The CCA-adding Enzyme Has a Single Active Site*

The CCA-adding enzyme (tRNA nucleotidyltransferase) synthesizes and repairs the 3′-terminal CCA sequence of tRNA. The eubacterial, eukaryotic, and archaeal CCA-adding enzymes all share a single active-site signature motif, which identifies these enzymes as belonging to the nucleotidyltransferase superfamily. Here we show that mutations at Asp-53 or Asp-55 of theSulfolobus shibatae signature sequence abolish addition of both C and A, demonstrating that a single active site is responsible for addition of both nucleotides. Mutations at Asp-106 (and to a lesser extent, at Glu-173 and Asp-215) selectively impaired addition of A, but not C. We have previously demonstrated that the tRNA acceptor stem remains fixed on the surface of the CCA-adding enzyme during C and A addition (Shi, P.-Y., Maizels, N., and Weiner, A. M. (1998)EMBO J. 17, 3197–3206). Taken together with this new evidence that there is a single active site for catalysis, our data suggest that specificity of nucleotide addition is determined by a process of collaborative templating: as the single active site catalyzes addition of each nucleotide, the growing 3′-end of the tRNA would progressively refold to create a binding pocket for addition of the next nucleotide.

The CCA-adding enzyme (ATP (CTP):tRNA nucleotidyltransferase) adds the essential 3Ј-terminal CCA sequence to tRNA (1,2). This enzyme is present in all three living kingdoms (3). In organisms that do not encode the 3Ј-terminal CCA sequence, including eukaryotes, many eubacteria, and some archaea, the CCA-adding enzyme carries out an essential step in tRNA biosynthesis. In organisms like Escherichia coli where all tRNA genes encode CCA, the CCA-adding enzyme repairs 3Ј termini depleted by exonucleolytic attack, but is not absolutely essential for survival (4). The CCA-adding enzyme can use tRNA-CC, tRNA-C, or tRNA-N (containing only the discriminator base) as substrate. It completes the terminal CCA sequence by adding one nucleotide at a time. Despite the fact that the CCA-adding enzyme has no nucleic acid template, incorporation is highly specific and apparently error-free under physiological conditions: neither UTP nor GTP is incorporated into tRNA, and the enzyme does not add oligo(C) to tRNA-C when deprived of ATP or oligo(A) to tRNA-CC when deprived of CTP (3,5).
We recently cloned the CCA-adding enzyme from the ther-mophilic archaeon Sulfolobus shibatae (3). The archaeal CCAadding enzyme contained a clear match with the active-site signature of the nucleotidyltransferase superfamily (6), but it exhibited no strong homology to the eubacterial and eukaryotic CCA-adding enzymes (which are very similar to one another). This enabled us to divide enzymes of the nucleotidyltransferase superfamily into two distinct classes (3). Class I enzymes include the archaeal CCA-adding enzymes (Fig. 1), eukaryotic poly(A) polymerases, DNA polymerase ␤, and kanamycin nucleotidyltransferase. Class II enzymes include eubacterial and eukaryotic CCA-adding enzymes and eubacterial poly(A) polymerases. Class II enzymes share 25 kDa of strong Nterminal homology, but diverge at the C termini (3,7). In contrast, Class I enzymes exhibit little obvious similarity to each other or to class II enzymes outside of the active-site signature.
The nucleotidyltransferase superfamily active-site signature (6) contains two carboxylates (DXD(E)) that are thought to occupy the same positions in space as seen in the crystal structures of DNA polymerase ␤ (8,9) and kanamycin nucleotidyltransferase (10). These two carboxylates, together with a third contributed by a more distant region of the protein, coordinate two divalent metal cations that stabilize the trigonal bipyramidal transition state generated by S n 2 attack of a hydroxyl group on a phosphoester bond (11). Three carboxylates are necessary for DNA polymerase ␤ activity:  within the conserved signature (12) and the distant Glu-256 (13). Similarly, three carboxylates are necessary for bovine poly(A) polymerase activity: Asp-113 and Asp-115 within the signature and the distant Asp-167 (7). These data suggest that class I and II enzymes of the nucleotidyltransferase superfamily use the same reaction mechanism as many other polymerases (14,15). The alignment of Martin and Keller (7) predicted that the third carboxylate should fall within an RRD motif in class II enzymes and within an RXD motif in class I enzymes as previously shown for DNA polymerase ␤ and poly(A) polymerase. However, no RXD motif could be identified by inspecting the sequences of the four known archaeal CCAadding enzymes. We therefore set out to mutate systematically each of the conserved carboxylates in the S. shibatae CCAadding enzyme, keeping in mind that the archaeal enzyme might have dispensed with the need for a third carboxylate, as have a number of DNA and RNA polymerases (see "Discussion").
Sequence alignments predicted that the aspartate residues at positions 53 and 55 within the active-site signature of the S. shibatae CCA-adding enzyme would be critical to catalysis. Here we report that mutation of either Asp-53 or Asp-55 abolishes addition of both CTP and ATP by this enzyme. No third carboxylate seems to be absolutely required, although mutation of Asp-106 (and to a lesser extent, Glu-173 and Asp-215) diminishes ATP but not CTP addition. These data demonstrate that one catalytic center is responsible for addition of both CTP and ATP. In principle, a single catalytic center could translo-cate along the tRNA molecule to add all three nucleotides of the terminal CCA sequence. However, we have recently found that the tRNA acceptor stem does not translocate or rotate as each nucleotide is added to the 3Ј-end of tRNA-C, but remains fixed upon the surface of the enzyme (16). Taken together, the observations that there is a single active site and that RNA polymerization occurs without translocation argue that the specificity of nucleotide addition is determined by a process of collaborative templating in which the enzyme and the bound tRNA substrate jointly specify the identity of the next nucleotide to be added. At each step of addition, the growing 3Ј-end of the tRNA would progressively refold to create a binding pocket for addition of the next nucleotide, until addition ceases when the CCA-binding pocket is full.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis and Deletion-To facilitate protein purification and site-directed mutagenesis, the S. shibatae cca gene (3) was cloned into the pET22x(bϩ) vector (Novagen), which supplies a hexahistidine tag. First, plasmid pEt22cca ( Fig. 2) was generated by excising the cca gene from pET17cca by digestion with EcoRI and partial digestion with NdeI and cloning this fragment into the corresponding sites of pET22x(bϩ). To bring the carboxyl-terminal hexahistidine tag of the vector into frame with the CCA coding region, pET22cca-His was constructed by replacing sequences between the ClaI site in the gene and the HindIII site in the vector with a polymerase chain reaction fragment generated using primer oligonucleotides 1 and 3 and pET17cca as template (Table I); this fused the CCA coding region to the hexahistidine tag through a six-residue tether (LAAALE). The resulting con-struct was sequenced to ensure that no mutations had been introduced by polymerase chain reaction.
Site-directed mutagenesis was carried out according to Kunkel et al. (17). E. coli strain CJ236 (dut -, ung -) was transformed with pET22cca-His and infected with helper phage M13k07, and single-stranded phage DNA (corresponding to the CCA coding strand) was prepared by multiple phenol and chloroform extractions. Single-stranded DNA was annealed with mismatched oligonucleotides (synthesized at the W. M. Keck Biotechnology Center, Yale School of Medicine) ( Table I) by heating to 90°C, followed by slow cooling to 25°C in a 10-l reaction containing 50 mM NaCl, 10 mM Tris-HCl (pH 7.4), and 2 mM MgCl 2 . Mutant strands were elongated by incubation with T4 DNA polymerase (100 units/ml) and T4 DNA ligase (300 units/ml) at 37°C for 90 min in buffer containing 50 mM NaCl, 17.5 mM Tris-HCl (pH 7.4), 3.75 mM MgCl 2 , 0.5 mM dithiothreitol, 0.4 mM each dNTP, and 0.75 mM ATP. Reactions were stopped by addition of EDTA, and plasmids were transformed into E. coli strain BL21(DE3) (dut ϩ , ung ϩ ). Mutants were identified by DNA sequencing and retransformed into BL21(DE3), and retransformants were resequenced.
Two C-terminal mutants were constructed, ⌬C30 and ⌬C135, lacking 30 and 135 C-terminal residues, respectively. For ⌬C30, the region between the ClaI and HindIII sites in pET22cca-His was replaced with a polymerase chain reaction fragment that was generated using primer oligonucleotides 2 and 3 and pET17cca as template and then digested with ClaI and HindIII. The reconstructed region of ⌬C30 was sequenced to rule out mutations. For ⌬C135, the region between the two SacI sites in pET22cca was deleted, and the plasmid was recircularized, bringing the truncated CCA coding region into frame with the C-terminal His tag from the vector.
Enzyme Expression and Purification-His-tagged protein was expressed in E. coli BL21(DE3) by growth in 0.5 mM isopropyl-␤-D-thio-FIG. 1. Alignment of archaeal CCAadding sequences. Similarity between the CCA-adding enzymes of classes I (archaeal) and II (eubacterial and eukaryotic) appears to be confined to the boxed active-site signature (3). Conserved or partially conserved acidic residues that were mutated in the S. shibatae enzyme are indicated in blue. The GXGXXG motif in the S. shibatae is indicated in pink; the same motif can be found in the E. coli enzyme, but is only partially conserved in other class II enzymes (3). A potential amphipathic RNA-binding ␣-helical motif (RXXKXXXK) is highlighted in yellow; a possible functional homolog (RXXRXXR-XXXR) is conserved among class II enzymes (we thank J. Wang for pointing out this feature). The S. shibatae (3), M. jannaschii (21), M. thermoautotrophicum (22), and A. fulgidus (23) CCA-adding sequences were aligned by PILEUP and adjusted by eye. The apparently overlapping MJ1111 and MJ1112 reading frames (21) turned out to contain a solitary sequencing error (TTTT for TTT) that introduced a frameshift in the middle of the activesite signature (16). Gaps introduced for alignment are indicated by ellipses.
galactopyranoside for 3 h at 37°C to induce gene overexpression. Native protein was purified over a nickel column as recommended by the manufacturer (Novagen). About 2 mg of pure protein could be obtained from 1 g of wet cells. Purified protein was stored at -20°C in buffer containing 100 mM NaCl, 25 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, and 50% glycerol. Protein concentrations were measured using the Bio-Rad dye method. Tagged and untagged recombinant CCA-adding enzymes exhibited comparable activity and thermostability (data not shown).
Enzyme Assays-Standard assays measured incorporation of [␣-32 P]CTP or [␣-32 P]ATP (Amersham Pharmacia Biotech) into unfractionated E. coli tRNA trimmed with venom phosphodiesterase (see below). The 10-l reactions contained 100 mM glycine/NaOH (pH 8.5), 10 mM MgCl 2 , 50 M CTP, 100 M ATP, 40 pmol of tRNA, and 3 ng of enzyme (for CTP addition reactions) or 10 ng of enzyme (for ATP addition reactions). Reactions were terminated by addition of 5 l of 95% formamide and 20 mM sodium EDTA (pH 8.0) containing bromphenol blue and xylene cyanol, resolved by denaturing 10% polyacrylamide gel electrophoresis, and quantified by phosphoimager analysis. Reaction rates were measured by withdrawing 10-l aliquots from a 60-l reaction, and initial reaction rates were determined while the reaction was in the linear range. As the S. shibatae CCA-adding enzyme has an optimal temperature of 70°C (3), we compared the relative activities of various mutants at 56°C in order to reduce potential complications due to globally destabilizing mutations (Ref. 18; for review, see Ref. 19). The wild-type enzyme has about one-third the activity at 56°C as at 70°C (3). We also assayed mutants under low substrate "challenge" conditions; the conditions for CTP addition ([␣- tRNA Substrate Preparation-Addition of both C and A by the CCAadding enzyme can be assayed using crude unfractionated E. coli tRNA (Sigma, grade XXI), but the sensitivity of the assay is low and suffers from product inhibition. A more efficient substrate was prepared by trimming the 3Ј-ends of the unfractionated tRNA (100 mg total in 25 ml) with snake venom phosphodiesterase I (1.2 units/25 ml; Sigma) for 1 h at 20°C in 100 mM Tris-HCl (pH 9.0), followed by phenol/chloroform extraction. Under these conditions, partial digestion by the 3Ј-exonuclease leaves a tRNA mixture with N, NC, and NCC ends, where N is the discriminator base; addition of both CTP and ATP increases by ϳ10-fold after this treatment. To remove any rRNA fragments and degradation products, 3Ј-end-trimmed tRNA was bound to Q-Sepharose (2 mg of tRNA/ml column; Amersham Pharmacia Biotech) in 20 mM Tris-HCl (pH 7.5), washed with the same buffer containing 0.4 M NaCl, and eluted with a 0.4 -1.0 M NaCl gradient, and the pooled tRNA fractions were precipitated with ethanol. All mutant CCA-adding enzymes were assayed using a single batch of 3Ј-end-trimmed tRNA substrate. Bacillus subtilis tRNA Asp lacking the 3Ј-CA sequence was transcribed in vitro by T7 RNA polymerase using FokI-digested pmBsDCCA as template (a kind gift of N. Pace, University of California, Berkeley, CA) (20). The runoff transcript was gel-purified as described (16). tRNA was quantified assuming an A 260 nm of 24 at 1.0 mg/ml.

RESULTS
Signature Sequence Carboxylate Mutants-All CCA-adding enzymes belong to the nucleotidyltransferase superfamily and share a similar active-site signature, which contains a DXD motif. Based on crystal structures for two enzymes in this family, DNA polymerase ␤ and kanamycin nucleotidyltransferase, the carboxylates in the signature motif are thought to chelate the two metal ions involved in phosphoester bond transfer (11). Martin and Keller (7) demonstrated by mutational analysis that the two carboxylates in the active-site signature motif are indeed essential for catalysis by another enzyme in this family, bovine poly(A) polymerase. The S. shibatae CCA-adding enzyme contains a clear DXD motif in the predicted active site (3). To determine if the aspartate residues in this motif are critical for catalysis, we generated four mu-  tants: D53A, D55A, D55E, and the D53A/D55A double mutant. We expressed and purified His-tagged proteins carrying these mutations and assayed activity. As shown in Table II, both CTP-and ATP-adding activities were drastically reduced in each of the mutants. D53A exhibited no ATP-adding activity and only 0.6% residual CTP-adding activity, and D55A had no activity at all. Even the relatively conservative substitution of glutamate for aspartate was not tolerated at position 55: D55E retained only 0.3% ATP-adding activity and 1% CTP-adding activity. These data confirm the identity of Asp-53 and Asp-55 as active-site carboxylates and imply that the active site is highly structured.
Searching for a Third Active-site Carboxylate-Most enzymes that catalyze nucleotide transfer reactions, including the only two members of the nucleotidyltransferase superfamily whose structures are known, DNA polymerase ␤ (8,9) and kanamycin nucleotidyltransferase (10), contain three activesite carboxylate residues that chelate two metal ions. Sequence comparisons with other nucleotidyltransferase superfamily members predicted the identity of the third carboxylate in bovine poly(A) polymerase, and the function of this residue was then established by systematic mutational analysis (7). However, sequence comparisons with other members of the nucleotidyltransferase superfamily did not reveal a likely candidate for the third active-site carboxylate in the S. shibatae CCAadding enzyme. We therefore decided to test all carboxylates that are well conserved among the four known archaeal CCAadding enzyme sequences, S. shibatae (3), Methanococcus jannaschii (21), Methanobacter thermoautotrophicum (22), and Archaeoglobus fulgidus (23). As shown in Fig. 1, alignment of the sequences of these enzymes identified 12 conserved carboxylates, in addition to the two within the active-site signature. We systematically replaced each of the conserved carboxylates in the S. shibatae enzyme with alanine, creating mutants E104A, D106A, E114A, D124A, E139A, D143A, E144A, E161A, E173A, E209A, D215A, and D218A. Surprisingly, none of the mutations in any of these positions had a significant effect on CCA-adding activity assayed under standard conditions with excess substrate and saturating nucleotides (data not shown).
Challenging the Enzyme-Our failure to identify a third active-site carboxylate residue by systematic alanine scanning mutagenesis suggested that saturating levels of all three substrates (tRNA, CTP, and ATP) might have obscured any defects caused by mutation. We therefore retested the mutants at lower substrate concentrations (challenge conditions) in the hope of revealing subtle defects in substrate binding or catalysis. We defined the challenge conditions empirically and, as our goal was to compare the CCA-adding activities of many different mutants, we carried out all assays at 56°C to reduce the risk of thermal instability or denaturation that might occur at 70°C, the temperature optimum for the wild-type enzyme.
We first sought a tRNA substrate concentration that gave half-maximal CTP-adding activity at saturating ATP and CTP. CTP-adding activity was assayed using 0.125-4 M tRNA at high (200 M ATP and 50 M CTP) and low (50 M ATP and 12.5 M CTP) nucleotide concentrations. As shown in Fig. 3, 1 M tRNA was nearly saturating at both NTP concentrations, and we therefore used 0.25 M tRNA for subsequent assays.
Next we sought ATP and CTP concentrations that gave halfmaximal CTP-and ATP-adding activities at fixed 0.25 M tRNA. We assayed CTP-adding activity using 3-50 M ATP at fixed 25 M CTP (Fig. 4A), and using 1.5-25 M CTP at fixed 50 M ATP (Fig. 4B). This experimental protocol was designed to test whether the ATP-and CTP-binding sites of the archaeal enzyme interact with each other, as appeared to be the case for the rabbit liver CCA-adding enzyme (24,25). We found that CTP-adding activity was unaffected by up to 50 M ATP (Fig.  4A) and was half-maximal at 3 M CTP (Fig. 4B). Subsequent CTP addition assays were therefore performed at 3 M CTP ϩ 50 M ATP.
Similarly, we assayed ATP-adding activity using 6.2-100 M ATP at fixed 50 M CTP and found that 100 M ATP was not saturating (Fig. 5A). ATP was therefore fixed at 100 M, and ATP-adding activity was assayed by varying CTP from 3 to 50 M (Fig. 5B). ATP addition was reduced at 3 M CTP, the lowest CTP concentration tested. This was expected, as the tRNA substrates prepared by partial digestion of mature tRNA with venom phosphodiesterase lack the 3Ј-terminal CA and require addition of C before A. Subsequent ATP addition assays were therefore performed at 100 M ATP ϩ 3 M CTP.
All mutant CCA-adding enzymes, including point mutations and both deletions, were then retested for ATP-and CTPadding activities under challenge conditions (Fig. 6, left and right panels, respectively). Some level of CTP-and ATP-adding activities (5-90%) was evident in all mutants, except those in the signature sequence (D53A, D55A, D55E, and D53A/D55A) and the two C-terminal deletions (⌬C30 and ⌬C135). A number of the other mutants retained considerable activity under these conditions. However, in three mutants, activity was reduced to 20% or less of wild-type levels. D215A was defective in both CTP-and ATP-adding activities (16 and 13% of wild-type levels, respectively), and E173A was also defective in both CTPand ATP-adding activities (19 and 7% of wild-type levels, respectively). Most intriguingly, D106A retained 90% of normal CTP-adding activity, but was only 14% as active as the wildtype enzyme when assayed for ATP-adding activity.
A Shared GXGXGG Motif?-The E. coli CCA-adding enzyme TABLE II Activity of C-terminal deletion and signature sequence mutants Initial reaction rates are shown and were measured as described under "Experimental Procedures" using standard conditions with 5-120 ng of His-tagged purified enzyme/60-l reaction. All rates were normalized to the wild-type enzyme concentration.  contains a potential ATP-binding motif (GXGXXG) just downstream of the active-site signature sequence, and a mutation in the second G in the motif (G70D) had previously been shown to diminish ATP-but not CTP-adding activity by 7-10-fold in crude extract (26). However, this GXGXXG motif is not conserved in any other of the highly homologous class II nucleotidyltransferases such as the CCA-adding enzymes of a fungus (Saccharomyces cerevisiae), a higher plant (Lupinus albus), a metazoan (Caenorhabditis elegans), and a variety of other eubacterial CCA-adding enzymes and poly(A) polymerases (Acidominococcus fermentans, B. subtilis, Hemophilus influenzae, Mycobacterium leprosae, and Thermus aquaticus). Curiously, the S. shibatae CCA-adding enzyme also contains a GXGXXG sequence downstream of the signature sequence, and this motif is partially conserved in other highly homologous class I CCA-adding enzymes (M. jannaschii, M. thermoautotrophicum, and A. fulgidus) (Fig. 1). Although similarity between class I and II nucleotidyltransferases is apparently confined to the activesite signature sequence, we nonetheless asked whether the GXGXXG motif of S. shibatae (a class I CCA-adding enzyme) might be functionally equivalent to that of E. coli (a class II CCA-adding enzyme). As shown in Fig. 6, neither CTP-nor ATP-adding activity was significantly affected by a G156D mutation in the second G of the S. shibatae sequence. Thus, the GXGXXG motif is not conserved, either structurally or functionally, in class I or II CCA-adding enzymes. Most likely, the G70D mutation (26) affects ATP addition indirectly, perhaps by altering overall protein structure rather than the ATP-binding site.
D106A Preferentially Affects ATP Addition-To confirm the selective deficiency in A addition by mutant D106A, mutant enzymes that retained significant activity using venom-treated E. coli tRNA as substrate (Fig. 6) were re-assayed for CTP-and ATP-adding activities using a defined tRNA substrate. Unlabeled B. subtilis tRNA Asp lacking the 3Ј-terminal CA was synthesized as a runoff transcript by T7 RNA polymerase using FokI-digested pmBsDCCA as template (20). Assays were performed using 0.25 M tRNA Asp -C under challenge conditions for CTP addition ([␣-32 P]CTP, 3 M CTP, and 50 M ATP) or ATP addition ([␣-32 P]ATP, 100 M ATP, and 1.5 M CTP), and products were resolved by denaturing polyacrylamide gel electrophoresis to separate molecules differing by a single nucleotide. These experiments confirmed the observation that the D106A mutation preferentially affects ATP but not CTP addition: Asp-106 generates tRNA Asp -CC with efficiency approaching that of the wild-type enzyme, but almost no A addition is evident (Fig. 7). Although considerably less active than D106A, mutants E173A and D215A may also preferentially add C. In addition, the mutant enzymes generally seem more defective than the wild-type enzyme when assayed with the defined tRNA Asp -CC substrate (Fig. 7) compared with venom-treated E. coli tRNA (Fig. 6). This may reflect mild tRNA specificity, as might be expected for an enzyme that must be able to use Ͼ50 different cellular tRNA species as substrate. DISCUSSION We have identified two active-site carboxylates in the CCAadding enzyme from the thermophilic archaeon S. shibatae. Asp-53 and Asp-55 both fall within the conserved nucleotidyltransferase signature sequence, and four different mutations at these residues (D53A, D55A, D55E, and D53A/D55A) abolish addition of both CTP and ATP. Thus, the CTP-and ATP-adding reactions appear to use the same catalytic center, although strictly speaking, we cannot rule out distinct but overlapping catalytic sites.
To explain how the CCA-adding enzyme can accurately polymerize CCA without recourse to a nucleic acid template, Masiakowski and Deutscher (24,25) proposed that the active site contains multiple "subsites" recognizing the body of the tRNA, the 3Ј terminus of the tRNA primer, and each of the donor nucleotides. This model was supported by extensive kinetic data using partially purified rabbit liver enzyme and state-of-the-art oligonucleotide and tRNA-derived substrates. Using powerful new tools for protein overexpression and purification, site-directed mutagenesis, oligonucleotide synthesis, and in vitro transcription, we have now been able to revisit these earlier conclusions. Our results make it very unlikely that the CCA-adding enzymes contain multiple subsites.
The active sites of many polymerases (including the three best studied members of the nucleotidyltransferase superfam-ily: DNA polymerase ␤, poly(A) polymerase, and kanamycin nucleotidyltransferase) contain three key carboxylates that chelate two divalent metal ions (7-10, 15, 27, 28). Typically, two of these carboxylates are adjacent to one another or separated by a single residue in the primary sequence and can usually be identified by inspection of the primary sequence. The third carboxylate is more distant in primary sequence, although close in three-dimensional space, and it is often less obvious from sequence analysis. In the two other class I enzymes examined so far, DNA polymerase ␤ and bovine poly(A) polymerase, a third carboxylate is clearly essential for activity (7,13). An alignment of rat DNA polymerase ␤, bovine poly(A) polymerase, and kanamycin nucleotidyltransferase (7) suggested that the third carboxylate in the S. shibatae CCA-adding enzyme should fall within an RXD motif. As no conserved RXD motif was present in the four known archaeal CCA-adding enzymes (S. shibatae, M. jannaschii, M. thermoautotrophicum, and A. fulgidus), we performed alanine scanning mutagenesis of all 12 conserved or partially conserved carboxylates in the archaeal enzymes. None of the mutations in these conserved carboxylates abolished activity under standard assay conditions (saturating tRNA and CTP and nearly saturating ATP). A third carboxylate therefore appears not to be essential for CCAadding activity.
Three mutants exhibited significant defects (Ͻ20% activity) under challenge conditions, in which concentrations of tRNA, CTP, and ATP were limiting. E173A and D215A were defective in both CTP-and ATP-adding activities. This could indicate a global effect on enzyme structure, or because our data suggest that there is only a single catalytic site, one or both mutations could influence the structure of the active site. The most interesting mutant was D106A, which retained normal CTP-adding activity, but exhibited greatly reduced ATP-adding activity on both heterogeneous and defined tRNA substrates. Asp-106 cannot be involved in catalysis as predicted by Thurlow et al. (29) because this mutant has normal CTP-adding activity. We therefore propose that Asp-106 may be a component of the ATP-binding pocket and may contribute to ATP-adding specificity.
The S. shibatae CCA-adding enzyme is unusual among class I nucleotidyltransferase family enzymes analyzed thus far in not requiring a third carboxylate for catalysis. Nonetheless, the third carboxylate is known to be more important for some polymerases than for others. In the polymerase ␣ family, mutations in the invariant Asp-249 of 29 DNA polymerase reduce activity drastically (30), whereas mutations in the corresponding Asp-860 of human polymerase ␣ have little effect on activity, but do reduce dNTP binding (31). In contrast, the structurally equivalent Asp-705 of Klenow fragment and Asp- FIG. 7. D106A preferentially affects ATP-adding activity. Mutant enzymes that retained significant CTP-and ATPadding activities using venom-treated E. coli tRNA as substrate in Fig. 6 were reassayed under the same challenge conditions for ATP-adding activity (left panels) and CTP-adding activity (right panels) using in vitro transcribed B. subtilis tRNA Asp -C as substrate. Upper and lower panels represent two different gels, and asterisks indicate labeled nucleotide. WT, wild-type enzyme.
110 of reverse transcriptase play critical roles in both dNTP binding and catalysis (32)(33)(34). T7 RNA and DNA polymerases provide two other exceptions to the rule. Only two active-site carboxylates have been identified in T7 RNA polymerase (27,35). These two residues (Asp-537 and Asp-812) are far apart in the protein sequence, but both are important for metal binding and catalysis (36), and there is no candidate for a third carboxylate nearby in the crystal structure (37). The high resolution structure of T7 DNA polymerase complexed with a primertemplate and a deoxynucleoside triphosphate revealed two metals interacting with the strictly conserved Asp-475 and Asp-654, but not with Glu-655 (38). Thus, two carboxylates suffice for binding two active-site metals that stabilize the trigonal bipyramidal transition state required for phosphoester bond transfer.
If two carboxylates suffice for other polymerases, one must then ask what is the function of the third carboxylate in other enzymes of the nucleotidyltransferase superfamily? This residue could participate in catalysis, nucleotide triphosphate binding, or both as discussed below. We speculate that the third carboxylate in these other class I enzymes may be a component of a dNTP-or ATP-binding pocket. If so, mutations in the third carboxylate might have a greater effect on poly(A) polymerase (a template-independent enzyme) than on DNA polymerase ␤, where the template strand can help to select the incoming nucleotide.
Our phylogenetic analysis suggests that the CCA-adding enzyme and poly(A) polymerase may have interconverted at least once over the course of evolution (3): the eubacterial poly(A) polymerase and CCA-adding enzymes are highly homologous class I enzymes, whereas the eukaryotic poly(A) polymerase and CCA-adding enzymes belong to different classes of the nucleotidyltransferase superfamily and share no obvious similarity outside the active-site signature sequence (3). Thus, the eukaryotic CCA-adding enzyme could have "degenerated" into a poly(A) polymerase, or poly(A) polymerase might have acquired the ability to add CCA. The intraconversion of these two enzyme activities might also explain why some RNA and DNA polymerases preferentially add untemplated adenosines during runoff transcription (39 -41) or polymerase chain reaction amplification (42,43).
Finally, we propose that the CTP-and ATP-binding pockets of the CCA-adding enzyme represent a collaboration between the enzyme and the tRNA primer. Our reasoning is straightforward. 1) We have shown here that the S. shibatae CCAadding enzyme has only a single active site because CTP and ATP additions are both disabled by mutations in either carboxylate (Asp-53 or Asp-55) of the nucleotidyltransferase activesite signature sequence. 2) We have recently shown that the tRNA acceptor stem remains fixed upon the surface of the S. shibatae and E. coli CCA-adding enzymes as each new nucleotide is added to the growing 3Ј-end; the tRNA does not rotate or translocate relative to the enzyme (16). 3) Judging by the compact crystal structures of two other class I enzymes, DNA polymerase ␤ (8,9) and kanamycin nucleotidyltransferase (10), the active site is unlikely to move relative to the tRNA-binding site.
If neither the tRNA nor the active site moves as each nucleotide is added to the growing 3Ј-end of the tRNA, the 3Ј-end itself must move. We therefore propose the following model for RNA polymerization without translocation. As the active site catalyzes addition of each new nucleotide, the growing 3Ј-end of the tRNA would progressively refold to create a binding pocket for addition of the next nucleotide. Specifically, binding of tRNA-N (where N is the discriminator base) or tRNA-NC to the enzyme would create a binding pocket for CTP, whereas bind-ing of tRNA-NCC to the enzyme would create an ATP-binding pocket. This would be consistent with earlier kinetic studies showing that tRNA-N and tRNA-NC are poor substrates for ATP addition, tRNA-NCC is a poor substrate for CTP addition, and mutations in the 3Ј-terminal Cys-75 in tRNA-NCC prevent ATP addition (10,24,25). The enzyme and the growing 3Ј-end would collaboratively template CCA addition, and addition would cease when the CCA-binding pocket was full. We wish to emphasize that this model is likely to apply to both class I (archaeal) and class II (eubacterial and eukaryotic) CCAadding enzymes because the active sites of the two classes are highly homologous (3), and the tRNA does not rotate or translocate relative to either class of enzyme during CCA polymerization (16).
Collaborative templating is not radical or unprecedented. Selection of the correct nucleotide by template-directed polymerases also represents an obvious, although seldom acknowledged, collaboration between enzyme and substrate. The incoming nucleotide fits into a flat, narrow pocket; the incoming nucleotide forms hydrogen bonds with the edge of the template base on the floor of the pocket, and it stacks between the two flat hydrophobic walls of the pocket defined by the previous base pair (n Ϫ 1) and a critical tyrosine that behaves like the next base pair (n ϩ 1). Functionally equivalent tyrosines are found in Taq (44), T7 (38), and E. coli (45) DNA polymerases and human DNA polymerase ␣ (31). Thus, although the primer-template is a major determinant specifying the incoming nucleotide, template-dependent polymerases recognize the incoming sugar (ribose or deoxyribose) and provide one of the two hydrophobic walls that sandwich the incoming nucleotide, conferring greater stability and specificity on this key hydrogenbonding interaction. The possibility that the specificity of CCA addition is determined by collaborative templating provides a mechanistically plausible answer to the long-standing question of how the CCA-adding enzyme specifically adds the CCA sequence without an external nucleotide template.