The RNA-binding PUA Domain of Archaeal tRNA-Guanine Transglycosylase Is Not Required for Archaeosine Formation*

Bacterial tRNA-guanine transglycosylase (TGT) replaces the G in position 34 of tRNA with preQ1, the precursor to the modified nucleoside queuosine. Archaeal TGT, in contrast, substitutes preQ0 for the G in position 15 of tRNA as the first step in archaeosine formation. The archaeal enzyme is about 60% larger than the bacterial protein; a carboxyl-terminal extension of 230 amino acids contains the PUA domain known to contact the four 3′-terminal nucleotides of tRNA. Here we show that the C-terminal extension of the enzyme is not required for the selection of G15 as the site of base exchange; truncated forms of Pyrococcus furiosus TGT retain their specificity for guanine exchange at position 15. Deletion of the PUA domain causes a 4-fold drop in the observed kcat (2.8 × 10–3 s–1) and results in a 75-fold increased Km for tRNAAsp(1.2 × 10–5 m) compared with full-length TGT. Mutations in tRNAAsp altering or abolishing interactions with the PUA domain can compete with wild-type tRNAAsp for binding to full-length and truncated TGT enzymes. Whereas the C-terminal domains do not appear to play a role in selection of the modification site, their relevance for enzyme function and their role in vivo remains to be discovered.

Transfer RNA from all three kingdoms of life contains a unique series of highly modified nucleosides derived from 7-deazaguanine (1). In eukaryotes and bacteria, the modified nucleoside queuosine (Q) 2 (Fig. 1) occurs in position 34 in tRNAs with a GUN anticodon (tRNA Asp , tRNA Asn , tRNA His , and tRNA Tyr ) (reviewed in Refs. 2 and 3). The replacement of G by Q in the first anticodon position affects tRNA efficiency in codon recognition during translation (4,5). The 7-deazaguanine derivative, archaeosine ( Fig. 1), is found at position 15 in most archaeal tRNAs (6 -8); this position is often involved in long range tertiary base interactions with C48 (9). Archaeosine at this position may therefore strengthen this tertiary interaction and further stabilize the overall tRNA structure.
The biosynthesis of these hypermodified 7-deazaguanine derivatives begins with GTP and requires the action of several enzymes only recently uncovered (10,11). Through the action of the tRNA-guanine transglycosylase (TGT) enzyme, guanine is replaced with a 7-deazaguanine precursor, forming preQ 0 at position 15 in archaeal or preQ 1 at position 34 in bacterial tRNAs (Fig. 1). Once incorporated into the tRNA, these precursors are further modified by additional enzymes to form the mature archaeosine or queuosine bases. Only one of the enzymes acting downstream of TGT in the bacterial pathway has been identified to date (12), and its crystal structure was recently solved (13). Once the mature Q base is formed, further modifications occur such as glycosylation in eukaryotes (14 -17) and even glutamylation by the glutamyl-tRNA synthetase paralog, YadB, present in many bacteria (18 -22).
The TGT enzyme in the bacterial pathway to Q formation (QueTGT) requires its substrate G34 to be flanked by U in positions 33 (conserved in all tRNAs) and 35 (the second base of the anticodon) (23)(24)(25). This requirement explains the presence of Q in the subset of tRNAs containing a 5Ј-34 GUN 36 -3Ј anticodon sequence (9). The first archaeal TGT enzyme (ArcTGT) was purified from Haloferax volcanii (26); it incorporates preQ 0 or guanine into tRNA at position 15. Subsequently, the recombinant enzymes from Methanocaldococcus jannaschii (27) and Pyrococcus horikoshii (28) were shown to have similar properties. In contrast to QueTGT, which requires three nucleotides for proper tRNA recognition, the P. horikoshii enzyme was shown to need only G15 as a tRNA identity element (28). Most archaeal tRNA genes known to date contain G15 (7,29), and half of the archaeal tRNAs whose post-transcriptional modifications have been determined contain archaeosine at this position (6,7).
The archaeal and bacterial TGTs differ in size and domain architecture. QueTGT is only 61% of the archaeal enzyme's size, due to the addition of three small domains to the C terminus of the ArcTGT catalytic core ( Fig. 2A, C1-C3). The structure of the ArcTGT catalytic domain is superimposable upon that of QueTGT from Zymomonas mobilis with a main chain root mean square deviation of ϳ1.4 Å (30). The C3 domain, proximal to the ArcTGT carboxyl terminus ( Fig. 2A), was identified as a potential RNA binding domain based on sequence analysis (31). This domain is present in TGT proteins from all branches of the archaeal domain; it was named PUA due to its occurrence in several pseudouridine synthases and archaeosine-tRNA transglycosylases. An interesting subclass of split TGTs is found in the genomes of at least 11 archaeal organisms including Methanosarcina barkeri. This form is encoded by two distinct ORFs, with the "split" always occurring between domains C1 and C2 ( Fig. 2A; split ArcTGT ).
It was suggested that the ability of the archaeal TGT enzymes to recognize a modification site in the tRNA different from their bacterial counterparts might be attributed to the presence of the C-terminal domains of ArcTGT ( Fig. 2A) (30,32). The crystal structure of P. horikoshii TGT complexed to unmodified tRNA Val (33) revealed the tRNA in an alternate structure (termed the -form). In this form, the tRNA D-loop is extremely distorted, and base pairing along the entire D-stem is disrupted. As a consequence, the normally buried G15 is exposed, and the rearranged D-loop and stem are bound primarily by the catalytic core of the enzyme, whereas canonical structures of the acceptor stem, T-arm, and anticodon arm remain largely intact. Based on this structure, a model for the modification site selection of ArcTGT was proposed that necessitates the use of the C-terminal extension and PUA domain to anchor the acceptor stem of the tRNA (33). With one end anchored, the enzyme would then be able to bind the -form, allowing the positioning of G15 into the active site due simply to its nucleotide distance from the 5Ј-end of the tRNA chain. Whereas this model provides a compelling explanation for the ability of ArcTGT to recognize G15 in tRNAs with an apparent lack of sequence requirements (28), there is little experimental evidence to support the contributions made by the C terminus.
Here we show that the entire C-terminal extension (C1, C2, and C3), including the PUA domain, is dispensable for P. furiosus TGT activity and for proper modification at G15. Thus, the ArcTGT catalytic core, like its bacterial homolog, is sufficient to effectively target a specific site in the tRNA for modification. Cloning-The gene encoding the P. furiosus TGT (PF1046) was PCRamplified from genomic DNA using the Expand Hi Fidelity PCR system (Roche Applied Science). Restriction sites for NcoI and XhoI were introduced at the 5Ј-and 3Ј-ends of the gene, respectively, and the terminal stop codon was deleted, to allow translation of the His 6 tag present in the recipient vector, pET28a (Novagen, Madison, WI). The primers used include the forward primer PFCA3FW (5Ј-CATGCCATGGGCATGCTAAGATTT-GAAATAAAGGAC-3Ј) and reverse primer PFCA3RV (5Ј-GGGATC-CTCGAGCTTTTTCTCCTCTACA-C-3Ј). The PCR product was digested with NcoI and XhoI overnight. The digested fragment was then ligated into pET28a previously digested with NcoI/XhoI and treated with calf intestinal alkaline phosphatase. The sequence of the gene was verified, and the resulting plasmid, pJSPFCA3, was transformed into BL21(DE3) RIL for overexpression.

EXPERIMENTAL PROCEDURES
The truncations of the P. furiosus gene were PCR-amplified as above using pJSPFCA3 as the template. The positions at which the protein was truncated were chosen to be in looped regions separating defined domains. This was based on sequence conservation among archaeal TGT enzymes as well as the existing structural data for the P. horikoshii enzyme (32) with which P. furiosus TGT shares 93% amino acid similarity. The correspond-  ing DNA sequences were amplified using the forward primer (5Ј-GGAAT-TCCATATGCTAAGATTTGAAATAAAGGAC-3Ј) that hybridizes to the same sequence as PFCA3FW above but introduces an NdeI site into the 5Ј-end of the gene. The reverse primers used to determine the end point of the sequence amplified are as follows: amino acids 1-506, 5Ј-CGGGATC-CTCGAGTTATGGATATGGGAGAGTTCGT-3Ј; amino acids 1-424, 5Ј-CGGGTCTCGAGTTAGTCTTCCTCACTCTCGCTTT-3Ј; amino acids 1-363, 5Ј-CGGGATCCTCGAGTTAAGTTATTGGTTCGTA-TTCCT-3Ј. Each reverse primer introduced an in-frame stop codon to terminate translation after the position indicated as well as an XhoI cut site. PCR products were gel-purified, digested with NdeI/XhoI, and further processed as described above. Each fragment was then ligated into pET15b (Novagen), which was previously digested with NdeI/XhoI and treated with calf intestinal alkaline phosphatase. The sequence of the resulting plasmids were verified and transformed into the BL21(DE3) CodonPlus-RIL for overexpression. The free PUA domain (Fig. 2, domain C3) was cloned by PCR amplification using the plasmid pJSPFCA3 as the template. The forward primer (5Ј-CATGCCATGGGCATGAGAGTTGTGGTTAATA-AGGAA-3Ј) along with PFCA3RV (see above) were used to amplify the sequence corresponding to amino acids 508 -585, which was cloned as described above into pET28a.
The M. barkeri MS ORFs encoding TGT were PCR-amplified from genomic DNA using primers designed against the M. barkeri fusaro sequence (accession number NZ_AAAR02000001). The ORFs encoding the N-terminal fragment (accession number ZP_00296567; annotated as queuosine/archaeosine tRNA-ribosyltransferase) as well as the C-terminal fragment (accession number ZP_00297604; annotated as a member of the family of molecular chaperones implicated in de novo protein folding) in the M. barkeri fusaro genome were identified as most similar to the split TGT genes of Nanoarchaeum equitans (35). The forward primer (5Ј-GGGAAT-TCCATATGTCAGCAATATTCGAAAT-3Ј) and the reverse primer (5Ј-CGGGATCCTCGAGTTATTCTTTTTTCCATCTT-3Ј) were used to amplify the N-terminal portion, whereas the forward (5Ј-GGGAATC-CATATGAATAATGATACCGAAG-3Ј) and reverse (5Ј-CGGGATC-CTCGAGCTTTTCTTTTTTCGAAGCTA-3Ј) primers were used to amplify the C-terminal peptide containing the PUA domain. The PCR product corresponding to the N-terminal peptide was cloned into the NdeI/XhoI sites of pET15b, resulting in the addition of a His 6 tag on the N terminus. The PCR product corresponding to the C-terminal peptide was cloned into pET20b after digestion with NdeI and XhoI. This construct appends a His 6 tag to the C terminus of the protein. The sequence of each fragment was compared with that published from M. barkeri fusaro strain. Only 73 and 33 base changes, resulting in 7 and 8 amino acid changes, were noted for the coding sequence of the N-and C-terminal peptides, respectively. The sequence of the N-terminal (accession number DQ104433) and C-terminal (accession number DQ104434) portions of this TGT have been deposited in GenBank TM .
Overexpression and Purification-The E. coli TGT was overexpressed in DH5␣ harboring the pTGT5 vector and purified using ammonium sulfate precipitation and anion exchange chromatography on a Hi-Trap Q-Sepharose column (Amersham Biosciences) essentially as described (36). Overnight cultures of cells harboring plasmid pJSPFCA3 expressing a C-terminally His 6 -tagged full-length P. furiosus TGT were diluted 20-fold into Luria broth containing 100 g/ml ampicillin and 34 g/ml chloramphenicol. Cultures were shaken (250 rpm) at 37°C until reaching A 600 ϭ 1. Expression was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and the ampicillin level was refreshed to 100 g/ml. Induction was allowed to continue for 3 h, and cells were pelleted by centrifugation at 2700 ϫ g for 15 min and then stored at Ϫ80°C. The frozen pellets were thawed on ice for 30 min and resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 10% glycerol) containing 5 mM 2-mercaptoethanol and 0.1 mM phenylmethylsulfonyl fluoride. Lysozyme was then added to 1 mg/ml. After incubation on ice for 30 min, the cell suspension was sonicated (Branson Ultrasonics, Danbury, CT) five times for 30 s each at 50% duty cycle, 60% output. The lysate was cleared by centrifugation at 20,000 ϫ g for 90 min. The entire protein extract was incubated at 60°C for 30 min, and the flocculated proteins were removed by centrifugation at 8000 ϫ g for 5 min. To the supernatant, imidazole was added to 10 mM as well as 1 ml of Ni 2ϩ -nitrilotriacetic acid-agarose slurry (50% ethanol, 50% matrix). The subsequent purification steps were performed according to the manufacturer's recommendations.
Cell growth, overexpression, lysate preparation, and Ni 2ϩ -nitrilotriacetic acid affinity purification of the free PUA domain, the N-terminally His 6 -tagged, truncated P. furiosus enzymes, and the N-terminal peptide of the M. barkeri enzyme was performed as described above. The mutant enzymes, however, precipitate during flocculation; thus, this step was omitted. Growth of the strain overexpressing the C-terminal portion of M. barkeri TGT was carried out as described with the following alterations. Upon reaching A 600 ϳ1, ethanol was added to a final content of 2% (w/v), and the cell culture was shifted from 37 to 15°C and allowed to acclimate for 10 min. After this incubation, protein production was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside, and the ampicillin was refreshed as outlined above. The cultures were incubated for an additional 6 h at 15°C before the cells were harvested. The remaining steps were performed identically to those described above for the truncated enzymes.
Mutant enzymes used in kinetic analyses were further purified by anion exchange chromatography on a HiTrap Q-Sepharose (5 ml) column (Amersham Biosciences). Before being loaded onto the column, the Ni 2ϩ -nitrilotriacetic acid-purified samples were diluted 10-fold into start buffer (50 mM Tris-HCl, pH 7.5, 5 mM 2-mercaptoethanol, 10% glycerol); proteins were eluted with a gradient from 0 to 1 M NaCl. The majority of the active enzyme was found in the flow-through, which was concentrated and dialyzed into 50 mM HEPES, pH 7.0, 5 mM 2-mercaptoethanol, 10% glycerol. The sample was subsequently loaded onto a MonoS HR 5/5 (Amersham Biosciences) column equilibrated in the same buffer. The TGT enzyme was eluted with a gradient of 0 -1 M NaCl in the start buffer. The resulting protein was Ͼ90% pure as judged by SDS-PAGE and Coomassie Brilliant Blue staining. Enzyme-containing fractions were pooled, concentrated, and dialyzed against 50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 5 mM 2-mercaptoethanol, and 10% glycerol and stored at 4°C.
The fraction of active molecules in each sample of truncated TGT was estimated based on RNA binding ability using a filter-binding assay carried out under conditions described below. In this assay, a constant concentration of each mutant protein (15 M) measured by the Bradford method (37) was incubated with increasing concentrations of tRNA Asp transcripts (up to 20 M) isotopically labeled by the addition of 32 P-labeled tRNA Asp . The fraction of protein bound to tRNA at saturation was normalized against that for the full-length enzyme (data not shown) and used to correct the protein concentration obtained by Bradford assay for the presence of inactive protein in the sample.
Preparation of Unfractionated Q-deficient tRNA-Unfractionated tRNA lacking the Q base (34) was prepared from strain JE7350 by phenol/chloroform extraction and isopropyl alcohol precipitation as described (38). Additional workup using a Qiagen-tip 100 column (39) ensured the removal of any contaminating DNA and small RNAs. The resulting tRNA could be modified to a level representing 40 and 10% of The Role of C-terminal Domains in Archaeal TGT Activity the total tRNA by the wild-type P. furiosus and E. coli TGTs, respectively (data not shown).
In Vitro Transcription and Purification of E. coli tRNA Asp -Construction of the plasmid containing the T7 RNA polymerase promoter upstream of the sequence for E. coli tRNA Asp has been described previously (22). The correct end for run off transcription with T7 RNA polymerase was generated by digestion with BstNI overnight at 55°C, and the tRNA gene was transcribed according to published procedures (40). The tRNA transcripts were purified from the reaction mixture by anion exchange chromatography essentially as described (41) using an Amersham Biosciences ⌬kta FPLC system and MonoQ HR 5/5 column (Amersham Biosciences). Fractions containing full-length tRNA were pooled, ethanol-precipitated, and resuspended in water. On average, transcripts prepared in this manner could be modified, using the activity assay described below, to levels reaching 90 -95% of the predicted maximum based on absorbance at 260 nm (data not shown).
The in vitro transcription cassette for E. coli tRNA Asp lacking the 3Ј bases 73-76 comprising G73 at the discriminator base position and the 5Ј-C 74 C 75 A 76 -3Ј end ( Fig. 4; ⌬GCCA) was constructed by annealing two DNA oligonucleotides (5Ј-GATCCTAATACGACTCACTATAGG-AGCGGTAGTTCAGTCGGTTAGAATACCTGCCTGTCACGCA-GGGGGTCGCGGGTTCGAGTCCCGTCCGTTCCGTTTAAACT-CATCCA-3Ј and 5Ј-AGCTTGGATGAGTTTAAACGGAACGGAC-GGGACTCGAACCCGCGACCCCCTGCGTGACAGGCAGGTAT-TCTAACCGACTGAACTACCGCTCCTATAGTGAGTCGTATTA-G-3Ј) containing the tRNA gene and a downstream FokI recognition site (underlined) positioned to allow cleavage of the template strand immediately 5Ј to the template base for C72 (boldface type). The annealed duplex DNA was directly ligated into pUC19 digested with BamHI/HindIII. The ligated plasmid was transformed into DH5␣ for amplification, and the sequence was verified. In order to generate the proper 3Ј-end of the template, prior to run-off transcription, the plasmid was digested overnight at 37°C using FokI. Subsequent steps were identical to those described above.
Modified Guanine Exchange Assay-The assay used in this work is modified from previously published methods (36,42). Incorporation of  Table 2. Each reaction was started by the addition of enzyme, and a thin layer of mineral oil was added to prevent evaporation over the course of the experiment. Aliquots of 15 to 60 l (as noted) were removed from the reaction at the appropriate time and spotted onto glass fiber filters (Whatman GF/C) that had been prespotted and dried with 10% trichloroacetic acid plus 0.05 mM cold guanine. After drying for 5 min at room temperature, each spotted filter was placed into ice-cold 10% trichloroacetic acid plus 0.05 mM guanine solution (200 ml) until completion of the time course. All filters were sequentially added to this trichloroacetic acid solution and washed en masse for 10 min at 4°C on a rotary shaker set to 225 rpm. A second wash step was carried out with 5% trichloroacetic acid plus 0.025 mM cold guanine for an additional 10 min at 4°C. The filters were rinsed with 95% ethanol and dried at 72°C for 10 min. The quantity of [ 14 C]guanine incorporated into acidprecipitable material on each filter was measured by liquid scintillation counting and provided results equivalent to those obtained with the pub- respectively. Reactions were extracted with 1 volume of citrate-buffered phenol (pH 4.5) and then 1 volume of chloroform and were ethanolprecipitated. The pellet was resuspended in 10 l of 50 mM Tris-HCl, pH 7.5, and RNase A (Sigma) was added to 1 mg/ml. After 30 min at 37°C, calf intestinal alkaline phosphatase was added (5 units), and the sample was incubated for an additional 30 min at the same temperature. Following calf intestinal alkaline phosphatase treatment, the sample was dried down under vacuum and resuspended in water, and 1-2 l (corresponding to roughly 2000 cpm) were spotted onto chromatography paper (Whatman No. 1; 23 ϫ 45 cm) and allowed to dry. The samples were developed with 2-propanol/ammonium hydroxide/water (7:1:2) for 15 h in a sealed, descending paper chromatography tank (30 ϫ 30 ϫ 60 cm) pre-equilibrated with the solvent. In that time, the solvent front traveled roughly 30 cm. The radiolabeled spots were visualized by a PhosphorImager, and the R F values for GpU (0.73) and ApGpU (0.51) were calculated relative to the migration of a free guanine standard using the ImageQuaNT software package (Amersham Biosciences).
Measurement of tRNA Binding-The affinity of the TGT variants for a tRNA Asp transcript was measured using the filter-binding method described previously (43,44). The E. coli tRNA Asp transcript was 3Јlabeled using the CCA-nucleotidyl transferase enzyme from E. coli as previously described (45). The radiolabeled tRNA was purified away from unincorporated [␣-32 P]ATP and other reaction components by two rounds of gel filtration on a BioSpin 30 Tris column (Bio-Rad) as well as phenol/chloroform extraction. The resulting sample was precipitated with ethanol and resuspended in water.
Dissociation constants for wild-type tRNA Asp binding to TGT variants were determined by incubating increasing concentrations of each protein on ice in 10 l of binding buffer (50 mM HEPES, pH 7.5, 50 mM NaCl, 5% glycerol) for 30 min after the addition of tRNA Asp transcript to 4.5 nM (3 ϫ 10 3 cpm [A76-␣-32 P]tRNA Asp ). A 96-well vacuum manifold (Hybri-dot 96; Whatman Biometra, Germany) was used to spot aliquots of the binding reaction onto an upper nitrocellulose membrane (MF-Millipore; Millipore Corp., Billerica, MA) and a lower nylon membrane (Hybond-N ϩ ; Amersham Biosciences) previously soaked in binding buffer at 4°C. Aliquots of 3 l from each binding reaction were spotted in triplicate and washed with 100 l of ice-cold binding buffer. The levels of radiolabeled tRNA on each filter were quantified by phosphorimaging and used to determine the ratio of RNA bound to RNA total after correction for nonspecific binding to the nitrocellulose (44).
Determination of the K d for several mutant tRNA Asp transcripts was performed in a competitive binding reaction. Radiolabeled tRNA Asp transcript (40 -50 nM), in the presence of increasing concentrations of unlabeled mutant transcript (0 -5 M), was incubated in 10 l of binding buffer on ice with full-length or truncated P. furiosus TGTs. The concentration of each enzyme in these reactions was chosen based on K d measurements to result in Ͻ50% binding of labeled tRNA in the absence of unlabeled inhibitor. After 30 min on ice, aliquots of 8 l were spotted in duplicate as described above. The fraction of labeled wildtype tRNA Asp that remained bound in each sample was quantified as above and corrected for nonspecific binding. The plot of fraction bound versus concentration of inhibitor added was fit to the following equation, where f represents the fraction of labeled tRNA bound at a given concentration of inhibitor tRNA (I), and f 0 is the fraction of tRNA bound in the absence of inhibitor.

The C-terminal Extension of ArcTGT Is Not Required for Activity-
Initial tests of ArcTGT activity revealed that the modification site selection for this enzyme was different from its well studied bacterial counterpart (26). The chemistry involved in guanine exchange by either enzyme is similar due to a highly homologous catalytic core domain ( Fig. 2A). To investigate the role played by the C-terminal extension, several successive C-terminal truncations of the P. furiosus TGT enzyme were produced (Fig. 2B) and analyzed using the guanine exchange assay. The results indicate that, compared with the full-length enzyme (Fig. 3A, open circles), the truncated forms (Fig. 3A, solid lines) possess a reduced level of activity similar to that of E. coli TGT, although the temperature of the assay for the mesophilic and thermophilic enzymes were different (37 versus 65°C).
We then attempted to reconstitute the activity of the ⌬PUA enzyme by the addition of a 2-fold excess of the isolated PUA domain. The significant restoration of activity in this experiment (Fig. 3B) suggests that the domain architecture of the TGT enzymes is highly modular. The observed activity was greater than that of the full-length P. furiosus TGT (data not shown). In order to test a naturally split enzyme, we cloned the two separate genomic ORFs encoding the N-and C-terminal portions of the M. barkeri MS TGT and purified the proteins separately. Whereas each protein by itself was inactive, activity was restored when the proteins were brought together in the reaction (Fig. 3B, inset).
The TGT Core Domain Correctly Selects the Modification Site-As discussed above, E. coli TGT consists solely of the catalytic core domain and discriminates against tRNAs that lack the 5Ј-U 33 G 34 U 35 -3Ј sequence in the anticodon loop. On the other hand, ArcTGT is less selective and modifies any tRNA containing G at position 15. Although the model proposed for ArcTGT recognition of G15 provides a persuasive argument for the function of the C-terminal extension (33), our observation that the truncated TGT enzymes were active (Fig. 3A) made an examination of the modification site obligatory.
For this experiment, E. coli tRNA Asp was labeled with [ 14 C]guanine by base exchange with full-length and truncated TGTs. An RNase A digest of the radioactive tRNA Asp should indicate conclusively the site of trans-glycosylation, since a radioactive G15 will be found in the trinucleotide ApGpUp, whereas labeled G34 will be contained in the dinucleotide GpUp (Fig. 5A, outlined ). After RNase A digestion, followed by phosphomonoesterase treatment, the oligonucleotide mixture was separated by descending paper chromatography. The results show that full-length or truncated ArcTGT enzymes act only on G15, whereas the E. coli enzyme is specific for G34 (Fig. 5B, top). Thus, deleting up to 222 amino acids (corresponding to domains C1, C2, and C3) in ArcTGT has no measurable effect on the enzyme's ability to select the correct modification site in tRNA.
The crystal structure of P. horikoshii TGT complexed to a tRNA Val transcript (33) revealed extensive interactions of the tRNA ACCA 3Ј terminus with the PUA domain ( Fig. 2A, C3 domain). To determine the importance of these bases for TGT specificity, a tRNA Asp transcript was constructed lacking these nucleotides ( Fig. 4; ⌬GCCA). This tRNA was also labeled with [ 14 C]guanine and then digested with RNase A. The results (Fig. 5B, bottom)  indicate that both full-length and truncated enzymes modify this mutant tRNA at position 15. Thus, the 3Ј terminus of tRNA Asp is not required for correct selection of the site for guanine exchange.
tRNA Binding Is Only Moderately Affected by Deletion of PUA and C-terminal Domains-Although the archaea-specific extension may not be involved in selection of the modification site, it might be important for the overall reaction. We therefore measured the ability of the truncated P. furiosus TGT proteins to bind a tRNA Asp transcript in a double filter-binding assay (44, 46) (see "Experimental Procedures"). Progressive deletion of one (⌬PUA), two (CA1), or all three (CA) of the C-terminal domains resulted in a 1-3-fold K d change when compared with full-length ArcTGT (WT in Table 1). The largest effect was seen for the CA1 truncation (3-fold increased K d ), which may result from a higher degree of misfolding of the C1 domain at its C terminus. This domain is thought to provide a connection between the catalytic core and remaining domains and therefore may not be able to fold properly under these binding conditions. In addition, these experiments illustrate the first biochemical measurement of the ability of the PUA domain to bind RNA (Fig. 6). This 79-amino acid protein folds into an autonomous domain capable of binding tRNA Asp with a 0.2 M affinity ( Table 1, Fig. 6B).
Mutant tRNA Transcripts Can Compete with Wild-type tRNA for Binding to TGTs-The ability of mutant tRNA transcripts to compete for binding to ArcTGT variants is consistent with the notion that binding is not dependent on interactions between the enzyme and the end of the tRNA acceptor stem. The K d for each tRNA Asp mutant was within 3-4-fold of the K d measured above for a wild-type transcript (Table 1). Compared with the complete tRNA transcript, the absence of the 3Ј-terminal GCCA sequence in the ⌬GCCA tRNA Asp variant led to a nearly 2-fold decrease in K d with full-length ArcTGT (Table 1). Another mutant tRNA lacking the single-stranded CCA overhang to which an additional G-C base pair was added at the top of the acceptor stem (5PBP in Fig. 4) was assayed for its ability to compete for binding to this set of proteins. It was evident from the co-crystal structure that such an addition would clash sterically with residues in the PUA domain upon binding (33). However, the 5PBP tRNA binds with only a 2-3-fold decreased affinity to ArcTGT as well as to the isolated PUA domain (5PBP in Table 1). This illustrates that perturbing the interactions between the tRNA and PUA domain produces only moderate effects.
The presence of G15 is known to be the single most important determinant for ArcTGT activity, since a G15A mutation in a tRNA Val transcript resulted in a Ͼ5000-fold loss of catalytic efficiency (28). Our work demonstrates that all TGT variants are able to bind this substrate with comparable affinity to wild-type tRNA Asp (G15A in Table 1).
The PUA Domain Contributes Significantly to ArcTGT Efficiency-The role of this domain in ArcTGT activity was further analyzed by steady-state kinetics comparing the ⌬PUA truncation ( Fig. 2B) with full-length P. furiosus and E. coli TGTs. Deletion of the PUA domain from the full-length enzyme (⌬PUA) has marked effects on enzyme activity toward tRNA Asp and results in a 300-fold reduction in catalytic efficiency compared with wild-type ArcTGT ( Table 2, column 4). The most significant contribution to this drop comes from a 75-fold increase in K m for its tRNA substrate ( Table 2, column 3). The reaction rate observed was reduced nearly 4-fold compared with the wild type enzyme under the experimental conditions employed ( Table 2, column 2).
The kinetic parameters were also determined using the ⌬GCCA tRNA Asp as a substrate. Deleting these residues at the 3Ј-end of the tRNA had little effect on the kinetic parameters determined for the ⌬PUA enzyme, since these nucleotides are recognized by the protein  domain missing from this truncation (33). The K m value determined for full-length ArcTGT increased 2-fold (Table 2; compare columns 3 and  5), mirroring the effects of this tRNA mutation on binding affinity measured above ( Table 1). The loss of interactions between the PUA domain of ArcTGT and the tRNA in this case, however, cannot account for the 75-fold increase in K m observed after deletion of the PUA domain from ArcTGT (⌬PUA in Table 2).

DISCUSSION
Previous biochemical studies of ArcTGT from P. horikoshii using an in vitro transcribed tRNA Val substrate indicated that the enzyme was capable of modifying this transcript at position 15 regardless of the surrounding nucleotide sequence (28). Furthermore, it was determined that G15 was the only sequence element essential for modification by the enzyme. The crystal structure of this enzyme in complex with a tRNA substrate provided a possible explanation of how recognition could be accomplished. In this structure, the majority of the RNA/protein interactions are mediated through the phosphate backbone, bound by positively charged amino acids on the surface. Bases bound by the enzyme were accommodated in nonspecific hydrophobic pockets. The model proposed necessitated the recognition of an alternate -form tRNA, mediated primarily by the C-terminal domains of the enzyme (33). The precise positioning of G15 into the catalytic site was suggested to proceed by a counting mechanism, which begins with the PUA domain anchoring the tRNA acceptor stem and 3Ј overhang. With the end of the tRNA thus located, the nucleotide distance from 1 to 15 was thought to be measured via contacts with the phosphate backbone and nonspecific hydrophobic accommodation of the bases themselves.
This model necessitates the presence of the counting surface provided by the C-terminal domains. In light of the evidence presented here, it may need reevaluation. Our results demonstrate that these domains and their interactions are not required for TGT activity, since truncated enzymes can still incorporate guanine to appreciable levels (Fig. 3). Even in the absence of one or all of the C-terminal domains, the catalytic core of the ArcTGT is able to locate and modify the G at position 15 (Fig. 5B, top). Proper positioning of the acceptor stem and 3Ј nucleotides by the C-terminal domains is not a prerequisite for correct modification, since eliminating these interactions has little effect (Fig.  5B, bottom).
ArcTGT Binding to tRNA Is Mediated by the Catalytic Core-The C-terminal domains, although not contributing to modification site selection, make many interactions with the tRNA, as shown in the structure (33). Under the conditions tested, the presence or absence of these domains was found to have only modest effects on binding the tRNA substrate (only 2-3-fold). In isolation, this domain has an affinity for tRNA Asp comparable with the enzyme as a whole (Table 1). In addition, the introduction of mutations designed to interfere with (5PBP in Fig. 4) or eliminate (⌬GCCA in Fig. 4) interactions with the PUA domain changes K d values only slightly (Table 1).
Nonsubstrate tRNAs Can Efficiently Compete for Binding to ArcTGT-Currently, the only anti-determinant for ArcTGT activity is having a base other than G at position 15. tRNAs mutated at this position were shown to display Ͼ5000-fold reduction in catalytic efficiency (28). Our results show that the catalytic core domain, in the presence or absence of any number of C-terminal domains, is still capable of binding to a G15A mutant tRNA Asp with an affinity comparable with the wildtype transcript (G15A in Table 1). This type of inherent tRNA affinity has been noted previously for pseudouridine synthetase Pus1 in yeast. In this case, it was shown that the difference in affinity between a substrate and a nonsubstrate tRNA was minimal, whereas affinity for different cognate tRNAs could vary by 1 order of magnitude (47). Thus, under the   experimental conditions employed, ArcTGT does not distinguish between substrate and nonsubstrate tRNAs at the level of binding. This is in contrast to QueTGT from Z. mobilis and E. coli, which do not bind to tRNAs carrying the equivalent G34A or any other mutation in its recognition sequence (25,48,49). The PUA Domain Accelerates the Reaction Rate but Primarily Influences the K m -The effect on the catalytic efficiency caused by deleting the PUA domain is substantial (a 300-fold decrease), based primarily on a 75-fold increase in the K m . The resulting 12 M value may not preclude such an enzyme from a possible in vivo function as may be the case for the split ArcTGTs; for instance, E. coli SerRS aminoacylates selenocysteine tRNA with a comparable K m (50).
The moderate decrease in the k cat value of the ⌬PUA mutant results in a catalytic rate similar to that of E. coli TGT as measured here (2.5 ϫ 10 Ϫ3 s Ϫ1 ) and by others (51). It is also comparable with that reported for pseudouridine synthetases from yeast (47) and E. coli (52,53), which also catalyze a transglycosylation step during their reaction. In general, all TGTs from archaea and bacteria, whether full-length or truncated have a slow rate of catalysis compared with other enzymes that act on tRNA, such as the M. barkeri seryl-tRNA synthetase, which can catalyze the formation of seryl-tRNA Ser at a rate as fast as 1-4 s Ϫ1 (54).
From these data, it is clear that whereas deletion of the PUA domain may not significantly affect tRNA binding (Table 1), it substantially reduces the catalytic efficiency of transglycosylation (Table 2, column 4) through a significant increase in K m for its tRNA substrate. Taken together, the differing contributions of the PUA domain to the kinetics and binding suggest that, perhaps, it is during the steps following association, leading up to and including catalysis, where the influence of the C-terminal domains is most significant for the archaeal enzyme.
Why Are There Split Forms of the Archaeal Enzyme?-The presence of a split form of ArcTGT in a few organisms may represent an interesting evolutionary intermediate on the way toward the full-length enzyme or an adaptation of the ArcTGTs to a specific niche. In addition, the separately encoded catalytic and RNA binding domains found in M. barkeri may enable this organism to tune the levels of archaeosine modification by independently modulating the expression of the C-terminal portions, which are then capable of enhancing the activity of the Nterminal catalytic domain (Fig. 3B). These organisms would therefore be able to balance the energetic costs of producing these proteins (as well as the consumption of GTP in the biosynthesis of archaeosine precursor preQ 0 ) with the requirement for the presence of archaeosine in the tRNA. In environments where nutrients are limited, the ability to conserve amino acid and energy stores is essential for survival. Alternatively, in these organisms, the tRNA binding properties of the Cterminal fragment could be recruited to serve another role in tRNA processing in addition to archaeosine modification.