Genetic and Biochemical Analysis of the Functional Domains of Yeast tRNA Ligase*

Yeast tRNA ligase (Trl1) converts cleaved tRNA half-molecules into spliced tRNAs containing a 2′-PO4, 3′-5′ phosphodiester at the splice junction. Trl1 performs three reactions: (i) the 2′,3′-cyclic phosphate of the proximal fragment is hydrolyzed to a 3′-OH, 2′-PO4 by a cyclic phosphodiesterase (CPD); (ii) the 5′-OH of the distal fragment is phosphorylated by an NTP-dependent polynucleotide kinase; and (iii) the 3′-OH, 2′-PO4, and 5′-PO4 ends are sealed by an ATP-dependent RNA ligase. Trl1 consists of an N-terminal adenylyltransferase domain that resembles T4 RNA ligase 1, a central domain that resembles T4 polynucleotide kinase, and a C-terminal CPD domain that resembles the 2H phosphotransferase enzyme superfamily. Here we show that all three domains are essential in vivo, although they need not be linked in the same polypeptide. We identify five amino acids in the adenylyltransferase domain (Lys114, Glu266, Gly267, Lys284, and Lys286) that are essential for Trl1 activity and are located within motifs I (114KANG117), IV (266EGFVI270), and V (282FFKIK286) that comprise the active sites of DNA ligases, RNA capping enzymes, and T4 RNA ligases 1 and 2. Mutations K404A and T405A in the P-loop (401GXGKT405) of the central kinase-like domain had no effect on Trl1 function in vivo. The K404A and T405A mutations eliminated ATP-dependent kinase activity but preserved GTP-dependent kinase activity. A double alanine mutant in the P-loop was lethal in vivo and abolished GTP-dependent kinase activity. These results suggest that GTP is the physiological substrate and that the Trl1 kinase has a single NTP binding site of which the P-loop is a component. Two other mutations in the central domain were lethal in vivo and either abolished (D425A) or severely reduced (R511A) GTP-dependent RNA kinase activity in vitro. Mutations of the signature histidines of the CPD domain were either lethal (H777A) or conferred a ts growth phenotype (H673A).

Intron-containing tRNAs are widespread in the archaeal and eukaryal domains of the universal phylogenetic tree (1). The intron is usually located in the anticodon loop of the pre-tRNA and must be removed precisely for the tRNA to function in protein synthesis. tRNA splicing occurs in two stages: (i) intron excision and (ii) joining of the broken tRNA halves (2, 3) (Fig.  1). Unlike pre-mRNA splicing in eukarya, which relies on ribonucleoprotein catalysts, all of the reactions of the tRNA splic-ing pathway are performed by protein enzymes. The intron removal phase of tRNA splicing requires two incisions of the pre-tRNA at the exon-intron borders. The chemistry of the reaction entails breakage of the phosphodiester backbone by transesterification to yield 2Ј,3Ј-cyclic phosphate and 5Ј-OH termini at both incision sites (2). The breakage reactions are catalyzed by a tRNA splicing endonuclease that specifically recognizes the fold of the pre-tRNA (4 -10). The specificity and fidelity of tRNA splicing are largely governed by the endonuclease component, which is conserved in structure and mechanism among archaea and lower and higher eukaryal species (4 -10).
The joining phase of the tRNA splicing pathway has been studied most extensively in yeast, where a single multifunctional tRNA ligase enzyme (Trl1) catalyzes a series of chemical transformations at the ends of the broken tRNA half-molecules that eventuates in the formation of a ligated tRNA molecule containing a 2Ј-PO 4 , 3Ј-5Ј phosphodiester structure at the junction (3,11). Trl1 performs three reactions: (i) the 2Ј,3Ј-cyclic phosphate terminus is hydrolyzed to a 3Ј-OH, 2Ј-PO 4 terminus by a 2Ј,3Ј-cyclic phosphodiesterase (CPD) 1 activity; (ii) the 5Ј-OH terminus is phosphorylated by an NTP-dependent polynucleotide kinase activity; and (iii) the resulting 3Ј-OH, 2Ј-PO 4 , and 5Ј-PO 4 ends are sealed by an ATP-dependent RNA ligase activity (11)(12)(13)(14)(15)(16). The mechanism of the ligase component of yeast tRNA ligase resembles that of bacteriophage T4 RNA ligase, whereby RNA joining entails three nucleotidyl transfer steps: (i) ligase reacts with ATP to form a covalent ligase-(lysyl-N)-AMP intermediate plus pyrophosphate; (ii) AMP is transferred from ligase-adenylate to the 5Ј-PO 4 RNA end to form an RNA-adenylate intermediate (AppRNA); and (iii) ligase catalyzes attack by an RNA 3Ј-OH on the RNA-adenylate to seal the two ends via a phosphodiester bond and release AMP (11,12,(17)(18)(19)(20). Trl1 is capable of modifying and/or ligating the ends of artificial RNA substrates such as oligo(A) 16 , oligo(U n G), 5Ј-(AUCUCG) n AUCUCG, and 5Ј-GGGCGAAUU (11,14,15), which implies that Trl1 activity is not obligately linked to the processing of tRNAs. The final step in yeast tRNA splicing is the removal of the 2Ј-PO 4 at the splice junction by the essential 2Ј-phosphotransferase Tpt1, which catalyzes the transfer of the tRNA 2Ј-PO 4 to NAD ϩ to form ADP-ribose 1Љ-2Љ cyclic phosphate (21)(22)(23) (Fig. 1).
We are interested in the end-remodeling and strand-joining steps of tRNA splicing, which exemplify a general RNA repair pathway entailing the healing and sealing of broken RNA ends. Another well characterized example of RNA repair is the tRNA restriction/repair phenomenon elicited by infection of Escherichia coli prr strains with bacteriophage T4 (24,25). T4 infec-tion activates a latent anticodon nuclease (PrrC) encoded by the host bacterium. PrrC specifically incises tRNA Lys at a single site in the anticodon loop. Depletion of tRNA Lys blocks phage protein synthesis and arrests the infection before it can spread. However, the T4 enzymes polynucleotide kinase and RNA ligase 1 repair the broken tRNAs and thereby thwart the host defense mechanism. The enzymatic steps in tRNA splicing and tRNA restriction/repair are generally similar. The incision steps in both cases result in the formation of 2Ј,3Ј-cyclic phosphate and 5Ј-OH termini (25). tRNA splicing requires two breaks in the backbone of the pre-tRNA to excise the intron, whereas tRNA restriction involves a single break in the mature tRNA. The healing and sealing steps of the phage-encoded tRNA repair pathway are performed by two separate enzymes (Pnk and Rnl1), whereas a single enzyme Trl1 performs these functions in yeast tRNA splicing. The structural basis for the T4 RNA repair reactions has been illuminated recently by extensive site-directed mutagenesis and x-ray crystallography (26 -30). In contrast, the yeast Trl1 has received little attention during the past decade.
Renewed interest in yeast tRNA ligase is warranted for two reasons. First, the discovery that yeast tRNA ligase is responsible for nonspliceosomal splicing of mRNA in the unfolded protein response pathway (31, 32) extends the RNA repair paradigm to mRNA metabolism and holds out the prospect that there are other instances (yet to be discovered) in which RNA primary structure is altered by breakage and repair. Second, the phylogenetic distribution of Trl1-like proteins is surprisingly narrow, given the wide occurrence of tRNA introns. Trl1 homologs are found in several genera of fungi, including Saccharomyces, Candida, Schizosaccharomyces, and Aspergillus (32) (Fig. 2). Trl1-like proteins are absent from the proteomes of archea and nonfungal eukarya. This may be because archaea and nonfungal eukarya use a different end-joining mechanism for tRNA splicing than do fungi (34 -36). Indeed, there is evidence that mammals have two different pathways of tRNA splicing, a yeast-like mechanism yielding a 2Ј-phosphate at the splice junction and a second pathway in which the 2Ј,3Ј-cyclic phosphate of the incised pre-tRNA is retained as the 3Ј-5Ј phosphodiester of the spliced tRNA product (37). The latter pathway does not require phosphorylation of the 5Ј-OH of the cleaved tRNA (34 -36). Nonetheless, mammalian cells possess a nuclear 5Ј-OH polynucleotide kinase activity that could participate in the yeast-like pathway (38). None of the specific proteins responsible for tRNA ligation in archaea or higher eukarya have been identified, nor have their genes been cloned. A multifunctional Trl1-like enzyme has been purified to near homogeneity from wheat germ (39 -43), but the gene encoding wheat germ RNA ligase has not been identified, and there is no obvious Trl1 homolog in the proteome of the plant Arabidopsis thaliana. This scenario, in which the Trl1-like pathway in metazoa and plants is either redundant or performed by enzymes without recognizable structural similarity to fungal tRNA ligases, recommends Trl1 as an excellent target for antifungal drug discovery.
As a trifunctional enzyme with an essential role in vivo, Trl1 is ripe for structure-function analysis. Soon after the Trl1 gene was cloned, the Greer and Abelson laboratories (13,14) demonstrated an autonomous CPD domain within the C-terminal portion of Trl1 and an autonomous adenylyltransferase domain within the N-terminal portion of Trl1. The site of covalent AMP attachment was mapped to Lys 114 (13), which is located within a conserved sequence motif (KX(D/N)G; motif I) that defines a superfamily of covalent nucleotidyl transferases that includes T4 RNA ligases 1 and 2, DNA ligases, and mRNA capping enzymes (44 -47). Although the kinase function of Trl1 has not been assigned to an autonomous segment, the middle third of the Trl1 polypeptide bears some resemblance to the kinase domain of T4 Pnk (14), especially with respect to the presence of the Walker A-box motif GXGKT (also known as the "P-loop") that comprises the NTP binding site of T4 Pnk and numerous other NTP-dependent phosphotransferases (26 -29) (Fig. 2). The CPD domain of Trl1 resembles the so-called "2H" superfamily of phosphoesterases, which is defined by two copies of a histidine-containing motif HT (where is a hydrophobic residue) (Fig. 2).
Here we address the following outstanding questions. (i) Does Trl1 activity in vivo depend on all three putative domains? (ii) Is the physical linkage of the domains within a single polypeptide essential for Trl1 function in vivo? (ii) Where is the boundary of the adenylyltransferase domain? (iii) Is the limited primary structure similarity between Trl1 and other covalent nucleotidyl transferases functionally important? (iv) Is the limited similarity between Trl1 and T4 Pnk functionally relevant? (v) Are the histidines of the 2H motifs relevant to Trl1 function in vivo? To answer these questions, we established a plasmid shuffle assay to gauge mutational effects on Trl1 function in vivo. Our studies show that all three domains are essential, although they need not be linked in the same polypeptide. We identify five individual residues in the adenylyltransferase domain that are essential for Trl1 activity and conserved in all members of the covalent nucleotidyl transferase family. In contrast, single mutations in the putative NTP binding P-loop of the kinase domain had no effect on Trl1 function in vivo. Our finding that single P-loop mutations severely decreased or eliminated ATP-dependent kinase activity but preserved GTP-dependent kinase activity supports the prior suggestion that GTP is the real substrate for the Trl1 kinase (15,16). However, a double-alanine mutant in the Ploop was lethal in vivo and abolished GTP-dependent kinase activity, suggesting that the Trl1 kinase domain has a single NTP binding site of which the P-loop is a component. We report that alanine mutations of the signature histidines of the CPD domain were either lethal (at His 777 ) or conferred a temperature-sensitive growth phenotype (at His 673 ).

EXPERIMENTAL PROCEDURES
TRL1 Genomic Clone-A DNA fragment containing the TRL1 gene plus 550 bp of 5Ј-flanking DNA and 220 bp of 3Ј-flanking genomic DNA was amplified by PCR using total yeast genomic DNA as template and primers that introduced a SphI site at the 5Ј-end of the amplified fragment and a SacI site at the 3Ј-end. The amplified gene was cloned into pUC18 to generate pUC-TRL1. A BamHI site was introduced immediately 5Ј of the start codon of the TRL1 open reading frame (ORF), which, together with an endogenous BglII site overlying codon 338 and an endogenous SacII site immediately 3Ј of the TRL1 stop codon (TAGACCGCGG), facilitated the cloning of truncated and mutated versions of TRL1 under the control of the natural TRL1 promoter and flanked by the natural TRL1 3Ј-untranslated region.
Deletions and Alanine Mutations of TRL1-Deleted versions of TRL1 were constructed by PCR amplification using oligonucleotide primers that introduced either a start codon and a BamHI site or a stop codon and a SacII site at the desired positions in the TRL1 ORF. The PCR products were digested with BamHI and SacII and inserted into BamHI/SacII-cut pUC-TRL1. Alanine mutations and overlapping diagnostic restriction sites were introduced via the two-stage PCR overlap extension method (65). The mutated PCR products were digested with either BamHI and BglII or BglII and SacI (depending on the location of the mutation) and inserted into BamHI/BglII-cut or BglII/SacI-cut pUC-TRL1 in lieu of the wild-type gene fragment. The inserts were sequenced completely to confirm the presence of the desired mutations and to exclude the acquisition of unwanted coding changes during PCR amplification and cloning.
Plasmids for Expression of TRL1 in Yeast-Wild-type and mutated versions of the TRL1 gene plus 5Ј-and 3Ј-flanking genomic DNA were excised from the respective pUC-TRL1 plasmids with SphI and SacI and inserted into yeast shuttle vectors pSE360 (CEN URA3), pSE358 (CEN TRP1), or pSA360 (CEN ADE2). Selected TRL1 alleles were cloned into the PstI site of pRS413 (CEN HIS3).
Test of Trl1 Function by Plasmid Shuffle-YRS1 was transformed with CEN plasmids bearing wild-type or mutant alleles of TRL1. Transformants were selected on appropriate drop-out media. Two individual colonies were transferred to drop-out agar medium, and cells from each isolate were then streaked on agar containing 0.75 mg/ml 5-FOA. The plates were incubated at 18, 25, 30, and 37°C. Lethal mutations were those that did not allow formation of 5-FOA-resistant colonies after 7 days at any of the temperatures tested. Other mutated alleles supported 5-FOA-resistant colony formation within 4 days at one or more of the growth temperatures. Two individual colonies from each streak were picked from the 5-FOA plate, transferred to yeast extract/peptone/ dextrose (YPD) medium, and then tested for growth on YPD agar at 18, 25, 30, and 37°C.
Recombinant Trl1 Proteins-Cultures (600 ml) of E. coli BL21(DE3)/ pET28-His 10 -TRL1 (wild-type or mutant) were grown at 37°C in Luria-Bertani medium containing 0.06 mg/ml kanamycin until the A 600 reached 0.5. The cultures were adjusted to 0.4 mM isopropyl-␤-D-thiogalactopyroanoside, and incubation was continued at 17°C overnight. Cells were harvested by centrifugation, and the pellet was stored at Ϫ80°C. All subsequent procedures were performed at 4°C. Thawed bacterial pellets were resuspended in 10 ml of buffer A (50 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 10% sucrose) and left on ice for 10 min. The suspensions were then sonicated for 30 s. Lysozyme was added to a final concentration of 50 g/ml, and the suspensions were further incubated on ice for 30 min. Triton X-100 was added to a final concentration of 0.1% and sonication was repeated for 1 min to reduce viscosity. Insoluble material was removed by centrifugation in a Sorvall SS34 rotor at 18,000 rpm for 45 min. The soluble extracts were applied to 1.5-ml columns of Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) equilibrated with buffer A containing 0.1% Triton X-100. The column was washed with the same buffer and then eluted stepwise with buffer B (50 mM Tris-HCl, pH 8.0, 0.1 M NaCl, 10% glycerol) containing 50, 100, 200, 500, and 1000 mM imidazole. The polypeptide compositions of the column fractions were monitored by SDS-PAGE. The recombinant Trl1 proteins were retained on the column and recovered in the 200 mM imidazole eluates. The 200 mM imidazole fraction was dialyzed against buffer C (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM dithiothreitol, 10% glycerol, 0.05% Triton X-100). The Trl1 preparations were stored at Ϫ80°C. Protein concentrations were determined using the Bio-Rad dye binding assay with bovine serum albumin as a standard.
Adenylyltransferase Assay-Reaction mixtures (20 l) containing 50 mM Tris-HCl (pH 8.0), 5 mM dithiothreitol, 10 mM MgCl 2 , 50 M [␣-32 P]ATP (3-5 Ci/mmol), and Trl1 as specified were incubated for 10 min at 30°C. The reactions were quenched with SDS and the products were analyzed by SDS-PAGE. The ligase-[ 32 P]AMP adduct was visualized by autoradiography of the dried gel and quantitated by scanning the gel with a Fujix BAS-2500 phosphor imager.

RESULTS
All Three Domains of Trl1 Are Required for Cell Viability, but They Need Not Be Linked in the Same Polypeptide-Trl1 consists of three modules: an N-terminal adenylyltransferase domain, a central kinase-like domain, and a C-terminal CPD domain (Fig. 3). Although the isolated domains have been characterized in vitro as dihydrofolate reductase fusion proteins (14), their individual functions in vivo have not been formally tested by site-directed mutagenesis, although it is well established that TRL1 is an essential gene. Here we used the plasmid shuffle method to gauge the effects of structural alterations on Trl1 function in vivo. To establish the assay, TRL1 was disrupted in a diploid strain by replacement of the entire coding sequence with a kanMX cassette. The TRL1 trl1::kanMX diploid was transformed with a CEN URA3 plasmid containing TRL1 under the control of its own promoter. The diploid was sporulated, and tetrads were dissected to obtain G418 R trl1⌬ haploids. The trl1⌬ strain was unable to grow on medium containing 5-FOA (a drug that selects against the URA3 TRL1 plasmid) but was able to grow on 5-FOA after being transformed with a CEN TRP1 TRL1 plasmid. 5-FOAresistant trl1⌬ isolates containing the plasmid-borne TRL1 gene grew as well as the parental wild-type yeast strain on rich medium (YPD agar) at all temperatures tested (25,30, and 37°C; scored as ϩϩϩ growth in Fig. 3).
The first question we asked was whether all three modules of Trl1 are required for yeast growth. Therefore, cDNAs encoding the adenylyltransferase domain (aa 1-388), the kinase-like domain (aa 389 -561), and the CPD domain (aa 562-827) were cloned into CEN TRP1, CEN ADE2, and CEN HIS3 plasmids, respectively, under the control of the natural TRL1 promoter and then tested by plasmid shuffle for trl1⌬ complementation. None of the domains per se was able to support cell growth (i.e. no 5-FOA-resistant transformants were recovered during selection at either 25 or 30°C) (scored as Ϫ in Fig. 3). To test if any two modules together sufficed for viability, we constructed a CEN TRP1 plasmid expressing the N-terminal adenylyltransferase kinase domain (aa 1-561) and a CEN ADE2 plasmid expressing the C-terminal kinase-CPD domain (residues 389 -827). Neither construct was able to support growth on 5-FOA (Fig. 3). Cotransformation with plasmids expressing the adenylyltransferase domain on a CEN TRP1 plasmid and the CPD domain on a CEN HIS3 plasmid also failed to complement trl1⌬ (Fig. 3).
The biologically active adenylyltransferase domains Trl1-(1-388) and Trl1-(1-376) and the noncomplementing N-terminal truncations Trl1-(21-388), Trl1-(41-388), and Trl1-(59 -388) were produced in bacteria as His 10 -tagged fusion proteins and then purified from soluble bacterial lysates by nickel-agarose chromatography. SDS-PAGE analysis showed that the preparations were highly enriched with respect to the adenylyltransferase polypeptides and that the extents of purification were similar in each case (Fig. 5A). The sizes of the recombinant proteins and the electrophoretic mobility differences between the truncated versions were consistent with the values calculated by translation of the open reading frames. (His 10 -tagged proteolytic fragments of the adenylyltransferase domain were also detected in each of the affinity-purified protein preparations.) The C-terminal truncation mutants Trl1-(1-362) and Trl1-(1-320) were extensively proteolyzed when expressed in E. coli, and we were unable to purify the intact His-tagged recombinant polypeptides (not shown). Thus, these two deletion mutants were not available for biochemical characterization.
The adenylyltransferase activity of recombinant Trl1-(1-388) was evinced by label transfer from 50 M [␣-32 P]ATP to the polypeptide to form a covalent enzyme-adenylate adduct (Fig. 5B). The reaction was divalent cation-dependent and optimal at 5-20 mM MgCl 2 (not shown). The divalent cation requirement was also satisfied by either 10 mM manganese or calcium; we detected no activity with 10 mM copper or zinc (not shown). The yield of Trl1-(1-388)-[ 32 P]AMP was proportional to input enzyme and optimal at pH 8.0 in 50 mM Tris-HCl buffer (not shown). The C-terminal truncated derivative Trl1-(1-376) was active in autoadenylylation, but the N-terminal truncations Trl1-(21-388), Trl1-(41-388), and Trl1-(59 -388) were apparently inert (Fig. 5B). We conclude that the N-terminal peptide segment of Trl1 is essential for both in vivo function and adenylyltransferase catalytic activity.
Identification of Individual Essential Amino Acids in the Adenylyltransferase Domain-Xu et al. (13) mapped the AMP attachment site of Trl1 to Lys 114 , but there has been no report of the effect of mutating this residue on Trl1 function. The lysine nucleophile of Trl1 is located within a conserved sequence element referred to as motif I ( 114 KX(D/N)G 117 in Fig. 2) that defines a superfamily of covalent nucleotidyl transferases, which includes DNA ligases, RNA ligases, and mRNA capping enzymes (44 -49). DNA ligases and capping enzymes have a common tertiary structure composed of five peptide motifs (I, III, IIIa, IV, and V) that contain essential amino acids responsible for nucleotide binding and catalysis (46 -54). It has been suggested that DNA ligases and capping enzymes evolved from a common ancestral nucleotidyl transferase (55), possibly from an ancient RNA strand-joining enzyme. This model remains speculative, because there is no atomic structure available for a nucleotide-dependent RNA ligase. However, a recent mutational analysis of T4 RNA ligase 2 underscored the conservation of motifs I, III, IIIa, IV, and V and the essentiality of the conserved side chains within these motifs for RNA ligation in vitro (56,57). The adenylyltransferase domain of fungal tRNA ligase has scant similarity to T4 Rnl2, DNA ligases, or mRNA capping enzymes outside of motif I. However, perusal of the aligned primary structures of tRNA ligases from multiple fungal species highlights potential counterparts in Trl1 of nucleotidyl transferase motifs IV ( 266 EG 270 , where is a hydrophobic side chain) and V ( 282 KK 286 ). In order to delineate if any of the conserved amino acids are functionally relevant, we performed an alanine scan of the eight positions of the Trl1 adenylyltransferase domain indicated by a vertical line in Fig.  2. The targeted residues included Lys 114 and Asn 116 in motif I, Glu 266 and Gly 267 in motif IV, and Lys 284 and Lys 286 in motif V. We also mutated Gly 152 and Glu 153 , which are located between motifs I and IV and are conserved among fungal tRNA ligases (Fig. 2).
TRL1-Ala alleles were cloned into a CEN TRP1 plasmid under the control of the native TRL1 promoter and transformed into the trl1⌬ plasmid shuffle strain. The K114A, E266A, G267A, K284A, and K286A transformants failed to give rise to 5-FOA-resistant colonies at 18, 25, 30, or 37°C; thus, these five alanine mutations were lethal in vivo. The N116A, G152A, and E153A alleles supported growth on 5-FOA, and the resulting TRL1-Ala strains grew as well as TRL1 cells on YPD agar at all temperatures (Table I). These results show that that amino acids within putative equivalents of nucleotidyl transferase motifs I, IV, and V are essential for tRNA ligase function in vivo.
The wild-type adenylyltransferase domain Trl1-(1-388) and the K114A, N116A, G152A, E153A, E266A, G267A, K284A, and K286A mutants thereof were expressed in bacteria as His 10 -tagged fusion proteins and then purified from soluble bacterial lysates by Ni 2ϩ -agarose chromatography. SDS-PAGE analysis showed that the preparations were highly enriched with respect to the ϳ48-kDa His-Trl1-(1-388) polypeptide and the extents of purification were similar in each case (Fig. 6A). Whereas recombinant wild-type Trl1-(1-388) reacted with [␣-32 P]ATP to form a covalent enzyme-adenylate adduct, the K114A mutant was inert in ligase adenylation (Fig. 6B), con- sistent with the assignment of this lysine as the active site nucleophile for AMP transfer (13). The concordant loss of adenylyltransferase activity in vitro and Trl1 function in vivo provides evidence that catalysis of the adenylyltransferase reaction is an essential part of the repertoire of Trl1.
Mutants E266A and K284A were severely defective in ligase adenylation (4 and 7% of wild-type activity, respectively), thereby explaining their inability to sustain yeast growth. In contrast, the N116A, G152A, and E153A mutants retained substantial adenylyltransferase activity in vitro (78, 130, and 33% of wild-type activity, respectively), consistent with their apparently normal growth phenotypes in vivo. It was noteworthy that the G267A and K286A mutations had little effect on ligase adenylation (Fig. 6B), despite being lethal in vivo. These results hint that the G267A and K286A changes may affect a downstream component of the three-step RNA ligase pathway.
Effects of Mutations in the Putative Kinase Domain-The central domain of the Trl1 polypeptide resembles the kinase domain of T4 Pnk with respect to the presence of a Walker A-box motif 401 GCGKT 405 (the P-loop) (Fig. 2). The A-box motif of T4 Pnk ( 12 GSGKS 16 ) comprises part of the NTP-binding site, together with a second peptide motif 122 RNSKR 126 located 106-aa downstream of the A-box (28,29). Yeast Trl1 has a similar element 507 RVIKR 511 placed 102 aa downstream of its A-box. The lysine and serine of the A-box and the second arginine of the RXXXR motif are each essential for the T4 polynucleotide kinase reaction (26,27). Here we queried by alanine scanning whether the corresponding side chains in the central kinase domain of Trl1 (Lys 404 , Thr 405 , and Arg 511 ) are relevant to Trl1 function in vivo or kinase activity in vitro. We also mutated Asp 425 and Asp 454 , which are located between the A-box and the RXXXR motif and are conserved among fungal tRNA ligases (Fig. 2).
The five TRL1-Ala alleles were cloned into a CEN TRP1 plasmid under the control of the native TRL1 promoter and tested for trl1⌬ complementation. The D425A and R511A transformants failed to give rise to 5-FOA-resistant colonies at 18, 25, 30, or 37°C; thus, these two alanine mutations were lethal in vivo (Table I). The K404A, T405A, and D454A alleles supported growth on 5-FOA, and the resulting TRL1-Ala strains grew as well as TRL1 cells on YPD agar at all temperatures (ϩϩϩ growth in Table I). These results show that the defining Lys and Thr residues of the A-box motif are dispensable for tRNA ligase function in vivo. This initially surprising result was verified by recovering the TRP1 plasmid from 5-FOA-selected K404A and T405A yeast strains, amplifying them in vivo by transformation in E. coli, and then sequencing the TRL1 genes of the clonal isolates, which revealed that the plasmid alleles retained the K404A and T405A mutations.
The wild-type Trl1 protein and the K404A, T405A, D425A, and R511A mutants were expressed in bacteria as His 10 -tagged fusion proteins and then isolated from soluble bacterial lysates by nickel-agarose chromatography. SDS-PAGE analysis showed that the material eluted from nickel-agarose with imidazole consisted of a mixture of full-length 90-kDa His-Trl1 (indicated by the arrow in Fig. 7A) plus a collection of smaller His-tagged N-terminal fragments of Trl1, presumably arising via proteolysis during recombinant protein production in E. coli, as described previously by Abelson and colleagues (11,13). The polypeptide compositions of the affinity-purified His-Trl1-Ala mutants were essentially identical to that of wild-type Trl1 (Fig. 7A). The adenylyltransferase activity of the fulllength Trl1 polypeptide was evinced by label transfer from [␣-32 P]ATP to form a ϳ90-kDa covalent Trl1-adenylate adduct (Fig. 7B). A ϳ65-kDa N-terminal fragment of Trl1 seen in the Coomassie Blue stained gel (Fig. 7A) was also labeled with [ 32 P]AMP (Fig. 7B). The full-length mutant enzymes K404A, T405A, D425A, and R511A retained adenylyltransferase activity (Fig. 7B), suggesting that these mutations in the central kinase domain did not result in global unfolding of Trl1.
The 5Ј polynucleotide kinase activity of wild-type Trl1 was assayed by the transfer of 32 P i from 100 M [␥-32 P]ATP to the 5Ј-OH terminus of an 18-mer RNA oligonucleotide to form a 5Ј 32 P-labeled RNA product that was resolved from free ATP by polyacrylamide gel electrophoresis (Fig. 7C). The extent of label  transfer from ATP to the 5Ј-OH RNA acceptor was reduced to Ͻ10% of the wild-type value by the K404A mutation in the Walker A-box. Kinase activity was abolished by the T405A mutation in the A-box and by the D425A and R511A changes (Fig. 7C). The loss of in vitro kinase activity elicited by the single mutations in the NTP-binding P-loop motif was paradoxical in light of the finding that the K404A and T405A mutations had no apparent effects on yeast cell growth (Table I). Prior studies by the Greer and Abelson laboratories (15,16) had shown that Trl1 is capable of using GTP as the phosphate donor in the polynucleotide kinase reaction; indeed, their work suggested that GTP is the preferred substrate for the Trl1associated kinase activity in vitro. We assayed the ability of wild-type and mutated versions of Trl1 to transfer 32 P i from 100 M [␥-32 P]GTP to the 5Ј-OH terminus of an 18-mer RNA oligonucleotide. The instructive findings were that the K404A and T405A mutants, which were active in vivo, retained substantial GTP-dependent kinase activity in vitro (35 and 72% of wild-type Trl1, respectively). Thus, the lysine and threonine side chains of the P-loop are critical for ATP-dependent kinase activity but not for the GTP-dependent reaction. The R511A mutant, which was lethal in vivo and inactive as an ATP-dependent kinase, was ϳ14% as active as wild-type Trl1 in GTPdependent RNA phosphorylation (Fig. 7C). Conceivably, the low residual GTP-dependent activity of R511A did not suffice for Trl1 function in vivo. The lethal D425A mutation abolished GTP-dependent RNA kinase activity, just as it abolished the ATP-dependent kinase. These results suggest that GTP rather than ATP is the physiological substrate for the Trl1 kinase activity in vivo and that the central domain of Trl1 contributes to the kinase active site.
The role of the central region of Trl1 in polynucleotide kinase function was investigated further by purifying the Trl1 domains and testing them for GTP-dependent RNA kinase activity in vitro. The polypeptide compositions of the full-length Trl1 (aa 1-827), the adenylyltransferase domain (aa 1-388), the kinase-CPD domain (aa 389 -827), and the CPD domain (aa 562-827) are shown in Fig. 8A. Whereas the full-length Trl1 protein and the kinase-CPD domain readily catalyzed phosphoryl transfer from GTP to the 18-mer 5Ј-OH RNA oligonucleotide, the isolated adenylyltransferase and CPD domains were inert with respect to kinase activity (Fig. 8B). We conclude from the deletion analysis that the adenylyltransferase domain plays no significant role in the kinase function of Trl1 and that the central domain is essential for kinase function. The latter inference is consistent with the observed inactivation of the kinase by the D425A mutation in the central domain. We were unable to produce the isolated central domain (aa 389 -561) as an intact soluble protein in bacteria; thus, its biochemical properties could not be evaluated.
The findings that single mutations K404A and T405A in the P-loop signature GXGKT did not eliminate Trl1 function in vivo or GTP-dependent kinase activity in vitro raised the prospect that ATP and GTP might bind to separate sites on the enzyme, with ATP utilizing the P-loop and GTP binding elsewhere. Alternatively, the Trl1 kinase might have a single NTP site that includes the P-loop, at which ATP binding is more acutely dependent on contacts with the lysine and threonine side chains of the P-loop than is GTP binding. To explore the latter scenario, we constructed a double mutant of the P-loop, K404A-T405A, in which both side chains were replaced by alanine. We found that the K404A/T405A mutation was lethal in vivo (Table I). The purified recombinant His-Trl1-(K404A/ T405A) protein was inert in catalysis of phosphoryl transfer from [␥-32 P]GTP to the 5Ј-OH terminus of the 18-mer RNA oligonucleotide acceptor (not shown). We infer from these results that utilization of GTP as the physiological substrate for the Trl1 kinase depends on the P-loop element within the central kinase-like domain and that the lysine and threonine mutations synergize to inactivate the GTP-dependent Trl1 kinase.
Delineating the Margins of the CPD Domain and Essential Amino Acids Therein-Complementation of trl1⌬ could also be achieved by coexpressing the adenylyltransferase-kinase domain Trl1-(1-561) and the CPD domain Trl1-(562-827) as unlinked polypeptides (Fig. 9). The TRL1-(1-561) TRL1-(562-827) strain grew as well as wild-type TRL1 on YPD agar at 25 and 30°C (ϩϩϩ growth at these temperatures) but formed only pinpoint-sized colonies at 37°C (scored as ϩ growth). Thus, the kinase module can function in vivo when linked either to the adenylyltransferase domain or the CPD domain.
The CPD domain of yeast tRNA ligase belongs to the socalled "2H" superfamily of phosphoesterases, which is defined by the presence of two copies of the histidine-containing motif, HT, spaced 70 -110 aa apart in the primary structure (58,59). The crystal structure of a plant CPD enzyme, which hydrolyzes both nucleoside 2Ј,3Ј-cyclic phosphates and ADP-ribose 1Љ,2Љ-cyclic phosphate, revealed that the defining histidines are located close together in the tertiary structure and together comprise the active site (60). Here we replaced the individual histidines of the 673 HITL and 777 HITL motifs of Trl1 with alanine. The H673A and H777A changes were made in the context of the full-length Trl1 protein, and the mutant alleles were tested by plasmid shuffle for complementation of trl1⌬. We found that the H777A mutation was lethal (Table I). The H673A allele supported growth of trl1⌬ on 5-FOA; the resulting 5-FOA-resistant TRL1-H673A strain grew as well as wild-type yeast on YPD agar at 25 and 30°C but was unable to form any colonies at 37°C (Table I). The finding that elimination of the distal and proximal histidines of the putative CPD active site resulted in constitutive and conditional lethality, respectively, is consistent with CPD catalytic activity being important for Trl1 function in vivo. DISCUSSION Three enzymatic activities of the yeast tRNA ligase are organized in a modular fashion within a single 827-aa polypeptide. The physical order of the active sites within the primary structure of Trl1 (H 2 N-adenylyltransferase-kinase-CPD-COOH) is conserved in other fungal tRNA ligase orthologs. Compaction of three RNA end-processing enzymes within a single protein echoes the case of the poxvirus mRNA capping enzyme, in which the RNA triphosphatase, RNA guanylyltransferase, and RNA (guanine-N7)-methyltransferase active sites are arranged in a modular fashion within an 844-aa polypeptide (63). The emergence of such multifunctional enzymes probably occurred via gene duplication and fusion events, in which the individual modules first acquired their biological specificity for tRNA splicing (or capping) via gene duplication and divergence from an ancestral catalytic domain and then later fused to form a single unit dedicated to a particular RNA processing pathway. Gene fusion is the most plausible scenario for tRNA ligase, given the existence in other cellular and viral niches of freestanding monofunctional adenylyltransferase/ligase and CPD enzymes that resemble the Trl1 domains. The present study sheds light on the relationships of the three Trl1 domains to structurally related catalytic modules in other systems and the issue of whether the covalent connections between the Trl1 domains are critical for in vivo function.
We show here that all three domains and enzyme activities are essential for yeast viability. The essentiality of all three enzymes had not been demonstrated previously and was not a foregone conclusion. For example, Phizicky et al. (62) reported that conditional repression of Trl1 production in vivo resulted in the accumulation of unligated tRNA half-molecules that contained processed 5Ј-PO 4 ends on the distal half-fragment and decyclized monophosphate ends on the proximal half-fragment. Their results provided strong evidence that the ligase function of Trl1 was essential but raised the possibility that other cellular enzymes might be able to perform the kinase and CPD reactions and thereby be redundant to these two functions of Trl1. The present findings that point mutations in the central kinase domain of Trl1 abolished kinase activity in vitro and were lethal in vivo argue against the existence of a functionally redundant RNA kinase activity in budding yeast. The lethal effects of deleting the CPD domain or of a point mutation within the CPD domain suggest that there is no backup enzyme available to convert the cyclic phosphate end to the 3Ј-OH/2Ј-PO 4 terminus required for sealing by the RNA ligase component. Thus, the earlier findings concerning conditional repression may be attributable to residual levels of Trl1 kinase and CPD activity.
The N-terminal adenylyltransferase and C-terminal CPD fragment comprise autonomous catalytic domains in vitro (14). Here we show that they are functionally autonomous in vivo as well (i.e. they can complement trl1⌬ when separated genetically from the rest of the Trl1 protein). The in vivo function of the isolated domains was evident without resorting to increased gene dosage or the use of a non-native promoter to drive expression of the isolated domains. These results suggest that either there is no need for physical interactions between the N-and C-terminal portions of Trl1 or the separately expressed N-and C-domains are able to interact in trans. The kinase domain was functional in vivo whether fused to the adenylyltransferase domain or to the CPD domain. The recombinant kinase-CPD protein was catalytically active in RNA phosphorylation. We were unable to produce a recombinant version of the isolated central kinase domain, which suggests that fusion to one of the flanking domains might be critical for proper folding or activity. In the case of bacteriophage T4 Pnk, where the N-terminal kinase domain is fused to a C-terminal 3Ј-phosphatase domain, severing the covalent connection between the two domains results in a 10-fold decrement in kinase activity for the isolated kinase domain compared with the full-length enzyme (26,28). We have not tested whether coexpression of the three isolated Trl1 domains in yeast could support cell growth. Little is known at present about the quaternary structure of yeast Trl1, its component domains, or its repertoire of protein-protein interactions; these will be the subjects of future studies.
The N-terminal segment Trl1-(1-376) represents a minimal adenylyltransferase domain that suffices for yeast growth at all temperatures and catalytic activity in vitro. Apostol et al. (14) reported that an internally deleted version of Trl1 of residues constructed by eliminating residues 379 -396 and fusing the flanking segments together was inert in adenylyltransferase activity. Thus, they suggested that amino acids 378 -396 were required for adenylyltransferase activity. The present study shows that this is not the case. It is possible that the internal deletion analyzed by Apostol et al. (14) was deleterious in its own right to the proper folding of the upstream adenylyltransferase domain.
To better understand the structural requirements for Trl1 adenylyltransferase activity, we initiated an alanine-scanning mutational analysis of selected residues of Trl1 that are conserved in other fungal tRNA ligases and in bacteriophage T4 RNA ligase 1, an enzyme dedicated to tRNA repair in vivo and the closest homolog of fungal tRNA ligases. We found that residues in nucleotidyl transferase motifs I (Lys 114 ), IV (Glu 266 and Gly 267 ), and V (Lys 284 and Lys 286 ) were essential for Trl1 activity in vivo, whereas mutations at three other conserved residues (Asn 116 , Gly 152 , and Glu 153 ) did not affect cell growth. Mutations K114A, E266A, and K284A either abolished or suppressed adenylyltransferase activity in vitro. These effects are concordant with previous studies of the equivalent side chains of DNA ligases, mRNA capping enzymes, and bacteriophage T4 RNA ligases 1 and 2 (30,(51)(52)(53)(54)(55)(56)(57)64). Our results indicate that the structural basis for nucleotidyl transfer is at least partially conserved among tRNA ligases, phage RNA ligases, DNA ligases, and mRNA capping enzymes. Such conservation is consistent with the speculation that RNA-joining enzymes that evolved during a primordial RNA/protein world are the ancestors of present day DNA ligases and mRNA capping enzymes (55).
The primary structure of the central portion of yeast Trl1 resembles the kinase domain of T4 polynucleotide kinase. The crystal structures of T4 Pnk (28, 29) guided our mutational analysis of the central domain of Trl1. Single mutations K404A and T405A in the putative NTP binding P-loop ( 401 GXGKT 405 ) had no effect on Trl1 function in vivo. The K404A and T405A mutations eliminated ATP-dependent kinase activity but preserved GTP-dependent kinase activity. A double alanine mutant in the P-loop was lethal in vivo and abolished GTP-dependent kinase activity. These results suggest that GTP is the physiological substrate in vivo (as suggested originally by the in vitro studies of Abelson and Greer (15,16)) and that the Trl1 kinase has a single NTP binding site of which the P-loop is a component. Two other mutations in the central domain were lethal in vivo and either abolished (D425A) or severely reduced (R511A) GTP-dependent RNA kinase activity in vitro.
We can make reasonable predictions about the roles of Trl1 residues Lys 404 , Thr 405 , Asp 425 , and Arg 511 based on the crystal structures of T4 Pnk, which have been solved with either sulfate or ADP occupying the NTP binding site (28,29). Lys 404 , Thr 405 , and Arg 511 are likely to coordinate the ␤-phosphate of the NTP substrate. The differential effect of single alanine mutations on the ATP-versus GTP-dependent RNA kinase activities of Trl1 suggests that there are specific contacts between Trl1 and GTP that do not apply to ATP and that such GTP contacts compensate for the loss of either one of the P-loop side chains Lys 404 or Thr 405 but not for the simultaneous elimination of both P-loop residues. This model is consistent with available biochemical evidence that the apparent K m of the Trl1 kinase for ATP is 800-fold greater than the K m for GTP (15,16). Asp 425 of Trl1 is the putative equivalent of the essential Asp 35 residue of T4 Pnk. It is proposed that this aspartate is located near the 5Ј-OH of the RNA acceptor and that it either serves as a general base catalyst to promote attack of the 5Ј-OH on the ␥-phosphorus of the NTP substrate (28) or comprises part of a divalent cation binding site (29).
Given the biochemical similarities between tRNA splicing and tRNA restriction repair and the apparent mechanistic and structural similarities between the adenylyltransferase and kinase components of the yeast and phage T4 repair systems, we propose that Trl1 and T4 Rnl1/Pnk have a shared evolutionary history whereby they descend from ancestral ligases and kinases devoted to repairing broken tRNAs. Where the two systems differ is the covalent connection (yeast) or lack thereof (T4) between the adenylyltransferase and kinase modules and the characteristics of the enzymes that process the 3Ј-end of the tRNA break. It has been proposed that the CPD domain of Trl1 belongs to the 2H family of phosphotransferases (58,59). Our findings that the two signature histidines that define this family are either essential or important for Trl1 function in vivo supports this proposal. The 3Ј-phosphatase domain of T4 Pnk belongs to the DxD family of phosphotransferases (26 -29), which are structurally and mechanistically unrelated to the 2H enzymes.