Biogenic ammonia modifies antibiotic resistance at a distance in physically separated bacteria


  • Steve P. Bernier,

    1. Institut Pasteur, Unité de Génétique des Biofilms, CNRS URA 2172, 25 rue du Dr. Roux, 75724 Paris CEDEX 15, France
    Search for more papers by this author
    • Present address: Department of Medicine, Faculty of Health Sciences, McMaster University, Ontario, Canada.

    • These authors contributed equally to this work.

  • Sylvie Létoffé,

    1. Institut Pasteur, Unité de Génétique des Biofilms, CNRS URA 2172, 25 rue du Dr. Roux, 75724 Paris CEDEX 15, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Muriel Delepierre,

    1. Institut Pasteur, Unité de Résonance Magnétique Nucléaire, CNRS URA 2185, 28 rue du Dr. Roux, 75724 Paris CEDEX 15, France
    Search for more papers by this author
  • Jean-Marc Ghigo

    Corresponding author
    1. Institut Pasteur, Unité de Génétique des Biofilms, CNRS URA 2172, 25 rue du Dr. Roux, 75724 Paris CEDEX 15, France
    Search for more papers by this author

E-mail:; Tel. (+33) 1 40 61 34 18; Fax (+33) 1 45 68 80 07.


Bacteria release low-molecular-weight by-products called secondary metabolites, which contribute to bacterial ecology and biology. Whereas volatile compounds constitute a large class of potential infochemicals, their role in bacteria–bacteria interactions remains vastly unexplored. Here we report that exposure to gaseous ammonia released from stationary-phase bacterial cultures modifies the antibiotic resistance spectrum of all tested Gram-negative and Gram-positive bacteria. Using Escherichia coli K12 as a model organism, and increased resistance to tetracycline as the phenotypic read-out, we demonstrate that exposure to ammonia generated by the catabolism of l-aspartate increases the level of intracellular polyamines, in turn leading to modifications in membrane permeability to different antibiotics as well as increased resistance to oxidative stress. We show that the inability to import ammonia via the Amt gas channel or to synthesize polyamines prevent modification in the resistance profile of aerially exposed bacteria. We therefore provide here the first detailed molecular characterization of widespread, long-range chemical interference between physically separated bacteria.


Bacteria produce and sense a previously unsuspected diversity of small molecules used as chemical cues or signals to adapt to changing environments (Straight and Kolter, 2009). Bacterial communication via diffusible molecules has been extensively studied in the last decade, leading to active re-examination of the potential functions of secondary metabolites, a broad class of small molecules long considered as inessential by-products of the main metabolic reactions (Monds and O'Toole, 2008; Surette and Davies, 2008). Secondary metabolites are involved in a wide range of functions, from monitoring of cell density to growth co-ordination, virulence and other unexplored biological processes (Price-Whelan et al., 2006; Monds and O'Toole, 2008; Surette and Davies, 2008).

Recent investigations revealed that a neglected class of secondary metabolites corresponding to volatile compounds of fungi, plants, mammals, insects and nematodes could influence microbial activity and virulence (Cugini et al., 2007; Kai et al., 2009; Minerdi et al., 2009). Volatile compounds of bacterial origin were shown to contribute to cross-kingdom interactions, such as promotion of plant growth and defences or colonization of nematode hosts (Ryu et al., 2003; Farag et al., 2006; Niu et al., 2010).

However, volatile compounds have been studied primarily in a context of inter-kingdom responses and little is known about their potential roles in bacteria-to-bacteria interactions. Several studies suggested that volatile molecules emitted by bacteria could influence bacterial phenotypes such as colony morphogenesis, biofilm and pigment production (Kai et al., 2009; Cepl et al., 2010; Nijland and Burgess, 2010). In addition, an intriguing brief study reported that an unknown volatile compound produced by Escherichia coli could increase resistance to ampicillin and tetracycline antibiotics in physically separated E. coli recipient bacteria (Heal and Parsons, 2002).

Here, we investigated the molecular bases of the latter phenomenon using an original experimental design allowing remote aerial induction of antibiotic resistance. Using this phenotype as a read-out, we show that the implicated volatile compound is gaseous ammonia produced by the catabolism of l-aspartate in E. coli K12. When ammonia diffusing from bacterial cultures reaches a concentration threshold, its uptake via the E. coli Amt gas channel leads to increased levels of intracellular polyamines. The resulting changes in membrane permeability induce modifications of antibiotic sensitivity profiles and resistance to oxidative stress in physically distant bacterial cells. Although the involved molecular mechanisms may differ in other bacteria, we also show that exposure to ammonia produced by all tested microorganisms also induce non-inherited modification of the resistance profile to tetracycline, ampicillin and aminoglycosides in other aerially exposed bacteria, including Pseudomonas aeruginosa, Bacillus subtilis and Staphylococcus aureus. Our report therefore demonstrates that long-range bacteria-to-bacteria chemical interactions mediated by gaseous ammonia extend beyond its sole role as a nitrogen source.


Aerial exposure to E. coli bacterial volatiles alters antibiotic resistance profile in Gram-positive and Gram-negative bacteria

In order to determine the nature and mode of action of previously reported E. coli volatile(s) in the induction of antibiotic resistance (Heal and Parsons, 2002), we developed a 2-Petri-dish experimental design to investigate this phenotype (Fig. 1A). This allowed us to confirm that aerial exposure to volatile molecule(s) released from spent medium of an overnight (O/N) culture in LB permitted growth of E. coli BL21 on inhibitory concentrations of ampicillin and tetracycline (Figs 1B and C and 2A). In contrast, E. coli resistance to non-aminopenicillin β-lactams (ticarcillin), chloramphenicol, ofloxacin and vancomycin was not increased (data not shown). Similar results were obtained with all tested E. coli strains used either as source of spent medium or as exposed bacteria (Table S1).

Figure 1.

Two-Petri-dish assay. Evaluation of volatile-mediated modulation of antibiotic resistance between physically separated bacteria. A. A small lidless Petri dish is placed inside a larger one, closed by its lid. Bacteria spotted on the antibiotic-containing external agar ring are exposed to volatile molecules released from the solution placed in the central small Petri dish. B. Top view, close-up: growth of E. coli BL21 on 1.0 µg ml−1 ampicillin upon exposure to a volatile compound emitted by E. coli K12 MG1655 spent medium filtered (s.m.) from a LB O/N culture, or to sterile LB medium. Growth was monitored after 24 h of incubation at 37°C. C. Top general view (the larger Petri dish lid was removed to take the picture): absence of growth when the middle plate is covered and sealed with parafilm, indicative of a volatile-dependent phenotype (bottom panel).

Figure 2.

Aerial exposure to E. coli spent medium increases resistance to tetracycline in all tested bacteria. Growth of exposed E. coli BL21 (A), P. aeruginosa Lm1 (B), B. subtilis (C) and S. aureus Xen 36 (D) on inhibitory concentrations of tetracycline previously determined in the absence of a volatile source. Source of volatile compound: Exp to E. coli s.m.: Exposition to spent medium from E. coli K-12 MG1655 ΔtnaA grown in LB. Exp ø: Exposition to sterile LB. Pictures were taken after 24 h of aerial exposure to the volatile source in the 2-Petri-dish experimental design presented in the top part of the figure.

We then tested the ability of representative Gram-negative and Gram-positive bacteria, including P. aeruginosa, S. aureus and B. subtilis, to grow in the presence of inhibitory concentrations of tetracycline when exposed to E. coli spent medium as the source of volatile molecules. All three bacterial species displayed increased tetracycline resistance when exposed to E. coli spent medium using the 2-Petri-dish assay (Fig. 2B–D). Finally, we observed that, reciprocally, volatile compound emitted by spent medium extracted from various cultures of Gram-positive and Gram-negative bacteria also promoted growth of E. coli on tetracycline (Table S1).

Identification of the biosynthesis pathway of the E. coli volatile molecule

To identify the volatile compound(s) and/or the biosynthesis pathways responsible for the observed modifications of antibiotic resistance profile, we used the increased tetracycline resistance of E. coli upon exposure to spent media from different E. coli mutant strains as phenotypic read-out. We first established that spent medium from a tnaA mutant unable to synthesize the volatile metabolite indole still triggered aerial antibiotic resistance, therefore eliminating indole as a possible candidate (Fig. 2). Next, we observed that growth phase was critical, since spent media extracted from exponentially or early stationary growing cultures (OD600 < 2) did not promote volatile-dependent antibiotic resistance (data not shown). Finally, whereas spent supernatant originating from E. coli culture grown in various widely used rich media (LB, TSB or BHI) were all active, exposure to spent media from E. coli grown in minimal or rich medium supplemented with glucose did not induce tetracycline resistance at distance (Table S2 and data not shown). These latter observations suggested that biosynthesis of the sought volatile molecule could be regulated by carbon catabolite repression (CCR) (Deutscher, 2008). Consistently, we obtained similar results when we used emitting solutions extracted from media supplemented with carbon sources known to induce CCR (Table S2). Moreover, carbon sources that do not elicit CCR, such as citrate, succinate and pyruvate, had no inhibitory effect on the volatile-dependent antibiotic resistance (Table S2). In support of a role for CCR in the volatile compound synthesis, we observed that spent medium extracted from cultures of E. coli crp or cyaA, two mutants of the CRP–cAMP transcription dual regulator of CCR-sensitive operons, did not aerially induce tetracycline resistance. We then systematically tested spent media from E. coli K-12 mutants corresponding to genes regulated by CCR or contributing to either glycolysis or gluconeogenesis as well as mutants of major E. coli regulators (Gutierrez-Rios et al., 2007) (Table 1). This analysis led to the identification of 13 mutants unable to aerially induce E. coli BL21 growth on tetracycline, all of which corresponded to genes closely associated with CCR and the tricarboxylic acid cycle (Fig. 3 and Table 1).

Table 1.  Complete list of E. coli K-12 mutants tested in this study.
  • a. 

    Activity of volatile molecules emitted by different E. coli K-12 mutants was determined by growth of E. coli BL21 on 1.5 µg ml−1 tetracycline after 24 h of aerial exposure using the 2-Petri-dish assay (see Fig. 1).

  • b. 

    Underlined mutants had no growth defects monitored by OD600 density from O/N culture.

  • c. 

    Perception of volatile compound by different E. coli K-12 mutants was determined by their growth on 3.0 µg ml−1 tetracycline after 24 h of aerial exposure to an O/N culture of E. coli using the 2-Petri-dish assay. Quantification of growth using luminescent recipient strains is indicated in the main text and Experimental procedures.

Mutants lacking volatile activitya,b
cyaA, crp, mdh, gltA, sdhA, sdhB, sdhC, aspA, aspC, lpd, sucA, sucB, purA
Mutants with partial volatile activitya
pta, ackA, acsA/gadC, aldA, aldB, gadA, gadB 
Mutants with wild-type volatile activitya
tnaA, rpoS, luxS, hns, fnr, fis, lro, relA, ihfA, ihfB, soxR, rfaH, lysR, oxyR, cpxR, dam, arcA, flhC, flhD, ntrC, frdA, frdB, frdC, frdD, dmsA, dmsB, dmsC, torA, torC, torD, torY, torZ, ynfE, ynfF, ynfG, ynfH, dmsD, narG, napF, narZ, narY, nrfA, nrfB, aceE, aceF, ubiC,E,F,H, menB, menC, pykA, pykF,ppc, prpC, pckA, maeA, tatA,B,C,D,E, ycfL, ycfR, ybjF, ybeD, yebE, yggN, lsrB, ydcI, yqeC, yoaB, fadJ, yqcC, fimA, recA, pspF, fdhF, lctR, spy, cyoC, sulA, yfhN, yfhO, ygiB, cutC, rseA, rseB, yghO, yhhY, mqo, maeB, ppsA, ldhA, poxB, betA, betB, mlc/dgsA, fabH, fabF, fabR, pflB, tdcE, eutB, eutC, aldH/puuC, puuA, puuE, ydcW, gabD, tynA, paaK, fucO, glcE, yiaW, mhpE, mhpD, alkB, tyrB, acs, acnA, acnB, icd, sucC, sucD, sdhD, fumA, fumB, fumC, aceA, aceB, glcB, pspABCDE, pspG, nadB, allB, allC, gabP, gabT, yneI, putA, astD, maoC, adhE, ygiG/folB, argE, atoA, atoD, actP, nagA, speE, tdcD, prpE, prpD, atoB, ydiF, adhC, eutG, gdhA, glnA pspG, dinI, fadB, gadX, rbsB, sixA, ecO, malM, malG, sodC, yccA, ycfJ, bssS, msrA, glnL, glnG, glnK, amtB, nsrR, scpC, asnA, asnB, iaaA, ansA, ansB, gltB, gltD, gltS, gltJ nirB, nirD, ybaS, yneH, malY, metC, cynT, cynS, dadA, gdhA, glnA, yqhD, pepA, ggt, carB, astB, panD, purC, pyrC, pyrB, pyrL, argG, argB, argA, argD, argH, metL, lysC, thrA, argC, purB, ndK, nrdD, nrdE, nrdF, purH, purE, purK, eutA, ygcM, ybeM, yafV, frmA, fdnG, fdnI, fdoG,H,I, folX, ygfH, ygfA, hcp, yahI, add
Mutants impaired in volatile receptionb,c
speA,B,C,D,E, amtB, gdhA, glnB, glnA, carA, carB
Mutants not impaired in volatile receptionc
argA,B,C,G,H, hisP,M,Q, hns, tynA, feaB, paak, gltB, gltD, glnK
Figure 3.

Biosynthesis pathway leading to volatile-dependent antibiotic resistance in E. coli. A. E. coli BL21 growth on 1.5 µg ml−1 tetracycline when aerially exposed for 24 h to spent medium corresponding to E. coli aspC (aspartate aminotransferase), purA (adenylosuccinate synthetase) or aspA (aspartate ammonia lyase) grown in presence of fumarate (Fum.). WT: wild-type; s.m: spent medium. Pictures were taken after 24 h of aerial exposure to the volatile source in the 2-Petri-dish experimental design presented in the top part of the figure. B. Metabolic pathways showing 13 mutations leading to absence of volatile activity (see Table 1). Of note: since a purB mutation converting adenylosuccinate into AMP has no impact on volatile-dependent antibiotic resistance (see Table 1) our data are consistent with a contribution of PurA to the intracellular pool of aspartate by reverse conversion from adenylosuccinate as suggested as possible by the EcoCyc database (see also Fig. 4). Boxed genes correspond to genes playing a central role in production of ammonia.

Two of these mutants, aspC and purA, displayed no growth defects, but had totally inactive spent medium that did not aerially increase tetracycline resistance of exposed E. coli (Fig. 3A). Both genes are involved in reactions that incorporate or produce aspartate (Fig. 3B), and chemical complementation of aspC and purA mutants by direct addition of aspartate into the LB growth medium restored the volatile activity of extracted spent media (data not shown). This demonstrated the central role of aspartate in the synthesis of the active volatile compound. We also observed that spent culture media of aspC and purA mutants grown in LB supplemented with fumarate were inactive (data not shown).

Conversion of aspartate into fumarate by the AspA aspartate ammonia lyase produces the volatile compound ammonia (NH3) as a by-product (Fig. 3B). Therefore, we tested the activity of an aspA mutant and observed that it did not display any volatile activity, even when fully complemented for growth in LB supplemented with fumarate (Fig. 3A). These genetic results strongly suggest biogenic ammonia as a candidate molecule involved in the investigated volatile-dependent antibiotic resistance phenotype.

Aerial exposure to pure ammonia promotes antibiotic resistance in different bacteria

To determine whether biogenic ammonia was the active volatile compound, we used different dilutions of aqueous ammonium solutions placed in the centre plate of the 2-Petri-dish assay as a source of gaseous ammonia (Fig. 4A). In this experimental design, ammonia is released from the ammonium solution in the gas phase of our 2-Petri-dish assay. After 24 h, we observed that a 5 mM ammonium emitting solution (pH 8.5) promoted growth of exposed E. coli on tetracycline similarly to that induced upon exposure to E. coli wild-type spent medium (Fig. 4A). Increasing the concentration of the emitting ammonium solution to up to 25 mM further improved growth of exposed recipient bacteria on tetracycline in a dose-dependent manner without any apparent bacterial toxicity (Fig. 4B and Fig. S1), which only occurred upon exposure to concentrations of 50 mM and above (Fig. S1). We also showed that exposure to ammonium solutions induced volatile-dependent antibiotic resistance phenotype in P. aeruginosa, S. aureus and B. subtilis (Fig. S2B–D).

Figure 4.

Exposure to volatile ammonia promotes antibiotic resistance in different bacteria. A. E. coli BL21 growth on 1.5 µg ml−1 tetracycline upon exposure to emitting solutions composed either by E. coli spent medium (s.m.) or by increasing concentrations of ammonium solutions releasing volatile ammonia, using the 2-Petri-dish experimental design presented in the top part of the figure. B. Dose–response curve of E. coli BL21 growth on increasing concentration of tetracycline upon exposure to 0–25 mM ammonia-emitting ammonium solutions. C. Experimental design used to compare the amount of ammonium in a water cup exposed for 24 h at 37°C to various emitting solutions. D. Ammonium concentrations in water exposed to spent media of E. coli (E. c), P. aeruginosa (P. a), S. aureus (S. a) and B. subtilis (B. s) grown in LB supplemented or not with 0.4% l-aspartate. E. Ammonium concentrations in water exposed to spent media from O/N cultures of different E. coli K12 mutants. fum: fumarate 0.1% (see text). Reduced bacterial growth correlates with a reduced ammonia concentration in water exposed to spent medium from E. coli cultures taken at an OD600 of 2 compared with OD600 of 4. The data reported represent the mean (n = 3) ± standard error (SD). Statistical analysis: P ≤ 0.0001 (*) by one-tailed unpaired Student's t-test in comparison with medium without aspartate (D) or wild-type (WT) (E).

The gaseous ammonia released by ammonium solutions (NH4+ + OH- inline image NH3 + H2O) was determined according to the relationship pH = pKa + Log [NH3/NH4+] with pKa = 9.25. In order to determine the concentration of ammonia actually present in the gas phase of the 2-Petri-dish assay after 24 h of exposition to either 5 mM ammonium or E. coli spent medium, we first quantified the ammonium concentration in the emitting solution using a highly specific enzyme-based ammonium detection assay (see Experimental procedures). We observed that the concentration of the 5 mM ammonium emitting solution rapidly decreased, but stabilized after 15 h and reached 0.51 ± 0.005 mM (pH 6.5) after 24 h. This concentration is compatible with the concentration measured after 24 h in the emitting solution composed of E. coli spent medium (0.37 ± 0.005 mM; pH 8.2). We then measured the ammonium concentration in a small cup containing 80 µl of pure H2O exposed at the same distance from the emitting source than recipient bacteria spotted in our experimental setting (Fig. 4C). After 24 h, the ammonium concentration of the exposed water originating from dissolution of ammonia present in the gas phase of our assay was 0.58 ± 0.001 mM (pH 6.2) in the case of exposure to the solution initially containing 5 mM ammonium and 0.4 ± 0.001 mM (pH 6.5) upon exposure to E. coli spent medium.

The similarities in ammonium concentrations between the different ammonia-emitting sources and their respective exposed water indicated that equilibrium was rapidly reached in our experimental design. These data allowed us to determine that, after 24 h, the growth-promoting concentration of ammonia present in the gas phase of bacteria exposed to either spent medium or 5 mM ammonium solution was 7.1 × 10−4 mM and 5.16 × 10−4 mM, respectively, according to the relationship [NH3] = 10pH−pKa[NH4+]. Consistently with the activity of spent media extracted from cultures of P. aeruginosa, B. subtilis and S. aureus the ammonium concentrations in water exposed to corresponding spent media displayed similar ammonium concentrations to those exposed to E. coli one (Fig. 4C). Moreover, addition of l-aspartate into growth medium resulted in an increased production of ammonia by E. coli, P. aeruginosa, S. aureus and B. subtilis, therefore indicating that these bacteria have the ability to produce ammonia via l-aspartate catabolism (Fig. 4C).

Finally, we observed that the ammonium concentrations in water cups exposed to inactive spent media extracted either from E. coli aspC, purA or aspA mutants or from low-density exponential-phase wild-type E. coli cultures (OD600 ≤ 2) were significantly reduced compared with those from late stationary growth phase (Fig. 4D).

Ammonia import and assimilation are essential for the volatile-dependent antibiotic resistance phenotype

Our current genetic and biochemical analyses results suggested that the investigated volatile molecule was ammonia produced by the conversion of l-aspartate into fumarate (Fig. 3B). In order to support this hypothesis with additional genetic evidence, we reasoned that volatile-dependent antibiotic resistance phenotype could be reduced in E. coli mutants impaired in ammonia transport and assimilation. Consistently, an E. coli amtB lacking the Amt transporter that is important for growth at low external concentrations of ammonia or ammonium (Fig. 5A) (Soupene et al., 2002; Tremblay and Hallenbeck, 2009) displayed a reduced ability to grow on tetracycline when exposed to E. coli spent medium or ammonium solutions (Fig. 5C and data not shown). Moreover, since transported ammonium is incorporated into glutamate by the glutamate dehydrogenase GdhA (Reitzer, 2003) (Fig. 5A), we tested an E. coli gdhA mutant as recipient bacteria and observed a reduced growth on tetracycline when exposed to active ammonia emitting (Fig. 5C and data not shown). Both amtB and gdhA mutants could be meaningfully tested, since their tetracycline minimal inhibitory concentrations (MICs) were similar to wild-type E. coli (data not shown). Moreover, exposure of E. coli amtB and gdhA mutants to ammonia did not affect their growth on antibiotic-free medium (Fig. 5B and data not shown). This therefore demonstrated that the reduced growth of E. coli amtB and gdhA mutants on tetracycline was not caused by ammonia-dependent toxicity.

Figure 5.

Decreased ammonia import and assimilation impairs E. coli response to volatile ammonia. A. Metabolic pathways involved in assimilation of NH3/NH4+. B. Biomass quantification of ability to grow on antibiotic-free medium upon exposure to spent medium in mutants affected in NH3/NH4+ assimilation using E. coli K-12 luminescent derivatives after 24 h of growth at 37°C in the 2-Petri-dish experimental design. Similar results were obtained upon exposure to ammonia-emitting solutions. C. Biomass quantification of decreased ability to grow on tetracycline upon exposure to spent medium in mutants affected in NH3/NH4+ assimilation using E. coli K-12 luminescent derivatives after 24 h of growth at 37°C in the 2-Petri-dish experimental design presented in the top part of the panel. Similar results were obtained upon exposure to ammonia-emitting solutions. The data reported represent the mean (n = 3) ± SD. Statistical analysis: P ≤ 0.0001 (***) by unpaired Student's t-test in comparison with wild-type (WT).

These results further demonstrated that gaseous ammonia is responsible for the induced metabolic changes affecting antibiotic resistance in distant and physically separated recipient bacteria.

Exposure to ammonia increases the intracellular pool of polyamines

To investigate metabolic changes induced in bacteria aerially exposed to ammonia, we first confirmed that exposure to spent medium did not affect the growth rate of bacteria on tetracycline-free medium (data not shown). While this ruled out simple pleiotropic effects due to assimilation of inorganic nitrogen, it led us to hypothesize that antibiotic resistance could originate from the production of ammonia-derived molecules (Turano and Kramer, 1993). A particular class of intracellular amine compounds present in the millimolar concentration range, polyamines, was previously reported to modulate antibiotic resistance both in Gram-negative and in Gram-positive bacteria (Iyer et al., 2000; Tkachenko et al., 2006; Kwon and Lu, 2007). Interestingly, supplementation of LB agar medium with the polyamines spermidine and putrescine buffered at pH 7.5 allowed E. coli, P. aeruginosa, S. aureus and B. subtilis to grow on tetracycline independently of exposure to biogenic ammonia (Fig. 6A). Thin-layer chromatography analysis of the evolution of the E. coli polyamine content upon exposure to different active or inactive emitting solutions in our 2-Petri-dish assay revealed a significant increase in intracellular spermidine (2.27-fold; P ≤ 0.001) and putrescine (2.02-fold; P ≤ 0.01) in E. coli cells exposed to ammonia or wild-type E. coli spent medium (Fig. S3A). Consistently, E. coli exposure to increasing concentration of ammonia-emitting ammonium solutions led to an increase of both putrescine and spermidine intracellular content (Fig. S3B and C). Moreover, aerial exposition of an E. coli amtB mutant to either bacterial spent medium or ammonium solution did not increase the intracellular level of polyamines (data not shown). Finally, the use of combined multiple mutations of polyamine synthesis genes demonstrated that the reduction or absence of polyamines (Fig. 6B and C) significantly impaired the ability of E. coli to respond to volatile ammonia monitored by growth of luminescence-tagged bacteria on tetracycline (Fig. 6D).

Figure 6.

Exposure to biogenic ammonia increases the intracellular pool of E. coli polyamines. A. Effect of supplementation of LB agar with 2 mM spermidine upon resistance to tetracycline without exposure to spent supernatant. Exp ø: not aerially exposed to biogenic ammonia. Inset in top left-hand corner of each picture represents control plate without addition of buffered spermidine (pH 7.5). Similar results were obtained using buffered putrescine (pH 7.5). B. E. coli polyamine biosynthesis pathway. C. Analysis of polyamine content in E. coli polyamine mutants grown on antibiotic-free agar minimal medium using the 2-Petri-dish experimental design presented in panel D. D. Luminescence-based biomass quantification of the growth of E. coli polyamine mutants on tetracycline LB agar upon exposure to E. coli spent medium. Similar results were obtained upon exposure to a 5 mM ammonia-emitting solution. All mutant strains exhibited wild-type growth and tetracycline MIC in LB medium. The data reported represent the mean (n = 4) ± SD. Statistical analysis: P ≤ 0.01 (*), P ≤ 0.001 (**), P ≤ 0.0001 (***) by one-tailed unpaired Student's t-test in comparison with WT.

Ammonia-dependent increase of polyamine levels alters E. coli membrane permeability to antibiotics and increase resistance to oxidative stress

Polyamines were previously reported to modulate E. coli antibiotic resistance by blocking outer membrane porin channels and decreasing membrane permeability towards tetracycline and ampicillin (Iyer et al., 2000; Tkachenko et al., 2006; Kwon and Lu, 2007). Increased levels of intracellular polyamines also induce sensitivity to aminoglycosides due to higher translation of OppA, a periplasmic binding protein involved in aminoglycosides uptake (Kashiwagi et al., 1998; Chattopadhyay et al., 2003). In addition, polyamines also protect against oxidative stress induced by paraquat, possibly by binding to nucleic acids and protect against damages produced by oxygen radicals or by reducing membrane permeability to paraquat (Santiviago et al., 2002; Chattopadhyay et al., 2003). Consistently with our biochemical and genetic evidences, E. coli exposure to bacterial spent medium or ammonium solution resulted in decreased resistance to kanamycin (Fig. 7A and Fig. S4) as well as in polyamine-dependent increased resistance to oxidative stress induced by paraquat (Fig. 7B and C).

Figure 7.

Polyamine-associated phenotypes in bacteria aerially exposed to ammonia. A. Exposure to ammonia-emitting E. coli K-12 (ΔtnaA) spent medium in the 2-Petri-dish assay decreases resistance to kanamycin in E. coli BL21. B. E. coli K-12 polyamine mutants are more sensitive to oxidative stress induced by paraquat. Exp ø: no exposition to ammonia-emitting solutions. C. Resistance to oxidative stress and growth on increasing concentrations of paraquat in luminescent E. coli K-12 MG1655 exposed either to 5 mM ammonium emitting solution or to H2O. N.S., non-significative. Luminescence-based biomass quantification using the 2-Petri-dish experimental design presented in the top part of the panel. The data reported represent the mean (n = 3) ± SD. Statistical analysis: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.0001 (***) by one-tailed unpaired Student's t-test.

Although the multiple targets of the polyamine modulon prevented us from obtaining a strain totally unable to respond to ammonia, we nevertheless observed that this response was impaired in ompR, a mutant affecting both OmpF and OmpC porin synthesis (Fig. S5A and B). Similarly, we determined that an oppA mutant did not display a wild-type response to aminoglycosides when exposed to volatile ammonia (Fig. S5C and D). Interestingly, while these observations demonstrated that increased level of intracellular polyamines constituted the metabolic response to exposure to biogenic ammonia in E. coli, we also observed a reduced resistance to aminoglycosides using P. aeruginosa, S. aureus and B. subtilis as exposed recipient cells (Fig. S4). Owing to the absence of porins in the tested Gram-positive bacteria, this suggested that other polyamine-dependent transport or binding mechanisms also lead to altered antibiotic permeability in these bacteria.


Although bacteria release a wide diversity of volatile compounds that can diffuse in heterogeneous mixes of solids, liquids and gaseous milieu such as soil, organic tissues and microbial mats, little is known about their biological functions (Kai et al., 2009). Here we show that bacterial uptake and metabolism of biogenic ammonia increases the intracellular level of polyamines and influences antibiotic resistance and oxidative stress responses.

Ammonia production results from the metabolism of peptide and amino acid, the principal carbon sources in many environments, in which l-aspartate generally plays a central role in ammonia production as one of the most highly catabolysable amino acids in most bacteria (Reitzer, 2003; Sezonov et al., 2007; Alteri et al., 2009). Ammonia was previously shown to influence later stages of Dictyostelium discoideum development (Schaap et al., 1995) and to be involved in long-distance co-ordination of yeast growth leading to optimal distribution in their natural habitat (Palkováet al., 1997). While ammonia is an important source of nitrogen, our study demonstrates that ammonia can also mediate long-range bacteria-to-bacteria chemical interactions. Indeed, in addition to requiring specialized assimilation mechanisms (Amt transporter), specific physiological changes induced by non-toxic concentrations of exogenous ammonia extend beyond its mere catabolism (Monds and O'Toole, 2008). However, since it is not yet clear how ammonia sensing can be mutually beneficial for emitter and recipient bacteria, ammonia does not fully comply with the proposed definition of a bona fide intercellular signalling molecule (Monds and O'Toole, 2008). Nevertheless, our study shows that physically separated bacteria can ‘eavesdrop’ on ammonia released by neighbouring microorganisms, as proposed for certain quorum sensing molecules (Shank and Kolter, 2009).

Ammonia-rich environments correspond to high-cell-density bacterial populations developing in closed or semi-open systems such as certain parts of the intestinal tract, soils and microbial biofilm communities. In these complex and structured biological systems, ammonia concentration can be locally high and reached for instance 10–30 mM, depending of available peptide carbon sources (Wrong and Vince, 1984; Hughes et al., 2000).

Above a concentration threshold only reached in dense stationary-phase cultures, ammonia produced by bacteria could constitute a ‘universal’ biotic cue leading to acquisition of competitive capacities in densely populated environments, including modulation of antibiotic and stress resistance (Hibbing et al., 2010). We showed that all tested Gram-negative and Gram-positive bacteria displayed increased resistance to tetracycline upon aerial exposure to ammonia or addition of buffered polyamines. Hence, despite distinct nitrogen metabolism, one can speculate that different molecular events could result in still unknown polyamine-dependent mechanisms changing antibiotic permeability in non-E. coli Gram-negative and Gram-positive bacteria.

Polyamines are known to induce pleiotropic protection against various environmental stresses, and ammonia sensing could therefore improve the fitness of the recipient population, similarly to the induction of defences against predators by plant volatiles or ammonia-dependent co-ordinated stimulation of autophagy in mammalian tissues, leading to increased cell survival (Tabor and Tabor, 1985; Arimura et al., 2000; Rhee et al., 2007; Eng et al., 2010). Indeed, during the final stage of preparation of our study, Nijland and collaborators reported that biogenic ammonia from different bacterial species induced biofilm formation and pigment production of Bacillus licheniformis (Nijland and Burgess, 2010). Although we used a different experimental set-up and read-out, we believe that our study actually determined the nature of the sensing and metabolic responses involved in this process. Interestingly, polyamines were recently demonstrated to be critical for biofilm development of B. subtilis (Burrell et al., 2010). Altogether, these studies suggest that ammonia sensing has several phenotypic read-outs, which is consistent with the pleiotropic role of polyamines (Igarashi and Kashiwagi, 2000; Rhee et al., 2007).

Exposure to biogenic ammonia has potential clinical implications, since increase in MIC in ammonia-rich bacterial infection sites can affect pharmacokinetic and pharmacodynamic parameters during antibiotherapy using aminopenicillin or tetracycline (Craig, 1998). This could constitute one mechanism involved in development of transient and non-inherited antibiotic resistance observed in bacterial communities (Levin and Rozen, 2006). Moreover, even when modest, a local increase in MICs can provide a favourable background, in which selection of additional chromosomally encoded resistance mechanisms can occur (Lee et al., 2010). In conclusion, we provide here the molecular bases for long-range aerial metabolic interference between physically separated bacteria that potentially spans species and kingdoms. Volatile compounds must now be counted among the extraordinarily diverse array of small molecules of ecological relevance, and we anticipate that further studies will identify more bacterial biological processes influenced by biogenic volatile molecules.

Experimental procedures

Bacterial strains and growth conditions

Bacterial strains used in this study are described in Tables S1 and S2. Most E. coli K12 mutant strains originated from the Keio Collection (Baba et al., 2006). λ-red linear DNA gene inactivation method (Derbise et al., 2003) and P1 phage transduction were used to construct all other E. coli mutants. When required, kanamycin resistance markers flanked by two FRT sites were removed using Flp recombinase (Cherepanov and Wackernagel, 1995). All mutations were confirmed by PCR and sequencing analysis. Primers used in this study are listed in Table S3. Antibiotics were used as indicated, or as follows: kanamycin (50 µg ml−1); chloramphenicol (25 µg ml−1); tetracycline (15 µg ml−1). All experiments were performed in lysogeny broth (LB) medium and incubated at 37°C. Ammonia solution (2 M) in 100% ethanol was dissolved in water and aqueous dilutions were used as source of ammonia (Sigma-Aldrich, Reference 392685). All chemicals were purchased from Sigma-Aldrich. Polyamines were prepared by dissolution in sterile distilled water, adjusted to pH 7.5 with HCl and sterilized by filtering through 0.4-µm-pore-size disposable membranes (Millipore).

Screening for volatile-mediated modulation of antibiotic resistance

2-Petri-dish assay.  To evaluate the activity of a volatile compound, an uncovered 3.5 cm Petri dish was placed inside a 9 cm Petri dish. The resulting external ring was filled with 20 ml of antibiotic-containing LB agar (Fig. 1). To determine volatile-dependent induction of antibiotic resistance, 3 ml of either spent medium (adjusted to OD ≥ 3), LB medium, water or NH4+/NH3 emitting solutions were introduced in the middle Petri dish as sources of volatile molecules. Recipient test bacteria were spotted on the external agar ring as 20 µl drops of a 10−5 dilution of an O/N culture adjusted at OD600 of 1, which corresponded to approximately 100 cfu for E. coli, P. aeruginosa and S. aureus but only 20 cfu for B. subtilis. Due to the rapid spread and lawn formation of B. subtilis colonies on plate, this reduced inoculum was used. The large Petri dish was then closed and incubated for 24 h at 37°C. Volatile activity was qualitatively visually estimated as a function of the growth of recipient bacteria on the external ring containing an inhibitory concentration of an antibiotic previously determined in the absence of exposure to a bacterial volatile source. Quantification of bacterial growth upon exposure to volatile compounds was monitored by direct detection of luminescence of bacterial lux-tagged recipient bacteria.

Quantification of ammonium (NH4+) concentrations in solutions

Ammonium present in water exposed to spent supernatants (see experimental design in Fig. 4C) was quantified using an ammonia assay kit (Sigma-Aldrich, Reference AA0100). Briefly, in the presence of α-ketoglutarate and ammonium ions, nicotinamide adenine dinucleotide phosphate (NADPH) is oxidized in NADP upon addition of purified l-glutamate dehydrogenase (GdhA). The decrease in NADPH is monitored by decreased absorbance at 340 nm, which is proportional to the ammonia concentration.

Thin-layer chromatography analyses

Polyamine extraction and analysis were performed as previously described (Tkachenko et al., 2006). Briefly, 5 × 108 bacterial cfu were spotted on antibiotic-free M63B1 glucose minimal medium agar and exposed for 24 h to volatile sources. Bacterial colonies were collected, resuspended in M63B1 glucose minimal medium and centrifuged for 5 min at 14 800 r.p.m. Cell pellets were resuspended in 100 µl of 0.4 N HClO4 for 1 h with shaking. One hundred microlitres of HClO4 extract was centrifuged at 14 800 r.p.m. for 5 min. Fifty microlitres of supernatant was then mixed with 50 µl of saturated Na2CO3 (2 M) and 100 µl of dansyl chloride (1 dimethylamino-1-naphthalene sulphyonyl chloride, Sigma-Aldrich) in acetone (2.7 mg ml−1) and incubated in the dark for 2 h at 37°C. The resulting mixture was evaporated in a flow of cold air (Speed Vac) and extracted with 200 µl of benzene. Benzene extracts were quantitatively placed onto TLC silica gel plates (20 × 20 cm, Merck) for TLC and separated successively into two solvent systems: (i) benzene–triethylamine (20:2) and (ii) benzene–methanol (10:0.45). Fifty microlitres of a 0.2 mM solution of pure putrescine and spermidine were treated with dansyl chloride in the same way as HClO4 prepared cell extracts. Five microlitres of benzene extracts were placed onto a silica gel plate as standard controls. Dried chromatograms were photographed in ultraviolet light, which excites the blue-green luminescence of dansyl polyamine spots. The size and intensity of each fluorescent spot were proportional to the polyamine concentration, which was quantified using Image J software.

Statistical analysis

One-tailed unpaired Student's t-test analyses were performed using Prism 5.0 for Mac OS X (GraphPad Software). Each experiment was performed at least three times.


We thank M. Shumkov for his help with TLC analysis, and thank P. Bouloc for providing some of the mutants used in our study and J. Chamot-Rooke, A. Dulaurent and C. Starkeman for initial volatile compound analyses. We thank C. Beloin, D. Lebeaux, P. Delepelaire, C. Sakamoto, A. Ullmann, G. Cohen, J. Davies, I. Lasa and J. Penades for helpful discussions and critical reading of the manuscript. This work was supported by the Network of Excellence EuroPathoGenomics (LSHB-CT-2005-512061) and the Fondation BNP-PARIBAS (J.-M.G.). S.P.B. was the recipient of a postdoctoral fellowship from the Canadian Louis Pasteur Foundation.