Resistance to ketolide antibiotics by coordinated expression of rRNA methyltransferases in a bacterial producer of natural ketolides

Abstract

Ketolides are promising new antimicrobials effective against a broad range of Gram-positive pathogens, in part because of the low propensity of these drugs to trigger the expression of resistance genes. A natural ketolide pikromycin and a related compound methymycin are produced by Streptomyces venezuelae strain ATCC 15439. The producer avoids the inhibitory effects of its own antibiotics by expressing two paralogous rRNA methylase genes pikR1 and pikR2 with seemingly redundant functions. We show here that the PikR1 and PikR2 enzymes mono- and dimethylate, respectively, the N6 amino group in 23S rRNA nucleotide A2058. PikR1 monomethylase is constitutively expressed; it confers low resistance at low fitness cost and is required for ketolide-induced activation of pikR2 to attain high-level resistance. The regulatory mechanism controlling pikR2 expression has been evolutionary optimized for preferential activation by ketolide antibiotics. The resistance genes and the induction mechanism remain fully functional when transferred to heterologous bacterial hosts. The anticipated wide use of ketolide antibiotics could promote horizontal transfer of these highly efficient resistance genes to pathogens. Taken together, these findings emphasized the need for surveillance of pikR1/pikR2-based bacterial resistance and the preemptive development of drugs that can remain effective against the ketolide-specific resistance mechanism.

Keywords: antibiotics; ketolides; macrolides; resistance; ribosome.

Fig. 1.

S. venezuelae pikR resistance genes and ketolide antibiotics. (A) The structure of the MTM/PKM biosynthetic gene cluster in S. venezuelae ATCC 15439. The polyketide synthase (pikA) and desosamine biosynthesis (des) gene operons along with the pikC and pikD genes are required for production of the active MTM and PKM antibiotics. The putative resistance genes pikR1 and pikR2 precede the MTM/PKM biosynthesis operon. (B) Structures of the natural antibiotics MTM and PKM produced by S. venezuelae ATCC 15439, the semisynthetic clinical ketolide TEL, and the C3-cladinose containing macrolide ERY. The dotted box in the TEL structure highlights the keto group, which replaces the cladinose sugar.

Fig. S1.

Similarity of proteins encoded in the pikR1 and pikR2 genes in S. venezuelae ATCC 15439.

Fig. 2.

PikR1 and PikR2 RNA methyltransferases target A2058 in the 23S rRNA. (A) Primer extension analysis of m₂⁶A modification of rRNA extracted from WT E. coli cells (W; lane 3) or those constitutively expressing pikR1 (R1; lane 2) and pikR2 (R2; lane 1) genes. Sequencing lanes are marked C, U, A, G. Full gels are shown in Fig. S2A. (B) Primer extension analysis of the same samples as in A but carried out under conditions optimized for detection of m⁶A modification (Materials and Methods). The E. coli ΔrlmJ mutant, which lacks the native m⁶ modification of A2030 (33), was used as a control (Δ; lane 4). (C and D) MALDI-TOF analysis of the RNaseA-generated 23S rRNA fragment encompassing nucleotide A2058. rRNA samples were prepared from cells expressing (C) PikR1 or (D) PikR2.

Fig. S2.

Primer extension analysis of modification of rRNA extracted from (A) E. coli or (B) S. venezuelae. The segments of the full gels represented in Figs. 2 and 3 are boxed. Sequencing lanes are marked with the letters C, U, A, G according to the rRNA nucleotides. W, WT.

Fig. 3.

Expression of pikR2 in S. venezuelae is activated during antibiotic production. (A and B) Primer extension analysis of rRNA extracted from WT S. venezuelae ATCC 15439 carried under conditions specific for detection of (A) m₂⁶A or (B) m₂⁶A and m⁶A modification. Full gels are shown in Fig. S2B. RNA was extracted from WT cells or mutants unable to produce active antibiotics because of deletion of the pikAI-pikAIV (ΔpikA) or desI (ΔdesI) genes. Sequencing lanes in A and B are marked C, U, A, G. (C–E) MALDI-TOF analysis of 23S rRNA fragments from the WT or the ΔpikA and ΔδεσI KO mutants of S. venezuelae; the fragments were generated with RNaseA and encompass nucleotide A2058.

Fig. S3.

Primer extension on 23S rRNAs from (A) S. venezuelae and (B) E. coli. The extension reactions were performed with 1 mM dTTP and 1 mM ddCTP (H) or 0.01 mM dTTP and 1 mM ddCTP (L). The dideoxynucleotide completely stops reverse transcription at G2056 in S. venezuelae rRNA and at G2057 in E. coli rRNA. Gel band intensities were measured by scanning on a Typhoon FLA 9500 (GE Healthcare) and used to quantify the stops at A2058. The band at A2058 in extensions with 1 mM dTTP is caused by m⁶₂ dimethylation, whereas extensions with the lower dTTP concentration detect both monomethylated and dimethylated A2058. The proportion of A2058 nucleotides that were m⁶₂-dimethylated after ketolide induction was slightly above 30% (WT H lane). Note that, in contrast to adenine m⁶₂ dimethylation, m⁶A monomethylation does not completely arrest primer extension, even under low dTTP concentration conditions; therefore, the primer extension technique can be used to reliably quantify the extent of A2058 dimethylation but not monomethylation. G, A, U, and C are dideoxy sequencing lanes. K, E. coli cells with the empty vector; P, primer band. Sequencing lanes are marked G, A, U, C.

Fig. S4.

Primer extension analysis of mono- and dimethylation of A2058 in 23S rRNA extracted from different S. venezuelae mutants without (−) and with (+) preincubation for 10 h with one-fourth MIC of TEL. All of the strains used in the experiment were derivatives of the ΔpikA strain (WT*) containing only pikR1 (R1), only pikR2 (R2), or lacking both of the pikR genes (Δ).

Fig. 4.

The pikR2 regulatory region controls the inducible expression of the pikR2 resistance gene. (A) The putative leader ORF pikR2L precedes the pikR2 resistance gene. The nucleotide sequence of the ORF and the amino acid sequence of the encoded leader peptide are shown. (B) Antibiotic disk diffusion assay reveals inducibility of pikR2 in E. coli. In the reporter construct in E. coli cells, the pikR2 regulatory region controls expression of the lacZα reporter. The antibiotic disks contained TEL, MTM, PKM, ERY, or chloramphenicol (CHL). The clear areas around the disks contain antibiotic concentrations that inhibited cell growth. Blue halos around the ketolide-containing disks indicate drug-dependent induction of the reporter. (C, Upper) Toe-printing analysis shows ketolide-induced ribosome stalling at the Leu13 codon of the pikR2L ORF. The band of the ribosomes stalled by the control antibiotic thiostrepton (THS) at the initiation codon is indicated by black arrows. The ribosomes arrested with the Leu13 codon in their P site are shown by the red arrows. The ribosomes that reached the pikR2L 13th codon but failed to arrest translation were captured at the next Arg14 codon (blue arrows) because of the depletion of Ile-tRNA from the translation reaction by the presence of the Ile-tRNA synthetase inhibitor, mupirocin. The stop codon or the ORF is indicated with an asterisk. (C, Lower) Stalling efficiency calculated from the ratios of the intensity of the bands representing ketolide-dependent arrest (codon 13) vs. read through (codon 14). Error bars indicate data spreads in two independent experiments.

Fig. S5.

The models of the secondary structure of pikR2 mRNA in the (A) noninduced and (B) induced states. The nucleotide sequence of the leader ORF pikR2L is italicized. The RLR sequence in the PikR2L leader peptide encompassing the Leu13 stalling codon is boxed, and the codon is underlined. The Shine–Dalgarno region of the pikR2 gene is shown in red, and the initiator codon is blue. The models are based on mfold predictions (40).

Fig. S6.

Growth competition reveals low fitness cost associated with A2058 monomethylation in S. venezuelae. S. venezuelae strains Δ and R1 deficient in antibiotic production and lacking both resistance genes (Δ) or carrying only pikR1 gene (R1) were mixed at approximately equal amounts and grown for the indicated number of generations with four consecutive passages with 1:1,000 dilutions. After isolation of the total RNA, the fraction of the R1 strain was assessed by the extent of A2058 monomethylation using primer extension. If RT did not pause because of A2058 modification, it would completely stop at C2055 because of incorporation of ddGTP. Primer extensions carried out on RNA prepared from the pure cultures of the Δ and R1 strains were used for calibration. The estimated fraction of the R1 strain in the co-growth culture is plotted.

Fig. S7.

In contrast to the mechanism seen for pikR2, regulation of the pikR1 gene does not occur by stalling at its putative upstream ORF. (A) The sequence of the putative upstream ORF (pikR1L) and the encoded protein sequence. (B) Toe-printing analysis does not show any ketolide- or macrolide-specific stops during in vitro translation of the pikR1L ORF. The black arrows indicate the toe-printing band corresponding to the ribosome stalled by the control antibiotic thiostrepton (THS) at the pikR1L initiator codon (boxed). The green arrows show the band representing the ribosomes that were captured at the pikR1L 11th codon (boxed) because of the depletion of Ile-tRNA from the translation reaction by the presence of the Ile-RS inhibitor, mupirocin. Sequencing lanes are marked G, A, U, C.