TLR13 Recognizes Bacterial 23S rRNA Devoid of Erythromycin Resistance–Forming Modification

See allHide authors and affiliations

Science  31 Aug 2012:
Vol. 337, Issue 6098, pp. 1111-1115
DOI: 10.1126/science.1220363


Host protection from infection relies on the recognition of pathogens by innate pattern-recognition receptors such as Toll-like receptors (TLRs). Here, we show that the orphan receptor TLR13 in mice recognizes a conserved 23S ribosomal RNA (rRNA) sequence that is the binding site of macrolide, lincosamide, and streptogramin group (MLS) antibiotics (including erythromycin) in bacteria. Notably, 23S rRNA from clinical isolates of erythromycin-resistant Staphylococcus aureus and synthetic oligoribonucleotides carrying methylated adenosine or a guanosine mimicking a MLS resistance–causing modification failed to stimulate TLR13. Thus, our results reveal both a natural TLR13 ligand and specific mechanisms of antibiotic resistance as potent bacterial immune evasion strategy, avoiding recognition via TLR13.

Toll-like receptor 2 (TLR2), TLR4, and TLR9 are major host sensors of Gram-negative bacteria, and TLR2 is thought to be the central detector of Gram-positive bacteria, whereas other pattern-recognition receptors (PRRs) such as TLR7 contribute to bacteria sensing as well (17). However, the high sensitivity of mice lacking expression of these TLRs to Gram-positive bacteria implies that other TLRs or members of other classes of PRRs—such as C-type lectins, RIG-I–like helicases (RLHs), or nucleotide binding domain– and leucine-rich repeat–containing receptors [NOD-like receptors (NLRs)]—play a role in the detection of Gram-positive bacteria. We therefore compared the responsiveness of macrophages lacking the expression of molecules that signal downstream of these PRRs, including caspase recruitment domain (CARD) 9, receptor-interacting protein 2, apoptosis-associated speck-like protein containing a CARD, interleukin-1 (IL-1) receptor, IL-18, or MyD88, to heat inactivated Staphylococcus aureus (hiSa) or Streptococcus pneumoniae (both Gram-positive) in the presence of a TLR2-blocking antibody (see supplementary materials and methods section) (2, 810). We found that cytokine production strictly depends on MyD88 (fig. S1A), which suggests that TLRs rather than RLHs or NLRs are responsible for the detection of these bacteria. Moreover, analysis of ectopically expressed RLH function indicated a lack of RLH involvement in Gram-positive bacteria sensing (fig. S1B).

Next, we asked whether endosomal TLRs (TLR3, -7, -8, -9, -11, and -13) are involved in cell activation. We inhibited endosomal acidification with bafilomycin and analyzed UNC93B1-mutant (3D) macrophages that lack endoplasmic reticulum–endosome TLR trafficking and are susceptible to S. aureus infection (2, 11, 12). Bafilomycin treatment abrogated recognition of Gram-positive bacteria in Tlr2–/– macrophages (Fig. 1A). Furthermore, 3D/Tlr2/4–/– and 3D/Tlr2–/– mice or corresponding macrophages (but not those generated from 3D mice unless TLR2 was blocked) were unresponsive to a Gram-positive bacterial challenge (Fig. 1, B and C, and fig. S1C). Unexpectedly, Tlr23479–/– macrophages (or mice) responded well to a hiSa challenge, unless the bacterial preparations were subjected to ribonuclease A (RNase A) treatment, which did not impair TLR2-driven activation of wild-type (WT) controls, or endosomal TLR function was abrogated (Fig. 1, D to F). These data suggested that an endosomal RNA sensor besides TLR3 and TLR7 can act as cellular detector of hiSa.

Fig. 1

Gram-positive bacteria and their RNA activate Tlr23479–/– macrophages and DCs via an unknown TLR. (A) Macrophages were preincubated for 30 min with dimethyl sulfoxide (DMSO) alone or 50 nM bafilomycin A1 and were challenged for 8 hours with 109 colony-forming units (CFU)/ml heat-inactivated S. aureus (hiSa; DMSO) or 109, 108, 107, and 106 CFU/ml hiSa (bafilomycin A1; -, unchallenged). Supernatants were analyzed by enzyme-linked immunosorbent assay (ELISA). n.d., not detected. (B) Macrophages were challenged for 16 hours with 109 or 108 CFU/ml of bacterial suspensions, whereas (C) corresponding mice were challenged intravenously (i.v.) with 109 CFU hiSa or PBS only (-) at 2 and 4 hours, upon which serum was drawn. Supernatants and serum samples were analyzed by ELISA. Cyt., cytokine; TNF, tumor necrosis factor. One out of three experiments with similar results and, respectively, n = 3 mice per group is illustrated as mean ± SD (error bars). (D and E) Macrophages were challenged for 16 hours (D) or for the times indicated (E) with untreated (-) or RNase A–treated (+) hiSa suspensions. (D) 109, 108, 107, and 106 or (E) 108 CFU/ml hiSa was used for the challenge, upon which supernatants were analyzed by ELISA or lysates were analyzed by SDS–polyacrylamide gel electrophoresis and immunoblotting. P, phosphorylated; ERK, extracellular signal–regulated kinase. (F) Mice were challenged by injection with hiSa (1.6 × 1011 CFU/kg body weight) and α-d-galactosamine (800 mg/kg body weight) intraperitoneally 45 min after intravenous injection of IFN-γ (50 μg/kg body weight). Survival was monitored, and all deaths occurred within 16 hours of treatment (n = 6 deaths per WT and 3D/Tlr2/4–/– groups, n = 4 for Tlr23479–/– mice). (G) Flt3-ligand–derived DC subsets were challenged with untreated (white bars) or RNase A–treated (gray bars) hiSa at 5 × 106 CFU/ml for 16 hours. Supernatants were analyzed for cytokine contents by bead array. The respective TLR expression (expr.) in DC subsets is indicated (-, no detectable expression; +, expression). (A to E and G) For each panel, representative results from at least three experiments are shown, and each illustrated data point (A to D and G) represents mean ± SD (error bars) of duplicates.

Dendritic cell (DC) subsets express different sets of TLRs (13). We generated bone marrow–derived conventional (c) DCs and plasmacytoid (p) DCs in vitro. The responsiveness of these cells to hiSa was dependent on MyD88 and UNC93B1. Specifically, Tlr23479–/– CD8high (expressing TLR11, TLR12, and TLR13) and signal regulatory protein α (Sirp)high cDCs (expressing TLR13 but lacking TLR11 and TLR12) responded to hiSa, whereas Tlr23479–/– pDCs (expressing TLR12 but lacking TLR11 and TLR13) failed to do so (Fig. 1G). Together, these findings imply that TLR13 acts as a bacterial single-stranded (ss) RNA sensor, even though TLR13 has recently been linked with the recognition of vesicular stomatitis virus (14).

To identify the relevant RNA, we incubated hiSa with calf intestinal phosphatase, 5′-phosphate–specific phosphatase [to affect the integrity of 16S and 23S ribosomal RNA (rRNA)], or double-stranded RNA–specific RNase III or VI. These treatments did not alter the stimulatory activity of hiSa, in line with a recent report (fig. S2, A to C) (15). However, ssRNA-specific RNase A treatment abrogated the Tlr23479–/– cDC (and macrophage) stimulatory activity of hiSa, as did nucleic acid–degrading benzonase [Fig. 1, D, E, and G, and fig. S2B; note that Flt3L-expanded CD8+ cDCs do not produce IL-12p70 in response to TLR2 ligands that are contained in hiSa (16)]. We then treated total RNA with 5′-phosphate–dependent exonuclease (to degrade specifically large rRNAs, namely 16S and 23S rRNA) and purified large rRNAs (fig. S2C) to narrow down the stimulatory activity. After transfection, large rRNA isolates of both S. aureus and Escherichia coli triggered the activation of Tlr23479–/– macrophages and cDCs, whereas 16S/23S rRNA digestion abrogated stimulatory activity (Fig. 2A). Accordingly, low–molecular weight portions from total RNA lacked stimulatory activity, whereas high–molecular weight portions of Gram-positive and Gram-negative bacterial RNA activated Tlr23479–/– cells (Fig. 2B and fig. S2, D and E). These findings suggested that a fraction of large bacterial rRNAs activates macrophages and cDCs in a MyD88-dependent manner. We assume that the increased RNA-driven activation of Tlr23479–/– macrophages in comparison to WT cells reflects a lack of TLRs competing for downstream signal transduction molecules.

Fig. 2

Bacterial 23S rRNA is stimulatory unless it is methylated by erms. (A) Bacterial RNA preparations resulting from incubation of total RNAs with 5′-phosphate–specific exo RNase targeting large rRNAs (dig.) or precipitation of both large rRNAs (pur.), as well as untreated mouse fibroblast cell line (3T3) RNA (control) were transfected into macrophages using N-[1-(2,3-dioleoyloxy)]-N,N,N-trimethylammonium propane methylsulfate (Dotap, Roche Diagnostics, Mannheim, Germany). S.a., S. aureus; E.c., E. coli. (B) Bacterial total RNAs were separated by anion-exchange chromatography into low–molecular weight (lmw) and high–molecular weight (hmw) fractions, which were used to challenge macrophages with or without Dotap transfection. (C) Macrophages were challenged with 109 CFU/ml erythromycin-sensitive (std.) hiSa or five erythromycin-resistant clinical S. aureus isolates cultured in 10 mg/liter erythromycin (clin. isolat. +ery). (A to C) Supernatants were analyzed 16 hours poststimulation using ELISA. (D) Tlr23479–/– mice were infected i.v. with 108 CFU erythromycin-resistant S. aureus clinical isolate growing logarithmically in the presence (+) or absence (-) of erythromycin. Serum was drawn after 2 hours and analyzed for cytokines by cytometric bead array. Mean ± SD (error bars) for n = 6 mice per each group is shown. (E) Total (tot.) RNA from erythromycin-sensitive S. aureus (std.) and agarose gel–purified 16S (16) and 23S (23) rRNAs from clinical isolate 2 grown in the absence or presence of erythromycin (ery) were transfected into Tlr23479–/– macrophages using Lyovec (Cayla-InvivoGen, Toulouse, France). (F) E. coli BL21 was transformed with empty vector control (ctrl) or ermB or ermC expression plasmids. After 16 hours, culture of 16S and 23S rRNA was isolated and transfected into Tlr23479–/– macrophages. (E and F) Supernatants were analyzed 16 hours postchallenge using ELISA. (A to C, E, and F) Each panel illustrates a representative result of three independent experiments and depicts means ± SD (error bars) of duplicate samples. For (D), one statistically significant experiment has been performed (P ≤ 0.015).

To analyze whether rRNA modifications induced in antibiotic-resistant strains by antibiotic treatment [e.g., with erythromycin (17, 18)] would modify the immunostimulatory capacity of rRNA, we applied five clinical S. aureus isolates displaying various resistance phenotypes, including erythromycin resistance. Isolates grown in the presence of erythromycin largely lacked the capacity to activate Tlr23479–/– macrophages and induced lower amounts of serum cytokines early after infection (2 hours) of Tlr23479–/– mice (Fig. 2, C and D). In contrast, WT as well as Tlr23479–/– mice and corresponding macrophages responded largely normally toward the same isolate grown in the absence of erythromycin (Fig. 2, C and D, and fig. S2, F to H). The later (16 hours) increase and equalization of serum cytokine levels independent of erythromycin treatment (fig. S2H) suggested the loss of 23S rRNA methylation in the absence of erythromycin within the host. Together, these results demonstrate an erythromycin-driven camouflage of RNA from its receptor. Specifically, N6 methylation of rRNA adenosine (A) 2085 in S. aureus (corresponding to E. coli A2058) by the erythromycin resistance methyltransferase B (ermB) or ermC confers macrolide, lincosamide, and streptogramin group (MLS) antibiotic (including erythromycin) resistance (17, 18). Accordingly and also in line with the inducibility of erm expression by erythromycin (17, 18), 23S rRNA from S. aureus grown in erythromycin failed to stimulate Tlr23479–/– macrophages (Fig. 2E). In contrast, 23S rRNA from resistant S. aureus not grown in erythromycin and 23S rRNA from E. coli (including enterohemorrhagic E. coli) activated Tlr23479–/– macrophages, whereas the respective 16S rRNAs failed to do so (Fig. 2E and fig. S2, I and J). Moreover, overexpression of ermB and ermC (the latter being subcloned from cDNA of an erythromycin-grown S. aureus isolate) in E. coli and Bacillus subtilis strains not only conferred erythromycin resistance but also ablated 23S rRNA stimulatory activity (Fig. 2F and fig. S2K). These data indicate that resistance to MLS group antibiotics (including erythromycin) mediated by site-specific methylation (targeting A2085 in S. aureus and A2058 in E. coli 23S rRNA) rendered 23S rRNA nonstimulatory.

To address the immune stimulatory activity of 23S rRNA in more detail, we designed three oligoribonucleotides (ORNs) as analogs of S. aureus 23S rRNA segments, each of which carries an A in its center that becomes methylated constitutively or under growth restriction to modulate the docking of protein synthesis cofactors or antibiotics. The three ORNs named SaI, SaII, and SaIII represented S. aureus A1662 [E. coli A1616, methylation of which promotes fitness (19)], S. aureus A2530 [E. coli A2503, targeted by chloramphenicol, florfenicol, and clindamycin resistance RNA methyltransferase (20)], and S. aureus A2085 [E. coli A2058, modification of which costs fitness (17, 18, 21)], respectively (table S1).

Only SaIII (which mirrors S. aureus A2085) activated Tlr23479–/– cells (Fig. 3A). pDCs recognized SaIII via TLR7, but this activity was lost with 3′-terminal deletion (fig. S3). ORNs resulting from deletions of 3′- and 5′-termini (SaIIId3, SaIIId5, Sa23) equally activated Tlr23479–/– cDCs (Fig. 3B), whereas preincubation of S. aureus RNA or of ORN Sa23 with an antisense SaIII RNA strand (SaIIIas) abrogated the stimulatory activity (Fig. 3C). These results indicated single-strand structure and singularity of the stimulatory activity within the bacterial transcriptome. Successive terminal deletions toward a 12-mer ORN (Sa12, table S1) led to sequences that were identical in S. aureus and E. coli 23S rRNAs. Length-dependent reduction of stimulatory capacity could largely be compensated by terminal fill-ups (Sa12A19, Fig. 3D) (22). Upon N6 methylation at A6 (corresponding to S. aureus A2085 and mimicking erm-methylated 23S rRNA), Sa12 lacked stimulatory capacity, whereas N6 methylation at A7 merely caused a partial reduction (Fig. 3E). Consecutive single substitutions of Sa12 revealed “CGGAAAGACC” as the minimal stimulatory segment because ORNs with substitutions at position one or two of Sa12 (Sa12s1 and Sa12s2) were fully stimulatory, whereas further substitutions resulted in drastic loss (Sa12s10 and Sa12s12) or abrogation of the stimulatory activity (Fig. 3F and table S1).

Fig. 3

Oligoribonucleotides (ORNs) covering the erm target site in 23S rRNA (region around A2085 in S. aureus/2058 in E. coli) activate macrophages and cDCs. (A) Sequence motifs covering three separate methylation sites in S. aureus 23S rRNA were mirrored by ORNs (see table S1). Macrophages were challenged with 1, 10, and 100 pmol per well of the ORNs. NO, nitrite. (B) Tlr23479–/– FL-CD8+ cDCs were transfected with the ORNs indicated (amount per well [pmol]: black, 10; gray, 1; white, 0.1). (C) Tlr23479–/– Sirphigh cDCs were transfected with 100 ng per well of the S. aureus RNA preparations indicated or an ORN covering the SaIII core sequence (10 pmol per well), either in the absence of (none) or upon preincubation for 20 min with 100 pmol per well antisense RNA ORN (SaIIIas, +antisense). (D to F) Undifferentiated bone marrow cells were challenged with ORNs at the doses per well indicated in (D) and (E) or 100 pmol per well (F). (A to F) Cells were transfected [(A) Dotap, (B to F) Lyovec] for 16 hours with the indicated RNAs and ORNs. In each experiment, supernatants were assayed for nitrite content by Griess assay (A) or proinflammatory cytokine contents by bead assay (B and C) or ELISA (D to F). Each panel illustrates a representative result of three independent experiments and depicts means ± SD (error bars) of duplicate samples (A to E) or the mean ± SD of at least three independent experiments (F).

In contrast, Sa12 derivatives mimicking eukaryotic 28S rRNA or specific 23S rRNA mutations that render bacteria resistant to MLS antibiotics (S. aureus 23S rRNA A2085G, mimicked by ORN Sa12s6G or Sa12s6U) failed to stimulate bone marrow cells (Fig. 3F and table S1) (18, 23). These findings suggest that molecular mechanisms rendering bacteria resistant to naturally occurring antibiotics also impede MyD88-dependent host recognition by an ill-defined endosomal TLR.

To characterize the responsible TLR, we focused on TLR13, because analysis of Tlr8–/– macrophages ruled out the involvement of TLR8. Specifically, WT and Tlr8–/– macrophages exhibited comparable response to hiSa upon blockade of TLR7, TLR9, and TLR2. Moreover, responsiveness to 23S rRNA–derived SaIII was similar (fig. S4A). Notably, small interfering RNA (siRNA)–driven suppression of TLR13 mRNA accumulation impaired the recognition of stimulatory ORNs such as SaIII by Tlr23479–/– macrophages (Fig. 4A). Although recognition of low doses of hiSa by Tlr23479–/– macrophages treated with siRNA for TLR13 was strongly impaired, high-dose hiSa challenge activated not only control but also TLR13 siRNA-treated cells, presumably via unsuppressed TLR13 molecules (fig. S4B). In addition, knockdown of MAPK1 mRNA indicated involvement of MAPK1 in TLR13-driven signal transduction (Fig. 4A and fig. S4B). Furthermore, ectopic expression of TLR13 but not of CD14, TLR3, -7, -8, -9, or -12 conferred responsiveness of human embryonic kidney (HEK) 293 cells toward hiSa or the ORNs SaIII, Sa23, Sa17, or Sa12 (Fig. 4, B to D, and fig. S4, C and D). Other ORNs such as RNA40 (TLR7 ligand) or CpG-containing oligodeoxynucleotides (ODNs) (TLR9 ligands) were inactive (Fig. 4E).

Fig. 4

TLR13 recognizes heat-inactivated S. aureus and ORNs mirroring bacterial 23S rRNA segments covering A2085/2058. (A) 5 × 105 Tlr23479–/– macrophages were transfected with 50 pmol mRNA-specific siRNAs or scrambled control siRNA (scram.). After 48 hours, cells were challenged for 16 hours with 100 pmol per well ORN SaIII (black columns, bottom) or left untreated (white columns, bottom), and supernatants were analyzed by ELISA (bottom). Untreated cells were lysed to isolate mRNA, and levels of corresponding mRNAs were determined by reverse transcription polymerase chain reaction (top and middle). (B to E) HEK293 line cells were transfected with control, TLR2, or TLR13 expression and luciferase reporter plasmids. In general, cells were transfected with 15 ng empty vector (empty v.), 2 ng TLR2 (C), 15 ng TLR13 (D and E), or the amounts of TLR13 expression plasmid indicated in (B) and (C). At 24 hours posttransfection, cells were challenged with 109, 108, and 107 CFU/ml of hiSa (B); 109 CFU/ml hiSa (C); 100, 10, and 1 pmol per well ORN (C and D); 100 pmol per well ORN only (D and E); or 100 and 10 pmol per well ORN (E). Either 10 μM of oligodeoxynucleotides (ODN, 1668 and 2006) only or 10 and 1 μM of ODN was applied. ORN RNA40 was transfected with the reagent Dotap (E). After incubation for 16 hours, NF-κB–driven relative luciferase activity (rel. lucif. activ.) was analyzed. n.p., not performed; -, no challenge. (A to E) Each panel illustrates a representative result of three experiments and depicts means ± SD (error bars) of triplicate samples. (F) WT mice were challenged by i.v. injection of 10 nmol of ORN or ODN (n = 9 mice per group) in 200 μl PBS or PBS alone (-). Serum was drawn 6 hours later and analyzed for IL-12p70 content by cytometric bead arrays (IL-12). Combined data of three experiments in which three mice per group were applied are shown as the mean of individual results.

Having identified the conserved 23S rRNA sequence “CGGAAAGACC” as a ligand for TLR13, we set out to evaluate its importance in vivo. Therefore, we compared the cytokine storm induced by systemic application of TLR13-activating ORNs with that of TLR9-activating CpG-ODNs. Application of a nuclease-resistant phosphorothioate Sa19 variant (Sa19 PSO) in vivo triggered systemic proinflammatory cytokine release similar to that elicited by the PSO-CpG oligonucleotide 1668 (Fig. 4F and fig. S4, E and F). Consequently, systemic application of Sa19PSO to mice along with interferon-γ (IFN-γ) and d-galactosamine sensitization induced a fatal, septic shock–like syndrome in mice with functional TLR13 (WT and Tlr23479–/–), whereas the 3D/Tlr24–/– mice that lack responsiveness to TLR13 were resistant (fig. S4G), concordant with the genotype-selective fatal pathology elicited by systemic challenge with hiSa (Fig. 1F). In contrast to the ORN Sa19, an ODN version of Sa19 (Sa19DNA, containing two CpG motifs) lacked TLR13 stimulatory activity but activated TLR9 (fig. S4F). Together, these data indicate that TLR13 functions as an important bacteria sensor by recognizing an ssRNA segment within the peptidyl transferase loop of bacterial 23S rRNA that binds antibiotics of the MLS group.

Our data unravel an unanticipated link between antibiotic resistance and evasion from TLR13 recognition, because 23S rRNA modifications generating resistance toward MLS antibiotics also camouflaged bacteria from TLR13 recognition. MLS antibiotic-producing bacteria such as Saccharopolyspora erythraea were possibly first to express erms (to resist their own antibiotics) (17). Erm expression plasmids might have been acquired from S. erythraea by staphylococci, pneumococci, and mycobacteria (which seem to accompany or even correlate with the tuberculous property of the latter) (17, 24). Though macrolide resistance appears to be associated with fitness costs (21), the pathogenic recipients did gain invisibility to TLR13. We therefore speculate that widespread ancient antibiotic resistance (25) has subverted TLR13-driven antibacterial immune resistance, which may explain why TLR13 expression has been abandoned in certain mammalian species, including humans. If so, we anticipate that, in humans, the function of TLR13 has been replaced by an RNA-sensing PRR that is able to still recognize erythromycin resistance–forming RNA modifications.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Table S1

References (2632)

References and Notes

  1. Acknowledgments: We thank S. Schimanski, J. Pätzold, R. Kassub, T. Scholtysik, and P. Juszczak, as well as C. Chebrolu for technical work; M. Freudenberg, J. Steinmann, and J. Kehrmann for provision of mice and bacteria; R. Lang for advice; the German research foundation (Deutsche Forschungsgemeinschaft) for funding of KI 591/4-1 and GK1045 (to C.J.K.) and BA 1618/5-1 (to S.B.), as well as to U.K.; and the Japanese Government for funding to S.A. and T.K. The Behring Röntgen Stiftung funded grant 56-0034 (to S.B.). The Univ. of Duisburg-Essen (C.J.K.), Bavarian Nordic GmbH (H.H.), and the Philipps Univ. of Marburg (S.B.) filed a patent on TLR13 activators (U.S. Patent 61/597,063; agonists and antagonists of TLR13). H.W. and S.B. had an advisory contract with Coley Pharmaceuticals/Pfizer that ended in 2010. The data presented in this paper are tabulated in the main paper and in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article