Structure-Function Analysis of T4 RNA Ligase 2*

Bacteriophage T4 RNA ligase 2 (Rnl2) exemplifies a polynucleotide ligase family that includes the trypanosome RNA-editing ligases and putative RNA ligases encoded by eukaryotic viruses and archaea. Here we analyzed 12 individual amino acids of Rnl2 that were identified by alanine scanning as essential for strand joining. We determined structure-activity relationships via conservative substitutions and examined mutational effects on the isolated steps of ligase adenylylation and phosphodiester bond formation. The essential residues of Rnl2 are located within conserved motifs that define a superfamily of nucleotidyl transferases that act via enzyme-(lysyl-N)-NMP intermediates. Our mutagenesis results underscore a shared active site architecture in Rnl2-like ligases, DNA ligases, and mRNA capping enzymes. They also highlight two essential signature residues, Glu34 and Asn40, that flank the active site lysine nucleophile (Lys35) and are unique to the Rnl2-like ligase family.

Bacteriophage T4 RNA ligase 2 (Rnl2) exemplifies a polynucleotide ligase family that includes the trypanosome RNA-editing ligases and putative RNA ligases encoded by eukaryotic viruses and archaea. Here we analyzed 12 individual amino acids of Rnl2 that were identified by alanine scanning as essential for strand joining. We determined structure-activity relationships via conservative substitutions and examined mutational effects on the isolated steps of ligase adenylylation and phosphodiester bond formation. The essential residues of Rnl2 are located within conserved motifs that define a superfamily of nucleotidyl transferases that act via enzyme-(lysyl-N)-NMP intermediates. Our mutagenesis results underscore a shared active site architecture in Rnl2-like ligases, DNA ligases, and mRNA capping enzymes. They also highlight two essential signature residues, Glu 34 and Asn 40 , that flank the active site lysine nucleophile (Lys 35 ) and are unique to the Rnl2-like ligase family.
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). In step 3, attack by an RNA 3Ј-OH on the RNA-adenylate seals the two ends via a phosphodiester bond and releases AMP. Bacteriophage T4 RNA ligase 1 (Rnl1) is the founding member of the RNA ligase family (1). The active site lysine of Rnl1 is located within a conserved sequence element, KX(D/N)G (motif I), that defines a superfamily of covalent nucleotidyl transferases, which includes DNA ligases and mRNA capping enzymes (6 -8). DNA ligases and capping enzymes share a common tertiary structure composed of five conserved motifs (I, III, IIIa, IV, and V) that contain amino acid side chains responsible for nucleotide binding and catalysis (9 -12) (Fig. 1). It has been suggested that DNA ligases and capping enzymes evolved from a common ancestral nucleotidyl transferase, possibly from an ancient RNA strand-joining enzyme. Yet, the structural basis for catalysis by RNA ligases remains ill-defined because few RNA ligase enzymes have been studied, and there is scant mutational analysis available outside of the KX(D/N)G motif.
We recently identified and characterized a second T4 RNA ligase (Rnl2) encoded by T4 gene 24.1. Purified Rnl2 catalyzes intramolecular circularization and intermolecular dimeriza-tion of single-stranded RNA through ligase-adenylate and RNA-adenylate intermediates (13). Rnl2 exemplifies a distinct subfamily of RNA ligases defined by variant nucleotidyl transferase motifs that includes the RNA-editing ligases of Trypanosoma and Leishmania, putative RNA ligases encoded by eukaryotic viruses (baculoviruses and an entomopoxvirus), and putative RNA ligases encoded by many species of archaea (13)(14)(15)(16). Thus, the Rnl2-like ligases are present in all three phylogenetic domains. T4 RNA ligase 1 exemplifies a separate subfamily of RNA ligases with a narrower phylogenetic distribution.
Alignment of the primary structures of several of the Rnl2like ligases highlights two notable features: (i) a defining variant of motif I, EKX(H/D)XN, and (ii) the presence of putative counterparts of nucleotidyl transferase motifs III, IIIa, IV, and V found in DNA ligases and capping enzymes (Fig. 1). To understand the structural requirements for catalysis by the Rnl2-like ligases, we initiated an alanine-scanning mutational analysis of selected residues in the nucleotidyl transferase motifs of Rnl2 (13). We showed that Lys 35 in motif I, Glu 204 in motif IV, and Lys 225 and Lys 227 in motif V are required for overall strand joining and particularly for the enzyme adenylation reaction (step 1) of the ligation pathway. His 37 in motif I is not required for ligase adenylation but plays a critical role downstream of step 1. These initial results suggested that the partial reactions of the RNA ligation pathway may be catalyzed by distinct constellations of active site residues.
To further delineate which of the conserved side chains of the nucleotidyl transferase motifs are functionally relevant, we have extended the mutational analysis of T4 Rnl2, focusing on residues in motifs I, III, IIIa, IV, and V (indicated by q in Fig.  1). We identified individual amino acids that are required for strand joining, determined structure-activity relationships via conservative substitutions, and then stratified the mutational effects on the isolated steps of ligase adenylylation (step 1) and phosphodiester formation (step 3). We find that all five motifs are essential for covalent nucleotidyl transfer by Rnl2, as they are for DNA ligases and capping enzymes, and we locate essential structural "signatures" that are unique to Rnl2-like ligases.

EXPERIMENTAL PROCEDURES
Rnl2 Mutants-Amino acid substitution mutations were introduced into the rnl2 gene by PCR using the two-stage overlap extension method as described previously (13). The PCR products were digested with NdeI and BamHI and inserted into pET16b (Novagen). The inserts of the mutant pET-RNL2 plasmids were sequenced completely to exclude the acquisition of unwanted changes during amplification and cloning. pET-RNL2 plasmids were transformed into Escherichia coli BL21(DE3). Induction of Rnl2 production with isopropyl-1-thio-␤-Dgalactopyranoside and purification of Rnl2 from soluble bacterial extracts by nickel-agarose chromatography were performed as described previously (13). The wild-type and mutant Rnl2 preparations were stored at Ϫ80°C.
Adenylyltransferase Assay-Reaction mixtures (20 l) containing 50 * This work was supported by National Institutes of Health Grant GM63611. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. mM Tris acetate (pH 6.5), 5 mM DTT, 1 1 mM MgCl 2 , 20 M [␣-32 P]ATP, and Rnl2 as specified were incubated for 5 min at 37°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 PhosphorImager.
RNA Ligase Assay-An 18-mer oligoribonucleotide was 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 two-piece stem-loop substrate (see Fig. 2), 100 pmol of the gel-purified 32 P-labeled 18-mer strand was mixed with 500 pmol of a partially complementary 15-mer RNA in buffer containing 10 mM Tris-HCl (pH 8.0), 0.1 M NaCl, 1 mM EDTA. The mixture was heated at 60°C for 10 min, shifted to 37°C for 15 min, and then cooled slowly to 22°C. RNA ligation reaction mixtures (10 l) containing 50 mM Tris acetate (pH 6.5), 5 mM DTT, 1 or 2 mM MgCl 2 , 1 pmol of 5Ј 32 P-labeled RNA, and ATP and Rnl2 as specified were incubated for 15 min at 22°C. The reactions were quenched by adding 5 l of 90% formamide, 20 mM EDTA. The samples were analyzed by electrophoresis through an 18% polyacrylamide gel containing 7 M urea in 45 mM Tris borate, 1 mM EDTA. The ligation products were visualized by autoradiography.
Preparation of RNA-Adenylate-2 nmol of 32 P-labeled 18-mer RNA was incubated with 50 g of purified Rnl2 in the presence of 1 mM ATP and 7.5 mM MgCl 2 for 30 min at 22°C. The products were then treated with 2.5 units of calf intestine alkaline phosphatase (Roche Applied Science) for 60 min at 37°C. The digestion products were resolved by electrophoresis through a native 18% polyacrylamide gel in 90 mM Tris borate, 2 mM EDTA. The 32 P-labeled AppRNA strand was located by autoradiography of the wet gel and then eluted from an excised gel slice in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA. The two-piece stem-loop AppRNA substrate (see Fig. 4B) was formed by annealing the gelpurified 32 P-labeled AppRNA strand to an unlabeled 15-mer strand at a molar ratio of 1:5.
Ligation of RNA-Adenylate-Reaction mixtures (10 l) containing 50 mM Tris acetate (pH 6.5), 5 mM DTT, 2 mM MgCl 2 , 0.2 pmol of 32 Plabeled AppRNA, and Rnl2 as specified were incubated for 15 min at 22°C. The reactions were quenched by adding 5 l of 90% formamide and 20 mM EDTA, and the samples were analyzed by electrophoresis through a 18% polyacrylamide gel containing 7 M urea in 45 mM Tris borate, 1 mM EDTA. The products were visualized by autoradiography of the gel.

RESULTS
Two-piece Strand Joining by Rnl2-The strand joining activity of T4 Rnl2 was demonstrated initially with a 5Ј 32 Plabeled single-stranded 18-mer RNA substrate. Cyclization of the 18-mer RNA was the predominant outcome of the ligation reaction rather than formation of linear multimers (13). The same preference for cyclization is displayed by T4 Rnl1 with substrates of similar size (3) and is construed to reflect proximity of the intramolecular 3Ј-OH terminus to the active site. The function of Rnl1 in vivo is to repair a break in the anticodon loop of E. coli tRNA Lys triggered by phage activation of a host-encoded anticodon nuclease (17,18). Thus, the physiological substrate for Rnl1 is a folded two-piece RNA molecule with the reactive ends held in proximity by the secondary structure of the tRNA. It was of interest to determine whether Rnl2 was capable of joining such a substrate or whether its action was confined to single-stranded RNAs. We constructed a two-piece substrate designed to mimic the anticodon stem and broken anticodon loop of tRNA Lys (17) via hybridization of a 5Ј 32 Plabeled 18-mer RNA to a complementary 15-mer RNA with 5Ј-OH and 3Ј-OH termini (Fig. 2). The broken loop of the synthetic two-piece substrate consists of a two-nucleotide single-stranded tail on one strand of the stem duplex that provides the ligatable 3Ј-OH and a 5Ј-nucleotide single strand tail on the other strand that provides the ligatable 5Ј-PO 4 end (17). Reaction of this substrate with Rnl2 resulted in joining of the 5Ј 32 P-labeled 18-mer to the unlabeled 15-mer to form a 33-mer ligation product (Fig. 2, RNAЈpRNA). Thus, Rnl2 is capable of joining ends that are brought together by secondary structure. Enzyme titration experiments showed that ligation of the stemloop substrate occurred with efficiency comparable with circularization of an 18-mer single strand (data not shown).
The yield of the two-piece ligation product was optimal between pH 6.0 and 7.0. Reducing the pH to 5.5 or 5.0 suppressed strand joining while stimulating the formation of the RNAadenylate intermediate (Fig. 2). Similar pH effects on AppRNA accumulation were noted previously for Rnl2-mediated circularization of a single-stranded RNA substrate (13). We conclude that the step of phosphodiester bond formation becomes rate-limiting at mildly acidic pH, independent of the type of RNA substrate used. Raising the pH to Ն7.5 suppressed the 1 The abbreviation used is: DTT, dithiothreitol. two-piece ligation reaction without trapping the RNA-adenylate intermediate (Fig. 2).
We reported previously that the inclusion of 1 mM ATP in the strand joining reaction with a single-stranded RNA substrate promoted accumulation of AppRNA and suppressed formation of ligated circles (13). The explanation offered for the ATP effect is that Rnl2, like Rnl1 (3), is prone to dissociate from the newly formed RNA-adenylate product of step 2 and that an immediate reaction of Rnl2 with ATP to form ligase-adenylate precludes it from rebinding to the RNA-adenylate for subsequent catalysis of strand joining. The experiment in Fig. 3 shows a smooth transition from circular product to AppRNA product as the ATP concentration was increased, with a midpoint at ϳ10 M ATP. The ATP concentration dependence of the trapping of RNA-adenylate roughly parallels the ATP concentration dependence of ligase-adenylate formation, which is saturated at 20 M ATP (13). We considered the possibility that Rnl2 might be less prone to dissociate from the two-piece stemloop substrate in which the 3Ј-OH end is tethered in the vicinity of the 5Ј end, but found that inclusion of ATP in the twopiece ligation reaction had the same effect of trapping RNAadenylate as it did with the single-stranded RNA substrate (not shown). Thus, we surmise that Rnl2 is generally liable to dissociate from the step 2 product under the reaction conditions employed.
Phosphodiester Formation at a Preadenylated RNA 5Ј End-The production of high levels of RNA-adenylate by Rnl2 in the presence of ATP allowed us to synthesize and gel purify a preadenylated RNA substrate (AppRNA) for analysis of step 3 of the ligation pathway in isolation. Formation of a phosphodi-ester at the activated 5Ј end was manifest by the appearance of a sealed circular RNA product, the yield of which was proportional to the amount of input Rnl2 (Fig. 4A). More than 90% of the substrate was converted to circular RNA at saturating Rnl2 concentrations. Circularization of RNA-adenylate required a divalent cation cofactor (data not shown). A small fraction of the input AppRNA was apparently deadenylated during the reaction to yield pRNA, which migrated between RNA-adenylate substrate and the ligated circle (Fig. 4A). Deadenylation is the reverse of step 2 of the ligation pathway. Circularization of AppRNA was optimal at pH 6.0 to 7.0 and fell off sharply at a more acidic pH; activity declined gradually at alkaline pH such that the yield of circular product at pH 8.5 was ϳ25% of the value at pH 7.0 (data not shown).
We also prepared a preadenylated version of the two-piece stem-loop substrate. Rnl2 catalyzed efficient joining of the activated 5Ј end of the labeled 18-mer to the 3Ј-OH of the unla- FIG. 4. Ligation of a preadenylated RNA substrate. A, the 5Ј adenylated 32 P-labeled 18-mer RNA strand was synthesized and gelpurified as described under "Experimental Procedures." The two potential reactions of Rnl2 with the isolated RNA-adenylate are illustrated schematically.
Step 3 entails attack in cis of the 3Ј-OH on the RNAadenylate to form a circular RNA product; step Ϫ2 (reverse of step 2) is the transfer of AMP from RNA-adenylate to the ligase to form a linear 5Ј phosphate RNA. Reaction mixtures (10 l) containing 50 mM Tris acetate (pH 6.5), 5 mM DTT, 2 mM MgCl 2 , 0.2 pmol of 5Ј 32 P-labeled 18-mer AppRNA, and 5, 10, 20, 40, 80, or 160 ng of Rnl2 (increasing from left to right) were incubated for 15 min at 22°C. Rnl2 was omitted from a control reaction (lane indicated by dash). B, the 32 P-labeled 18-mer AppRNA was hybridized to an unlabeled 15-mer RNA to form the stem-loop RNA-adenylate substrate illustrated at the bottom. Reaction mixtures (10 l) containing 50 mM Tris acetate (pH 6.5), 5 mM DTT, 2 mM MgCl 2 , 0.2 pmol of 5Ј 32 P-labeled 18-mer AppRNA, and 10, 20, 40, or 80 ng of wild-type (WT) Rnl2 or the K35A mutant (increasing from left to right in each series) were incubated for 15 min at 22°C. A control reaction lacking Rnl2 is shown in lane indicated by dash. The reaction products were analyzed by PAGE and visualized by autoradiography. beled 15-mer to yield a 33-mer product (Fig. 4B). Again, a small fraction of the substrate was deadenylated to produce pRNA. Also, a trace amount of circular product was formed, most likely reflecting the presence of a small fraction of unhybridized 18-mer AppRNA strand in the substrate preparation.
Lys 35 is the site of covalent attachment of AMP to Rnl2; its replacement by alanine abolishes ligase-AMP formation and, perforce, the formation of the RNA-adenylate intermediate (13). An instructive finding was that the K35A mutant of Rnl2 was capable of sealing a preadenylated two-piece stem-loop RNA substrate, albeit less efficiently than did wild-type Rnl2 (Fig. 4B). As expected, the K35A mutant was unable to deadenylate the AppRNA substrate to form the pRNA species seen in the wild-type Rnl2 reaction. This result underscores that the lysine nucleophile is not strictly essential for the chemical step of phosphodiester bond formation. Similar results concerning the ability of motif I lysine mutants to catalyze phosphodiester bond formation have been reported for ATP-dependent DNA ligases (19 -21).
New Alanine-scanning Mutagenesis-The five putative nucleotidyl transferase motifs of Rnl2 and related ligases are highlighted in Fig. 1. Motif I, which contains the active site lysine, adheres to the consensus sequence EKX(H/D)GXN, which differs from the XKXDGXR sequence characteristic of ATP-dependent DNA ligases and most mRNA capping enzymes (22). To evaluate the functional relevance of the signature Glu 34 and Asn 40 side chains, we substituted each with alanine. Alanine changes were also introduced at five other conserved positions: Arg 55 , located between motifs I and III; Glu 99 in motif III; Phe 119 and Asp 120 in motif IIIa; and Lys 209 flanking motif IV. A single nonconserved residue, Lys 189 , was also changed to alanine. The E34A, N40A, R55A, E99A, F119A, D120A, K189A, and K209A mutants were produced in E. coli as His 10 -tagged fusions and purified from soluble bacterial extracts by nickel-agarose chromatography (Fig. 5A). The 42-kDa Rnl2 polypeptide was the predominant species detected by SDS-PAGE, and the extents of purification were comparable for mutant and wild-type Rnl2 (Fig. 5A). The D120A mutation elicited a reproducible increase in the electrophoretic mobility of Rnl2, apparently caused by the loss of the acidic carboxylate moiety (see below).
The adenylyltransferase activity of recombinant Rnl2 was assayed by label transfer from [␣-32 P]ATP to the Rnl2 polypeptide to form the covalent enzyme-adenylate intermediate. The extent of ligase-adenylate formation by wild-type Rnl2 was proportional to input protein (Fig. 5B). We estimated from the slope of the titration curve that 65% of the input enzyme molecules became labeled with [ 32 P]AMP. The residual fraction of the protein preparation likely comprises preadenylated Rnl2 (see below). The E34A, E99A, F119A, D120A, and E204A mutants were effectively inert over the same range of input enzyme, i.e. their specific activities were Յ1% of the wild-type value. Other mutants displaying significant defects in step 1 adenylation were R55A (2.3%), K209A (2.7%), K227A (4.1%), and N40A (15%). Only the K198A mutant displayed near wildtype adenylyltransferase activity ( Fig. 5B and Table I).
Mutational effects on ligation of a 5Ј 32 P-labeled singlestranded RNA substrate in the presence and absence of ATP under conditions of enzyme excess are shown in Fig. 6. Wildtype Rnl2 efficiently circularized the RNA in the absence of ATP (reflecting the catalysis of steps 2 and 3 by preformed Rnl2-AMP in the enzyme preparation) but generated predominantly AppRNA in the presence of ATP. The ligation activity of the K189A mutant was indistinguishable from that of wildtype Rnl2. Mutants E34A, E99A, and D120A failed to form any ligated RNA circles or RNA-adenylate intermediate. R55A and F119A formed only trace amounts of RNA-adenylate and no ligated circles (Fig. 6). The profound defects of the E34A, R55A, E99A, F119A, and D120A mutants in overall ligation were in keeping with their inability to form the ligase-adenylate intermediate in vitro. The failure to ligate in the absence of ATP suggests that these five mutants were also not adenylated to any significant extent during their production in E. coli (unlike the wild-type Rnl2). The K209A mutant generated high levels of RNA-adenylate in the absence of ATP but did not carry the reaction through to form circular products. Thus, although K209A was defective in ligase adenylation in vitro, it clearly did react with ATP to form ligase-adenylate in vivo in E. coli. The accumulation of AppRNA suggested that K209A might be defective in step 3 of the ligation pathway (see below). The N40A mutant displayed weak activity in RNA adenylation in the absence of ATP and did not generate a circular product.
Structure-Activity Relationships at Essential Residues of Rnl2-The present results, together with previous mutational data (13), highlight 11 individual amino acids in addition to the motif I lysine nucleophile (Lys 35 ) that are essential for Rnl2 activity: Glu 34 , His 37 , Asn 40 , Arg 55 , Glu 99 , Phe 119 , Asp 120 , Glu 204 , Lys 209 , Lys 225 , and Lys 227 . To better evaluate the contributions of these residues to the RNA ligase reaction, we tested the effects of conservative substitutions at 10 positions (all except His 37 , which was mutated previously to Asp). Arginine was replaced by lysine and glutamine, glutamate by glutamine and aspartate, aspartate by asparagine and glutamate, lysine by arginine and glutamine, and phenylalanine by leucine. Also, Asn 40 was replaced with arginine, which is present at the equivalent motif I position of ATP-dependent DNA ligases and most mRNA capping enzymes. Twenty new Rnl2 mutants were produced in E. coli and purified from soluble bacterial extracts by nickel-agarose chromatography. The Rnl2 polypeptide was the predominant species detected by SDS-PAGE, and the extents of purification were comparable for mutant and wild-type Rnl2 (Fig. 7). The D120N mutation caused the same increase in the electrophoretic mobility seen with D120A, which was rectified when Asp 120 was replaced by Glu (Fig. 7, middle panel). The K227R mutation also caused an increase in electrophoretic mobility, which was not seen with K227Q (Fig. 7, bottom panel) or K227A (13).
The adenylyltransferase activities of the mutants were determined by enzyme titration and normalized to the wild-type value (Table I). Conservative replacement of Glu 34 , Glu 99 , or Glu 204 with Asp or Gln elicited severe catalytic defects comparable with those seen with the respective Ala mutants. These data establish the requirement for a carboxylate residue at positions 34, 99, and 204 and a minimum distance from the main chain to the carboxylate that is met by glutamate but not aspartate. Note that glutamates are strictly conserved at these three positions in the trypanosome, baculovirus, and entomopoxvirus Rnl2 homologs (Fig. 1). Changing Asp 120 to Asn reduced ligase-AMP formation to 0.2% of the wild-type level,   Fig. 5B. The specific activities were calculated from the slopes of the titration curves and normalized to the specific activity of wild-type Rnl2 (defined as 100%). The right-hand column lists the side chain present at the equivalent position of DNA ligases and the relevant side chain contacts seen in the available crystal structures of DNA ligases and RNA capping enzymes. Contacts to the 5Ј-PO 4 are inferred from the crystal structure of Chlorella virus DNA ligase-adenylate, which contains a sulfate bound to the enzyme surface adjacent to the AMP phosphate. This sulfate is proposed to mimic the 5Ј-PO 4   similar to the D120A mutant, but activity was partially restored by the glutamate substitution (to 12% of wild type). Thus, the carboxylate functional group is critical at position 120, but there is some flexibility in accommodating the longer Glu side chain instead of Asp. This is notable in light of the fact that Glu is present at the corresponding motif IIIa position of the baculovirus and entomopoxvirus Rnl2 homologs (Fig. 1).
Whereas replacement of Arg 55 by Lys resulted in a significant gain of function (to 30% of wild type) compared with R55A (2.3%), the R55Q change had no salutary effect. Similarly, the introduction of Arg in lieu of Lys 209 restored activity to 25% of wild type, compared with 2.7% for K209A, whereas the glutamine mutant was 7.3% as active as wild type. We surmise that the positive charges at positions 55 and 209 are important for ligase adenylation. (Note that the position corresponding to Rnl2 Lys 209 is occupied by Arg in several of the Rnl2 homologs shown in Fig. 1.) In contrast, Lys 225 and Lys 227 in motif V are strictly essential and cannot be functionally replaced by either Arg or Gln (Table I), even though the second lysine is naturally an arginine in one of the trypanosome RNA-editing ligases.
Replacing Phe 119 in motif IIIa with leucine resulted in a slight gain of function (to 3.9% of wild type) compared with the F119A mutant (Ͻ0.1% activity), but F119L was still significantly compromised. This finding attests to the importance of the aromatic side chain. We eschewed changing Phe 119 to Tyr because tyrosine is present at the equivalent positions of the entomopoxvirus and baculovirus Rnl2 homologs (Fig. 1).
Instructive conservative mutational effects were seen at Asn 40 of motif I, where adenylyltransferase was restored to wild-type level by aspartic acid but not by glutamine. These findings suggest that: (i) there is steric constraint precluding function of the longer glutamine side chain and (ii) Asn 40 likely functions as a hydrogen bond acceptor during the adenylyltransferase reaction. This is in contrast to the hydrogen bond donor function of the conserved arginine at the equivalent motif I position of ATP-dependent DNA ligase and capping enzyme (10,11). Of note, we found that replacing Asn 40 of Rnl2 with Arg was considerably more deleterious than side chain removal, i.e. N40R had 0.2% of wild-type adenylyltransferase activity compared with 15% for N40A. This result highlights the importance of the motif I Asn as a signature feature of the active site of the Rnl2-like ligase family.
Effects of Conservative Substitutions on RNA Ligation-Conservative mutational effects on circularization of singlestranded RNA generally paralleled their impact on the ligase adenylation step of the reaction pathway (Fig. 8). For example, E34D (like E34A) failed to form any ligated RNA circles or RNA-adenylate intermediate, whereas E34Q formed only low levels of AppRNA, consistent with its minimally better adenylyltransferase activity in vitro compared with E34D and E34A. N40D was as active as wild-type Rnl2 in overall ligation, consistent with full restoration of adenylyltransferase activity. In contrast, N40Q accumulated AppRNA but did not efficiently form circles, whereas N40R was entirely unreactive. R55K restored the wild-type profile for overall ligation, just as it elicited a major gain of function in step 1 adenylyltransferase, whereas R55Q remained defective in ligation. E99D, E99Q, E204D, E204Q, K225Q, and K227R were unreactive in the pRNA ligation reaction, likely as a consequence of their step 1 defects. K225R and K227Q formed low levels of AppRNA in the absence of ATP, indicating that some ligase adenylation had occurred in vivo in E. coli, which is consistent with their marginal gains of in vitro adenylyltransferase activity compared with K225Q and K227R, respectively (Table I).
F119L, D120E, K209R, and K209Q accumulated fairly high levels of RNA-adenylate in the absence of ATP (56 -72% of the total labeled RNA as AppRNA) but did not proceed to form circular products (which comprised between 8 and 14% of total labeled RNA, compared with 90% circles in the wild-type Rnl2 reactions in Fig. 8). Thus, although F119L was defective in ligase adenylation in vitro, it obviously did react with ATP to form ligase-adenylate in vivo in E. coli. Adenylation in vivo of D120E, K209R, and K209Q is in keeping with their partial adenylyltransferase activities in vitro (between 7 and 25% of wild-type Rnl2). The accumulation of AppRNA seen in Fig. 8 suggested that F119L, D120E, K209R, or K209Q might be defective in step 3 of the ligation pathway.
Mutational Effects on Phosphodiester Formation at a Preadenylated RNA 5Ј End-Although the effects on step 1 sufficed to explain the loss of overall ligation activity of many of our Rnl2 mutants, we were interested in determining whether the enzyme functional groups implicated for step 1 are also required for phosphodiester formation (step 3). We showed above (Fig. 4) that step 3 can be studied in isolation by bypassing the requirement for steps 1 and 2 and gauging the ability of Rnl2 to seal a preadenylated RNA substrate. Thus, the 13 Rnl2-Ala mutants in our collection were reacted with a single-stranded 18-mer AppRNA substrate under conditions of enzyme excess (Fig. 9). Wild-type Rnl2 catalyzed nearly quantitative conversion of the input AppRNA strand to a circular product, as did K189A, the only one of the Rnl2-Ala proteins that retained near wild-type activity in the pRNA ligation pathway. Mutants E34A, E99A, F119A, D120A, and E204A were inert in the isolated step of phosphodiester formation, just as they were virtually inert in the isolated step 1 ligase adenylylation reaction. R55A and K209A, which were 2-3% as active as wild-type Rnl2 in step 1, formed only trace amounts of circular product in the isolated step 3 reaction (Fig. 9). A simple interpretation of these results is that Glu 34 , Arg 55 , Glu 99 , Phe 119 , Asp 120 , Glu 204 , and Lys 209 play equivalent roles in the first and third steps of the ligation reaction. The obvious common feature of the step 1 and 3 reactions is that they entail recognition of the adenylate moiety of the ATP and AppRNA substrates, respectively.
Nonetheless, as discussed above for the K35A mutation, it is not a forgone conclusion that any change that abrogates or severely affects step 1 will exert the same effect on step 3. For example, K225A is just as defective in step 1 as E34A or E204A (Table I), but it clearly can catalyze ligation of AppRNA, albeit less efficiently than wild-type Rnl2 (Fig. 9). K227A is more active in step 3 than K209A (Fig. 9), although they have comparable activity in step 1 (3-4% of wild type).
Are any of the putative active site constituents of Rnl2 specific for the step of phosphodiester bond formation? We showed previously that the motif I mutant H37A was fully active in step 1 ligase adenylylation and capable of transferring the adenylate to the 5Ј-PO 4 of RNA to form RNA-adenylate but seemingly unable to form RNA circles during the pRNA ligation reaction (13). Here we found that H37A was severely and selectively impaired at the isolated step of phosphodiester bond formation (Fig. 9). Thus, His 37 is specifically implicated in catalysis of step 3. Replacing His 37 by Asp rectified the pRNA strand joining defect of the H37A mutant (13). Here we see that the H37D change restored wild-type activity in the isolated step 3 reaction (Fig. 9). Note that Asp is normally present at the equivalent motif I position of the entomopoxvirus and baculovirus Rnl2 homologs (Fig. 1) and in motif I of the majority of known DNA ligase enzymes. The signature Asn 40 residue of the Rnl2 family may also play a key role in step 3, insofar as the N40A mutant appears to be more defective in step 3 ( Fig. 9) than it does in step 1 (15% of wild-type adenylyltransferase activity).
Additional insight into the requirements for phosphodiester bond formation were gleaned from an analysis of the effects of conservative mutations on the isolated step 3 reaction (Fig. 10). E34D, E34Q, E99D, E99Q, E204D, and E204Q were inert or severely defective in sealing AppRNA, just like the respective Glu-to-Ala mutants. The N40D and N40Q changes restored step 3 function compared with the defective N40A mutant, whereas N40R was unreactive in step 3. The R55K change revived phosphodiester bond formation relative to R55A, whereas the R55Q mutant remained defective. The K209R and K209Q mutations both resulted in gains of step 3 activity relative to the severely defective K209A protein.
F119L displayed feeble step 3 activity, which was nonetheless an improvement compared with F119A. The D120E change partially restored step 3 function relative to the D120A mutant, but the D120N mutant remained defective. The isolated step 3 defects of the F119L and D120E proteins may account for their accumulation of RNA-adenylate during the composite ATPindependent ligation reaction (Fig. 8).
In motif V, the K225R protein was partially active in step 3 (like K225A), whereas K225Q was severely defective. At position 227, however, the arginine substitution abolished step 3 activity, whereas the glutamine mutant displayed nearly wildtype step 3 activity. The disparate effects of conservative substitutions indicate that the two motif V lysines play distinct roles at the step of phosphodiester bond formation, i.e. the function of Lys 225 depends on its positive charge, whereas Lys 227 function likely depends on its hydrogen bonding capacity. DISCUSSION We have now identified 12 essential residues of bacteriophage T4 Rnl2 and defined structure-activity relationships via conservative substitutions. We find that all five nucleotidyl transferase motifs are essential in Rnl2, just as they are in DNA ligases and RNA capping enzymes (23)(24)(25)(26)(27)(28)(29)(30). These results underscore a shared structural basis for catalysis among Rnl2like ligases, DNA ligases, and capping enzymes. They also highlight two essential signature residues in motif I (Glu 34 and Asn 40 ) that are unique to Rnl2-like ligases.
We report mutational effects on overall RNA ligation and the isolated steps of ligase adenylylation and phosphodiester formation. Our inferences about which features of the individual side chains are required for activity have been discussed in detail above. Thus, we focus here on how the findings may be interpreted in light of the crystal structures available for other members of the covalent nucleotidyl transferase superfamily. In the absence of an atomic structure for any RNA ligase, we assume based on concordance of mutational data that the fold of the N-terminal nucleotidyl transferase domain of Rnl2 (which includes the five motifs) resembles that of Chlorella virus DNA ligase (the minimal eukaryotic ATP-dependent DNA ligase) and that there is a direct correspondence between the essential amino acids of the Rnl2-like RNA ligase family and the amino acids found at "equivalent" positions of DNA ligases, which are listed in Table I along with the atomic contacts made by these side chains, as revealed by the struc- tures of the ligase-AMP intermediate of Chlorella virus DNA ligase and the ATP-bound bacteriophage T7 DNA ligase (9,11). Note that eight of the ten essential Rnl2 positions listed in Table I are occupied by an identical or closely related amino acid in ATP-dependent DNA ligases. The two exceptions are the signature motif I residues of Rnl2, Glu 34 and Asn 40 .
Rnl2 residues Arg 55 , Glu 99 , Phe 119 , and Glu 204 are critical for the first and third steps of the RNA ligation reaction, which, by analogy to DNA ligase, will entail docking of the adenylate moiety of the ATP or the AppRNA substrate into an AMPbinding pocket of Rnl2. The counterparts of Rnl2 residues Arg 55 , Glu 99 , Phe 119 , and Glu 204 in the DNA ligases make direct contacts with constituents of the adenosine nucleotide. The motif III glutamate contacts the ribose sugar; the motif IIIa aromatic group engages in a stack on the adenine base; the motif IV glutamate is implicated in coordinating a divalent cation; the two motif V lysines coordinate the ␣-phosphate; and the conserved arginine located between motifs I and III is proposed to contact the ␥-phosphate of ATP in step 1 and the 5Ј-PO 4 of the nucleic acid substrate during subsequent steps. We impute similar functions to the respective essential side chains of Rnl2.
The counterparts of the essential Rnl2 residues Asp 120 (motif IIIa) and Lys 209 (motif IV) do not contact the nucleotide in any of the available DNA ligase (or capping enzyme) structures. Rather they form ion pairs with oppositely charged side chains of the respective enzymes. The motif IIIa Asp is located next to the essential aromatic amino acid that forms a hydrophobic pocket for the purine base. This position is occupied by aspartate in the majority of ATP-dependent DNA ligases and RNA capping enzymes. The participation of the aspartate in a salt bridge may be crucial for the active site architecture of Rnl2, insofar as the neutral asparagine substitution elicited defects in steps 1 and 3 of the Rnl2 reaction. It is remarkable that the positively charged partner in the ion pair formed by the conserved motif IIIa Asp is located at very different places in the amino acid sequences of the several ligases and capping enzymes for which atomic structures have been solved. In the Chlorella virus DNA ligase structure, the motif IIIa Asp forms an ion pair with the arginine-flanking motif IV, corresponding to the essential Lys 209 residue of Rnl2. The significant gain of function seen for the K209R mutant of Rnl2 is in keeping with a putative ionic interaction for Lys 209 , be it with Asp 120 (analogous to the ion pair of Chlorella virus DNA ligase) or with some other acidic residue in Rnl2.
Asn 40 is one of the signature amino acids of the Rnl2-like family. An essential arginine side is present at the equivalent position in motif I of ATP-dependent DNA ligases and capping enzymes; this motif I arginine interacts with the ribose sugar of the adenosine or guanosine nucleoside. Asn 40 of Rnl2 may also be concerned with recognition of the adenosine sugar of ATP or AppRNA, albeit functioning as a hydrogen bond acceptor. Alternatively, it may play different role in substrate binding or catalysis. In any event, it is clear that Rnl2 has evolved a unique structural requirement for this side chain, which cannot be fulfilled (but is instead antagonized) by the arginine normally present in other covalent nucleotidyl transferases.
Glu 34 occupies the position immediately preceding the active site lysine nucleophile. Whereas this residue is invariant in Rnl2-like proteins, the equivalent side chain in DNA ligases is typically hydrophobic. It is worth noting that T4 Rnl1 has no counterpart of Glu 34 in motif I; rather there is a threonine immediately preceding the motif I lysine. Thus, Glu 34 stands out as an essential and defining feature of the Rnl2 family (required for steps 1 and 3 of the Rnl2 ligation pathway). Delineation of the atomic contacts made by this Rnl2-specific side chain will obviously hinge on crystallization of Rnl2 or one of its orthologs.