Site of Action of a Ribosomal RNA Methylase Responsible for Resistance to Erythromycin and Other Antibiotics*

The enzyme which confers resistance to erythromycin in the producing organism Streptomyces erythraeus dimethylates a single adenine residue in Bacillus stearothermophilus 23 S rRNA. This corresponds to residue Ade 2058 in Escherichia coli 23 S RNA. The methylase responsible for resistance to macrolides, lincomycin, and streptogramin B-related antibiotics in Staphylococcus aureus also acts at this site.

11 Supported by Grant GM 26756 from the National Institutes of Health.

EXPERIMENTAL PROCEDURES
Bacterial Strains-The organisms used in this work were Streptomyces coelicolor A3 (2), S. erythraeus NRRL 2338, Bacillus stearothermophilus NCIB 8924, Escherichia coli MRE 600, and B. subtilis 430 containing staphylococcal plasmid pE194. The latter organism was kindly supplied by Dr. David Dubnau (Department of Microbiology, The Public Health Research Institute of the City of New York, Inc.). Plasmid pE194 specifies resistance to MLS antibiotics; this phenotype is inducible by erythromycin.
Preparation of Ribosomal Subunits and rRNA-Ribosomal subunits and total RNA from 70 S ribosomes were prepared as described elsewhere (4); 23 S rRNA was prepared as described below for [a2P] RNA. Preparation of Uniformly Labeled p"P/rRNA-This was carried out in 20 ml of buffered salt solution supplemented with 0.8 ml each of 10% (w/v) Difco casamino acids and 10% (w/v) Difco Bactopeptone plus 0.4 ml of 20% (w/v) glucose, 0.2 ml of 0.1 M MgSO,, and 0.2 ml of 0.13% (w/v) K,HPO,. Buffered salt solution contained, (per liter) 1.5 g of KCI, 5 g of NaCl, 1 g of NH,CI, 13.2 g of Tris-HCI, and 1.9 g of Tris base. The final pH was 7.4. Components of the medium were sterilized separately, and solutions of casamino acids and peptone were freed of phosphate prior to autoclaving (8). The medium was inoculated with approximately 2 X 10' cells of B. stearothermophilus in the exponential phase of growth, 10 mCi of i'"P] phosphate was added, and the culture was shaken at 60 "C. Cells were harvested after about six generations of growth (AGoo "~ = 0.3) and were washed by recentrifugation in buffer containing 10 mM Tris-HCI (pH 7.6 a t 20 "C), 10 mM MgC12,50 mM NH,CI, 0.5 mM EDTA, 3 mM 2-mercaptoethanol. They were then resuspended in 2 ml of buffer to which were added lysozyme, pancreatic DNase, and Triton X-I00 a t final concentrations of 2.5 mg/ml, 5 pg/ml, and 0.05% (w/v), respectively. After incubation a t 37 "C for 2 min, the lysate was centrifuged a t 200,000 X g for 90 min. The pellet was resuspended in 1 ml of buffer containing 10 mM Tris-HC1 (pH 7.6 at 20 "C), 10 mM Na,EDTA, 50 mM LiCI, 0.2% (w/v) sodium dodecyl sulfate and layered onto a 35-ml 10-30% (w/v) sucrose density gradient made in similar buffer lacking sodium dodecyl sulfate. Centrifugation was at 25,000 rpm for 20 h a t 15 "C in a Beckman SW 27 rotor. Fractions containing 23 S RNA were pooled, and the RNA was precipitated with 2.5 volumes of ethanol at -20 "C, reprecipitated twice from 0.5 M sodium acetate, and stored frozen at -20 "C. Typically, 1 mCi of 23 S [3'P] RNA was obtained at a specific activity of about 8 pCi/ pmol. tlnlabeled 23 S RNA was prepared similarly from cells grown in Tr-yptic Soy Broth (Difco). Methylation of 23 S rRNA in Vitro-Incubation mixtures (300 pl, pH 7.5) contained 50 mM HEPES-KOH, 10 mM Tris-HC1, 5 mM MgCI2, 200 mM NH,CI, 3% (w/v) glycerol, 250 pmol of 23 S RNA, 60 pCi of S-adenosyl-~-[methyl-~H]methionine (15 Ci/mmol), and 60 units of erythromycin-resistance methylase from S. erythraeus. This partially purified enzyme was prepared as previously described (4). Incubation was a t 37 " C for 30 min, after which the methylated RNA was extracted twice wit,h phenoI and three times with ether and precipitated with ethanol. Stoichiometries of methylation were determined as described elsewhere (4).
Partial Digestion of Methylated 23 S rRNA-Methylated RNA (250 pmol) was dissolved in 60 gl of buffer containing 30 mM Tris-HCI (pH 7.6 at 20 "C) 320 mM KC[; MgCIz (10 mM final concentration) was then added to give TMK buffer. After 15 min at 37 "C, the RNA was chilled to 0 "C, and 125 units of RNase TI (Sankyo) was added.
After 30 min a t 0 "C, 10 volumes of TMK buffer was added, followed by 0.1 volume of 2% (w/v) bentonite (Serval. The digest was then extracted three times with TMK-saturated phenol (with back extraction of the phenol layers) and precipitated with ethanol a t -20 "C prior to fractionation by gel electrophoresis.
Preparative Gel Electrophoresis-RNA fragments were resolved by two-dimensional polyacrylamide gel electrophoresis as described by Pedersen and Haseltine (9) except that the thickness of our gels was 1.5 mm in the first dimension and 1.0 mm in the second. Onedimensional electrophoresis in 8 M urea, 20% (w/v) acrylamide gels was used both preparatively and analytically (see Fig. 1). Such gels were 0.5 mm thick and 35 cm long. They contained (per 50 ml) 10 g of acrylamide, 0.4 g of bisacrylamide, 24 g of urea, 5 ml of buffer containing 0.75 M Tris, 0.68 M boric acid, and 10 mM Na2EDTA, and 0.3 ml of 4% (w/v) ammonium persulfate. Gels were polymerized by addition of 100 &I of 3-dimethylaminopropionitrile pre-electrophoresed for a t least 30 min and run a t 2.5 kV.
Nucleotide Sequence Analysis-Fingerprinting of uniformly labeled ["'PIRNA and determination of base compositions of oligonucleotides were performed as previously described (10). Gel sequencing of 5' end-labeled RNA was carried out according to Donis-Keller et al. (11) using reagents supplied by Bethesda Research Laboratories, except for RNase TI which was obtained from Sankyo. Sequencing gels were prepared as above except that the acrylamide concentration was 25% (w/v).

When the erythromycin-resistance methylase from
S.
erythraeus acts in vitro upon rRNA from various species of Bacillus or Streptomyces, the stoichiometry of methylation is usually within the range of 1.5-2.0 methyl groups incorporated per 23 S RNA molecule. Typical data are given in Table I. Since 23 S rRNA from Gram-positive bacteria in general (and from B. stearothermophilus in particular) does not contain monomethyladenine (6), we have argued previously (4) that such stoichiometries represent the conversion of a single adenine residue to NF,N"-dimethyladenine. Ideally, we wished to study the action of the enzyme upon E. coli 23 S rRNA, but, as shown in Table I, the latter was not such a good substrate for methylation. We have no ready explanation for this observation. However, since most of the nucleotide sequence of B. stearothermophilus 23 S RNA is known (7), we decided to determine the site of action of the S. erythraeus methylase using that RNA as substrate. Before doing so, we wished to ascertain whether the S. er~ythraeus methylase and the staphylococcal MLS-resistance methylase act at a common site. To do this, we prepared total ribosomal RNA from B. subtilis carrying the staphylococcal plasmid pE194 before and after induction of the MLS-resistance phenotype with erythromycin (20 pg/ml). As shown in Table I, rRNA from the uninduced culture was an excellent substrate for the S. erythraeus methylase, whereas that from induced cells was ' This assay contained 5 times the usual amount of enzyme. not. Since the Streptomyces enzyme acts at a single site within 23 S RNA, the simplest explanation for these data is that it and the staphylococcal methylase act at the same site. However, they do not exclude the possibility that the latter enzyme might also act at additional sites, although we have no reason to suspect that it does.
Preparation of a Methylated Fragment of 23 S RNA-In order to localize the site of action of the S. erythraeus methylase, it was first necessary to obtain a fragment of 23 S RNA containing the methylated residue and long enough to possess a unique sequence. Accordingly, 23 S RNA from B. stearothermophilus was radioactively methylated in vitro by incubation with the S. erythraeus enzyme together with S-adenosyl-L-[methyl-"Hlmethionine as cofactor. This material was then subjected to partial digestion by TI ribonuclease followed by electrophoresis in polyacrylamide-urea gels. Only one methylated oligonucleotide was evident under our chosen conditions (Fig.1, track 2). This was estimated to be about 20-25 residues long by comparison with the mobilities of tracker dyes. When this procedure was repeated using radioactively methylated RNA uniformly labeled with :32P, no single band in the TI digest obviously corresponded to the methylated fragment ( Fig. 1, track I ) . Therefore, this digest was subjected to two-dimensional gel electrophoresis, and the approximate position of the methylated fragment was predicted from its known mobility in the individual dimensions of the twodimensional system. Oligonucleotides were eluted from this portion of the two-dimensional gel and subjected to double label counting. A fraction greatly enriched in 'H radioactivity was thus identified and was further resolved in a polyacrylamide-urea gel. Autoradiography revealed a major 32P-labeled band containing all the "H radioactivity, plus several minor contaminants of closely similar mobility (results not shown). The major component was eluted and rerun on a similar gel. It appeared homogeneous and co-migrated with the original "H-labeled RNA frgment (Fig. 1, tracks 2 and 3). This purified oligonucleotide was then subjected to "fingerprint" analysis.
Fingerprint Analysis of the Methylated Oligonucleotide-The purified, methylated oligonucleotide was digested to completion with TI ribonuclease and subjected to two-dimensional electrophoresis, as previously described (10). The resultant fingerprint revealed six "'P-labeled oligonucleotides (Fig. 2). These were eluted and their base compositions determined following digestion with RNase TS. Double label counting of the six oligonucleotides showed that only one of them (AAAG) was methylated and that there was only 1 residue of dimethyladenine/tetranucleotide. This finding is consistent with the earlier observation that the staphylococcal MLS-resistance methylase also acts within the oligonucleotide sequence AAAG (12).
To establish which Ade residue had been methylated by the S. erythraeus enzyme, radioactively methylated 23 S RNA was first digested to completion with RNase TI to produce the methylated tetranucleotide ApApApGp. This was further digested with endonuclease PI, which liberates nucleoside 5'phosphates. During paper chromatography, all the [methyl-3H] radioactivity eo-migrated with N',N'-dimethyladenosine (RF of approximately 0.65) under conditions where nucleotides did not move (results not shown). In control experiments, when either nuclease was omitted, all the radioactivity remained at the origin. We therefore concluded that the 5'terminal residue of oligonucleotide AAAG had been methylated.
Gel Sequencing of the Methylated RNA Fragment-The methyl-"-fragment of 23 S RNA was prepared as described above using a small amount of the purified "P-labeled mate- " X C rial as tracer. This was added immediately prior to twodimensional gel electrophoresis. Material eluted from such a gel was subjected to 5' end labeling with [32P]phosphate and was fractionated on a polyacrylamide-urea gel. The major component was eluted and its purity checked by gel electrophoresis (Fig. 1, track 4). Since this pure oligonucleotide possessed an extra terminal phosphate group, it migrated slightly ahead of those visualized in Fig. 1, tracks 2 and 3. The nucleotide sequence was established on sequencing gels following partial enzymic digestion (data not shown)2 and is given in Fig. 3a. The data obtained by fingerprint analysis can be fitted perfectly to this sequence provided that residue 19 (from the 5' end), which was not identified on sequencing gels, is assumed to be Ap. This is a necessary assumption since the oligonucleotide AAAG cannot otherwise be accommodated. Presumably, this position in the sequence was not cleaved by RNase Uz or Phy M because the Ade residue in question was dimethylated.
The sequence determined here for the methylated fragment of B. stearothermophilus 23 S RNA (Fig. 3a) is obviously homologous with that of E. coli 23 S between residues 2040 and 2061 (Fig. 36). Part of the sequence of the corresponding region of B. stearothermophilus 23 S RNA has been published (7), and it is identical with our sequence. We therefore conclude that the adenine residue converted to N6,N6-dimethyladenine by the erythromycin-resistance methylase of S. erythraeus corresponds to residue Ade-2058 in E. coli 23 S rRNA.
' Sequence data were provided to the referee. Presumably, when the S. erythraeus methylase acts upon E. coli 23 S RNA (Table I), residue Ade-2058 is modified, although we have not established this directly.

DISCUSSION
Obvious similarities exist between the action and effects of the erythromycin-resistance methylase of S. erythraeus and those of the staphylococcal MLS-resistance methylase. However, there are also some apparent differences. The Streptomyces enzyme acts in vitro at a single site in 23 S rRNA, whereas results obtained in vivo suggested that the staphylococcal enzyme might act at more than one site (12). Some confusion also exists regarding substrate specificity. The Streptomyces enzyme, acting in vitro, methylates free 23 S rRNA quantitatively but is totally inactive on 50 S ribosomal subunits (Table I). Similarly,Weisblum (1) concluded that the staphylococcal enzyme probably does not methylate mature ribosomal subunits in uiuo. In contrast, Shivakumar and Dubnau (13) claimed that the staphylococcal methylase acts in vitro both on 23 S RNA and on 50 S subunits, although not on 70 S ribosomes. However, we consider that the rather low levels of methylation reported by the latter authors do not unequivocally establish the substrate specificity of the staphylococcal methylase. Considerable interest has been aroused by the translational attenuation model for inducibility of the MLS-resistance phenotype in Staphylococcus (14,15). It is therefore relevant to ask whether the Streptomyces methylase gene is similarly controlled. This question cannot, at present, be answered definitively, although resistance in S. erythraeus is not ob-viously inducible. Now that the methylase gene from S. erythraeus has been cloned in S. lividans (5), it may be possible to resolve this matter.
Co-resistance to the MLS antibiotics can be rationalized on the basis that such drugs all bind to closely related ribosomal sites. For example, erythromycin and lincomycin compete with each other and with chloramphenicol for binding to the 50 S ribosomal subunit (16). It is therefore not surprising that methylation of a single site in 23 S rRNA can lead to reduced affinity for the various MLS antibiotics although, interestingly, the binding of chloramphenicol is not obviously affected.
Given the extensive sequence homology between rRNA molecules from prokaryotes and mitochondria, it has been possible to identify in E. coli 23 S rRNA sites corresponding to those at which single base substitutions in mitochondrial rRNA result in antibiotic resistance. These sites are positions 2447, 2451, 2503, and 2504 in the case of chloramphenicol resistance (17,18) and, significantly in the present context, position 2058 for erythromycin resistance (19). As shown in Fig. 3, current models for the secondary structure of 23 S RNA place these 5 residues close together (7, 20, 21). Again, these data are compatible with a model whereby chloramphenicol and the MLS antibiotics bind to a common ribosomal domain which presumably includes the region of 23 S RNA represented in Fig. 3. Additionally, that domain should also include protein L16, which has been implicated in the ribosomal binding of chloramphenicol, erythromycin, and virginiamycin S (an MLS antibiotic), and protein L15 which binds erythromycin, albeit weakly, in free solution (22)(23)(24).