Transient ADP-ribosylation of a 2′-Phosphate Implicated in Its Removal from Ligated tRNA during Splicing in Yeast*

The last step of tRNA splicing in yeast is catalyzed by Tpt1 protein, which transfers the 2′-phosphate from ligated tRNA to NAD to produce ADP-ribose 1"-2"-cyclic phosphate (Appr>p). Structural and functional TPT1 homologs are found widely in eukaryotes and, surprisingly, also in Escherichia coli, which does not have this class of tRNA splicing. To understand the possible roles of the Tpt1 enzymes as well as the unusual use of NAD, the reaction mechanism of the E. colihomolog KptA was investigated. We show here that KptA protein removes the 2′-phosphate from RNA via an intermediate in which the phosphate is ADP-ribosylated followed by a presumed transesterification to release the RNA and generate Appr>p. The intermediate was characterized by analysis of its components and their linkages, using various labeled substrates and cofactors. Because the yeast and mouse Tpt1 proteins, like KptA protein, can catalyze the conversion of the KptA-generated intermediate to both product and the original substrate, these enzymes likely use the same reaction mechanism. Step 1 of this reaction is strikingly similar to the ADP-ribosylation of proteins catalyzed by a number of bacterial toxins.

The last step of tRNA splicing in yeast is catalyzed by Tpt1 protein, which transfers the 2-phosphate from ligated tRNA to NAD to produce ADP-ribose 1؆-2؆-cyclic phosphate (Appr>p). Structural and functional TPT1 homologs are found widely in eukaryotes and, surprisingly, also in Escherichia coli, which does not have this class of tRNA splicing. To understand the possible roles of the Tpt1 enzymes as well as the unusual use of NAD, the reaction mechanism of the E. coli homolog KptA was investigated. We show here that KptA protein removes the 2-phosphate from RNA via an intermediate in which the phosphate is ADP-ribosylated followed by a presumed transesterification to release the RNA and generate Appr>p. The intermediate was characterized by analysis of its components and their linkages, using various labeled substrates and cofactors. Because the yeast and mouse Tpt1 proteins, like KptA protein, can catalyze the conversion of the KptA-generated intermediate to both product and the original substrate, these enzymes likely use the same reaction mechanism. Step 1 of this reaction is strikingly similar to the ADP-ribosylation of proteins catalyzed by a number of bacterial toxins.
tRNA introns occur widely in Eukarya and Archaea (1,2). Splicing in these organisms is initiated by a highly conserved endonuclease that excises the intron (3)(4)(5), followed by joining of the two half-molecules by one of two different ligases (6 -13). In the yeast Saccharomyces cerevisiae, in which the process is best studied, ligation occurs by a four-step reaction, producing a splice junction with a 2Ј-phosphate (14,15).
A single essential gene (TPT1) encodes the 2Ј-phosphotransferase responsible for removal of the splice junction 2Ј-phosphate from ligated tRNA (16,17). This reaction is unusual because the 2Ј-phosphate is transferred to NAD (18), producing mature tRNA and ADP-ribose 1Љ-2Љ cyclic phosphate (ApprϾp) 1 (19). The TPT1 gene product is involved in this step in vivo because a conditional yeast tpt1 mutant, when depleted for the gene product, accumulates at least eight ligated tRNA species bearing a splice junction 2Ј-phosphate (17). This presumably is the essential function of Tpt1 protein. Examination of four of these tRNAs demonstrated that they are also undermodified specifically at the splice junction residue (17).
The yeast Tpt1 protein is part of a family of functional 2Ј-phosphotransferases found in Eukarya (Schizosaccharomyces pombe, Candida albicans, Arabidopsis thaliana and Mus musculus), and Eubacteria (Escherichia coli), with other likely members in another bacterial species and several Archaea (46). Expression of the eukaryotic phosphotransferase genes and the E. coli gene (kptA) complements a yeast tpt1 mutant, and the corresponding proteins catalyze the same reaction as the yeast protein, producing ApprϾp from ligated tRNA and NAD. The widespread occurrence of 2Ј-phosphotransferases in Eukarya is consistent with the ubiquitous presence of intron-containing tRNAs and ligases that generate 2Јphosphorylated substrates in eukaryotes. However, a functional 2Ј-phosphotransferase was not anticipated in E. coli and bacteria in general because they do not have the corresponding ligase or this class of tRNA splicing. Bacteria only have the distinctively different group I and group II self-splicing introns (20 -22). Moreover, phylogenetic analysis indicates that the bacterial gene is ancient. It is unclear what the function of the E. coli protein might be. One possible role for the E. coli protein is the catalysis of a related chemical reaction, which might also be catalyzed by other Tpt1 homologs.
To learn the spectrum of reactions that could be catalyzed by the Tpt1 enzyme family and to understand the unusual use of NAD in the phosphotransferase reaction, we undertook to learn the reaction mechanism. Two mechanistic pathways that could account for the formation of ApprϾp are illustrated in Fig. 1. In mechanism A (19), the first step is a phosphoryl transfer in which the 2Ј-phosphate at the splice junction is transferred to the 2Љ-hydroxyl of the NMN ribose of NAD, releasing the dephosphorylated RNA. This is followed by cyclization of the phosphate to the 1Љ position of NAD, with concomitant release of nicotinamide. In mechanism B, the phosphodiester bond is formed in step 1 by formation of a covalent bond between the 2Ј-phosphate of tRNA and the 1Љ position of NAD, releasing nicotinamide. The second step is then a simple transesterification reaction, where the 2Љ-hydroxyl of NAD displaces the tRNA 2Ј-OH in the phosphodiester linkage. In both mechanisms the energy for phosphodiester bond formation is derived from the hydrolysis of the C-N glycosidic bond joining ribose and nicotinamide.
We provide evidence here that mechanism B is correct, based on the identification and characterization of a reaction intermediate generated by the E. coli 2Ј-phosphotransferase, KptA. Further, we show that the bacterial mechanism is conserved for yeast (Tpt1), and mouse (mTpt1) phosphotransferase proteins.
Step 1 of mechanism B is related to a class of ADPribosylating reactions found in a variety of bacterial toxins and other ADP-ribosyl transferases.

EXPERIMENTAL PROCEDURES
Labeling of Substrates-ApA P pA was 5Ј-phosphorylated in 20-l reactions containing 100 M ApA P pA, 5 Ci of [␥-32 P]ATP (3,000 Ci/ mmol), and 1 unit of polynucleotide kinase (3Ј-phosphatase-free) in buffer (Boehringer Mannheim). Reactions were incubated for 30 min at 37°C and applied to silica thin layer chromatography plates (J. T. Baker), and products were separated in buffer containing n-propyl alcohol:NH 4 OH:H 2 O, 55:35:10 v/v/v and eluted in H 2 O. [ 32 P-adenylate]NAD was prepared from [␣-32 P]ATP (3,000 Ci/mmol) and NMN as described (23). Reaction mixtures were treated subsequently with 0.01 unit of calf intestinal phosphatase for 30 min, phenol extracted, and applied to silica thin layer chromatography plates; products were resolved in buffer containing ethanol and 1 M NH 4 OAc, pH 7.2, 7:3 v/v, eluted in water, and dried. [ 3 H]NAD was prepared as described by Little (24). p*-intermediate was made from 12-30 fmol of p*ApA P pA (or p*ApApA P POCH3 ) in the presence of 5 mM NAD, and products were purified by chromatography on silica thin layer plates developed in buffer containing ethanol and 1 M NH 4 OAc, pH 7.2, 7:4.5 v/v, followed by elution in water and drying. Although this procedure yields intermediate that is partially contaminated with some RNA substrate and product (because of the poor separation on these plates), the eluted intermediate lacks NAD and is active (see "Results"). Ap*pN intermediate was purified the same way, after synthesis as described in Fig. 6.
2Ј-Phosphotransferase-Phosphotransferase activity was assayed for 30 min as described (16) at 30°C for yeast (Tpt1) and mouse (mTpt1) phosphotransferases and at 37°C for KptA from E. coli. Analysis was by TLC on polyethyleneimine (PEI)-cellulose plates developed in 2 M sodium formate, pH 3.5, unless otherwise stated.
Proteins-KptA and mTpt1 were prepared from extracts of E. coli cells expressing the protein by Blue Sepharose column chromatography. 2 Yeast Tpt1 was prepared by M. Steiger as a Tpt1-His 6 fusion protein and was purified from E. coli extracts by nickel column affinity chromatography. Yeast cyclic phosphodiesterase protein has been described (23). Calf intestinal alkaline phosphatase (Boehringer Mannheim) was diluted to 0.01 units/l in 300 mM NaCl, 1 mM MgCl 2 , and 0.1 mM ZnCl 2 , and 1 l was added for 30 min at 37°C. Pyrophosphatase reactions contained 10 mM Tris-HCl, pH 9, 10 mM MgCl 2 , 1 g of carrier RNA, and 0.1-1 unit of nucleotide pyrophosphatase (type II: Crotalus adamanteus; Sigma) in 10 l of buffer and were incubated at 37°C for 60 min.

RESULTS
The E. coli 2Ј-Phosphotransferase Protein Produces a Reaction Intermediate-To examine the mechanism of the E. coli 2Ј-phosphotransferase (KptA protein), we used a 5Ј-32 P endlabeled synthetic RNA trimer containing an internal 2Ј-phosphate (p*ApA P pA, where the * indicates the position of the labeled phosphate). This substrate and several other synthetic RNAs have been characterized extensively by NMR and analysis with nucleases and phosphatases 2 ; a similar substrate has been used with the yeast enzyme (19).
As shown in Fig. 2, titration of p*ApA P pA with increasing amounts of the E. coli enzyme in the presence of NAD produces both the expected product (p*ApApA) and in some cases, an additional product. This additional product (designated p*intermediate) is most prominent in lane f, in which the amount of substrate and final product are roughly equal. It seems unlikely that this additional product is the result of a contaminating activity in the enzyme preparation because the amount of the product decreases as more enzyme is added to the reaction (lanes h-j). Rather, the additional product appears to be an intermediate because there is less of it at both lower and higher concentrations of enzyme.
If there is an intermediate, it should appear before product forms and disappear as product accumulates. This is demon-  Fig. 3, which displays the results of a time course of the reaction at several concentrations of the E. coli phosphotransferase. In the presence of 1 unit of enzyme activity (panel C), it is clear that at the early time points (up to 40 min), there is substantially more of the spot labeled p*-intermediate than of product, and at later time points (after 80 min), there is substantially less p*-intermediate and correspondingly more product. This is quantitated in Fig. 3 c-e with lanes m and n).
In addition, an intermediate is formed with labeled NAD. To show this we used [ 32 P-adenylate]NAD (Ap*pN) in a phosphotransferase reaction with unlabeled RNA (ApA P pA). As shown in Fig. 6 (lanes c-e), two products are formed. One of these comigrates with Ap*prϾp and increases with more protein in the reaction. The other is designated Ap*pN intermediate because, like the p*-intermediate, its levels are maximal when roughly half of the final amount of product is formed (lane d).
As with the p*-intermediate, the isolated Ap*pN intermediate is efficiently converted to product, in this case to Ap*prϾp, with or without NAD (Fig. 7, compare lane a with lanes c and d). As expected for Ap*prϾp, the product is resistant to phosphatase (lane e) and is a substrate for a yeast cyclic phosphodiesterase that converts Ap*prϾp to Ap*pr1Љp (lane f) (23).
The . Therefore, additional unlabeled NAD was used in the reaction with labeled NAD, which requires both extra substrate RNA (to convert a reasonable amount of the label to product) and correspondingly more protein to catalyze the reaction. Nonetheless, the amount of protein required to form the intermediate in each case conforms very closely to that expected from the kinetic parameters. 2 The fact that both RNA and NAD comprise the intermediate eliminates mechanism A as the catalytic pathway for the phosphotransfer reaction (Fig. 1). The results are consistent with mechanism B, and the experiments described below, which further probe the structure of the intermediate, lend support to this conclusion.
The RNA 2Ј-Phosphate Is Linked to the NMN Portion of NAD-To determine if the 2Ј-phosphate of the RNA is involved in the NAD-RNA linkage, as predicted by mechanism B, we treated both p*-intermediate and the Ap*pN intermediate with phosphatase. As shown in Fig. 7 (lane b), there is no change in the mobility of the Ap*pN intermediate after phosphatase treatment (although a minor contaminant is sensitive), indicating that the 2Ј-phosphate is involved in the linkage. By contrast, treatment of p*-intermediate results in the formation of inorganic phosphate (Fig. 4, lane h), demonstrating that the 5Ј-32 P-labeled phosphate of the RNA substrate is not involved in the linkage.
As predicted by mechanism B, the 2Ј-phosphate is linked to the NMN moiety of NAD. To show this, we treated the p*intermediates formed with NAD and NGD (see Fig. 5) with pyrophosphatase to remove the AMP (GMP) portion of the molecule. Because this removes the distinguishing portion of NAD and NGD, the position of RNA attachment to NAD can be determined simply by investigation of the migration properties of the released labeled products. Since the pyrophosphatase digestion products comigrated in our TLC system, this indicates that the attachment is to NMN (data not shown).
Nicotinamide Is Released during Intermediate Formation-Two approaches were used to demonstrate that nicotinamide is released as intermediate is formed, as predicted in mechanism B. First, we demonstrated that phosphotransferase can catalyze the reversal of step 1 in the presence of nicotinamide (Fig.  8, panel A).  and f). Quantitation at 10 mM nicotinamide indicates that there is conversion of the intermediate to Ϸ70% substrate and Ϸ30% product. Similarly, the Ap*pN intermediate can be converted to the original substrate (Ap*pN) with excess nicotinamide (Fig. 8, panel B, lanes h and  i). Because nicotinamide addition reverses the phosphotrans- Second, we showed directly that nicotinamide release occurs well before formation of product (data not shown). This was accomplished by conducting a time course of the release of [ 3 H]nicotinamide from [ 3 H]NAD. To correlate nicotinamide release with reaction progress, we used a second labeled substrate, p*ApA P pA, in the same reaction, at comparable concentrations (and at similar intensity after autoradiography). The [ 3 H]nicotinamide was clearly formed (10% at 10 min) well before formation of product p*ApApA (7% at 40 min). Presumably nicotinamide release occurs during intermediate formation.
Because nicotinamide is absent from the intermediate and the 2Ј-phosphate is engaged in the linkage to NAD, the most reasonable structure of the intermediate has the RNA 2Ј-phosphate linked at the 1Љ position. If so, the adjacent 2Љ-and 3Љ-hydroxyls should be sensitive to periodate. To test this, we used p*ApApA P POCH3 as substrate because it has no vicinal hydroxyls on its riboses. 2 After forming the intermediate with AppN, we treated it with pyrophosphatase to release the AMP moiety (Fig. 9, lane c); this removed the vicinal hydroxyls on the ribose of AMP. As shown in lane d, the remaining material is periodate-sensitive, which could only occur if the 2Љ-hydroxyl is free. We conclude that step 1 involves attachment of the RNA 2Ј-phosphate to the 1Љ position as depicted in mechanism B of Fig. 1. (The alternative explanation, that the RNA 2Ј-phosphate is attached to the 3Љ position of ADP-ribose, would make the reaction too complicated to form the final product ApprϾp). It follows that step 2 is then a simple transesterification reaction, resulting in cyclization of the phosphate at the 2Љ position of the ribose.

Yeast and Mouse 2Ј-Phosphotransferases Can Catalyze
Step 2 of the Reaction and the Reverse of Step 1-Because the E. coli protein is part of a highly conserved family that includes the yeast (Tpt1) and mouse (mTpt1) phosphotransferases, the mechanism of the reaction should be similarly conserved. Three lines of evidence support this claim. First, in the presence of high concentrations of NAD, a small amount of intermediate is formed with the yeast and mouse phosphotransferases, which comigrates with the intermediate formed with KptA protein (data not shown). Second, both Tpt1 and mTpt1 proteins can catalyze the conversion of the E. coli-produced intermediate to product (Fig. 10, lanes b and c; data not shown for mouse). Third, both yeast and mouse can catalyze the reverse of step 1: the KptA intermediate is converted to the original substrate when an excess of nicotinamide is added to the reaction (Fig. 11, lanes c and d). Because the yeast and mouse phosphotransferases can catalyze both the forward and the reverse reactions from the intermediate, they must be able to catalyze the reaction by the same mechanism.

DISCUSSION
Isolation and characterization of a reaction intermediate provides evidence that removal of the splice junction 2Ј-phosphate from ligated tRNA during the last step of splicing involves transient ADP-ribosylation of that phosphate. Individual labeling of the RNA substrate, the adenylic acid phosphate of NAD, and nicotinamide indicates that both RNA and NAD are part of the intermediate and that nicotinamide is absent. Consistent with this, the intermediate can be converted to product in the absence of both RNA substrate and NAD and is converted back to substrate in the presence of excess nicotinamide. As depicted in mechanism B of Fig. 1, these results suggest that step 1 of the reaction is a nucleophilic attack of the RNA 2Ј-phosphate on the 1Љ position of NAD. Evidence for this attachment in the intermediate includes the phosphatase resistance of the RNA 2Ј-phosphate after intermediate formation and the periodate sensitivity of the ribose that released nicotinamide. This attachment is also consistent with energetic requirements: release of nicotinamide during step 1 would provide sufficient energy for phosphodiester bond formation between the RNA 2Ј-phosphate and the NMN ribose of NAD because the free energy of hydrolysis of the N-glycosidic bond of NAD is Ϫ8.2 kcal/mol (25,26).
Step 2 is then a transesterification reaction, resulting in cyclization of the phosphate to the 2Љ position of ADP-ribose and the release of RNA. Our evidence also indicates that the yeast and mouse phosphotransferases use the same mechanism as the E. coli protein because both Tpt1 and mTpt1 proteins are able to catalyze the conversion of the isolated E. coli intermediate both forward to product and back-ward to substrate. Presumably, other members of this widespread family of 2Ј-phosphotransferase enzymes catalyze this reaction in the same way. Although we have not yet demonstrated that the intermediate forms with tRNA, it seems probable that the mechanism is the same with this substrate.
The intermediate described here is one of a surprisingly large number of chemical alterations that are formed one base 3Ј of the anticodon (the hypermodified position) in tRNAs that are spliced. During the first two steps of splicing, this nucleotide bears a 2Ј-3Ј-cyclic phosphate (4,27), and then a 2Ј-phosphate (28) before the half-molecules are joined to generate a splice junction phosphodiester with a 2Ј-phosphate (28). As described here, the 2Ј-phosphate is then modified by addition of an ADP-ribose adduct before removal of the phosphate as ApprϾp (19) and formation of tRNA with a splice junction 2Ј-OH (17). Finally, the base of this residue can be modified efficiently to its hypermodified state (17). The significance of the multiple steps at this one residue is unclear; perhaps they occur to ensure that immature tRNA is not used inappropriately in the cell (1).
Step 1 of the reaction is strikingly similar to the ADP-ribosylation catalyzed by a number of well studied bacterial toxins that exert their effects by modification of protein synthesis factors, structural proteins, or signal transduction proteins. In the case of the 2Ј-phosphotransferase, the nucleophile is the 2Ј-phosphate of the RNA. By contrast, the nucleophile for the toxins can be a variety of amino acid residues: diphthamide (modified histidine) for Pseudomonas exotoxin A and diphtheria toxin (29,30); arginine for choleragen toxin, Clostridium botulinum C2, and dintrogenase reductase ADP-ribosyl transferase (31-34); cysteine for pertussis toxin (35); and asparagine for C1 botulinum toxin (36). We noted previously that the family of 2Ј-phosphotransferases shares sequence similarity with the NAD binding site of diphtheria toxin and Pseudomonas exotoxin A (16). Two other reactions that are similar to step 1 include the first step of poly(A)DP-ribose synthetase, in which glutamate acts as the nucleophile (37), and NAD glycohydrolases, which hydrolyze NAD to ADP-ribose and nicotinamide (38,39).
In the second step, KptA presumably catalyzes a transesterification reaction, in which the 2Љ-O of ADP-ribose displaces the 2Ј-O of tRNA in the phosphodiester bond, generating a cyclic phosphate.
Step 2 is comparable to the first step of a class of RNases where hydrolysis of the phosphodiester bond is catalyzed by attack of an adjacent hydroxyl to form a cyclic phosphate intermediate (40). An identical reaction occurs in tRNA splicing, in which the endonuclease generates a 2Ј-3Ј-cyclic phosphate terminus on the 5Ј-half molecule during hydrolysis to excise the intron (27). This step is also similar to the transesterification reactions of RNA-catalyzed self-splicing (41).
It is possible that KptA protein functions in E. coli as an ADP-ribosylating enzyme. Such a role would be consistent with the observation that the bacterial protein catalyzes the second step of the reaction poorly compared with the yeast and mouse phosphotransferases. This is why it is difficult to observe the intermediate with the yeast and mouse proteins. Furthermore, we argued previously that for an ancient bacterial protein still to be functional in yeast, it must have retained a high degree of substrate specificity and chemical reactivity, implying that its role is related to its catalytic function (46). Several cases of ADP-ribosylation of endogenous proteins have been reported in bacteria (E. coli, Pseudomonas maltophilia and Streptomycin triseus) (42)(43)(44), as well as the well studied regulatory ADPribosylation of dintrogenase reductase which occurs in the photosynthetic bacterium, Rhodospirillum rubrum (45). The E. coli protein might ADP-ribosylate a protein, as do the other proteins with this activity, or perhaps RNAs (as described here) or small molecules. Such a role might also be conserved in the eukaryotic 2Ј-phosphotransferases. Alternatively, there may be some other metabolite or small molecule that requires removal of its phosphate in E. coli, either to effect dephosphorylation or to produce ApprϾp.