A single catalytically active subunit in the multimeric Sulfolobus shibatae CCA-adding enzyme can carry out all three steps of CCA addition.

The CCA-adding enzyme ATP(CTP):tRNA nucleotidyltransferase builds and repairs the 3'-terminal CCA sequence of tRNA. Although this unusual RNA polymerase has no nucleic acid template, it can construct the CCA sequence one nucleotide at a time using CTP and ATP as substrates. We found previously that tRNA does not translocate along the enzyme during CCA addition (Yue, D., Weiner, A. M., and Maizels, N. (1998) J. Biol. Chem. 273, 29693-29700) and that a single nucleotidyltransferase motif adds all three nucleotides (Shi, P.-Y., Maizels, N., and Weiner, A. M. (1998) EMBO J. 17, 3197-3206). Intriguingly, the CCA-adding enzyme from the archaeon Sulfolobus shibatae is a homodimer that forms a tetramer upon binding two tRNAs. We therefore asked whether the active form of the S. shibatae enzyme might have two quasi-equivalent active sites, one adding CTP and the other ATP. Using an intersubunit complementation approach, we demonstrate that the dimer is active and that a single catalytically active subunit can carry out all three steps of CCA addition. We also locate one UV light-induced tRNA cross-link on the enzyme structure and provide evidence suggesting the location of another. Our data rule out shuttling models in which the 3'-end of the tRNA shuttles from one quasi-equivalent active site to another, demonstrate that tRNA-induced tetramerization is not required for CCA addition, and support a role for the tail domain of the enzyme in tRNA binding.

The 3Ј-terminal CCA sequence of mature tRNA is essential for efficient aminoacylation (1,2) as well as for translation because it base pairs with the large rRNA near the peptidyltransferase center (3,4). The CCA-adding enzyme ATP(CTP): tRNA nucleotidyltransferase builds and repairs the 3Ј-terminal CCA sequence of tRNA one nucleotide at a time using CTP and ATP as substrates (5). CCA-adding activity has been identified in all three kingdoms (6) and is an essential biosynthetic activity in eukaryotes (7) and presumably in Archaea where some or all tRNA genes lack encoded CCA, but it is partially dispensable in those eubacteria that encode all CCA sequences (8). The CCA-adding enzyme is a member of the nucleotidyltransferase superfamily defined by a signature 16-kDa structure that uses a two-metal ion mechanism (9, 10) to transfer a mononucleotide from a nucleotide triphosphate donor to an acceptor hydroxyl group (11). The nucleotidyltransferase superfamily can be divided into at least two distinct classes (6,12,13), namely class I nucleotidyltransferases, which include archaeal CCA-adding enzymes and eukaryotic poly(A) polymerases, and class II nucleotidyltransferases, which include eukaryotic and eubacterial CCA-adding enzymes as well as eubacterial poly(A) polymerases (6). The observation that both the class I and class II nucleotidyltransferase families include CCA-adding enzymes and poly(A) polymerases suggests that these two activities may interconvert over evolutionary time (6). The discovery of pairs of CC-and A-adding enzymes in three widely different species of eubacteria is consistent with the postulated interconversion (14,15). Although class I and class II CCA-adding enzymes share little sequence similarity outside the signature structure, the mechanisms of action may be similar because tRNA binds almost identically to class I and class II enzymes (16), and the two classes make very similar use of nucleotide analogs (17).
Crystal structures are known for the class I archaeal Archaeoglobus fulgidus CCA-adding enzyme (18,19) and the class II CCA-adding enzymes from the eubacterium Bacillus stearothermophilus and humans (20,21). As expected (6,(11)(12)(13), these structures reveal homologous nucleotidyltransferase domains (head domains), but the other domains (neck, body, and tail) are completely different, suggesting that the mechanisms of nucleotide selection, the recognition of tRNA substrates, and possibly the orientation of the tRNA minihelix relative to the nucleotidyltransferase domain may differ between the two classes (18,20). Soaking the class II B. stearothermophilus apoprotein crystals in CTP or ATP yielded cocrystals exhibiting a single nucleotide binding site that could function as a protein template by recognizing both CTP and ATP; this protein template, consisting of aspartate (Asp-154) and arginine (Arg-157) residues contributed by the neck domain, appears to "base pair" with the Watson-Crick face of both nucleotides (20); however, the class I archaeal A. fulgidus enzyme apparently binds nucleotides nonspecifically, at least in the absence of tRNA substrates (18).
The CCA-adding enzyme is unique among members of the nucleotidyltransferase superfamily because the enzyme synthesizes a precise sequence without using a nucleic acid template (16,22). Five different models have been proposed to explain CCA addition in the absence of a nucleic acid template (5,16,20,(23)(24)(25)(26). First, the two-site model postulates at least two nucleotide binding sites, one for CTP and the other for ATP (5,23). Second, the poly(C) polymerase model argues that the CCA-adding enzyme is a poly(C) polymerase that undergoes a significant conformational change after two rounds of CTP addition; this conformational change would block further poly(C) synthesis and allow the addition of a single, possibly prebound ATP (25)(26)(27). Third, the "collaborative templating" model proposes that the growing 3Ј-end of the tRNA is progressively sequestered within a pocket near the active site; progressive packing of this pocket would create a binding site for the incoming nucleotide, and CCA synthesis would cease when the pocket was full (16,28). Fourth, the "scrunching-shuttling" model proposes that two identical subunits of the tRNA-induced tetrameric enzyme would function nonequivalently (24). One subunit would add C74 and then "scrunch" or bulge C74 to allow the addition of C75; the CC terminus would then be long enough to "shuttle" to a nonequivalent subunit, which would add A76. Fifth, the "protein-assisted scrunching" model (20) proposes that the scrunching of the 3Ј-terminal CC (but not C74 alone) would cause a conformational change that switches the specificity of the nucleotide binding pocket from CTP to ATP.
The scrunching-shuttling model of CCA addition attempts to explain the otherwise puzzling observation that the dimeric Sulfolobus shibatae CCA-adding enzyme is a tRNA-inducible tetramer that binds two tRNA molecules (24). To test whether the growing 3Ј-end of tRNA can shuttle within the tetramer from one quasi-equivalent active site that adds CTP to a second quasi-equivalent active site that adds ATP (24), we performed intersubunit complementation experiments. Using the wild type S. shibatae CCA-adding enzyme as well as three different mutants that block the addition of C74 (Y95V), the addition of A76 (H93V), or the addition of both CTP and ATP (K153A), we generated heterodimers by the reassortment of tagged and untagged subunits under partially denaturing conditions. We found that a single wild type subunit in an immobilized heterodimer can still add CCA; moreover, by cross-linking the tRNA substrate to the enzyme before or after subunit reassortment, we were able to show that the inactive subunit in an active heterodimer contacts the tRNA substrate. Thus, our data rule out shuttling models in which the 3Ј-end of the tRNA shuttles from one quasi-equivalent active site to another (24) but are consistent with a protein-assisted scrunching model for a CCA addition 1 (20) in which the scrunching of CC (but not C74 alone) switches the specificity of nucleotide binding site from CTP to ATP.

MATERIALS AND METHODS
Preparation of tRNA Substrates-tRNA substrates lacking A, CA, or CCA (tRNA-DCC, tRNA-DC, or tRNA-D, where D is the discriminator base) were prepared as described previously by in vitro transcription of FokI-or BbsI-digested pmBsDCCA encoding Bacillus subtilis tRNA Asp (17). Following purification by electrophoresis in a 12% polyacrylamide gel containing 8 M urea, the tRNA band was visualized by UV shadowing, excised, eluted, and concentrated by ethanol precipitation.
Untagged and Histidine-tagged CCA-adding Enzymes-Untagged and histidine-tagged S. shibatae CCA-adding enzymes were expressed and purified as described previously 1 (6,16,17). Individual mutant H93V, Y95V, and K153A are described in detail elsewhere, 1 and untagged mutants (H93V, Y95V, and K153A) were constructed by the fragment exchange using the DraIII and BamHI restriction sites. All constructs expressing mutant enzymes were confirmed by DNA sequencing. Hexahistidine-tagged enzymes were purified by affinity chromatography on nickel-nitrilotriacetic acid (Ni-NTA) 2 resin (Qiagen), and untagged enzymes were purified by ion exchange chromatography on SP-Sepharose (Amersham Biosciences) as described previously (16). An untagged S. shibatae CCA-adding enzyme migrated as a single band of 48 kDa upon SDS-10% PAGE, and the hexahistidine-tagged enzyme migrated as a single band of 50 kDa.
Heterodimers of the S. shibatae CCA-adding Enzyme-To prepare heterodimers of the S. shibatae CCA-adding enzyme, a 4:1 mixture of untagged and hexahistidine-tagged enzyme (100 g/ml total protein concentration) in Buffer B (20 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 5% glycerol, and 0.1 mM phen-ylmethylsulfonyl fluoride) was denatured by the addition of solid urea to 8 M followed by incubation at 20°C for 1 h. The enzyme was renatured and the subunits were reassorted by diluting the urea to 6, 4, 2, and 0 M with Buffer B; heterodimers and hexahistidine-tagged homodimers were purified by affinity chromatography on Ni-NTA resin and recovered from the column as described (16).
Use of Ni-NTA HisSorb Plates to Prevent tRNA-induced Tetramer Formation-To avoid tRNA-induced tetramer formation, heterodimers generated as described above were adsorbed on Ni-NTA HisSorb plates essentially as described by the manufacturer (Qiagen). Briefly, 200 l of CCA-adding enzyme in phosphate-buffered saline (50 mM potassium phosphate, pH 7.2, 150 mM NaCl) at a concentration of 1 mg/ml was added to each well of an Ni-NTA HisSorb plate and incubated for 1 h at room temperature. After incubation, the wells were washed four times with phosphate-buffered saline containing 0.05% Tween detergent to remove nonspecifically bound proteins. The reaction volume was increased from 10 to 100 l when the assay was performed on Ni-NTA HisSorb plates. The 100 l reactions were identical to the standard 10-l reactions, except that CTP and [␣-32 P]CTP were increased to 500 and 0.5 M, respectively; ATP and [␣-32 P]ATP were increased to 1 and 0.5 M, respectively; tRNA was increased to 2 M; and the CCA-adding enzyme was increased from 10 nM (in solution) to 10 pmol (immobilized on the HisSorb plate), which would be equivalent to 100 nM enzyme if it were free to diffuse throughout the 100-l reaction volume. After incubation at 55°C for 3 min the reactions were withdrawn from the Ni-NTA HisSorb plates, and the tRNA was purified by phenol extraction and concentrated by ethanol precipitation using glycogen as a carrier. Reaction products were resolved by denaturing 12% PAGE and quantified by phosphorimaging (Amersham Biosciences).
CCA Addition to UV Light-cross-linked tRNA-Enzyme Complexes-UV light cross-linking was carried out essentially as described previously (16); S. shibatae CCA-adding enzyme was denatured with 8 M urea for 1 h, diluted into 4 M urea, and incubated for 10 min at room temperature with 5 g of unlabeled tRNA-D in standard reaction buffer lacking CTP and ATP. After irradiation at 254 nm with a hand-held UV light (Spectronics Corp.) for 10 min on ice, uncross-linked tRNA was removed by Amicon YM-30 filtration (Millipore). To generate heterodimers bearing cross-linked tRNA, 5 g of cross-linked tRNA-enzyme complex was mixed with 20 g of histidine-tagged wild type or mutant enzyme, and the reactions were denatured and renatured; heterodimers were recovered on a Ni-NTA affinity column, eluted with imidazole in buffer A, dialyzed into buffer B, and concentrated by Amicon YM-30 filtration. The CTP addition to UV-cross-linked tRNAenzyme complexes was assayed using 0.5 M [␣-32 P]CTP and 50 M ATP; ATP addition was assayed by using 0.5 M [␣-32 P]ATP and 50 M CTP. All reactions were incubated at 70°C for 10 min, analyzed by 10% SDS-PAGE, and quantified by phosphorimaging (Amersham Biosciences).
tDNA Minihelix for UV Cross-linking-tDNA minihelices with a 13-base pair stem were generated by annealing equimolar amounts of two gel-purified oligonucleotides, namely 5Ј-CC*CGTCATCACCC-ArC-3Ј (where C* is 5-bromocytidine and rC is a cytidine ribonucleotide) and 5Ј-GGGTGATGGCGGG-3Ј (DNAgency). The oligonucleotides were brought to a concentration of 15 M in 5 mM HEPES (pH 7.9), 50 mM NaCl, and 10 mM MgCl 2 , denatured by heating to 80°C for 3 min, and quick-cooled on ice for 10 min (29). A 200-l reaction containing 6 M (1200 pmol) tDNA minihelix and 6 M (1200 pmol) S. shibatae enzyme was incubated in 100 mM glycine/NaOH (pH 9), 10 mM MgCl 2 , and 1 mM dithiothreitol for 15 min at room temperature and then on ice for 10 min before irradiation. The 200-l reaction was spotted in 10-l aliquots onto a sheet of Parafilm © resting on an aluminum block in ice and irradiated at 312 nm for 30 min at a distance of 2 cm using a hand-held Spectroline medium wavelength UV transilluminator (Spectronics Corporation) as described (30 -32). When required, the tDNA minihelix was 5Ј-labeled with T4 polynucleotide kinase and [␥-32 P]ATP (3000 Ci/ mmol; Amersham Biosciences) as described by the supplier (New England Biolabs).
Tryptic Digestion and Purification of the Cross-linked Peptide-After 312-nm irradiation, cross-linked samples were subjected to Amicon YM-30 filtration (Millipore) to remove uncross-linked tRNA. The crosslinked complexes were brought to 2 ml with 50 mM Tris-HCl (pH 7.6), 1 mM CaCl 2 , and 50 mM NH 4 HCO 3 (pH 7.8). After the addition of 3 g (a 1:20 weight ratio) of sequencing grade modified trypsin (Promega), digestion was allowed to proceed for 2 h at 37°C, followed by incubation with an additional 3 g of trypsin for another 1 h. The reaction was stopped by quick freezing, diluted with buffer (50 mM KCl, 20 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 5% glycerol) to a salt concentration of 50 mM and passed over a 300-l DEAE-Sepharose ion exchange column (Amersham Biosciences). The column was washed with low salt buffer (50 mM KCl, 20 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 5% glycerol) to remove free peptides, and cross-linked peptides were then recovered by stepwise elution with 100 -1000 mM KCl in the same buffer. The 400 mM KCl fraction contained cross-linked tDNA-peptide and was sent for microsequencing by Edman degradation (Yale Howard Hughes Medical Institute/Keck Biotechnology Resource Laboratory).

RESULTS AND DISCUSSION
Generating Heterodimers That Contain a Single Active Subunit-We established previously that the S. shibatae CCAadding enzyme has only a single active site (28) and that the enzyme can faithfully add CTP and ATP to tRNA substrates that have been cross-linked to the enzyme by UV light irradiation (16). These data ruled out conventional translocation of the tRNA minihelix along the enzyme as CCA is added and suggested instead that the 3Ј-end of the tRNA substrate refolds (16) or "scrunches" (20) to reposition the growing 3Ј-end in a single active site. However, subsequent discovery that the dimeric S. shibatae enzyme is a tRNA-inducible tetramer suggested a very different model in which the growing 3Ј-end could shuttle within the tetramer from one quasi-equivalent active site that adds CTP to a second quasi-equivalent site that adds ATP (24). The quasi-equivalent sites would have different nucleotide specificity either because the tetramer is intrinsically asymmetric or because tRNA binding induces an asymmetric conformational change. In the simplest version of this shuttling model, the tRNA minihelix would remain fixed on the enzyme; however, the 3Ј-end of the shorter tRNA-N and tRNA-NC substrates could only reach the CTP addition site, whereas the 3Ј-end of the longer tRNA-NCC substrate could reach the ATP addition site. We set out to test this alternative model by asking whether an active subunit can complement inactive subunits within a heterodimeric enzyme.
To generate heterodimers of mutant and/or wild type enzymes, hexahistidine-tagged and untagged homodimers were combined pairwise, denatured in urea, and renatured by stepwise dilution, and dimers bearing the hexahistidine tag were recovered on Ni-NTA resin (Fig. 1A). In addition to wild type, three different S. shibatae mutant enzymes were used as follows. 1) Y95V blocks the addition of C74 but not the addition of C75. 2) H93V blocks the addition of A76; and 3) K153A blocks both CTP addition and ATP addition. 1 As shown in Fig. 1B, 6 M urea is sufficient to cause a complete reassortment of subunits. A 1:4 mixture of tagged and untagged subunits was denatured in 0, 2, 4, 6, or 8 M urea and renatured by dilution, and tagged dimers were recovered on a Ni-NTA column. Above 6 M urea, the ratio of tagged to untagged subunits is constant and roughly 5:4, as expected for complete subunit reassortment. Confirming the selectivity of the purification, only tagged subunits are retained on Ni-NTA resin when the urea denaturation step is omitted (Fig. 1B). Although complete subunit reassortment requires 6 M urea, significant reassortment is seen at 4 M urea, a concentration that is only mildly denaturing for most proteins; thus, it seems likely that urea preferentially weakens the dimer interface rather than completely denaturing the subunits.
Heterodimers were generated as in Fig. 1, eluted from the Ni-NTA resin with imidazole, concentrated by Amicon YM-30 filtration, and assayed for CCA-adding activity (Fig. 2). Intriguingly, heterodimers of a wild type subunit with the C74defective Y95V mutant, the A76-defective H93V mutant, and the catalytically inactive K153A mutant were all able to add CCA. We cannot conclude from this experiment alone that CCA is added by a single catalytically active subunit, because the addition of the tRNA substrate induces heterodimers to form tetramers (24). These tetramers could, in principle, juxtapose two wild type subunits as postulated by the "shuttling" model, where one quasi-equivalent active site would add CTP and the other ATP. We can conclude, however, that the heterodimers are well behaved; heterodimerization does not inactivate the wild type subunit, nor is activity of the heterodimer mutantspecific (Fig. 2).
Use of Immobilized Enzymes to Generate Functional Heterodimers-Our dissociation/reassociation protocol allows functional subunit reassortment (Fig. 2), but we could not distinguish whether CCA was added by a single wild type subunit in the tRNA-induced heterotetramer (a dimer of heterodimers) or by sequential activity (possibly CC and A addition) of the two wild type subunits within a tRNA-induced heterotetramer (24). To test whether a single wild type subunit in a heterodimer can add CCA, we took advantage of Ni-NTA HisSorb plates (Qiagen) to immobilize histidine-tagged dimers in the absence of tRNA, thus preventing tRNA-induced tetramer formation when tRNA substrate is added. Immobilization is gentle, and the activity of bound proteins can be assayed directly on the plate.
No functional heterodimers were observed when tagged mutants (A76-defective, C74-defective, or inactive) were mixed with untagged wild type enzyme and applied directly to HisSorb plates omitting the subunit reassortment step. Immobilization did not interfere with enzymatic activity (Fig. 3A,  lanes 1-4), and wild type subunits did not reassort with tagged mutants under these conditions (Fig. 3A, lanes 5-7). When a subunit reassortment step was included before immobilization, all three heterodimers displayed wild type activity (Fig. 3B,  compare lanes 1-3 with lanes 5-7).
The ability of a wild type subunit to complement an inactive subunit (Fig. 3B, lane 7) strongly suggests that the active site of a single subunit is responsible for all three steps of CCA addition; but the remote possibility remained that tRNA could induce tetramer formation between two immobilized heterodimers. We can exclude this possibility, however, because two observations argue that enzyme activity on the HisSorb plates is dilution-independent. First, half-saturation of subunit binding sites on the HisSorb plate results in roughly halfmaximal enzyme activity, not quarter-maximal as would be expected for a bimolecular reaction (Fig. 3C, right panel). Second, equivalent picomoles of subunits have comparable activity whether immobilized as heterotetramers (Fig. 3C, left panel) or heterodimers (Fig. 3C, right panel).
tRNA Cross-linked to One Subunit of the Heterodimer Serves as Substrate for the Other Subunit-To ask whether the inactive subunit in an active heterodimer (Fig. 3) plays a direct role in CCA addition, tRNA was UV light-cross-linked to heterodimers before subunit reassortment. This analysis was greatly simplified by the observation that UV light cross-linking of tRNA to the enzyme either in the presence of 4 M urea or following exposure to 4 M urea during the renaturation protocol generates only one of the two cross-linked complexes seen when cross-linking was performed in the absence of urea (Fig. 4; compare Ref. 16). Although we do not understand why exposure to 4 M urea prevents the formation of complex 1 or why it slightly reduces the yield of complex 2 (Fig. 4), both complex 1 and complex 2 are sensitive to proteinase K and RNase T1 digestion as expected for covalent tRNA/protein complexes (data not shown). Complexes 1 and 2 may be cross-linked to different sites on the tRNA, different sites on the enzyme subunit, or different subunits within the enzyme dimer (see Fig. 6). In any event, tRNA cross-linked to the enzyme was nearly as good a substrate as free tRNA under standard assay conditions (16), adding C75 and A76 with Ͼ80% efficiency (data not shown).
To prepare heterodimers in which tRNA was cross-linked to the inactive subunit, we cross-linked tRNA to inactive subunits before subunit reassortment but in the presence of 4 M urea to generate complex 2 (Fig. 5A). Uncross-linked tRNA was removed by Amicon YM-30 filtration (Millipore), the cross-linked inactive subunits were reassorted with tagged wild type subunits by stepwise dilution from 8 M urea, and heterodimers were recovered by affinity chromatography on Ni-NTA resin (Fig. 5A). Heterodimers cross-linked to tRNA were eluted with imidazole and assayed under standard conditions with labeled CTP or ATP, and the products were resolved by 10% SDS-PAGE (Fig. 5B). A single wild type subunit conferred full CCAadding activity on a heterodimer containing tRNA cross-linked to either a C74-or A76-defective subunit (Fig. 5B, first lanes of all four panels), whereas a fully defective subunit did not confer activity on the equivalent heterodimer (Fig. 5B, second lanes of  all four panels). This experiment argues against the possibility that association with a wild type subunit can allosterically reactivate the K153A mutant; allosteric reactivation of the Y95V and H93V mutants is also unlikely because these S. shibatae mutations, when mapped onto the crystal structure of the highly homologous A. fulgidus CCA-adding enzyme (18,19), lie on the surface of the nucleotide binding pocket. 1 Because a single active subunit in a heterodimer can add CCA (Fig. 3), we conclude that an active subunit can add CCA to tRNA that is cross-linked to the inactive subunit within a heterodimer.
Identifying the Binding Site for the tRNA TC Stem-Loop Region-The CCA-adding enzyme recognizes primarily the top half of tRNA (acceptor and TC stem-loop) (16), and the tDNA minihelices corresponding to the top half of Escherichia coli tRNA Val are good substrates for the E. coli CCA-adding enzyme as long as the 3Ј-terminal residue is a ribonucleotide (29). To identify residues within the CCA-adding enzyme that are involved in binding the top half of tRNA, we synthesized a tDNA minihelix that spans the acceptor stem and the TC stem, carries a 3Ј-terminal ribonucleotide corresponding to C74, and is site-specifically labeled with a photoactivatable bromocyti- dine residue at position 61 near the TC loop (Fig. 6A). UV cross-linking of the 5Ј-labeled tDNA minihelix to the S. shibatae CCA-adding enzyme in the absence of urea generated a single cross-linked complex (Fig. 6B, lane 2); no cross-linked complex was formed when the same tDNA minihelix lacked bromocytidine (Fig. 6B, lane 4). After trypsin digestion, crosslinked peptides were purified by ion exchange over DEAE-Sepharose (Amersham Biosciences) and subjected to aminoterminal microsequencing (Yale Howard Hughes Medical Institute/Keck Biotechnology Resource Laboratory), which identified the tryptic peptide NIGQYYLNIGPQYYSETID-DFIQK. As discussed below, this peptide may be near the TC loop when the tDNA is docked to the structure of the highly  (16). However, although reassorted subunits retain full CCA-adding activity (Fig. 2), only complex 2 is seen when cross-linking is performed on reassorted subunits that have been subjected to partial denaturation in 4 M urea during the reassortment protocol.
FIG. 5. A role for both subunits of the S. shibatae dimer in CCA addition; tRNA cross-linked to one subunit of a heterodimer can serve as substrate for the other subunit. A, preparation of soluble heterodimers in which tRNA is cross-linked to a single subunit as in Fig. 4. B, tRNA-D cross-linked to a C74-or A76-defective mutant subunit can serve as substrate for CTP addition (left panel) or ATP addition (right panel) by a wild type (WT) subunit in the same heterodimer or heterotetramer.  1-4), nor does an immobilized tagged enzyme spontaneously reassociate with untagged enzyme (lanes 5-7). B, immobilization of heterodimers on Ni-NTA HisSorb plates following the dissociation/reassociation protocol does not inhibit C or A addition (lanes 1-4), and one wild type (WT) subunit within the heterodimer is sufficient for full activity (lanes 5-7). C, immobilized heterodimers are as active as immobilized tRNA-induced heterotetramers, and activity is dilution-independent. The protein binding capacity of each HisSorb well is 10 pmol. Tagged wild type enzyme with or without equimolar tRNA was immobilized, washed as usual, and assayed for CCA addition on the plate, and the products were resolved by denaturing 12% PAGE. Assays were performed as described under "Materials and Methods." The average of two independent experiments is shown; error bars indicate S.D.
The S. shibatae CCA-adding enzyme is highly homologous to the A. fulgidus CCA-adding enzyme (28), which is known to dimerize through its body and tail domains (18,19). The dimensions of the resulting cleft in the A. fulgidus dimer and the distribution of electrostatic surface potential suggest that both the body and tail domains may be involved in tRNA binding (18). Indeed, when tRNA is docked onto the A. fulgidus structure (18) using tRNA protection and interference data as a strong constraint (16), the acceptor and TC stem of tRNA are located within the cleft, whereas the tail domain of the catalytically active subunit may interact with the TC loop (Fig.   6C). This docking also places the TC loop close to a region of the catalytically inactive subunit in the A. fulgidus dimer (18,19).
We mapped the cross-linked S. shibatae peptide onto the structure of the A. fulgidus apoenzyme docked to tRNA ( Fig.  6C; adapted from Ref. 18), because the class I A. fulgidus and S. shibatae CCA-adding enzymes are highly homologous in sequence (28) and function (17) and, thus, presumably in structure. The cross-linked peptide spans region ␤14-␣N of the active subunit (18,19) and thus may correspond to complex 1 obtained with intact tRNA (Fig. 4). The TC loop could also potentially interact with helix ␣L or region ␣H-␤8 of the inactive subunit (18,19), perhaps corresponding to complex 2 ob-FIG. 6. UV cross-link between a tDNA minihelix and the S. shibatae CCA-adding enzyme. A, tDNA minihelix used in UV cross-linking experiment. Left, alkylation protection and interference data for the class I S. shibatae enzyme with B. subtilis tRNA Asp as substrate; larger symbols indicate stronger effects (16). Middle, E. coli tRNA Val used as model for the tDNA minihelix (29). Right, data from left panel mapped onto the tDNA minihelix used in the cross-linking experiment. B, identification of a tryptic peptide that can be cross-linked near the TC loop of the tDNA minihelix substrate. Upper, UV light cross-linking between the S. shibatae CCA-adding enzyme and the tDNA minihelix substrate in panel A. The tDNA minihelix was 5Ј-labeled with T4 polynucleotide kinase before UV light cross-linking. Selection was done on DEAE-Sepharose for peptides bearing a cross-linked tDNA oligonucleotide tag (see "Results and Discussion") enriched for the indicated tryptic peptide. C, location of the cross-linked S. shibatae peptide on the structure of the A. fulgidus apoenzyme docked to tRNA (18). For simplicity, only the A. fulgidus monomer is shown in ribbon representation. The head, neck, body, and tail domains are colored magenta, green, cyan, and yellow, respectively. tRNA is shown in stick representation with the backbone path highlighted by a lavender coil. Magenta spheres indicate phosphates that are protected from alkylation or display enhanced alkylation when tRNA binds to the enzyme. Gold spheres indicate phosphates that interfere with CCA-adding activity when alkylated. The 5-bromocytidine at position 61 in the tDNA minihelix is mapped onto the intact tRNA structure. The cross-link between 5-bromocytidine and the ␤14-␣N region of the tail domain of the catalytically active subunit could reflect fraying at the end of the tDNA minihelix T-stem, which is unconstrained by a TC loop; alternatively, the tDNA minihelix may be able to slide along the binding cleft, functioning as substrate for CCA addition at one extreme (29) and as a substrate for cross-linking at the other. See "Results and Discussion" for additional details. tained with intact tRNA (Fig. 4). Although evidence has been published for both 1:1 and 2:1 complexes of class I CCA-adding enzymes with tRNA (18,19,24), the ratio may depend on the organism of origin, the concentrations of enzyme and tRNA, and the conditions used for the gel shift analysis. 3 For simplicity, we have assumed a 2:1 complex in Fig. 6C, but similar arguments could be made for a 1:1 complex. We still do not understand why the stable S. shibatae CCA-adding enzyme dimer forms a tetramer upon binding of one tRNA per dimer (24), and the crystal structure of the homolgous A. fulgidus enzyme does not provide any hints regarding the nature of the dimer/dimer interface in the tRNA-induced tetramer (18).
Implications for the Mechanism of CCA Addition-We have presented the following two kinds of evidence that a single catalytically active subunit within a dimeric S. shibatae CCAadding enzyme can add CCA. 1) A single catalytically active subunit in an immobilized heterodimer can carry out all three steps of CCA addition (Fig. 3); and 2) a single active subunit can carry out all three steps of CCA addition in a heterodimer containing tRNA that is cross-linked to the inactive subunit (Fig. 5B). These data rule out shuttling models in which one subunit adds CTP and a quasi-equivalent subunit adds ATP (24). Just as importantly, the data demonstrate that tRNAinduced tetramer formation (24) is not required for CCA addition. It will be interesting to see whether other class I archaeal CCA-adding enzymes are tRNA-inducible tetramers (but see Ref. 19), as this may help to explain whether tRNA-inducible tetramerization of the S. shibatae enzyme is fortuitous or essential for some activity other than CCA addition (24).
We have also presented evidence based on UV cross-linking experiments with intact tRNA (Fig. 5) and a site-specifically labeled tDNA minihelix (Fig. 6) that identify one UV lightinduced cross-link and suggest the location of a second. The cross-link between tRNA and the catalytically inactive subunit (Fig. 5) is likely to be between the TC loop and helix ␣L or region ␣H-␤8 (18,19). The cross-link between 5-bromocytidine at position 61 of the tDNA minihelix and the enzyme involves region ␤14-␣N of the active subunit (18,19). Although the 5-bromocytidine is remote from ␤14-␣N when mapped onto intact tRNA (Fig. 6C), it could approach ␤14-␣N more closely if the end of the tDNA minihelix T-stem were to fray or if the tDNA minihelix itself were to slide along the binding cleft, enabling the minihelix to function as a substrate for a CCA addition in one extreme register (29) and as a substrate for UV cross-linking in the other.
The UV cross-links support a role for the tail domain in tRNA binding, consistent with the observation that deletion of the tail of the homologous A. fulgidus CCA-adding enzyme generates a monomer with lower affinity for tRNA substrate (19). Indeed, effective tRNA binding may require not only the presence of the tail domain but stabilization of this domain by interactions across the dimer interface between the juxtaposed body and tail domains (18,19). It is surprising that the tDNA minihelix can serve as an efficient substrate for CCA addition (29) and also cross-link to the ␤14-␣N region, because a 13-base pair stem is not long enough to reach between the active site and the cross-linking site. The simplest interpretation is that the tDNA minihelix can slide along the tRNA binding cleft, implying that the tail domain of catalytically active subunit or the helix ␣L region of the inactive subunit or both may interact with the TC loop so as to block tRNA translocation after each nucleotide addition (Fig. 6C). This possibility could potentially explain why class I archaeal CCA-adding enzymes are dimers but class I eukaryotic poly(A) polymerases function as monomers (12,33).