Bacterial strains and primers used in this study are shown in Supplementary Table S1. P. putida KT2440 and P. aeruginosa strains were grown at 30°C and 37°C in Luria-Bertani (LB) and modified M9 media [Na₂HPO₄7H₂O (6.8 g/l), KH₂PO₄ (3 g/l), NaCl (0.5 g/l), NH₄Cl (1 g/l), MgSO₄ (2 mM), and CaCl₂ (0.1 mM)] (Sambrook et al., 1989) containing 10 mM glucose and 10 mM succinate (Lee Y. et al., 2010) with aeration and shaking. Reagents in media were purchased from Sigma (USA). Growth was monitored by measuring the optical density of cultures at 600 nm (OD₆₀₀) using a biophotometer (Eppendorf, Germany) or by counting CFU.

The following chemicals were purchased from Sigma (USA): indole, phenyl-arginine-beta-naphthylamide (PAβN), indole-acetic acid, tryptophan, tetracycline (Tet), carbenicillin (Car), rifampicin (Rif), chloramphenicol (Chl), norfloxacin (Nor), and kanamycin (Kan). Ticarcillin (Tic) and apramycin (Apr) were purchased from RPI. Ampicillin (Amp) was purchased from AMRESCO. Gentamicin (Gen) was purchase from Gibco. Indole-acetic acid, Amp, Car, Nor, Kan, Tic, Apr and Gen were dissolved in distilled water. Indole, tryptophan, Tet, and Chl were dissolved in ethanol. PAβN was dissolved in 0.5% dimethyl sulfoxide (DMSO), and Rif was dissolved in methanol. Solvent effects can generally be ignored, because they were not significantly affected by changes in the solvent.

The MICs were determined using two fold dilution method (Irith et al., 2008). MICs were defined as the antibiotic concentration that inhibited growth after 24 h of incubation in LB liquid medium at 30°C. Overnight cultures were collected and washed two times with PBS. Approximately 10⁶ CFU/ml cells were inoculated into 96-well microtiter plates containing fresh LB medium and antibiotics. Microtiter plates were incubated for 24 h at 30°C.

Pseudomonas putida KT2440 and P. aeruginosa strains were grown in LB liquid media with shaking at 30°C and 37°C, respectively. The stationary growth phase cells were diluted 100 fold in fresh media and incubated until the cells reach an OD₆₀₀ of the exponential phase (OD₆₀₀ of ~0.4). The exponentially growing cells were harvested by centrifugation and washed twice with phosphate-buffered saline (PBS). Cells were inoculated into PBS at approximately 10⁷ CFU/ml and serially diluted. Each dilution was spotted on an LB agar plate and incubated at the optimal temperature for the specific strain for 24 h. Agar plates were supplemented with indole, antibiotics, or both indole and antibiotics. To test volatile indole-mediated alterations in Amp resistance, indole was provided from an emitting source (25 μ moles on a paper disk) of a small plate placed inside of a square plate separated from the medium.

The primers used in this study are listed in Supplementary Table S1. A 311 bp fragment of the internal region of the ampC gene was amplified using the KT ampC SC-F and KT ampC SC-R primers. The polymerase chain reaction (PCR) product for the ampC mutant was digested with the EcoRI and KpnI restriction enzymes. Fragment was subsequently inserted into a pVIK112 vector via ligation. The constructed plasmids were then transformed into the E. coli S17-1 λ pair. Conjugation was performed using the biparental filter mating method.

β-lactamase activity was measured by hydrolysis of nitrocefin (O’Callaghan et al., 1972; Cavallari et al., 2013; Liu et al., 2016). Cells were grown to the exponential phase at 30°C with aeration. The cells were then treated with or without Amp or indole for 3 h. Following incubation, 5 ml culture was pelleted, washed with PBS (pH 7.0), and resuspended in the same buffer. Samples were placed on ice and then lysed by sonication, centrifuged, and supernatants were collected. The reaction was initialized by adding cell lysate contained β-lactamase to the reaction mixture containing 25 μg nitrocefin (abcam) and PBS (pH 7.0), and the total volume for each reaction was 1 ml. Enzymatic reaction was performed at 30°C for 10 min, and the change at 486 nm was measured over 1 min on a UV/visible spectrophotometer. The extinction coefficient for degraded nitrocefin was 20.5 mM^-1 cm^-1. Enzyme activity was measured in nmol min^-1 mg^-1 nitrocefin hydrolyzed.

Pseudomonas putida KT2440 cells were grown overnight in LB medium and then diluted 100 fold. When the diluted cells reached the exponential phase (OD₆₀₀ ~0.4), the cells were collected and washed two times with PBS. Appropriate dilutions of the cells were spread on LB plates containing 1 mM indole, 50 μg/ml Amp, or both 1 mM indole and 50 μg/ml Amp. After 12 h of incubation at 30°C, cells were collected from the plates, which contained approximately 100 colonies per plate. Total RNA was isolated using the RNAprotect Bacteria Reagent (Qiagen, Valencia, CA, USA) and RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. The following procedure was conducted, as previously described Kim et al. (2013). Genes that showed increases of more than two fold (upregulated genes) or decreases more than 0.5 fold (downregulated genes) were selected. The microarray data were deposited in the National Center for Biotechnology Information (NCBI) GEO site (accession number GSE 86617). cDNA probes for the cDNA microarray analysis were prepared by reverse transcription of total RNA (50 μg) in the presence of aminoallyl-dUTP and 6 μg of random primers (Invitrogen, Carlsbad, CA, USA) for 3 h. The cDNA probes were cleaned using a Microcon YM-30 column (Millipore, Billerica, MA, USA), followed by coupling to the Cy3 dye (for reference) or Cy5 dye (for test samples) (Amersham Biosciences Pharmacia, Amersham, UK). The dried Cy3- or Cy5-labeled cDNA probes were then resuspended in hybridization buffer containing 30% formamide (v/v), 5 × saline-sodium citrate, 0.1% sodium dodecyl sulfate (SDS) (w/v), and 0.1 mg/mL salmon sperm DNA. The Cy3- or Cy5-labeled cDNA probes were mixed and hybridized onto a microarray slide. The hybridization images on the slides were scanned using an Axon 4000B microarray scanner (Axon Instruments, Union City, CA, USA) and analyzed with GenePix Pro software (version 3.0, Axon Instruments) to determine the gene expression ratios (control versus test samples).

Pseudomonas putida KT2440 cells were grown overnight in LB medium and then diluted 100 fold. Appropriate dilutions of exponentially growing cells were spread on LB plates containing indole, Amp, or both indole and Amp. After 36 h of incubation at 30°C, cells were collected from the plates, and total RNA was isolated using an RNeasy Mini Kit according to the manufacturer’s instructions. Northern blot analysis was then performed as described previously (Kim and Park, 2013). Samples of total RNA (2.5 μg) were loaded onto denaturing agarose gels containing 0.25 M formaldehyde, separated, and then stained with ethidium bromide to visualize 23S and 16S rRNA. The fractionated RNA was transferred to nylon membranes (Schleicher and Schuell, Germany) using a TurboBlotter (Schleicher and Schuell, Germany). The amount of mRNA was determined by hybridizing the membrane with a specific ³²P-labeled probe (Takara, Japan), prepared by PCR amplification with the respective primer pairs. Autoradiography was conducted using an IP plate (Fujifilm, Japan) and Multiplex Bio-Imaging System (Fujifilm, Japan).

Overnight cultures were collected and washed two times with PBS. Approximately 10⁵ CFU/ml cells were inoculated into fresh LB medium containing 100 μg/ml Amp and indole (0, 50, 100, 250, 500, and 1000 μM) and incubated for 30 h at 30°C and 200 rpm/min. To verify whether adaptive transitional resistance occurred or not, cells were transferred to fresh media and incubated before determining of the percentage of mutated cells. Cells were collected and washed twice with PBS to remove remaining Amp and indole and approximately 10⁶ CFU/ml cells were inoculated into fresh LB medium and incubated for 24 h at 30°C. Appropriate dilutions of the cells were spread on LB plates containing 50 μg/ml Rif or 200 μg/ml Amp. The total number of CFUs was determined on LB agar plates, with colonies counted after 24 h of incubation at 30°C. To measure the percentage of mutated cells grown on agar plates, 10⁵ CFU were spotted on LB agar plates containing 75 μg/ml Amp. Indole was provided from an emitting source (25 μ moles in paper disk) of a small plate placed inside of a square plate separated from the medium. After 24 h of incubation at 30°C, cells were sampled from each region, based on the distance from the indole source, and incubated in fresh LB medium for 24 h at 30°C. And then, incubated cells were diluted and spread on Amp- or Rif-containing LB plates. The total number of CFUs was determined on LB agar plates, with colonies counted after 24 h of incubation at 30°C. The percentage of mutated cells was determined based on the relative percentage of CFU/ml [(CFUs obtained from the antibiotic plate/total number of CFUs obtained from the LB plate)*100].

Approximately 10⁵ CFU/ml P. putida KT2440 cells grown overnight were inoculated into fresh LB medium containing 100 μg/ml Amp and indole (0, 50, 100, 250, 500, and 1000 μM) and incubated for 8 h at 30°C. Total RNA was isolated using an RNeasy Mini Kit according to the manufacturer’s instructions, and cDNA was synthesized from 1 μg of RNA used as a template with primers for the target gene. The PCR mixture contained 12.5 μl of iQ SYBR Green Supermix (Bio-Rad, Hercules, CA, USA), 1 μl of each primer (0.5 μM) (Supplementary Table S1), and 2 μl of cDNA in a total volume of 25 μl. The PCR conditions were 95°C for 3 min, followed by 40 cycles of 45 s at 95°C, 45 s at 60°C, and 45 s at 72°C. To normalize the expression of each gene, the expression level of 16S rDNA was quantified with primers used previously (Watanabe et al., 2001). The results were determined from experiments performed in triplicate.

Indole has been known to increase AMR by enhancing persister cell formation in E. coli (Lee H.H. et al., 2010; Vega et al., 2012, 2013). To examine whether P. putida KT2440 showed increased survived cells with indole treatments, survived cells were quantified after cells were exposed to a high concentration of Amp (200 μg/ml) in the presence of indole. We did not observe that indole induces resistance in P. putida KT2440 (Supplementary Figure S1). Neither the simultaneous addition of indole and Amp, nor indole pretreatment before exposure to Amp led to increased numbers of survived cells. We determined the minimum inhibitory concentrations (MICs) of various antibiotics in P. putida KT2440 and found that those MICs were identical in the absence or presence of 1 mM indole except for ticarcillin MIC (Supplementary Table S2). However, indole supplied at the beginning of growth increased Amp-resistant cells in LB agar (Figure 1A). We also verified the acquisition of Amp resistance in different minimal media (Figure 1B). Resistance to Amp in the presence of indole also appeared in P. aeruginosa PAO1, Acinetobacter oleivorans DR1, and E. coli O157:H7, regardless of their ability to produce indole (Figure 1C, Supplementary Figure S2A). Indole also increased resistance to tetracycline, kanamycin, other β-lactam antibiotics (ticarcilllin and carbenicillin), and apramycin in LB agar plates (Figure 1D, Supplementary Figure S2B). But, the acquisition of resistance was not observed with rifampicin, gentamicin, chloramphenicol, and norfloxacin resistance in the presence of indole (Figure 1D). Therefore, it seems likely that long-term exposure to indole can enhance AMR in many bacterial species. For a more detailed study of the effects of indole on AMR, Pseudomonas species and the β-lactam antibiotic, Amp, were chosen as a model species and drug, respectively.

To identify genes that are important for Amp resistance in P. putida KT2440 on indole-agar plates, a microarray analysis was conducted. Detailed procedures for the preparation of cells are shown in Supplementary Figure S3. Two genes, trpA and trpB (encoding tryptophan synthase), were highly upregulated by continuously supplied indole, indicating that the tryptophan pathway is induced by indole (Supplementary Figure S4, Supplementary Table S3). Several oxygenases and cytochrome C oxidases required for degradation of indole in non-indole-producing bacteria (Lee and Lee, 2010; Lee et al., 2015b) were highly expressed in the presence of indole (Supplementary Table S4). Ring-cleaving dioxygenase (encoded PP_3328) was increased over 10 fold by indole, indicating that this gene might be important to modifying or degrading indole in P. putida. Genes involved in tryptophan metabolism and oxidation were upregulated by indole, which suggested that indole was degraded under our tested conditions. Genes responsible for the tricarboxylic acid cycle (TCA cycle) and chaperones/proteases that immediately responded to indole during a 10-min treatment in liquid media, as shown in our previous report (Kim et al., 2013), showed mild upregulation with long-term exposure to indole in the transcriptome data, suggesting that the effects of indole were not very different between liquid and plate growth conditions (Supplementary Table S3). Among the genes involved in β-lactam resistance (Nakae et al., 1999; Quale et al., 2006; Kong et al., 2010), genes encoding resistance nodulation cell division (RND) efflux pumps/transporters (ttgABC, acrB2, and acrB3) and bacterial secretion systems, specifically, were increased by indole plus Amp or only indole (Supplementary Tables S5–S7), suggesting that they might contribute primarily to the acquisition of indole-induced Amp resistance. Two genes (PP_1239 and PP_5084) encoding β-lactamase and penicillin-binding protein were upregulated 1.5 fold by indole with Amp; however, most of genes encoding β-lactamases and penicillin binding proteins were not expressed even in Amp alone, probably because the concentration of Amp is a factor controlling the induction of these genes (Supplementary Table S5). Expression of almost 40 genes were increased more than 1.5 fold following treatment for a short time with 1 mM indole in our previous microarray study (Kim et al., 2013). However, 189, 193, and 162 genes were upregulated by indole, Amp, and indole plus Amp, respectively, indicating that both chemicals can cause a broad range of transcriptional responses during long-term exposure of P. putida.

According to our transcriptome analysis and previous reports (Kim et al., 2013), genes encoding efflux pumps and transporters were upregulated by indole (Supplementary Tables S5, S6). To check the importance of RND-type efflux pumps in AMR acquired in response to indole treatment, phenyl-arginine-beta-naphthylamide (PAβN), which is a well-studied broad-spectrum RND-type efflux pump inhibitor (Lamers et al., 2013; Li et al., 2015; Opperman and Nguyen, 2015), was used. The MIC of PAβN was 256 μg/ml in liquid LB media and over 100 μg/ml in LB agar plate, but P. putida KT2440 experienced severe growth defects with over 8 μg/ml of PAβN. A low concentration of PAβN, without any toxic effects, reduced the acquisition of indole-mediated AMR (Figure 2A). Our data demonstrated that indole increased AMR through the action of RND-type efflux pumps in P. putida KT2440. A wide range of β-lactamases in Pseudomonas confer resistance to β-lactam antibiotics (Thomson and Bonomo, 2005; Zeng and Lin, 2013). In our transcriptome analysis, the expression of several genes encoding β-lactamases and penicillin binding proteins were increased by indole (Supplementary Table S5). However, ampC, encoding one of the best characterized β-lactamases that confers ampicillin resistances to many bacteria (Zeng and Lin, 2013), was not induced by any conditions assessed in this study. We checked the expression level of ampC (PP_ 2876) by Northern blot assay and β-lactamase activity with different concentrations of Amp and showed that the expression of ampC increased with long exposure to Amp, and, surprisingly, only the addition of indole enhanced ampC expression and β-lactamase activity in P. putida (Figure 2B). The ampC mutant showed low level of β-lactamase activity in P. putida and increased Amp sensitivity compared to its parental strain in P. putida and P. aeruginosa (Figure 2C). Thus, the expression of several β-lactamases, including ampC, might be essential for AMR induction by indole, along with efflux pumps.

To confirm the contribution of genetic factors identified by our transcriptome analysis to indole-induced AMR, Amp susceptibility tests using P. aeruginosa mutants were performed (Supplementary Figure S5, Supplementary Table S8). Contrary to our expectation, most mutant strains showed increased Amp resistance with exposure to indole. Three genes (PA1413 encoding the LysR family transcriptional regulator, PA4353, and PA4284 exodeoxyribonuclease V beta subunit) are highly expressed by indole plus Amp and conserved among Pseudomonas species. These deletion mutants showed high sensitivity to Amp (Supplementary Figure S6), which indicated that these genes could contribute to Amp resistance in P. aeruginosa. The VgrG protein (encoded by PA1511), which participates in type VI secretion systems (T6SS) in P. aeruginosa, was highly expressed in indole plus Amp conditions (Supplementary Table S7). However, the mutant strain survived in the presence of high concentrations of Amp (Supplementary Figure S7). Pseudomonas strains can carry several copies of vgrG and T6SS clusters in their genomes (Barret et al., 2011), and only three VgrG proteins (encoded by PA0091, PA0095, and PA2685) have been characterized (Hachani et al., 2011). Disruption of the PA1511 gene can result in Amp resistance, providing evidence to clarify the function of novel and uncovered T6SS clusters. Interestingly, three mutant strains, RND transporter mutants (ΔoprM and ΔmexB) and a β-lactamase mutant (ΔampC), lost indole-induced AMR (Figures 2C,D). However, these mutants also acquired indole-induced Amp resistance with an adjustment in the concentration of indole (Supplementary Figures S6, S7). Thus, effective concentrations of indole might contribute to increased AMR in bacteria, even in strains with different susceptibilities. Therefore the acquisition of resistance in response to indole could not fully be explained by the action of efflux pumps or β-lactamase alone in P. aeruginosa PAO1.

Indole is a bacterially produced volatile compound (Kai et al., 2009; Kim and Park, 2013). In order to determine whether volatile indoles can give rise to AMR, we designed experiments in which volatile indole could be supplied from a distance (Figure 3). The indole emission point was separated from the agar media, but located within the upper left corner of a square petri dish (Figure 3A). Indole was slowly released from a paper disk containing 25 μ moles indole, and the same number of P. putida cells (10⁵ CFU per spot) was inoculated at constant intervals. After 24 h of incubation, results showed that cells located closer to the indole-emitting point grew well, even in the presence of Amp at a concentration that prevented the growth of P. putida cells on a control plate (Figure 3A, Plate C). Cells distant from the indole source could also survive, but they had irregular shapes and generated colonies of various sizes (Figure 3A, Plate C). This suggested that heterogeneity might be associated with indole concentration gradients and that cells grown far from the indole source might show phenotypic variation. To verify whether indole presence would cause mutational resistance or temporary adaptive resistance, cells were sampled from each region, transferred to fresh media, and incubated for 24 h before determining of mutated cells. Distant cells survived at high concentrations of Amp (Figure 3B) and showed high percentages of mutated cells compared to cells grown close to the indole source (Figures 3C,D). We verified that cells grown in single spots showed differences in temporal adaptation or mutations depending on the distance from the indole source. Thus, we speculated that cells could develop AMR by mutagenesis in the presence of low indole concentrations through spontaneously formed concentration gradients resulting from the volatility of indole.

Different properties of the cells acquiring Amp resistance in response to volatile indole during plate growth led us to monitor the growth patterns of P. putida KT2440 in various concentrations of indole (0, 50, 100, 250, 500, and 1000 μM) with Amp (0, 50, 75, 100, 150, and 200 μg/ml) in microtiter plates (Supplementary Figure S8). No apparent indole toxicity was observed at ~500 μM without Amp, as shown in our previous report (Kim et al., 2013), and cells failed to grow in 200 μg/ml Amp (data not shown). At 50–1000 μM indole, in addition to Amp, the initial growth rates were greater than that in Amp-only conditions, but cultures started to collapse at 10–15 h (Supplementary Figure S8). Growth was scaled up and monitored by counting CFUs in the presence of indole with 100 μg/ml Amp (Figure 4A). Growth curves based on CFUs of cells grown in indole also contained the part of sticking out from the normal curve during 4–12 h (Figure 4A). This suggested that some cells died, but that surviving cells continued to grow, and after 30 h, the cell densities reached in each condition were nearly identical (Figure 4A). This two-stage growth curve indicated that rapid adaptation or mutation induced by both compounds allowed cultures of P. putida to withstand Amp. In addition, initial growth rates increased with the addition of indole (Figure 4B). At the initial growth stage, cells showed improved growth in the presence of a low rather than a high concentration of indole, which inhibited growth and caused cellular toxic effects in cells by increasing the NADH/NAD⁺ ratio and reducing the ATP concentration because of perturbations in the membrane potential when indole crossed the membrane (Kim et al., 2013). Small amounts of indole can be rapidly consumed with contributing to cell growth, because many non-indole-producing bacteria degrade millimolar concentrations of indole within hours or 10s of hours (Lee et al., 2009, 2015a; Barret et al., 2011; Hachani et al., 2011). Thus, we hypothesized that heterogeneous responses might be induced by indole degradation when cells are exposed to antibiotics.

Heterogeneity was assessed by a susceptibility test (Figure 4C) of cells sampled at each time point from agar plates containing various concentrations of Amp and percentages of mutated cells with Amp and rifampicin (Rif) resistance (Figures 4D,E). To verify whether adaptive transitional resistance occurred or not, cells were transferred to fresh media and incubated for 24 h before determining of the percentage of mutated cells. We confirmed that cells grown in the absence of indole or presence of low concentrations (0, 50, or 100 μM) of indole could become more resistant to Amp because they were able to grow on 200 μg/ml Amp-containing plates (Figure 4C). However, cells incubated in high concentrations of indole (500 and 1000 μM) with Amp failed to grow in 200 μg/ml Amp-containing plates (Figure 4C), even though the number of them were similar with cells grown in other conditions after 30 h of incubation (Figure 4A). We also determined that the percentage of mutated cells decreased as the concentration of indole increased (Figures 4D,E). There was a small or no significant difference in mutated cells grown in the absence and or presence of low concentrations of indole, indicating that the effects of indole in these conditions might be attenuated by metabolism of indole and that mutagenesis induced by Amp could facilitate resistance. Thus, populations grown in the absence of indole or at low concentrations of indole showed more mutated cells (Figures 4C–E). However, the low percentage of mutated cells might be caused by antibiotics being pumped from the inside of cells, which could be essential for the acquisition of resistance, along with the generation of mutations in the presence of high concentrations of indole (~1000 μM) (Figures 4C–E), as confirmed in Figure 2. The percentage of mutated cells in high concentrations of indole (1000 μM) increased 1000 fold in the presence of the RND-type efflux pumps inhibitor PAβN (Figure 4F), suggesting that RND-type efflux pumps, rather than mutagenesis, were the main contributors to resistance when indole concentrations were sufficient to upregulate RND-type efflux pumps. Therefore, indole might induce heterogeneous responses in cells under antibiotic stress.

To identify paths to resistance against Amp, the expression of genes implicated in the tryptophan pathway (trpB), Amp resistance (ampC, ttgA, and oprD), and DNA repair/mutagenesis (rpoS, recA, and mutS) was verified under various concentrations of indole in the presence of Amp (Figure 5). The expression of trpB showed that the tryptophan pathway was induced by indole, regardless of the addition of Amp (Figure 5A). Two genes involved in Amp resistance, ampC (encoding a β-lactamase) and ttgA (encoding an RND transporter), were expressed highly and in an indole-concentration dependent manner (Figures 5B,C). Expression of oprD, which encodes the porin protein (Quale et al., 2006; Kumita et al., 2009), was increased in all treatment conditions relative to the control; however, there were no significant difference based on the presence of indole, Amp, and both (Figure 5D). The high levels of oprD expression encoding a porin that facilitates β-lactam antibiotics entrance into the cell (Thomson and Bonomo, 2005) implied that the effects of Amp could be similar in all conditions, even if the concentrations of indole varied. Thus, porin reduction was unlikely to contribute to the heterogeneity of cells in response to indole. It was speculated that indole might affect mutation evaluated by the expression of rpoS and mutS, which engages in bacterial mutagenesis in the presence of β-lactam antibiotics (Gutierrez et al., 2013), under indole treatments regardless of the concentration (Figures 5E,G). recA gene expression was increased over five fold in the presence of low concentrations of indole, suggesting that DNA repair may be necessary in these conditions (Figure 5F). Our gene expression analysis demonstrated that indole reduces Amp-induced mutagenesis by promoting the expression of TtgAB efflux pumps and an indole-induced β-lactamase in P. putida KT2440, and that indole alone does not alter the expression of many mutagenesis-related genes such as rpoS, recA, and mutS.