NusG-Dependent RNA Polymerase Pausing and Tylosin-Dependent Ribosome Stalling Are Required for Tylosin Resistance by Inducing 23S rRNA Methylation in Bacillus subtilis

Antibiotic resistance is a growing health concern. Resistance mechanisms have evolved that provide bacteria with a growth advantage in their natural habitat such as the soil. We determined that B. subtilis, a Gram-positive soil organism, has a mechanism of resistance to tylosin, a macrolide antibiotic commonly used in the meat industry. Tylosin induces expression of yxjB, which encodes an enzyme that methylates 23S rRNA. YxjB-dependent methylation of 23S rRNA confers tylosin resistance. NusG-dependent RNA polymerase pausing and tylosin-dependent ribosome stalling induce yxjB expression, and hence tylosin resistance, by preventing transcription termination upstream of the yxjB coding sequence and by preventing repression of yxjB translation.

RNA polymerase (RNAP) transcribes into the downstream genes. These transcription attenuation mechanisms typically involve overlapping antiterminator and terminator structures that can form in the nascent transcript (1). RNAP pausing can participate in transcription attenuation by providing sufficient time for RNA structure formation and/or regulatory factor binding (4)(5)(6). Regulation of translation initiation is another strategy used by bacteria to control gene expression (2,7). RNA structures that sequester the Shine-Dalgarno (SD) sequence prevent ribosome binding, leading to translational repression.
During translation, polypeptides travel through the peptide exit tunnel (PET) within the large ribosomal subunit (7)(8)(9). Macrolide antibiotics bind exclusively to 23S rRNA within the PET (10). Once bound, these antibiotics can cause stalling of the translating ribosome when an appropriately positioned macrolide arrest motif is encountered in the nascent peptide. The known arrest motifs are typically three amino acids in length, with the downstream residue corresponding to the ribosome A site (11). Only a subset of any of the arrest motifs leads to translation arrest (11), suggesting that additional amino acids within the nascent polypeptide contribute to ribosome stalling (11,12).
Resistance to a macrolide antibiotic can occur via methylation of its 23S rRNA target in the PET. For example, Erm methylates the N6 position of A2058, leading to resistance to several macrolides (13,14). Gram-negative bacteria contain RlmA I , which methylates the N1 position of G745, whereas Gram-positive organisms contain RlmA II , which methylates the N1 position of G748. Tylosin is a macrolide antibiotic produced by Streptomyces fradiae. Synergistic high-level tylosin resistance in S. fradiae is conferred by methylation of A2058 by TlrD and by methylation of G748 by RlmA II (15). Methylation of G748 is specific to tylosin resistance because the mycinose sugar moiety contacts this residue (14,15).
We determined that yxjB encodes RlmA II in B. subtilis, that YxjB-specific methylation of G748 confers tylosin resistance, and that tylosin induces yxjB expression. The induction mechanism requires NusG-dependent RNAP pausing in the yxjB leader as well as tylosin-dependent ribosome stalling during translation of a leader peptide, while translation of the leader peptide is required to reduce transcription termination in the yxjB leader region and to overcome a translation attenuation mechanism that represses YxjB synthesis.

RESULTS
yxjB encodes RlmA II in B. subtilis. RlmA I from Escherichia coli methylates G745 of 23S rRNA, while RlmA II from Gram-positive bacteria methylates G748 within the same helix (14,16). B. subtilis YxjB is 30% identical to RlmA II from S. fradiae (16). To determine whether YxjB methylates G748 in B. subtilis, primer extension inhibition experiments were performed on RNA isolated from wild-type (WT), ΔyxjB, and yxjB overexpression strains. In this assay, reverse transcriptase (RT) terminates at the nucleotide preceding a residue containing a methyl group on the Watson-Crick (WC) face of the nucleobase. Similar experiments were carried out on RNA isolated from E. coli as a control. RT stops were observed at positions 749 and 746 on B. subtilis and E. coli 23S rRNA, respectively, indicating that 23S rRNA was methylated at position G748 in B. subtilis and position G745 in E. coli ( Fig. 1A and B). Deletion of B. subtilis yxjB eliminated methylation at G748, while overexpression of yxjB increased methylation (Fig. 1B). On the basis of previous studies of other Gram-positive organisms, we conclude that yxjB encodes RlmA II in B. subtilis.
In addition to G748, an RT stop corresponding to U747 was observed for B. subtilis (Fig. 1B). Deletion and overexpression of yxjB eliminated and increased the RT stops at the two positions, respectively. While we infer that the RT stops were caused by modification on the WC face of both nucleobases, we did not investigate the molecular basis for the RT stop at U747.
YxjB-mediated methylation of G748 in 23S rRNA confers resistance to tylosin. S. fradiae produces the macrolide antibiotic tylosin. Methylation of G748 by RlmA II in S. fradiae confers low-level resistance to tylosin, whereas high-level resistance also re-quires TlrD-mediated methylation of A2058 in 23S rRNA (15). Both of these residues line the PET and, when methylated, inhibit tylosin interaction (14)(15)(16). Although B. subtilis does not produce tylosin, since both S. fradiae and B. subtilis are soil organisms, it was conceivable that B. subtilis had evolved resistance to tylosin produced by S. fradiae. We found that WT B. subtilis grew well with tylosin concentrations at or below 0.5 g/ml and stopped growing above 1 g/ml. To determine whether overexpression of yxjB would lead to increased resistance, strains containing the yxjB overexpression plasmid or an empty vector were grown with or without tylosin. While both strains grew in the absence of tylosin, induction of yxjB expression was required for growth in the presence of 4 g/ml tylosin (Fig. 1C).
As yxjB overexpression resulted in increased methylation of G748 in 23S rRNA and resistance to tylosin, we reasoned that growth in the presence of subinhibitory tylosin concentrations could lead to increased methylation of this residue. Hence, primer extension was performed on 23S rRNA extracted from WT cells grown in the absence and presence of increasing tylosin concentrations. Methylation of G748 was greatly Primer extension inhibition was used to detect methylated residues in helix 35 of 23S rRNA in WT and ΔyxjB strains, as well as a strain in which yxjB was overexpressed from an IPTG-inducible promoter (pYxjB). Methylation of 23S rRNA from E. coli is shown as a control. (C) Overexpression of plasmid-borne yxjB from an IPTG-inducible promoter confers tylosin resistance. pVector was used as a control. (D) Primer extension inhibition was used to detect the position of methylated residues in 23S rRNA when cultures were grown in the presence of the indicated tylosin concentration. Methylation of 23S rRNA from E. coli was used as a control. (E) Growth in the presence of subinhibitory tylosin concentrations leads to increased tylosin resistance. Cultures were grown with the tylosin concentrations indicated at the left, and then serial dilutions were spotted onto plates containing the tylosin concentration shown below each plate. Experiments were performed at least twice with comparable results.
Mechanism of Tylosin Resistance in Bacillus subtilis ® increased in the presence of the lowest concentration of tylosin tested, with methylation gradually increasing as the tylosin concentration was increased further (Fig. 1D).
We next tested whether growth in the presence of subinhibitory tylosin concentrations increased resistance to the antibiotic. The WT strain was grown in the absence or presence of 0.125, 0.25, or 0.5 g/ml tylosin. Aliquots of 10-fold serial dilutions were then spotted onto plates containing 0, 1, or 2 g/ml tylosin. The results of this analysis indicated that growth in the presence of subinhibitory concentrations of tylosin leads to increased resistance to this antibiotic (Fig. 1E). We conclude that growth in the presence of tylosin increases YxjB activity, leading to increased methylation of G748 in 23S rRNA and thereby conferring resistance to this antibiotic.
A NusA-dependent terminator and an SD-sequestering hairpin are present in the yxjB leader. Since growth in tylosin resulted in increased methylation of 23S rRNA and resistance to this antibiotic, we set out to determine the molecular mechanism(s) responsible for this phenomenon. Primer extension analysis led to the identification of a single yxjB transcription start site 177 nucleotides (nt) upstream of the yxjB translation initiation codon (see Fig. S1 in the supplemental material), which allowed us to identify a A -dependent promoter upstream ( Fig. 2A). The presence of the 177-nt-long leader suggested that yxjB expression could be controlled posttranscriptionally. RNA structure predictions (17) identified a potential intrinsic transcription terminator in the yxjB leader region, suggesting that yxjB expression could be controlled by transcription attenuation (Fig. 3A). However, an overlapping antiterminator was not identified. Additional computer modeling led to the prediction that transcripts that failed to terminate would form a structure that would sequester the yxjB SD sequence and initially translated region, suggesting that yxjB could also be under the control of a translational regulatory mechanism (Fig. 3B).
Results from in vitro transcription experiments demonstrated that termination occurred over a 4-nt window, with the strongest termination occurring at position ϩ142. Although the basal termination efficiency was low, the addition of NusA, a general transcription elongation factor known to stimulate intrinsic termination in vitro and in vivo (18), resulted in a 4-fold increase in termination (Fig. 3A). We also tested the effect of terminator mutations predicted to strengthen or weaken the terminator. The G71A: G72A mutant resulted in a shortened terminator hairpin followed by two C residues. These substitutions virtually eliminated termination (Fig. 3A), consistent with previous observations that U residues immediately downstream from the hairpin are crucial (18,19). Similarly, deleting C128, which eliminated a G-C base pair and introduced a single nucleotide bulge in the hairpin, resulted in a severe termination defect. In contrast, the U131C mutation strengthened the hairpin by replacing a G-U with a G-C base pair, In vitro transcription termination assays using wild-type (WT) or mutant templates were performed in the absence (-) or presence (ϩ) of NusA. Terminated (T) and runoff (RO) transcripts are marked. Percent termination is shown below each lane. The structure of the terminator is shown below with point mutations indicated in red. A termination assay adjacent to a sequencing ladder generated with 3= dC or 3= dG is also shown. Numbering is with respect to the start of yxjB transcription. (B) Structure mapping of the SD-sequestering hairpin. yxjB RNA was subjected to limited RNase T1 digestion (ϩ). -, control without RNase T1. Partial alkaline hydrolysis (OH) and RNase T1 digestion (T1) ladders are shown. Residues that were cleaved by RNase T1 are indicated on the right. The numbering on the left corresponds to G residues in yxjB. The positions of the yxjB SD sequence and start codon (M) are marked. The structure of the yxjB SD-sequestering hairpin is on the right with cleaved G residues marked in red. Experiments were performed at least twice with comparable results. resulting in more-efficient termination (Fig. 3A). Importantly, this NusA-dependent terminator could be involved in a transcription attenuation mechanism.
In vitro RNA structure mapping experiments were also conducted to determine whether the yxjB SD-sequestering hairpin formed as predicted. A transcript extending from ϩ1 to ϩ190 was subjected to limited RNase T1 digestion, which cleaves RNA following single-stranded G residues. We observed strong cleavage at positions G85, G105, G107, and G168 and weak cleavage at G185 (Fig. 3B). These results are consistent with the structure predicted by Mfold, with four exceptions. An apparently noncanonical A46-G163 base pair forms (20), while the U30-G185, U41-G168, and U101-G107 base pairs do not form. Importantly, this structure would be capable of repressing yxjB translation (Fig. 3B).
Tylosin is required for induction of yxjB expression. P yxjB -yxjB-lacZ transcriptional and P yxjB -yxjB=-=lacZ translational fusions were constructed and integrated into the chromosomal amyE locus to examine yxjB expression. Both fusions would be subject to transcription attenuation, while the translational fusion would also be subject to any mechanism regulating yxjB translation (Fig. 2B). Expression of the transcriptional fusion was observed throughout growth, with the highest level of expression occurring during the transition to and within stationary phase ( Table 1, row 1). Introduction of the G71A and G72A substitutions that eliminated termination in vitro resulted in 2-fold-higher expression ( Table 1, rows 1 and 2). In contrast to the transcriptional fusion results, expression of the translational fusion was not observed (Table 1, row 4). The latter result was consistent with the possibility that translation was repressed by an SDsequestering hairpin.
The lack of expression of the translational fusion led us to test whether there was an interrelationship between tylosin and yxjB expression. Thus, the P yxjB -yxjB=-=lacZ translational fusion strain was grown in the presence of subinhibitory concentrations of tylosin. Expression of this fusion was observed at the lowest concentration of tylosin tested, while expression gradually increased as the concentration of tylosin was increased (Table 1, rows 4 to 7). Remarkably, the gradual increase in expression correlated precisely with the gradual increase in 23S rRNA methylation observed under the same a ␤-Galactosidase activity was measured during the mid-exponential phase (Mid-exp), the late exponential phase (Late-exp), the transition between exponential and stationary phase (Transition), and the stationary phase (Stationary). Cells were grown in the absence or presence of the indicated concentration (in micrograms per milliliter) of tylosin (Tyl) or erythromycin (Ery). Each experiment was performed at least 3 times. Values are given in Miller units Ϯ standard deviations. b yxjB-lacZ TXN, P yxjB -yxjB-lacZ transcriptional fusion; yxjB-lacZ TXN T-mut, P yxjB -yxjB-lacZ transcriptional fusion with G71A and G72A terminator mutations; yxjB=-=lacZ TLN, P yxjB -yxjB=-=lacZ translational fusion; LP=-=lacZ TLN, P yxjB -LP=-=lacZ translational fusion; yxjB=-=lacZ TLN LP-mut, P yxjB -yxjB=-=lacZ translational fusion with the LP ATG start codon mutated to ACG; yxjB=-=lacZ TLN LP-AYA, P yxjB -yxjB=-=lacZ translational fusion with the LP RYR ribosome stalling motif mutated to AYA; yxjB=-=lacZ TLN Pmut, P yxjB -yxjB=-=lacZ translational fusion with the T131A pause site mutation. c No measurable expression was observed when cells were grown in the presence of erythromycin at 0.0125, 0.025, or 0.05 g/ml. Data are shown for the highest concentration only. growth conditions (Fig. 1D). Expression of the transcriptional fusion also increased about 2-fold in the presence of tylosin (Table 1, rows 1 and 3). We conclude that tylosin induces yxjB expression and that tylosin is required for expression when the entire leader region is present, as is the case for the translational fusion.
To determine whether induction of yxjB expression was a general feature of macrolide antibiotics, we tested the effect of erythromycin. We found that cells grew well in the presence of 0.05 g/ml erythromycin but grew poorly at a concentration of 0.1 g/ml. Expression of the P yxjB -yxjB=-=lacZ translational fusion was not induced when cells were grown in the presence of subinhibitory concentrations of erythromycin (Table 1, rows 4 and 8), indicating that induction of yxjB expression is not a general feature of macrolide antibiotics and may be specific to tylosin.
Tylosin-dependent induction of yxjB expression requires translation of a leader peptide. The results described above indicate that tylosin-dependent induction of yxjB expression leads to increased methylation of 23S rRNA, thereby conferring resistance to tylosin. In addition to the terminator and SD-sequestering hairpin, we identified a 9-amino-acid open reading frame in the yxjB leader. This coding sequence was preceded by an SD sequence and followed by tandem termination codons ( Fig. 2A). Thus, expression of a chromosomally integrated P yxjB -LP=-=lacZ translational fusion was examined to determine whether the leader peptide was expressed (Fig. 2B). We found that this fusion was expressed and that expression increased throughout growth (Table 1, row 9). We then introduced an ATG-to-ACG mutation in the start codon of the leader peptide in the context of the P yxjB -yxjB=-=lacZ translational fusion to determine whether expression of the leader peptide affected yxjB expression. Preventing translation of the leader peptide by this mutation eliminated tylosin-dependent induction of yxjB expression (Table 1; compare rows 4 and 7 with rows 10 and 11). We conclude that translation of the leader peptide is required for tylosin-dependent induction of yxjB expression.
Tylosin-dependent ribosome stalling during leader peptide synthesis is required for yxjB expression. Macrolide antibiotics bind within the PET of the ribosome (9, 10). In conjunction with specific sequence motifs within the nascent peptide, bound macrolides are capable of causing ribosome stalling. These stalling sites are enriched in prolines and charged residues and can be found throughout the coding region (11). One such stalling motif, R/K-X-R/K, was identified in a transcriptomic study investigating the effect of macrolides in Staphylococcus aureus (11). The last three residues of the yxjB leader peptide (RYR) match this motif ( Fig. 2A). To determine whether this motif was involved in tylosin-dependent induction, nucleotide substitutions were introduced into the P yxjB -yxjB=-=lacZ translational fusion such that the RYR motif was changed to AYA. Other than generating an internal loop consisting of C54 to G61 and C147 to G155, these substitutions did not affect the predicted structure of the SD-sequestering hairpin (Fig. 3B). Importantly, this mutation virtually eliminated tylosin-dependent induction ( Table 1; compare rows 4 and 7 with rows 12 and 13), suggesting that tylosindependent ribosome stalling is responsible for induction of yxjB expression.
We next conducted an in vitro toeprinting assay using a coupled transcriptiontranslation assay to provide direct evidence for tylosin-dependent ribosome stalling during leader peptide synthesis. The position of stalled ribosomes was detected by primer extension inhibition. A tylosin-dependent toeprint was identified at position U74 in the WT yxjB leader but not in the AYA mutant (Fig. 4A). A toeprint was not observed when experiments were carried out with erythromycin (data not shown). Prior toeprinting studies performed with 30S ribosomal subunits identified toeprints 15 to 16 nt downstream from the A of AUG initiation codons (11,21). This distance places the tyrosine and the second arginine of the RYR motif at the P and A sites of the stalled ribosome, respectively (Fig. 4B). We conclude that induction of yxjB expression requires tylosin-dependent ribosome stalling at the RYR motif of the leader peptide.
NusG-dependent RNAP pausing is required for tylosin-dependent induction of yxjB expression. The position of the stalled ribosome provides an explanation for how tylosin induces yxjB expression. A ribosome stalled at this position would prevent completion of the terminator hairpin, resulting in transcriptional readthrough. Moreover, the stalled ribosome would prevent formation of the yxjB SD sequestering hairpin such that the yxjB SD sequence would be available for ribosome binding (Fig. 3B and  4B). This issue poses a dilemma because translation initiation of the leader peptide must occur before RNAP finishes transcribing the entire SD-sequestering hairpin, as this structure would also inhibit translation of the leader peptide. Thus, the position of actively transcribing RNAP and the timing of leader peptide translation would be critical for this regulatory mechanism. One way to circumvent this potential problem would be for RNAP to pause at an appropriate position during transcription of the yxjB leader region and thereby provide sufficient time for initiation of leader peptide translation.
Because of our interest in NusG-dependent pausing, we identified NusG-dependent pause sites throughout the B. subtilis transcriptome (A. Yakhnin, M. Kashkev, and P. Babitzke, unpublished data). This method, which combines native elongating transcript sequencing (NET-seq) with RNase footprinting of the transcripts and is called RNET-seq (22), identified three adjacent NusG-dependent pause sites in the yxjB leader that are preceded by an appropriately positioned pause hairpin ( Fig. 2A) (see also Fig. 5A and B). Note that the pause hairpin forms the apex of the terminator and SD-sequestering hairpins (Fig. 3). Importantly, pausing at these positions (U129, A130, and U131) could provide sufficient time for initiation of leader peptide translation. Interestingly, previous Term-seq studies identified what was assumed to be a terminator in the yxjB leader, when in fact the identified 3= ends actually correspond to the nascent 3= end of the stable NusG-dependent pause site (18,23).
In vitro transcription pausing assays were performed to obtain additional information on the NusG-dependent pause sites identified in vivo. Paused RNAP is visualized in this assay by an initial accumulation of a band corresponding to the paused transcription complex and then subsequent elongation to longer transcripts (24). We observed in vitro the same three pause sites as were identified by RNET-seq, and NusG greatly stimulated pausing at each position (Fig. 5C). The time course of the pausing assay indicated that RNAP paused first at position U129, then a second time at A130, and then a third time at U131. Our prior studies on the trp leader NusG-dependent pause site demonstrated that NusG contacts a T-rich region within the nontemplate DNA strand of the paused transcription bubble (24)(25)(26)(27)(28). On the basis of our prior studies, we tested the effect of mutations that were predicted to interfere with NusG-dependent pausing in the yxjB leader ( Fig. 5B and D). Simultaneously changing T123, T124, and T125 to A residues eliminated NusG-dependent pausing, while deletion of C128 resulted in a single pause at the position corresponding to U131. Changing U131 to C had no effect on pausing, whereas A and G substitutions virtually eliminated pausing at this position.
We compared the levels of expression of the P yxjB -yxjB=-=lacZ translational fusion in WT and ΔnusG strains and found that tylosin-dependent induction was nearly abolished in the ΔnusG strain (Table 1; compare rows 4 and 7 with rows 14 and 15). We also tested the effect of the T131A mutation on expression of the fusion. Consistent with our in vitro results (Fig. 5D), the T131A mutation reduced, but did not eliminate, tylosindependent induction (Table 1; compare rows 4 and 7 with rows 16 and 17). We conclude that NusG-dependent RNAP pausing contributes to tylosin-dependent induction of yxjB expression. We further infer that pausing provides sufficient time for translation of the leader peptide before the yxjB SD-sequestering hairpin has a chance to form.

DISCUSSION
Macrolide antibiotics bind within the PET of the translating ribosome and contribute to ribosome stalling when an appropriate macrolide arrest motif is encountered, such as R/K-X-R/K. In the general translation attenuation mechanism, macrolide-dependent ribosome stalling during leader peptide synthesis disrupts a downstream SDsequestering hairpin for the cognate resistance gene, thereby leading to specific antibiotic-induced expression. This general mechanism is thought to control expression Mechanism of Tylosin Resistance in Bacillus subtilis ® of several macrolide resistance genes, including 23S rRNA methyltransferases, efflux proteins, and enzymes that inactivate the antibiotic (9). A similar translation attenuation mechanism controls expression of a gene conferring resistance to chloramphenicol (8,29). A recent transcriptomics study in Listeria monocytogenes identified an antibiotic resistance mechanism that appears to combine macrolide-dependent ribosome stalling and transcription attenuation to control expression of a putative ribosome-splitting factor (30). Our results establish that yxjB encodes RlmA II , an enzyme that methylates G748 of 23S rRNA, and that modification of this residue confers resistance to tylosin (Fig. 1). Thus, we are renaming B. subtilis yxjB as tlrB to be consistent with established nomenclature (15).
Regulation of B. subtilis tlrB expression combines tylosin-dependent ribosome stalling, transcription attenuation, and translation attenuation mechanisms. The mechanism regulating B. subtilis tlrB expression is far more complex than what has been described for other macrolide-induced ribosome stalling mechanisms. In addition to transcription attenuation and translation attenuation, NusG-dependent RNAP pausing is required for tylosin-dependent induction (Fig. 5) (Table 1). Perhaps RNAP pausing participates in other macrolide-dependent induction mechanisms. Our results are consistent with the following regulatory model (Fig. 6). In the absence of tylosin, NusG-dependent pausing provides sufficient time for translation initiation of the tlrB leader peptide. Once the ribosome approaches the stop codon, it is able to interfere with formation of the terminator hairpin provided that RNAP has escaped the pause state and resumed transcription. However, once translation of the leader peptide is completed, the released ribosome is no longer able to interfere with termination. Readthrough transcripts that fail to terminate would be subject to translation attenuation because the tlrB SD-sequestering hairpin would prevent ribosome binding. The combined effects of transcription termination and translation repression result in very low levels of RlmA II synthesis. Although expression of the translational fusion was not detectable when cells were grown in the absence of tylosin (Table 1), some RlmA II must be produced under this condition because G748 modification was detected when cells were grown without tylosin (Fig. 1B). tlrB expression is induced when cells are grown in the presence of subinhibitory concentrations of tylosin. In this case, RNAP pausing provides time for translation initiation of the leader peptide. Once the translating ribosome reaches the RYR motif in the leader peptide, the combination of tylosin bound to 23S rRNA in the PET and the RYR motif in the peptidyltransferase center causes the ribosome to stall prior to incorporation of the final arginine residue positioned at the ribosome A site. The positive charge of the two arginine residues and the length of the side chains interfere with peptide bond formation (31). The stalled ribosome would prevent transcription termination because nucleotides that are required for forming the base of the terminator hairpin would be bound by the stalled ribosome. Furthermore, the stalled ribosome would prevent formation of the SDsequestering hairpin and thereby alleviate translational repression. The induced RlmA II levels would lead to increased methylation of G748 in 23S rRNA and to tylosin resistance. Tylosin-dependent induction is self-limiting, since ribosomes modified at G748 would not stall during translation of the leader peptide, resulting in tight control of the level of 23S rRNA modification. Since there is no known enzyme capable of removing the methyl on G748, the level of methylation of cells no longer exposed to tylosin would slowly diminish as the organism continued to grow. This regulatory system ensures that methylation of 23S rRNA occurs only when B. subtilis comes in contact with tylosin. Presumably, the tight control of tlrB expression minimizes a fitness cost associated with methylation of G748 when tylosin is not present.
The complex regulatory mechanism for tylosin resistance in B. subtilis appears to be conserved in other Bacillus species (see Fig. S2 in the supplemental material). In each case, the intrinsic terminator, the tlrB SD-sequestering hairpin, the general position and length of the leader peptide, and the RXR motif are conserved. Potential NusGdependent pause sites were also identified. The most significant difference in the three RNA structures is that the leader peptide SD sequence in B. pumilus is predicted to be sequestered in a weak hairpin.
Only a fraction of arrest motifs leads to macrolide-dependent ribosome stalling, indicating that additional nascent peptide residues traversing the PET participate in the stalling mechanism (11,12). This context-dependent stalling is critical for regulating expression of resistance genes, which are often specific for a particular antibiotic (9). For example, a previous study demonstrated that erythromycin is capable of causing ribosome stalling at only a subset of R/K-X-R/K motifs (11). The inability of erythromycin to cause stalling at the RYR motif in the tlrB leader peptide provides further evidence that additional sequence information on the nascent peptide is required for the specificity of macrolide-induced ribosome stalling. Since the B. subtilis tlrB leader In the presence of tylosin, the ribosome stalls at the RYR motif such the ribosome remains bound to the nascent transcript. Once RNAP resumes transcription, the tlrB SD sequence is single stranded and translation is activated. RlmA II then methylates G748 in 23S rRNA, leading to tylosin resistance. (C) Schematic representation of the series of events beginning with exposure to tylosin and culminating in tylosin resistance.

Mechanism of Tylosin Resistance in Bacillus subtilis
® peptide is only 9 amino acids in length ( Fig. 2A), when the ribosome is stalled the entire nascent peptide resides within the PET. The leader peptide from B. licheniformis is identical in sequence to that from B. subtilis; however, the leader peptide in B. pumilus differs in two positions. Hence, we generated a maximum likelihood phylogeny from a multiple-sequence alignment of 180 leader peptides predicted to be located directly upstream of tlrB homologs in Gram-positive organisms and to contain the R/K-X-R/K stalling motif at the C terminus ( Fig. S3 and S4). The leader peptides are identical in several species, suggesting that the tylosin-dependent stalling mechanism is conserved in these organisms. It is likely that at least one of the first 6 amino acids in the leader peptide sequence (MIIQFIRYR) provides specificity for tylosin-dependent stalling. Perhaps as the leader peptide sequence diverges, the specificity of induction changes to include additional or alternative macrolides. B. subtilis, S. fradiae, and several other organisms in our phylogenetic analysis are soil microbes. The ability of other species to protect themselves from the harmful effects of tylosin produced by S. fradiae would provide a distinct growth advantage relative to other organisms that are incapable of inducing resistance to this antibiotic.

MATERIALS AND METHODS
Bacterial strains and plasmids. The B. subtilis strains used in this study are listed in Table 2, the plasmids used are listed in Table S1 in the supplemental material, and the oligonucleotides used are listed in Table S2. Strain and plasmid constructions are described in Text S1 in the supplemental material.
Primer extension assays. For determination of the yxjB transcription start site, RNA was isolated from a late-exponential-phase culture of B. subtilis PLBS338 grown in LB. RNA was hybridized to a 32 P-end-labeled oligonucleotide complementary to the yxjB leader. Primer extension reaction mixtures were incubated for 15 min at 42°C using SuperScript III reverse transcriptase (Thermo Fisher Scientific). Samples were denatured by heating for 5 min at 90°C and then fractionated through 6% polyacrylamide sequencing gels. Radiolabeled bands were imaged on a Typhoon 8600 Phosphorimager (GE Healthcare Life Sciences).
For identification of 23S rRNA methylation sites, RNA was isolated from B. subtilis strains PLBS338, PLBS852, and PLBS867, and E. coli strain MG1655 was grown in LB. RNA and hybridized to a 32 P-endlabeled oligonucleotide complementary to nt 798 to 818 of E. coli 23S rRNA and to nt 845 to 865 of B. subtilis 23S rRNA. RT reaction procedures were identical to those described above except that primer extension was performed for 30 min. Details of the procedures are described in Text S1.
Transcription termination and pausing assays. Analysis of RNAP pausing was performed as described previously (24,32) with modifications. Briefly, DNA templates contained WT or mutant yxjB leader sequences driven by a A promoter as well as a 29-nt C-less cassette. Halted elongation complexes were formed by the exclusion of CTP. Elongation was resumed by the addition of all four nucleoside triphosphates (NTPs) (final concentration, 150 M) and 100 g/ml heparin, with or without 1 M NusG. Pausing reaction mixtures were incubated at 23°C, and aliquots were removed at various times. Transcription of the last aliquot was chased for 10 min at 37°C with a 0.5 mM concentration of each NTP. Transcription termination assays were performed as described for pausing except that the extension reaction mixtures were incubated at 37°C for 10 min. The 3= ends of paused and terminated transcripts were mapped using sequencing reactions performed by in vitro transcription in the presence of one of four 3= deoxynucleoside triphosphates (dNTPs). RNA bands were visualized with a phosphorimager and quantified using ImageQuant (GE Healthcare Life Sciences). Details of the procedures are described in Text S1. RNase T1 structure mapping. yxjB RNA (ϩ1 to ϩ190) was synthesized with an RNAMaxx kit (Agilent Technologies). RNA was subjected to 5= end labeling using T4 polynucleotide kinase (New England BioLabs) and [␥-32 P]ATP. Labeled RNAs were renatured by heating for 1 min at 90°C followed by cooling to room temperature. Reaction mixtures (10 l) contained 2 nM RNA, 10 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 100 mM KCl, 40 ng of yeast RNA, 7.5% glycerol, 0.1 mg/ml xylene cyanol, and 200 g/ml bovine serum albumin (BSA). RNA cleavage was performed by addition of 0.016 U RNase T1 (Thermo Fisher Scientific) followed by incubation for 15 min at 37°C. Reactions were stopped by adding 10 l stop solution. Samples were heated for 5 min at 90°C and fractionated through standard 6% sequencing gels. Cleaved patterns were examined using a phosphorimager.
␤-Galactosidase assay. B. subtilis cultures containing transcriptional or translational fusions were grown at 37°C in LB. When appropriate, growth media also contained 5 g/ml chloramphenicol, 12.5 g/ml kanamycin, 12.5 g/ml tetracycline, and various concentrations of tylosin or erythromycin. ␤-Galactosidase activity was determined as described previously (33).
Cell growth with tylosin. The PLBS867 and PLBS868 strains were grown at 37°C in LB with 0 or 4 g/ml tylosin with or without 0.5 mM IPTG (isopropyl-␤-D-thiogalactopyranoside). A photograph of the culture tubes was taken after 20 h. yxjB overexpression was confirmed by primer extension of total cellular RNA extracted from the same strains. Strain PLBS338 was grown in LB media containing various tylosin concentrations until mid-exponential phase. Cultures were serially diluted 10-fold and then spotted onto LB plates containing various tylosin concentrations. Photographs of the plates were taken the next day.
Toeprint of tylosin-induced ribosome stalling. This analysis represents a modified version of a previously published procedure (11) performed using a PURExpress system (New England Biolabs). Briefly, a DNA template containing a T7 promoter and the yxjB leader region was added to PURExpress reactions with or without tylosin or erythromycin and incubated for 1 h at 37°C. A labeled DNA toeprint primer complementary to positions 140 to 166 of yxjB transcription was added to each reaction, and primer extension was carried out for 1 h at 37°C with SuperScript III. Reactions were terminated by the addition of stop solution. Samples were denatured prior to fractionation through 6% sequencing gels. Radioactive bands were visualized using a phosphorimager. Details of the procedure are described in Text S1.