Signaling-Mediated Bacterial Persister Formation · Supplementary Methods and Results for: Signaling-Mediated Bacterial Persister Formation . Nicole M. Vega,1 Kyle R. Allison,1 Ahmad

Supplementary Methods and Results for:

Signaling-Mediated Bacterial Persister Formation Nicole M. Vega,1 Kyle R. Allison,1 Ahmad S. Khalil,1 and James J. Collins1,2,3*

1Howard Hughes Medical Institute, Department of Biomedical Engineering, and Center for BioDynamics, Boston University, Boston, MA 02215, USA 2Boston University School of Medicine, 715 Albany Street, Boston, MA 02118, USA 3Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, MA 02118, USA *Corresponding author: James J. Collins; email: [email protected]; phone: (617) 353-0390.

This file includes:

Supplementary Methods

Supplementary Results

Supplementary Figures 1-20

Supplementary Movie Legends 1-3

Supplementary Tables 1-5

Supplementary References

Nature Chemical Biology: doi:10.1038/nchembio.915

1

Supplementary Methods

Bacterial Strains

All experiments were performed in laboratory strains of E. coli. Ancestral wild-type E.

coli K-12 EMG2 (F+) obtained from Yale E. coli Genetic Stock Center (ECGC 4401) was the

reference wild-type strain used in all experiments1,2. Single-gene knockout mutants were

constructed from an E. coli KanR knockout library3 via lambda Red transduction4 into wild-type.

Mutants not available in the knockout library were constructed using the Datsenko-Wanner PCR

products method4. PCR primers for the Datsenko-Wanner method were designed using

Oligocalc5 and obtained from Integrated DNA Technologies. Strains and primers used in this

study are presented in Supplementary Table 3.

The ∆tnaA, ∆mtr, and ∆mtr∆tnaA strains were created to study persistence in indole

metabolism mutants (Supplementary Fig. 1). Knockouts of the AcrEF-TolC system were not

used, as this system’s role in drug efflux6 could confound results in survival assays.

As the ∆oxyR mutant self-aggregates due to de-repression of the flu gene, serial dilutions

with this strain cannot be used to accurately determine cell viability. We therefore constructed a

∆flu∆oxyR mutant, which can be used in serial dilutions47. The PspBC transmembrane proteins,

which are responsible for activation of the phage shock response under most conditions48, were

deleted to prevent activation of the phage shock response via transmembrane signaling.

Antibiotics and Chemicals

The following concentrations of antibiotics were used in this study: 100 µg/mL

ampicillin, 10-60 µg/mL kanamycin, and 1-5 µg/mL ofloxacin. Strains containing kanamycin-

resistance plasmids were grown with 60 µg/mL kanamycin for selection. Otherwise, antibiotic


2

treatments were ≥ 10X MIC to ensure killing of sensitive cells7. All antibiotics were purchased

from Sigma-Aldrich. Indole (≥99%, FCC) and hydrogen peroxide (ACS reagent, stock strength

30%) were purchased from Sigma-Aldrich.

Media and Growth Conditions

All experiments were performed in accordance with standard protocols unless otherwise

stated. Briefly, cultures were grown in light-insulated shakers at 37°C (or 30°C) with shaking at

300 RPM (14 mL Falcon tubes, Fisher Scientific) or 900 RPM (96-well, clear, flat-bottom

culture plates, Fisher Scientific; with Breathe-Easy® adhesive gas-permeable membrane, USA

Scientific). Cultures were grown in tryptophan-free minimal medium (M9 + 0.2% Casamino

acids + 0.4% glucose, M9CG, pH >7.2) to minimize endogenous indole production or in rich

media (Luria-Bertani, LB) to enable production of high concentrations of indole. 500 µM of

indole was added to cultures to approximate physiological levels in the mammalian gut8.

Antibiotic tolerance was assessed by incubating cultures for at least 3 hours with

antibiotic to allow full killing of sensitive cells9,10. For stationary phase cultures, ofloxacin

treatment was used for killing of sensitive cells, as quinolones have bactericidal activity against

stationary phase E. coli11. 10 µL of culture medium was removed before addition of antibiotic

and after 3 hours of antibiotic challenge for serial dilution in 1x PBS, and serial dilutions were

spot-plated onto LB agar (10 µL spots). Serial dilution plates were allowed to grow overnight at

37°C, and colony counts were used to calculate CFU/mL. CFU measurements before and after

antibiotic treatment were compared to determine survival.

Indole-responsive growth assays were performed as follows. Cultures were grown

overnight in LB from glycerol stocks, then inoculated 1:200 in 1 mL M9 + 0.2% glucose and


3

incubated with shaking at 37°C. After two hours incubation, 0 or 50 µM indole was added.

Growth was determined by serial dilution and spot plating.

Minimum Inhibitory Concentration (MIC)

Minimum inhibitory concentration of ofloxacin was determined for all strains used in this

study. Briefly, cultures were inoculated 1:200 in M9CG containing ofloxacin concentrations of

0.025 µg/mL to 0.4 µg/mL (log2 scale) in 96-well plates, covered with a gas-permeable

membrane, and allowed to grow overnight at 37°C. OD600 was measured at inoculation and after

overnight incubation to determine growth.

Survival Assays

Cultures were inoculated 1:200 from overnight LB cultures into 1 mL M9CG in 14-mL

Falcon tubes and incubated with shaking at 37°C. For exponential phase experiments, cultures

were incubated for 4 hours before treatment; for stationary phase experiments, cultures were

incubated for 23 hours before treatment. After the growth period, cultures were treated with

indicated concentrations of indole and allowed to incubate 1 hour before challenge with

antibiotic (5 µg/mL ofloxacin, 25 µg/mL kanamycin, 100 µg/mL ampicillin; ofloxacin treatment

was used in stationary phase cultures because of the antibiotics used in this study, only ofloxacin

has bactericidal activity against stationary phase E. coli9,12).

We evaluated the effects of duration and intensity of antibiotic treatment on the observed

response to indole in stationary phase cultures. Stationary phase cultures were incubated 1 hour

with 500 µM indole before treatment with ofloxacin. Cultures were treated with 5 µg/mL

ofloxacin for 3 or 8 hours to test the effects of treatment duration, and cultures were treated with


4

1, 5 or 10 µg/mL ofloxacin for 8 hours to test the effects of treatment intensity. Survival was

determined by serial dilution plating.

Hydrogen peroxide treatment was performed as follows. Cultures were inoculated 1:200

from overnight LB cultures into 1 mL M9CG in 14-mL Falcon tubes and allowed to grow to

stationary phase. Cultures were incubated 1 hour with the indicated concentration of hydrogen

peroxide before treatment with 5 µg/mL ofloxacin. Serial dilution plating was performed before

and after 1-hour hydrogen peroxide incubation and after 3 hours ofloxacin treatment to

determine survival at each stage.

Complementation of Genetic Knockouts using Rescue Plasmids

Rescue plasmids were constructed using the pZ system13. Purification of restriction

digest products and PCR products was performed using commercially available kits (Qiagen).

Cloning primers contained an ~8-bp GC-rich leader sequence, a restriction site for plasmid

insertion, and a ~26-bp homology sequence overlying the START or STOP codon of the gene of

interest (Supplementary Table 3). Genes for complementation were cloned from E. coli EMG2

using hot-start Phusion PCR (Sigma-Aldrich). PCR products and the parent plasmid pZA21-

GFP were digested using KpnI/BamIII and ligated using T4 DNA ligase (Sigma-Aldrich),

replacing the GFP cassette with the gene of interest.

Knockout strains were transfected with the spectinomycin-resistant Pro cassette to enable

activation from the PLtet0-1 promoter. Strains were then made electrocompetent by washing twice

in ice-cold sterile water and twice in ice-cold sterile 10% glycerol. Electrocompetent strains

were incubated 20 minutes with plasmid in cold 10% glycerol before electroporation at 2400V.

Plasmid-transformed cells were inoculated in SOC media and incubated 1 hour at 37°C to allow


5

expression of plasmid-based resistance genes before addition of the appropriate antibiotic for

selection.

In all experiments, cultures were grown in media containing antibiotics for plasmid

selection (35 µg/mL kanamycin), and 25-50 ng/mL anhydrotetracycline (aTc) was added after 4

hours growth for induction of plasmid-borne genes.

Flow Cytometry and Fluorescence Activated Cell Sorting (FACS)

E. coli K-12 bearing the plasmid pUA66-ptnaC-gfpmut2 (PtnaC) from a GFP reporter

plasmid library14 were inoculated 1:200 from overnight LB cultures into 1 mL selective media

(M9CG + 60 µg/mL kanamycin) and allowed to grow 4 hours before treatment with 500 µM

indole. Cultures were returned to shaker for 1 hour before sampling. Sample aliquots were

diluted ~1:1000 in 1X PBS before flow cytometry.

Data collection and fluorescence-based sorting were performed using a BD Biosciences

FACSAria system and BD FACSDiva software. BD cytometer setup and tracking beads were

used for instrument calibration, and BD Accudrop beads were used for sort calibration.

Fluorescence data were collected at 530 nm emission (FL1) using a 488 nm argon laser

for excitation. Data were recorded for least 10,000 cells, and 500,000 cells were collected in

each bin during sorting experiments. Data were processed using MATLAB (Mathworks, Natick,

MA).

Microfluidic Device Design and Construction

Our microfluidic devices were designed to enable single-cell observation of E. coli

microcolonies and automated control of on-chip, environmental liquid switching. The design of


6

the cell trapping scheme used in these microfluidic devices was adapted from the previously

described Tesla microchemostat15,16, in which cell trapping chambers were constructed with

heights that match E. coli cylindrical diameters to support the growth of single monolayers. To

this core design, we overlaid a control layer of integrated elastomeric valves to allow for on-chip

switching of liquid inputs and outputs (Supplementary Fig. 14a-b).

Multilayer microfluidic devices were fabricated out of the silicone elastomer

polydimethylsiloxane (PDMS/Sylgard 184, Dow Corning) using multilayer soft lithographic

techniques, as described previously17-19. Briefly, two photoresist-based molds were used to

fabricate the multilayer device, corresponding to flow and control layers patterned with

respective microchannel structures. The flow layer mold was constructed by first patterning SU-

8 2 negative photoresist (MicroChem Corp.) at the appropriate feature height onto a silicon wafer

and transferring the cell trapping chamber pattern from a high-resolution transparency

photomask (CAD/Art Services, Inc.). Next, AZ4620 positive photoresist (Capitol Scientific,

Inc.) was patterned at the larger flow channel feature height and aligned to the trapping chamber

pattern before transferring. The completed flow layer mold was placed on a hotplate at 145ºC

for 1 minute to reflow the photoresist and round the channel profiles for complete valve closure

in assembled devices. The control layer mold was constructed by patterning SU-8 10 negative

photoresist (MicroChem Corp.) onto a second silicon wafer and similarly transferring the control

layer pattern from a photomask.

Devices were fabricated via replica molding from the microfabricated master molds.

PDMS/Sylgard 184 was mixed in a 10:1 ratio of elastomer base:curing agent, poured on the

control layer mold to a thickness of ~5 mm, and baked at 80ºC for 2 hours. Elastomer was

similarly prepared, spun onto the flow layer master (3500 rpm for 60 s), and baked at 80ºC for


7

~10 min. The cured control layer was peeled from the master and aligned over the flow mold

under a microscope. The multilayer assemblies were then baked for an additional 2 hours,

peeled from the master, cleaned, and finally sealed to pre-cleaned No. 1 glass coverslips (Fisher

Scientific) to realize monolithic devices.

Flow Field Simulation

Flow fields were simulated using commercial finite element analysis software (COMSOL

4.0, MA) (Supplementary Fig. 14c). In these simulations, only the cell chamber and adjacent

flow channel segments were considered to simplify the computation (the hydraulic inlet pressure

for the model was estimated based on relative fluidic resistances). It was assumed that medium

flow was steady and incompressible (with density 1000 kg/m3 and dynamic velocity 0.001 Pa·s).

The velocity profile along the flow channel (Reynolds number << 1) becomes fully developed

within ~100 µm of channel length, indicating that the channel segment in this model is

sufficiently long for accurate computation of developed flow profile within the cell chamber.

Microfluidic Device Operation

Integrated devices were operated by a custom-built microfluidic platform that combines

plumbing (i.e., fluidic tubing and manifolds), hardware, and software for the delivery of liquids

to the device and for off-chip control of the elastomeric valves. Briefly, the control layer of the

final assembled device was pressurized to and maintained at 15 psig in order to deflect the

PDMS membrane downward into the flow channel and thus prevent fluid from passing. Off-

chip control of these elastomeric valves, which are pressurized by a nitrogen gas source, was

carried out by miniature solenoid check valves (The Lee Co.) that are driven by an analog-to-


8

digital logic board (National Instruments PCI-DIO-32HS) and an open-source JAVA operating

software20. The JAVA library and interface abstracts the underlying device architecture, and

thus allows complex device operations to be implemented and automated with simple logical

commands20.

Microfluidic Single-Cell Indole Response

In each experiment, E. coli bearing the plasmid pUA66-ptnaC-gfpmut2 from a GFP

reporter plasmid library14 were inoculated 1:200 from overnight LB cultures into 1 mL selective

media (M9CG + 60 µg/mL kanamycin) and allowed to grow 2-3 hours (OD ~0.3) before seeding

the microfluidic device. Meanwhile, the microfluidic device was mounted to the microscope

stage, elastomeric valves were pressurized and checked, and flow channels were wetted with

PBST (PBS + 0.05% Tween 20). Sample inputs, corresponding to (1) selective media, (2)

selective media + 500 µM indole, and (3) selective media + 500 µM indole + 100 µg/mL

ampicillin (Supplementary Fig. 14a), were connected to the device through Tygon tubing (500

µm id) and primed. Cells were then loaded into the device from the cell input port, and upon

trapping a small number of cells in the trapping chamber, flow was reversed to perfuse sample

inputs through the chamber and support growth of several microcolonies. Cells received growth

media (input (1)) for > 2 hours so that the total number of analyzed cells exceeded 50, then were

treated with indole (input (2)) for 1 hour, followed by a 1-hour ampicillin treatment (input (3)),

and finally an outgrowth with input (1) for ≥ 2 hours. Conditions throughout the experiment

were maintained at 37ºC and 30% humidity with a LiveCell environmental chamber (Pathology

Devices, Inc.).


9

Timelapse Microscopy and Image Analysis

During growth and treatment of the monolayer colonies, images were collected at 100x

magnification (Plan Apo VC 100x) using an Eclipse Ti epifluorescence inverted microscope

(Nikon Instruments, Inc.) equipped with the Perfect Focus System, a ProScan II XYZ-motorized

stage (Prior Scientific, Inc.), and a CoolSNAP HQ2 charge-coupled device (CCD) camera

(Photometrics), and controlled by the NIS-Elements Advanced Research software (Nikon

Instruments, Inc.). Images were acquired in the differential interference contrast (DIC)

configuration and in the GFP fluorescent channel, using a motorized fluorescence excitation and

emission filter turret with a GFP filter set (Chroma Technology Corp.), a Lambda SC

SmartShutter (Sutter Instrument), and an Intensilight C-HGFIE mercury fluorescent lamp (Nikon

Instruments, Inc.). Specifically, we acquired DIC images at 3-minute intervals and GFP images

at 6-minute intervals while perfusing inputs (1) and (2), and increased the DIC image acquisition

frequency to 30-second intervals during ampicillin treatment.

Following each experiment (N = 3 replicates), the time series of DIC images was

manually analyzed to identify the surviving cells from the population (i.e., cells that exhibited

regrowth following ampicillin treatment). Single-cell fluorescence data were extracted from the

time series of GFP images using the CellTracer21 image analysis platform written for Matlab

(Mathworks, Natick, MA). Mean fluorescent values for each segmented cell were subtracted

from background, and maximum values across the time series were normalized with respect to

the lowest fluorescing cell in the analyzed population and grouped into surviving or dying sub-

populations.


10

High Performance Liquid Chromatography (HPLC)

Indole quantification was performed via HPLC using a protocol based on Lee et al.22,

Kobayashi et al.23 and Hirakawa et al.24. Cultures were inoculated 1:200 from overnight LB

cultures into 1 mL of LB or M9CG, and incubated 5 hours (exponential phase) or 24 hours

(stationary phase) before harvesting. Cell-free medium for quantification of extracellular indole

was obtained by centrifugation of whole culture (8K RPM, tabletop centrifuge) and purification

of the resulting supernatant on Spin-X filter columns. HPLC was performed using a Waters C18

(5 µm, 4.6 x 150) column with H20 + 0.1% formic acid/acetonitrile (1:1) as the isocratic mobile

phase. Under these conditions, in M9 + 0.2% Casamino acids + 0.4% glucose, indole eluted at

approximately 2.9 minutes as a single peak at 271 nm. In LB, indole eluted at approximately 4

minutes as a single peak at 271 nm. A series of indole standards (M9CG:5100 µM, LB:

50500 µM) in sterile growth medium was used to estimate indole concentration as a linear

function of peak height and peak area. Each culture-time point combination was analyzed using

three (exponential phase) or four (stationary phase) biological replicates; estimates for indole

concentration were generated using both linear equations. Error bars represent mean ± SD of all

estimated values. All HPLC experiments were performed in the BU Chemical Instrumentation

Center (Agilent 1090 HPLC system, ChemStation software).

We used high-performance liquid chromatography (HPLC) to quantify indole, as this

method is selective, sensitive and precise25 and is well-established in the indole literature6,22-

24,26,27. We measured indole concentration in the extracellular space, rather than attempt to

measure indole associated with the cell pellet, as distinguishing between indole in the cytoplasm,

indole in the periplasm, and indole that might stick to the outer membrane is difficult using

HPLC. There are other methods to measure indole concentration, such as intracellular


11

measurements using styrene monooxygenase28,29, which hold promise for future understanding

of indole signaling.

Gene Expression Analysis

RNA was collected from indole-treated and untreated cultures of E. coli K-12. Cultures

were inoculated 1:200 from overnight LB cultures in 1 mL of M9CG in 14 mL Falcon tubes and

incubated 4 hours (exponential phase) or 23 hours (stationary phase) before treatment with

indole. After 30 minutes incubation with 0 or 500 µM indole, cultures were stabilized with

RNAprotect Bacteria Reagent (Qiagen) according to the manufacturer’s protocol, and the

resulting cell pellets were stored overnight at -80°C. RNA extraction was performed using

RNeasy Mini Kit (Qiagen) and DNA contamination was eliminated using TURBO DNA-Free

(Ambion, Austen, TX) according to the manufacturers’ protocols. Sample concentration was

estimated using the ND-1000 NanoDrop spectrophotometer. Standard PCR using Taq

polymerase and qPCR primers (see below for primer information) was used to test for DNA

contamination in RNA samples after TURBO DNA-Free digestion (4% RNA suspension by

volume, 30-35 cycles). RNA was stored at -80°C.

cDNA for qPCR was synthesized from RNA using the Superscript III First Strand

Synthesis kit (Invitrogen) and stored at -20°C. rrsC and rrsH were used as reference genes.

Quantitative PCR primers for each transcript of interest and the reference transcripts were

designed based on the NCBI E. coli K-12 MG1655 genomic sequence (Refseq NC_000913)

using Primer3Plus software30 (Supplementary Table 3). Primers were designed under the

following constraints: amplicon size was 130-160 bp, the calculated primer melting temperature

was ~55˚C, GC content was 45-55%, and probabilities of primer-dimer/hairpin formations were


12

minimized. Primer specificity was confirmed via standard Taq polymerase PCR on genomic

DNA.

qPCR reactions using LightCycler 480 SYBR Green I Master Kit (Roche Applied

Science) were prepared manually according to the manufacturer’s instructions. qPCR reactions

were performed with a total volume of 20 µl, containing 0.5 µM of forward primers and 0.5 µM

of reverse primers and 10 µl 2X 480 SYBR Green Master Mix. Reactions were carried out in

white LightCycler 480 96-well plates (Roche). One negative control (replacing cDNA with PCR

H2O) was performed for each primer set in all qPCR runs. PCR parameters were: denaturation

(95˚C for 10 minutes), 30-35 cycles of three-segment amplification (95˚C for 10 seconds, 50˚C

for 10 seconds, 72˚C for 10 seconds). The thermal cycling program was concluded with a

dissociation curve (65˚C ramped to 95˚C, 10 seconds at each 1˚C interval) to detect non-specific

amplification or primer-dimer formation. Cp values were obtained from the Roche LightCycler

480 software set with default parameters for basic relative quantification analysis. Data were

evaluated for consistency using previously described metrics31, and results were recorded as

magnitude of the change in expression after normalization, using reference gene expression as

detailed previously32. qPCR was performed using 3-5 biological replicates for each growth

phase/treatment combination reported; technical duplicates were performed on separate days to

assess variability.

Gene expression profiles were determined from mRNA using Affymetrix E. coli

Antisense 2 genome arrays. The resulting microarray *.CEL files were combined with *.CEL

files from Affymetrix E. coli arrays archived in the Many Microbe Microarrays (M3D) database

and arrays previously generated by our laboratory (http://m3d.bu.edu/; Ecoli_b17) 33, and RMA

normalized with RMAExpress, for a total of 809 RMA-normalized E. coli arrays. Standard


deviation of expression (σ) was calculated from the arrays using wild-type strains (E. coli K-12

MG1655 and EMG2); separate calculations were made for exponential phase (462 arrays) and

stationary phase (27 arrays). Z-scale differences for each gene were calculated based on the

difference between normalized expression under an experimental condition (indole treatment)

and a control (no treatment):

σcontrolXX

Z−

=Δ expexp

Under each experimental condition (exponential phase or stationary phase), genes

showing significant differences in gene expression (|z| ≥ 2, p ≤ 0.05) were identified as being

significantly affected by indole treatment34. Analysis of normalized data and creation of heat-

map plots were performed in MATLAB. Visualization of gene expression data was performed

in Cytoscape35.

Hypergeometric test-based Gene Ontology enrichment analysis was performed on sets of

genes for which expression was altered at least a factor of 1.8 by indole treatment in exponential

phase and stationary phase, respectively. These analyses were based on Gene Ontology (GO)

biological process category annotations obtained from EBI

(http://www.ebi.ac.uk/GOA/proteomes/18.E_coli_K-12.goa). Genes lacking functional

annotations were excluded. These analyses were performed in R36 using GOstats in

Bioconductor37,38.

13


Supplementary Results

Supplementary Figure 1. Metabolism of the stationary phase signal molecule indole in E. coli. The tryptophanase enzyme TnaA produces indole during catabolism of tryptophan, and the reverse reaction is highly unfavorable39. Indole is exported from the cell in part by the AcrEF-TolC complex6, and the Mtr transporter imports indole from the periplasm40. The tnaA gene is in an operon also encoding leader peptide TnaC and tryptophan permease TnaB; expression is induced under nutrient stress41 (controlled by Crp and cAMP) and alkaline stress42 (controlled by TorR), and the TnaC leader peptide enforces Rho-dependent termination in the absence of tryptophan43.

14


Supplementary Figure 2. Indole increases multi-drug tolerance in exponential phase. Percent survival of exponential-phase wild-type E. coli pre-incubated for 1 hour with and without 500 μM indole and treated for 3 hours with 5 µg/mL ofloxacin, 100 μg/mL ampicillin, or 10 μg/mL kanamycin.

15


Supplementary Figure 3. Indole-induced tolerance does not depend on duration of antibiotic treatment. Stationary phase cultures of wild-type E. coli grown in M9CG were pre-incubated for 1 hour with (purple bars) and without (grey bars) 500 µM indole prior to treatment with 5 µg/mL ofloxacin. Survival was determined after 3 or 8 hours ofloxacin treatment. Error bars represent mean ± SD of six biological replicates.

16


Supplementary Figure 4. Indole-induced tolerance does not depend on antibiotic dose. Stationary phase cultures of wild-type E. coli grown in M9CG were pre-incubated for 1 hour with (purple line) and without (grey line) 500 µM indole prior to treatment with 1, 5, or 10 µg/mL ofloxacin. Survival was determined after 8 hours ofloxacin treatment. Error bars represent mean ± SD of three to six biological replicates.

17


Supplementary Figure 5. Extracellular indole in cell-free stationary phase supernatant of E. coli grown in minimal media. Concentration of extracellular indole was determined by HPLC for stationary phase cultures of wild-type E. coli (grey bars) and indole import-deficient mutant Δmtr (green bars) grown in tryptophan-free medium (M9CG). No indole was detected in cultures of ΔtnaA E. coli. Error bars represent mean ± SD of three biological replicates.

18


Supplementary Figure 6. Extracellular indole in cell-free supernatant of E. coli grown in rich media. Concentration of extracellular indole was determined by HPLC for exponential phase and stationary phase cultures of wild-type E. coli (grey bars) and indole import-deficient mutant Δmtr (green bars) grown in rich medium (LB). No indole was detected in cultures of ΔtnaA E. coli. Error bars represent mean ± SD of three biological replicates.

19


Supplementary Figure 7. Complementation of knockout mutants with rescue plasmids. Cultures of wild-type, ∆tnaA, and ∆mtr E. coli bearing no plasmid (grey bars), the kanamycin-resistant plasmid pZA21-GFP (pGFP, green bars), or a plasmid carrying an inducible copy of the missing gene (p(Rescue); pZA21-tnaA or pZA21-mtr, purple bars) were grown to stationary phase (24 hours) in (a,c) M9CG or (b,d) LB media containing 35 µg/mL kanamycin for selection and 25 ng/mL anhydrotetracycline for induction of plasmid-borne genes. Cultures were incubated with 0 (a-b) or 500 µM (c-d) indole for one hour before treatment with ofloxacin. Error bars represent mean ± SD of at least three biological replicates.

20


Supplementary Figure 8. Indole-induced tolerance is affected by temperature. Cultures of wild-type and ∆tnaA E. coli were grown to stationary phase in LB at 30°C (12 hours) or 37°C (24 hours) before treatment with ofloxacin. Survival was determined after 4-6 hours (37°C) or 20-24 hours (30°C) ofloxacin treatment to allow full killing of susceptible cells; ranges represent day-to-day variation, and indole-treated and untreated within-strain pairs were performed on the same day. Error bars represent mean ± SD of at least three biological replicates.

21


Supplementary Figure 9. Mtr is required for high-efficiency transport of indole. All experiments were performed in M9 + 0.2% glucose at 37°C. (a) Cultures of ∆trpA and ∆trpA∆mtr E. coli were inoculated 1:500 in M9 + glucose and incubated for two hours before addition of 0 or 50 µM indole. Error bars represent mean ± SD of at least three biological replicates. (b) Cultures of ∆mtr E. coli were inoculated 1:500 in M9 + glucose and incubated for two hours before addition of 0, 50, or 500 µM indole. Error bars represent mean ± SD of at least three biological replicates.

22


Supplementary Figure 10. Long-term incubation with indole increases persistence in E. coli. Cultures of wild-type, ∆tnaA, and ∆mtr E. coli were grown for 4 hours in M9CG before incubation with (light purple bars) or without (grey bars) 15 µM indole. Cultures were incubated with indole for 20 hours before treatment with ofloxacin. Error bars represent mean ± SD of at least three biological replicates.

23


Supplementary Figure 11. Elimination of indole production in a Δmtr mutant abolishes the high-persistence phenotype. Cultures of wild-type, Δmtr, ΔtnaA, and ΔtnaAΔmtr E. coli were grown to stationary phase in M9CG or LB before treatment with ofloxacin. Error bars represent mean ± SD of at least six biological replicates.

24


Supplementary Figure 12. Indole-induced persistence is dose-dependent and saturates at high concentrations of indole. Black line indicates percent survival of cultures after incubation for 1 hour with indole. Grey bars indicate percent survival of the same cultures after subsequent treatment with ofloxacin, relative to survival after incubation with indole.

25


26


27

Supplementary Figure 13. Indole signaling induces transcription from the native tnaC promoter and correlates with increased tolerance on the single-cell level. (a) Wild-type, (b) ∆tnaA, and (c) ∆mtr E. coli bearing the pUA66-ptnaC-gfpmut2 (PtnaC) fluorescent reporter plasmid were grown to exponential phase in selective media (M9CG + 60 µg/mL kanamycin) and incubated for 1 hour with (red line) and without (blue line) 500 µM indole before fluorescence quantification via flow cytometry. (d) Untreated wild-type and (e) indole-treated ∆tnaA exponential phase E. coli containing PtnaC were sorted into sub-populations based on relative fluorescence (high, top 10% of population; low, bottom 10%) and resuspended in 1 mL 1X PBS; survival was determined after 0, 30, and 60 minutes treatment with 1 µg/mL ofloxacin. Wild-type E. coli bearing the lac reporter plasmid pZA12G (f), which does not respond to indole, was used as a control.


28


29

Supplementary Figure 14. Multilayer microfluidic chemostat device. (a) Schematic of the full microfluidic device. The flow layers are shown as grey channels. The control layer, consisting of integrated elastomeric valves for on-chip pneumatic valving of the flow channels, is shown as red channels. W1 and W2 denote waste ports. Cells loaded from port C are directed toward the cell trapping chamber (light grey channels; boxed (b)), which has been fabricated to the height of a single monolayer of E. coli. Once cells are seeded, flow is reversed to perfuse one of three liquid inputs (denoted 1, 2, and 3) that can be toggled ON and OFF in automated fashion. (c) Simulation of flow velocity fields within the cell chamber and adjacent flow channel segments for an input pressure of 1 psig. Flow fields were simulated using commercial finite element analysis software (COMSOL 4.0, MA).


Supplementary Figure 15. Representative kill curve for microfluidic-based ampicillin treatment. Time 0 represents addition of antibiotic. Data were obtained from a single experiment in the microfluidic chemostat.

30


31


32

Supplementary Figure 16. Microarray results show up-regulation of phage shock and OxyR regulons following indole treatment in stationary phase cultures. (a) Z-score ranking of significantly up- and down-regulated genes in indole-treated cultures (|z| ≥ 2, p ≤ 0.05). Up-regulated genes are shown in red and down-regulated genes are shown in blue. Significance was determined by calculating z-scores for gene expression after indole treatment as compared to an untreated baseline; variance was estimated from microarray compendium data (see Supplementary Methods). Fold-change expression of genes in the (b) phage shock regulon and (c) OxyR regulon after indole treatment.


33


34

Supplementary Figure 17. Microarray results show up-regulation of phage shock and OxyR regulons following indole treatment in exponential phase cultures. (a) Z-score ranking of significantly up- and down-regulated genes in indole-treated cultures (|z| ≥ 2, p ≤ 0.05). Up-regulated genes are shown in red and down-regulated genes are shown in blue. Significance was determined by calculating z-scores for gene expression after indole treatment as compared to an untreated baseline; variance was estimated from microarray compendium data (see Supplementary Methods). Fold-change expression of genes in the (b) phage shock regulon and (c) OxyR regulon after indole treatment.


Supplementary Figure 18. qPCR validation of microarray results for selected genes. All experiments were carried out in 1 mL cultures in M9CG. (a) Fold-change expression values were obtained for selected genes in the phage shock and OxyR regulons in wild-type cultures (500 µM indole vs. untreated control). Error bars for qPCR data represent mean ± SD for four (exponential phase) or five (stationary phase) biological replicates. (b) Fold-change expression values were obtained for selected genes in the phage shock and OxyR regulons in wild-type (300 µM hydrogen peroxide vs. untreated control) and ΔtnaAΔmtr (500 µM indole vs. untreated control) stationary phase cultures. Error bars represent mean ± SD for three biological replicates.

35


Supplementary Figure 19. The phage shock and OxyR regulons are involved in indole-induced persistence. Cultures of wild-type, Δflu, ΔpspBC, ΔfluΔoxyR, and ΔfluΔpspBCΔoxyR E. coli were grown to stationary phase in M9CG, then incubated for 1 hour with and without 500 µM indole prior to treatment with ofloxacin. Error bars represent mean ± SD of at least three biological replicates.

36


Supplementary Figure 20. Un-annotated indole-responsive genes are not necessary for indole-induced persistence. Single-gene knockout E. coli strains were created for un-annotated genes identified as being significantly up-regulated by indole treatment in stationary phase. 24-hour stationary phase cultures of wild-type and single-gene knockout E. coli grown in M9CG were pre-incubated for 1 hour with and without 500 µM indole prior to treatment with ofloxacin. Results are presented as (a) percent survival and (b) fold change survival of indole-treated vs. untreated cultures. Error bars represent mean ± SD of at least three biological replicates.

37


38

Supplementary Movies 1-3. Time-lapse of optical (DIC) and fluorescence (GFP) images of E. coli bearing indole-responsive fluorescent plasmid PtnaC. Each movie represents a time series of images obtained from one independent experiment. In all experiments, cells received selective growth media (M9CG + 60 µg/mL kanamycin) for 2 hours so that the total number of analyzed cells exceeded 50 (‘Media’), then were treated with selective media + 500 µM indole for 1 hour (‘Indole’), followed by a 1-hour ampicillin treatment (M9CG + 500 µM indole + 100 µg/mL ampicillin, ‘Ampicillin’), and finally an outgrowth with selective media for ≥ 2 hours (‘Media’). DIC images were taken at 30-second intervals during ampicillin treatment and 3-minute intervals otherwise, and GFP images were taken at 6-minute intervals across all conditions. Time (in hours:minutes since beginning of experiment) is displayed in the lower right corner of each movie, while media or treatment is displayed in the lower left corner.


39

Supplementary Table 1. Survival of indole-treated and untreated stationary phase cultures in M9CG and LB. Row headers indicate the media type and amount of indole added to the culture (0 or 500 µM). These data correspond to Figure 1a in the main text and are presented as mean (SD).

WT ΔtnaA Δmtr

LB, 0 1.00 (0.06) 0.12 (0.04) 4.45 (0.82) LB, 500 1.12 (0.20) 2.09 (1.33) 5.11 (2.68)

M9CG, 0 0.21 (0.06) 0.22 (0.12) 2.06 (1.33) M9CG, 500 2.42 (0.87) 2.63 (1.07) 5.20 (3.03)


40

Supplementary Table 2. Fold change expression of multi-drug transporter genes in indole-treated (500 µM) vs. untreated cultures of E. coli K-12 EMG2. Data from Hirakawa et al.44 (500 µM indole, LB medium, E. coli strain MC4100) are presented for comparison.

Gene

Exponential Phase

M9CG, EMG2

Stationary Phase

M9CG, EMG2

Exponential Phase

LB, MC410063

acrA 1.52 0.86 1.2

acrD 1.05 1.17 1.6

acrE 0.93 0.62 0.8

acrF 0.91 1.16 ND

bcr 1.19 0.67 1.6

cusB 0.91 0.91 1.0

emrA 1.21 0.95 1.4

emrD 0.94 1.20 1.2

emrE 0.99 1.17 1.0

emrK 1.03 0.86 1.2

fsr 1.02 1.12 1.6

macA 0.83 0.84 1.5

mdfA 1.08 0.70 1.5

mdtA 1.30 1.03 1.7

mdtB 1.01 1.09 ND

mdtC 1.11 1.07 ND

mdtD 0.99 0.83 ND

mdtE 1.15 0.79 2.4

mdtF 1.11 0.94 ND

mdtG (yceE) 1.05 1.00 1.0

yceL 0.96 0.85 1.5

ydgF 0.89 1.09 0.9

ydhE 0.97 1.00 1.6

yidY 0.94 0.86 1.7

yjiO 1.01 0.98 1.2


41

Supplementary Table 3. Bacterial strains, primers, and plasmids used in this study. Strains and primers Genotype or sequence

E. coli K12 strains EMG2 Wild-type EMG2 tnaA K12 EMG2 ∆tnaA EMG2 mtr K12 EMG2 ∆mtr EMG2 tnaA mtr K12 EMG2 ∆tnaA ∆mtr Ω KmR EMG2 pspBC K12 EMG2 ∆pspBC Ω KmR EMG2 flu K12 EMG2 ∆flu EMG2 flu oxyR K12 EMG2 ∆flu ∆oxyR Ω KmR qPCR primers ahpC Forward TGCAGAAACTGGGCGTAGAC ahpC Reverse TGTCGAAGTTACGGGTCAGG ahpF Forward CGGCAGAAGAGCTGAACAAG ahpF Reverse CGAGGATCTGACCACCAAAA dps Forward GCGCTAACTTCATTGCCGTA dps Reverse CGGGGTTTTGCTGTTGATAA grxA Forward CGGGTTGCCCTTACTGTGT grxA Reverse CGGGTTTACCTGCCTTTTGT pspA Forward GTGAGGTTGAATGGCAGGAA pspA Reverse CAGCGTCACTTCATGTTCCA rrsC Forward AGCGGTGAAATGCGTAGAGA rrsC Reverse TTTACGGCGTGGACTACCAG rrsH Forward ATCAGATGCAGTTCCCAGGT rrsH Reverse GTAATACGGAGGGTGCAAGC Knockout confirmation primers tnaA Forward CCAGAGCCAAACCGATTAGA tnaA Reverse ACCATAACACCCCAAAATGC mtr Forward TGGTGATGGTCGTCGTTTA mtr Reverse AGCTGGCGATGGATGTTAAG pspBC Forward AGAAAGCCACAGCTTCGGTA pspBC Reverse CACAGCCAGCAACATTTTCA flu Forward CAGGGAACAACAGAACCACA flu Reverse GCTCTGAACAAAAGGCGAAG oxyR Forward AACTCTCGAAACGGGCAGT oxyR Reverse TTCATATCGGTCAGGCGATT Primers for phage-mediated PCR product knockoutspspBC Forward GGAGTACTTATGAGCGCGCTATTTCTGGCTATTCCGGTGTAGGCTGGAGCTGCTTCG pspBC Reverse TAACTTTCCTCACAGTTGACGGAAACGGCTACGTAACATATGAATATCCTCCTTAGTTCCTATTC Rescue plasmid construction tnaA cloning F CCGCGCCGTGCGGTACCATGGAAAACTTTAAACATCTCCCTG tnaA cloning R CGCGGCGTCGTCTAGAGCTAACATCCTTATAGCCACTCTG mtr cloning R GCGGTCGATCTAGAGTCGGATAAGGCACCGCTGATTAC mtr cloning F CGCGATGCGTCGGTACCATGGCAACACTAACCACCAC Rescue plasmids pZA21G KanR; PltetO-1-GFP pZA21-mtr derivative of pZA21G; mtr gene cassette replaces GFP (KpnI/HindIII) pZA21-tnaA derivative of pZA21G; tnaA gene cassette replaces GFP (KpnI/HindIII)


42

Supplementary Table 4. Hypergeometric test results for GO category enrichment of genes for which expression was changed at least 1.8-fold by indole treatment in stationary phase (see Analysis of Microarray Results in Supplementary Note 1). GOBPID P value Term Genes

GO:0006412 9.97E-25 translation

rpsT, rpmF, rpmG, pheM, rplQ, rpsM, rpsK, rpsD, rpsE, rplE, rpsN, rpsH, rplF, rplR, rplP, rpmC, rpsQ, rplN, rplX, rpsS, rplV, rpsC, rpsJ, rplC, rplD, rplW, rplB, rplM, rpsL, rpsG, rpsF, rplU

GO:0019538 2.83E-14 protein metabolic process

dnaK, rpsT, rpmF, rpmG, pheM, rplQ, rpsM, rpsK, rpsD, rpsE, rplE, rpsN, rpsH, rplF, rplR, rplP, rpmC, rpsQ, rplN, rplX, rpsS, rplV, rpsC, rpsJ, rplC, rplD, rplW, rplB, rplM, rpsL, rpsG, malE, groS, rpsF, rplU

GO:0043284 9.42E-09 biopolymer biosynthetic process

cpxP, rpsT, pspC, rpmF, pspA, pspB, rpmG, pheM, ygeV, glgS, rplQ, rpsM, rpsK, rpsD, rpoA, rpsE, rplE, rpsN, rpsH, rplF, rplR, rplP, rpmC, rpsQ, rplN, rplX, rpsS, rplV, rpsC, rpsJ, rplC, rplD, rplW, rplB, rplM, rpsL, rpsG, rpoB, rpsF, rplU

GO:0009228 8.92E-03 thiamin biosynthetic process thiE, thiC, thiH GO:0006097 9.53E-03 glyoxylate cycle aceB, aceA

GO:0006081 1.31E-02 cellular aldehyde metabolic process aceE, aceB, aceA

GO:0032268 1.92E-02 regulation of cellular protein metabolic process rpsH, rplD

GO:0000162 4.63E-02 tryptophan biosynthetic process trpL, tnaB


43

Supplementary Table 5. Hypergeometric test results for GO category enrichment of genes for which expression was changed at least 1.8-fold by indole treatment in exponential phase (see Analysis of Microarray Results in Supplementary Note 1). GOBPID P value Term Genes

GO:0000162 3.00E-11 tryptophan biosynthetic process trpD, trpC, trpB, trpE, trpL, trpA, aroH, mtr

GO:0046417 2.11E-08 chorismate metabolic process trpD, trpC, trpB, trpE, trpL, trpA, aroH, pheL, mtr

GO:0042401 2.03E-06 biogenic amine biosynthetic process trpD, trpC, trpB, trpE, trpL, trpA, aroH, mtr

GO:0008652 1.39E-05 cellular amino acid biosynthetic process

thrL, trpD, trpC, trpB, trpE, trpL, trpA, aroH, hisL, pheL, sdaC, mtr, asnA, glnA, aspA

GO:0016053 6.60E-05 organic acid biosynthetic process

thrL, trpD, trpC, trpB, trpE, trpL, trpA, aroH, hisL, pheL, sdaC, mtr, asnA, glnA, aspA

GO:0006221 3.64E-03 pyrimidine nucleotide biosynthetic process pyrD, pyrI, pyrL

GO:0019752 6.19E-03 carboxylic acid metabolic process

thrL, trpD, trpC, trpB, trpE, trpL, trpA, aroH, hisL, pheL, sdaC, mtr, tnaA, tnaC, asnA, glnA, aspA

GO:0006529 7.05E-03 asparagine biosynthetic process asnA, aspA

GO:0006805 1.57E-02 xenobiotic metabolic process ahpF, marA, katG, soxS

GO:0009266 1.85E-02 response to temperature stimulus cspA, ibpB, ibpA

GO:0009117 1.94E-02 nucleotide metabolic process pyrD, ndk, purT, purH, pyrI, pyrL

GO:0015949 2.42E-02

nucleobase, nucleoside and nucleotide interconversion codB, codA, ndk, uraA, purH

GO:0009058 2.57E-02 biosynthetic process

thrL, crl, trpD, pspC, trpC, pyrD, ndk, trpB, marR, trpE, trpL, pspA, pspB, trpA, aroH, purT, hisL, pheL, sdaC, marA, mtr, rpmC, cspA, asnA, glnA, purH, soxS, aspA, rplI, pyrI, pyrL, rpsI, csgD

GO:0006807 3.09E-02 nitrogen compound metabolic process

thrL, codB, codA, crl, pspC, pyrD, ndk, marR, pspA, pspB, purT, hisL, uraA, pheL, sdaC, marA, cspA, rnpA, tnaA, tnaC, asnA, glnA, purH, soxS, aspA, pyrI, pyrL, csgD, mcrB

GO:0046483 3.21E-02 heterocycle metabolic process pyrD, ndk, purT, hisL, purH, pyrI, pyrL


44

Supplementary References

1. Bachmann, B.J. Bacteriol. Rev. 36, 525-557 (1972). 2. Hayashi, K. et al. Mol. Syst. Biol. 2, 2006.0007 (2006). 3. Baba, T. et al. Mol. Syst. Biol. 2, 2006.0008 (2006). 4. Datsenko, K.A. & Wanner, B.L. Proc. Natl. Acad. Sci. U S A 97, 6640-6645 (2000). 5. Kibbe, W.A. Nucl. Acids Res. 35, W43-W46 (2007). 6. Kawamura-Sato, K. et al. FEMS Microbiol. Lett. 179, 345-352 (1999). 7. Wiuff, C. & Andersson, D.I. J Antimicrob Chemother 59, 254-263 (2007). 8. Jensen, M.T., Cox, R.P. & Jensen, B.B. Appl. Environ. Microbiol. 61, 3180-3184 (1995). 9. Keren, I. et al. FEMS Microbiol. Lett. 230, 13-18 (2004). 10. Keren, I. et al. J. Bacteriol. 186, 8172-8180 (2004). 11. Eng, R.H. et al. Antimicrob. Agents Chemother. 35, 1824-1828 (1991). 12. Ashby, M.J., Neale, J.E., Knott, S.J. & Critchley, I.A. J. Antimicrob. Chemother. 33, 443-

452 (1994). 13. Lutz, R. & Bujard, H. Nucl. Acids Res. 25, 1203-1210 (1997). 14. Zaslaver, A. et al. Nature Methods 3, 623-628 (2006). 15. Cookson, S. et al. Mol. Syst. Biol. 1, 2005.0024 (2005). 16. Stricker, J. et al. Nature 456, 516-519 (2008). 17. Duffy, D.C., McDonald, J.C., Schueller, O.J.A. & Whitesides, G.M. Anal. Chem. 70,

4974-4984 (1998). 18. Unger, M.A. et al. Science 288, 113-116 (2000). 19. Thorsen, T., Maerkl, S.J. & Quake, S.R. Science 298, 580-584 (2002). 20. Urbanski, J.P. et al. Lab Chip 6, 96-104 (2006). 21. Wang, Q. et al. Cytometry Part A 77A, 101-110 (2010). 22. Lee, H.H., Molla, M.N., Cantor, C.R. & Collins, J.J. Nature 467, 82-85 (2010). 23. Kobayashi, A. et al. J. Bacteriol. 188, 5693-5703 (2006). 24. Hirakawa, H. et al. Microbiology-(UK) 155, 541-550 (2009). 25. Kema, I.P., de Vries, E.G.E. & Muskiet, F.A.J. J. Chromatogr. B. 747, 33-48 (2000). 26. Han, T.H. et al. Res. Microbiol. 7, 108-116 (2010). 27. Lee, J.T., Jayaraman, A. & Wood, T.K. BMC Microbiol. 7, 42 (2007). 28. Lacour, S. & Landini, P. J. Bacteriol. 186, 7186-7195 (2004). 29. Pinero-Fernandez, S., Chimerel, C., Keyser, U.F. & Summers, D.K. J. Bacteriol. 193,

1793-1798 (2011). 30. Untergasser, A. et al. Nucl. Acids Res. 35, W71-W74 (2007). 31. Gardner, T.S., di Bernardo, D., Lorenz, D. & Collins, J.J. Science 301, 102-105 (2003). 32. Pfaffl, M.W. Nucl. Acids Res. 29, 9.e45 (2001). 33. Faith, J.J. et al. PLOS Biology 5, 54-66 (2007). 34. Storey, J.D. & Tibshirani, R. Proc. Natl. Acad. Sci. U S A 100, 9440-9445 (2003). 35. Shannon, P. et al. Genome Res. 13, 2498-2504 (2003). 36. Fica, A.E., Horowitz, H.W., Lior, H. & Cabello, F.C. J. Clin. Microbiol. 32, 2327-2330

(1994). 37. Falcon, S. & Gentleman, R. Bioinformatics 23, 257-258 (2007). 38. Audia, J.P., Webb, C.C. & Foster, J.W. Int. J. Med. Microbiol. 291, 97-106 (2001). 39. Watanabe, T. & Snell, E.E. Proc. Natl. Acad. Sc.i U S A 69, 1086-1090 (1972).


45

40. Yanofsky, C., Horn, V. & Gollnick, P. J. Bacteriol. 173, 6009-6017 (1991). 41. Botsford, J.L. & DeMoss, R.D. J. Bacteriol. 105, 303-312 (1971). 42. Bordi, C., Theraulaz, L., Mejean, V. & Jourlin-Castelli, C. Mol. Microbiol. 48, 211-223

(2003). 43. Gong, F. & Yanofsky, C. J. Bacteriol. 185, 6472-6476 (2003). 44. Hirakawa, H. et al. Mol. Microbiol. 55, 1113-1126 (2005).


Supplementary Note 1

Background

Indole has been identified as a stationary phase signal in E. coli (Supplementary Fig. 1).

Indole is produced by E. coli under conditions known to increase persistence, such as nutrient

limitation1, stationary phase growth2, and biofilm formation3.

Bacterial persisters are dormant cells within an isogenic bacterial population that tolerate

antibiotic treatment4 and have been implicated in chronic and recurrent infections5. Persister

cells are thought to be dormant, dividing less frequently than non-persisters or not at all, and

emerging more slowly from stationary phase than non-persisters6. Persister formation occurs

heterogeneously within an antibiotic-susceptible population, predominantly at the transition to

stationary phase7-10. This process has been viewed as a bacterial “bet-hedging” strategy in

fluctuating environments11. Unlike antibiotic resistance, persistence is not associated with

growth in the presence of antibiotics (increased MIC), and survival of persisters is robust to

intensity and duration of antibiotic treatment10,12.

Indole Induces Persister Formation

We sought to verify that these experiments measured persister levels and not resistance.

We verified that these experiments measured persister levels and not resistance by extending the

duration of antibiotic treatment and increasing the concentration of antibiotic (Supplementary

Fig. 3 and 4), neither of which affected survival. Further, we confirmed that, the ofloxacin MICs

for wild-type, ΔtnaA, Δmtr, and ΔtnaAΔmtr strains were identical (~0.1 µg/mL), indicating that

none of the genetic perturbations used affected intrinsic antibiotic resistance.


Indole is a Stationary Phase Signal Molecule in E. coli

We sought to confirm that indole was produced as a stationary phase signal in our

system. We observed very low levels of indole production in tryptophan-free minimal media

(M9CG, Supplementary Fig. 5) but observed high levels of indole production in rich media (LB,

Supplementary Fig. 6). Extracellular indole levels were higher in stationary phase than in

exponential phase in both media conditions.

Indole Signaling Increases Survival After Antibiotic Treatment at 30°C

Previous work has shown that survival after antibiotic treatment increases at lower

temperatures in the presence of indole13. Consistent with this earlier work, we found that the

∆tnaA mutant showed a greater deficit in persister formation relative to wild-type when growth

temperature was decreased from 37°C to 30°C (Supplementary Fig. 8).

Mtr is Required for High-Efficiency Import of Indole to the Cytoplasm

Recent work14 has called the function of Mtr15 into question. We investigated the role of

Mtr in indole uptake using high-performance liquid chromatography (HPLC) and indole

auxotrophy studies. Our HPLC measurements demonstrate that the extracellular indole

concentration is significantly higher in a ∆mtr strain than in wild-type (Supplementary Fig. 5),

suggesting that wild-type cells import indole more efficiently than cells lacking mtr.

Additionally, we examined growth in minimal media of a ∆trpA mutant, which is

auxotrophic for tryptophan but can synthesize this amino acid from indole16,17, and in the

∆trpA∆mtr mutant. Cultures were grown in minimal media (M9 + 0.2% glucose) with and

without addition of 50 µM indole. We found that ∆trpA∆mtr cells respond more slowly to indole


than ∆trpA cells (Supplementary Fig. 9a), suggesting that Mtr is important for import of indole to

the cytoplasm. It has been suggested that growth inhibition of the ∆mtr mutant by indole could

account for this observation14; however, we found that the growth of ∆mtr cultures was not

affected by physiological levels of indole (Supplementary Fig. 9b). Cultures of the ∆trpA∆mtr

strain eventually reached the same density as cultures of the ∆trpA strain, likely as a result of

passive indole import to the cytoplasm or activity of lower-affinity transporters in the absence of

Mtr.

Our results, together with previous studies, suggest that Mtr actively transports indole to

the cytoplasm and that in the absence of Mtr, indole can still be imported by less-efficient means.

We observed Mtr-independent indole-responsive growth beginning after three hours' incubation

with indole, consistent with the recent report by Pinero-Fernandez et al., which showed Mtr-

independent indole import after four hours of incubation14. A previous report indicated that the

mtr mutant was entirely unable to utilize indole for growth15; however, the previous study used a

background strain lacking the entire trp operon, where conversion of indole to tryptophan by the

highly inefficient TnaA reverse reaction18 was required for growth. It is plausible that the more

efficient conversion of indole to tryptophan by TrpB2 can support growth at lower concentrations

of intracellular indole.

Extracellular Indole Levels in ∆mtr Cultures can Account for High-Persistence Phenotype

We tested whether the increased extracellular indole in Δmtr cultures grown in M9CG

(~15 µM) could account for the increase in persistence relative to wild-type (~5 µM, Fig. 1a,

Supplementary Fig. 5). To do so, we incubated wild-type cultures in M9CG overnight with 15

µM indole to mimic indole concentrations in Δmtr cultures. We found that wild-type persister


levels were increased by an order of magnitude by long-term incubation with low concentrations

of indole (Supplementary Fig. 10), suggesting that the difference in persistence between the

wild-type and Δmtr strains in M9CG could be accounted for by increased indole in the periplasm

or extracellular space.

Elimination of Indole Production in a Δmtr Mutant Abolishes High-Persistence Phenotype

We hypothesized that extracellular indole levels were responsible for the high persistence

observed in Δmtr, and that elimination of indole production in the Δmtr strain should therefore

eliminate its high-persistence phenotype. To test this hypothesis, we measured persister levels in

a ΔtnaAΔmtr strain that was unable to produce indole. We found that untreated cultures of the

ΔtnaAΔmtr strain and ΔtnaA strain had similar persister levels (Supplementary Fig. 11),

indicating that the Δmtr strain is not a high-persistence mutant in the absence of indole, and

suggesting that the increased persistence of the Δmtr strain is a function of indole signaling. As

indole-induced persistence did not require high-efficiency active transport into the cytoplasm by

Mtr or native indole production by TnaA, we surmised that the mechanism responsible for the

observed increase in persister formation did not involve an increase in cytoplasmic indole.

Indole-Induced Persistence is Dose-Dependent

Interestingly, we found that wild-type persister levels were approximately five-fold

higher in LB than in M9CG (Fig. 1a, Table 1), and that incubation with indole increased wild-

type persister formation in M9CG but did not further increase persister levels in LB (Fig. 1a,

Table 1). These findings suggest that the increased persister levels observed in wild-type

cultures grown in LB compared with those grown in M9CG may be, in part, accounted for by


indole signaling. Further, these results indicate that indole-induced persister formation may be

saturated in LB. To determine whether the indole response could be saturated, we incubated

wild-type E. coli with varying concentrations of indole in M9CG (10-500 µM, Supplementary

Fig. 12). We found that indole-induced persister formation reaches a plateau at physiological

concentrations of indole (250-500 µM). These results suggest that physiological concentrations

of indole, which are well below the cytotoxic level, serve as an “inoculant” to induce persister

formation.

Validation of Microarray Data

Multiple genes in the OxyR and phage shock regulons showed significantly increased

expression in response to indole. The entire known phage shock regulon, consisting of the

pspABCDE and pspG operons, was up-regulated in indole-treated cultures (Supplementary Fig.

16 and 17). The E. coli phage shock response is induced by stationary phase conditions and a

variety of other stresses, and is important for long-term survival under alkaline stress and

nutrient limitation19-21.

Within the OxyR regulon, multiple genes were up-regulated in both stationary and

exponential phase indole-treated cultures (Supplementary Fig. 16 and 17). OxyR responds to

oxidative stress, and its activation has been observed to protect cellular components from

oxidative damage and increase survival during stasis22,23.

We used qPCR to validate our microarray results for indole-induced expression of genes

in the OxyR and phage shock regulons in wild-type E. coli (Supplementary Fig. 18a).

Additionally, using qPCR, we verified that indole induces expression of genes in the OxyR and

phage shock regulons in a ΔtnaAΔmtr strain (Supplementary Fig. 18b). These results indicate


that the OxyR and phage shock responses are induced by indole even in strains impaired in

production and active import of indole.

Analysis of Microarray Results

Hypergeometric test-based GO enrichment analyses were performed on sets of genes

where expression was changed by at least a factor of 1.8 following indole treatment. GO

categories showing significant enrichment (p ≤ 0.05) were considered, and for each significantly

enriched GO category, the corresponding genes within the indole-responsive gene sets (at least

1.8-fold change expression) were listed. GO analysis of stationary phase data (Supplementary

Table 4) and exponential phase data (Supplementary Table 5) indicated that indole treatment

induced stress response pathways and decreased expression of genes involved in anabolic

metabolism.

Treatment with indole has been previously observed to increase expression of OxyR-

controlled genes24 and decrease expression of genes in curli synthesis and amino acid

biosynthesis pathways25. As “oxidative stress” and “phage shock” are not currently included in

the set of GO biological processes, enrichment analysis was performed based on regulon

contents as represented in EcoCyc26. Hypergeometric tests confirmed over-representation of

OxyR-controlled genes (p ≈ 0.0006) and phage shock genes (p < 0.0001), and indicated over-

representation of curli fiber-associated genes (p ≈ 0.0033) in the set of genes affected by indole.

The extracytoplasmic stress response system CpxAR has been implicated in indole-

controlled expression of multi-drug transporters and motility genes24,27-29. Our data showed that

indole treatment results in increased expression of the highly inducible regulator cpxP

(exponential phase, 1.54-fold; stationary phase, 3.66-fold), consistent with activation of CpxAR


in the absence of membrane stress30. However, the observed increase in cpxP was not

statistically significant, and as the phage shock and OxyR regulons can account for indole-

induced persistence (Fig. 2a, Supplementary Fig. 19), it is unlikely that the Cpx system plays a

critical role in this response. Though the BaeSR two-component stress response system has been

observed to respond to high levels of indole29, we did not observe differential expression of

BaeSR-associated genes after treatment with physiological indole concentrations. Likewise, we

did not observe indole to induce SdiA expression, which was previously observed in biofilm

cultures in rich media31.

High levels of indole have been shown to induce expression of multi-drug efflux

pumps27,28,32, resulting in resistance to antibiotics (increased MIC). However, we observed an

increase in survival after antibiotic challenge without observing any significant increase in

expression of drug exporter systems (P ≈ 0.28) (Supplementary Table 2), consistent with the

hypothesis that the increase in survival after incubation with indole is due to an increase in

persister formation. Our results are broadly consistent with those reported for exponential phase

planktonic cultures, where indole-based induction of multi-drug pumps was observed during

growth with toxic levels of indole (2 mM) but not with lower levels of indole (0.51 mM)27.

These results are in accord with earlier reports showing that indole signaling affects

genes involved in oxidative stress responses, curli synthesis, amino acid synthesis, and indole

production and transport13,24,25,27-29,31. However, the phage shock system has not previously been

shown to respond to indole. Furthermore, we did not observe significant activation of multi-drug

pumps by indole. The differences between our observations and earlier indole-responsive

transcriptome data may be due to differences in E. coli strains and/or test conditions. Whereas

previous studies were performed in rich media, typically on biofilms, at high indole


concentrations, and at longer exposure times13,24,25,27-29,31, we used minimal media, planktonic

cultures, physiologically relevant indole concentrations, and shorter exposure times to detect the

initial transcriptional response to indole.

Un-Annotated Genes Induced by Indole Are Not Required for Indole-Induced Persistence

Several of the genes found to be significantly induced by indole treatment in stationary

phase cultures lack functional annotation (Supplementary Fig. 16). The biological functions and

regulatory properties of these genes are not known. To determine whether these un-annotated

genes represent a biological function or functions important for indole-induced persistence, we

created genetic knockout strains. Stationary phase cultures of these knockout strains were tested

for indole-induced persister formation as previously described. We found that none of these

genes were required for indole-induced persistence (Supplementary Fig. 20). However, the ydcO

knockout shows lower persister levels than wild-type; this gene has been shown to increase

resistance to the toxin bromoacetate when over-expressed in E. coli33.


1. Wang, D.D., Ding, X.D. & Rather, P.N. J. Bacteriol. 183, 4210-4216 (2001). 2. Bordi, C., Theraulaz, L., Mejean, V. & Jourlin-Castelli, C. Mol. Microbiol. 48, 211-223

(2003). 3. Di Martino, P. et al. Can. J. Microbiol. 49, 443-449 (2003). 4. Lewis, K. Nat. Rev. Microbiol. 5, 48-56 (2007). 5. Smith, P.A. & Romesberg, F.E. Nat. Chem. Biol. 3, 549-556 (2007). 6. Balaban, N.Q. et al. Science 305, 1622-1625 (2004). 7. Levin, B.R. & Rozen, D.E. Nat. Rev. Microbiol. 4, 556-562 (2006). 8. Gefen, O. & Balaban, N.Q. FEMS Microbiol. Rev. 33, 704-717 (2009). 9. Sufya, N., Allison, D.G. & Gilbert, P. J. Appl. Microbiol. 95, 1261-1267 (2003). 10. Keren, I. et al. FEMS Microbiol. Lett. 230, 13-18 (2004). 11. Rotem, E. et al. Proc. Natl. Acad. Sc.i U S A 107, 12541-12546 (2010). 12. Kapoor, B., Sharma, C. & Sharma, R. Indian J. Med. Sci. 59, 120-129 (2005). 13. Lee, J. et al. Isme J. 2, 1007-1023 (2008). 14. Pinero-Fernandez, S., Chimerel, C., Keyser, U.F. & Summers, D.K. J. Bacteriol. 193,

1793-1798 (2011). 15. Yanofsky, C., Horn, V. & Gollnick, P. J. Bacteriol. 173, 6009-6017 (1991). 16. Jackson, E.N. & Yanofsky, C. J. Mol. Biol. 71, 149-161 (1972). 17. Jackson, E.N. & Yanofsky, C. J. Bacteriol. 116, 33-40 (1973). 18. Watanabe, T. & Snell, E.E. Proc. Natl. Acad. Sc.i U S A 69, 1086-1090 (1972). 19. Brissette, J.L., Weiner, L., Ripmaster, T.L. & Model, P. J. Mol. Biol. 220, 35-48 (1991). 20. Weiner, L. & Model, P. Proc. Natl. Acad. Sc.i U S A 91, 2191-2195 (1994). 21. Shah, D. et al. BMC Microbiol. 6, 53 (2006). 22. Dukan, S. & Nystrom, T. Genes Dev. 12, 3431-3441 (1998). 23. Storz, G., Tartaglia, L.A. & Ames, B.N. Antonie Van Leeuwenhoek 58, 157-161 (1990). 24. Garbe, T.R., Kobayashi, M. & Yukawa, H. Arch. Microbiol. 173, 78-82 (2000). 25. Bansal, T. et al. Infect. Immun. 75, 4597-4607 (2007). 26. Keseler, I.M. et al. Nucl. Acids Res. 37, D464-D470 (2009). 27. Hirakawa, H. et al. Mol. Microbiol. 55, 1113-1126 (2005). 28. Nishino, K., Honda, T. & Yamaguchi, A. J. Bacteriol. 187, 1763-1772 (2005). 29. Bury-Mone, S. et al. Plos Genetics 5, 17 (2009). 30. DiGiuseppe, P.A. & Silhavy, T.J. J. Bacteriol. 185, 2432-2440 (2003). 31. Lee, J.T., Jayaraman, A. & Wood, T.K. BMC Microbiol. 7, 42 (2007). 32. Davin-Regli, A. et al. Curr. Drug Targets 9, 750-759 (2008). 33. Desai, K.K. & Miller, B.G. Proc. Natl. Acad. Sc.i U S A 107, 17968-17973 (2010).


Documents

Signaling-Mediated Bacterial Persister Formation · Supplementary Methods and Results for: Signaling-Mediated Bacterial Persister Formation . Nicole M. Vega,1 Kyle R. Allison,1 Ahmad