Potential New Antibiotic Sites in the Ribosome Revealed by Deleterious Mutations in RNA of the Large Ribosomal Subunit*

The ribosome is the main target for antibiotics that inhibit protein biosynthesis. Despite the chemical diversity of the known antibiotics that affect functions of the large ribosomal subunit, these drugs act on only a few sites corresponding to some of the known functional centers. We have used a genetic approach for identifying structurally and functionally critical sites in the ribosome that can be used as new antibiotic targets. By using randomly mutagenized rRNA genes, we mapped rRNA sites where nucleotide alterations impair the ribosome function or assembly and lead to a deleterious phenotype. A total of 77 single-point deleterious mutations were mapped in 23 S rRNA and ranked according to the severity of their deleterious phenotypes. Many of the mutations mapped to familiar functional sites that are targeted by known antibiotics. However, a number of mutations were located in previously unexplored regions. The distribution of the mutations in the spatial structure of the ribosome showed a strong bias, with the strongly deleterious mutations being mainly localized at the interface of the large subunit and the mild ones on the solvent side. Five sites where deleterious mutations tend to cluster within discrete rRNA elements were identified as potential new antibiotic targets. One of the sites, the conserved segment of helix 38, was studied in more detail. Although the ability of the mutant 50 S subunits to associate with 30 S subunits was impaired, the lethal effect of mutations in this rRNA element was unrelated to its function as an intersubunit bridge. Instead, mutations in this region had a profound deleterious effect on the ribosome assembly.

Antibiotics are among the most important medicines ever found by humans. They have saved millions of lives and helped to curb many devastating diseases. Since the advent of antibiotics in the middle of the 20th century, the main challenge of antibiotic therapy has been fighting infections caused by bacterial pathogens that have developed resistance to the available drugs. Not a single class of antibiotics has escaped the inevitable emergence of resistance. Cur-rently, the constant development of new drugs is the only available strategy to keep up with an ever-growing variety of antibiotic-resistant strains of pathogens.
The development of microbial genomics brought new hope to the drug discovery field. Several hundred bacterial genes were found to be essential (1,2), and many others were identified as virulence genes (3). The products of such genes are viewed as promising drug targets. Until now, however, very few successful drug candidates have been produced with the help of genomics (4).
An alternative approach is to follow the lead of nature and to develop new drugs that act on the targets "experimentally validated" in the course of evolution. Among such targets, the ribosome is the preferred one; more than half of all the known antibiotics arrest cell growth by interfering with the ribosome functions and inhibiting protein synthesis.
The ribosome, which has a molecular weight of 2.5 million daltons, is one of the largest enzymes in the cell. Each of the two ribosomal subunits is built of RNA and proteins, with ribosomal RNA being the main structural and functional component of the ribosome. The ability of the ribosome to carry out efficient and accurate translation depends on the structural integrity and operational competence of its multiple functional centers. Perturbing rRNA structure, its conformational flexibility, or its interaction with the ligands should have a profound effect on translation. Therefore, not surprisingly, most antibiotics that act on the ribosome interact directly with rRNA (see Refs. 5-7 for reviews). The sites of action of many drugs have been mapped to the main functional centers, and their mechanisms of action have been elucidated (8 -10). However, it is generally unknown which of the functional sites, besides the main ones, can be targeted by antibiotics.
In the small ribosomal subunit, antibiotics act upon a variety of sites where the drugs may affect binding of tRNAs, displace mRNA, or interfere with translocation (11)(12)(13)(14)(15)(16). In a striking contrast, the distribution of sites of antibiotic action in the large ribosomal subunit is highly constrained. Despite the enormous size of the large subunit, the multiplicity of activities in which it is involved, and the diversity of drugs that inhibit the large subunit functions, all clinically relevant antibiotics act on essentially one site: the peptidyltransferase center (PTC) 3 and the adjacent region of the nascent peptide exit tunnel (5,7,8,17,18). Only two other sites in the bacterial large ribosomal sub-unit, the GTPase-associated center and the translation factor binding region, are targeted by a few other known antibiotics (19 -23). Nonetheless, one would expect that binding of small molecules to other, yet undiscovered, sites could inhibit functions of the large ribosomal subunits (24).
Protein synthesis inhibitors act on critical centers of the ribosome, where structural perturbations have a detrimental effect on the ribosome functions. Similar perturbations can be caused by mutations. As a result, there is a clear correlation between the distribution of the known deleterious mutations in rRNA and the location of antibiotic-binding sites in the ribosome. We therefore reasoned that the clustering of deleterious mutations at a particular ribosomal locale may help in identifying new centers that can be targeted by antibiotics. With the general goal to identify new putative sites of antibiotic action and to map new functionally and structurally critical centers in the ribosome, we used random mutant rRNA libraries to isolate and characterize a variety of deleterious mutations in rRNA of the small and large ribosomal subunits. In our previous work, we reported a collection of deleterious mutations in 16 S rRNA that highlighted several sites as potential antibiotic targets (25). Furthermore, several regions of previously unrecognized functional importance were found where rRNA mutations had a strong deleterious effect on translation. A conceptually similar but complementary approach has been applied to 16 S rRNA by Cunningham and co-workers (26).
Here we describe the results of mutational studies of rRNA of the large ribosomal subunit. Clusters of deleterious mutations in 23 S rRNA pinpointed several regions in the large ribosomal subunit that are sensitive to structural alterations and thus may be targeted by new protein synthesis inhibitors.
Mutagenesis and Construction of Segment-Mutant Libraries-Mutagenesis and fragment exchange were carried out essentially as described previously (25). Briefly, the E. coli mutator strain XL-1 Red was transformed with the plasmid pLG857 (30) that carries a copy of the temperature-sensitive repressor gene cI857 and a kanamycin resistance marker. The plasmid pLK45 was introduced into pLG857-containing XL-1 Red cells and propagated for 1 day at 30°C (under conditions where pLK45-borne rRNA operon is not expressed). The mutated pLK45 plasmid was isolated and used as a template for PCR amplification. rDNA segments corresponding to the rrnB segments flanked by the restriction sites unique in pLK45 were PCR-amplified using low mutation frequency Triple Master enzyme mixture (Eppendorf). Primers GTGATAAGCAAT-TTTCGTGTCCCC and CACACGCCTAAGCGTGCTCC-CACT were used to amplify the XbaI-SnaBI segment (library C), CCGGCGAGGGGAGTGAAAAAGAAC and GTCAG-CATTCGCACTTCTGATACC primers were used for amplification of the SnaBI-SphI segment (library D), AAG-CCTGCGAAGGTGTGCTGTGAG and GAGACAGCCTG-GCCATCATTACGC primers were used to amplify the SphI-BsgI segment (library E), and ACGGTCCTAAGGTA-GCGAAATTCC and TGAATCCGTTAGCGAGGTGC-CGCC primers were used to generate the BsgI-BamHI segment (library F) (see Fig. 1). PCR products were cut with the corresponding restriction enzymes and cloned into unmutated pLK45 cut with the same enzymes. Segment-mutant libraries were transformed into POP2136 cells and plated onto LB agar/ ampicillin plates that were incubated at 30°C. The resulting segment-mutant libraries were enriched in clones carrying deleterious mutations by one round of negative selection as described previously (25).
Screening of the Segment-Mutant Libraries and Mapping of the rRNA Mutations-A robotic colony picker (Biopick BP600; BioRobotics) was used to inoculate 8,000 colonies from library C and 12,000 colonies from each of libraries D, E, and F into 90-l LB/ampicillin broth cultures in 384-well plates. After growing at 30°C for 48 h, each plate was replicated using a 384-pin replicator (Boekel) into two new plates, one with LB/ampicillin medium and one with LB/ampicillin supplemented with 15 g/ml erythromycin. The former (control) plate was incubated overnight at 30°C; the latter plate was incubated overnight at 42°C. Culture densities were read at 600 nm using the Spectra Max Plus 384 plate reader (Amersham Biosciences). Clones that exhibited poor growth at 42°C were rearranged into 96-well plates with fresh medium, and the phenotypes were retested. Plasmids were isolated from clones that reproducibly exhibited poor growth at 42°C. The mutagenized rrn segments were sequenced using the same primers that were used for PCR amplification. The severity of the deleterious phenotype was assessed by testing the ability of the mutant plasmids to transform E. coli JM109 cells at 37°C or POP2136 cells at 42°C.
Introduction of the Mutations into the pLK35 Plasmid-The QuikChange XL site-directed mutagenesis kit (Stratagene) was used to generate selected mutations in the pLK35 plasmid. Mutagenesis was carried out using the supplier's protocol (Stratagene) with the following PCR conditions: 95°C for 2 min (95°C for 1 min, 60°C for 1 min, 68°C for 25 min) ϫ 18 cycles. The generated PCR products were treated with DpnI restriction enzyme and transformed into POP2136 cells. Fragment exchange was used for introduction of the 34-nt deletion in helix 38 in the pLK35 plasmid. The XbaI-SphI segment was excised from the plasmid pLK1192U that carried a deletion of the 872-905 segment of 23 S rRNA (31), purified in 1% Sea-Plaque agarose gel (Cambrex), and cloned into pLK35 vector cut with the same restriction enzymes. The U860C mutation was then introduced into the resulting plasmid using the QuikChange XL site-directed mutagenesis kit.
Polysome Profile Analysis-Polysomes were analyzed as described previously (32,33) with minor modifications as described in Ref. 25. The ratio of mutant 23 S rRNA to wild-type 23 S rRNA in the gradient fractions was assessed with a primer extension inhibition assay (34) using the primer GTAAAGGT-TCACGGGGTC specific for the pLK45-resident mutation A2058G. Primer extension products were separated on a 12% polyacrylamide gel and quantified using a PhosphorImager (Amersham Biosciences).
Preparation of the Mutant 50 S Subunits-The cultures of POP2136 cells carrying the pLK35 plasmid or its derivatives were grown overnight at 30°C in LB/ampicillin and then diluted 1:600 to 1:2500 into 1 liter of fresh LB/ampicillin medium. After shaking for 2 h at 30°C, the temperature was raised to 42°C, and cultures were grown for about 4 h until A 600 reached 0.5. Cells were pelleted at 8,000 rpm in a JLA-10.5 rotor (Beckman) for 10 min at 4°C, and cell pellets were kept frozen at Ϫ70°C. Ribosomes were prepared following the published protocols (35) with minor modifications. Frozen pellets were thawed in an ice-water bath, washed with 30 ml of buffer A (20 mM Tris-HCl, pH 7.5, 100 mM NH 4 Cl, 10 mM MgCl 2 , 0.5 mM EDTA, 6 mM BME), pelleted by centrifugation in a JA-25.5 rotor for 15 min at 4°C, and then resuspended in 20 ml of cold buffer A. Cells were lysed by passage through a French press at 18,000 p.s.i. 10 l (1 unit/l) of RQ1 DNase (Promega) was added, and the lysates were clarified by centrifugation two times in the JA-25.5 rotor at 4°C (16,000 rpm for 15 min and then 24,000 rpm for 10 min). The supernatant was layered onto a 10-ml 1.1 M sucrose cushion in buffer B (20 mM Tris-HCl, pH 7.5, 500 mM NH 4 Cl, 10 mM MgCl 2 , and 0.5 mM EDTA) in a Beckman Optiseal polyallomer tube, and the tube was topped off with buffer B. Ribosomes were pelleted by centrifugation in a Beckman Ti-70 rotor at 33,000 rpm for 22 h at 4°C. Ribosome pellets were washed three times with 1 ml of buffer D (50 mM Tris-HCl, pH 7.5, 100 mM NH 4 Cl, 1 mM MgCl 2 , and 6 mM BME) and then resuspended in 300 l of the same buffer. Optical density (A 260 ) was determined, and samples were aliquoted, snap-frozen in a dry ice/ethanol bath, and stored at Ϫ70°C. For subunit preparation, 150 A 260 of the sample was layered on 36 ml of 10 -40% sucrose gradient prepared in buffer D, and the gradients were centrifuged in a Sorvall AH 629 rotor at 23,500 rpm for 18.5 h. Gradients were fractionated using a gradient fractionator (Brandel), and absorbance was monitored at 280 nm using an ISCO UA-6 absorbance detector. 1-ml fractions were collected into Eppendorf tubes. Fractions corresponding to 30, 42, and 50 S peaks were pooled, the MgCl 2 concentration was adjusted to 10 mM, and buffer C (50 mM Tris-HCl, pH 7.6, 100 mM NH 4 Cl, 10 mM MgCl 2 , and 6 mM BME) was added to bring the sample volume to 9 ml. Ribosomal subunits were pelleted by centrifugation in a Ti-50 rotor at 40,000 rpm for 16 h at 4°C. Subunit pellets were washed twice with 1 ml of buffer C and resuspended in 500 l of the same buffer. Optical density (A 260 ) of the resuspended subunits was determined, and the samples were aliquoted into Eppendorf tubes, snap-frozen, and stored at Ϫ70°C.

Selection of Deleterious Mutations in 23
S rRNA-Our strategy for identification of critical structural and functional sites in the large ribosomal subunit was based on mapping deleterious mutations in the 50 S subunit rRNA. The experimental approach involved preparing libraries of clones in which randomly mutated plasmid-borne rRNA genes would be conditionally expressed, enriching the libraries in clones with deleterious mutations, and mapping the positions of the mutations.
Random mutations were generated in the E. coli rrnB rRNA operon expressed under the control of the P L promoter in the pLK45 plasmid (28). The plasmids were propagated in the E. coli strain POP2136 that carries the temperature-sensitive repressor. In this strain, transcription of the plasmid-borne rRNA operon is activated at 42°C and is repressed at 30°C. The plasmid-borne rrnB carries spectinomycin resistance mutation C1192T in the 16 S rRNA gene and erythromycin resistance mutation A2058G in 23 S rRNA genes. The presence of these mutations facilitates monitoring the expression of the plasmidborne rRNA genes by primer extension analysis (34,37). According to this analysis, 3.5-4 h after shifting the temperature to 42°C, the plasmid-encoded 23 S rRNA accounts for 40 -60% of the total 23 S rRNA in POP2136 cells harboring pLK45 plasmids (38) (data not shown). As previously described in detail, the plasmid pLK45 was randomly mutagenized by propagating it in the mutator E. coli strain (25). Individual segments of the mutated rrnB were PCRamplified and cloned back into unmutated pLK45. In the resulting "segment-mutant libraries," only a specific segment of the plasmid-borne rRNA operon carried the mutations. This experimental setup facilitated subsequent mapping and analysis of the mutations. Two of the segment-mutant libraries that span the 16 S rRNA gene have been described previously (25). We now completed the rrnB mutagenesis by preparing four new segment-mutant libraries (C, D, E, and F) that covered the entire 23 and 5 S rRNA genes (Fig. 1). The mutated 781-bp segment in the library C bounded by the unique XbaI and SnaBI restriction sites included 259 bp of the 16 -23 S intergenic spacer and a 522-bp 5Ј-terminal portion of the rrlB gene comprising domain I of 23 S rRNA. Library D, flanked by the SnaBI and SphI restriction sites, carried mutations in a 709-bp segment of the 23 S gene representing domain II. The SphI-BsgI 735-bp mutated segment in library E comprised domains III and IV. Finally, library F contained a 1497-bp mutated BsgI-BamHI segment that encoded 937 nt of domains V and VI of 23 S rRNA, the 23 S/5 S spacer, and the entire 5 S rRNA gene and included a 348-bp region downstream of the rrnB operon.
After subjecting each library to one round of negative selection, which increased the representation of deleterious mutations (25), the libraries were plated on agar/ampicillin plates that were incubated at 30°C. Individual clones were then tested for their ability to grow in liquid culture at 42°C when mutant rRNA is expressed.
Eight thousand clones from library C and 12,000 clones from each of libraries D, E, and F were individually tested, and ϳ1,200 clones that grew poorly at 42°C were initially selected. The phenotypes of the clones were retested, and the mutated rDNA segments were sequenced in 360 clones using the primers flanking the corresponding cloning sites. The analysis of sequencing data showed that ϳ8% of the sequenced clones contained more than one mutation. Most of those clones were excluded from further analysis, since their investigation would necessitate the segregation of mutations. A number of the clones containing cloning defects, such as cloning site mutations, insertions or deletions of rDNA segments, or other plasmid rearrangements, were also excluded.
As a result of this analysis, a total of 77 individual deleterious mutations were identified in 23 S rRNA ( Fig. 2 and Table 1). Of these, 69 were single-base substitutions, and eight were singlenucleotide deletions occurring in runs of identical nucleotides. Three clones carried a mutation in the 23 S rRNA gene and an additional mutation in one of the transcribed spacers. One such clone from library C had a G380A mutation in 23 S rRNA and a mutation in the 16 S/23 S spacer (273 nt downstream from the 3Ј end of the 16 S rRNA gene, in the intergenic spacer separating the tRNA Glu and 23 S rRNA genes). Two clones from library F (A2453G and ⌬C 2556) each had an additional mutation, 356 and 321 nt downstream from the 3Ј end of the 5 S rRNA gene, respectively. Because no clones with deleterious mutations exclusively in the intergenic spacers were found, we assumed that poor growth of these three clones was due to the expression of the mutant rRNA and tentatively included them in our analysis. No deleterious mutations were found in the 5 S rRNA gene.
Ranking Deleterious Mutations in 23 S rRNA-The mutations in 23 S rRNA were ranked according to the severity of their deleterious phenotype. The classification was based on the ability of the mutant plasmids to transform two E. coli strains, POP2136 or JM109, in which the level of expression of mutant rRNA is substantially different. In the repressor-containing POP2136 cells, the fraction of plasmid-encoded rRNA accounts for 40 -60% of the total ribosomal RNA upon full induction at 42°C. The expression of the plasmid-borne operon in JM109 cells that lack the repressor is higher, and at 37°C, the level of mutant 23 S rRNA reaches up to 80% of the total cellular rRNA (25). A mutation was classified as strongly deleterious if the plasmid carrying the mutation failed to transform POP2136 cells at 42°C or JM109 cells at 37°C. Fourteen such mutations were identified (shown in red in Fig. 2). Twenty-three plasmids with moderately deleterious mutations (blue in Fig. 2) failed to produce JM109 transformants (at 37°C) but could transform POP2136 cells at 42°C. Finally, 40 mutations were classified as mildly deleterious, because the corresponding plasmids could transform both POP2136 and JM109 cells (green in Fig. 2). Importantly, however, expression of the plasmid-borne rRNA operon with even mildly deleterious mutations in 23 S rRNA notably inhibited growth of cells in liquid culture.
Testing Effects of Deleterious Mutations in the Absence of pLK45-resident Mutations-The plasmid pLK45 carries two latent mutations, C1192T in the 16 S rRNA gene and A2058G in the 23 S rRNA gene, that confer antibiotic resistance. Although these mutations are commonly assumed to be silent in terms of their effect on ribosome function and cell growth, on a few occasions, they were shown to enhance deleterious effects of other rRNA mutations (39). To test whether the resident pLK45 mutations had any significant effect on the extent of the deleterious phenotype conferred by mutation from our collection, six of the mutations, T328C, A804G, T860C, G864A, A1342G, and G1423A, representing mildly, moderately, and strongly deleterious mutations and belonging to different mutation clusters (see "Discussion") were engineered in plasmid pLK35. The pLK35 plasmid is a close relative of pLK45 but lacks the resident C1192T and A2058G mutations in 16 and 23 S rRNA genes (27). The mutations were engineered in the 23 S rRNA gene in pLK35 by site-directed mutagenesis, and the phenotypes of POP2136 cells transformed with the mutant pLK35 plasmids were tested in liquid culture, by transformation assay  and replica plating. The severity of the phenotypes conferred by 23 S rRNA mutations engineered in pLK35 was identical to those observed when these mutations were present in the pLK45 plasmids. Because these six mutations were scattered through the 23 S rRNA gene and represented mutations of different classes, this result led us to believe that the majority of the mutations in our collection inhibit protein synthesis irrespective of the presence of the resident mutations in plasmid pLK45.
Effect of Mutations on the Large Subunit Assembly and Function-An rRNA mutation can affect cell growth by interfering with either the assembly or the function of the ribosome. To distinguish between these two general scenarios, we analyzed polysome profiles in cells expressing several 23 S rRNA mutations from our collection. Three of the tested mutations had a strong deleterious phenotype; they were located on the interface side of the large ribosomal subunit in the sites with known (the PTC, A2451U) or suspected (helix 38, U860C and G864A) functional importance (Figs. 2 and 7). Three other tested mutants, one with a moderate phenotype (G1423A) and two with a mild phenotype (U328C and A1342G), had mutations located in functionally unremarkable regions at the solvent side (U328C) or near the "bottom" (G1423A and A1342G) of the subunit. POP2136 cells harboring the mutant plasmids were induced for 4 h at 42°C, and cell lysates were fractionated by sucrose gradient centrifugation (Fig. 3). The fraction of mutant 23 S rRNA relative to the total amount of 23 S rRNA in the major ribosomal peaks (individual subunits, 70 S ribosomes, and polysomes) was then determined by primer extension.
In "wild-type" cells that carried unmutated pLK45, the plasmid-encoded 23 S rRNA represented 35-45% of the total 23 S rRNA in the free 50 S subunits, 70 S ribosomes, and polysomes (Fig. 3). In contrast, in most of the tested mutants, the fraction of plasmid-encoded 23 S rRNA in active ribosomes (70 S and polysomes) was notably reduced compared with the free subunits. Thus, it appears that all of the tested mutant 50 S subunits participate in translation only reluctantly. Furthermore, the G1423A and U860C mutants accumulate more free subunits and show a decreased amount of polysomes relative to 70 S ribosomes, another strong indication of a severe functional defect.
In addition, all of the six tested mutants experienced problems with assembly of the mutant ribosomes. The G1423A, U860C, and A2451G mutants accumulated a high proportion of mutant 23 S rRNA in the 30 S peak, which indicates deficiency in the early assembly steps of the large ribosomal subunit (40,41). A late assembly step was apparently affected in the other three tested mutants, as could be judged by the decreased size of the 50 S peak (the U328C mutant) or its distorted shape (the G864A and A1342G mutants). In the G864A and A1342G mutants, a significant amount of the ribosomal material containing mutant 23 S rRNA sedimented between the 30 and 50 S peaks.
The U860C Mutation Dramatically Affects the Assembly of the Large Ribosomal Subunit-Mutations in helix 38 (H38) (U860C and G864A) formed the only cluster of severely deleterious mutations outside of the peptidyltransferase center (see "Discussion"). Therefore, we investigated in more detail the reason mutations at this site have such a profound effect on translation. In the ribosome, H38, also known as the A-site finger, protrudes toward the small ribosomal subunit and participates in the formation of the intersubunit bridge B1a (see Fig. 7) (42,43). Interestingly, however, deletion of up to 34 nt at the tip of H38 that disrupts the bridge results in only mild phenotypes (31,44,45). To investigate whether the deleterious effect of H38 point mutations we observed in our experiments is related to the function of this rRNA element as an intersubunit bridge, we engineered two new mutants. One construct carried a 34-nt deletion (⌬872-905) studied previously by Sergiev et al. (31) (we will refer to this deletion as ⌬B1a, because it eliminated the B1a bridge). The second construct combined the ⌬B1a deletion with the U860C mutation. Both constructs were prepared on the basis of the pLK35 plasmid and therefore lacked any additional rRNA mutations. In agreement with the previous observations, the ⌬B1a deletion alone had only a mild effect on cell growth when expression of mutant 23 S rRNA was induced at 42°C (Fig. 4). In contrast, the U860C mutation, even when it was combined with the ⌬B1a deletion, still had a profound dominant negative effect on cell growth. This result argued that the deleterious effect of the U860C mutation is unrelated to functions of H38 as an intersubunit bridge.
We next investigated how the U860C mutation (in the absence of the ⌬B1a deletion) affects the integrity and properties of the 50 S subunit. When the ribosomes prepared from the cells carrying the pLK35 plasmid with the U860C mutation (in the context of the otherwise intact H38) were dissociated into subunits and analyzed by sucrose gradient centrifugation in a buffer containing 1 mM MgCl 2 , an additional peak with an apparent sedimentation coefficient of 42 S was observed (Fig.  5). The 42 S peak was enriched in the mutant 23 S rRNA, confirming that the U860C mutation impairs the ribosome assembly. However, mutant rRNA was also present in the 50 S peak (20% of the 23 S rRNA in this peak). Even taking into account poor separation of 42 and 50 S peaks, this result indicates that a fraction of mutant 23 S rRNA can still assemble into compact 50 S subunits.
No obvious difference in the protein content between the 42 S particles and 50 S subunits was observed in an SDS-polyacrylamide gel (data not shown), although the latter result could not rule out some variation in protein composition. Analysis of the RNA composition of the 42 S peak showed the presence of approximately equimolar amounts of 23 S rRNA and 5 S rRNA. On the basis of its anomalous sedimentation behavior, we concluded that the 42 S peak represents improperly or incompletely assembled 50 S subunits. When the Mg 2ϩ concentration in the gradient buffer was increased from 1 mM (as in gradients shown in Fig. 5) to 10 mM, the sedimentation coefficient of the 42 S particles increased to 50 S (data not shown). This result explained why the 42 S peak was not observed in the polysome profiles of the U860C mutant (Fig. 3) and indicated that the compactness of the 42 S particles is strictly dependent on Mg 2ϩ concentration, compatible with the notion of their generally loose nature.
We then investigated how mutant 50 S subunits and the 42 S material associate with small ribosomal subunits. The mutant 50 S showed reduced propensity for association with 30 S subunits in the absence of mRNA and tRNA as compared with 50 S subunits prepared from wild-type cells; a substantial amount of 50 S subunits containing mutant 23 S rRNA failed to form 70 S ribosomes (Fig. 6D). However, the addition of mRNA and initiator tRNA enhanced association of mutant 50 S subunits into 70 S ribosomes (Fig. 6E). The fraction of mutant 23 S rRNA in the 70 S peak in the gradient shown in Fig. 6E was essentially indistinguishable from that in the 50 S subunits used in the association experiment (ϳ20%), indicating that in the presence of mRNA and tRNA mutant, mutant 50 S subunits are indeed capable of forming 70 S ribosomes. Furthermore, the mutant (U860C) ribosomes can participate in translation, as follows from the presence of mutant 23 S rRNA in the polysome fractions (Fig. 3). Importantly, although the sucrose gradient centrifugating revealed a subunit association defect related to the U860C mutation, the behavior of the mutant U860C 50 S subunits was essentially indistinguishable from that of the ⌬B1a subunits (in Fig. 6, compare D and F or E and G). Given the viability of the ⌬B1a mutant, the defect in subunit association is unlikely to be the major cause of lethality of the U860C mutation.  In contrast to the mutant 50 S subunits, the 42 S particles completely failed to associate with 30 S subunits in the absence or in the presence of mRNA and tRNA (Fig. 6, H and I), suggesting that these particles that accumulate in the mutant U860C cells are functionally inactive.

DISCUSSION
The aim of this project was to map functionally or structurally critical sites in the large ribosomal subunit that can be used as new antibiotic targets. The underlying principle was to detect sites in rRNA where nucleotide alterations would result in the loss of ribosome activity and would manifest themselves in a form of deleterious phenotype.
Most of the previous mutational studies of rRNA have been targeted to specific sites with well known or suspected functional importance (46). In addition, several genetic selection experiments were designed to identify rRNA sites involved in specific ribosomal functions (e.g. those involved in controlling the accuracy of translation) (47,48). The main difference of our approach compared with the previous studies is its intrinsic randomness. It was not built on any preconceived notion of which ribosomal site is functionally important, and it does not pursue any specific ribosomal site or any specific function. Instead, our technique addresses the general importance of an rRNA site for any aspect of ribosome function or biogenesis. The utility of such a strategy is demonstrated by the fact that we were able to identify deleterious rRNA mutations in several ribosomal sites whose functional importance was not revealed by other techniques.
A total of 77 mutations including 40 mildly deleterious, 23 moderately deleterious, and 14 strongly deleterious mutations were identified in 23 S rRNA. Despite the randomness of the mutagenesis selection scheme, the distribution of 77 deleterious mutations in 23 S rRNA was far from random. The bias in the distribution of the mutations is already evident in the 23 S rRNA primary and secondary structures and is even more striking when viewed in the context of the ribosome spatial structure (Fig. 2, A-C). Mildly deleterious mutations are clustered in the 5Ј portion and the middle of the molecule corresponding to domains I and III of the 23 S rRNA secondary structure. Moderately deleterious mutations are distributed more evenly along 23 S rRNA and form clusters in domains II, III, IV, and V. Most of the strongly deleterious mutations are located toward the 3Ј end of 23 S rRNA and clustered in domain V. The 5Ј-3Ј gradient of the increasing severity of deleterious phenotypes of the 23 S rRNA mutations probably reflects the general distribution of functional properties of rRNA. The 5Ј-proximal 23 S rRNA regions, being located toward the solvent side of the subunit, are involved in the overall organization of the subunit structure. Local effects of the mutations here are tolerated better than nucleotide alterations in the functionally critical peptidyltrans- ferase center composed of domain V segments, where mutations may have a profound effect on the binding or proper positioning of the ribosomal substrates. In the spatial structure of the 50 S subunit, all of the strongly deleterious mutations are located on the functionally charged interface side (Fig. 2C). Many of the moderately deleterious mutations are found within rRNA elements that form the outer shell of the PTC and buttress its structural organization. The identity of these nucleotides may allosterically affect the orientation of the key PTC nucleotides. Mildly deleterious mutations are localized toward the "lower" portion of the solvent side of the subunit and might represent regions involved in interactions with the secretion machinery and chaperones (49,50).
It should be emphasized that our mutagenesis scheme was not saturating, and our selection procedure was not exhaustive. As a result, we have mapped some, but not all, deleterious mutations. Consequently, we have "missed" a few of the known lethal mutations in the PTC, in the sarcin-ricin loop (H95), or in the GTPase-associated region (H42-H44) (51,52). Importantly, however, we did find other deleterious mutations in both the PTC and the sarcin-ricin loop ( Fig. 2B and Table 1). Noteworthy was the lack of deleterious mutations in 5 S rRNA, which is generally important for the ribosome function and assembly (53)(54)(55). In general, because of the presence of the erythromycin resistance mutation in 23 S rRNA of the pLK45 plasmid and selection of mutants in the presence of erythromycin, our technique would detect both dominant and recessive deleterious mutations in 23 S rRNA but only dominant mutations in 5 S rRNA. However, we are unaware of any severely deleterious point mutations in 5 S rRNA (51,56). It is possible that small changes in the 5 S rRNA structure do not perturb sufficiently its functions in the ribosome to be manifested in our experimental system. The lack of deleterious mutations in the spacer tRNA Glu gene was not surprising, given the redundancy of the gene in the E. coli chromosome.
Phylogenetic conservation is usually taken as an indicator of functional relevance of rRNA residues. In general agreement with this assumption, all of the 14 strongly deleterious mutations, most of the moderately deleterious mutations, and many mildly deleterious mutations coincide with the nucleotide residues that are more than 98% conserved in bacteria (Table 1) (57). It should be kept in mind, however, that a mere conservation of a nucleotide residue is not an absolute predictor of its functional role; mutations of some of the universally conserved nucleotides have no effects or have only minor effects on cell growth (58 -60). This notion underscores the importance of experimental validation of the functional significance of individual rRNA residues provided by our approach. Of the 77 rRNA positions affected by the deleterious mutations in bacterial 23 S rRNA, 30 residues are different between bacterial and human ribosomes (Table 1), which opens a possibility of using the corresponding rRNA sites as targets for selective antibiotics.
One of the main foci of this work was to identify 23 S rRNA sites that could be used as targets for new antibiotics. In our previous work that focused on 16 S rRNA mutations, we used spatial proximity of mutations in the context of the 30 S subunit to identify putative new sites of antibiotic action. Such sites, similar to the sites used by known antibiotics, are often composed of remote rRNA elements brought together only in the three-dimensional structure of the ribosome. However, from the standpoint of development of new antibiotics, much more attractive are the sites represented by well defined, "self-standing" rRNA elements that would have a better chance to assume the authentic structure even out of the whole-ribosome context. The utility of such an approach has been demonstrated by impressive progress in understanding the mechanism of binding and action of aminoglycosides, which interact with a defined segment of helix 44 of 16 S rRNA with a distinctive structure. The ability to chemically synthesize the aminoglycoside binding site in isolation and use it in structural and functional studies paved the way for development of new drug prototypes that act on this site (61)(62)(63)(64)(65)(66). With these considerations in mind, we identified five elements of 23 S rRNA (sites I-V; Fig.  7) that we view as promising antibiotic targets. These sites are not targeted by known ribosomal antibiotics and are characterized by clustering of deleterious mutations within defined rRNA elements of irregular structure.
Site I-Only four strongly deleterious mutations in our collection were located outside of the PTC, the site of action of most antibiotics that target the large ribosomal subunit. Two of these mutations, a deletion of guanine residues in H91 (positions 2523-2526) and the G2664A substitution in the sarcinricin loop (H95), are within the sites of action of other known protein synthesis inhibitors: oligosaccharide antibiotics evernimicin and avilamycin (H91) (21)(22)(23) or ribosome-inactivating proteins (H95) (67)(68)(69). The only other two strongly deleterious mutations outside of the PTC, U860C and G864A, are located in H38 (Fig. 7A), which forms the intersubunit bridge B1a and is dubbed the "A site finger" because it contacts the A site tRNA (42,43). The 858 -868/909 -919 segment of the helix shows a much higher degree of conservation than the rest of H38 (57). Remarkably, two conserved adenine residues, A910 and A911, are extruded from the helix (Fig. 7). Their flippedout conformation resembles the orientation of A1492 and A1493 in the 16 S rRNA decoding center in the presence of cognate tRNA in the A-site. The bases A910 and A911 form class II and class I A-minor contacts, respectively, with H81 in domain V, which in turn connects with H80 and fixes the orientation of the P-site tRNA. Thus, the conformation of A910 and A911, which is directly linked to the orientation of U860, may be part of a communication line between the PTC and other ribosomal centers. This segment of H38 also makes a close contact with 5 S rRNA (70). Although functions of 5 S rRNA in the ribosome are unclear, it is important for the assembly of the large ribosomal subunit (55,71,72). Thus, another possible role of the central segment of H38 could be that of one of the pivotal assembly centers of the large ribosomal subunit.
The latter possibility agrees well with the results of our experimental investigation of functional effects of the H38 mutations. Although the U860C and G864A mutations are lethal, the disruption of the A site finger by deletion of the tip of H38 (the ⌬B1a mutation) has only a mild phenotype (31,44,45). Furthermore, the U860C mutation was severely deleterious even in the context of the ⌬B1a deletion (Fig. 4), showing that the lethal effect of this mutation is unrelated to H38 interaction FIGURE 7. Putative new antibiotic target sites in the large ribosomal subunit. The sites were identified on the bases of the constellation of deleterious mutations, confinement to a defined RNA structural element, and distinct structural features of the RNA element. The sites of mutations are colored according to the severity of the phenotype of the mutation (as in Fig. 2). The general location of each site in the 50 S subunit is shown on the schemes representing the spatial structure of 23 and 5 S rRNA in the context of the E. coli large ribosomal subunit (74). To facilitate orientation, 5 S rRNA is colored in cyan, the GTPase-associated center is shown in magenta, and the protein L1-binding 23 S rRNA site (H78) is shown in orange. In site IV, the H96-H95 stem-loops are colored green. Note the 180°C degree rotation of the 50 S subunit in the site V panel compared with the panels representing the other sites.
with the A site-bound tRNA or the function of H38 as an intersubunit bridge. Although the 50 S subunits prepared from the U860C mutant cells were somewhat impaired in association with the 30 S subunits in the absence of mRNA and tRNA, their association properties closely matched those of the 50 S subunits that carried the rather benign ⌬B1a deletion; thus, this defect is an unlikely cause of the lethality of the U860C (and probably G864A) mutation. However, the observed functional imperfection of the mutant 50 S subunits might contribute to the overall deleterious phenotype of these mutations and to the altered polysome profile in the mutant cells (Fig. 3).
The most dramatic difference in the behavior of the ribosomal material from the U860C mutant cells compared with the wild-type cells or with the ⌬B1a deletion mutant was the appearance of a prominent 42 S peak in the sucrose gradient run at a low (1 mM) concentration of Mg 2ϩ ions. At higher Mg 2ϩ concentration, the sedimentation coefficient of this material increased to 50 S, and the RNA and protein composition of this peak revealed it as incompletely or improperly assembled large ribosomal subunits. These particles failed to associate with 30 S subunits even in the presence of tRNA and mRNA and thus are likely to be nonfunctional in protein synthesis. This result suggests that the possible cause of the deleterious effect of the U860C mutation (and the G864A mutation) could be the major defect in the ribosome assembly. We are puzzled, however, that such a defect would be revealed as a dominant phenotype, because a sufficient amount of functional 50 S subunits capable of protein synthesis appears to be present in the mutant cells as shown by polysome profiling (Fig. 3). One possibility is that accumulation of assembly intermediates depletes the cellular pool of the ribosome assembly chaperons, so that even 50 S subunits that carry wild-type 23 S rRNA do not assemble properly and are impaired in some aspects of their function. Therefore, antibiotics targeted against 23 S rRNA site I are expected to inhibit cell proliferation by interfering with the ribosome assembly (73).
Site II-Four moderately deleterious mutations, G674A, A804G, C806U, and U807C, are clustered in H32, which buttresses the PTC wall below the central protuberance of the 50 S subunit. The pyrimidine-pyrimidine pair U807-C673 is stacked on the G674-C806 base pair, and unpaired A804 is held inside the helix due to its cross-strand stacking interaction with G674. The flipped out G805 base may open sufficient space for a hypothetical small organic molecule to squeeze into the internal loop of H32 and perturb its structure. Clustering of several deleterious mutations in this site clearly indicates that its structure plays an important role either in the ribosome function or in assembly of the subunit. This conclusion is further supported by the fact that the nucleotides forming the internal loop of H32, including the sites of mutations, show a notably higher degree of evolutionary conservation compared with the surrounding nucleotides.
Site III-This site is defined by a constellation of five moderately and mildly deleterious mutations, four single-nucleotide substitutions (G1421A, G1423A, A1569G, and A1572C), and one single-nucleotide deletion (⌬G1421-1426) in H55 in domain III (the deletion was arbitrarily assigned to the position 1425; Fig. 4C). This site is located in the lower section of the 50 S subunit at the L1 side. The importance of this region of 23 S rRNA for the ribosome function is unclear. However, high conservation of nucleotide residues in the internal loop connecting H55 and H56 (positions 1426 -1428/1568 -1571) supports the notion that this region may play a prominent functional or structural role. Clustering of several deleterious mutations within H55 and the adjacent internal loop suggests that even a small alteration in the structure of this rRNA element brought about by binding a putative antibiotic molecule would have a detrimental effect on cellular protein synthesis. Sucrose gradient analysis showed a dramatic accumulation of the mutant assembly precursors or dead end assembly debris in the 30 S peak (Fig. 3), indicating that the RNA structure within site III is critical for the early assembly steps of the large ribosomal subunit. In addition, the elevated ratio of free subunits to 70 S ribosomes suggests that even fully assembled 50 S subunits are defective in their ability to associate with the 30 S subunits and participate in translation. The characteristic S-twist of one of the RNA strands within the internal loop and flipping out of the A1569 residue opens space where small organic molecules with possible antibiotic functions could potentially fit.
Site IV-Two moderately deleterious mutations and one mildly deleterious mutation affect the structure of H61 in domain IV of 23 S rRNA (Fig. 4D). This RNA element, located at the interface side of the 50 S subunit, establishes a close contact with the helices H96-H95 (shown in green in Fig. 4D), which present the sarcin-ricin loop (the H95 loop) for the binding of translation elongation factors. The distortion of the H61 shape at the site of juxtaposed A1664 and C1996 leads to a partial extrusion of A1664, allowing it to form a stacking interaction with A2726 of H96 (shown in light blue in Fig. 4D). It is conceivable that mutations in H61 could affect the interaction between H61 and the H96-H95 stack, resulting in the displacement of the sarcin-ricin loop and the associated defects in binding elongation factors.
Site V-A constellation of five mildly deleterious mutations in H19 and H20 mark a putative antibiotic site at the solvent side of the 50 S subunit (Fig. 4E). This RNA element has a distinctive structure, with H19 and H20 loops coming into close proximity to each other due to multiple interloop interactions. Two of the residues whose alterations interfere with protein synthesis, A310 and A330, are stacked on each other and help to hold the H19 and H20 loops in place. U328 in H20 forms a reverse Hoogsteen base pair with A332, which helps to maintain the characteristic shape of the H20 loop. The other two positions, C335 and A324, are part of H20 stem; their mutations are expected to affect the overall structure of H20. The H19-H20 RNA element is involved in interaction with ribosomal proteins L4 and L24, both of which are early assembly proteins (74,75). Nevertheless, although the relative amount of the 50 S peak was somewhat decreased, the sucrose gradient profile did not reveal any dramatic effects in the assembly of the mutant 50 S subunits (Fig. 3). On the other hand, the representation of mutant subunits in the polysome fractions of the gradient was notably decreased compared with the extract from cells bearing the unmutated pLK45 plasmid, indicating that mutations in site V affect function properties of the large ribosomal subunit. Protein L4, which contacts the H19-H20 element, extends into the nascent peptide exit tunnel and was proposed, along with other elements, to communicate the information from the tunnel to the surface of the ribosome in the vicinity of the exit tunnel opening (76 -78). Conceivably, the functional role of site V might be mediated by its influence on the conformation of L4.
The main objective of this study was identification of new antibiotic sites in the ribosome. It is difficult, however, to ignore the fact that functional involvement of most of the sites that we view as possible antibiotic targets is poorly understood. Investigation of the role of each of the identified sites in the ribosome assembly or function is well beyond the scope of this work. Nevertheless, we believe that future studies of these sites, some of which are under way in this laboratory, may bring a better understanding of the function of ribosomal RNA in translation and its role in ribosome assembly and may lead to development of new potent antibiotics.