Indole Reverses Intrinsic Antibiotic Resistance by Activating a Novel Dual-Function Importer

Recently, signaling molecules were found to play a role in mediating antibiotic resistance. In this study, we demonstrated that indole reversed the intrinsic antibiotic resistance (IRAR) of multiple bacterial species by promoting the expression of a novel dual-function importer. In addition, population-dependent behavior induced by 13-methyltetradecanoic acid, a quorum sensing signal molecule designated LeDSF, was involved in the IRAR process. This study highlights the dynamic regulation of bacterial antibiotic resistance by small signaling molecules and provides direction for new therapeutic strategies using traditional antibiotics in combination with signaling molecules.

established that efflux pumps encoded by bacterial genes can confer multidrug resistance, and therefore, the structures and working mechanisms of several multidrug efflux pumps in bacteria have been well characterized (3)(4)(5). Meanwhile, the production of specific enzymes results in the inactivation of beta-lactam and aminoglycoside antibiotics by hydrolysis or formation of derivatives (6). Moreover, the formation of bacterial biofilms and their inherent resistance to antibiotics are the root cause of many persistent and chronic bacterial infections (7,8).
Recent studies have shown that signaling molecules could also mediate antibiotic resistance by promoting the expression of specific genes such as those encoding antioxidant enzymes and efflux pumps (10)(11)(12)(13)(14). Here, we focused our study on the interkingdom signal indole that tightens epithelial cell junctions (15,16). Previous studies have shown that indole, a small molecule that is widely expressed throughout the bacterial kingdom, affects bacterial antibiotic tolerance (17)(18)(19). Indole induces the expression of a variety of xenobiotic exporter genes in Escherichia coli (17). Moreover, additional evidence shows that indole reduces persistent formation of E. coli (20)(21)(22). YafQ, a specific endoribonuclease, significantly reduced expression of both RpoS and TnaA, which resulted in reduced levels of indole and an increased number of persister cells (20). It was also demonstrated that halogenated indoles inhibited persister and biofilm formation by E. coli and Staphylococcus aureus (21). Phosphodiesterase DosP decreased the activity of tryptophanase, which converts tryptophan to indole, leading to increased persistence formation (22). The intestinal pathogen Salmonella enterica serovar Typhimurium enhances antibiotic tolerance in response to exogenous indole via a process mediated primarily by the oxidative stress response (18). Previous studies also suggest that S. Typhimurium effectively received an indole signal produced by cocultured E. coli to enhance its antibiotic tolerance in the intestinal environment (19). By repeated transfer of E. coli in the presence of increasing levels of antibiotic, it was found that indole induces population-dependent antibiotic resistance in E. coli, which suggests that bacterial density may also influence changes in antibiotic resistance caused by small molecules (11). As mentioned above, there are plenty of studies on indole enhancing microbial antibiotic resistance. Relatively few studies have been conducted on the mechanism of indole reducing antibiotic resistance.
Lysobacter spp. are common environmental bacteria that have emerged recently as a new source of antibiotics (23)(24)(25)(26)(27). For example, heat-stable antifungal factor (HSAF) and analogs from Lysobacter enzymogenes are a group of polycyclic tetramate macrolactams with potent antifungal activity and a distinct mode of action (28). WAP-8294A from L. enzymogenes OH11 and Lysobacter sp. strain WAP-8294 is a cyclic lipodepsipeptide compound with promising activity against methicillin-resistant Staphylococcus aureus (27,29). Another salient feature of Lysobacter spp. is their intrinsic resistance to multiple antibiotics (30). However, the molecular mechanism underlying this intrinsic antibiotic resistance is not well understood, possibly because Lysobacter spp. produce multiple natural antibiotics. In this study, we describe a phenomenon in which indole reverses the intrinsic antibiotic resistance of Lysobacter spp. (indole reversal of antibiotic resistance [IRAR]) by promoting the expression of a novel dual-function membrane importer.

RESULTS
Indole reverses the intrinsic antibiotic resistance of Lysobacter spp. IRAR was observed in all tested species of the Lysobacter genus. In traditional plating experiments, the addition of 0.5 mM indole rendered Lysobacter spp. sensitive to antibiotic treatment (Fig. 1A). We also monitored the dynamics of bacterial growth under a microscope for bacteria with different treatments. L. enzymogenes YC36 cells were able to elongate and proliferate normally with or without antibiotic treatment. However, when both an antibiotic and indole were added to the culture, cells ceased growth or died from cell lysis (Fig. 1B). For a negative control, we determined whether exogenous indole had any toxic effect on bacterial growth. In the absence of antibiotics, indole alone did not result in any adverse effect on cell  (Fig. 1C), confirming that the combination of indole and antibiotic was responsible for the observed cell death.
Indole upregulated vitamin B 12 gene cluster during the process of IRAR. To explore the biological mechanism underlying IRAR, genome-wide transcriptional profiling of L. enzymogenes YC36 treated with or without exogenous indole was performed. Gene profiling showed that 257 genes were upregulated upon indole treatment, while 111 genes were downregulated (P Ͻ 0.005). A number of regulatory genes were upregulated by indole (see Fig. S1 in the supplemental material), including tetR, which encodes tetracycline resistance repressor protein, and luxR, which encodes HTH-type transcriptional regulator LuxR. The high expression levels of these regulatory genes helped cells to sense exogenous indole signals and regulate the expression of a series of downstream genes. Notably, a cluster of 10 genes, which has been annotated to be associated with synthesis and transport of vitamin B 12 , was significantly upregulated by indole. Expression of the btuD gene, which encodes an ATP-binding protein, was upregulated by 10-fold, and the genes flanking btuD were similarly upregulated to various degrees ( Fig. 1D and E). This 10-gene cluster is common to all species in the Lysobacter genus (Fig. 1F). In this cluster, orf1, orf2, orf6, orf7, orf8, and orf10 encode the outer membrane vitamin B 12 receptor ButB, ABC transporter ATP-binding protein BtuD, a threonine kinase involved in vitamin B 12 biosynthesis, cysteine synthase, carboxylase, and MFS superfamily export protein YceL, respectively. orf3, orf4, orf5, and orf9 encode hypothetical proteins. BtuD possesses a conserved P loop/Walker A, Walker B, ABC signature domain, and a Switch domain (31). The amino acid sequence of BtuD showed less than 35% identity to any known ABC transporter ATP-binding protein, and the best hit was BtuD from Salmonella enterica subsp. enterica serovar Typhimurium (identity of 32.0%) (Fig. S2).
IRAR was facilitated by a novel BtuD-associated dual-function importer that can transfer both vitamin B 12 and antibiotics. To confirm that BtuD is responsible for importing vitamin B 12 , we performed a bioinformatic analysis and biochemical assays. We first tested vitamin B 12 uptake in L. enzymogenes YC36 cells with or without exogenous indole. Vitamin B 12 content was determined by enzyme-linked immunosorbent assay (ELISA) and high-performance liquid chromatography (HPLC). We found that 0.5 mM indole significantly improved the absorption efficiency of vitamin B 12 ( Fig. 2A and Fig. S3). While the ΔbtuD mutant had weak vitamin B 12 absorption efficiency, the efficiency of vitamin B 12 uptake was restored in the btuD complementary ΔbtuD::btuD strain ( Fig. 2A), which demonstrated that the uptake of vitamin B 12 was related to the ButD-associated importer. The addition of indole promoted bacterial growth by increasing the efficiency of vitamin B 12 uptake (Fig. 2B). The ΔbtuD mutant exhibited very slow growth under vitamin B 12 -deficient conditions. Sequence analysis indicated that residues Gly48 and Lys49 presumably make extensive hydrogen bonding contacts with the phosphate groups of ADP in the P-loop domain (Fig. S4). Next, we induced point mutations in these residues. The mutant with G48Y/K49D double-site substitution showed slow growth, especially under vitamin B 12 -deficient conditions. The mutant with K49D single-site substitution possessed decreased vitamin B 12 uptake ability and showed slower growth compared to that of wild-type cells. The growth of the G48Y substitution mutant was unaffected by the tested range of vitamin B 12 conditions (Fig. 2B). Interestingly, the ΔbtuD strain and the mutant with G48Y/K49D double-site substitution did not show IRAR, while IRAR was observed in G48Y and K49D single-site substitution mutants ( Fig. 2C and D). Therefore, we propose that BtuD is a dual-function importer that can transfer both vitamin B 12 and antibiotics. Indole stimulated BtuD overexpression and promoted efficient absorption of external vitamin B 12 ; meanwhile, the weak selectivity of the importer caused cells to take up high doses of antibiotics that resulted in cell death. Consistent with this hypothesis, mass spectrometry showed that indole treatment enhanced the accumulation of antibiotics in cells (Fig. S5). In order to monitor the dynamic entry of antibiotics into cells, we linked the fluorescent probe CFDA-SE (carboxyfluorescein diacetate, succinimidyl ester) to kanamycin to produce a fluorescent antibiotic construct, Kana-CFDA ( Fig. S6A to C). Dynamic imaging of Kana-CFDA-SE showed that antibiotics accumulated in indole-treated cells, but the entry of antibiotics into the ΔbtuD mutant cells was inhibited, which confirmed that BtuD was responsible for cellular uptake of antibiotics ( Fig. 2E and F).
IRAR is common across multiple bacterial species. In subsequent experiments, we found that IRAR is not limited to Lysobacter spp. and is shown by several bacterial species. Pseudoalteromonas is a common pathogenic bacteria and natural product producer that is intrinsically resistant to multiple antibiotics. Our experiments showed that exogenous indole enabled antibiotics to enter Pseudoalteromonas antarctica cells and accumulate efficiently (Fig. 3A). The resistance of Stenotrophomonas maltophilia, a common clinical pathogenic bacterial species, to a variety of antibiotics makes clinical treatment particularly difficult. IRAR was found to greatly improve the therapeutic effects of antibiotics on S. maltophilia (Fig. 3B). Indole also significantly improved the sensitivity of Xanthomonas cucurbitae, a common pathogenic bacterium in agriculture, to traditional antibiotics (Fig. 3B). Bioinformatic analyses showed that the BtuD proteins of different bacterial strains showing IRAR presented certain obviously similar sequence characteristics. A previous study reported that the glutamine (Q) around the Q-loop of BtuD dominates the surface of the protein that interfaces with membrane-embedded BtuC. However, for IRAR strains, glutamic acid (E), rather than Q, is located around the Q-loop area (Fig. 3C). Although the atomic structure of BtuD has not been resolved, we Indole-Induced Antibiotic Resistance Elimination ® speculate that this novel feature of the Q-loops of IRAR species changes the manner in which BtuD and BtuC interact.
LeDSF-induced population-dependent behavior is involved in IRAR. Further investigation revealed that the IRAR phenomenon depended sensitively on bacterial population density. When the Lysobacter species cell density reached a certain threshold (late exponential phase and stationary phase), indole was no longer able to affect the survival state of Lysobacter cells under antibiotic treatment (Fig. 4A). In other words, the intrinsic antibiotic resistance of Lysobacter spp. was restored when the cell density was sufficiently high. Microscopic observation revealed that stationary-phase cells grew and divided normally under treatment with indole and antibiotics (Fig. 4B). However, individual cells isolated from the stationary phase could not survive under the same culture conditions (indole with antibiotics) after gradient dilution to a certain threshold. Growth assays in liquid culture confirmed that L. enzymogenes YC36 could not grow with antibiotics if indole was added at the beginning of cultivation (OD 600 of 0). If indole was added at an OD 600 of 0.4, cells grew slowly, but indole significantly attenuated the survival rate. In contrast, the cultured cells were completely unaffected when indole was supplied at an OD 600 of 0.7 (Fig. 4C).
To understand the cell density dependence of the IRAR process, we carried out bioinformatic analyses. All of the sequenced Lysobacter strains contained a special quorum sensing system induced by LeDSF, a diffusible signaling factor-like molecule (Fig. 4D). In a previous study, the chemical formula of LeDSF was found to be 13-methyltetradecanoic acid (25). The LeDSF biosynthesis gene cluster contains rpfC, rpfG, rpfF, and rpfB. The two-component regulatory system encoded by rpfC and rpfG  Indole-Induced Antibiotic Resistance Elimination ® is responsible for sensing LeDSF and triggering subsequent regulatory processes, whereas rpfF and rpfB encode acyl CoA synthetase and enoyl CoA hydratase, respectively. Coincidently, in the presence of indole and antibiotics, the restored antibiotic resistance of high-density Lysobacter was affected if the LeDSF-associated genes were deleted. The restored antibiotic resistance was completely lost in the Δrpf mutant (Fig. 4E). The ΔrpfC mutant could not survive with antibiotics even when no indole was added, and it could not grow even when supplied with exogenous LeDSF. The survival ability of the ΔrpfF mutant under antibiotic treatment was decreased in comparison with that of the wild-type cells. However, in contrast with the ΔrpfC mutant, supplemental LeDSF fully restored the growth of ΔrpfF cells under antibiotic treatment. The ability to survive under antibiotic treatment was restored in the rpfC and rpfF complementary strains (Fig. 4F). The ΔrpfG and ΔrpfB mutants showed growth similar to that of the wild-type strain, which suggested that the functions of rpfG and rpfB can be replaced by homologs in the genome. Gene transcription analysis showed that btuD was significantly positively regulated by indole during the lag phase and early exponential phase (low cell density, OD of 0.1) in the wild-type strain and the ΔrpfC mutant (Fig. 4G). In the ΔrpfC mutant, btuD expression was increased by 10-fold. However, btuD was not obviously upregulated by indole in the wild-type strain during the late exponential phase and stationary phase (high cell density, OD of 1.0), during which btuD was upregulated by indole in the ΔrpfC mutant (Fig. 4G). On the basis of this evidence, we hypothesize that efficient expression of quorum sensing molecule LeDSF beginning in the late exponential phase inhibited btuD transcription, thereby inhibiting the IRAR process ( Fig. 4H and Fig. S7).

DISCUSSION
Indole signaling is an important means of bacterial communication that has been studied by many research groups. Previous studies reported that indole affected bacterial antibiotic tolerance of E. coli (17,(20)(21)(22). In particular, it has been shown that bacterial communication through indole signaling induces bacterial antibiotic resistance by activating stress responses (18). However, compared with the study of indole-induced antibiotic resistance, the mechanism through which indole reduces antibiotic resistance is largely unknown.
In this study, we describe a novel phenomenon in which indole reverses the intrinsic antibiotic resistance (IRAR) of multiple bacterial species. These species were able to elongate and proliferate normally with antibiotic treatment. However, when both an antibiotic and indole were added to the culture, bacterial cells ceased growth or died from cell lysis. Using L. enzymogenes YC36 as a model system, we reveal that exogenous indole activates a vitamin B 12 importer system and improves the absorption of external nutrients. At the same time, exogenous antibiotics are efficiently pumped into the cells and eventually lead to cell death (Fig. 4H). This process explains the IRAR observed in Lysobacter cells at low cell density. We report for the first time that the vitamin B 12 importer system plays a role in xenobiotic transport. Interestingly, our results show that the BtuD homologs of IRAR strains show similar sequence characteristics; glutamic acid (E), rather than Q, is located around the Q-loop area (Fig. 3C). As we mentioned, BtuD in L. enzymogenes showed less than 35% identity to the well-studied BtuD from Salmonella enterica subsp. enterica serovar Typhimurium. On the basis of the atomic structures of BtuD homologs, we speculate that this novel feature of the Q-loops of IRAR species changes the manner in which BtuD and BtuC interact. It will be interesting to study whether the IRAR process can be abolished if the Q-loop E of IRAR strains is mutated to Q.
When the cell density is high, Lysobacter spp. sense nutrition depletion and therefore secrete LeDSF, a quorum sensing signal, to ensure the survival of the bacterial population. LeDSF effectively suppresses the expression of the vitamin B 12 importer and thereby reduces the uptake of extracellular antibiotics, allowing cells to survive antibiotic treatment. Quorum sensing is a population-dependent mechanism that enables bacteria to communicate with their neighbor cells and to regulate the levels of expression of multiple genes. Early studies revealed that quorum sensing via N-acyl homoserine lactones is closely related to the development of antibiotic resistance and virulence factor production in multiple pathogens (32)(33)(34). However, it was unclear whether quorum sensing via LeDSF was related to antibiotic resistance. In this work, we demonstrate that LeDSF-induced population-dependent behavior is involved in antibiotic resistance. It will be interesting to further characterize whether the IRAR process could also play a role in regulating Lysobacter's population-dependent social activities and its antibiotic production.

MATERIALS AND METHODS
Bacterial strains, plasmids, and general methods. Lysobacter strains and the derived mutants were grown in 40% strength TSB medium. Davis minimal medium without methionine was used for the vitamin B 12 utilization assay (35). The concentration of indole in all experiments was 0.5 mM. The supplemental concentration of LeDSF (13-methyltetradecanoic acid) in the experiments was 5 M. E. coli strains DH5␣ and S17-1 were used for DNA manipulation and conjugation assays, respectively. Additional bacterial strains and plasmids used in this study are described in Table S1 in the supplemental material. Extraction of plasmids and DNA fragments was performed following the instructions included with the kits purchased from Omega (plasmid mini kit I and gel extraction kit, Omega USA). All molecular manipulations were carried out according to methods described previously (30,36). Restriction enzymes and molecular biology reagents were purchased from TaKaRa (TaKaRa Bio Group, Japan). PCR primers were synthesized by Tsingke Biological Technology Company.
Generation of in-frame gene deletion, gene complementary, and site-specific mutants. To construct vectors for in-frame gene deletion in L. enzymogenes YC36, upstream and downstream fragments were amplified using the primer pairs listed in Table S2. Genomic DNA was extracted and used as the PCR template. The upstream and downstream fragments of each gene were cloned into pEX18 to generate in-frame deletion vector pEX18-T. The resulting vectors were transferred into L. enzymogenes YC36 according to a method described previously (37), after which target colonies were selected using PCR verification. The confirmed single-crossover colonies were then subjected to double crossover to produce gene deletion mutants. To construct vectors for site-specific amino acid mutants, fragments containing mutation sites were amplified using the primers listed in Table S2. The procedure was identical to that described above for in-frame gene deletion. Plasmid pHmgA-P was used for the gene complementation assay. The target gene was amplified and linked to pHmgA-P to generate vector pHmgA-P-G. pHmgA-P-G was transferred into L. enzymogenes by conjugation according to a method described previously (36). All of the mutants were verified by PCR and sequencing verification ( Fig. S8A to C).
Bioinformatic analyses. Gene sequences were analyzed by BLAST (http://blast.ncbi.nlm.nih.gov/ Blast.cgi). Annotation and bioinformatic analyses were carried out by genome sequencing and EMBOSS (The European Molecular Biology Open Software Suite) (http://emboss.open-bio.org/). ENDscript 2 software was used to compare BtuD proteins (38). Primers for real-time PCR and gene manipulation assays were designed using Primer Premier 5 (39).
Vitamin B 12 content analysis. Vitamin B 12 content in L. enzymogenes was determined by enzymelinked immunosorbent assay (ELISA) and high-performance liquid chromatography (HPLC). The ELISA experimental procedure was determined according to the instructions of the microbial vitamin B 12 testing kit from Kanglang Biotechnology Company (Shanghai, China). L. enzymogenes was cultured in 40% TSB medium to an OD of 1.0 and transferred to Davis minimal medium with excess vitamin B 12 for 12 h of cultivation. Due to the weak growth of the ΔbtuD mutant strain, multiple ΔbtuD cultures (each with the same volume) were used to ensure a uniform final cell number. The bacterial cells were collected and weighed, and samples of equal weight were used for resuspension and cell fragmentation. The supernatant was removed completely, after which the cell pellet was resuspended with 1 ml ddH 2 O. The content of vitamin B 12 in the cells was calculated by OD 450 . For the HPLC assay, each bacterial strain was cultured in 40% TSB medium to an OD of 1.0 and transferred to Davis minimal medium with excess vitamin B 12 for 12 h of cultivation. Cells were collected, resuspended in 50 ml ethanol, lysed, and dried. The precipitate was resuspended in 5 ml methanol. After centrifugation, a 50-l aliquot of each supernatant was analyzed by HPLC. Pure vitamin B 12 was used as the positive control. Water/0.1% TFA (solvent A) and acetonitrile/0.1% TFA were used as the mobile phases with a flow rate of 1.0 ml/min. The HPLC program was as follows: 5% solvent B at 0 min, increased to 60% solvent B at 10 min, and reduced to 5% B at solvent 11 min. Vitamin B 12 was detected at 359 nm.
Indole-Induced Antibiotic Resistance Elimination ® RNA extraction, reverse transcription-PCR, and real-time PCR. L. enzymogenes YC36 cells were cultured under different conditions, after which RNA was extracted at various time points using an RNA extraction kit (Omega) according to the manufacturer's instructions. After the RNA samples were reverse transcribed to cDNA, real-time PCR was performed in a total reaction mixture volume of 20 l containing 250 nM primers, 10 l of Eva Green 2ϫ qPCR master mix, 8.5 l of RNase-free water, and 0.5 l of 10-fold-diluted cDNA template. 16S rRNA was used as the reference gene. The primers used for qPCR are listed in Table S2. Real-time PCR was performed with a StepOne real-time PCR System (AB Applied Biosystems). The program was designed as described previously (40).
Preparation of fluorescent antibiotic. First, 50 mg of kanamycin (MW ϭ C 18 H 38 N 4 O 15 S ϭ 582.58, 0.0858 mmol) was added to 10 ml of anhydrous DMF, after which 1 ml of triethylamine was added to the reaction system. Magnetic stirring was carried out under nitrogen protection. Next, 47.8 mg of CFDA-SE [5,(6)-carboxyfluorescein diacetate, succinimidyl ester] was dissolved in 5 ml of DMF. CFDA-SE was added to the kanamycin solution for a 3-h reaction. Finally, thin-layer chromatography detection and highperformance liquid chromatography purification were performed. The chemical structure of Kana-CFDA was verified by mass spectrometry.
Cell staining for fluorescence microscopy. For Kana-CFDA (fluorescence antibiotic) staining, cells were collected, washed three times with 40% TSB, and resuspended in 40% TSB buffer. Kana-CFDA was added to a final concentration of 50 g/ml. For the experimental group, 50 g/ml Kana-CFDA was added with 0.5 mM indole. Cells were incubated for 5 h in the dark at 30°C with shaking, followed by observation under a microscope.
Bright-field and fluorescence microscopy. All images were collected on an inverted microscope (Zeiss Observer Z1). Illumination was provided by solid-state laser (Coherent). The fluorescent signal was collected with an EMCCD camera.
Time-lapse recording of bacterial growth under a microscope. We used the FCS2 flow cell system (Bioptechs) to record time-lapse images. Cells were cultured overnight, collected, diluted to a suitable OD value, and washed three times with 40% TSB medium. Next, cells were imaged on a gel pad containing 2% low-melting-temperature agarose. Finally, cells were observed at 30°C under a microscope.
To assess bacterial growth in the presence of antibiotics, cells were resuspended in 40% TSB with ampicillin or kanamycin. To record bacterial antibiotic resistance in the presence of indole, an antibiotic and indole were both added.

ACKNOWLEDGMENTS
We are grateful to Yaoyao Li of Shandong University and Guoliang Qian of Nanjing Agriculture University for supplying bacterial strains. This work was supported by the Marine S&T Fund of Shandong Province for Pilot National Laboratory for Marine Science and Technology (Qingdao) (grant