The 2'-5' RNA ligase of Escherichia coli. Purification, cloning, and genomic disruption.

An RNA ligase previously detected in extracts of Escherichia coli is capable of joining Saccharomyces cerevisiae tRNA splicing intermediates in the absence of ATP to form a 2′-5′ phosphodiester linkage (Greer, C., Javor, B., and Abelson, J. (1983) Cell 33, 899-906). This enzyme specifically ligates tRNA half-molecules containing nucleoside base modifications and shows a preference among different tRNA species. In order to investigate the function of this enzyme in RNA metabolism, the ligase was purified to homogeneity from E. coli lysate utilizing chromatographic techniques and separation of proteins by SDS-polyacrylamide gel electrophoresis. A single polypeptide of approximately 20 kilodaltons exhibited RNA ligase activity. The amino terminus of this protein was sequenced, and the open reading frame (ORF) encoding it was identified by a data base search. This ORF, which encodes a novel protein with a predicted molecular mass of 19.9 kDa, was amplified from E. coli genomic DNA and cloned. ORFs coding for highly similar proteins were detected in Methanococcus jannaschii and Bacillus stearothermophilus. The chromosomal gene encoding RNA ligase in E. coli was disrupted, abolishing ligase activity in cell lysates. Cells lacking ligase activity grew normally under laboratory conditions. However, moderate overexpression of the ligase protein led to slower growth rates and a temperature-sensitive phenotype in both wild-type and RNA ligase knockout strains. The RNA ligase reaction was studied in vitro using purified enzyme and was found to be reversible, indicating that this enzyme may perform cleavage or ligation in vivo.

RNA ligases have been detected in organisms throughout all major divisions of the phylogenetic spectrum. Specific metabolic functions, however, have been assigned to few of these enzymes. T4 RNA ligase, for example, repairs nicks introduced into the anticodon loops of tRNAs in phage-restrictive Escherichia coli strains upon T4 infection (1); the eukaryotic tRNA splicing RNA ligase joins tRNA half-molecules resulting from endonucleolytic removal of introns from eukaryotic tRNA precursors (2)(3)(4), and an RNA ligase is apparently responsible for the joining of guide RNA molecules to mRNA during RNA editing in kinetoplastids (5). At least two other species of RNA ligase, the metazoan-specific "animal pathway" RNA ligase (6) and the archaeal stable RNA splicing ligase (7), have been identified, but the precise in vivo substrate(s) and function(s) of these enzymes have yet to be determined.
The existence of RNA ligase in bacteria in the absence of bacteriophage infection was discovered in this laboratory in 1983 (8). An activity capable of performing the ligation step of eukaryotic tRNA splicing was detected in extracts of a wide range of bacteria including members of the Alpha and Gamma subdivisions of proteobacteria, green sulfur bacteria, and low G ϩ C content Gram-positive bacteria (8). In extracts of E. coli the ligase activity had a substrate specificity restricted to 4 of the 10 Saccharomyces cerevisiae tRNA splicing intermediates: tRNA Tyr, Phe, Lys 2 , and Trp half-molecules. The reaction mechanism of the E. coli RNA ligase apparently differed from that of known RNA ligases since it did not require a nucleoside triphosphate cofactor, and the product contained an unusual 2Ј-5Ј phosphodiester bond at the ligated junction (8).
The discovery of RNA ligase in E. coli implies the existence of a novel form of bacterial RNA processing. No intervening sequences of the type found in eukaryotic nuclear or archaeal tRNA genes (which require enzymatic excision and religation) occur in known bacterial tRNA genes, although self-splicing introns are found in the tRNA genes of certain cyanobacteria (9) and some proteobacteria (10). In fact, no introns of any kind are found in the full genomic complement of tRNA genes in E. coli, Mycoplasma capricolum, and Hemophilus influenzae (11)(12)(13). To date no RNA processing event that would require the action of an RNA ligase enzyme has been observed to occur in any bacteria, indicating that elucidation of the substrate and function of the 2Ј-5Ј RNA ligase should reveal a previously unknown step in bacterial RNA metabolism.
A genuine in vivo function for the E. coli RNA ligase activity observed in vitro is suggested by the occurrence of 2Ј-5Ј linkages in native E. coli RNA. Several forms of 2Ј-5Ј-linked oligoadenylates have been detected in acid-soluble extracts of E. coli (14). The most abundant species of these oligoribonucleotides observed was 2Ј-5Ј-linked adenosine dinucleotide 3Јmonophosphate, which was estimated to exist at an intracellular concentration over 100 nM. This dinucleotide could not be an intermediate in oligoadenylate synthesis, and it has been suggested that this species may be a degradation product of RNAs containing individual 2Ј-5Ј bonds among standard 3Ј-5Ј linkages (14). Since the formation of 2Ј-5Ј linkages is not catalyzed by RNA polymerase, these bonds must be added in a posttranscriptional processing event by an enzyme such as the 2Ј-5Ј RNA ligase.
In order to study the function and mechanism of the 2Ј-5Ј RNA ligase, the enzyme was purified to homogeneity from E. coli extracts. A single polypeptide was found to contain RNA ligase activity. This protein was partially sequenced, and the ORF 1 encoding it was identified. This ORF was amplified from E. coli genomic DNA and cloned. The chromosomal locus containing the ligase gene was disrupted, abolishing ligase activity in cellular extracts. Cells completely lacking ligase activity grow similarly to the parent strain. E. coli strains overexpressing recombinant 2Ј-5Ј RNA ligase are temperature-sensitive for growth. The ligase reaction was also studied using purified enzyme and was found to be reversible in vitro.
Preparation of RNA Substrates for Ligation-RNA polymerase III transcription of pre-tRNA Tyr in yeast extract was performed by the method of Evans and Engelke (15), using extracts prepared as described. The template for Pol III transcription of pre-tRNA Tyr was supercoiled pYSUP6 (16). T7 RNA polymerase transcription was performed as described by Sampson and Saks (17). pre-tRNA Phe was transcribed from the artificial pre-tRNA gene of Reyes cloned into pUC13 (18). The template for pre-tRNA Tyr T7 transcription was created by PCR amplification from pYSUP6, adding a canonical T7 promoter at the 5Ј end and a BstNI restriction site at the 3Ј end of the pre-tRNA, as well as altering the acceptor stem base pairs: C 1 3 G, T 2 3 G, A 87 3 C, and G 88 3 C to improve transcription yields. This construct was cloned into the pBlueScript vector (Stratagene). tRNA substrates were modified by incubation at 24°C for 45 min in the same extract and reaction conditions as were used for Pol III transcription. tRNA precursors were cleaved as described by Peebles et al. (2), using partially purified S. cerevisiae tRNA splicing endonuclease fractions from the hydroxyapatite step or later as described in the Rauhut protocol (2,19).
All RNA transcripts were gel purified by polyacrylamide gel electrophoresis (PAGE) in 1 ϫ TBE (Tris borate EDTA buffer), 7 M urea and visualized by autoradiography. RNAs were eluted from crushed gel slices in 0.6 M NH 4 OAc, 2 mM EDTA, 0.005% Nonidet P-40 at room temperature with vortexing for 20 min. RNA eluates were extracted with phenol/chloroform (1:1, pH 4.5) and chloroform and then ethanolprecipitated in the presence of glycogen and resuspended in distilled water.
RNA Ligation Assay-tRNA half-molecules were annealed for ligation in 2 ϫ ligation buffer (1 ϫ ϭ 40 mM HEPES, pH 7.8, 3 mM MgCl 2 , 2 mM spermidine, 5% glycerol) by heating to 85°C and slowly cooling to 30°C over 20 min. Ligation reactions were typically performed in 4 l of 1 ϫ ligation buffer with 250 nM annealled tRNA half-molecules substrate. E. coli lysates or purification fractions (typically 1 l) were added and incubated for 2-5 min at 30°C. Reactions were stopped by the addition of 1.5 mg/ml proteinase K, 0.04% SDS, 5 mM EDTA, and 0.001 mg/ml E. coli total RNA for 20 min at 30°C. An equal volume of 90% deionized formamide plus tracking dyes was added, and reactions were incubated at 65°C for 5 min before cooling on ice and loading onto 8% acrylamide, 7 M urea, 1 ϫ TBE gels for electrophoresis. RNA bands were visualized by autoradiography or by exposure to a Molecular Dynamics storage phosphor plate for quantification and analysis on a PhosphorImager using Imagequant software.
Large-scale Cell Growth-Cells were grown in media containing 16 g/liter Bacto-tryptone, 6 g/liter yeast extract, 50 mM phosphate buffer, pH 6.8, 4 g/liter (NH 4 ) 2 SO 4 , 0.5 g/liter MgSO 4 (anhydrous), and 20 ml/liter glycerol with aeration and constant feeding of 2 ϫ media for 24 h. 5% methanol (which was found to induce ligase activity in extracts to approximately 1.5 ϫ that found in extracts of uninduced cells) was added to media 2 h before harvest, and harvested cell paste was stored at Ϫ70°C.
E. coli RNA Ligase Purification-All purification procedures were performed in extraction buffer (EB) containing 10% glycerol, 40 mM HEPES, pH 7.8, 2 mM EDTA, 1 mM Pefabloc (Boehringer Mannheim), 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and stated concentrations of KCl. 200 g to 1 kg of HB101 cells frozen at Ϫ70°C were thawed in a final volume of 1.6 ml/g cells of EB plus 125 mM KCl and dispersed with a hand blender. The cell suspension was sonicated on ice with a Branson Sonifier large tip at 80% power for 15 1-min periods with cooling on ice between. Cell lysate was centrifuged at 35,000 rpm in a Beckman 45Ti rotor at 3°C for 1 h. The supernatant was mixed batchwise with 500 ml of DEAE-Sepharose CL-6B (equilibrated in EB ϩ 125 mM KCl) for 30 min on ice. All subsequent steps were performed at 4°C. The DEAE/extract slurry was poured into a column, and the DEAE effluent was loaded at approximately 1 ml/min onto a 200-ml bed of cellulose phosphate P11 (Whatman) equilibrated in EB ϩ 125 mM KCl.
The phosphocellulose column was washed extensively with EB ϩ 500 mM KCl and eluted with a linear KCl gradient from 0.5 to 1.5 M in EB. Peak ligase activity fractions were pooled, diluted to 150 mM KCl, and loaded onto 25 ml of heparin Hyper-D (Biosepra) column. Ligase activity was eluted from the heparin column with a 150 -700 mM KCl gradient in EB, and peak fractions were pooled. The heparin pool was diluted to 150 mM KCl in EB and loaded onto a 15-ml bed column of E. coli tRNA linked to a Sepharose support. tRNA Sepharose (3 mg of RNA per ml of gel) was prepared as described by Rauhut et al. (19) but using E. coli tRNA (Sigma) instead of yeast tRNA. Ligase was eluted from this column using a gradient from 150 mM to 1 M KCl in EB, followed by a 2 M KCl wash. The ligase eluted in a broad peak beginning around 400 mM KCl, trailing into the 2 M wash.
Active fractions were pooled, diluted to 150 mM KCl, loaded onto a 2-ml column of heparin hyper-D, and eluted with EB ϩ 600 mM KCl. This concentrated activity pool was then passed through a 115-ml bed of Superdex 75 gel filtration medium (Pharmacia Biotech Inc.) at a flow rate of 0.04 ml/min. Peak active fractions were diluted to 150 mM KCl in EB and loaded onto a column of S. cerevisiae tRNA linked to Sepharose (19). Ligase activity was eluted with a 150 -800 mM KCl gradient. Note: For the 4-kg preparative purification mentioned under "Results," the purification procedure followed was the same as described above except that an S. cerevisiae tRNA affinity column was substituted for the initial E. coli tRNA step, and the tRNA pool was passed through Superdex 75 twice, sequentially.
SDS-PAGE, Elution, and Renaturation-Electrophoresis for preparative elution, as well as all analytical protein electrophoresis, was performed on 16% acrylamide/piperazine diacrylamide (2.67% crosslinking) gels under the conditions of Schagger et al. (20). Bands were visualized by staining with 0.3 M CuCl 2 for 15 min at room temperature, excised, and destained by successive washes in 0.25 M Tris-HCl, pH 8.8, 0.25 M EDTA, and distilled water. Excised bands were crushed and extracted overnight at 4°C in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% SDS, and 0.1 mg/ml acetylated bovine serum albumin (Life Technologies, Inc.). Eluted protein was precipitated with 4 volumes of cold acetone, incubated for 30 min at 0°C, and pelleted by centrifugation at 15,000 rpm, 4°C, for 20 min in an SS-34 rotor. Pellets were resuspended in EB containing 6 M guanidine HCl and microdialyzed versus EB plus 125 mM KCl at 4°C. Silver staining of SDS-PAGE gels was performed according to the method of Wray et al. (21).
Mapping Blot Hybridization-Labeled PCR product was heated to 95°C in 6 ϫ SSC, 10 mM phosphate buffer, pH 6.8, 1 mM EDTA, 0.5% SDS, 100 g/ml sonicated calf thymus DNA, and 0.1% dry milk, and then cooled on ice. The probe was hybridized to an E. coli Gene Mapping Membrane (PanVera) in the same buffer at 37°C, overnight. The blot was washed with 6 ϫ SSC, 0.1% SDS, and with 5 ϫ SSC at room temperature and then exposed to film for autoradiography.
Genomic Knockout-The E. coli RNA ligase gene was cloned by PCR amplification from HB101 genomic DNA using primers corresponding to sequences located 213 bp upstream of the 5Ј end and at the exact 3Ј end of the predicted ORF, adding EcoRI and BamHI restriction sites, respectively. Gene disruption was performed according to the method of Hamilton et al. (22). For genomic disruption, the Kan r Genblock (Pharmacia) cassette was inserted into a PstI site located at ϩ144 bp in the ligase ORF clone. A ts plasmid bearing the interrupted ligase gene was created as follows. The origin-bearing PvuII fragment of pFC20 (23) (bp 950-3838) was cloned into pMAK705 (22), replacing the PvuII fragment spanning bp 5458 -3722. The Kan cassette-interrupted ligase gene was then subcloned into the EcoRI-BamHI sites of the pMAK polylinker in this construct to create pTSIL. pTSIL was transformed into a RecA ϩ strain of E. coli (HSD947) and plated at 43°C on chloramphenicol to select for cointegration into the chromosome. Plasmid cointegrants were resolved by growth for several generations in liquid culture at 30°C, at which temperature the plasmid replicon interferes with chromosomal replication. The chromosomal recA locus was then disrupted by P1 phage transduction to prevent further homologous recombination 1 The abbreviations used are: ORF, open reading frame; PCR, polymerase chain reaction; PAGE, polyacrylamide gel electrophoresis; PNK, polynucleotide kinase; pre-tRNA, intron-containing tRNA precursor; bp, base pair(s); ts, temperature-sensitive; kbp, kilobase pair(s). events. Cointegrant cultures were pelleted and resuspended in 0.1 M MgSO 4 , 5 mM CaCl 2 and mixed with an equal volume of P1 lysate JC10240 2 for 20 min at 30°C and plated on LB ϩ tetracycline, chloramphenicol, and kanamycin to select for transductants. P1 transductants were screened for resolution of the ts plasmid by streaking separately on chloramphenicol, kanamycin, and nonselective LB plates at 43°C.

RESULTS
E. coli 2Ј-5Ј RNA Ligase Requires Base Modifications in tRNA Substrates-In previous studies of bacterial RNA ligase, activity had been assayed using substrates derived from tRNA precursors (pre-tRNAs) transcribed either in vivo in yeast or by endogenous RNA polymerase III activity in an S. cerevisiae extract (8). It was subsequently reported that pre-tRNA Phe transcribed from an artificial tRNA Phe gene by T7 RNA polymerase is an excellent substrate for the S. cerevisiae tRNA splicing enzymes (18), and this was also tested for ligation by E. coli RNA ligase. tRNA Phe half-molecules produced by digestion of T7-transcribed pre-tRNA Phe with S. cerevisiae tRNA splicing endonuclease were, however, poorly ligated in E. coli extracts (data not shown). Since tRNA Phe half-molecule substrates produced in vivo in yeast had previously been demonstrated to be ligated in E. coli extract (8), a requirement for some modification of substrate transcripts was implied.
In order to test the putative modification requirement, an S. cerevisiae tRNA Tyr gene was cloned under the control of a T7 RNA polymerase promoter. tRNA Tyr half-molecules derived from unmodified transcripts of this gene were poor substrates for the E. coli ligase, although they were utilized effectively by the S. cerevisiae tRNA splicing ligase. Fig. 1A shows that when pre-tRNA Tyr transcripts (lane M) were incubated in the same S. cerevisiae extracts utilized for RNA polymerase III transcription, tRNA Tyr half-molecules derived from them (lane 1) became substrates for the E. coli RNA ligase (lane 4). E. coli RNA ligase ligation of half-molecules derived from pre-tRNA Tyr modified in vitro produced a product of equivalent size to that produced by S. cerevisiae tRNA ligase or T4 RNA ligase and polynucleotide kinase (PNK) (lanes 2-3), although the bacterial enzyme is less efficient at joining these substrates. tRNA halfmolecule substrates derived from modified T7 RNA polymerase transcripts were utilized by the E. coli RNA ligase as efficiently as yeast polymerase III transcripts, as quantified in a substrate titration experiment shown in Fig. 1B. This was presumably due to the formation of modified nucleosides in these transcripts, as observed by two-dimensional thin layer chromatography (TLC) of nuclease digests of substrate transcripts (data not shown). The identity and location of these modifications were not investigated further. The presence of a 2Ј-5Ј linkage in ligated tRNA products was also observed by nuclease digestion and TLC (not shown), confirming that the reaction had been catalyzed by the previously described E. coli 2Ј-5Ј ligase activity.
Pre-tRNA Phe T7 RNA polymerase transcripts modified by incubation in yeast extract were also used to produce substrates for ligation by the E. coli enzyme, although modified tRNA Tyr half-molecules were preferred as substrates by a factor of 2-3-fold over modified tRNA Phe half-molecules (Fig. 1C). Because tRNA Tyr half-molecules produced by endonuclease digestion of T7-transcribed, yeast extract-modified pre-tRNA Tyr were the most active substrates tested for ligation by the E. coli RNA ligase, they were chosen as substrates for quantitative assays of ligase activity during subsequent procedures.
Purification of the Ligase-A protocol for purification of the E. coli RNA ligase was developed and followed as detailed under "Experimental Procedures." A quantitative profile of RNA ligase activity throughout the purification procedure is given in Table I. The complexity of the protein population at each purification step was assayed by SDS-PAGE and silver staining and is shown in Fig. 2. Briefly, cells were disrupted by sonication, and the lysate (Fig. 2, lane 1) was subjected to 2 M. Saks, personal communication. The indicated molar amounts of tRNA Tyr half-molecules were incubated with 1 l of E. coli lysate each at 30°C for 15 min. Halfmolecule substrates were derived from pre-tRNA Tyr : transcribed by T7 RNA polymerase (j), transcribed by yeast RNA polymerase III in yeast extract (Ⅺ), or transcribed by T7 polymerase and subsequently incubated in yeast extract (f). C, titration of tRNA Tyr and tRNA Phe halfmolecule substrates for ligation. tRNA Phe half-molecules (E) and tRNA-Tyr half-molecules (Ⅺ) were derived from T7-transcribed, yeast-modified pre-tRNAs. The indicated molar amounts of substrate were incubated with 0.5 l of partially purified E. coli RNA ligase (cellulose phosphate pool, see below) for 2 min at 30°C. centrifugation at 100,000 ϫ g. The S-100 supernatant (lane 2) was mixed batchwise with a DEAE anion-exchange resin, and the unbound fraction (lane 3) was loaded onto a cellulose phosphate column. RNA ligase activity was eluted from the cellulose phosphate by an increasing gradient of KCl, and fractions containing peak levels of RNA ligase activity were pooled for further purification by sequential binding to, and salt gradient elution from, heparin and E. coli tRNA affinity matrices. Pooled active fractions were subjected to gel filtration through Superdex 75 media. Peak RNA ligase activity fractions after gel filtration contained a mixture of at least six polypeptides (Fig.  2, lane 8). Superdex 75 peak fractions were then pooled for binding to a column of S. cerevisiae tRNA-Sepharose. RNA ligase activity eluted from this matrix with a salt gradient still contained several polypeptide species (lane 9).
Reconstitution of Ligase Activity from a Single Polypeptide Following SDS-PAGE-Since the optimized purification procedure did not yield a single homogeneous polypeptide, separation by denaturing electrophoresis was also utilized. A large scale purification was undertaken using 4 kg of E. coli cells as starting material. Final peak ligase activity fractions were pooled and concentrated by dialysis, and an aliquot of the concentrate was subjected to SDS-PAGE. Successive regions of the SDS-PAGE gel lane were excised and individually extracted. Eluted protein from each gel slice was precipitated and resuspended in guanidine HCl. After dialysis, eluates were assayed for RNA ligase activity (Fig. 3A) and examined by SDS-PAGE (Fig. 3B). Only a single eluate showed significant reconstituted activity (Fig. 3A, lane 5). The level of RNA ligase activity in this eluate was not affected by mixing with eluates of other gel slices (not shown). Fig. 3B shows that the active eluate contains only a single E. coli protein (of approximately 20 kDa) in addition to the bovine serum albumin carrier protein added during extraction (compare lane 5 to lane 2). A small amount of a polypeptide of about 20 kDa is present in the low molecular weight eluate (Fig. 3B, lane 3) which may explain the trace ligase activity observed in this fraction (Fig. 3A, lane  3). Due to its ability to reconstitute RNA ligase activity, the 20-kDa protein alone was presumed to be the E. coli RNA ligase.
Ligase Protein Sequencing and Identification of the RNA Ligase Gene-A second aliquot of the concentrated RNA ligase activity pool was subjected to SDS-PAGE and transferred to a nylon membrane, from which individual protein bands were excised for sequencing. 15 residues of amino-terminal sequence of the 20-kDa RNA ligase protein were obtained and are given in Fig. 4A. This sequence did not match any known protein or predicted ORF in the then current Genbank/EMBL data bases. A set of degenerate oligonucleotide primers corresponding to possible coding sequences for an internal segment of the aminoterminal sequence (shown in Fig. 4A) was synthesized for use in PCR. Fig. 4B shows that these oligonucleotides successfully amplified a DNA fragment of approximately 55 bp from E. coli cells or genomic DNA (lanes 1 and 2) but produced no product using S. cerevisiae cells as a template, or in reactions lacking either primers or template (lanes 3-5). The amplified DNA was cloned and sequenced.
In order to facilitate the cloning of the E. coli RNA ligase gene, its chromosomal location was determined using a genomic mapping blot. A radiolabeled DNA probe was created by PCR using the degenerate primers described above and hybridized to a membrane containing the Kohara "mini-set" of ordered, overlapping E. coli genomic phage clones (24). Hybrid-    (25), the determined amino-terminal protein sequence and 14 bp of unambiguous genomic DNA sequence were submitted for matching against their sequence data base. Exact matches to both sequences were found in a theoretical ORF located at 2.8 min on the E. coli chromosome, and these researchers kindly provided the nucleotide sequence of a 3-kbp region surrounding this ORF.
Analysis of the 2Ј-5Ј RNA Ligase Gene-Theoretical translation of the DNA sequence of the ligase ORF predicts a protein of 176 amino acids with a molecular mass of 19,934 Da (Fig.  5A). A methionine residue is encoded immediately prior to the first residue of the amino-terminal sequence obtained from the purified protein and is likely to be removed in vivo. A putative Shine-Delgarno ribosomal recognition site is located 14 bp upstream of the RNA ligase ORF (Fig. 5A). A possible match to the so-called "gearbox" transcriptional promoter consensus, which is typically found upstream of genes whose expression is inversely proportional to growth rate, can be found beginning at Ϫ119 upstream bp of the ORF (26). A possible s Ϫ35 region (often found without a Ϫ10 consensus), which may be recognized by E. coli RNA polymerase bearing the stationary phase sigma factor, can be seen at Ϫ293 bp (27). The nearest potential 70 housekeeping-type promoter consensus is found at Ϫ749 bp (not shown). Which promoter elements are utilized in vivo will need to be determined experimentally.
The RNA ligase ORF is closely flanked by two other ORFs. The previously characterized sfs1 gene (28) lies immediately downstream of the ligase ORF with the same polarity and is transcriptionally regulated by control elements that overlap the ligase coding sequence (Fig. 5A). A theoretical ORF encoding a putative RNA helicase of the DEAH family is found immediately upstream of the RNA ligase ORF but with the opposite polarity (25).
Comparison of the predicted RNA ligase protein sequence to translations of the GenBank/EMBL data base using the BLAST algorithm (29) identified two highly similar protein sequences, predicted ORFs of unknown function from Methanococcus jannaschii (30) and Bacillus stearothermophilus (31). The E. coli, M. jannaschii, and B. stearothermophilus sequences are at least 23% identical and almost 50% similar over the entire length of the E. coli ligase protein (Fig. 5B). The Bacillus protein bears an extension of 129 amino acids at its carboxyl terminus, but this additional sequence does not have any significant matches in the available sequence data bases. Alignment of the three sequences reveals three highly conserved regions (a, b, and c in Fig. 5B) which may represent important functional domains of these proteins.
Genomic Disruption of the RNA Ligase Gene-A genomic disruption of the putative RNA ligase gene was performed in order to confirm that the protein identified by purification of RNA ligase activity was in fact the genuine ligase enzyme and to observe any phenotypes caused by a lack of RNA ligase function. The disruption was performed according to the method of Hamilton et al. (22) utilizing homologous recombination and subsequent resolution of a temperature-sensitive (ts) plasmid bearing an interrupted copy of the RNA ligase gene. First, the ligase ORF and 200 bp of upstream flanking sequence were amplified from genomic DNA using unique oligonucleotide primers and cloned into the pBlueScript vector to create the plasmid pBS-lig. A kanamycin resistance gene cassette was inserted into a unique restriction site at ϩ45 bp in the ligase ORF, and the entire interrupted gene was subcloned into a plasmid containing a ts replicon and chloramphenicol resistance to create the plasmid pTSIL. After plasmid integration and resolution in a suitable RecA ϩ E. coli parent strain (as described under "Experimental Procedures"), 64 candidate colonies were recovered. One of these was found to be both kanamycin-resistant and chloramphenicol-sensitive at 43°C, indicating stable chromosomal insertion. Insertion of the kanamycin cassette into the correct chromosomal locus was confirmed by PCR amplification of genomic DNA from individual colonies of the knockout isolate (Fig. 6A). Amplification using primers hybridizing either 1 kbp upstream and at the 3Ј end of the ORF (set a) or at the 5Ј end of the ORF and 1.5 kbp downstream (set b) gave a product which in the disrupted isolates was increased by 1200 bp, precisely the size of this cassette. When whole cell extracts of isolates of this knockout strain were assayed for RNA ligase activity, none was detected (Fig. 6B, compare lanes 3-5 with 6 and 7).
Ligase Knockout Growth-Effects of the disruption of RNA ligase expression in E. coli on overall fitness were examined by assaying bacterial growth under a variety of conditions. RNA ligase knockouts were viable and showed wild-type growth rates at temperatures ranging from 23 to 43°C (not shown). The growth curve of knockout isolates at 37°C was essentially identical to that of the parent strain (Fig. 7A). The effects of moderate amounts of extra chromosomal expression of the 2Ј-5Ј RNA ligase in knockout and wild-type strains were also tested. E. coli strains were transformed with the pBS-lig construct which fortuitously supported expression of RNA ligase activity at approximately 10 times wild-type levels (as assayed by measuring specific activity in whole cell extracts) but still at a level of protein undetectable in crude extracts by SDS-PAGE and silver staining (not shown). All RNA ligase overproducing strains (Fig. 7B, B ϩ D), but not those transformed with vector alone (A ϩ C), were temperature-sensitive, being viable at 37°C but unable to grow at 43°C. These overproducing strains also showed a slow growth rate and reduced carrying capacity at stationary phase at 37°C, as shown in Fig. 7A. Thus, the overproduction of E. coli RNA ligase has a toxic effect.
Enzyme Equilibrium-The equilibrium of the RNA ligation reaction was studied in vitro in order to gain insight into the in vivo function of this enzyme. The time course of action of purified E. coli RNA ligase on tRNA Tyr half-molecules and on tRNA Tyr produced by ligation of tRNA half-molecules with E. coli RNA ligase (creating a 2Ј-5Ј linkage in the anticodon loop) was assayed. tRNA Tyr produced by ligation of half-molecules using T4 RNA ligase and PNK (to produce a 3Ј-5Ј junction) was also tested as a control. Fig. 8A shows that purified E. coli RNA ligase specifically cleaved 2Ј-5Ј-linked substrates to fragments comigrating with authentic Tyr half-molecules (lanes 0 -5, E. coli lig) with approximately the same kinetics as ligation of half-molecules by that enzyme (lanes 0 -5, Tyr 1/2's). tRNA Tyr with a 3Ј-5Ј linkage at the ligation junction was not cleaved (lanes 0 -5, T4 lig.). The identity of the tRNA cleavage products was confirmed by the ability of purified E. coli RNA ligase to rejoin them, as demonstrated in Fig. 8A (E. coli dig.). Beginning with either pure half-molecules or ligated tRNA, at 5 min of incubation the molar ratio of substrates to products ap-proached the same value (4 -5:1, halves:full-length), as quantified in Fig. 8C. DISCUSSION Fig. 1 demonstrate that the E. coli 2Ј-5Ј RNA ligase requires modified nucleosides in artificial ligation substrates. This suggests that the in vivo substrate(s) of this enzyme is modified and therefore is likely to be a stable RNA as these modifications occur exclusively in stable RNAs in E. coli (32). Modified nucleosides may be recognized directly by the enzyme, as has been shown to occur in the interactions between some tRNAs and tRNA aminoacyl synthetases (33), or these base modifications may be required to stabilize the tRNA splicing substrates in a conformation that can be recognized by this enzyme. Modified nucleosides have been demonstrated to stabilize biologically active conformers of tRNAs in other systems (34,35). The apparent requirement of E. coli 2Ј-5Ј ligase for nucleoside modifications Ϫ35 and Ϫ10 ϭ consensus E. coli promoter sequences, gbox Ϫ35 and gbox Ϫ10, potential gearbox promoter consensus sequences; s Ϫ35, potential stress-induced promoter consensus sequence. B, alignment of E. coli RNA ligase protein sequence with potential homologs from Methanococcus and Bacillus. Identical residues are boxed in black and similar residues are boxed in gray. The species of origin of each ORF is indicated. and the preference shown by this enzyme for a subset of S. cerevisiae tRNA splicing substrates suggest that the E. coli ligase is likely to act upon a tRNA or tRNA-like molecule in vivo. Comparison of the four S. cerevisiae tRNA species that are ligated by the E. coli RNA ligase to the six that are not (8) does not reveal any obvious consensus of sequence or base modifications that might be recognized. The preference of the E. coli RNA ligase for yeast tRNA Tyr half-molecules over tRNA Phe half-molecules (Fig. 1C) suggests that this enzyme has the ability to discriminate among individual tRNA species.

E. coli RNA Ligase Substrates-The results shown in
Ligase Purification-Purification of the E. coli 2Ј-5Ј RNA ligase over 1000-fold from crude extracts provided highly purified protein fractions but not a single homogeneous polypeptide. Contaminating proteins remaining at the final stages of purification probably represent molecules with properties very similar to the 2Ј-5Ј RNA ligase, but are not likely to be components of a macromolecular complex as their concentrations peak in different fractions during gel filtration (data not shown). The tight binding of E. coli RNA ligase to immobilized prokaryotic and eukaryotic tRNA provides additional evidence that the ligase recognizes a tRNA or tRNA-like substrate in vivo. The ability of the RNA ligase protein to refold and reconstitute enzymatic activity after SDS-PAGE suggests a stable, self-folding structure for this protein.
Ligase Gene Sequence-Theoretical translation of the nucleotide sequence of the RNA ligase gene predicts a polypeptide with size, charge, and other biochemical properties in excellent agreement with those observed for the ligase protein. The ligase does not appear to be expressed as part of a multicistronic operon as the nearest upstream ORF with the same polarity is located about 12 kbp away, and the adjacent downstream ORF (encoding the sfs1 protein) has its own promoter and regulatory elements (28). The cis-acting sequences controlling RNA ligase expression therefore remain to be determined.
The apparent conservation of the RNA ligase protein sequence between such distantly related bacterial species as E. coli and B. stearothermophilus and across kingdoms to the archaeote M. jannaschii suggests an ancient origin for this enzyme. This is in agreement with the observation of RNA ligase activity in extracts of a wide variety of bacterial species (8). The short blocks of high similarity between these predicted proteins as well as the lower dispersed similarity throughout suggest that the alignment is meaningful and is likely to represent a homologous origin and function for these proteins. If this alignment truly means that the 2Ј-5Ј RNA ligase is highly conserved between proteobacteria and Archaea, an important metabolic function for this enzyme is implied.
The RNA ligase enzyme may have been lost from some evolutionary branches between the low G ϩ C Gram-positive bacteria and the Gamma division proteobacteria however, since RNA ligase activity was not originally detected in Desulfurovibrio, Paracoccus, or Rhodopseudomonas species, although this may have been due to some artifact of extract preparation or ligation assay (8). No obvious homolog to the 2Ј-5Ј RNA ligase protein can be found in the completed sequence of the H. influenzae chromosome which, although closely related to E. coli by rRNA sequence comparisons, has an extremely streamlined genome less than half the size of the chromosomes of other proteobacteria (13). Haemophilus may have discarded the ligase activity or transferred this function to another polypeptide due to the selection pressures that caused the  drastic decrease in size of its genome. This would imply (as did the lack of detectable activity in several bacterial species) that the RNA ligase may not perform a function absolutely necessary for bacterial survival but may be conditionally required under growth conditions encountered by a wide variety of species.
Genomic Disruption-Disuption of the genomic locus encoding the putative ligase protein confirmed that the correct polypeptide had been purified, since a complete loss of ligase activity in cell extracts ensued. This protein therefore appears to be the only enzyme in E. coli capable of ligating yeast tRNA half-molecules. The fact that genomic ligase knockout isolates were viable demonstrates that RNA ligase is not absolutely required for survival under laboratory growth conditions. Although the disrupted strains do not display a lethal phenotype, they can be examined for more subtle effects on growth and RNA metabolism. The availability of viable knockouts will also provide a useful null background for the expression of affinitytagged or mutagenized ligase protein for use in further biochemical experiments. The toxic effects of moderate levels of ligase overexpression also imply some sort of interaction between the ligase enzyme and other cellular factors required for growth.
Examination of Enzyme Equilibrium-The reaction catalyzed by the E. coli ligase enzyme was shown to be fully reversible with an apparent equilibrium constant near unity, but favoring cleavage of 2Ј-5Ј bonds. The tendency toward cleavage may perhaps be explained by the thermodynamics of phosphodiester bond cleavage and formation, which have been investigated thoroughly in the hammerhead ribozyme system. The favorable entropy of bond cleavage causes the internal equilibrium of the hammerhead ribozyme to favor cleavage of 3Ј-5Ј phosphodiesters to 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl termini despite the unfavorable enthalpy associated with cyclic phosphate formation (36). For the cleavage reaction catalyzed by the E. coli RNA ligase, an increase of entropy in what is essentially a unimolecular reaction (given the tight structural association of tRNA half-molecules) may be due to the additional degrees of freedom available to released termini and to the disruption of water ordered in and about the closed, structured anticodon loop.
This observed equilibrium of cleavage and ligation appears to explain the maximum extent of ligation observed in in vitro activity assays. However, it begs the question of whether the function of the enzyme is to catalyze ligation or cleavage in vivo. Despite the fact that the equilibrium observed in vitro favors cleavage, the direction of the equilibrium in vivo will depend on the effective concentrations of substrates available for each reaction. If the true in vivo substrate is tRNA, and ligated tRNA products are utilized for translation and thereby removed from the pool of substrates available to the ligase, then the ligation reaction will be favored. If, however, the cleavage products are removed by some process such as ribonucleolytic degradation, then the equilibrium will favor cleavage. To propose a cleavage function for the ligase enzyme in vivo, however, it is necessary to posit a source of substrates with 2Ј-5Ј bonds, presumably in the context of a tRNA. No other E. coli enzyme is known or proposed that might form internal 2Ј-5Ј linkages in a tRNA structure. A variety of endoribonuclease activities, however, could theoretically produce substrates for ligation by the E. coli RNA ligase, and in fact an activity capable of doing so has been observed in E. coli extracts. 3 An enzyme capable of cyclizing free 3Ј-phosphates to 2Ј,3Ј-cyclic phosphates has also recently been discovered in E. coli. 4 Thus the available evidence, while circumstantial, favors a ligation function for this enzyme in vivo.