Clinical use of antibiotics is based on their capacity to inhibit bacterial growth via bacteriostatic or bacteriocidal effects. In this article, we show that the aminoglycoside antibiotic neomycin, the cyclic lipopeptide antibiotic polymyxin B, and the cyclic peptide antibiotics gramicidin and tyrothricin can induce IL-1β secretion in bone marrow dendritic cells and macrophages. LPS priming was required to trigger the transcription and translation of pro–IL-1β but was independent of TNFR or IL-1R signaling. All four antibiotics required the NLRP3 inflammasome, the adaptor ASC, and caspase-1 activation to secrete IL-1β, a process that depended on potassium efflux but was independent of P2X7 receptor. All four antibiotics induced neutrophil influx into the peritoneal cavity of mice, which required NLRP3 only in the case of polymyxin B. Together, certain antibiotics have the potential to directly activate innate immunity of the host.
The morbidity and mortality related to bacterial infections has dramatically improved by the use of antibiotics. Antibiotics are naturally occurring toxins that are secreted by microorganisms to suppress the growth of competitors in the same ecological niche. Antibiotics in clinical use encompass semisynthetic compounds that have been chemically optimized to improve biostability and efficacy. The clinical use of antibiotics is based on their capacity to kill bacteria (bactericidal effect) or to inhibit bacterial growth (bacteriostatic effect), which supports the host’s immune system to eradicate the pathogen. As such, antibiotics and the host’s immune defense have synergistic effects in controlling bacterial infections. Currently, the clinical use of antibiotics is based on the concept that they reduce the bacterial load, enabling the host’s immune defense to control those that survive or that are resistant to the antibiotic drug. A direct immunomodulatory effect of antibiotics on host defense is speculative.
Generally, some bacterial products can activate the host’s immunity through innate pathogen recognition receptors such as TLR or the inflammasomes (1, 2). For example, bacterial cell wall components activate TLR1, TLR2, TLR4, TLR6, and NLRP1; bacterial nucleic acids activate TLR9; and the AIM2 inflammasome or bacterial proteins activate TLR5 and the NLRC4 inflammasome. TLR activation triggers the secretion of multiple proinflammatory chemokines or cytokines and induces pro–IL-1β (1). In contrast with all other cytokines, the secretion of IL-1β requires inflammasome-mediated activation of caspase-1, also referred to as IL-1β–converting enzyme (2). Four different inflammasomes have been described to integrate the various endogenous and exogenous triggers of caspase-1 activation: NLRP1, NLRP3, NLRC4, and AIM2 (2–5). Recently, distinct bacterial exotoxins have been reported to specifically trigger the NLRP1 inflammasome-mediated (bacillus anthracis lethal toxin) or the NLRP3 inflammasome-mediated caspase-1 activation (6–8), including the bacterial ionophore nigericin (9, 10). We therefore questioned whether antibiotics, another class of bacterial toxins, might have the potential to directly activate inflammasome- and caspase-1–mediated release of IL-1β, and thereby enhance innate immunity as part of the host defense against invading pathogens.
Materials and Methods
Bone marrow-derived dendritic cells (BMDCs) and macrophages were isolated from 6-wk-old C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME) by established protocols. In some experiments, BMDCs were prepared from NLRP3- (11), ASC- (9), P2X7- (12), TNFR1/2- (The Jackson Laboratory), or IL-1R1–deficient mice (The Jackson Laboratory). Human PBMCs were isolated from whole human blood of healthy, voluntary donors by Ficoll-Hypaque density-gradient centrifugation (Biochrom). Experiments involving human materials were in accordance with precepts established by the Helsinki Declaration. Cells were cultured in RPMI medium 1640/GlutaMAX-I medium (Invitrogen) supplemented with 10% FBS (v/v; Biochrom KG), 1% of penicillin and streptomycin (PAA Laboratories). All cells were stimulated in serum-free RPMI 1640 medium at a density of 1× 106 cells/ml. Cells were prestimulated for 3 h with ultrapure LPS (10 ng/ml; Invivogen); then later stimulated for 6 h with antibiotics (50 μg/ml; Sigma), ATP (5 mM; Invivogen), and monosodium urate (250 μg/ml; Invivogen). In some experiments, cells were prestimulated for 6, 12, and 24 h. For all conditions, cell-free supernatants were analyzed for cytokine secretion by ELISA. Supernatants were concentrated with Amicon ultrafilters (Millipore), or cells were lysed for immunoblot analysis. Inhibitors such as Z-VAD-FMK (20 μM; Invivogen), cytochalasin D (5 μM), CA-074-Me (10 μM; Calbiochem), and N-acetyl cysteine (10 mM; Sigma) were added before 30 min of antibiotic stimulation. Cycloheximide (10 μM; Sigma) and Bay11-7082 (20 μM; Invivogen) were added 1 h before LPS prestimulation. KCl (75 mM) was used to increase extracellular K+ concentration. NaCl (75 mM) was used in control experiments.
Cell supernatants were analyzed for IL-1β secretion by ELISA (BD Biosciences). Precipitated media supernatants or cell extracts were analyzed by standard immunoblot techniques as described previously (13). Primary Abs were polyclonal goat Ab to mouse–IL-1β (R&D Systems) and to anti–caspase-1 (sc-514; Santa Cruz Biotechnology).
Mouse peritonitis model
Six-wk-old wild-type or NLRP3-deficient C57BL/6 mice received a single i.p. injection with 100 μg LPS (Invivogen). Three hours later, second injection with 100 μl of either PBS or 250 μg of selected antibiotics was given through the same route. After 6 h, all mice were sacrificed and peritoneal cavities were washed with 5 ml PBS. The lavage fluids were analyzed for neutrophil recruitment by FACS using Ly-6G and 7/4 (BD Biosciences and AbD Serotec).
Data are expressed as the mean ± SD. Comparison between two groups was performed by two-tailed t test. A p value <0.05 was considered to be statistically significant. All statistical analyses were calculated using GraphPad Prism.
Distinct polypeptide or aminoglycoside antibiotics activate dendritic cells and macrophages to secrete IL-1β
To address a putative immunostimulatory potential of antibiotics, we exposed LPS-primed BMDCs to selected members of commonly used classes of antibiotics and measured IL-1β production in cell culture supernatants after 6 h of stimulation. Among all antibiotics tested, only polymyxin B, tyrothricin, gramicidin, and neomycin induced IL-1β release in BMDCs (Fig. 1A). This effect was dependent on priming with LPS (Supplemental Fig. 1A), which provides the necessary signal for the induction of pro–IL-1β (2). Increased time duration of LPS priming augmented IL-1β secretion and caspase-1 activation (Supplemental Fig. 1B). However, 24-h LPS priming is toxic to cells (data not shown). NF-κB inhibition with Bay 11-7082 and inhibition of protein translation with cycloheximide abrogated secretion of mature IL-1β, as well as caspase-1 activation (Supplemental Fig. 1C), suggesting that de novo transcription and translation of pro–IL-1β by LPS-driven NF-κB signaling as a mandatory step. Secondary TNF or IL-1 release was not required for this process as polymyxin B, tyrothricin, gramicidin, and neomycin induced the secretion of similar IL-1β levels and caspase-1 activation in TNFR1/2 or IL-1R1–deficient mice (Fig. 1B, Supplemental Fig. 1D) (14). Similar results were obtained in bone marrow-derived macrophages (Supplementary Fig. 2A). However, tyrothricin and gramicidin showed less induction of IL-1β secretion in human PBMCs (Supplemental Fig. 2B). Interestingly, polymyxin B, tyrothricin, and gramicidin are cyclic polypeptide antibiotics, whereas neomycin is an aminoglycoside antibiotic. Furthermore, structurally related antibiotics showed different capacity to induce IL-1β secretion, for example, cyclic polypeptides polymyxin B and E (colistin) or the aminoglycosides neomycin and gentamicin (Fig. 1A).
Polymyxin B, tyrothricin, gramicidin, and neomycin trigger IL-1β release via the NLRP3 inflammasome
Various compounds such as crystals, pore-forming toxins, or ATP induce IL-1β secretion by activating the NLRP3 inflammasome (15). We therefore exposed polymyxin B, tyrothricin, gramicidin, and neomycin to BMDCs isolated from NLRP3-deficient or wild-type mice. Lack of NLRP3 substantially decreased IL-1β release on stimulation with all four antibiotics (Fig. 2A). However, polymyxin B and gramicidin elicited a partial response in NLRP3-deficient BMDCs as detected by an IL-1β ELISA, which might also detect extracellular pro–IL-1β (Fig. 2A). To address this possibility, we performed immunoblotting of cell culture supernatants to dissect the pro–IL-1β from the cleaved IL-1β (p17). In NLRP3-deficient cells, cleaved p17 was completely absent also on polymyxin B and gramicidin exposure. In addition, immunoblotting for the p10 subunit of caspase-1 in cell isolates of wild-type and NLRP3-deficient cells was performed to study caspase-1 activation. Stimulation with all four antibiotics induced the caspase-1 cleaving product p10 in wild-type cells but much less in NLRP3-deficient cells, showing that NLRP3 is required for antibiotics-induced caspase-1 activation (Fig. 2B). NLRP3 needs the adaptor ASC to activate caspase-1 (15). In fact, lack of ASC, as well as the pancaspase inhibitor Z-VAD-FMK, abrogated IL-1β release in BMDCs on stimulation with all four antibiotics (Fig. 2C, 2D). Thus, IL-1β secretion induced by polymyxin B, tyrothricin, gramicidin, and neomycin involves the NLRP3/ASC/caspase-1 pathway.
Antibiotics-induced NLRP3 activation depends on potassium flux but not on P2X7
Several models of NLRP3 activation have been described, that is, P2X7- and potassium flux-dependent pore formation (9), endosomal rupture involving cytosolic cathepsin B activity (16), and oxidative stress (17). An extracellular potassium concentration of 75 mM suppressed potassium efflux and abrogated IL-1β secretion on ATP, as well as on polymyxin B, gramicidin, tyrothricin, and neomycin (Fig. 3A). However, ATP-induced IL-1β secretion requires the P2X7 receptor, but all four antibiotics could induce IL-1β release in P2X7-deficient BMDCs (Fig. 3B). Potassium flux is obviously important for antibiotics-induced NLRP3 activation but is different from ATP because these antibiotics do not require P2X7. To test the role of the phagocytosis-dependent endosomal leakage pathway of NLRP3 activation, we exposed BMDCs to all four antibiotics in the presence or absence of the phagocytosis inhibitor cytochalasin D. Interestingly, cytochalasin D reduced IL-1β release by tyrothricin, gramicidin, and neomycin, but not by polymyxin B (Fig. 3C). In addition, the cathepsin B inhibitor CA-074-Me and the antioxidant N-acetylcysteine significantly suppressed tyrothricin- and neomycin-induced IL-1β secretion (Fig. 3D). These findings suggest a contribution of the phagocytosis-cathepsin B–dependent and the reactive oxygen species (ROS)-dependent modes of NLRP3 activation for tyrothricin and neomycin. However, gramicidin and polymyxin B were independent of cathepsin B- and ROS-induced NLRP3 activation. Together, polymyxin B, gramicidin, tyrothricin, and neomycin activate the NLRP3 inflammasome through different pathways, but all were dependent on potassium efflux.
Antibiotics induce peritonitis after i.p. injection
NLRP3 inducers can activate neurophil influx in in vivo peritonitis models (11). Therefore, we tested the potential of the antibiotics to induce neutrophil influx after LPS prestimulation into the peritoneal cavity within 6 h after a single i.p. injection. Polymyxin B, gramicidin, tyrothricin, and neomycin significantly increased neutrophil influx as compared with LPS exposure alone (Supplemental Fig. 2C). Without LPS priming these antibiotics did not induce neutrophil influx (data not shown). Polymyxin B-induced neutrophil influx was substantially reduced in Nlrp3-deficient mice, whereas NLRP3 was not required for the recruitment of neutrophils on injection of gramicidin, tyrothricin, and neomycin (Supplemental Fig. 2C). Polymyxin B, gramicidin, tyrothricin, and neomycin enhance neutrophil recruitment, but this process involves additional pathways beyond NLRP3 signaling.
Our data show that distinct cyclic polypeptide or aminoglycoside antibiotics activate the NLRP3-inflammsome– and caspase-1–mediated release of IL-1β. IL-1β secretion will elicit secondary effects after IL-1R activation, hence these antibiotics have the potential to trigger innate immunity. This is remarkable in three ways. First, these microbial toxins are known to inhibit bacterial growth directly (bactericidal or bacteriostatic effect). As a new entry, they also may suppress bacterial growth indirectly via activating host defense. Second, the immunostimulatory effects might also affect the growth of the antibiotic-producing microorganism itself. Third, when used as an antibiotic drug clinically, the immunostimulatory potential might contribute to drug toxicity. In fact, polymyxin B’s, gramicidin’s, tyrothricin’s, and neomycin’s toxicity profiles do not recommend systemic treatment, and all four drugs are in topical use only (18–20). The cyclic polypeptides and aminoglycosides have different toxicity profiles; therefore, it is unlikely that drug toxicity is solely mediated by IL-1β secretion.
We found all four immunostimulatory antibiotics to activate the NLRP3 inflammasome. This is not surprising because NLRP3 integrates diverse classes of stimuli for IL-1β secretion. It is interesting that all antibiotics require potassium efflux to activate NLRP3, whereas they differ in terms of the involvement of additional ways to activate NLRP3. Tyrothricin, gramicidin, and neomycin, but not polymyxin B, involve phagocytosis like monosodium urate crystals. Furthermore, tyrothricin and neomycin, but not gramicidin or polymyxin B, involve cathepsins and ROS production. The reasons for these diverse mechanisms remain unclear. It is reported that gramicidin dimers can form membrane-spanning pores, which may explain its dependency on potassium flux (19).
Cyclic polypeptide and aminoglycoside antibiotics act as antibacterial toxins; this is why they are in clinical use as antibiotics. Notably, all of these antibiotics cannot be used systemically because of drug toxicity, but the role of IL-1β induction in this context remains speculative. At least i.p. injection caused neutrophil influx, suggesting that these antibiotics can, in fact, elicit proinflammatory effects in vivo. Polymyxin-driven neutrophil recruitment partially depended on NLRP3 also in vivo, whereas this was not the case for the other antibiotics, suggesting that the process of antibiotic drug-induced neutrophil recruitment involves additional mechanisms. In addition, the LPS-binding capacity of polymyxin B is much appreciated by researchers to neutralize LPS contamination, but IL-1β induction might affect the results of such in vitro experiments. Together, our data document that polymyxin B, gramicidin, tyrothricin, and neomycin have the potential to trigger IL-1β secretion by activating the NLRP3 inflammasome through potassium efflux. Thus, these antibiotics add on to the list of microbial toxins that directly activate innate immunity of the host.
We thank Dr. M. Schnurr (University of Munich, Munich, Germany) for providing the P2X7-deficient mice, Dr. V. Vielhauer (University of Munich) for providing TNFR-deficient mice, and Dr. V.M. Dixit (Genentech, San Francisco, CA) for providing the ASC-deficient mice.
This work was supported by the Deutsche Forschungsgemeinschaft (Grant GRK1202).
The online version of this article contains supplemental material.
Abbreviations used in this article:
bone marrow-derived dendritic cell
reactive oxygen species.
The authors have no financial conflicts of interest.