Dual Mechanisms whereby a Broken RNA End Assists the Catalysis of Its Repair by T4 RNA Ligase 2*

T4 RNA ligase 2 (Rnl2) efficiently seals 3′-OH/5′-PO4RNA nicks via three nucleotidyl transfer steps. Here we show that the terminal 3′-OH at the nick accelerates the second step of the ligase pathway (adenylylation of the 5′-PO4 strand) by a factor of 1000, even though the 3′-OH is not chemically transformed during the reaction. Also, the terminal 2′-OH at the nick accelerates the third step (attack of the 3′-OH on the 5′-adenylated strand to form a phosphodiester) by a factor of 25–35, even though the 2′-OH is not chemically reactive. His-37 of Rnl2 is uniquely required for step 3, providing a ∼102 rate acceleration. Biochemical epistasis experiments show that His-37 and the RNA 2′-OH act independently. We conclude that the broken RNA end promotes catalysis of its own repair by Rnl2 via two mechanisms, one of which (enhancement of step 3 by the 2′-OH) is specific to RNA ligation. Substrate-assisted catalysis provides a potential biochemical checkpoint during nucleic acid repair.


Jayakrishnan Nandakumar and Stewart Shuman ‡
From the Molecular Biology Program, Sloan-Kettering Institute, New York, New York 10021 T4 RNA ligase 2 (Rnl2) efficiently seals 3-OH/5-PO 4 RNA nicks via three nucleotidyl transfer steps. Here we show that the terminal 3-OH at the nick accelerates the second step of the ligase pathway (adenylylation of the 5-PO 4 strand) by a factor of 1000, even though the 3-OH is not chemically transformed during the reaction. Also, the terminal 2-OH at the nick accelerates the third step (attack of the 3-OH on the 5-adenylated strand to form a phosphodiester) by a factor of 25-35, even though the 2-OH is not chemically reactive. His-37 of Rnl2 is uniquely required for step 3, providing a ϳ10 2 rate acceleration. Biochemical epistasis experiments show that His-37 and the RNA 2-OH act independently. We conclude that the broken RNA end promotes catalysis of its own repair by Rnl2 via two mechanisms, one of which (enhancement of step 3 by the 2-OH) is specific to RNA ligation. Substrate-assisted catalysis provides a potential biochemical checkpoint during nucleic acid repair.
T4 RNA ligase 2 (Rnl2) typifies a family of structurally homologous RNA-joining enzymes that includes protozoan RNA-editing ligases, Deinococcus radiodurans RNA ligase, vibriophage and mycobacteriophage RNA ligases, and putative RNA ligases of eukaryotic viruses and Archaea (1)(2)(3)(4)(5). Rnl2 joins 3Ј-OH and 5Ј-PO 4 RNA termini through a series of three nucleotidyl transfer steps (6) analogous to those employed by DNA ligases (21). Step 1 is the reaction of ligase with ATP to form a covalent ligase-(lysyl-N)-AMP intermediate and pyrophosphate. In step 2, the AMP is transferred from ligaseadenylate to the 5Ј-PO 4 RNA end to form an RNA-adenylate intermediate (AppRNA). 1 In step 3, attack by an RNA 3Ј-OH on the RNA-adenylate seals the two ends via a phosphodiester bond and releases AMP. Rnl2 and its homologs are particularly adept at joining RNA termini splinted together by a bridging template strand (5,(7)(8)(9).
T4 Rnl2 is a 334-amino-acid monomer composed of two domains. The N-terminal nucleotidyl transferase domain, Rnl2- , is able to perform steps 1 and 3 of the ligation pathway but is defective in the composite 3Ј-OH/5Ј-PO 4 sealing reaction, because the C-terminal domain is needed for Rnl2 to bind to the 3Ј-OH/5Ј-PO 4 nick during the 5Ј-adenylylation step (2,10). The crystal structures of the nucleotidyl transferase domains of T4 Rnl2, Trypanosoma brucei RNA-editing ligase 1, and several DNA ligases highlight a common fold and a conserved AMPbinding pocket composed of five peptide motifs (I, III, IIIa, IV, and V) (reviewed in Ref. 11). The motif I peptide, KX(D/H)G, contains the lysine to which the AMP becomes covalently attached in step 1 of the ligase pathway. Studies of exemplary DNA ligases have highlighted a requirement for covalent adenylylation of the enzyme to recognize the 5Ј-PO 4 nick structure (12,13); this requirement has been attributed to a direct role of the adenylate ribose sugar in coordinating the 5Ј-PO 4 at the nick (14). T4 Rnl2 must also be adenylated at Lys-35 of the 35 KXHG motif to stably bind to a nicked duplex ligand (10).
T4 Rnl2 is incapable of sealing nicked duplex DNA (9, 10). The RNA specificity of Rnl2 arises from a requirement for at least two ribonucleotides immediately flanking the 3Ј-OH of the nick (10). The rest of the nicked duplex can be replaced by DNA. The two key ribonucleotides play distinct roles in the ligation reaction, as surmised from the effects of sugar modifications on the ability of Rnl2-AMP to seal 3Ј-OH/5Ј-PO 4 nicks under single-turnover conditions. Whereas the penultimate 2Ј-OH is important for nick recognition, the terminal 2Ј-OH at the nick is important for the attack of the 3Ј-OH on the 5Јadenylated strand to form a phosphodiester, but seemingly dispensable for nick recognition and adenylylation of the 5Ј-PO 4 strand (10). Because the 2Ј-OH is not a chemical reactant in the sealing step, these results provided initial evidence for substrate-assisted catalysis during ligation by Rnl2. What role does the enzyme play in the sealing step? Preliminary studies indicated that the H37A mutation of motif I ( 35 KXHG) selectively impaired the sealing step of the reaction of Rnl2-AMP with an 18-mer single-stranded 5Ј-PO 4 RNA substrate, resulting in accumulation of the AppRNA intermediate (1,6).
Here, we use chemically modified nicked substrates, mutant enzymes, and transient-state kinetic analysis to address the following mechanistic questions: What are the quantitative contributions of His-37, the ribose hydroxyls at the nick 3Ј terminus, and the divalent cation during the catalysis of the 5Ј-adenylylation and/or phosphodiester synthesis steps of the ligase reaction? What are the contributions of Lys-35 and the C-terminal domain of Rnl2 to the step 3 reaction? Is the requirement for a ribonucleotide on the 3Ј-OH side of the nick inherent to the N-terminal catalytic domain, or is it imposed by the C-terminal domain? Do the functional groups implicated in step 2 and step 3 catalysis act in concert to facilitate a single rate-limiting event or in parallel to facilitate different component substeps of the reactions?

EXPERIMENTAL PROCEDURES
Recombinant Rnl2 Proteins-Full-length wild-type Rnl2 and mutants R266A and D292A, and the N-terminal domain Rnl2-(1-249) were produced in Escherichia coli BL21(DE3) as His 10 -tagged fusions and purified from soluble bacterial extracts by nickel-agarose chromatography as described previously (2,9). Gene fragments encoding N-terminal domain mutants Rnl2-(1-249)-K35A and Rnl2-(1-249)-H37A were amplified by PCR from plasmid templates containing the full-length K35A and H37A mutants (1) using an antisense primer designed to introduce a stop codon in lieu of the Ala-250 codon and a BamHI restriction site immediately 3Ј of the new stop codon. The PCR products were digested with BamHI and NdeI and then inserted into pET16b (Novagen). The inserts of the mutant pET-RNL2 (1-249)-Ala plasmids were sequenced completely to exclude the acquisition of unwanted changes during amplification and cloning. The His 10 -tagged Rnl2-(1-249)-K35A and Rnl2-(1-249)-H37A proteins were produced and purified in parallel with "wild-type" Rnl2-(1-249).
Ligase Substrates-Oligoribonucleotides were purchased from Dharmacon (Lafayette, CO) and deprotected as instructed by the vendor. Oligodeoxyribonucleotides were purchased from BIOSOURCE International. The DNA strands were 5Ј-32 P-labeled using T4 polynucleotide kinase and [␥-32 P]ATP and then purified by electrophoresis through a nondenaturing 18% polyacrylamide gel. To form the nicked substrates, mixtures of 5Ј-PO 4 (or AppDNA) strand, 3Ј-OH strand, and complementary template strand were annealed at a molar ratio of 1:5:2 in 150 mM NaCl, 10 mM Tris-HCl (pH 8.0), 1 mM EDTA by incubation for 10 min at 65°C, followed by incubation for 15 min at 37°C, and then 30 min at 22°C. The ligase substrates were stored at Ϫ20°C and thawed on ice immediately prior to use. Preadenylated 32 P-labeled DNA strands (AppDNA) were prepared by enzymatic adenylylation of 5Ј-32 P-labeled nicked duplexes and gel-purification of the adenylylated strands, as follows. The nicked duplexes were formed by annealing 200 pmol of 5Ј-32 P-labeled 12-mer DNA (pCACTATCGGAAT) or 24-mer DNA (pCACTATCGGAATAAGGGC-GACACG), 1000 pmol of a deoxysubstituted 18-mer RNA, r(UUUAAU-CAAUUGCGACC)(dC), and 400 pmol of a complementary 24-mer DNA d(ATTCCGATAGTGGGGTCGCAATTG) in 120 l of the annealing buffer described above. The annealed substrates were reacted with 1000 pmol of Rnl2-H37A in 150-l mixtures that had been adjusted to 50 mM Tris acetate (pH 6.5), 2 mM MgCl 2 , and 5 mM dithiothreitol. The reactions were quenched after 90 s at 22°C by adding 75 l of 95% formamide, 20 mM EDTA. The mixtures were then heated at 95°C for 5 min, and the products were resolved by electrophoresis through a 40-cm 17% native polyacrylamide gel containing 45 mM Tris borate, 1 mM EDTA buffer. The separated 32 P-labeled AppDNA and pDNA strands were located by autoradiography; a gel slice containing the AppDNA was excised, and the strand was eluted by incubating the slice overnight in 250 l of 10 mM Tris-HCl, 1 mM EDTA.
Single-turnover Ligation at a 5Ј-PO 4 Nick-Reaction mixtures (80 l) containing 50 mM Tris acetate (pH 6.5), 40 mM NaCl, 5 mM dithiothreitol, 1 mM MgCl 2 , 250 nM Rnl2 (corresponding to 125 nM Rnl2-AMP), and 50 nM 5Ј-32 P-labeled nicked duplex substrate as specified were incubated at 22°C. Aliquots (10 l) were withdrawn at the times specified and quenched immediately by adding 10 l of 90% formamide, 20 mM EDTA. The samples were analyzed by electrophoresis through a 14-cm 18% polyacrylamide gel containing 7 M urea in 45 mM Tris borate, 1 mM EDTA for 75 min at 15 watt constant power. The distribution of ligated product and the 5Ј-adenylated intermediate, expressed as the fraction of the total 32 P-labeled material, was quantified by scanning the gel with a Fujix BAS2500 imager.
Ligation at a Preadenylated Nick-Reaction mixtures (80 l) containing 50 mM Tris acetate (pH 6.5), 40 mM NaCl, 5 mM dithiothreitol, 1 mM MgCl 2 , 100 nM Rnl2 or Rnl2-(1-249), and 20 nM 5Ј-adenylated 32 P-labeled nicked duplex substrate were incubated at 22°C. Aliquots (10 l) were withdrawn at the times specified and quenched immediately with formamide/EDTA. The products were resolved by denaturing PAGE, as described above. The accumulation of the ligated product, expressed as the fraction of the total 32 P-labeled material, was quantified by scanning the gel.

RESULTS
The RNA 2Ј-OH and His-37 Promote Phosphodiester Synthesis by Distinct Mechanisms-We showed previously that Rnl2 requires only two terminal ribonucleotides on the 3Ј-OH strand for optimal nick sealing activity and that the penultimate ribonucleotide can be replaced by a 2Ј-OCH 3 nucleotide with no ill effect (10). Here, we prepared singly nicked 24-bp substrates in which the 12-mer 5Ј-PO 4 strand and the 24-mer template strand were all DNA, and the 12-mer 3Ј-OH strand consisted of 10 deoxynucleotides at the 5Ј-end, a 2Ј-OCH 3 nucleotide at the penultimate position flanking the 3Ј-OH end, and a terminal 3Ј-OH nucleoside containing either a 2Ј-OH or a 2Ј-H (Fig. 1). We examined the kinetics of single-turnover nick ligation of these substrates by Rnl2-AMP under conditions of enzyme excess, in the absence of exogenous ATP. The omission of ATP ensures that sealing will be limited to a single round of catalysis by preadenylated Rnl2-AMP, which comprised 50% of the protein in the purified Rnl2 preparation. The sealing of the 2Ј-OH substrate by wild-type Rnl2 was rapid and efficient; 98% of the input 32 P-labeled 12-mer DNA strand was converted to a ligated 24-mer product within 1 min, and no DNA-adenylate intermediate was detected (not shown). The reaction attained 82% of the end point value after 5 s, which was the earliest time analyzed. From this datum, we calculated an apparent rate constant of 0.37 s Ϫ1 for the composite step 2 ϩ 3 reaction.
The H37A mutation of Rnl2 slowed step 3 of the singleturnover ligation reaction at a 2Ј-OH nick, resulting in the transient accumulation of high levels of the DNA-adenylate intermediate (Fig. 1B). DNA-adenylate comprised 78% of the total labeled DNA after 15 s, at which time only 5% of the substrate was sealed. Because the extent of consumption of the input 12-mer substrate DNA strand (the sum of DNA-adenylate and ligated product) was nearly the same after a 5-s reaction of the 2Ј-OH nicked substrate with Rnl2(H37A)-AMP as it was with wild-type Rnl2, we inferred that the loss of the histidine did not interfere significantly with the chemistry of the 5Ј-adenylation step. High levels of AppDNA nick persisted at 30 s and declined steadily thereafter to 7% of the total labeled material at 5 min, concomitant with a progressive increase in abundance of the 24-mer ligation product, which comprised 85% of the labeled material at 5 min. The kinetic data in Fig. 1B were fit using the CKS kinetic simulation program (version 1.0; IBM Corporation) to a unidirectional twostep reaction mechanism with 5Ј-adenylylation and phosphodiester formation rate constants of 0.12 s Ϫ1 and 0.008 s Ϫ1 , respectively. The step 2 rate constant for H37A was one-third of the overall rate of single-turnover nick ligation by wild-type Rnl2, indicating that the loss of the histidine side chain had only a modest impact on step 2 catalysis. In contrast, the step 3 rate constant for H37A was two orders of magnitude slower than lower limit value of 0.9 s Ϫ1 calculated previously for step 3 catalysis by wild-type Rnl2 (10).
The selective effect of the H37A mutation on step 3 resembles the step 3 interference elicited by a 2Ј-H substitution at the terminal nucleoside of the nick (10), which also resulted in transient accumulation of high levels of AppDNA intermediate (Fig.  1A). From a simulation of the data in Fig. 1A, we derived 5Јadenylylation and phosphodiester formation rate constants of 0.30 s Ϫ1 and 0.035 s Ϫ1 , respectively. Thus, the 2Ј-H substitution had little or no effect on step 2, but slowed step 3 by a factor of 25.
Do the RNA 2Ј-OH and His-37 promote phosphodiester synthesis through a common mechanism, or do they separately affect the step 3 reaction? To answer this question, we performed a "biochemical epistasis" experiment in which we examined the kinetics of single-turnover sealing of the 2Ј-Hnicked substrate by the H37A mutant. The logic is as follows. If the 2Ј-OH and His-37 affect the same phase of the step 3 reaction then combining the nick modification with the protein mutation should not elicit a further decrement in the step 3 rate constant, whereas, if the two step 3 catalysts acts via independent mechanisms, we can expect to see additive effects on step 3 when both moieties are removed. We observed (Fig.  1C) that the combination of H37A and a 2Ј-H drastically slowed the overall nick-sealing reaction (note the x axis units are minutes) without affecting the shape of the product distribution curve. The DNA-adenylate intermediate comprised 79% of the total labeled DNA at 15 s and increased to 90% at 0.5-2 min before decaying steadily to 4% after 120 min. From a simulation of the data in Fig. 1C, we determined the 5Ј-adenylylation and phosphodiester formation rate constants of 0.06 s Ϫ1 and 0.0004 s Ϫ1 , respectively. The H37A ϩ 2Ј-H combination resulted in a 2250-fold decrement in step 3 catalysis, a value that agrees well with the theoretical additive effects (110 ϫ 25 ϭ 2750) of the individual 2Ј-OH and H37A modifications. We surmise that the 2Ј-OH and His-37 function as step 3 catalysts through distinct mechanisms.
Analysis of Phosphodiester Synthesis at a Preadenylated Nick-Step 3 of the Rnl2 ligation reaction was assayed in isolation using a preadenylated nicked duplex substrate (Fig.  2). The adenylated strand used to form this substrate was synthesized by ligase-mediated AMP transfer from the H37A mutant to the 5Ј-32 P-labeled 12-mer DNA strand at a nick containing a 3Ј-OH/2Ј-H terminus, i.e. using conditions that we showed were conducive to capturing high levels of the AppDNA intermediate. The radiolabeled AppDNA strand was then gelpurified and annealed to a 24-mer DNA template strand and a 12-mer 3Ј-OH strand containing two ribonucleotides at its 3Јend. Wild-type Rnl2 reacted with this preadenylated substrate in the absence of ATP to form a ligated 24-mer product ( Fig. 2A,  E). Although the reaction was efficient (95% of the input App-DNA strand was sealed in 1 min), the rate of the isolated step 3 reaction (k obs ϭ 0.042 s Ϫ1 ) was paradoxically slower than the rate of single-turnover ligation at a 3Ј-OH/5Ј-PO 4 nick (0. 37   FIG. 2. Kinetic analysis of phosphodiester formation at a  preadenylated nick (step 3). A, the structure of the three-piece preadenylated nicked substrate is illustrated, with the two terminal ribonucleotides of the 3Ј-OH strand shaded in gray. All other positions are deoxyribonucleotides, except for the 5Ј-adenylate. Reaction mixtures contained 20 nM 5Ј-adenylated 32 P-labeled substrate and 100 nM full-length wild-type Rnl2, mutant R266A or D292A, or 100 nM Nterminal domain Rnl2- (1-249). B, reaction mixtures contained 20 nM 5Ј-adenylated 32 P-labeled substrate depicted in A and 100 nM K35A or H37A mutants of Rnl2-(1-249). C, the structures of the preadenylated nicked substrates are shown with either one or two terminal ribonucleotides of the 3Ј-OH strand shaded in gray. Reaction mixtures contained 100 nM Rnl2-(1-249) and 20 nM 5Ј-adenylated 32 P-labeled substrate. The accumulation of the ligated 24-mer product, expressed as the fraction of the total 32 P-labeled material, is plotted as a function of time. s Ϫ1 ). The slow step 3 problem in sealing a preadenylated nick was noted previously for DNA ligases (15,16) and attributed to the imposition of a new rate-limiting conformational step for productive binding of the exogenous AppDNA, which does not pertain when the AppDNA intermediate is formed in situ on the ligase by catalysis of step 2.
An instructive finding was that the slow step 3 phenotype was alleviated by deleting the C-domain of Rnl2 ( Fig. 2A, Ⅺ). Sealing of the preadenylated nick by the isolated N-terminal domain Rnl2-(1-249) was efficient (95% ligation in 15 s) and rapid (90% of the end point value attained in 5 s). We calculated an apparent step 3 rate constant of 0.4 s Ϫ1 for Rnl2-(1-249). This value, which is a lower limit estimate, is ϳ10-fold faster than the rate observed for full-length Rnl2. We concluded that the C-terminal domain is an impediment to step 3 catalysis on an exogenous preadenylated substrate. Previously, we identified two mutations in the C-terminal domain, R266A and D292A, that phenocopied the C-domain deletion with respect to loss of overall activity in nick joining, despite retention of adenylyltransferase activity (9). Here we found that the R266A and D292A mutations elicited an acceleration of step 3 catalysis identical to that observed for Rnl2-(1-249) ( Fig. 2A).
To gauge the contribution of His-37 to the isolated step 3 reaction, we introduced the H37A change into the truncated Rnl2-(1-249) protein and performed a kinetic analysis of its ability to catalyze sealing at a preadenylated nick (Fig. 2B). The apparent step 3 rate constant of 0.0082 s Ϫ1 for the Rnl2-(1-249)-H37A mutant was in excellent agreement with the step 3 rate constant of 0.008 s Ϫ1 determined for full-length H37A-AMP in the single-turnover step 2 ϩ 3 reaction (Fig. 1B). In parallel, we tested the role of the motif I lysine in the step of phosphodiester formation by introducing the K35A mutation into the Rnl2-(1-249) protein. The apparent step 3 rate constant of the Rnl2-(1-249)-K35A mutant was 0.011 s Ϫ1 (Fig. 2B). This result shows that the lysine adenylylation site, which is strictly essential for steps 1 and 2 of the ligation pathway, is not absolutely required for step 3, although it does contribute a ϳ35-fold rate enhancement to phosphodiester synthesis.
Earlier studies highlighted the importance of a 2Ј-OH at the penultimate nucleotide on the 3Ј-OH side of the nick for nick recognition and catalysis of 5Ј-adenylylation (10). Here we addressed what contributions, if any, are made by the penultimate 2Ј-OH during the sealing of a preadenylated nick by preparing an otherwise identical 24-bp 5Ј-adenylylated nicked substrate that contained only a single ribonucleotide on the 3Ј-OH side of the nick (Fig. 2C, E). The penultimate 2Ј-H substitution had no effect on the extent of the step 3 reaction of Rnl2-(1-249) and exerted only a 2-fold effect on the step 3 rate constant (k obs ϭ 0.17 s Ϫ1 ) (Fig. 2C). We conclude that the penultimate 2Ј-OH is not involved directly in step 3 catalysis.
The Terminal 2Ј-OH Accelerates Phosphodiester Synthesis at a Preadenylated Nick-To gauge the contribution of the terminal 2Ј-OH to the isolated step 3 reaction, we prepared preadenylated nicked duplexes composed of a radiolabeled 24-mer AppDNA strand, a bridging DNA template strand, and either an all RNA 3Ј-OH strand (Fig. 3, f) or a 3Ј-OH RNA strand containing a single terminal 2Ј-H nucleoside (Fig. 3, E). Although RNA-to-AppDNA joining by Rnl2-(1-249) was efficient and rapid (k obs ϭ 0.32 s Ϫ1 ), a single deoxynucleotide at the 3Ј terminus of the nick slowed step 3 by a factor of 35 (k obs ϭ 0.009 s Ϫ1 ) without affecting the end point (Fig. 3). The 35-fold rate enhancement by the 2Ј-OH in the catalysis of step 3 in isolation was in good agreement with the rate acceleration of 25-fold determined for phosphodiester synthesis starting with a 5Ј-PO 4 nick.
Contributions of the Divalent Cation during Catalysis of Steps 2 and 3-The composite Rnl2 ligation reaction requires a divalent cation cofactor. To assess the contributions of the metal to catalysis of steps 2 and 3, we reacted excess Rnl2-AMP with the 24-bp substrate containing a 3Ј-ribonucleotide/5Ј-PO 4 nick in the absence of exogenous divalent cation. The Rnl2 enzyme preparation used in this experiment had been dialyzed against a buffer containing EDTA. In addition, we included 10 mM EDTA in the reaction mixture. We found that Rnl2 was capable of adenylylating the nicked substrate to form AppDNA under these conditions (Fig. 4A). The extent of the step 2 reaction was high (95% of the input pDNA strand was converted to AppDNA in 4 h), but the rate was very slow. The data in Fig. 4A are fit to a single exponential with a step 2 rate constant of 0.00097 s Ϫ1 . By comparing this value to the rate of single-turnover nick sealing in the presence of magnesium (0.34 s Ϫ1 ), we concluded that the metal cofactor accelerates step 2 by a factor of 380. We detected no formation of ligated 24-mer in the ϪMg/ϩEDTA reaction, even when the incubation was extended to 24 h (data not shown). Given that virtually all of the nicks were adenylylated by 2-4 h, this result suggests that the ligase has a more stringent requirement for a metal cofactor during the step of phosphodiester formation than during the 5Ј-adenylylation step. We confirmed this inference by analyzing the ability of excess Rnl2-(1-249) to seal a preadenylated nicked 24-bp duplex in the absence of added magnesium, whereby we detected no ligation product even after a 5-h reaction (data not shown).
The Nick 3Ј-OH Is Critical for Catalysis of the 5Ј-Adenylylation Reaction-The experiment in Fig. 4B shows that elimination of the 3Ј-OH at the nick completely blocked step 3 catalysis (as expected) but still permitted the execution of step 2, with a conversion of Ͼ90% of the input 5Ј-PO 4 DNA strand to the AppDNA intermediate. However, the rate constant of the step 2 reaction at a 3Ј-H terminus (3.4 ϫ 10 Ϫ4 s Ϫ1 ) was slower by a factor of 1000 than the step 2 rate at a nick containing a 3Ј-terminal ribonucleotide. This result is revealing mechanistically, insofar as the 3Ј-OH is itself not chemically reactive during the step of polynucleotide 5Ј-adenylylation. Thus, we have evidence for significant substrate-assisted catalysis by the 3Ј-OH during step 2, in addition to the substrate-assisted catalysis of step 3 by the 2Ј-OH described above.
Given the similar effects on step 2 catalysis elicited by the elimination of the metal cofactor and loss of the 3Ј-OH, we considered the possibility that the 3Ј-OH acts by coordinating the metal, and we conducted an epistasis experiment to test this idea, i.e. by analyzing the reaction of Rnl2 with the 3Ј-H nick in the absence of magnesium. We observed no formation of App-DNA during a 24-h reaction (data not shown), signifying that the metal ion and the 3Ј-OH likely promote step 2 catalysis independently.

DISCUSSION
The RNA specificity determinants for Rnl2 reside solely on the 3Ј-OH side of the nick. Here we showed that the ribose 3Ј-OH and 2Ј-OH at the broken RNA end promoted catalysis of the 5Ј-adenylylation and strand sealing reactions, respectively. Because neither hydroxyl is chemically transformed during the step that it promotes, we concluded that Rnl2 employs a mechanism of substrate-assisted catalysis. Substrate-assisted catalysis has been invoked for a variety of natural and engineered enzymes, including serine proteases, GTPases, acylphosphatases, DNA glycosylases, and the ribosome (22)(23)(24)(25)(26).
What distinguishes Rnl2 is its reliance on its nucleotide and polynucleotide substrates to drive at least three phases of its reaction with damaged RNA. First, Rnl2 must be adenylylated to recognize the nicked duplex, i.e. nick binding was virtually abolished by an alanine mutation of the Lys-35 nucleophile that forms the Rnl2-AMP intermediate (10). Recognition of DNA nicks by DNA ligases is similarly abolished when their active site lysines are changed to alanine (12,13). This mechanism ensures that ligase-adenylate is not competing with ligase apoenzyme for binding to sites in need of repair. Ligase-AMP is poised to catalyze phosphodiester formation, whereas binding the apoenzyme to the nick would block ingress of ATP to the adenylate binding pocket. Nick recognition by Rnl2 also depends on the C-terminal domain of the protein and the penultimate ribonucleotide flanking the 3Ј-OH side of the nick (10). Here we showed that the requirements for recognition of the 3Ј-OH/5Ј-PO 4 nick can be circumvented, either completely or partially, when the nick is preadenylylated. This is illustrated most clearly by the ability of the N-terminal domain, which is defective in binding and ligation of a 3Ј-OH/5Ј-PO 4 nick (2,10), to rapidly seal a nicked polynucleotide-adenylate substrate. We surmised from this result that occupancy of the adenylate binding pocket of Rnl2-(1-249) suffices to position the covalently tethered nicked duplex in the active site for phosphodiester synthesis, which resides entirely with the Nterminal domain. Indeed, we observed that the isolated Nterminal domain was more active on the exogenous nicked polynucleotide-adenylate than was full-length Rnl2. Point mutations R266A or D292A in the C-terminal domain of Rnl2 elicited the same gain of function in sealing a preadenylated nick observed for a complete deletion of the domain. Thus, the C-terminal domain of Rnl2 is responsible for much of the ratelimiting impediment to binding and sealing the exogenous 5Јadenylated substrate. A rate-limiting impediment to the isolated step 3 reaction by the C-terminal protein segment was also observed for the Chlorella virus DNA ligase (15).
Second, the transfer of the adenylate from Rnl2-AMP to the 5-PO 4 end of the nicked duplex is critically dependent on the terminal 3Ј-OH of the nick. Loss of the 3Ј-OH group slows the rate of step 2 by three orders of magnitude (without affecting the final extent of 5Ј-adenylylation), even though the 3Ј-OH is not chemically transformed during the step 2 reaction. This phenomenon can be viewed as a kinetic checkpoint mechanism, which helps ensure that 5Ј-end activation is confined to sites where a reactive 3Ј-OH is in place and the ligation reaction can continue through the step of phosphodiester synthesis. Absent this checkpoint, Rnl2 (or other ligases) would run the risk of adenylylating 5Ј-PO 4 termini at gaps in duplex structures or at single-stranded ends. The resulting AppDNA or AppRNA ends might be difficult to repair, given that (i) polynucleotide ligases are predominantly in the adenylylated state at physiological ATP concentrations, (ii) ligase-adenylate can neither seal nor deadenylate an AppDNA/RNA terminus, and (iii) the AppDNA/ RNA end would be resistant to processing by 5Ј-exonucleases. Thus, substrate-assisted catalysis of step 2 by the 3Ј-OH focuses Rnl2 function on nucleic acid repair and diverts it from generating potentially deleterious 5Ј-lesions.
The 3Ј-OH step 2 checkpoint mechanism invoked for Rnl2 likely extends to DNA ligases, insofar as the 5Ј-adenylylation reactions of Chlorella virus, vaccinia virus, and human DNA ligases were blocked by a 3Ј-H substitution at the nick (17,18). Although an extended kinetic analysis of the reaction of the viral and human DNA ligases with a 3Ј-H nick was not conducted, early studies of E. coli DNA ligase noted that the rate of 5Ј-adenylylation of a substrate containing a 3Ј-H end was at least three orders of magnitude slower than the rate of strand joining with a 3Ј-OH substrate (16). Thus, Rnl2 and bacterial DNA ligase display a similar kinetic dependence on the 3Ј-OH for substrate-assisted catalysis of step 2. The recent report that the nicked DNA-adenylate intermediate was captured during crystallization of human DNA ligase I in the presence of ATP, magnesium, and a nicked ligand containing a 3Ј-H terminus suggests that the human enzyme was able to slowly adenylylate a 3Ј-H nick (19).
Third, Rnl2 relies on the terminal 2Ј-OH for catalysis of phosphodiester synthesis. This phase of substrate-assisted repair is unique to RNA ligase. Loss of the 2Ј-OH results in an accumulation of the polynucleotide 5Ј-adenylate intermediate during single-turnover ligation of a 3Ј-OH/5Ј-PO 4 nick by Rnl2, reflecting at least a 25-fold decrement in the rate of step 3 compared with the lower limit estimate of the step 3 rate constant determined previously (10). Our finding that a single deoxy substitution at the 3Ј nick terminus slowed the rate of FIG. 4. Kinetic analysis of nick 5-adenylylation. A, step 2 catalysis in the absence of magnesium. Reaction mixtures contained 250 nM Rnl2 and 50 nM 5Ј-32 P-labeled 24-bp nicked duplex containing a 3Ј-OH strand with a single penultimate 2Ј-OCH 3 nucleotide and a 3Ј-terminal ribonucleotide (see Fig. 1). Magnesium was omitted, and the mixture was supplemented with 10 mM EDTA. B, effect of a terminal 3Ј-deoxy modification. Reaction mixtures contained 250 nM Rnl2, 1 mM MgCl 2 , and 50 nM 5Ј-32 P-labeled 24-bp nicked duplex containing a 3Ј-OH strand with a single penultimate 2Ј-OCH 3 nucleotide and a 2Ј-OH/3Ј-H terminal nucleotide at the nick. The accumulation of the 5Ј-adenylylated 12-mer strand, expressed as the fraction of the total 32 P-labeled material, is plotted as a function of time.
phosphodiester synthesis by Rn2-(1-249) by a factor of 35 provides clear evidence that RNA specificity is inherent to the N-terminal catalytic domain.
His-37 in motif I contributes an ϳ100-fold rate enhancement of step 3 catalysis. An epistasis experiment showed that the 2Ј-OH and His-37 act independently to promote phosphodiester synthesis. It is noteworthy that the corresponding motif I Asp side chain of Chlorella virus DNA ligases accelerates phosphodiester synthesis by a factor of 60 (20), which attests to the conservation of function of the motif I Asp/His groups during step 3 of the RNA and DNA ligase reactions. However, Rnl2 and Chlorella virus DNA ligase diverge sharply in their reliance on the motif I Asp/His during step 2. Although loss of His-37 had almost no impact on the rate of the 5Ј-adenylylation reaction of Rnl2, the equivalent mutation of the Chlorella virus DNA ligase motif I Asp slowed the rate of single-turnover ligation by a factor of 6000 (17).
Finally, our experiments highlighted a quantitatively distinct dependence of the 5Ј-adenylylation and phosphodiester synthesis reactions on an exogenous divalent cation. In the absence of magnesium, step 2 catalysis is slowed by a factor of 380, whereas step 3 catalysis was undetectable. An epistasis experiment showed that the metal ion and the nick 3Ј-OH act independently to promote the 5Ј-adenylylation reaction.