Chemokines Kill Bacteria by Binding Anionic Phospholipids without Triggering Antimicrobial Resistance

Abstract

Classically, chemokines coordinate leukocyte trafficking during immune responses; however, many chemokines have also been reported to possess direct antibacterial activity in vitro. Yet, the bacterial killing mechanism of chemokines and the biochemical properties that define which members of the chemokine superfamily are antimicrobial remain poorly understood. Here we report that the antimicrobial activity of chemokines is defined by their ability to bind phosphatidylglycerol and cardiolipin, two anionic phospholipids commonly found in the bacterial plasma membrane. We show that only chemokines able to bind these two phospholipids kill Escherichia coli and Staphylococcus aureus and that they exert rapid bacteriostatic and bactericidal effects against E. coli with a higher potency than the antimicrobial peptide beta-defensin 3. Furthermore, our data support that bacterial membrane cardiolipin facilitates the antimicrobial action of chemokines. Both biochemical and genetic interference with the chemokine-cardiolipin interaction impaired microbial growth arrest, bacterial killing, and membrane disruption by chemokines. Moreover, unlike conventional antibiotics, E. coli failed to develop resistance when placed under increasing antimicrobial chemokine pressure in vitro. Thus, we have identified cardiolipin and phosphatidylglycerol as novel binding partners for chemokines responsible for chemokine antimicrobial action. Our results provide proof of principle for developing chemokines as novel antibiotics resistant to bacterial antimicrobial resistance mechanisms.

Keywords: Antimicrobial peptides; antibiotics; cardiolipin; multidrug-resistant microorganisms; phosphatidylglycerol; phospholipids.

Figure 1.. Antimicrobial chemokines bind CL and PG phospholipids.

a) Chemokines are more potent antimicrobials than classic antimicrobial peptides. Increasing doses (x-axis) of hBD3 (blue) and the antimicrobial chemokines CCL20 (magenta) and CXCL9 (orange) were incubated with 1 × 10⁺ cfu of E. coli for 2 h at 37°C. The non-antimicrobial chemokine CXCL8 (black) was used as a negative control. Samples were plated on agar plates and incubated overnight at 37°C. Data points represent the mean ± SEM cfu of 3 independent experiments analyzed in triplicate. Color-coded asterisks indicate statistically significant differences between each chemokine and hBD3 analyzed by two-way ANOVA with Dunnett’s multiple comparison test (**, p<0.01; ***, p<0.001). b) Antimicrobial assays. Increasing doses (x axis) of the human chemokines indicated above each graph were incubated with E. coli (magenta) or S. aureus (blue) as in panel a. Data points represent mean ± SD of the cfu counted from triplicates for each chemokine concentration. c) Chemokine binding to liposomes analyzed by biolayer interferometry (BLI). Top row, chemokine binding to liposomes replicating the phospholipid composition of E. coli (lpEC, magenta) or S. aureus (lpSA, blue). Bottom row, chemokine binding to phosphatidylcholine liposomes containing 30% of the phospholipids indicated on the inset of the left graph. Liposomes immobilized on BLI biosensors were incubated with 500 nM of the human chemokines indicated above each graph column. After 500 s, chemokine dissociation was recorded by incubating the biosensor with buffer alone. Lines represent the binding response in nm (y axis) over time (x axis). Data in b and c are from one experiment representative of 3 independent experiments. hBD3, human beta-defensin 3; cfu, colony formy units; CL, cardiolipin; PG, phosphatidylglycerol; lysPG, lysyl-phosphatidylglycerol; PE, phosphatidylethanolamine.

Figure 2.. Antimicrobial chemokines share common bacterial binding sites and localize to the bacterial cell poles.

a) Antimicrobial chemokines bind directly to bacteria. Representative images of the binding of 0.3 μM of the AZ-647-labeled antimicrobial chemokines CCL20, CCL21, CXCL9 and CXCL11 or the non-antimicrobial chemokines CCL3 and CXCL8, as indicated above each image, to E. coli (W3110 strain). b and c) Binding of antimicrobial chemokines to bacteria can be competed with other antimicrobial chemokines. In b, Representative images of CXCL11-AZ647 (0.3 μM) bound (magenta) to E. coli (W3110 strain) bacteria preincubated with buffer or the unlabeled chemokines indicated above each image column. In a and b, the top and bottom image rows show the staining for chemokine alone (magenta) or merged with DAPI (green), respectively. A white scale bar (20 μm) is inserted in the bottom left image. In c, quantification of the fluorescence intensity of CXCL11-AZ647 staining in bacteria preincubated with buffer alone or unlabeled chemokines as indicated on the x axis. Each dot corresponds to one bacterium (n ≈ 100). Data are from one experiment representative of 2 independent experiments. Statistical differences between each chemokine group and the “buffer” treatment group were analyzed by ANOVA with Tukey’s test for multiple comparisons (ns, not significant; ****, p < 0.0001). d) Antimicrobial chemokines bind to the PG/CL-rich membrane domains at the bacterial cell poles. Representative Airyscan confocal images of CXCL9-AZ647 binding to E. coli (W3110 strain) co-stained with NAO. Graph on the right side shows the normalized fluorescence intensity profile along a bacterium, as indicated by the dashed line in the “merge” panel, of NAO-524 (green), NAO-630 (orange) and CXCL9-AZ647 (magenta). DAPI, 4′,6-diamidino-2-phenylindole; NAO, nonyl acridine orange.

Figure 3.. Liposomes containing CL or PG neutralize microbial binding and killing by chemokines.

a and b) Liposomal PG and CL block chemokine binding to bacteria. In a, E. coli (W3110 strain) bacteria were incubated with CCL20-AZ647 in the presence of PC liposomes (100 μM) containing 30% PE, PG or CL or buffer alone as indicated above each column. Top and bottom micrograph rows show the staining for the chemokine alone or merged with DAPI, respectively. A white scale bar (20 μm) is inserted in the bottom left image. In b, quantification of the fluorescence intensity of CCL20-AZ647 bound to E. coli in the presence of buffer alone or liposomes containing PE, PG or CL as indicated on the x axis. Each dot corresponds to one bacterium (n ≈ 100). All liposome-treated groups were compared to “Buffer” by ANOVA with Tukey’s test for multiple comparisons (ns, not significant; ****, p < 0.0001). c and d) PG and CL-containing liposomes protect bacteria from antimicrobial chemokines. The chemokines (5 μg/ml) indicated above each graph were incubated for 2 h with E. coli (c) or S. aureus (d) in the presence of increasing doses (x axis) of liposomes containing 30% PE, PG or CL. Samples were plated on agar plates and cfu were counted after 18 h at 37°C. Lines represent the mean ± SD cfu from triplicates of one experiment representative of 3 independent experiments. The top and bottom horizontal dashed lines indicate the number of cfu counted when bacteria were incubated with buffer alone or with chemokine in the absence of liposome, respectively. DAPI, 4′,6-diamidino-2-phenylindole; cfu, colony forming units; CL, cardiolipin; PG, phosphatidylglycerol; PE, phosphatidylethanolamine.

Figure 4.. Cardiolipin-deficient bacteria are more resistant to antimicrobial chemokines.

a) E. coli strain BKT12 lacks CL. TLC plate showing the phospholipid composition of E. coli strains W3110 and BKT12 grown to exponential phase. b and c) Antimicrobial chemokines bind E. coli W3110 and BKT12 strains. Bacteria (1 × 10⁺⁶ cfu) were incubated with buffer alone or AZ647-labeled chemokines (0.3 μM) and stained with SYTO24 to distinguish bacteria from debris in the flow cytometer. In b, color-coded histograms showing the binding of the chemokines indicated in the inset to W3110 and BKT12 bacteria (SYTO24⁺ events). The background signal recorded in cells incubated with buffer alone (“Buffer”) is shown with a solid gray histogram. In c, quantification of the median fluorescence intensity (MFI) of the binding of the chemokines indicated on the x-axis to W3110 and BKT12 bacteria. Data are the mean ± SD of triplicates from one experiment and are representative of 3 independent experiments. d) CL-deficient bacteria are more resistant to CXCL9 and CCL20. W3110 and BKT12 bacteria (1 × 10⁺ cfu) were incubated for 2 h at 37°C with buffer or 1.2 μM of CXCL9 or CCL20 and then plated on agar plates. Surviving cfu were counted the next day and are plotted as the % relative to the number of cfu in samples incubated with buffer alone for each strain. Bars represent the mean ± SEM of data combined from 2 independent experiments. Three biological replicates were analyzed in triplicate in each experiment. Each dot corresponds to the average % cfu of one biological replicate. Data were analyzed by unpaired t test (***, p < 0.001). e) CXCL9 selectively kills W3110 bacteria. W3110 and BKT12 cfu were mixed 1:1 for a total of 1 × 10⁺ cfu and incubated for 2 h at 37°C with buffer alone or 1.2 μM of CXCL8 or CXCL9. Samples were then plated in agar plates with or without kanamycin for cfu counting of total bacteria or BKT12 bacteria (kanamycin resistant), respectively. W3110 cfu were calculated as the difference between the cfu counted in agar (total bacteria) and agar-kanamycin (BKT12 bacteria) plates for each sample. Bars represent mean ± SD of 3 biological replicates analyzed in triplicate for each strain (inset) and treatment (x-axis). Data are from one experiment representative of 3 independent experiments. Differences between the buffer and the chemokine groups within each bacterial strain were analyzed by 2-way ANOVA with Bonferroni test for multiple comparisons. Statistical analysis results are color-coded and indicated above the corresponding bars (ns, not significant; **, p < 0.01). cfu, colony forming units.

Figure 5.. Cardiolipin promotes rapid bacteriostatic and bactericidal action by antimicrobial chemokines.

a and b) CL is required for early bacteriostatic and bactericidal action by chemokines. FACS-based time-to-kill assays of the effect of chemokines and hBD3 on the growth (a) and killing (b) of parental W3110 and the CL-deficient E. coli strain BKT12. Bacteria (1 × 10⁺ cfu) were incubated at 37°C with 1.2 μM of the protein indicated above each graph, and live (a) and dead (b) bacteria were quantified at the indicated time points (x axis) by FACS after co-staining with SYTO24 (stains live and dead bacteria) and SYTOX-Orange (stains only dead bacteria). In a, the growth ratio was represented as the ratio between the number of live bacteria (SYTO24⁺ SYTOX⁻) detected at each time point and the initial number of live bacteria recorded immediately before the start of the incubation (time 0 min) with buffer alone or the corresponding protein (as indicated in the inset of the “CXCL8” graph). In b, the % of dead bacteria (SYTO24⁺ SYTOX⁺) relative to the number of total bacteria (SYTO24⁺) for each strain (as indicated in the inset of the “CXCL8” graph) and at each time point is shown. In a and b, data are summarized as the mean ± SEM of 3 independent experiments combined with 3 biological replicates performed in each experiment. Data were analyzed by two-way ANOVA with Bonferroni test for multiple comparisons (ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001). Results of statistical analyses for the comparison of buffer vs chemokine/hBD3 for each bacterial strain are color-coded and indicated above each graph. Results of statistical analyses for the comparison of W3110 vs BKT12 for each protein treatment are indicated above each x axis in black in panel a and above each graph in panel b. c) CL facilitates rapid bacteriostatic effects by low doses of antimicrobial chemokines. W3110 and BKT12 bacteria (as indicated in the inset of the bottom right graph) were incubated with decreasing doses (x axis) of CXCL11 and CCL21 and the number of live and dead bacteria were analyzed 1 or 2 h post treatment (hpt) (as indicated on the left of each graph row) by FACS, as in a and b. The number of live bacteria (left y-axis, black lines) detected at different chemokine concentrations is represented as % relative to the number of live bacteria recorded in the buffer-treated group for each strain and time point. The % of dead bacteria (right y-axis, blue lines) for each chemokine concentration and bacterial strain was calculated as in panel b. Data are shown as mean ± SD of three biological replicates from one experiment representative of 3 independent experiments. Data were analyzed by 2-way ANOVA with Bonferroni test for multiple comparisons. Results of the statistical analysis of the bacteriostatic (black lines) and killing activity (blue lines) for the W3110 vs BKT12 comparison are color-coded and indicated above each graph (ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001). d) CL is required for direct killing by antimicrobial chemokines. W3110 and CL-deficient BKT12 bacteria were incubated with buffer alone or 4.8 μM of CXCL9, CCL21 or protamine (Protam), and dead bacteria (SYTOX⁺) were analyzed 20 min after treatment by FACS, as in panel b. On top, representative contour plots showing the % of SYTOX⁺ cells detected after treatment of W3110 or BKT12 bacteria (as indicated above each plot column) with CXCL9 or protamine (as indicated on the right of each plot row). At the bottom, bar graph showing the quantification of the % of SYTOX+ W3110 or BKT12 bacteria (as indicated in the inset) in each treatment group. Bars show the mean ± SD of triplicates from one experiment representative of 3 independent experiments. Data were analyzed by two-way ANOVA with Bonferroni test for multiple comparisons (ns, not significant; **, p < 0.01; ***, p < 0.001). hBD3, human beta-defensin 3; hpt, hours post treatment; Protam, protamine.

Figure 6.. Antimicrobial chemokines lyse phospholipid bilayers in an anionic phospholipid-dependent manner.

a) Antimicrobial chemokines disrupt liposomal membranes that mimic the phospholipid composition of the E. coli plasma membrane. Calcein-leakage assay showing the release of calcein from PE/PG/CL liposomes upon injection (arrowhead) of 1.2 μM of the indicated chemokines or protamine. Curves represent the % of calcein released over time (x axis) by each protein relative to the maximum calcein release observed after incubation of the liposomes with 0.1% Triton X-100. Solid lines represent the mean of 3 biological replicates. Colored shaded area around the lines represents the SD. Data are from one experiment representative of 3 independent experiments. b) Radially integrated SAXS spectra of chemokines CCL5, CCL19 and CXCL11 interacting with 20:80 PG:PE liposomes at neutral pH. A peptide-to-lipid (P/L) charge ratio scan from 1/6 to 3/2 was performed as indicated for each chemokine. Bragg structure peaks were indexed for the observed phases. CCL19 and CXCL11 formed cubic phases (Q) (arrowed indices). All spectra showed a lamellar phase (L), and CCL5 displayed inverted hexagonal phases (H). To facilitate visualization, spectra have been manually offset in the vertical direction by scaling each trace by a multiplicative factor. c) Linear fits of the peak position indexations for the cubic phases indexed in b. Pn3m, Im3m and Ia3d cubic phases were observed; illustration of geometric surfaces next to symbol key. Estimation of mean NGC, <k>, from linear fits is displayed next to each plot. Each color represents a P/L ratio as noted in b. d) Anionic phospholipids are required for the membrane lytic activity of antimicrobial chemokines. Calcein-leakage assays showing the release of calcein from 3 different types of liposomes, PE/PG/CL, PE/PG, and PEPC, as indicated in the inset of the CCL19 panel, upon injection of 1.2 μM of the proteins indicated in the inset of each graph. Data were analyzed and graphed as in a and correspond to 3 biological replicates from one experiment representative of 3 independent experiments. e) CL facilitates membrane disruption by antimicrobial chemokines. Calcein release from 3 different types of liposomes (as indicated in the inset of the CXCL9 graph) at 20 min after addition of increasing doses (x axis) of the proteins indicated above each graph. Calcein leakage assays were performed as in d and the % of released calcein at 20 min after protein injection was recorded. Data are represented as mean ± SD of 3 biological replicates from one experiment representative of 3 independent experiments. Data were analyzed by two-way ANOVA with Tukey’s test for multiple comparisons. Statistical differences between the PE/PG/CL and the PE/PG liposome groups are indicated above each concentration point (ns, not significant; *, p < 0.05). CL, cardiolipin; PG, phosphatidylglycerol; PE, phosphatidylethanolamine; PC, phosphatidylcholine; NGC, negative gaussian curvature.

Figure 7.. CCL20 kills wild type and antibiotic-resistant E. coli without triggering chemokine resistance.

a) Coomassie Blue-stained gel (4–12% Bis-Tris) showing the SDS-PAGE analysis of commercially available CCL20 (Peprotech) and in-house produced CCL20-His. b) The His-tag of recombinant CCL20-His does not interfere with chemokine binding to PG and CL. BLI assays showing the binding of 0.5 μM CCL20-His to PC liposomes containing 30% of the phospholipids indicated in the inset. c) Microbroth dilution method for the determination of the MIC of CCL20-His. W3110 bacteria (5 × 10⁺⁵ cfu/ml) were incubated with increasing doses of CCL20-His in MHB and the % viable bacteria after 18 h incubation at 37°C was calculated using the BacTiter-Glo Kit. Graph shows mean ± SEM % of the bacterial viability at each chemokine concentration relative to bacteria treated with buffer alone. Data are combined from 3 independent experiments. d) Bacteria develop resistance against conventional antibiotics but not against CCL20-His. W3110 bacterial cultures (5 × 10⁺⁵ cfu/ml) were maintained in MHB for 2 weeks in the presence of a sublethal dose (0.5 × MIC) of three different antimicrobial compounds: tetracycline (Tet), ampicillin (Amp) or CCL20-His, as indicated on the inset. On selected dates, MICs for each antimicrobial agent on the corresponding culture was recalculated and bacteria were subcultured adjusting the compound concentration to 0.5 × MIC. Graph shows the MIC fold change compared to the initially calculated MIC of two independent bacterial cultures, represented by circle and triangle symbols, for each agent. e) CCL20-His kills Tet- and Amp-resistant bacterial strains. Parental W3110 bacteria or the three new bacterial strains generated in d, W3110-Amp, W3110-Tet or W3110-CCL20His, were incubated in MHB with Amp, Tet or CCL20-His at a concentration equivalent to the MIC of each compound on parental W3110. Bacteria (5 × 10⁺⁵ cfu/ml) were incubated at 37°C for 18 h and then bacterial viability in each sample was calculated as in b. Bars represent mean ± SEM % survival data combined from three independent experiments. kDa, kilodalton; MIC, minimum inhibitory concentration.