Novel Mechanism of RNA Repair by RtcB via Sequential 2′,3′-Cyclic Phosphodiesterase and 3′-Phosphate/5′-Hydroxyl Ligation Reactions*

Background: RtcB is a new RNA ligase implicated in tRNA splicing and RNA repair. Results: RtcB has 2′,3′-cyclic phosphodiesterase and 3′-phosphate/5′-hydroxyl ligase activities. Conclusion: RtcB seals broken RNAs via a two-step reaction pathway. Significance: The structure, active site, and chemical mechanism of RtcB are unique. RtcB enzymes are a newly discovered family of RNA ligases, implicated in tRNA splicing and other RNA repair reactions, that seal broken RNAs with 2′,3′-cyclic phosphate and 5′-OH ends. Parsimony and energetics would suggest a one-step mechanism for RtcB sealing via attack by the O5′ nucleophile on the cyclic phosphate, with expulsion of the ribose O2′ and generation of a 3′,5′-phosphodiester at the splice junction. Yet we find that RtcB violates Occam's razor, insofar as (i) it is adept at ligating 3′-monophosphate and 5′-OH ends; (ii) it has an intrinsic 2′,3′-cyclic phosphodiesterase activity. The 2′,3′-cyclic phosphodiesterase and ligase reactions both require manganese and are abolished by mutation of the RtcB active site. Thus, RtcB executes a unique two-step pathway of strand joining whereby the 2′,3′-cyclic phosphodiester end is hydrolyzed to a 3′-monophosphate, which is then linked to the 5′-OH end to form the splice junction. The energy for the 3′-phosphate ligase activity is provided by GTP, which reacts with RtcB in the presence of manganese to form a covalent RtcB-guanylate adduct. This adduct is sensitive to acid and hydroxylamine but resistant to alkali, consistent with a phosphoramidate bond.

bution of RtcB fortifies its candidacy as a eukaryal tRNA splicing enzyme.
In support of a broad role for RtcB ligases in RNA repair, we've shown that E. coli RtcB is competent and sufficient for tRNA splicing in vivo, by virtue of its ability to complement growth of yeast cells that lack the endogenous healing-andsealing-type tRNA ligase Trl1 (19). RtcB also protects yeast trl1⌬ cells against a fungal ribotoxin that incises the anticodon loop of cellular tRNAs. Moreover, RtcB can replace Trl1 as the catalyst of HAC1 mRNA splicing during the unfolded protein response (19).
RtcB is unique with respect to its tertiary structure and the architecture of its active site (20), thereby suggesting an unprecedented mechanism of end joining. The simplest chemical scheme for the RtcB sealing reaction would entail attack by the O5Ј nucleophile on the cyclic phosphate, with expulsion of the ribose O2Ј (Fig. 1A). Our findings that RtcB is manganesedependent distinguishes the RtcB sealing activity from the metal-independent ligation of a 5Ј-OH strand to 2Ј,3Ј-cyclic phosphate strand catalyzed by DNA topoisomerase IB (21). We've speculated that manganese might promote RtcB catalysis by coordinating the O5Ј nucleophile to lower its pK a and/or engaging the cyclic phosphate oxygens to stabilize the transition state (18). RtcB has an active site pocket lined by essential histidines and a cysteine (19,20), suggestive of a metal-binding site.
The one-step transesterification ligation mechanism depicted in Fig. 1A is attractive from an energetics perspective, insofar as the cyclic phosphodiester is a strained energy-rich reactant with a favorable leaving group (the ribose O2Ј). For this mechanism to be plausible, RtcB would need to exclude water as a nucleophile and have some means to accept a polynucleotide 5Ј-OH as the only idoneous nucleophile. The mechanism also engenders an explicit prediction that RtcB ought not to ligate a 3Ј-monophosphate RNA terminus. Here we interrogate and overturn the simple model in Fig. 1A, in favor of a less parsimonious, but mechanistically more interesting, two-step ligation pathway depicted in Fig. 1B, in which RtcB first hydrolyzes the 2Ј,3Ј-cyclic phosphodiester end to a 3Ј-monophos-phate and then joins the 3Ј-monophosphate RNA end to a 5Ј-OH end to form a 3Ј35Ј phosphodiester.

EXPERIMENTAL PROCEDURES
Purification of RtcB-pET28b-His 10 Smt3-RtcB plasmids encoding wild-type RtcB and mutants C78A and H168A were transformed into E. coli BL21-CodonPlus(DE3). Cultures (1 liter) derived from single kanamycin-resistant transformants were grown at 37°C until the A 600 reached 0.6 -0.8. The cultures were chilled on ice for 30 min and then adjusted to 0.1 mM isopropyl ␤-D-thiogalactoside. Incubation was continued at 17°C for 16 h with constant shaking. Cells were harvested by centrifugation and stored at Ϫ80°C. Preparation of soluble lysates, recovery of the His 10 Smt3-RtcB proteins by nickel-agarose chromatography, excision of the tags by treatment with the Smt3-specific protease Ulp1, separation of the tag-free RtcB proteins from His 10 Smt3 by a second round of Ni-agarose chromatography, and final purification of RtcB to by Superdex 200 gel filtration were performed as described previously (18). Protein concentrations were determined using the Bio-Rad dye reagent with BSA as the standard.
Assays of RNA Ligase and Cyclic Phosphodiesterase Activities-Reaction mixtures (10 l) containing 50 mM Tris-HCl (pH 8.0), 2 mM MnCl 2 , 100 M GTP, 20 nM RNA repair substrate as specified, and 200 nM RtcB were incubated for 30 min at 37°C. The reactions were quenched by adding 1 l of 40 mM EDTA. The products were either analyzed directly by PAGE or digested for 15 min at 37°C with 1000 units of RNase T1 (Fermentas) prior to PAGE. The samples were mixed with 10 l of 90% formamide, 0.01% bromphenol blue/xylene cyanol, 50 mM EDTA and then resolved by electrophoresis (at 50 watts of constant power) through a 40-cm 20% polyacrylamide gel containing 8 M urea in 45 mM Tris borate, 1.2 mM EDTA. The 32 P-labeled RNAs were visualized by autoradiography of the gel and, where specified, quantified by scanning the gel with a Fuji Film BAS-2500 imager.

RESULTS
RtcB Seals Stem-loops with 3Ј-Monophosphate/5Ј-OH Breaks-We showed previously that RtcB was adept at ligating broken stem-loop structures with 2Ј,3Ј-cyclic phosphate and 5Ј-OH termini to restore a 3Ј,5Ј-phosphodiester at the repair junction (18). Here we queried whether RtcB could seal a 3Ј-monophosphate end. We used T4 RNA ligase 1 to transfer [5Ј-32 P]pCp to the 3Ј end of a 19-mer RNA oligonucleotide to generate a 20-mer strand ( HO UGGCUCCGAUAUCACGCU-UpCp) with 5Ј-OH and 3Ј-monophosphate ends and a single radiolabel between the 3Ј-terminal and penultimate nucleosides. We used purified E. coli RNA 3Ј-phosphate cyclase RtcA (24,25) to convert the 3Ј-phosphate end of the labeled 20-mer RNA into a 2Ј,3Ј-cyclic phosphodiester ( HO UGGCUCCGAU-AUCACGCUUpCϾp). We also converted the 5Ј-OH termini of these labeled strands into 5Ј-PO 4 ends by reaction with cold ATP and a mutated (phosphatase-dead) version of T4 polynucleotide 5Ј-kinase/3Ј-phosphatase (22). The four radiolabeled strands were annealed to a complementary unlabeled 5Ј-OH terminated 20-mer RNA to form the broken stem-loop substrates depicted in Fig. 2. These substrates were then reacted with either RtcB or plant tRNA ligase (AtRNL) and the products were analyzed by denaturing PAGE. Sealing activity was manifest by the conversion of the labeled 20-mer into a slowermigrating 40-mer stem-loop (Fig. 2). As expected, AtRNL catalyzed efficient sealing of substrates with a 2Ј,3Ј-cyclic phosphate end (lanes 6 and 12), but was inert with otherwise identical 3Ј-monophosphate substrates (lanes 3 and 9). By contrast, RtcB readily ligated breaks with either 3Ј-monophosphate ends (lanes 2 and 8) or 2Ј,3Ј-cyclic phosphate ends (lanes 5 and 11).
Intramolecular Joining of 3Ј-Monophosphate and 5Ј-OH Ends-The radiolabeled 20-mer RNAs were reacted with RtcB in the absence of a complementary strand and the products were analyzed by denaturing PAGE (Fig. 3, top panel). The substrates with 5Ј-OH ends were efficiently converted to a more rapidly migrating circular RNA species as a consequence of intramolecular ligation (lanes 2 and 4). Low levels of more slowly migrating multimers (products of intermolecular ligation) were also generated. There was little difference in the extents of circularization of 5Ј-OH terminated single strands with 3Ј-monophosphate versus cyclic phosphate ends. The two substrates with 5Ј-PO 4 ends were not sealed by RtcB (lanes 6 and 8).
The input 3Ј end structures and the RNA splice junctions were characterized further by digesting the samples with RNase T1 prior to their electrophoresis through a polyacrylamide gel (Fig. 3, bottom panel). RNase T1 incised the 20-mer RNA substrates 3Ј of the most distal guanosine to yield 32 P-labeled tetranucleotides: either HO CUUpCp or HO CUUpCϾp -with distinct electrophoretic mobilities (lanes 1 and 3). (This analysis verified that the efficiency of cyclization of the -UpCp end of the 20-mer during the preparative reaction with RtcA was 95%, i.e. 5% of the RNase T1 product was HO CUUpCp in lane 3.) The key finding was that the reactions with RtcB resulted in depletion of the T1 fragments derived from the substrate strands and the appearance of a new, more slowly migrating T1 fragment (Fig. 3, bottom panel, lanes 2 and 4) that corresponded to a 6-mer oligonucleotide, HO CUUpCpUGp, released by T1 incision at the guanosines flanking the predicted ligation junction ( Fig. 3). Note that the T1 junction fragment did not accumulate when RtcB was reacted with non-ligatable substrates with 5Ј-PO 4 ends (lanes 6 and 8).
We further queried the specificity of RtcB by asking whether the enzyme could join a 2Ј-monophosphate end to a 5Ј-OH. The 2Ј-monophosphate 20-mer substrate was prepared by treating the radiolabeled HO UGGCUCCGAUAUCACGCUU-pCϾp strand with the isolated C-terminal end-healing domain of plant tRNA ligase (6), which hydrolyzes the 2Ј,3Ј-cyclic phosphodiester to yield a 2Ј-phosphate, 3Ј-OH terminus. (By digesting the labeled 20-mer with RNase T1 before and after treatment with the plant ligase healing domain, and then analyzing the the T1 fragments by PAGE, we verified that the initial 2Ј,3Јcyclic phosphate terminus had been completely converted to a more rapidly migrating phosphomonoester derivative; not shown.) The radiolabeled 2Ј,3ЈϾp and 2Ј-phosphate 20-mer strands were then reacted in parallel with RtcB and AtRNL and the products were analyzed by PAGE (Fig. 4). As expected, AtRNL catalyzed efficient sealing of both substrates, yielding a mixture of circular and multimeric ligation products (Fig. 4, lanes 3 and 6). By contrast, RtcB sealed the 2Ј,3ЈϾp strand to yield a predominant circular ligation product (lane 2) but was   DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 unreactive with the 2Ј-monophosphate substrate (lane 5). The slight differences in electrophoretic mobility of the circular products made by AtRNL versus RtcB reflect the distinct outcomes of the RtcB and AtRNL ligation reactions. Whereas RtcB seals the 2Ј,3ЈϾp end to yield a splice junction with a conventional 3Ј35Ј phosphodiester linkage (18), AtRNL sealing at either a 2Ј,3ЈϾp or 2Ј-phosphate end yields a splice junction that retains the 2Ј-phosphate (1). We conclude that RtcB and AtRNL have antipodal RNA end requirements for sealing, indicative of stark differences in the chemistry of end joining.

Two-step Pathway of End-joining by RNA Ligase RtcB
RtcB Has 2Ј,3Ј-Cyclic Phosphodiesterase Activity-The RNase T1 analysis provided the initially surprising observation that incubation of RtcB with the 5Ј-phosphorylated 20-mer strand with a cyclic phosphate end, although unproductive with respect to ligation, resulted in a chemical transformation at the 3Ј terminus, whereby the original 2Ј,3ЈϾp was converted to a 3Ј-monophosphate (Fig. 3, bottom panel, lanes 7 and 8). Thus, in the absence of a polynucleotide 5Ј-OH, RtcB catalyzes hydrolytic attack on the cyclic phosphodiester. In this experiment, 75% of the input cyclic phosphate was hydrolyzed. We also observed that incubation of RtcB with the 5Ј-phosphorylated 20-mer strand with a 3Ј-monophosphate end resulted in a low but appreciable conversion to a T1 fragment that comigrated with the cyclic phosphate species (Fig. 3, bottom panel, lanes 5  and 6).
The time course of the 2Ј,3Ј-cyclic phosphodiesterase (CPDase) activity of RtcB is shown in Fig. 5B. Opening of the cyclic end was relatively rapid and efficient (65% of the input 2Ј,3ЈϾp substrate was hydrolyzed in 2 min), and the reaction attained an end point with 80% of the cyclic ends opened. The inference that the CPDase reaction is reversible was consistent with the kinetic profile of the reaction of RtcB with a 5Ј-phosphorylated, 3Ј-monophosphate strand, wherein we saw a slow but steady conversion of the 3Ј-phosphate end to a species that comigrated with the 2Ј,3ЈϾp T1 fragment (Fig. 5D). The extent of this conversion was 15% at 45 min. The RtcB reaction mixtures did not include ATP, which is the requisite high energy substrate for RNA 3Ј-cyclization catalyzed by bacterial RtcA (24,25). Given that RtcA gene expression is repressed in E. coli under standard growth conditions (25), and the purity of the recombinant RtcB preparation (Fig. S2), we thought it unlikely that the 3Ј-phosphate cyclization observed for RtcB could be caused by RtcA contamination. To confirm this idea, we purposefully included 0.1 mM ATP in the RtcB reaction mixtures and examined the kinetic profile of cyclization, assuming that ATP would strongly stimulate the generation of cyclic ends if RtcA was the culprit (25). However, we found no effect of ATP on the rate or yield of cyclic phosphate product (not shown). Thus, we surmise that RtcB has an intrinsic CPDase activity that favors the forward hydrolysis reaction under the conditions of our assays.
The temporal profile of the RtcB intramolecular ligation reaction with the 5Ј-OH/2Ј,3ЈϾp 20-mer substrate was consistent with the intermediacy of a 3Ј-monophosphate end in the sealing pathway (Fig. 5A). To wit, the 3Ј-phosphate T1 fragment was more abundant than the ligated junction fragment at the 2-min time point, at which the 3Ј-phosphate and ligated species comprised 27 and 18% of the total radioactivity, respec- tively. The 3Ј-phosphate species declined steadily thereafter (to 11% at 45 min), concomitant with an increase in the abundance of the ligated product (to 52% at 45 min) (Fig. 5A). These results show that RtcB readily executes the hydrolytic CPDase reaction even in the presence of a ligatable 5Ј-OH RNA end.
The kinetic profile for intramolecular sealing of the 5Ј-OH/ 3Ј-monophosphate 20-mer revealed a simple decay of the 3Ј-monophosphate substrate concomitant with accumulation of the ligated junction fragment (Fig. 5C). The ligated species comprised 92% of the radiolabeled products at 45 min. Only scant conversion of the 3Ј-monophosphate to a species comigrating with the 2Ј,3ЈϾp fragment (to ϳ4 -5% of the total label) was detectable from 2 to 45 min (Fig. 5C). These data are consistent with a two-step mechanism of RtcB sealing of 2Ј,3ЈϾp and 5Ј-OH ends, whereby the CPDase activity of RtcB hydrolyzes the 2Ј,3ЈϾp to a 3Ј-monophosphate, after which the ligase activity of RtcB joins the 5Ј-OH end to the 3Ј-phosphate end (Fig. 1B).
We queried the ability of RtcB to use an alternative nonwater nucleophile for its attack on the 2Ј,3Ј-cyclic phosphate terminus, by supplementing the reaction mixtures with increasing concentrations of glycerol (10, 20, and 30% v/v). An apparent glycerololysis reaction was evinced by the appearance of a novel radiolabeled T1 fragment, concomitant with a decline in the abundance of the hydrolysis reaction product HO C-UUpCp (supplemental Fig. S1). The yield of this novel species (presumably HO CUUpC 3Ј p-glycerol) was proportional to glycerol concentration and dependent on RtcB. We surmise that glycerol can compete effectively with water as the nucleophile for the RtcB CPDase reaction, i.e. in the presence of 30% glycerol the distribution of radiolabeled CPDase products was 55% hydrolysis versus 45% glycerololysis (supplemental Fig. S1).
Metal Requirements for the RtcB CPDase and Ligase Activities-We showed previously that sealing of 2Ј,3Ј-cyclic phosphate and 5Ј-OH termini by RtcB depended on manganese (18). Here we examined the metal requirements for the component CPDase and 3Ј-phosphate/5Ј-OH ligation reactions. The CPDase reaction per se was metal-dependent; the metal requirement was best satisfied by 2 mM manganese (Fig. 6A). Zinc, calcium, and magnesium were less effective, nickel and cobalt supported feeble activity, and copper was ineffective (Fig. 6A).
RtcB was more fastidious in its metal requirement for 3Ј-phosphate/5-OH ligation, as assayed by the formation of intramolecular circles (Fig. 6B, top panel) and of RNase T1 fragments derived from the substrate and the ligation junction (Fig.  6B, bottom panel). The ligase metal requirement was satisfied by 2 mM manganese, but not by zinc. Copper, nickel, and cobalt were inactive in ligation, whereas magnesium and calcium supported only low levels of sealing activity above that of the nometal control reaction (Fig. 6B).
Effect of RtcB Mutations on CPDase and Ligase Activities-Previously, we used the crystal structure of Pyrococcus horikoshii RtcB (20) to guide a mutational analysis of conserved residues in E. coli RtcB that we thought might comprise an active site. A water in the putative archaeal RtcB active site, coordinated by the equivalents of essential E. coli RtcB side chains Asp-75, Cys-78, and His-168 (19), was regarded as a potential mimic of the enzyme-bound metal cofactor. Here we purified the mutant C78A and H168A RtcB proteins to apparent homogeneity via the same sequence of chromatography steps and gel-filtration used to purify the wild-type RtcB enzyme (supplemental Fig.  S2) and then tested the wild type and mutants for their ability to execute the component CPDase and 3Ј-phosphate ligation steps of the proposed RtcB RNA repair pathway.
An RtcB titration experiment established that the extent of intramolecular sealing of 3Ј-phosphate and 5Ј-OH ends was proportional to the concentration of wild-type RtcB from 50 to 500 nM, attaining an end point of 98% end joining at 500 -1000 nM (Fig. 7A). By contrast, C78A was inert in 3Ј-phosphate ligation even at 1000 nM, the highest level of protein tested (Fig.  7A). The H168A mutant displayed 3Ј-phosphate ligation activity only at 1000 nM input protein (17% of substrate ligated).

Two-step Pathway of End-joining by RNA Ligase RtcB
Mutational effects on the CPDase reaction are shown in Fig.  7B. Whereas opening of the cyclic phosphate by wild-type RtcB increased in proportion to enzyme concentration and was nearly complete at 500 -1000 nM RtcB, the C78A and H168A mutants were virtually inert at up to 1000 nM input protein. We infer that the CPDase and 3Ј-phosphate ligase reactions are catalyzed by a common active site.
A Role for GTP in 3Ј-Phosphate/5Ј-OH Ligation-In our initial characterization of RtcB, we noted that 0.1 mM GTP stimulated the sealing of a broken RNA stem-loop with 2Ј,3Ј-cyclic phosphate and 5Ј-OH termini (18). Consequently, we have included GTP in the RtcB reaction mixtures in all experiments presented above. The new findings that RtcB ligates a 3Ј-phosphate end to a 5Ј-OH end raised the question of whether GTP might serve as an energy-rich substrate in the ligation reaction, analogous to the way classic RNA ligases and RNA cyclase use ATP for end activation. Thus, we assayed 3Ј-phosphate/5Ј-OH sealing by RtcB in the presence and absence of GTP. The extents of product formation are plotted as a function of input RtcB in supplemental Fig. S3, wherein each datum is the average of three separate enzyme titration experiments. We found that RtcB could execute the 3Ј-phosphate ligation reaction in the absence of GTP; ligation proceeded to near completion at 1000 nM enzyme. However, 3Ј-phosphate ligase activity was stimulated 3-fold by 0.1 mM GTP (supplemental Fig. S3). By contrast, 0.1 mM GDP had little or no stimulatory effect (supplemental Fig. S3).
If the analogy between RtcB and classic ligases is valid, then one predicts that RtcB would react with GTP to form a covalent enzyme-guanylate intermediate (see below), in which case the observed GTP-independent 3Ј-phosphate sealing activity would be executed by a population of pre-formed RtcB-GMP (or RtcB⅐GTP) in the recombinant enzyme preparation. To explore this point further, we modified the purification protocol for recombinant RtcB in an effort to reduce or eliminate any residual nucleotide or metals. To wit: (i) during the second Niagarose step, the flow-through fraction containing tag-free RtcB was immediately adjusted to 20 mM EDTA and incubated for 1 h, and (ii) the EDTA-treated RtcB was then purified by gel filtration through a column of Superdex 200 in buffer containing 1 mM EDTA. This new RtcB enzyme preparation was strictly dependent on added manganese and either GTP or dGTP (0.1 mM) for its 3Ј-phosphate/5Ј-OH ligation activity (Fig. 8). ATP, CTP, and UTP (0.1 mM) were ineffective in lieu of GTP (Fig. 8).
RtcB Reacts with GTP to Form a Covalent RtcB-guanylate Adduct-Incubation of RtcB with 10 M [␣-32 P]GTP in the presence of 2 mM manganese resulted in the transfer of radiolabel to the RtcB polypeptide to form an SDS-resistant adduct that migrated as a single discrete species of ϳ45 kDa during SDS-PAGE (Fig. 9C). The extent of label transfer was proportional to input RtcB concentration (Fig. 9A); ϳ10 -15% of the added RtcB protomers were 32 P-labeled in vitro under these conditions. No label transfer from [␣-32 P]GTP to RtcB occurred in the absence of manganese (Fig. 9A).
We found that no label was transferred to RtcB when the enzyme was reacted with 10 M [␥-32 P]GTP (Fig. 9B). Therefore we surmise that the radiolabeled species is most likely a covalent RtcB-[ 32 P]GMP adduct. There was also no detectable label transfer from 10 M [␣-32 P]ATP to RtcB (Fig. 9B). As a control for the labeling specificity experiment, we tested the [ 32 P]NTPs for label transfer to Chlorella virus DNA ligase (ChV-Lig), a classic ATP-dependent DNA ligase (26) that reacted with 10 M [␣-32 P]ATP to form a covalent ChV-Lig-[ 32 P]AMP intermediate (Fig. 9B). As expected, there was no detectable label transfer to ChV-Lig from either 10 M [␣-32 P]GTP or 10 M [␥-32 P]GTP (Fig. 9). (Note that the specific radioactivity of the GTP and ATP substrates used in these experiments was the same, allowing direct comparisons of the extents of label transfer.) Manganese was the only divalent cation among the seven tested that was effective in supporting the RtcB guanylylation reaction (Fig. 9C). No label transfer to RtcB from [␣-32 P]GTP was seen when manganese was replaced by 2 mM magnesium, calcium, cobalt, nickel, copper, or zinc (Fig. 9C). We proceeded to conduct metal mixing experiments, in which RtcB guanylylation reactions containing 2 mM manganese were supplemented with 2 mM of another divalent cation. Cobalt, nickel, copper, and zinc virtually abolished RtcB guanylylation in the presence of manganese (Fig. 9D), echoing their inhibitory effects on the overall pathway of manganese-dependent 2Ј,3ЈϾp/5Ј-OH RNA sealing (18). By contrast, magnesium and calcium had no effect on RtcB guanylylation in combination with manganese (Fig. 9D), just as they had no negative impact on the overall RNA repair pathway (18). These findings fortify our earlier inferences that the inhibitory "soft" metals (cobalt, nickel, copper, and zinc) out-compete manganese for the His/ Cys-rich active site on the enzyme, wherein engaged they are unable to support reaction chemistry. The C78A active mutation that ablated RtcB ligase activity (see above) also abolished the reaction of RtcB with [␣-32 P]GTP to form the radiolabeled adduct (not shown). Thus, in multiple respects, including metal cofactor specificity, nucleotide specificity, and reliance on Cys-78 in the active site, the properties of the RtcB guanylylation reaction and the RtcB ligation reaction are concordant. Thus, we speculate that the RtcB-GMP complex is a likely intermediate in the 3Ј-phosphate ligation reaction.
Chemical Sensitivity of the RtcB-guanylate Adduct-The nature of the linkage between RtcB and the 32 P-guanylate was evaluated by chemical sensitivity tests (27), whereby the product of the label-transfer reaction was treated with EDTA and SDS and then exposed to acid, alkali, or hydroxylamine. The protein-guanylate adduct was stable to heating for 5 min at 75°C in 0.15 M NaOH (Fig. 10). By contrast, 89% of the adduct was hydrolyzed by heating for 5 min at 75°C in 0.15 M HCl, a treatment to which the RtcB polypeptide was itself stable (Fig.  10). The RtcB-[ 32 P]guanylate adduct was eliminated by exposure for 30 min at 37°C to 0.6 M hydroxylamine (pH 5.0) but was resistant to treatment in parallel with 0.6 M sodium acetate (pH 5.0) (Fig. 10). The chemical lability of the adduct to acid and hydroxylamine and its resistance to alkali are characteristic of a phosphoramidate bond (27). In the case of RtcB, where the active site is surmised from the crystal structure of the Pyrococcus homolog (20) and the mutational analysis inspired thereby (19), we suspect that the guanylate is attached to one of the essential histidine residues.

DISCUSSION
The present study establishes two facts that weight heavily against the one-step direct ligation mechanism depicted in Fig.  1A: (i) RtcB is adept at joining a 3Ј-monophosphate to a 5Ј-OH and (ii) RtcB has a CPDase activity that hydrolyzes a 2Ј,3Ј-cyclic phosphate end to a 3Ј-monophosphate. The available data favor the two-step pathway of RNA sealing depicted in Fig. 1B, in which the CPDase acts upstream of the 3Ј-phosphate ligase and both activities are intrinsic to RtcB. The CPDase of RtcB is envisioned as a straightforward in-line attack by water (or hydroxyl ion) on the cyclic phosphate apical to the ribose O2Ј leaving group. This step is metal-dependent, with manganese being the optimal cofactor. The structure of archaeal RtcB suggests a binding site for one or more metals that is rich in essential histidines and an essential cysteine. The one or more metals could promote CPDase catalysis by coordinating the water/hydroxyl nucleophile, coordinating and promoting expulsion of the ribose 2Ј oxyanion leaving group, and/or coordinating the non-bridging phosphate oxygens and stabilizing the transition state.
The 3Ј-phosphate ligation step of the RtcB pathway presents a challenge to phosphoryl transfer orthodoxy. It is unappealing on energetic grounds to suppose that the ligation reaction occurs by the direct attack of the terminal O5Ј nucleophile on the 3Ј-monophosphate to yield a 3Ј,5Ј-phosphodiester, because such a mechanism calls for the expulsion of one of the non-bridging phosphate oxygens (as hydroxide or, eventually, water) while the bridging phosphomonoester linkage remains intact. Rather, it is more appealing, and less of a departure from classic ligases, to invoke a mechanism entailing RNA 3Ј end activation via one or more high energy inter- mediates. Indeed, we found here that 3Ј-phosphate ligation by RtcB depends on GTP and that RtcB reacts with GTP in the absence of nucleic acid to from a covalent RtcB-guanylate adduct. These properties are reminiscent of the many classic polynucleotide ligases that act via covalent ligaseadenylate intermediates. Extending the analogy further, we speculate that RtcB-guanylate is the first of two activated ligation reaction intermediates, the second being RNA(3Ј)pp(5Ј)G generated by transfer of guanylate from RtcB to the RNA 3Ј-monophosphate terminus. Guanylylation of the RNA 3Ј-phosphate would activate it for attack by an RNA 5Ј-OH end to form the sealed phosphodiester and liberate GMP. This mechanism echoes how RNA cyclase enzymes transfer AMP from a covalent enzyme-adenylate intermediate to an RNA 3Ј-phosphate to form RNA(3Ј)pp(5Ј)A, thereby activating the 3Ј-phosphate for an attack by the vicinal ribose O2Ј to yield a 2Ј,3Ј-cyclic phosphate and liberate AMP (24,28). In this vein, we speculate that the apparent reversal of the CPDase reaction of RtcB (seen when RNA ligation is prevented by a 5Ј-phosphate) reflects the use of the ribose O2Ј as a nucleophile (instead of an RNA 5Ј-OH end) for attack on RNA(3Ј)pp(5Ј)G.
Additional studies will be required to map the site of covalent RtcB guanylylation, validate the intermediacy of RtcB-GMP, and capture the putative (and presumably evanescent) guanylylated RNA intermediate. A complete mechanistic picture will ultimately hinge on obtaining cocrystal structures of RtcB in complexes with manganese, GTP, GMP, and a 3Ј-phosphate RNA substrate.
Finally, the two-step RtcB pathway in Fig. 1B has important implications for the spectrum of RNA repair reactions that RtcB might perform. Specifically, our findings indicate that the substrates for sealing by RtcB need not be limited to RNAs with 2Ј,3ЈϾp ends. RNA-damaging ribotoxins and tRNA-splicing endonucleases characteristically generate 2Ј,3ЈϾp termini as their end-products, and these broken RNAs have been considered the most plausible substrates for splicing or repair by RtcB ligases. However, many other ribonucleases generate 3Ј-monophosphates, which can now be considered potential RtcB substrates. Indeed, our finding of an intrinsic 3Ј-phosphate ligase activity in E. coli RtcB provides a plausible explanation for the results of Popow et al. (17), who partially purified a ligase complex containing mammalian RtcB from a HeLa cell extract based on its ability to join a 3Ј-monophosphate end to a 5Ј-OH end. These investigators documented by mass spectrometry the presence of human RtcB in the ligase preparation, but they did not detect the presence of the human RNA 3Ј-phosphate cyclase protein (Rtc1). Based on our findings, we speculate that human RtcB may also exploit a two-step reaction pathway.