Multiple Amino Acid Substitutions Allow DNA Polymerases to Synthesize RNA*

DNA and RNA polymerase exhibit similarities in structures and catalytic mechanisms, suggesting that both classes of enzymes are evolutionarily related. To probe the biochemical and structure-function relation-ship between the two classes of polymerases, a large library (200,000 members) of mutant Thermus aquaticus DNA polymerase I ( Taq pol I) was created containing random substitutions within a portion of the dNTP binding site (motif A; amino acids 605–617), and a fraction of all selected active Taq pol I (291 of 8000) was tested for the ability to incorporate successive ribonucleotides; 23 unique mutants that added rNTPs into a growing polynucleotide chain were identified and sequenced. These mutants, each containing one to four substitutions, incorporate ribonucleotides at a efficiency approaching 10 3 -fold greater than that of wild type Taq pol I. Several mutants added successive ribonucleotides and thus can catalyze the synthesis of RNA. Sequence analysis of these mutants demonstrates that at least two amino acid residues are involved in excluding ribonucleotides from the active site. Interestingly, wild type DNA polymerases from several distinct families selectively discriminate against rUTP. This study suggests that current DNA and RNA polymerases could have evolved by divergent evolution from an ancestor that shared a common mechanism for polynucleotide synthesis. Both DNA and RNA polymerase can catalyze chain elongation reaction guided by single-stranded DNA templates to gen-erate polynucleotide products (1). The order of nucleotide addition proceeds in a 5 9 3 3 9 direction via metal-mediated were (50 m Refs. 14 and 15 that allows efficient ( . 50%) purification of Taq pol I while removing endogenous polymerase and nuclease activities. Biochemical analysis of DNA-dependent DNA polymerase activity of these 350 separate pols showed that 291 pols are active at elevated temper-atures ( . 10% activity relative to WT enzyme). Each of the selected Taq pols that retain at least 10% activity relative to WT enzyme at 72 °C (291 total) were tested for the ability to incorporate ribonucleotides. Primer-template constructs were prepared by hybridizing 5 9 - 32 P end-labeled 23-mer primer (5 9 -cgc gcc gaa ccg cta gca at) with 46-mer template (5 -gcg using a 1:2 primer:template ratio (16). The primer/template (5 n M ) was incubated in a reaction mixture containing 50 m M KCl, 10 m M Tris-HCl (pH 8), 0.1% Triton X-100, 2.5 m M MgCl 2 , and 1 m l of partially purified Taq pols (0.1 to 0.01 units) in 10- m l volumes in the presence of 0–250 m M each rNTP. Reactions were terminated after 30 min of incubation at 55 °C with the addition of 2 m l of formamide containing stop (Amersham Pharmacia Biotech). Products were analyzed by 14% dena- turing PAGE (16). Incorporation of ribonucleotides relative to deoxyribonucleotides onto nascent primers results in a slower migration on PAGE gels; only mutants that produced products with such migration patterns were considered rNTP incorporating polymerases. The RNA polymerase activity for several of these enzymes was confirmed by assaying for activity in a 20- m l reaction mixture containing 50 m M KCl, 10 m M Tris-HCl

DNA and RNA polymerase exhibit similarities in structures and catalytic mechanisms, suggesting that both classes of enzymes are evolutionarily related. To probe the biochemical and structure-function relationship between the two classes of polymerases, a large library (200,000 members) of mutant Thermus aquaticus DNA polymerase I (Taq pol I) was created containing random substitutions within a portion of the dNTP binding site (motif A; amino acids 605-617), and a fraction of all selected active Taq pol I (291 of 8000) was tested for the ability to incorporate successive ribonucleotides; 23 unique mutants that added rNTPs into a growing polynucleotide chain were identified and sequenced. These mutants, each containing one to four substitutions, incorporate ribonucleotides at a efficiency approaching 10 3 -fold greater than that of wild type Taq pol I. Several mutants added successive ribonucleotides and thus can catalyze the synthesis of RNA. Sequence analysis of these mutants demonstrates that at least two amino acid residues are involved in excluding ribonucleotides from the active site. Interestingly, wild type DNA polymerases from several distinct families selectively discriminate against rUTP. This study suggests that current DNA and RNA polymerases could have evolved by divergent evolution from an ancestor that shared a common mechanism for polynucleotide synthesis.
Both DNA and RNA polymerase can catalyze chain elongation reaction guided by single-stranded DNA templates to generate polynucleotide products (1). The order of nucleotide addition proceeds in a 5Ј 3 3Ј direction via metal-mediated phosphoryl transfer reaction resulting in the formation of phosphodiester bond and release of pyrophosphate (2). In addition, both DNA and RNA polymerases resemble in morphology a cupped human right hand and bind DNA template and the incoming nucleotide within the active site cleft (3,4). DNA polymerases differ from RNA polymerases in utilizing 2Ј-deoxynucleotides (dTTP, dCTP, dGTP, and dATP) rather than ribonucleotides (rUTP, rCTP, rGTP, and rATP). A detailed analysis of the polymerase active site is crucial to understanding how these polymerases distinguish between dNTPs and rNTPs, as well as to provide insights on how the two types of polymerases co-evolved to adopt similar mechanisms.
Despite the similarity in protein structure and function, there is almost no sequence identity between DNA and RNA polymerases. For example, nearly all of the over 40 prokaryotic and eubacteria DNA pol Is 1 sequenced (including Thermus aquaticus pol I, Chlamydia trachomatis pol I, and Escherichia coli pol I) contain the DYSQIELR sequence within the dNTP binding site (motif A; Ref. 5), yet RNA pols only have in common the catalytically essential aspartic acid residue (6). If evolution proceeded from an "RNA world" containing RNA synthesizing enzymes to a "DNA world" with genomes replicated by DNA synthesizing enzymes (7,8), it might be possible to gain insights into this process by substituting random sequences within the active site of a polymerase.
Analyses of high resolution x-ray crystal structures suggest that a single side chain prevents diverse polymerases from incorporating ribonucleotides (9 -11). Rational approaches involving site-directed mutagenesis of this residue result in polymerases that facilitate ribonucleotide incorporation, but substitutions at this residue could not produce polymerases that efficiently incorporate successive ribonucleotides (12,13). A disadvantage of structure-guided site-directed studies is that, by substituting a catalytically important residue to a neutral amino acid, the overall enzyme activity can be impaired. In this report, we offer an alternative to structure based site-directed mutagenesis approach for creating enzymes with unique properties. Assuming that current RNA and DNA polymerases evolved from a common ancestor and share a basic mechanism, it should be possible to evolve one of these enzymes into the other, following extensive rounds of mutagenesis and stringent selection protocols. To explore this concept, we randomly mutated a portion of the active site of a eubacteria DNA polymerase, Taq pol I, selected for functioning mutants, and tested 291 mutant enzymes for the ability to synthesize RNA. Twentythree different mutant polymerases containing substitutions in predominantly one of two amino acids were identified that incorporated ribonucleotides at a rate approaching 10 3 -fold greater than that of WT Taq DNA polymerase. We show that the active site within several families of WT DNA polymerases is especially evolved to exclude rUTP.

EXPERIMENTAL PROCEDURES
Screen for Ribonucleotide Incorporation Activity-The construction of the random Taq pol I library (containing 200,000 individual clones) and genetic selection protocol that yields functional mutants in vivo has been described in detail in a previous publication (5). The 350 colonies from the 8000 that complemented pol I temperature sensitive phenotype at 37°C were isolated and grown in nutrient broth individually overnight at 30°C. Each culture was grown to A 595 of 0.3 at 30°C in 10 ml, and Taq pol I expression was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside and incubations continued for 4 h. Taq pols were partially purified (50 l total volume) using a modified protocol from Refs. 14 and 15 that allows efficient (Ͼ50%) purification of Taq pol I while removing endogenous polymerase and nuclease activities. Biochemical analysis of DNA-dependent DNA polymerase activity of these 350 separate pols showed that 291 pols are active at elevated temperatures (Ͼ10% activity relative to WT enzyme). Each of the selected Taq pols that retain at least 10% activity relative to WT enzyme at 72°C (291 total) were tested for the ability to incorporate ribonucleotides. Primer-template constructs were prepared by hybridizing 5Ј-32 P endlabeled 23-mer primer (5Ј-cgc gcc gaa ttc ccg cta gca at) with 46-mer template (5Ј-gcg cgg aag ctt ggc tgc aga ata ttg cta gcg gga att cgg cgc g) using a 1:2 primer:template ratio (16). The primer/template (5 nM) was incubated in a reaction mixture containing 50 mM KCl, 10 mM Tris-HCl (pH 8), 0.1% Triton X-100, 2.5 mM MgCl 2 , and 1 l of partially purified Taq pols (0.1 to 0.01 units) in 10-l volumes in the presence of 0 -250 M each rNTP. Reactions were terminated after 30 min of incubation at 55°C with the addition of 2 l of formamide containing stop solution (Amersham Pharmacia Biotech). Products were analyzed by 14% denaturing PAGE (16). Incorporation of ribonucleotides relative to deoxyribonucleotides onto nascent primers results in a slower migration on PAGE gels; only mutants that produced products with such migration patterns were considered rNTP incorporating polymerases. The RNA polymerase activity for several of these enzymes was confirmed by assaying for activity in a 20-l reaction mixture containing 50 mM KCl, 10 mM Tris-HCl (pH 8), 0.1% Triton-X, 2.5 mM MgCl 2 , 0.4 mg of activated calf thymus DNA, 10 M each dCTP, dATP, and dTTP, 500 M rGTP, 0.25 mCi of [␣-32 P]dTTP, and 1 l of WT or mutant Taq pols (0.1 units). Incubations were at 72°C for 5 min, and reactions were stopped with the addition of 100 l 0.1 M sodium pyrophosphate, followed by 0.5 ml of 10% trichloroacetic acid. Polymerase activity was quantified by collecting precipitated radioactive DNA onto glass filter papers, and the radioactive counts were measured by scintillation.
Kinetic Analysis-A 47-mer template (3Ј-gcg cgg ctt aag ggc gat cgt tat agc tta agg cct tta aag ggc cc-5Ј) was hybridized with one of four primers: 23-mer (5Ј-cgc gcc gaa ttc ccg cta gca at), 24-mer (5Ј-cgc gcc gaa ttc ccg cta gca ata), 25-mer (5Ј-cgc gcc gaa ttc ccg cta gca ata t), or 26-mer (5Ј-cgc gcc gaa ttc ccg cta gca ata tc). All single nucleotide incorporation with dATP or rATP required 23-mer/47-mer primer/template; reactions with dTTP, dUTP, or rUTP required 24-mer/47-mer; reactions with dCTP or rCTP required 25-mer/47-mer; those with dGTP or rGTP required 26-mer/47-mer. The steady state Michaelis-Menten parameters V max and K m were determined following incubations with limiting amounts of Taq pol in the presence of 5 nM primer/template and varying concentration of each dNTP or rNTP for 10 min at 55°C for reactions containing Taq pol I or at 37°C for reactions containing other polymerases as described in Ref. 18. The kinetic rate parameter k cat was calculated by dividing V max by the enzyme concentration. All products were analyzed by 14% PAGE and quantified by phosphorimager analysis.

Barriers for WT Taq pol I to Incorporate
Ribonucleotides-We measured the capacity of several subclasses of DNA polymerases to incorporate ribonucleotide triphosphates relative to dNTPs. DNA polymerases from the thermophile eubacterium T. aquaticus, from hyperthermophile archaea Thermococcus litoralis, from prokaryote E. coli, and from retrovirus MMLV were analyzed. These DNA polymerases represent three major subclasses (6): prokaryotic pol I family (Taq pol I and E. coli pol I), mammalian pol ␣ family (Vent pol), and RT (MMLV RT). Incorporation of individual ribonucleotides by WT Taq pol I requires greater amounts of polymerase and at least 1000-fold higher concentrations of each rNTP relative to the corresponding dNTP (Fig. 1). It is important to note that polynucleotide products containing a 3Ј ribonucleotide migrate slower within polyacrylamide gels relative to 3Ј deoxynucleotide products. This property is useful to monitor the nature of products, as well as to ensure the purity of dNTP and rNTP substrates. Hyperbolic curve fit of steady state rates plotted as a function of nucleotide concentration yields Michaelis-Menten parameters V max and K m , and the initial slope of this plot reflects the catalytic efficiency of each enzyme. WT Taq pol I incorporates dGTP, dATP, and dCTP up to 30,000 times more efficiently (k cat /K m ) than the respective ribonucleotides (Tables  I and III). The low efficiency of rNTP incorporation is largely due to a 1000-fold greater K m for rG, rC, and rA relative to the respective deoxynucleotides. Other DNA polymerases from pol I, pol ␣, and RT families also incorporate ribonucleotides inefficiently (also largely because of high K m for rNTPs relative to dNTPs; Table I). DNA polymerases are especially adept at incorporating dTTP over rUTP. Although Taq pol I incorporates dTTP and dUTP opposite template dA residue with approximately equal efficiencies, the WT enzyme incorporates dTTP 10 6 -fold more efficiently than rUTP (Table I). Selection against rUTP by Taq pol I is 100-fold more stringent than the other three ribonucleotides. DNA polymerases from pol I, pol ␣, and RT classes also exhibit a high propensity to exclude ribouracil. For example, Klenow pol (3Ј-5Ј exoϪ) incorporates rCTP and rUTP 3100 and 130,000 times less efficiently relative to dCTP and dTTP, respectively, and ribonucleotide incorporation by Klenow pol (exoϩ) is less efficient, indicating the 3Ј-5Ј proofreading domain excises ribonucleotides off the nascent primer. Efficiency of rATP incorporation opposite template dT is comparable with that of rCTP incorporation opposite template dG for Taq pol, Klenow pol, Vent pol, and MMLV RT (data not shown). Astatke et al. (12) reported that Klenow (exoϪ) discriminates against rCTP and rUTP by similar factors and that discrimination against rNTPs lies exclusively at the level of k cat , where as we find the differences between rNTPs and dNTPs are at the K m level. In assays used by Astatke et al. (12), k cat refers to the rate of the conformation change and/or nucleotide incorporation; k cat in our steady state experiments corresponds to the rate of enzyme turnover. The nucleotide incorporation efficiency val-ues (k cat /K m in steady state experiments and k cat /K d in single turnover assays) in the two studies are consistent. These data indicate that several families of DNA polymerases have evolved a sophisticated mechanism to exclude ribonucleotides, especially uracil, from its catalytic site.
Testing a Large Library of Taq pol I Mutants for rNTP Incorporation-Assuming that DNA and RNA polymerases are related mechanistically and in structure-function, it should be  a Discrimination represents the efficiency (V max /K m ) of dNTP incorporation divided by efficiency of the corresponding rNTP. For dUTP, discrimination represents efficiency of dTTP incorporation divided by that of dUTP. V max (rel) represents steady state rates relative to that of dTTP or dCTP incorporation.
b Primer/template sequence used with each nucleotide is defined under "Experimental Procedures." possible to distinguish residues responsible for dNTP versus rNTP substrate specificity following extensive mutagenesis of the polymerase active site and genetic selection of active mutants. We previously created a Taq pol I library composed of 8,000 active members containing substitutions within active site (motif A) residues. To determine whether this plasticity within the catalytic site architecture can confer altered substrate specificity, we further analyzed all 291 selected Taq pol I, including 27 with wild type amino acid sequence, for the ability to incorporate ribonucleotides. Briefly, this analysis involved incubation of 0.01-0.1 unit of Taq pol I with 32 P-labeled oligonucleotide primer-template in the presence of all four rN-TPs. Products were separated by polyacrylamide gel electrophoresis and analyzed by autoradiography. Incorporation of ribonucleotides relative to deoxyribonucleotides onto nascent primers results in a slower migration on PAGE gels; only mutants that produced products with such unique migration patterns were considered rNTP incorporating polymerases. This screen identified that a small subset of the selected enzymes (23 of 291) can incorporate rNTPs efficiently (Table II). None of the 27 WT Taq pols containing silent mutations could efficiently catalyze ribonucleotide incorporation above the sensitivity of this assay. The 23 ribonucleotide incorporating mutants contain amino acid substitutions at distinct loci and can be divided into two major classes. 1) Those encoding a hydrophilic substitution at Ile-614; these enzymes constitute the majority of rNTP incorporating mutants containing from 0 to 1 additional substitutions (suggesting that alterations at this site do not require secondary compensatory amino acid substitutions to maintain physiologic activity). 2) Those that encode a E615D substitution; these enzymes contain 1-3 other substitutions (suggesting that substitutions at this site require compensatory mutations to maintain physiologic activity). Mutants (mutants 53, 65, 75, 94, 230, 265, 300, and 346) were purified to homogeneity and tested for the ability to incorporate rGTP into a biologically relevant gapped calf thymus DNA substrate. Each of these polymerases conducts efficient synthesis in the presence of rGTP (as judged by incorporation of [␣-32 P]dTTP, whereas mutant polymerases containing other types of substi-tutions and WT Taq pol I controls produced significantly lower product yields. These data indicate that amino acid residues Ile-614 and Glu-615 within the evolutionarily conserved motif A are the primary amino acids involved in discriminating against ribonucleotide incorporation. Efficient rNTP Incorporation by Mutant Taq pol Is-To characterize the efficiency of dNTP relative to rNTP incorporation by WT and mutant enzymes, the steady state k cat and K m values of nucleotide incorporation were determined. The plots of nucleotide incorporation rates as a function of nucleotide concentration exhibited typical Michaelis-Menten saturation kinetics for all reactions containing either dNTPs or rNTPs with either mutant or WT enzymes. The k cat /K m value obtained from fitting a hyperbolic curve directly measures the efficiency of nucleotide incorporation. Mutant homogenous enzymes containing specific substitutions at position 614 and 615 incorporate rGTP, rATP, and rCTP at an efficiency (k cat /K m ) approaching 1 ⁄10 that of the corresponding dNTP (Table III; dNTP/rNTP discrimination). The steady state rate (k cat ) for each dNTP incorporation is similar to that of rNTP incorporation for these mutants, thus indicating that rate-limiting step (i.e. enzyme turnover) is comparable for the mutant enzymes during either dNTP or rNTP incorporation reactions. In contrast, the WT enzyme incorporates each of the rNTPs 10 -100 times slower relative to the respective dNTPs. In addition, whereas the mutant enzymes exhibit 10 -100-fold higher K m for rGTP, rATP, and rCTP incorporation relative to the respective dNTPs, the WT enzyme exhibits Ͼ1000-fold higher K m for these rNTPs relative to the respective dNTPs, and the WT enzyme exhibits Ϸ50,000-fold higher K m for rUTP relative to dTTP or dUTP. Overall, the mutants incorporate each ribonucleotide up to 3 orders of magnitude more efficiently than the WT polymerase (Table III; dNTP/rNTP Discrimination). Mechanistically, the ability of the mutant polymerases to efficiently incorporate ribonucleotide results from a Ͼ10-fold faster incorporation rate relative to WT enzymes and the ability to incorporate rNTPs at concentrations that are significantly below physiologic levels. The relatively fast catalytic rate of rNTP incorporation, coupled with the K m for rNTP significantly lower a Underlining denotes mutants exhibit WT activity. b These two mutants differ in nucleotide sequence.

Evolution of the Polymerase Active Site
than the in vivo ribonucleotide concentration (approximately 1000 M each), suggest these enzymes should be able to incorporate ribonucleotides at physiologic conditions. To determine whether these polymerases can function as RNA polymerases by incorporating multiple ribonucleotides sequentially, we incubated purified WT Taq pol I, a mutant containing a substitution at Ile-614, and a mutant containing a substitution at Glu-615 in the presence of increasing amounts of all four rNTPs (Fig. 2). Although the WT enzyme inefficiently incorporates and extends ribonucleotides, both rNTP utilizing mutant enzymes polymerize multiple ribonucleotides, even at rNTP concentrations below that found in cells. Interestingly, the strong pause sites produced at runs of template dAs is exactly what one would predict from the kinetic data (Tables I  and III), demonstrating a specific decrease in incorporation (and perhaps a decreased extension of 3Ј-rU sequences). Extension past these runs is facilitated by either increasing incubation time or increasing ribonucleotide concentration. In the presence of a trace amount of a metal cofactor Mn ϩ2 (0.5 mM Mn ϩ2 and 2.5 mM Mg ϩ2 ), this pause site is markedly reduced, and elongation proceeds up to the 5Ј end of the template even in presence of low rNTP concentrations. Elongated RNA products are also efficiently generated by mutant polymerases containing substitutions at either Ile-614 or Glu615 using templates containing minimal template dA residues. Using such a template, RNA polymerization to the 5Ј template terminus by specific mutant polymerases occurs in a time-dependent manner and within minutes of incubation (Fig. 3). RNA synthesis is inefficient with the WT Taq pol I even after prolonged incubation times, and the limited products formed by WT enzyme are resistant to alkali degradation. In contrast, addition of alkali degraded elongated products synthesized by mutant enzymes to regenerate the initial substrate, illustrating the products are RNA (Fig. 2, lane 32 C, and Fig. 3, lanes OH). From the time-dependent accumulation of products, we estimate that the mutant Taq pols can synthesize extended RNA products (containing Ͼ10 nucleotides) at a Ͼ1000-fold faster rate relative to WT Taq pol I. The ability to synthesize long RNA products correlates with the DNA polymerase activity of the mutant enzyme. For example mutant 94 (A608S,I614N) with low DNA polymerase activity as measured on activated calf thymus DNA does not efficiently synthesize long stretches of RNA. Mutant Taq pol Is exhibiting high DNA polymerase activity (mutants 75 (I614M), 265 (I614N,L616I), and 346 (A608D,E615D)) can synthesize nucleic acids containing as many as 100 successive ribonucleotides on templates with limited dA residues (data not shown). To our knowledge, this is the first report of a highly active polymerase that can efficiently synthesize long stretches of either DNA or RNA. DISCUSSION There are two major conclusions from this study. First, DNA polymerases selectively exclude rUTP. Second, amino acid substitutions at two different positions within the DNA polymerase catalytic site can result in enzymes that incorporate ribonucleotides and synthesize RNA. These findings were not predicted by analysis of high resolution x-ray crystal structures. These observations link DNA and RNA polymerization mechanisms and may provide insights into the evolution of DNA polymerases during the postulated progression form an RNA into DNA world (7,8).
Taq pol I incorporates rUTP 1,000,000-fold less efficiently relative to dTTP or dUTP (Tables I and III). Joyce and coworkers (12) also showed that WT E. coli pol I discriminates against rUTP relative to dTTP by 10 6 -fold. We extended this finding by comparing dNTP relative to rNTP incorporation by DNA polymerases from different families. Our results indicate that ribouracil triphosphate exclusion is a general property of many DNA polymerases. With WT Taq pol I, the 10 6 -fold difference in nucleotide incorporation efficiency between dTTP or dUTP relative to rUTP incorporation is the highest discrimination factor that we have encountered with natural substrates. Although the concentration of each deoxynucleotide including dTTP is 25-100 M in organisms, that of each ribonucleotides, including rUTP, is approximately 1000 M (19). We estimate that if WT Taq pol I were to replicate the entire 2 ϫ 10 6 base pair eubacteria genome (20), assuming the rUTP concentration is 10-fold greater than that of dTTP, WT Taq pol I should introduce approximately 10 rUMP into genomic DNA per replication cycle, and these substitutions would presumably be corrected by DNA repair. Other polymerases from prokaryotic pol I, eukaryotic pol ␣, and RT families also discriminate against ribouracil (Table I). This DNA polymerase property of specifically discriminating against ribouracil might have resulted from natural selection. Cells have evolved mechanisms to exclude uracil from DNA. When formed by deamination of cytadines (21), uracil is a potent promutagen inducing G⅐C3 A⅐T transition mutations (22). Uracil lesions are also formed by the incorporation of dUTP by DNA polymerases because dUTP can readily substitute for dTTP (Table I and Ref. 23). Removal of uracil residues in DNA is accomplished by uracil-DNA glycosylase, and repair synthesis involves the generation of single-stranded breaks, a potential source of mutagenesis (24,25). In cells, rUTP is present in millimolar concentrations, well in excess of dTTP and dUTP concentrations (on average, 40 and 0.2 M, respectively; Refs. 19 and 23) and thus can be a potent source of damage (e.g. single-stranded breaks) if introduced into DNA in amounts that exceed the capacity of the cell for DNA repair.
Analysis of a high resolution crystal structure of Taq pol I bound to DNA and dNTP (9) suggested that a single residue can sterically exclude the 2Ј-OH of an incoming rNTP (Fig. 4). This glutamic acid residue is conserved in over 35 prokaryotic and eubacteria pol I sequences, including that of E. coli pol I (GenBank TM ; Ref. 5). Polymerases in the pol ␣ and RT families contain a planar ringed amino acid (usually Tyr or Phe) at this locus; within the HIV-1 RT structure bound with DNA and dNTP the homologous Tyr-115 residue is positioned adjacent to the incoming nucleotide and likely also prevents incorporation of ribonucleotides by sterically interfering with 2Ј-OH group (11,26). Modeling the selected substitutions into WT Taq pol I structure, followed by energy minimization suggests there are at least two mechanisms by which the steric interference conferred by Taq pol I Glu-615 with the 2Ј-OH group of the incoming ribonucleotide can be alleviated. First, modeling studies in Taq pol I suggests alterations that reduce the length of the Glu-615 side chain should permit ribonucleotide binding ( Fig. 4 and Table III). Joyce and co-workers (12) demonstrated that E710A substitution within E. coli pol I (Klenow), at a site identical to Taq pol I Glu-615 residue, permits ribonucleotide incorporation. We find substitutions to aspartic acid for Taq pol I Glu-615 facilitates ribonucleotide incorporation. So far, we have not yet detected a Glu 3 Ala substitution in the Taq pol I motif A active mutant library, as well as within an extensive motif A mutant library of E. coli pol I, presumably because of reduced catalytic activity of mutants not containing either Asp or Glu at this site (5). 2 Second, we find that diverse substitutions at the adjacent residue Ile-614 allow ribonucleotide incorporation and that evolved enzymes containing substitutions at position 614 utilize rNTPs at efficiencies equivalent to enzymes containing E615D substitutions ( Fig. 3 and Table III). Analysis of Taq pol I structure model bound with DNA and a rNTP shows the ribose ring of the rNTP packs closely against Ile-614. Residue Ile-614 is located at a junction of a highly conserved ␤ strand and ␣ helix and is highly mutable. Energy minimizations of models of substitutions at this residue, which confers the ability to incorporate rNTPs, suggest that these substitutions cause this junction to be located further from the incoming nucleotide, thus allowing 2Ј-OH to fit.
In summary, we find that motif A within the DNA polymerase active site is highly plastic and can tolerate numerous substitutions while preserving physiologic DNA polymerase activity, and we have used this flexibility in structure to evolve a set of enzymes with altered substrate specificity. Mutant polymerases that can synthesize both DNA and RNA may be useful for biotechnology and allow automated coupled polymerase chain reaction amplification and transcription by cycling nucleotide triphosphates within the reaction mixture. These polymerases could also be used for DNA sequence analysis, following specific incorporation of a ribonucleotide during polymerase chain reaction and subsequent alkali cleavage at the position of ribosubstitution (27). In addition, these polymerases allow one to introduce nucleotide analogs containing adducts (e.g. fluorophores) attached to the 2Ј ribose ring. The plasticity of the DNA polymerase active site should facilitate evolution within the laboratory of other polymerases for the incorporation of specific nucleotide analogs that are of mechanistic or medical importance.