymerases. The results have supported a model of the active site of ␣-like polymerases. In human polymerase ␣, Asp 1002 and Asp 1004 , and Thr 1003 located in an anti-parallel ␤-sheet (region I) chelate with the metal activator cation, Mg 2ϩ , which in turn chelates to the oxygen moiety of ␤and ␥-phosphate of the incoming dNTP (8,9). The phenyl ring side chain of Tyr 865 (in region II) interacts with the nucleotide base of the incoming dNTP to properly position the incoming dNTP for Watson-Crick base pairing (5). The oxygen moiety of the Ser 867 hydroxyl side chain in region II forms a hydrogen bond either directly or indirectly with the 3Ј-OH terminus of the primer. The hydrogen bond formation might enhance the oxygen moiety at the 3Ј-OH-primer terminus for nucleophilic attack at the ␣-phosphate of the incoming dNTP (4). The positively charged side chain of Lys 950 located in an ␣-helix in region III of human polymerase ␣ interacts either directly or indirectly with the oxygen group of the ␣-phosphate of the incoming dNTP. This interaction of a positively charged side chain will neutralize the negative charge on the ␣-phosphate to facilitate nucleophilic attack of the incoming primer 3Ј-hydroxy group. This proposed model of the ␣-like polymerase active site (Fig. 3) is based entirely on biochemical data and can only be verified in the future by crystallographic data of a ternary complex of polymerase ␣ with primer-template and dNTP substrates. Shown are the proposed functions of residues of human DNA polymerase ␣ studied by site-directed mutagenesis. The three region I residues, Asp 1002 , Thr 1003 , and Asp 1004 , shown here bound to the metal-nucleotide complex are adopted from Ref. 9. The phenyl ring of Tyr 865 in region II that interacts with the nucleotide base moiety of the incoming dNTP and the hydroxyl side chain of the Ser 867 residue that hydrogen bonds to the 3Ј-OH terminus of the primer shown here are adopted from Ref. 4. The positive charged side chain of Lys 950 of region III is shown here to interact with the oxygen group of the ␣-phosphate of the incoming dNTP. An X is shown here to depict the unknown side chain(s) of residue(s) in the active site that might participate in chelating the Mg 2ϩ metal activator.