Functional Effect of Deletion and Mutation of theEscherichia coli Ribosomal RNA and tRNA Pseudouridine Synthase RluA*

The Escherichia coli generluA, coding for the pseudouridine synthase RluA that forms 23 S rRNA pseudouridine 746 and tRNA pseudouridine 32, was deleted in strains MG1655 and BL21/DE3. The rluA deletion mutant failed to form either 23 S RNA pseudouridine 746 or tRNA pseudouridine 32. Replacement of rluA in trans on a rescue plasmid restored both pseudouridines. Therefore, RluA is the sole protein responsible for the in vivo formation of 23 S RNA pseudouridine 746 and tRNA pseudouridine 32. Plasmid rescue of bothrluA − strains using an rluA gene carrying asparagine or threonine replacements for the highly conserved aspartate 64 demonstrated that neither mutant could form 23 S RNA pseudouridine 746 or tRNA pseudouridine 32 in vivo, showing that this conserved aspartate is essential for enzyme-catalyzed formation of both pseudouridines. In vitro assays using overexpressed wild-type and mutant synthases confirmed that only the wild-type protein was active despite the overexpression of wild-type and mutant synthases in approximately equal amounts. There was no difference in exponential growth rate between wild-type and MG1655(rluA −) either in rich or minimal medium at 24, 37, or 42 °C, but when both strains were grown together, a strong selection against the deletion strain was observed.

Ribosomal RNA, considered to be the functional heart of the ribosome (1), contains a variety of modified nucleosides of unknown function (2). The most common single modification is the conversion of uridine to pseudouridine (⌿), 1 the 5-ribosyl isomer of uridine (3). ⌿ is formed by isomerization of specific uridines after the RNA chain is formed. The mechanism of the reaction, which involves breaking of the N 1 -glycosyl bond, rotation of the uracil ring, and formation of a C 5 -glycosyl bond is unknown but is thought to involve an active site carboxyl group of an aspartate residue (4). ⌿ is found in the rRNA of all organisms so far examined (5), and in Escherichia coli, which has one ⌿ in the 16 S RNA (6) and nine in the 23 S RNA (7,8), it is the most prevalent of the modified nucleosides. The single ⌿ in the 16 S RNA is found adjacent to the "530" loop, whose sequence has been almost completely conserved in all organisms and is known to be involved in the fidelity of codon recognition (reviewed in Refs. 9 and 10). In 23 S RNA, the nine ⌿ residues are distributed among three distinct areas, which, despite their separation in two-dimensional secondary structure representations, are at or near the peptidyl transferase center when in the ribosome (11). However, despite the congruence of the ⌿ residues with the two functional centers of the ribosome, namely decoding and peptide bond formation, there is so far no known role for ⌿ in the process of protein synthesis.
To approach this problem, we have initiated a program to identify the genes for the specific synthases that make the 10 ⌿ in E. coli rRNA on the assumption, subsequently shown to be correct, that distinct synthases are used to form ⌿ at the different sites in the rRNA molecule. Once identified, gene inactivation will result in the loss of a specific synthase and should therefore cause the loss of specific ⌿ residues for which the effect on cell physiology can then be assessed. Thus far, three ⌿ synthase genes have been inactivated in this manner. One, rluC, codes for a synthase solely responsible for formation of ⌿ residues 955, 2504, and 2580 in 23 S RNA (12) and a second, rluD, codes for the synthase that makes 23 S RNA ⌿1911, 1915, and 1917 (13). The third, rsuA, codes for the synthase that forms 16 S RNA ⌿516 (14). Inactivation of rluC and rsuA and the consequent loss of their respective ⌿ had no physiological effect. However, disruption of rluD with the loss of its three ⌿ severely inhibited cell growth.
Another synthase, RluA, was shown to form only ⌿746 when in vitro transcripts of 23 S RNA were the substrate. The synthase also specifically catalyzed the formation of ⌿32 in E. coli tRNA Phe (15). The ability to be highly specific for a single site in more than one class of RNA, a property termed "dual specificity" (15), has since been reported for another ⌿ synthase (16) as well as for a ribose methylating enzyme, although in the latter case the dual specificity resides in the guide RNA (17). A question left open by the work on RluA was whether it is the only synthase in E. coli capable of forming rRNA ⌿746 and tRNA ⌿32 and what the effect of its absence would be on the cell. This issue has now been addressed by deleting rluA and comparing the growth rate with wild type both separately and in a competition experiment. In addition, by mutating aspartate 64, which was predicted to be an essential residue by virtue of its location in a conserved sequence, HRLD (4), we have shown that RluA, as well as RsuA (14), RluD, 2 and TruA (4), requires this residue for function.

EXPERIMENTAL PROCEDURES
rluA and miaA Ϫ rluA Ϫ Strains-The rluA gene was deleted by the method of Hamilton et al. (18). The insert, cloned into the XbaI and KpnI sites of pMAK705, was prepared by PCR as described by Nelson and colleagues (Fig. 2 in Ref. 19). It contained 818 bases 5Ј to the AUG start and 785 bases 3Ј to the UAA termination codon. Sixteen bases of the N-terminal portion of the gene and 52 bases of the C terminus were retained with the remainder being replaced by the kanamycin resistance gene, obtained by PCR amplification from pUC4K (Amersham Pharmacia Biotech, catalog no. 27-4958-01). The host strain for pMAK705 was the leucine auxotroph MC1061, as described by Hamilton et al. (18). The deleted rluA gene was moved into strains MG1655 (Ref. 20; the gift of Dr. Kenneth Rudd, this department) and BL21/DE3 (Novagen, Inc.) by bacteriophage P1 transduction (21). Selection was done on either rich (LB, Ref. 22) or minimal (M9 ϩ ) medium containing 0.05 mg/ml kanamycin. The miaA deletion was moved from strain NU426 carrying the miaA::cat insertion (the kind gift of Malcolm Winkler, University of Texas, Houston Medical School) by P1 transduction into the MG1655 and MG1655(rluA Ϫ ) strains with selection on LB containing 34 mg/ml chloramphenicol in addition to kanamycin.
Wild-type Rescue Plasmids-The preparation of wild-type rescue plasmid pET15b/rluA has been described previously (15). Wild-type rescue plasmid pTrc99A/rluA was constructed by insertion into the NcoI and HindIII sites of pTrc99A (Amersham Pharmacia Biotech, catalog no. 275007-01) of a segment of DNA that was PCR-amplified from pET15b/rluA and consisting of the rluA gene starting from the initiator AUG and ending at the terminator UAA. The N-terminal primer had an NcoI site adjacent to the initiating AUG, whereas in the reverse orientation, the C-terminal primer incorporated a HindIII site after the terminator UAA.
Mutant Rescue Plasmids-These plasmids were prepared by the megaprimer PCR mutagenesis procedure (23). PCR reactions were performed using the pTrc99A/rluA rescue plasmid as template and three oligonucleotide primers, two outer primers, which were upstream and downstream of the mutation site, and one mutagenic primer. The upstream and downstream primers contained the restriction sites NcoI and HindIII, respectively, so that the product could be ligated directly into pTrc99A. Mutagenesis was carried out in three steps. The initial PCR reaction was performed with either mutagenic primer 5Ј-CATCG-TCTGACTATGGCTACC-3Ј for the D64T mutation or 5Ј-CATCGTCT-GAATATGGCTACC-3Ј for the D64N mutation (mutation sites shown in bold) and with the downstream primer 5Ј-GGGAAGCTTTTAAAAA-TCCGCTGGCGC-3Ј having a HindIII site (underlined). A 100-l reaction contained 50 ng of template plasmid DNA, 15 pmol each of the mutagenic primer and downstream primer, 3 units of Pfu DNA polymerase (Promega), 0.2 mM dNTPs, 20 mM Tris, pH 8.75, 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 2 mM MgCl 2 , 0.1% Triton X-100, and 0.1 mg/ml bovine serum albumin. The mixture was denatured at 95°C for 60 s, and then 10 cycles of amplification (95°C, 30 s; 47°C, 60 s; 72°C, 70 s) were performed, followed by a 5-min extension at 72°C. Fifty pmol of the upstream primer was added (5Ј-GGGGCCATGGATGGGGATGGAAA-ACTAC-3Ј, NcoI site underlined), and the reaction mixture was subjected to the same amplification program. Finally, 50 pmol of downstream primer was added, and the sample was subjected to the same amplification program again. The amplified product was purified by gel electrophoresis, digested with NcoI and HindIII, and ligated with similarly digested and purified pTrc99A for 16 h at 16°C. The ligation mixture was transformed into Novablue cells (Novagen, Inc.) by standard methods yielding 4 positive clones of 5 tested for D64T and 3 positive of 4 tested for D64N. DNA sequencing of the isolated plasmids verified that the expected mutation had been produced at the desired site. Transfer of the mutant rluA genes into pET15b was done by PCR amplification of the mutant rluA genes in pTrc99A. The N-terminal primer extended from Ϫ9 to ϩ18, where the A of the initiating AUG of rluA is ϩ1 with changes at Ϫ2 to Ϫ5, to create an XhoI site adjacent to the initiating AUG. The C-terminal primer, in the reverse orientation, extended from ϩ643 to ϩ669, where the last sense nucleotide is 657, and contained mismatches at ϩ661 to ϩ666 to create a BamHI site. The amplified product was purified, digested with XhoI and BamHI, and subsequently ligated with identically treated pET15b vector for 16 h at 16°C. DNA sequence analysis verified the constructs (data not shown).
rRNA and tRNA Isolation and Sequencing-5-[ 3 H]Uridine-labeled transcripts of full-length 23 S RNA (specific activity 168 dpm/pmol uridine residues) were prepared as described previously (13). The 5-[ 3 H]uridine-labeled transcripts of tRNA Phe (199 dpm/pmol), tRNA Cys (1269 dpm/pmol), and tRNA 4 Leu (454 dpm/pmol) were prepared as described for tRNA Phe (15) but using pTFMa-ECys (the gift of Ya-Ming Hou, Thomas Jefferson University, Philadelphia, PA) as template for tRNA Cys and pUC19/tRNA 4 Leu (UAA) for tRNA 4 Leu . Ribosomal RNA for ⌿ sequencing was prepared according to King and Schlessinger (24) with omission of the LiCl precipitation step. ⌿ sequencing of rRNA was performed as described previously (7,25). tRNA for ⌿ sequencing was isolated as described (26) from cells grown to an A 550 of 1.0 in LB medium. ⌿ sequencing of tRNA Cys was done exactly as for rRNA using a primer complementary to residues 61-76.
Growth Experiments-For the individual exponential phase growth experiments, overnight cultures at 37°C in the medium to be tested were diluted 50-fold (minimal medium) or 100-fold (rich medium) and placed at the testing temperature. Cell density was monitored at 600 nm. Viable cells in the mixed competition experiment were determined by plating on LB, in which both wild-type and mutant grow, and on LB plus kanamycin, in which only rluA Ϫ grows. For the 1:10 3 dilution series (Fig. 8), the first four cycles were analyzed by first plating aliquots on LB and then direct transfer of individual colonies to LB plus kanamycin (patch analysis). 100 colonies originating from each of four individual flasks (400 colonies total) were analyzed per time point. For the last two cycles, direct plating of three aliquots per each of four flasks to LB with and without kanamycin was used for a total of 12 platings per medium per time point. For the 1:1.6 ϫ 10 6 dilution series, direct plating of two aliquots per each of three flasks (six platings per time point) to LB with and without kanamycin was used for the rluA Ϫ cells. Patch analysis using 100 colonies from each of two flasks (200 colonies total) was used for the rsuA Ϫ and rluC Ϫ cells. The fraction of rluA Ϫ in the mixture is the number of colonies on LB plus kanamycin divided by the number on LB alone. The number of cell doublings (G) is calculated from the dilution factor (DF) at each cycle using the relation DF ϭ N/N o ϭ 2 G . Thus, to reach the same cell density after a 1:10 3 dilution requires 9.97 cell doublings, and the 1.6 ϫ 10 6 dilution corresponds to 20.6 cell doublings.
Other Methods and Materials-Transformants of wild-type, MG1655(rluA Ϫ ), and MG1655(miaA Ϫ rluA Ϫ ) strains with pTrc99A and pTrc99A/rluA as well as wild-type and BL21/DE3(rluA Ϫ ) with pET15b and pET15b/rluA were selected on LB plates containing 0.1 mg/ml carbenicillin. All subsequent growth media for the transformants also contained 0.1 mg/ml carbenicillin to retain the plasmid in the carbenicillin-sensitive host cells. In vitro assays of ⌿ synthase activity were done as described previously (15). Pfu DNA polymerase was from Promega. All other enzymes and primers were obtained and polyacrylamide gel electrophoresis was performed as described previously (13).

RESULTS
Identification of RluA as the Only Synthase for Formation of ⌿746 in 23 S RNA-Overexpressed and purified RluA converts U746 in E. coli 23 S rRNA transcripts to ⌿746 and U-32 in tRNA Phe transcripts to ⌿32 (15). In vitro, the enzyme was highly specific for these two sites. Comparison of the sequence surrounding the sites of the two U residues selected for conversion to ⌿ revealed that both possessed the same sequence immediately 3Ј to the U in question, thus providing a rationale for the dual specificity exhibited by this enzyme (15). These experiments did not, however, show whether additional enzymes existed in the cell that were also capable of ⌿ formation at these sites, nor did they show the effect of deletion of these ⌿ residues. Therefore, the gene was deleted by insertion of the kanamycin resistance gene (18) in strain MC1061. The deletion was confirmed by PCR amplification from the N and C terminii of the rluA gene in the chromosomal DNA of the deletion mutant. The wild-type control had the expected 670-base pair band, whereas the mutant having a kan insert was 1.4 kilobase pairs in size. The presence of the kan gene was further confirmed by amplification from the N and C termini of the kan gene. The mutant produced the expected 1.3-kilobase pair band, whereas no band was obtained from the wild type.
To assess the physiological effects of this gene deletion un-complicated by the other mutant genetic loci present in MC1061 (18), the deletion was transferred by bacteriophage P1 transduction into MG1655 in which the sequenced genome (20) provided a well defined background. Transductants were selected by resistance to kanamycin. PCR amplification confirmed the presence of the kanamycin insert, and ⌿ sequencing analysis of the ribosomal RNA from the mutant strain showed unequivocally that ⌿746 was absent (data not shown).
To prove that the loss of ⌿ attendant on the deletion of rluA was a direct consequence of the deletion and not because of some downstream polarity or other indirect effect, the gene was replaced in trans by transformation of the deletion strain with a rescue plasmid that contained only the rluA gene inserted into pTrc99A. Wild type and MG1655(rluA Ϫ ) were transformed with both the rescue plasmid and the control vector pTrc99A and selected on carbenicillin plates. Ribosomal RNA was isolated and sequenced for the presence of ⌿ (Fig. 1). Comparing the rluA ϩ lanes with the rluA Ϫ lanes, it is clear that the stop at residue 746 in the ϩ CMC lane of the rluA ϩ set is absent in the ϩ CMC lane of the rluA Ϫ pair. Recall that in this method of sequencing, reverse transcriptase halts one residue 3Ј to the CMC-⌿ (7,25). However, when the rescue plasmid was introduced into the rluA Ϫ strain, ⌿746 was again found. The stop seen in all lanes results from m 1 G745, because all of the RNAs were isolated from cells. We conclude that the loss of ⌿746 is a direct result of deletion of rluA and that RluA is the sole gene product capable of synthesis of ⌿746. Additional sequencing analyses verified that only ⌿746 was absent from the rluA Ϫ strain (data not shown).
Identification of a Synthase Amino Acid Essential for ⌿746 Formation-Recently, it has been shown that the replacement of Asp-60 in a conserved (G/H)(R/a)(L/t)(D) motif (lowercase identifies a rare event), by Ala, Asn, Glu, Lys, or Ser in the pseudouridine synthase TruA resulted in the loss of catalytic activity while retaining binding ability (4). There is an equivalent residue, Asp-64, in a similarly conserved motif, HRLD, in the RluA synthase, and it is the only Asp residue in such a motif in the molecule. To test the possibility that Asp-64 could be an essential residue of this enzyme, we mutated it to Thr and Asn. This was done by megaprimer mutagenesis (23). The two mutated rluA were cloned in pTrc99A and transformed into MG1655(rluA Ϫ ) cells to assess the function of these mutant enzymes in vivo. The wild-type rescue plasmid served as a control. Ribosomal RNA was isolated and sequenced for the presence of ⌿ (Fig. 2). It is clear that the only strain able to make ⌿746 is the strain carrying the wild-type rescue plasmid. Neither plasmid carrying a mutant rluA was any more effective than the vector alone. Thus, in vivo, the single mutation D64T or D64N is sufficient to block synthesis of an active ⌿746 synthase. One might conclude from this experiment that Asp-64 is an amino acid that actively participates in catalysis, but it could also be that its role is in the maintenance of the correct conformation of the enzyme. In the latter event, the replacement of Asp-64 by another amino acid might make the protein susceptible to protease degradation.
To address this question, we turned to the BL21/DE3 strain and pET15b to obtain stable overexpression of the mutant proteins. The rluA Ϫ gene was transferred into BL21/DE3 by P1 transduction from MC1061 with selection by kanamycin resistance. PCR amplification confirmed the presence of the kanamycin insert, and ⌿ sequencing analysis of the ribosomal RNA from the mutant strain showed the absence of ⌿746 (data not shown, but see Fig. 4 for an equivalent result). Both the wildtype and mutant rluA constructs were subcloned into pET15b. DNA sequencing analysis (data not shown) confirmed that the desired mutants had been produced in pET15b. The BL21/ DE3(rluA Ϫ ) cells were then transformed with vector alone or with the rluA constructs in pET15b. Transformants were selected on carbenicillin plates. The BL21/DE3 cells carrying either the vector or the various rluA constructs were then induced. After a 3-h induction with isopropyl-1-thio-␤-D-galactopyranoside, samples from each cultures were taken out for protein analysis on SDS-polyacrylamide gels as well as for ribosomal RNA isolation and ⌿ sequencing analysis. Fig. 3 shows that a strongly overexpressed protein band at about 27 kDa, the expected size, was found in the cells carrying both wild-type and mutant rluA constructs, whereas there was no such overexpressed protein band in the cells carrying the vector only. Furthermore, induction was required to produce the band. The intensity of the 27-kDa band appeared the same in both wild-type and mutant constructs. ⌿ sequencing analysis of the rRNA showed that, as with the results obtained in Fig. 2, the mutant rescue plasmids were unable to form ⌿746 (Fig. 4). We conclude that the two mutant rluA constructs produced stable proteins that had, nevertheless, lost the capability to isomerize U746 to ⌿ as a result of the replacement of Asp-64 by Thr-64 or Asn-64.
Affinity purification of the overexpressed proteins shown in Fig. 3 and assay of in vitro activity using 23 S [ 3 H]RNA transcripts as substrate gave the same result, namely that whereas unit stoichiometry of ⌿ formation could be obtained for the wild-type construct, as reported previously (15), both the D64T and D64N mutants were totally inactive (Fig. 5). RluA Is the Only Protein Capable of ⌿32 Formation in tRNA-⌿32 is found in four tRNAs of E. coli, tRNA Phe , tRNA Cys , tRNA 4 Leu (UAA), and tRNA 5 Leu (CAA) (28,29). We previously showed that RluA formed ⌿32 on a transcript of tRNA Phe in an in vitro reaction (15), a result that led to the concept of dual specificity for this enzyme. All five of these RNAs, the 23 S RNA and the four tRNAs, share a common sequence surrounding the ⌿ residue, namely (A/G)⌿UN-(A/C)AAA. Therefore, it seemed reasonable that these other tRNAs could also serve as a substrate for RluA. To test this hypothesis, tRNA Cys and tRNA 4 Leu transcripts were assayed for their ability to react with RluA (Fig. 6). Both transcripts were active. The rate and yield with tRNA Cys was virtually identical to that with tRNA Phe , whereas tRNA 4 Leu was somewhat less reactive for unknown reasons.
To determine whether RluA is the only protein in E. coli capable of tRNA ⌿32 formation, tRNA from the rluA Ϫ strain was analyzed. However, before the reverse transcription assay could be used, obstacles created by the presence of other modified nucleosides in the tRNAs that block reverse transcriptase had to be overcome. ms 2 i 6 A37, present in all three tRNAs, is a strong inhibitor of reverse transcription. Moreover, tRNA Phe also has acp 3 U47, another strong blocker, and tRNA 4 Leu has cmnm 5 Um34 (29) only two residues away from ⌿32. To replace ms 2 i 6 A37 by A37, a deletion of miaA, the gene responsible for the enzyme that forms i 6 A (30, 31), was transduced into MG1655(rluA Ϫ ) by bacteriophage P1. To avoid the other modified nucleosides, tRNA Cys was chosen for analysis. ⌿ sequencing of tRNA Cys by the reverse transcription procedure is shown in Fig. 7. Note that this method obviated the need to purify tRNA Cys from the total tRNA preparation. In the miA Ϫ rluA ϩ tRNA, CMC-dependent stops were found at positions 33, 40, and 56 corresponding to the ⌿32, ⌿39, and ⌿55 residues known to be in this tRNA. No stop corresponding to ⌿32 was found in the miA Ϫ rluA Ϫ strain transformed with the pTrc99A vector only, although both ⌿39 and ⌿55 were present. However, when plasmid carrying the rluA gene was used, ⌿32 reappeared. Clearly, RluA is the only protein able to form ⌿32 in E. coli. Furthermore, when the two rluA genes mutant at Asp-64 were used, no ⌿32 was formed, showing that both ⌿746 and ⌿32 formation requires the same essential Asp residue.
Effect of the rluA Deletion on Growth-Cells with ribosomes lacking 23 S RNA ⌿746 and tRNA ⌿32 were viable and appeared to grow normally. To better detect small metabolic defects, growth rates were measured at different temperatures in both rich and minimal glucose media. The growth experiments were done in the MG1655 genetic background after transduction of the rluA Ϫ gene from strain MC1061. Both wild type and MG1655(rluA Ϫ ) were transformed with both the rescue plasmid and its control, and exponential growth rates were measured for all four strains (Table I). Although both rich and minimal media were tested over a temperature range from 24 to 42°C, no significant difference in growth rate between the wild-type and rluA Ϫ strain was observed.
The above experiment did not, however, test the effect of the rluA mutation under more natural conditions, namely in competition with non-mutant cells. Moreover, only the exponential phase of growth was examined. To test these additional conditions, the following experiment was performed. Exponential phase wild-type and rluA Ϫ cells were mixed and grown for 24 h to stationary phase at 37°C in LB medium with aeration. A sample was taken for analysis of viable cells, and the cultures were diluted either 1:10 3 or 1:1.6 ϫ 10 6 and grown for an additional 24 h. In this way, the wild-type and mutant cells compete for nutrients through the entire cycle of growth, starting from stationary phase at the start of the experiment, then by dilution into lag phase and exponential phase, and again into stationary phase. Aliquots were diluted into fresh medium, and the process was repeated for a total of five or six cycles. The results are shown in Fig. 8 together with control experiments in which rsuA Ϫ and rluC Ϫ cells were also competed against wildtype cells. These strains, in which gene deletions also carry a kanamycin insert like the rluA Ϫ cells, control for an effect due to the presence of the kanamycin insert. Fig. 8A shows the relative percentage of loss of rluA Ϫ cells as a function of the number of cycles of growth to saturation followed by dilution into fresh medium. Compared with the slow decrease in the rsuA Ϫ and rluC Ϫ cells in mixed culture, the rluA Ϫ cells in an equivalent mixed population decreased sharply. Thus, when the rsuA Ϫ strain had decreased to 56% of its original value, the rluA Ϫ strain had gone down to 0.4% in one case and 0.5% in the other, which is a 124-fold decrease relative to the control rsuA Ϫ strain. Clearly rluA Ϫ cells are at a marked disadvantage when growing in competition with wild-type cells. Moreover, there was no effect of a 1600-fold difference in the extent of dilution between cycles. If selection against rluA Ϫ cells had occurred during the exponential phase of growth, a large effect would have been expected, because to reach the same cell density after a 1.6 ϫ 10 6 dilution requires 20.6 doublings, whereas a 10 3 dilution only requires 9.97 doublings. Twice as many doublings should have had twice the effect if selection had occurred in exponential phase. This effect is illustrated in Fig. 8B, where the same data are plotted versus the calculated number of cell doublings of the mixed culture. The discrepancy in the rate of decrease of the two dilutions of rluA Ϫ cells shows that the selection against rluA Ϫ cells does not occur during exponential phase but must take place elsewhere in the growth cycle.

DISCUSSION
Specificity-In a previous report, we showed that RluA in vitro was able to form ⌿746 in transcripts of E. coli 23 S RNA (15) but did not form any of the other known ⌿ residues. In the present work, we have demonstrated that RluA is the only gene product in E. coli that can carry out this reaction, because deletion of the gene caused only the loss of ⌿746 and replacement of this gene on a plasmid restored it. Nevertheless, it is possible that RluA in vivo might share the ability with another ⌿ synthase for formation of one or more of the other ⌿ sites in 23 S RNA. Only synthases for two such ⌿ are still possible candidates, namely ⌿2457 and ⌿2605, because we have shown that deletion of rluC results in the loss of ⌿ residues 955, 2504, and 2580 (12) and that deletion of rluD results in the loss of ⌿ residues 1911, 1915, and 1917 (13). Deletion experiments on candidate genes for these last two synthases are in progress.
RluA was also shown capable of forming ⌿32 in tRNA Phe in vitro (15), and in this work the reaction was extended to tRNA Cys and tRNA 4 Leu as well (Fig. 6). All three tRNAs are known to have ⌿ at position 32 when isolated from cells (28). The question then arose whether the deletion in rluA also caused the loss of ⌿32 from tRNA or whether a second synthase exists in E. coli that is able to catalyze this reaction. By sequencing tRNA Cys from the rluA Ϫ strain, we showed that ⌿32, but not ⌿39 or ⌿55, was absent. Therefore, RluA is the only protein in the cell able to carry out both reactions. Nevertheless, it is not clear whether only one of the ⌿ is the desired one, and the other is a benign by-product, or whether both are desired by the cell, and one protein has been co-opted to perform both functions.
Relationship of ⌿746 to m 1 G745 and m 5 U747-This segment of E. coli 23 S RNA is notable for its concentration of three modified nucleosides, m 1 G745, ⌿746, and m 5 U747 adjacent to each other in a small stem-loop structure. Because we previously showed that ⌿746 could be formed in vitro on rRNA transcripts, it is clear that ⌿ formation does not require the presence of either m 1 G745 or m 5 U747 (15). The present results, which show the existence of m 1 G745 in the deletion strains lacking ⌿746 (Figs. 1, 2, and 4), demonstrate that m 1 G745 synthesis is independent of prior ⌿746 formation. The existence of m 5 U747 in the deletion strains was not examined.
Reaction Mechanism-The minimum reaction required for the isomerization of uridine to pseudouridine involves cleavage of the uracil N 1 -ribosyl C 1Ј bond, rotation of the uracil ring either 180°about the N 3 -C 6 axis in the ring plane or 120°about an axis perpendicular to the ring plane, and formation of a uracil C 5 -ribosyl C 1Ј bond. Recently, a reaction mechanism was proposed for this type of isomerization that involves the pres- ence of the ␤-carboxyl of a conserved aspartate residue at the reaction center of the pseudouridine synthase TruA (4). The mechanism was proposed to be applicable to all ⌿ synthases because of the existence of a conserved sequence motif in all known or putative enzymes that takes the form in E. coli of (G/H)(R/a)(L/t)(D), where the use of lowercase letters identifies a rare event. There is a single such aspartate in RluA in the sequence HRLD. This aspartate residue, Asp-64, appears to play an important role in both 23 S RNA ⌿746 and tRNA ⌿32 formation because replacement of this aspartate by threonine or asparagine blocked formation of both ⌿ in vivo and blocked ⌿746 synthesis in vitro. ⌿32 formation by RluA mutants in vitro was not tested. Thus, the essential nature of the conserved aspartate has now been shown for both TruA and RluA. In other work (14), the conserved aspartate in the same sequence motif in RsuA, the synthase responsible for ⌿516 formation in 16 S RNA, has also been shown to be essential. Moreover, recent studies have shown that the aspartate in the same sequence motif in RluD is needed to make ⌿1911, ⌿1915, and ⌿1917 in 23 S RNA in vitro. 2 Function of ⌿-There was no difference in exponential phase growth rates when cells lacking 23 S RNA ⌿746 and tRNA ⌿32 were grown in separate cultures, even when both the medium and the temperature were varied (Table I). However, when the rluA Ϫ strain was grown in competition with wild-type cells, a marked selection against the mutant cells was observed (Fig.   8). Compared with the rsuA Ϫ strain, which also carries the same kanamycin resistance cassette, rluA Ϫ cells were more than 100 times more likely to die. Moreover, it is possible that there is little or no effect of the presence of the kanamycin resistance gene on survival. The slow decay observed for rsuA Ϫ cells may be the result of an intrinsic decrease in fitness because of the absence of the RsuA protein. This view is supported by the results with the rluC Ϫ strain, which, while also a much better survivor than the rluA Ϫ strain, was itself less fit than the rsuA Ϫ strain. Thus, the true lack of survival of rluA Ϫ cells may be even greater than indicated by the comparison with rsuA Ϫ cells.
At what stage of growth does the discrimination occur? The fact that a more than 1000-fold difference in dilution for each cycle in Fig. 8 did not change the decay rate argues strongly against the effect occurring in exponential phase. Other possibilities are the approach to stationary phase, stationary phase itself, or the lag phase before reinitiation of growth. Further experimentation will be required to answer this question. It is also not known whether the growth defect manifest in the absence of RluA is because of its inability to form ⌿746 in 23 S RNA and ⌿32 in four tRNAs or to some other unknown function of the protein. Such a situation was found for the E. coli RUMT enzyme, which catalyzes the m 5 U54 formation in tRNA. In this case, the protein was essential, and yet its methylation activity was dispensable (32). The two aspartate mutants de-  8. Growth competition between wild-type and mutant MG1655 strains. The competition experiment is described in the text. 20-ml cultures in LB medium were shaken at 37°C in 250-ml flasks. Aliquots were sampled and viable cells determined as described under "Experimental Procedures." The percentage of cells that were kanamycin-resistant at each cycle was normalized by dividing the values by the initial percentage of kanamycin-resistant cells. Initial percentages were 49 (rluA Ϫ , E), 39 (rluA Ϫ , q), 36 (rsuA Ϫ , ‚), and 46% (rluC Ϫ , OE). The values were plotted as the number of cycles of growth to saturation and dilution into fresh medium (panel A) and as the number of cell doublings calculated from the dilution factor (panel B). E, dilution factor of 1:10 3 between cycles; q, OE, and ‚, dilution factor of 1.6 ϫ 10 6 . scribed in this work should provide a means to test this possibility because even though normal amounts of mutant RluA were produced, the mutants had no ⌿ synthase activity.
This work adds RluA to the category of ⌿ synthases that can not be deleted without a serious effect on cell growth. Previously, we showed that disruption of the rluD gene, which codes for the synthase that makes ⌿1911, ⌿1915, and ⌿1917, severely inhibits the growth of E. coli (13). On the other hand, the lack of RluC, which makes ⌿955, ⌿2504, and ⌿2580, has much less of an effect on growth. Although no effect was observed when individual exponential growth rates were compared with wild type (12), Fig. 8 shows that in competition with wild type there is an effect, albeit much less than for rluA Ϫ cells. The 5-fold drop observed after only six cycles would still result in the eventual loss of the rluC Ϫ cell from the culture. There is an even smaller effect in the case of rsuA Ϫ cells, which also grew exponentially at the same rate as wild type (14). In competition, only a 2-fold drop was observed after six cycles (Fig. 8), some or all of which could be attributed to the presence of the kanamycin resistance protein and not to the absence of RsuA. Thus, there already appears to be a gradient of effects of depriving the cell of ⌿ synthases. The most severe effect was found by blocking RluD formation, which resulted in tiny colonies on plates and a readily detectable decrease in exponential growth rate (13). A strong RluA effect was found only in competition studies; the RluC effect was much less and was found only in competition, and the effect of the loss of RsuA was either very small or nonexistent.
In eukaryotes, which use a guide RNA system to specify those uridines in ribosomal RNA to be converted into ⌿, deletion of a number of guide RNAs, and thus the absence of those ⌿ residues, also had no apparent effect on cell growth or metabolism (reviewed in Ref. 5). However, so far no competition studies have been performed. It seems unlikely that two distinct systems for forming ⌿ in ribosomal RNA should have evolved without the driving force of a significant role in the survival of the cell. The discovery of strong effects in two cases in E. coli, RluD (13) and RluA (this work), marks the first two times any kind of role for ⌿ in cellular metabolism has been established. The exact nature of that role remains to be deciphered.