Metabolic control of porin permeability influences antibiotic resistance in Escherichia coli

Abstract

Porins mediate the passage of hydrophilic nutrients and antibiotics across the outer membrane but might contribute to proton leak from the periplasm, suggesting that their conductance could be regulated. Here we show, using single-cell imaging, that porin permeability in Escherichia coli is controlled by changes in periplasmic H⁺ and K⁺ concentration. Conductance through porins increases with low periplasmic H⁺ caused by starvation, promoting nutrient uptake, and decreases with periplasmic acidification during growth in lipid media, limiting proton loss. High metabolic activity during growth in glucose media, however, activates the inner membrane voltage-gated potassium channel, Kch, increasing periplasmic potassium and enhancing porin permeability to dissipate reactive oxygen species. This metabolic control of porin permeability explains the observed increase in ciprofloxacin resistance of bacteria catabolizing lipids and clarifies the impact of mutations in central metabolism genes on drug resistance, identifying Kch as a therapeutic target to improve bacterial killing by antibiotics.

Fig. 1. Bacterial porins are regulated by changes in internal proton and potassium concentrations.

a,b, Accumulation of the fluorescent glucose analogue 2NBDG into E. coli following 10 min incubation at a range of concentrations (a) and at 20 μg ml⁻¹ over a range of incubation times (b), showing significant 2NBDG uptake at 5 min (P = 2.54 × 10⁻⁵), 10 min (P = 2.53 × 10⁻⁵), 30 min (P = 2.47 × 10⁻⁵) and 45 min (P = 3.55 × 10⁻⁵); (n ≥ 10 biological replicates for each condition). c, Accumulation at 10 min of 20 μg ml⁻¹ 2NBDG in WT E. coli (black) or isogenic knockout strains for the major porins ompF and ompC, minor porins ompG, nmpC and phoE (grey), and the voltage-gated potassium channel kch (blue; n = 8 biological replicates for each condition). d, Effect on 2NBDG accumulation (left) in WT E. coli of changing external pH alone (white) or in the presence of CCCP (250 μM; red) which led to significantly increased permeability at pH 3 (P = 2.61 × 10⁻³), pH 4 (P = 8.86 × 10⁻⁹), pH 5 (P = 8.86 × 10⁻⁹), pH 6 (P = 1.42 × 10⁻⁸) and pH 7–8.5 (all P ≤ 8.86 × 10⁻⁹); and (right) in WT (black), ΔompF (yellow) and ΔompC (purple) E. coli in the presence of CCCP (n = 12 biological replicates for each condition). e, Effect on 2NBDG accumulation (left) in WT E. coli of changing external potassium concentrations (while maintaining monovalent cations constant) alone (white) or in the presence of valinomycin (100 μM; blue) which caused significant increase in permeability at 10 mM (P = 7.82 × 10⁻²), 20 mM (P = 3.24 × 10⁻²), 40 mM (P = 5.91 × 10⁻³), 60 mM (P = 6.19 × 10⁻³), 80 mM (P = 2.18 × 10⁻³), 100 mM (P = 3.06 × 10⁻³) and 150 mM (P = 2.76 × 10⁻²) external K⁺; and (right) in WT (black), ΔompF (yellow) and ΔompG (purple) E. coli in the presence of valinomycin (n = 3 biological replicates for each condition). y axes in d and e show normalised 2NBDG fluorescence. f, Structure of OmpC (PDB 2J1N) highlighting periplasmic residues (E2, E43, E189, K6, K308, D7, D48, D135, D141, D268) likely to be affected by periplasmic acidification (their −log₁₀ acid dissociation constants (pKa) are shown in red). g, Cross-sectional views of molecular dynamic simulations of the effect of protonation of residues shown in f on pore diameter, thereby modelling the potential impact of periplasmic acidification. The structure shown is the average of sampled conformers. h, The effect on porin permeability (assessed using 2NBDG uptake (shown as 2NBDG fluorescence normalised to wild type); left) and on periplasmic pH (measured using a pHuji-based fluorescent reporter (shown as pHuji fluorescence normalised to wild type); right) of expressing a mutant OmpC where all charged residues on the periplasmic surface of OmpC were replaced with alanines (E23A, K27A, D28A, E64A, D69A, D156A, D162A, D289A, K329A); Mutant OmpC (green) compared to isogenic WT control (white). i,j, Single-cell fluorescence imaging of E. coli for a single experiment grown in a microfluidic perfusion system (mother machine) shows 2NBDG accumulation over time in WT bacteria expressing empty vector (WT) or expressing the light-activated proton pump ArchT in the inner membrane in the presence (blue) or absence (red) of 541 nm light exposure showing representative images (i) and quantification (shown as background-corrected 2NBDG fluorescence in a.u.) (j). Scale bar, 1 μm. Data (mean ± s.e.m.) are representative of at least three independent experiments performed at least in triplicate, imaging at least 50 individual bacteria per condition on each occasion. All data (mean ± s.e.m.) are representative of at least three independent experiments performed at least in triplicate. *P < 0.05, **P < 0.01, ***P < 0.001 (two-sided Student’s t-test). Source data

Fig. 2. Fluctuations in periplasmic ions cause temporal changes in porin permeability.

a, Representative images (left) and traces (right) over time of different individual bacteria (growing in M9 + 1 g l⁻¹ casamino acids + 1 g l⁻¹ glucose and 1 mM tryptophan media) imaged in the mother machine microfluidics platform expressing fluorescence sensors for cytoplasmic pH (pHcyto; pHluorin; purple), periplasmic pH (pHperi; pelB-pHuji; green), cytoplasmic K⁺ (K⁺cyto; Ginko1; blue), periplasmic K⁺ (K⁺peri; pelBC-ginko2; red) or cytoplasmic calcium (Ca²⁺cyto; GCaMp6f; yellow). Data are representative of at least three independent experiments with at least 50 bacterial cells per condition. b, Left: effect of CCCP (250 μM; green; P = 8.5 × 10⁻⁸), oligomycin (3 μM; violet; P = 0.00019), valinomycin (100 μM; light blue; P = 1.6 × 10⁻⁸) or vehicle alone (white) on baseline inner membrane voltage of individual bacteria (measured using the sensor QuasAr2 (ref. ), which increases in fluorescence with membrane depolarization). Right: baseline inner membrane voltage in WT cells (white) or isogenic knockouts for kch (red), ompC (purple) or ompF (yellow). N = 3 biological replicates; all P < 2 × 10⁻¹⁶. y axes show QuasAr2 fluorescence in a.u. c, Single bacterial measurements of (i) periplasmic K⁺ (detected through changes in pelBC-GINKO2 fluorescence; P < 2 × 10⁻¹⁶), (ii) periplasmic pH (detected through changes in pelB-pHuji fluorescence; P = 6.4 × 10⁻⁷) and (iii) cytoplasmic reactive oxygen species (ROS) levels (detected through changes in HyPer7 fluorescence; P = 0.00037) in WT (white) or isogenic kch knockout strains (red) of E. coli. y axes show sensor fluorescence in a.u. d, Time-lapse montage and the trace of E. coli WT and Δkch cells expressing the membrane potential reporter QuasAr2. Representative images and traces of membrane voltage (measured by QuasAr2 fluorescence (a.u.) in WT E. coli (blue) or isogenic kch knockouts (red). e, Single-cell recordings of periplasmic pH over time (monitored by pelB-pHuji fluorescence) following treatment with 150 mM external K⁺ (0 mM external Na⁺) and 100 μM valinomycin (High K⁺ + valino; red), 150 mM external K⁺ (0 mM Na⁺) alone (High K⁺; blue) or 0 mM external K⁺ (150 mM external Na⁺) and 100 μM valinomycin (Low K⁺ + valino; green). Representative images after 1,500 s treatment. f,g, Representative images (f) and traces (g) of simultaneous recordings in single bacteria (WT E. coli) of membrane voltage (QuasAr2 fluorescence; black) and 2NBDG accumulation (green) over time. Data are representative of at least three independent experiments with at least 50 bacterial cells imaged per condition. h, Plot of membrane voltage and 2NBDG uptake in individual bacteria (n > 40) over time. Line shows the linear regression of QuasAr2 fluorescence against 2NBDG fluorescence. Data are representative of at least three independent experiments performed at least in triplicate per condition, imaging at least 50 individual bacteria per condition shown as mean ± s.e.m. (Student’s t-test) in b, or violin plots (Wilcoxon signed-rank test) in c. ***P < 0.001. In b and e, error bars and ribbon bands represent the s.e.m. All statistical tests were two-sided. Scale bars, 1 μm. Source data

Fig. 3. The impact of metabolism on periplasmic ions, membrane voltage and porin permeability.

a, Representative traces of periplasmic K⁺ (top, measured by pelBC-GINKO2 fluorescence (a.u.)) and periplasmic H⁺ (bottom, measured by pelB-pHuji fluorescence (a.u.)) in individual bacteria when external media is changed to minimal media (M9, green), glucose media (M9 + 4 g l⁻¹ glucose; blue) or lipid media (M9 + 0.014 g l⁻¹ DPPC). b, Representative traces of changes in membrane voltage over time (monitored by QuasAr2 fluorescence (a.u.)) in individual bacteria exposed from low (5 ng l⁻¹) to high (500 mg l⁻¹) fumarate (orange) or glucose (blue). c, Frequency of action potentials (average peaks per min) for WT E. coli following exposure to media with different concentrations of fumarate (orange), fructose (purple), glucose (blue) or pyruvate (dark blue) as the only carbon source. Statistical analysis was performed using a generalized linear model (Methods; P = 1. 71 × 10⁻⁹ between fumarate and fructose). d, Porin permeability (detected through Hoechst accumulation measured by flow cytometry) of WT E. coli exposed to M9 media alone (minimal media; green), M9 media with low glucose (0.04 g l⁻¹; light blue), high glucose (4 g l⁻¹; dark blue) or lipid (0.014 g l⁻¹ DPPC; red). Bacteria were incubated for 10 min at 37 °C with Hoechst at times indicated. Fluorescence normalized to low glucose fluorescence at 0 h. Data shown (mean ± s.e.m.) are representative of at least three independent experiments per condition, each performed at least in triplicate. ***P < 0.001 (two-sided Student’s t-test). Source data

Fig. 4. Metabolic control of porin permeability influences antibiotic resistance.

a, Accumulation of ciprofloxacin in individual bacteria over time (monitored by fluorescence imaging of wild type E. coli using a microfluidics platform) assessed for WT bacteria (black), and isogenic knockouts for kch (red), the efflux pump component tolC (green), ompF (yellow) or ompC (purple). Ciprofloxacin fluorescence normalized to 0 s value. The data shown here (mean ± s.e.m.) are representative of at least three independent experiments per condition with at least 50 individual bacteria imaged each time. ***P < 0.001 (Student’s t-test). b, EC₅₀ values for WT E. coli (black) and isogenic knockouts of kch (red), ompF (yellow) and ompC (purple) exposed to ciprofloxacin (left; n = 3) or colistin (right; n = 5). Data shown (mean ± s.e.m.) are representative of at least three independent experiments performed at least in triplicate per condition. NS, not significant (Student’s t-test). c, Bacterial growth (monitored by OD₆₅₀) of WT E. coli (left) or isogenic knockouts of ptsH (middle) or aceA (right) in lipid media (0.014 gl⁻¹ DPPC + 0.3 g of Bacto Casitone and 0.5 ml of Tyloxapol; middle row), glucose media (M9 + 1 g l⁻¹ glucose; top row), mixed media (0.014 g l⁻¹ DPPC + 0.3 g of Bacto Casitone and 0.5 ml of Tyloxapol + 1 g l⁻¹ glucose; bottom row) in the presence of ciprofloxacin (0.025 μg ml⁻¹; blue) or vehicle alone (white). d, Accumulation of ciprofloxacin in individual WT E. coli over time (monitored as in a) during exposure to media containing glucose (black) or glucose at pH 3.0 with CCCP (250 μM; blue) (left) or media containing lipid (black) or lipid at pH 7/4 with CCCP (250 μM; red; right). e, MICs of ciprofloxacin for WT (white) or isogenic knockouts for kch (red); ptsH, gltD, or ycgG (blue); yidA (light blue), ushA (dark blue), or ompF (yellow) grown in glucose (left) or lipid (right) media (MIC values normalized to WT for each media). Significant increases in relative MIC observed for Δkch (P = 0.0023), ΔptsH (P = 0.039), ΔycgG (P = 0.010), ΔyidA (P = 0.0029) ΔushA (P = 0.014) and ΔompF (P= 0.0033) in glucose media but only ΔompF (P = 0.039) when grown in lipid media (Mann–Whitney U-test). f, Accumulation of ciprofloxacin within individual E. coli (monitored as in a, shown as ciprofloxacin fluorescence (a.u.)) grown in glucose media (M9 + 4 g l⁻¹ glucose) comparing WT (black), ushA (dark blue; P = 2.26 × 10⁻²¹) or yidA (light blue; P = 9.61 × 10⁻²³) isogenic knockout strains. Data shown (mean ± s.e.m.) are representative of at least three independent experiments per condition with at least 150 individual bacteria imaged each time. ***P < 0.001 (Student’s t-test). Source data

Extended Data Fig. 1. Effect of treatments and gene deletions on porin permeability and viability in E. coli.

(A) Uptake of 2NBDG (quantified by flow cytometry) by wild type E. coli. Gating strategy shown. (B) Uptake of 2NBDG by wild type (WT; white) and ΔptsH (red) E. coli. (C) Effect of CCCP or Valinomycin on E. coli viability measured in colony-forming units (CFU) per ml under conditions identical to those shown in Fig. 1e. Cells were grown in glucose (left) or lipid-based media (right) and treated with vehicle (white), CCCP (250 μM; blue), or Valinomycin (100 μM; green) for 30 minutes. (D) Accumulation of bocillin (0.5 µg/mL; top) or Hoechst (10 µg/mL: bottom) in E. coli (quantified by flow cytometry) following 10 min incubation using wild type bacteria (WT; black) or isogenic knockouts for ompF (yellow), ompC (purple), minor porins (ompG, nmpC, phoE; grey), and the ion channels clcB and kch (blue). (E) Accumulation of bocillin (0.5 µg/mL; top) or Hoechst (10 µg/mL: bottom) in wild type E. coli (quantified by flow cytometry) following treatment with the protonophores CCCP (blue) or Indole (red) or the potassium ionophore valinomycin (green), normalised to vehicle alone (white). Data (mean ± SEM) are representative of at least three independent experiments performed in at least triplicate. *p < 0.05, ** p < 0.01, *** p < 0.001 (Student’s t-test). Source data

Extended Data Fig. 2. OmpF and OmpC mutations found across phylogenetic tree of E. coli.

(A-D) Left Phylogenetic relationships between E. coli lineages based on core genes13. Scale bar shows expected substitutions per site. Right Amino acids at each variable position within OmpF (A) or OmpC (D) for each lineage. (B and E) OmpF (PDB accession 2OMF) or OmpC (PDB accession 2J1N) molecular structure coloured by position entropy; a measure of sequence variation at the position. (C and F) Same molecular structures coloured with key sites labelled (pink) that have been implicated by our molecular dynamics simulations as being involved in periplasmic ion-dependent regulation of pore conductance.

Extended Data Fig. 3. Molecular dynamic (MD)simulations of the change in porin pore size during acidification.

(A) MD simulation of OmpC (left) and OmpF (right) trimers at (top) physiological pH, (middle) following protonation of periplasmic amino acid residues, and (bottom) following protonation of the L3 loop within the channel. (B) Calculated pore diameters (in angstroms) using HOLE software for a range of different simulation conditions.

Extended Data Fig. 4. Growth dynamics of wild type and mutant OmpC-expressing E. coli.

The effect on E. coli growth in a variety of different media of porin permeability (assessed using 2NBDG uptake; left) of expressing a mutant OmpC where all charged residues on the periplasmic surface of OmpC were replaced with alanines (E23A, K27A, D28A, E64A, D69A, D156A, D162A, D289A, K329A); Mutant OmpC; red) compared to isogenic wild type control (blue).

Extended Data Fig. 5. Optogenetic probes and fluorescent ion sensors expressed in E. coli.

(A) Changes in cytosolic pH (assessed by the fluorescent reporter pHluorin) in E. coli expressing ArchT light-activation proton pump (right) or controls (left) in the absence (blue) or presence (red) of 561 nm illumination. Activation of Arch T results in significant alkalinisation of the cytosol as a consequence of pumping protons into the periplasm. ***p-value < 0.001 (B) Left. Calibration curve for cytoplasmic pHluorin and periplasmic pHuji. Cells were grown overnight and then seeded in the mother machine for microscopy observation (See methods for detailed protocol). The sensors were calibrated by flowing PBS + 1 g/L glucose + 250 μM CCCP and buffered at a range of pH (5.5, 6.5, 7.5, 8.5). Data shown (mean ± SEM) is representative of at least three independent experiments per condition. Right. Calibration curve for cytoplasmic GINKO1 and periplasmic-GINKO1. Cells were grown overnight and then seeded in the mother machine for microscopy observation (See methods for detailed protocol). The sensors were calibrated by flowing HEPES + 1 g/L glucose + 100 μM Valinomycin. The HEPES buffer then was balanced with NaCl or KCl in order to maintain the ionic strength constant at 150 mM. The KCl concentration was set at 150, 15, 1.5, and 0.015 mM. (C) Representative image of cell pole tracking algorithm for periplasmic measurements. The algorithm consisted in two steps. First, it segmented all the cells tracked the cells as a whole. Then, it segmented the top and bottom regions (associated with the periplasmic localisation) and tracked the fluorescence of those regions. (D) Fluorescence signal variance over time in individual wild type E. coli for each ion sensor detecting cytoplasmic pH (pH cyto; purple), periplasmic pH (pH peri; green), cytoplasmic potassium ions (K+ cyto; blue), periplasmic potassium ions (K+ peri; red), and cytoplasmic calcium ions (Ca2+cyto; gold). Data are representative of at least three independent experiments performed in at least triplicate, imaging at least 30 individual bacteria per condition. ***p-value < 0.001 (Student’s t-test relative to Ca2+ cyto signal). Source data

Extended Data Fig. 6. Effect of different carbon source metabolism on periplasmic pH, action potentials, and porin permeability in E. coli.

(A) The relative change in pH (given by the background subtracted pHuji fluorescence signal in the cell pole compared to the cell body) in cells grown in lipid-based medium (red; n = 392) and cells grown in glucose-based medium (blue; n = 394). The black dashed line shows the time at which the media was switched from M9 to lipid-based or glucose-based medium. (B) Frequency of action potentials (average peaks/ minute) for wild-type E. coli following exposure to media with different concentrations of fumarate (orange), fructose (pink), glucose (blue), pyruvate (purple), or DPPC (red) as the only carbon source. (C) Porin permeability (detected through Hoechst accumulation measured by flow cytometry) of wild-type (white) and Δkch (red) E. coli in different carbon sources. The cells were exposed to M9 media alone (first panel), high glucose (4 g/L; second panel), M9 media with lipid (0.014 g/L DPPC; third panel) or low glucose (0.04 g/L; fourth panel). Cells were incubated for 10 minutes at 37 oC with Hoechst at times indicated. Fluorescence normalised to low glucose fluorescence at 0 h. The data shown (mean ± SEM) is representative of at least three independent experiments performed at least in triplicate per condition. Source data

Extended Data Fig. 7. Model for metabolic control of porin regulation.

Model explaining metabolic control of porin permeability. (A) During starvation, the permeability of porins (blue) is high due to low periplasmic H+ and K+ levels (resulting from minimal electron transport chain (ETC; grey) activity and no opening of Kch channels (pink), respectively). (B) Growth in lipids causes an increase in periplasmic H+ (through increased ETC activity) without activating Kch, resulting in low porin activity. (C) In contrast, growth in glucose media drives ETC activity and Kch channel opening, leading to fluctuating periplasmic H+ and high periplasmic K + , causing fluctuations also in porin permeability.

Extended Data Fig. 8. Effect of gene deletions on antibiotic susceptibility for E. coli grown in lipid media.

EC50 values for wild-type E. coli (WT; black) and isogenic knockouts of kch (red), ompC (purple) and ompF (yellow) exposed to ciprofloxacin (left) or colistin (right) when growing in M9 + Lipids media. Data shown (mean ± SEM) is representative of at least three independent experiments performed at least in triplicate per condition. ns not significant, *p < 0.05, ** p < 0.01, *** p < 0.001 (Student’s t-test). Source data