Cytotoxic rhamnolipid micelles drive acute virulence in Pseudomonas aeruginosa

ABSTRACT

Pseudomonas aeruginosa is an opportunistic human pathogen that has developed multi- or even pan-drug resistance toward most frontline and last resort antibiotics, leading to increasing frequency of infections and deaths among hospitalized patients, especially those with compromised immune systems. Further complicating treatment, P. aeruginosa produces numerous virulence factors that contribute to host tissue damage and immune evasion, promoting bacterial colonization and pathogenesis. In this study, we demonstrate the importance of rhamnolipid production in host-pathogen interactions. Secreted rhamnolipids form micelles that exhibited highly acute toxicity toward murine macrophages, rupturing the plasma membrane and causing organellar membrane damage within minutes of exposure. While rhamnolipid micelles (RMs) were particularly toxic to macrophages, they also caused membrane damage in human lung epithelial cells, red blood cells, Gram-positive bacteria, and even noncellular models like giant plasma membrane vesicles. Most importantly, rhamnolipid production strongly correlated with P. aeruginosa virulence against murine macrophages in various panels of clinical isolates. Altogether, our findings suggest that rhamnolipid micelles are highly cytotoxic virulence factors that drive acute cellular damage and immune evasion during P. aeruginosa infections.

INTRODUCTION

Pseudomonas aeruginosa is a Gram-negative opportunistic human pathogen that causes a variety of nosocomial infections, including ventilator-associated pneumonia, urinary tract infections, and soft tissue infections (1, 2). It is also one of the most common causes of secondary bacterial infection in influenza and COVID-19 patients (3). Unfortunately, P. aeruginosa infections are becoming increasingly difficult to treat due to the pathogen’s resistance to many front-line antibiotics and its ability to evade and subvert host immune responses (4, 5). During acute infection, P. aeruginosa notoriously secretes a panoply of virulence factors like the elastases LasA and LasB and proteases AprA and PrpL that degrade host immune proteins like immunoglobin G and complement factors (5–9). P. aeruginosa also produces various toxins, such as those secreted by the type III secretion system, that damage and kill phagocytic immune cells (10, 11). We recently demonstrated that the production of the siderophore pyoverdine was important for P. aeruginosa virulence against murine macrophages (12). During chronic infections, particularly in the lungs of cystic fibrosis patients, P. aeruginosa forms dense biofilms—bacterial communities reversibly attached to host tissue through the secretion of adhesion proteins, extracellular DNA, and exopolysaccharides—that are impervious to host immune cells and antimicrobial therapy (13–15).

While most pathogen immune evasion strategies rely on secreted proteins and small molecules, recent work suggests that lipid virulence factors can also target immune cells. For instance, lipopolysaccharides can trigger inflammasome activation and programmed cell death in macrophages (16, 17). In the presence of the quorum-sensing molecule 2-heptyl-3-hydroxy-4-quinolone (Pseudomonas quinolone signal, PQS) or under bacterial stress such as lysozyme or antimicrobial treatment, P. aeruginosa also produces membrane vesicles formed by blebbing of the outer membrane (18–22). In addition to activating the inflammasome in macrophages (23, 24), these vesicles can also directly deliver periplasmic content, including virulence factors, to host cells through membrane fusion (25, 26). Furthermore, a class of glycolipids called rhamnolipids that are regulated by quorum sensing has been shown to exert cytotoxicity toward phagocytic cells (27, 28). The contribution of these various P. aeruginosa virulence factors to immune cell death and host immune evasion remains unclear.

In this report, we demonstrate that rhamnolipids predominantly drive P. aeruginosa acute virulence against murine macrophages. We show that secreted rhamnolipids can form micelles that exhibit acute cytotoxicity, rupturing the macrophage plasma membrane and damaging intracellular organellar membranes within minutes. We also examine these rhamnolipid micelles’ structural and biochemical properties via transmission electron microscopy and liquid chromatography-mass spectrometry. Furthermore, we demonstrate that while these micelles are particularly toxic to macrophages, they are also capable of damaging a wide range of other cells, including human bronchial epithelial cells, red blood cells, and even Gram-positive bacteria. Finally, we report that rhamnolipid production in various panels of clinical isolates strongly correlates with P. aeruginosa virulence.

RESULTS

Lipid-rich material secreted by P. aeruginosa is toxic to murine macrophages

We previously showed that low-molecular-weight secreted material from P. aeruginosa is cytotoxic to murine macrophages and that this toxicity is partially dependent on siderophore pyoverdine (12). High-molecular-weight materials such as Pseudomonas Exotoxin A and protease IV have also been known to kill these cells (29–32), but a comprehensive evaluation of the relative impact of secreted virulence factors has not yet been performed. To determine which secreted factors contribute to virulence against murine macrophages, RAW264.7 cells were treated with supernatant from bacteria grown in a modified M9-casamino acid medium, which was previously used to study the acute virulence of P. aeruginosa against the nematode host Caenorhabditis elegans (33). Supernatants from these growth cultures were highly toxic to RAW264.7 cells, causing nearly complete cell death within 5 h (Fig. 1A). Based on our previous studies (12, 33, 34), a pyoverdine biosynthetic mutant PA14pvdF was used to test whether toxicity was mediated by pyoverdine. Comparable cell death was observed in this mutant and the wild-type PA14 (Fig. 1A).

Fig 1.

Lipid-rich material in the P. aeruginosa supernatant is highly toxic to murine macrophages. (A) Murine macrophage (RAW264.7) survival after exposure to supernatants from wild-type PA14 and PA14pvdF (grown in low-iron M9 medium). Survival was normalized to saline control. (B) Macrophage survival after exposure to PA14pvdF supernatants that have been pre-treated with 100 µg/mL proteinase K (inactivated by 5 mM PMSF) or depleted of lipids by chloroform extraction. Survival was normalized to saline control. (C) Quantification of lipids in the supernatant by fluorescence after 20 µg/mL FM 1–43 treatment. (D) Correlation between FM 1–43 fluorescence and toxicity toward murine macrophages for supernatants of 19 P. aeruginosa clinical and environmental isolates. Representative strains with high lipid content (PA14) and low lipid content (JJ692) are labeled in red. Survival was normalized to media control. (E) Schematic of purification pipeline for lipid-rich material. (F) Interactions between RAW264.7 cells and bacterial filtrate or purified lipid-rich material from PA14pvdF or JJ692pvdL in the presence of SYTOX Orange cell-impermeant nucleic acid stain (red). Secreted bacterial lipids were prelabeled with FM 1–43 (green). Cells were prelabeled with Hoechst 33342 cell-permeant nucleic acid stain (blue). Data in A, B, and C were analyzed via one-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not statistically significant.

Material from this mutant was used to further investigate which class(es) of secreted macromolecules (e.g., proteins, lipids) were responsible for cytotoxicity. Supernatant was treated with proteinase K or by using chloroform to extract lipids. Proteinase K treatment had no apparent effect on the toxicity of the supernatant (Fig. 1B), suggesting proteins were not responsible for the cell death. In contrast, lipid extraction dramatically reduced cytotoxicity, indicating that cell death was associated with a lipidaceous virulence factor.

FM 1–43, a probe that remains nonfluorescent in aqueous solution but becomes highly fluorescent when bound to lipid membranes or vesicles (35–37), was used to assess the presence of lipids in the supernatant. As expected, supernatants from both wild-type PA14 and PA14pvdF showed high FM 1–43 fluorescence, indicating the presence of substantial lipids (Fig. 1C). Not all P. aeruginosa isolates secreted considerable amounts of lipids; one of the isolates we tested, JJ692, had nearly 20-fold less FM 1–43 signal compared to PA14 (Fig. 1C). JJ692 is a strain isolated from urinary tract infections (38), which has been employed in research in flagellin glycosylation and bacteriocins (39–42). It is found to have a unique glycosylation of flagella, but it does not affect the virulence of P. aeruginosa (43). It also produces a class of pyocins displaying homology with colicin M, which could target peptidoglycan metabolism and exhibits some antibacterial activity (44). JJ692 exhibits moderate virulence in C. elegans (45).

We surveyed supernatant cytotoxicity and lipid content, as indicated by FM 1–43 fluorescence, in a well-characterized panel of 19 P. aeruginosa clinical and environmental isolates (45). A strong, negative correlation (r = −0.943) was observed in these strains between macrophage survival and the amount of secreted lipids (Fig. 1D). Approximately half of the strains, including JJ692 (Fig. 1D—labeled in red), exhibited minimal toxicity toward RAW264.7 cells. In the subsequent experiments, its pvdL mutant (JJ692pvdL) was employed as a negative control representing strains with low FM 1–43 fluorescence and limited cytotoxicity.

To characterize this unknown toxic lipid, a purification pipeline was developed (Fig. 1E; Fig. S1A). After ammonium sulfate precipitation and subsequent centrifugation, low-density floc was observed floating on top of the PA14pvdF filtrate (Fig. S1B right), which had significantly high FM 1–43 fluorescence (data not shown). However, in the identically treated material from the negative control strain (JJ692pvdL), the corresponding liquid at the top layer had virtually no FM 1–43 signal and no toxicity (data not shown). Instead, a distinct pellet was formed (Fig. S1B left), whose FM fluorescence was low as well. Hence, we performed further purification with segregated materials (floc for PA14pvdF and pellet for JJ692pvdL). Both materials later went through a discontinuous Nycodenz step gradient, from which a distinct layer with high lipid content in the PA14pvdF sample (Fig. S1C, highlighted in red) was collected and reconstituted in phosphate-buffered saline (PBS) prior to further use. This lipid-rich layer was absent from JJ692pvdL sample, so the layer at the same position of JJ692pvdL was collected and reconstituted in the same way, referred to as analogous prep (Fig. S1C, highlighted in blue) in the following experiments.

Confocal laser-scanning microscopy was used to observe the interaction between RAW264.7 cells and this FM 1–43-labeled, lipid-rich material or bacterial filtrate. Both PA14pvdF filtrate and purified material killed ~90% cells within 5 min, as indicated by staining with SYTOX Orange (a cell-impermeant nucleic acid stain) (Fig. 1F; Fig. S1D). Analogous material from JJ692pvdL had minimal effect on cell viability. Immediately, prior to the appearance of bright SYTOX Orange nuclear staining (i.e., cell death), we observed abrupt internalization of FM 1–43. The membrane dye quickly labeled cellular lipids that were soon released from cells, indicating that the purified material caused acute host membrane permeabilization (Fig. 1F; Fig. S1D; Movies S1 and S2).

Structural characterization of cytotoxic P. aeruginosa micelles

We characterized the purified lipid-rich material and macromolecule floc via transmission electron microscopy (TEM). Due to the differing requirements in sample processing for positive and negative staining of TEM, we applied positive staining for floc (solid) (Fig. 2A through C) and negative staining for purified material (aqueous) (Fig. 2D and E). TEM micrographs revealed micellar structures in both samples that were ~30 nm in diameter (Fig. 2F). No such structures were found in the prep from JJ692pvdL (Fig. 2G and H). Micelles were no longer present after chloroform extraction of lipids in PA14pvdF material (Fig. 2I); lipid depletion also ameliorated toxicity toward macrophages (Fig. 1B). These results suggest an association between these micelles and the cytotoxic behavior of P. aeruginosa spent medium and lipid-rich material.

Fig 2.

P. aeruginosa secretes cytotoxic micelles. (A–C) Positively stained macromolecule floc from PA14pvdF visualized by transmission electron microscopy. Digitally zoomed-in view in C. (D and E) Negatively stained purified lipid-rich material from PA14pvdF visualized by TEM. Digitally zoomed-in view in E. (F) Average micelle diameter in bacterial floc and lipid-rich material from PA14pvdF. (G and H) Negatively stained purified analogous prep from JJ692pvdL. Digitally zoomed-in view in H. (I) Negatively stained purified lipid-rich material from PA14pvdF after lipid extraction via chloroform.

P. aeruginosa micelles damage cellular and organellar membranes

We employed TEM to visualize cellular damage during micelle exposure. RAW264.7 cells treated with lipid-deficient material from JJ692pvdL showed a clearly identifiable nucleus, intact plasma membrane, and tubular mitochondria (Fig. 3A through C). In contrast, most cells exposed to purified micelles from PA14pvdF exhibited ruptured plasma membranes and severely disrupted mitochondrial membranes, causing the organelles to become engorged (Fig. 3D through F).

Fig 3.

P. aeruginosa micelles cause severe damage to the plasma membrane and mitochondrial membrane. (A–C) Fixed RAW264.7 cells visualized by TEM after exposure to purified sample from JJ692pvdL. A representative healthy cell in B. Uncompromised mitochondria (yellow arrows) in a healthy cell (C). (D–F) Fixed macrophages visualized by TEM after exposure to purified PA14pvdF micelles. A representative cell with ruptured plasma membrane in E. Detailed view of compromised mitochondria (yellow arrows) in the representative damaged cell (F). (G) Visualization of macrophage plasma membrane after 10 min exposure to purified micelles from PA14pvdF or material from JJ692pvdL. A representative cell (white square) was selected and enhanced for a detailed view of the plasma membrane (green arrow). Cells were prelabeled with Hoechst 33342 (blue) and CellMask Deep Red plasma membrane stain (red). Scale bar = 10 µm. (H) Visualization of macrophage mitochondria after 10 min exposure to purified micelles from PA14pvdF or material from JJ692pvdL. A representative cell (white square) was selected and enhanced for a detailed view of individual mitochondria (yellow arrows). Cells were pre-labeled with Hoechst 33342 (blue) and MitoTracker Red CMXRos (red). Scale bar = 10 µm.

To visualize this phenomenon in real-time, we pre-labeled the plasma membrane and mitochondria using a CellMask deep red plasma membrane stain and MitoTracker Red CMXRos, respectively, and performed time-course confocal microscopy during micelle exposure. The plasma membrane expanded and ruptured within 10 min (Fig. 3G; Fig. S2A). The compromised plasma membrane was consistent with what was observed in electron micrographs (Fig. 3E). Similarly, mitochondria rapidly fragmented and became engorged upon micelle exposure (Fig. 3H; Fig. S2B). This treatment also significantly reduced MitoTracker Red fluorescence (Fig. S2C), likely due to the loss of mitochondrial membrane potential. Together, these results indicate that P. aeruginosa micelles exert their cytotoxicity by damaging cellular and organellar membranes.

P. aeruginosa micelles damage a wide range of host membranes

In addition to murine macrophages, we tested micelle toxicity toward human bronchial epithelial cells (16HBE). PA14pvdF filtrate and purified micelles also killed these cells, while lipid-deficient JJ692pvdL samples remained largely nontoxic (Fig. 4A). Since these micelles damage plasma membranes, their effects on giant plasma membrane vesicles (GPMVs) derived from 16HBE cells (46) were tested. GPMVs are mainly composed of plasma membrane and limited cytosolic content and, thus, can be used to study the interactions between P. aeruginosa micelles and cellular membranes without triggering cell death pathways that may otherwise cause membrane rupture (e.g., necroptosis) (47, 48).

Fig 4.

P. aeruginosa micelles kill eukaryotic and prokaryotic cells. (A) Cytotoxicity of bacterial filtrate and micelles from PA14pvdF or material from JJ692pvdL against RAW264.7 and human bronchial epithelial cells (16HBE). Data in A were analyzed via two-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not statistically significant. (B) Visualization of giant plasma membrane vesicles derived from 16HBE cells after 5 min exposure to micelles from PA14pvdF or material from JJ692pvdL. GPMVs were prelabeled with CellMask Deep Red plasma membrane stain. Scale bar = 10 µm. (C) Hemolysis of erythrocytes on sheep blood’s agar after 10 min or 8 h exposure to micelles from PA14pvdF, material from JJ692pvdL, or PA14pvdF sample after lipid extraction via chloroform. (D) Visualization of Enterococcus faecalis OG1RF::GFP and Staphylococcus aureus USA300 after 4 h exposure to micelles from PA14pvdF or material from JJ692pvdL. (E–G) Comparison between PQS-regulated outer membrane vesicles (OMVs) and micelles (this study) from wild-type PA14. (E) FM 1–43 fluorescence reads in samples after normalized to their initial culture volume, respectively. (F) Cytotoxicity of samples toward macrophages when matching the fluorescence level. (G) Cytotoxicity of samples toward macrophages when matching to the respective initial culture volume. Error bars in A, E, F, and G represent SEM of three biological replicates. Data in A and G were analyzed via two-way ANOVA. Data in E and F were analyzed via t test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not statistically significant.

Purified micelles from PA14pvdF ruptured GPMVs within minutes (Fig. 4B; Fig. S3A), suggesting that micelles cause cell death by directly damaging host membranes rather than by triggering host cells to lyse themselves. We further tested micelle toxicity toward erythrocytes using sheep blood agar since these cells lack mitochondria and nuclei. We observed rapid hemolysis within 10 min upon PA14pvdF micelle treatment (Fig. 4C). Extracting lipids from this sample or using material from JJ692pvdL prevented hemolysis even after 8 h (Fig. 4C). Since P. aeruginosa is known to cause hemolysis through the secretion of hemolytic phospholipase C (PlcH), we also tested whether micelles from a plcH mutant (PA14plcH) could lyse red blood cells. Genetic disruption of plcH did not affect the hemolytic properties of purified micelles (Fig. S3B). Moreover, plcH mutant still had high FM 1–43 fluorescence in spent media and correspondingly high toxicity toward macrophages (Fig. S3C).

To investigate whether P. aeruginosa micelles damaged a wide range of host membranes, including that of noneukaryotic cells, we also measured micelle toxicity toward several microorganisms. The Gram-positive bacteria Enterococcus faecalis and Staphylococcus aureus are frequently co-isolated with P. aeruginosa from biofilms (49–51). Treatment with PA14pvdF-derived micelles caused significant bacterial death in E. faecalis, as indicated by the accumulation of propidium iodide staining. Bacteria treated with material from JJ692pvdL remained mostly viable (Fig. 4D). Micelles also exhibited bactericidal properties against S. aureus (Fig. 4D), albeit at a slightly lower rate. However, these micelles showed little activity against Escherichia coli and Candida albicans (Fig. S3D), likely due to structural differences in the cell walls of Gram-positive bacteria, Gram-negative bacteria, and fungal cells.

P. aeruginosa micelle-mediated damage is distinct from that of PQS-regulated outer membrane vesicles

One major lipidaceous virulence factor of Gram-negative bacteria is outer membrane vesicles (OMVs), which can be produced by P. aeruginosa in the presence of the PQS (20, 22) or under bacterial stress such as lysozyme or antimicrobial treatment (18, 19, 21). These vesicles can not only activate the inflammasome in macrophages (23, 24) but also deliver periplasmic content within the community or to host cells through membrane fusion (25, 26). The OMVs induced by the PQS in brain heart infusion (BHI) medium could package PQS and then traffic this signal molecule within a population, enabling intercellular communication and group behavior (20, 22). Cytotoxic micelles purified as described were compared to PQS-induced OMVs produced using previously established and validated protocols (52–54).

Consistent with previous studies, OMVs produced by P. aeruginosa under this condition were measurable by FM 1–43 fluorescence (Fig. 4E). When normalized to initial culture volume, FM fluorescence of our micelles was nearly 20 times higher than that of OMV sample, suggesting that considerably more lipid micelles are produced than OMVs. Micelle cytotoxicity was also considerably higher; >95% of cells were killed, while OMV cytotoxicity was not detected (Fig. 4F). If equivalent amounts of fluorescent materials were used, cytotoxic micelles and OMVs both exhibited substantial cytotoxicity (Fig. 4G). These data suggest that, although multiple kinds of bacterial lipids can be damaging to eukaryotic cells if present at sufficient concentration, micelles and OMVs are produced in different concentrations and possibly via different mechanisms.

Cytotoxic P. aeruginosa micelles are composed of rhamnolipids

Previous studies have demonstrated that secreted rhamnolipids from P. aeruginosa can induce hemolysis (55, 56), as was observed for the purified micelles (Fig. 4C). To ascertain whether these cytotoxic micelles were comprised of rhamnolipids, partially purified micelles were analyzed via liquid chromatography-mass spectrometry (LC-MS) and compared to two commercially sourced rhamnolipid standards, one enriched in mono-rhamnolipids and the other in di-rhamnolipids (Fig. 5A).

Fig 5.

Rhamnolipid biosynthetic mutants do not produce cytotoxic micelles. (A) Chemical structures of mono-rhamnolipids and di-rhamnolipids. (B) Rhamnolipid biosynthetic pathway in P. aeruginosa. (C) Lipid content in supernatants from PAO1 and PA14 transposon mutants measured by FM 1–43 fluorescence. (D) Cytotoxicity of supernatants from PAO1 and PA14 transposon mutants against RAW264.7 cells. Black labels indicate statistical significance compared to wild-type PAO1. Green labels indicate statistical significance compared to PA14pvdF. (E) Interactions between RAW264.7 cells and bacterial filtrate from wild-type PA14 or PA14rhlB in the presence of SYTOX Orange cell-impermeant nucleic acid stain (red). Secreted bacterial lipids were prelabeled with FM 1–43 (green). (F) Cytotoxicity of two commercially-sourced rhamnolipid products (enriched in mono-rhamnolipids and di-rhamnolipids, respectively) at rhamnolipid concentrations comparable to wild-type PA14 filtrate (standardized by FM 1–43 fluorescence). PA14rhlB filtrate was volume-matched to wild-type PA14 filtrate. Error bars in C, D, and F represent SEM of three biological replicates. Data in C and D were analyzed via one-way ANOVA. Data in F were analyzed via two-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not statistically significant.

More than 60% of the compounds detected in the sample were rhamnolipids (Table 1), two of which corresponded to the predominant species found in the standards (Rha-C10-C10 and Rha-Rha-C10-C10). Analysis identified only a limited number of common membrane lipids detected in LC-MS (Table S1), indicating that these micelles were not derived from membranes.

TABLE 1.

Cytotoxic P. aeruginosa micelles are composed of rhamnolipids

Rank	m/z	Mass	Molecular formula	Rhamnolipid components	Overall vol%	Relative amount%
1	531.35	532.36	C28H52O9	Rha-C10-C12/Rha-C12-C10	8.98	14.52
2	649.38	650.39	C32H58O13	Rha-Rha-C10-C10	8.01	12.95
3	529.34	530.35	C28H50O9	Rha-C10-C12:1/Rha-C12:1-C10	7.49	12.11
4	503.33	504.33	C26H48O9	Rha-C10-C10	6.94	11.22
5	677.41	678.42	C34H62O13	Rha-Rha-C10-C12/Rha-Rha-C12-C10	6.87	11.11
6	517.34	518.35	C27H50O9	Rha-C10-C10-CH3	6.23	10.07
7	475.29	476.30	C24H44O9	Rha-C8-C10/Rha-C10-C8	4.82	7.79
8	675.40	676.40	C34H60O13	Rha-Rha-C10-C12:1/Rha-Rha-C12:1-C10	3.22	5.21
9	489.31	490.31	C25H46O9	Rha-C9-C10/Rha-C10-C9	2.04	3.30
10	663.40	664.40	C33H60O13	Rha-Rha-C10-C10-CH3/Rha-Rha-C10-C11	2.03	3.28
11	557.37	558.37	C30H54O9	Rha-C10-C14:1	1.56	2.52
12	705.44	706.45	C36H66O13	Rha-Rha-C12-C12	1.36	2.20
13	559.39	560.39	C30H56O9	Rha-C12-C12	1.15	1.86
14	621.35	622.36	C30H54O13	Rha-Rha-C8-C10/Rha-Rha-C10-C8	1.14	1.84
Overall					61.84	100.00

To validate these findings, we measured the lipid content and cytotoxicity of spent media from several rhamnolipid biosynthetic mutants from PAO1 and PA14 harboring transposon insertions in rhlA, rhlB, rhlC, rmlA, rmlB, rmlC, or rmlD (data not shown) (Fig. 5B) (57–61). Among these mutants, only rmlB, rhlA, and rhlB failed to produce micelles, or secrete lipids in general, based on the low levels of FM 1–43 fluorescence. These strains also displayed significantly attenuated toxicity toward murine macrophages (Fig. 5C and D). Interestingly, spent media from rhlC mutants, which are unable to convert mono-rhamnolipids to di-rhamnolipids, exhibited neither decreased lipid content nor lowered cytotoxicity (Fig. 5C and D), suggesting that production of mono-rhamnolipids is sufficient for full toxicity. We also studied the effect of PA14rhlB filtrate on RAW264.7 cells via confocal microscopy. Unlike wild-type material, PA14rhlB filtrate did not rupture the host membrane (Fig. 5E; Fig. S4A). In addition, we processed PA14rhlB filtrate through our micelle purification pipeline (Fig. 1E). We observed that the spent medium did not form apparent floc following ammonium sulfate precipitation and centrifugation (Fig. S4B) or the distinct layer with high lipid content after ultracentrifugation (Fig. S4C). The fully purified material from the PA14rhlB mutant, prepared identically to micelles from wild-type (WT) PA14 or PA14pvdF, did not lyse red blood cells (Fig. S4D), affirming that rhamnolipid production is necessary for the secretion of cytotoxic micelles. At rhamnolipid concentrations comparable to that of purified WT PA14 micelles (less than 1 mg/mL, standardized by FM 1–43 fluorescence), commercially-sourced rhamnolipids induced cell death in RAW264.7 macrophages (Fig. 5F).

To investigate whether proteins were involved in micelle formation or contributed to cytotoxicity, LC-MS proteomics were used to analyze partially-purified micelles. Only three proteins from P. aeruginosa were detected. The most abundant was protease IV (Table S2), the iron-regulated, secreted protease PrpL (62). However, spent media from PA14∆prpL exhibited cytotoxicity comparable to that of WT PA14 (Fig. S5).

Rhamnolipid production correlates with P. aeruginosa virulence against murine macrophages

Finally, we investigated the clinical utility of targeting rhamnolipid production during P. aeruginosa infection. We first surveyed rhamnolipid micelle production and the toxicity of spent media from 12 hematological isolates (63). Eleven of these isolates produced detectable amounts of lipid micelles, seven of which were comparable to PA14 and were toxic to murine macrophages, showing 10% survival or less (Fig. 6A). Overall, we observed a significant negative correlation between P. aeruginosa supernatant rhamnolipid content and macrophage survival (Fig. 6A). No correlation was observed between bacterial growth and rhamnolipid production (Fig. S6A).

Fig 6.

Supernatant rhamnolipid content strongly correlates with cytotoxicity in P. aeruginosa clinical isolates. (A) Correlation between rhamnolipid micelle production (FM 1–43 fluorescence) and cytotoxicity against murine macrophages for supernatants from 12 hematological isolates. The three control strains are labeled in gray. Cell survival was normalized to saline control. (B) Correlation between rhamnolipid micelle production (FM 1–43 fluorescence) and cytotoxicity against macrophages for supernatants from 68 clinical isolates from pediatric cystic fibrosis patients. The three control strains are labeled in gray. Cell survival was normalized to saline control.

Taking advantage of a larger panel of 68 clinical isolates from pediatric cystic fibrosis patients (34), this analysis was expanded to a larger group of strains. Nearly half of the isolates in this panel lost rhamnolipid production and also showed little to no toxicity in RAW264.7 cells (Fig. 6B). Similarly, rhamnolipid production strongly correlated to supernatant cytotoxicity and was independent of bacterial growth (Fig. S6B).

We also performed whole-genome sequencing in selected isolates from these panels and combined these data with pre-existing sequencing data from the Broad Institute (45) for the clinical and environmental isolates we tested in Fig. 1D. We compared these strains’ protein sequences for rhamnolipid biosynthetic enzymes (RhlA, RhlB, RhlC, RmlA, RmlB, RmlC, and RmlD) and known quorum-sensing regulators (Vfr, LasR, RsaL, LasI, RhlI, and RhlR) to the reference strain PAO1. Polymorphisms were found in nearly all of these proteins for both high and low rhamnolipid-producing strains. There was an enrichment in rhlR mutations (mostly nonsense or frame-shift mutations) among isolates that had lost the ability to secrete rhamnolipid micelles (Table S3), suggesting that this quorum-sensing system is a crucial regulator of rhamnolipid micelle production.

DISCUSSION

In this study, we demonstrated that P. aeruginosa secretes rhamnolipids that are highly cytotoxic to murine macrophages and blood cells, rupturing their plasma membranes within minutes. TEM analysis showed that the secreted rhamnolipids formed micelles. Using confocal microscopy and TEM, we observed rhamnolipid-mediated destruction of cellular membranes, including plasma and mitochondrial membranes. We also characterized the structural and biochemical properties of rhamnolipid micelles using TEM and LC-MS. Importantly, rhamnolipid micelles damaged a wide-range of host membranes, including those from murine macrophages, human bronchial epithelial cells, erythrocytes, and Gram-positive bacteria, suggesting that these micelles could also modulate the permeability of the human airway epithelium or influence polymicrobial interactions (64, 65). Others have also reported that rhamnolipids exhibit broad-spectrum antimicrobial properties against bacterial (e.g., Klebsiella pneumoniae, Listeria monocytogenes) and fungal (e.g., Mucor circinelloides, Verticillium dahlia) pathogens (66–69), even synergizing with conventional antibiotics (70, 71). We did not notice significant toxicity of rhamnolipid micelles against E. coli or C. albicans, which might be correlated with the difference in the cell wall structures of Gram-positive bacteria, Gram-negative bacteria, and fungal cells. Gram-positive bacteria have a much thicker cell wall made up of peptidoglycan and lack the outer membrane with lipopolysaccharides compared to Gram-negative bacteria (72). Fungal cell wall is more different from these two, mainly composed of mannoprotein outer layer and β-glucan-chitin skeleton (73). These structural differences could lead to their varying vulnerability toward rhamnolipids. In addition, the variations in the original medium for rhamnolipid production, test environment (e.g., agar or liquid, pH, and microorganism density), and other parameters might also play a role in the effectiveness of rhamnolipids.

Rhamnolipids, a class of glycolipids produced by P. aeruginosa, composed of a rhamnosyl head group and 3-(hydroxyalkanoyloxy)alkanoic acid fatty acid tail, have previously been shown to play several roles in virulence. For example, rhamnolipids’ amphiphilic structure allows them to reduce water surface tension, making them a potent biosurfactant (74). They also facilitate P. aeruginosa immune evasion through supporting the development of biofilms (75) and through inhibiting phagocytosis by macrophages and polymorphonuclear leukocytes, even at sublethal concentrations (27, 28, 76). Rhamnolipid production has also been associated with the development of ventilator-associated pneumonia (77).

The rhamnolipids secreted in our study appeared to assemble into micelles. This is consistent with studies that have indicated that rhamnolipids self-assemble into various structures, including micelles, vesicles, lamellar structures, and even mesophases, depending on factors such as concentration, pH, temperature, presence of additives, and sample heterogeneity (congeners) (78, 79). However, the micelles we have characterized here are distinctly smaller (~30 nm in diameter) than those previously characterized (~100–1,000 nm) (80). This discrepancy may have been due to the different media used to grow the pathogens, the different methods for harvesting, and processing of the secreted products. Despite the difference in size, both groups of micelles have been shown to target and kill S. aureus.

Previously, rhamnolipid micelles have been shown to contain small P. aeruginosa metabolites (80). However, LC-MS analysis of our partially purified rhamnolipid material for small molecules and proteomic content suggests that these smaller micelles did not contain any cargo. Instead, toxicity appears to be a consequence of the lipid itself rather than any encapsulated cargo, though the precise mechanism of this phenomenon remains unclear. It will be important to further elucidate how rhamnolipids interact with host membranes, such as the role of surface glycoproteins or membrane domains (i.e., distribution of cholesterol and sphingolipids) and the molecular basis of membrane rupture. These mechanistic studies could provide some insights into how rhamnolipids relate to P. aeruginosa virulence.

Available evidence indicates that the connection of rhamnolipids to virulence appears complex. Since rhamnolipids have been detected in the sputum of cystic fibrosis patients in 1987 (81), several studies have reported rhamnolipid production by P. aeruginosa cystic fibrosis isolates (65, 82). However, while most P. aeruginosa isolates produce rhamnolipids during the acute, early stages of infection, many gradually lose this ability (82). Our observations from a panel of multidrug-resistant isolates from pediatric cystic fibrosis patients showed the same pattern: nearly half of the isolates in this panel failed to produce rhamnolipids (Fig. 6B), likely due to mutations in rhlR (Table S3). This is consistent with a well-known phenomenon, wherein P. aeruginosa undergoes a transition from an acute-to-chronic virulence pattern, often associated with mutations in key quorum-sensing regulators such as LasR or RhlR (82). This phenomenon has been well documented in cystic fibrosis patients, where P. aeruginosa frequently establishes chronic infections that can persist for decades (83, 84).

Interestingly, nearly all strains in a panel of isolates from hematological infections exhibited some level of rhamnolipid production (Fig. 6A). These infections typify acute virulence patterns, as the pathogen requires the secretion of various materials, such as toxins, proteases, and elastases, to colonize primary infection sites and to traverse into the bloodstream. Our surveys of hematological and cystic fibrosis isolates, where macrophage death was strongly correlated to rhamnolipid production, indicate that this material may play a crucial role during certain infections.

This result, combined with the previously described virulence roles for rhamnolipids, may indicate that targeting rhamnolipid production could have a beneficial effect, particularly during acute infection. Targeting virulence determinants during infection is a strategy that has increasingly received attention as a promising alternative or supplement to conventional antimicrobials. This has been particularly true in response to the emergence of multi- or even pan-drug-resistant strains, which has made treating P. aeruginosa infections increasingly challenging.

MATERIALS AND METHODS

Bacterial strains and growth conditions

P. aeruginosa strain PA14 (wild-type), pyoverdine biosynthetic mutant PA14pvdF, and mutant of hemolytic phospholipase C (PA14plcH), as well as the rhamnolipid biosynthetic mutant (PA14rhlB), were all obtained from the UCBPP-PA14 transposon mutant library (85). The mutant with protease IV deletion PA14∆prpL was obtained from Dr. Frederick Ausubel. P. aeruginosa strain PAO1 (wild-type), pyoverdine biosynthetic mutant PAO1pvdF, and its rhamnolipid biosynthetic mutants (rhlA, rhlB, rhlC, rmlA, rmlB, rmlC, and rmlD) were all obtained from PAO1 two-allele transposon mutant library (86, 87). JJ692pvdL was constructed using the pMAR2xT7 vector containing the mariner transposon as previously described (85). Nineteen P. aeruginosa clinical and environmental isolates in Fig. 1 were from Dr. Frederick Ausubel (45). Twelve P. aeruginosa hematological isolates patients were obtained from the OHSU Clinical Microbiology lab after being isolated from BSIs in OHSU HCT/HM patients, provided by Dr. Morgan Hakki (63). Sixty-eight deidentified P. aeruginosa isolates from pediatric cystic fibrosis patients were provided by Dr. Carolyn Cannon (34).

For all experiments, an overnight culture of P. aeruginosa was first grown in LB medium (#BP1426500, Fisher Scientific, ThermoFisher Scientific, Waltham, MA) for 12–14 h. The culture was diluted 1:100 into low-iron M9 medium [1% 5 × M9 salts (#248510, BD Difco, Franklin Lakes, NJ) with 1.3% low-iron Casamino Acids (#223050, Gibco, ThermoFisher Scientific, Waltham, MA)] supplemented with 1 mM MgSO₄ and 1 mM CaCl₂ and incubated for 16–20 h (37°C, 225 rpm). Bacterial growth (absorbance at 600  nm) and FM 1–43 (#T3163, Invitrogen, ThermoFisher Scientific, Waltham, MA) fluorescence were measured using a Cytation5 Multimode Reader (Biotek, Winnoski, VT).

To test micelle toxicity toward microorganisms, C. albicans fRS26::GFP, E. faecalis OG1RF::GFP, E. coli OP50::GFP, and S. aureus USA300 were used (88). fRS26::GFP and OG1RF::GFP were grown in BHI broth (#76345-080, Avantor, VWR, Radnor, PA) for 12–14 h (37°C, 225 rpm). OP50::GFP and USA300 were grown in LB medium for 12–14 h (37°C, 225 rpm).

Rhamnolipid micelle purification

For the preparation of bacterial supernatant, the M9 culture was centrifuged at 13,300 rpm for 15 min. Antibiotics, including amikacin, carbenicillin, and tobramycin (final concentration was 100 µg/mL each), were added to the collected supernatant.

The bacterial filtrate was used for rhamnolipid micelle purification. Similarly, to prepare bacterial filtrate, the overnight M9 culture was first centrifuged at 10,000 rpm for 40 min, later filtered through 0.2 µm PES membrane (#124-0045, Thermo Scientific, ThermoFisher Scientific, Waltham, MA), and supplemented with antibiotics. The macromolecules inside the filtrate would form a layer of floc floating on the top when adding ammonium sulfate to 75%. This material was collected and mixed with 80% Nycodenz (wt/vol) (#AN1002423, Accurate Chemical, Carle Place, NY) to make 40% Nycodenz solution. Nycodenz gradients were then layered into an ultracentrifuge tube at concentrations of 40% (the layer with bacterial material), 20%, 10%, and 0% (PBS only). Gradients were ultracentrifuged at 37,500 rpm for 4 h at 4°C. The harvested lipid-rich material went through 2 h dialysis within 2K molecular-weight cut-off cassettes (#A52961, Thermo Scientific, ThermoFisher Scientific, Waltham, MA) to be reconstituted into PBS. The final product was stored at −80°C.

For Proteinase K treatment, PA14pvdF supernatant was incubated with Proteinase K (#V3021, Promega, Fitchburg, WI) (working concentration: 100 µg/mL) at 37°C for 24 h. The reaction was later stopped by adding PMSF (#36978, Thermo Scientific, ThermoFisher Scientific, Waltham, MA) (working concentration: 5 mM) and incubating at room temperature for 1 h. For lipid extraction, PA14pvdF supernatant or rhamnolipid micelles were 1:1 (volume to volume) mixed with chloroform with vortex and later centrifuged at 13,300 rpm for 15 min. The top layer was collected as lipid-extracted material.

PQS-regulated OMV purification

PQS-regulated OMVs were generated from wild-type PA14 as described previously (52–54). Briefly, PA14 was grown in BHI broth for 12 h (37°C, 250 rpm). The culture was centrifuged at 15,000 × g at 4°C for 15 min to pellet the cells, and the supernatant was filtered through a 0.45-µm syringe filter. The supernatant was centrifuged at 200,000 × g at 4°C for 1.5 h to pellet out the OMVs. The pellet was resuspended in 500 µL of MV Buffer (50 mM Tris, 5 mM NaCl, 1 mM MgSO4, pH 7.4). Routine analysis of samples (as previously described) (52–54) for the presence of PQS and lipopolysaccharide (outer membrane marker) and the absence of succinate dehydrogenase activity (inner membrane marker) demonstrates that this procedure is effective at purifying bona fide OMVs that are not contaminated by cell lysis debris. Those rhamnolipid micelles used for comparison, which were labeled as “Micelles (this study),” were also from wild-type PA14.

Cell culture

Murine macrophages (RAW264.7) were maintained at 37°C in 5% CO₂ in RPMI-1640 medium (#R8758, Millipore Sigma, St. Louis, MO) containing 10% bovine calf serum (#12133C, Millipore Sigma, St. Louis, MO) and 1% penicillin-streptomycin (P/S) (#P4333, Millipore Sigma, St. Louis, MO). Human bronchial epidermal cells (16HBE), which have been immortalized by SV40 large T-antigen, were cultured at 37°C in 5% CO₂ in MEM medium (#M4655, Millipore Sigma, St. Louis, MO) supplemented with 10% fetal bovine serum (#35011CV, Corning, Corning, NY), 1% nonessential amino acids (NEAA) (#M7145, Millipore Sigma, St. Louis, MO), and 1% P/S.

Cell viability assay

The viability of each cell was quantified using alamarBlue HS Cell Viability Reagent (#A50100, Invitrogen, ThermoFisher Scientific, Waltham, MA). Cells were seeded in a 24-well plate (about 1 million/well) overnight and used when reaching 80% confluence. Before adding sample to cells, the culture medium was replaced with serum-free medium (RPMI-1640 medium containing 1% P/S or MEM medium with 1% NEAA and 1% P/S, respectively). In most experiments, tested samples were first diluted in PBS to designed concentrations, 150 µL of which was added into 350 µL serum-free medium in each well. The 24-well plate was incubated at 37°C in 5% CO₂ for 4 h. And then, 50 µL alamarBlue reagent was added into each well. After incubation for another 1 h, the fluorescence at 590 nm was measured using a Cytation5 Multimode Reader and normalized to the well with 150 µL PBS. The cytotoxicity data of 19 common isolates were normalized to M9 medium.

For comparison with PQS-regulated OMVs, when matching FM 1–43 fluorescence, 250 µL of sample was added to 250 µL serum-free medium in each well. The 24-well plate was incubated at 37°C in 5% CO₂ for 3 or 7 h. The cytotoxicity data were normalized to the well with 250 µL MV buffer. When matching initial culture volume, 25 µL rhamnolipid micelle sample or 42 µL OMV sample was added into serum-free medium in each well (final volume: 500 µL). The 24-well plate was incubated at 37°C in 5% CO₂ for only 1 h. The cytotoxicity data were normalized to the well with 42 µL MV buffer.

Fluorescence imaging

The fluorescence images of cells were taken via Zeiss LSM800 Airyscan fluorescence microscopy. RAW264.7 cells were seeded in an 8-well plate (#155409, Thermo Scientific, ThermoFisher Scientific, Waltham, MA) (about 0.75 million/well) overnight and used when reaching 90% confluence. The cell medium for imaging was similar to the one for viability assay, 70% serum-free medium and 30% test sample at designed concentrations.

The working concentrations of each dye for cells: Hoechst 33342 (#62249, Thermo Scientific, ThermoFisher Scientific, Waltham, MA), 40 µM in serum-free medium when imaging; SYTOX Orange (#S11368, Invitrogen, ThermoFisher Scientific, Waltham, MA), 10 µM in serum-free medium when imaging; CellMask deep red plasma membrane stain (#C10045, Invitrogen, ThermoFisher Scientific, Waltham, MA), 10 µg/mL in serum-free medium (rinsed before imaging); MitoTracker Red CMXRos (#M7512, Invitrogen, ThermoFisher Scientific, Waltham, MA), 1 µM in serum-free medium (rinsed before imaging). Working concentrations of FM 1–43 to pre-stain rhamnolipid micelle samples (including bacterial supernatant/filtrate, purified rhamnolipid micelles, or JJ692pvdL material) were 20 µg/mL.

The actual color of SYTOX Orange is orange (Ex: 547 nm, Em: 570 nm); FM 1–43 is orange as well (Ex: 473 nm, Em: 579 nm). Here we assigned SYTOX Orange as red and FM 1–43 as green to avoid confusion in images. Images were exported and quantified to determine the area of red fluorescence using an image-processing package Fiji.

Transmission electron microscopy

For PA14pvdF micelles and JJ692pvdL material, as well as lipid-extracted micelles, freshly-made samples using the purification pipeline above were negatively stained with 3% uranyl acetate (#22400, Electron Microscopy Sciences, Hatfield, PA). For TEM of macrophages, cells were incubated with PA14pvdF micelles and JJ692pvdL material for 1 min, fixed with Karnovsky’s Fixative (#15732-10, Electron Microscopy Sciences, Hatfield, PA) overnight and pelleted for TEM analysis. Pelleted samples were then post-fixed for 1 h in 1% osmium tetroxide (#19100, Electron Microscopy Sciences, Hatfield, PA), dehydrated in a graded series of ethanol, embedded in Embed812 epoxy resin (#14120, Electron Microscopy Sciences, Hatfield, PA), and heat polymerized overnight at 70°C. Samples were then sectioned at 100 nm thickness using a Leica EM UC7 ultramicrotome. Sections were positively stained with saturated methanolic uranyl acetate and Reynold’s lead citrate (#22410, Electron Microscopy Sciences, Hatfield, PA). Both negatively- and positively-stained samples were imaged using a JEOL JEM-1230 TEM operating with 80 kV of accelerating voltage and equipped with an AMT NanoSprint15 mKII sCMOS camera.

Toxicity toward microorganism

Toxicity of rhamnolipid micelles toward microorganisms was visualized using a fluorescent microscope (Zeiss Axio Imager M2). Twenty-five microliters of rhamnolipid micelles was added into 75 µL media from an overnight culture and mixed. The mixture was incubated at 37°C with 225 rpm shaking for 4 h and then centrifuged at 13,300 rpm for 5 min. The pellet was resuspended in 25 µL PBS with 2 µg/mL propidium iodide for fRS26::GFP, OG1RF::GFP, or OP50::GFP. The pellet of USA 300 was resuspended in 50 µL PBS with 40 µM acridine orange (#A1301, Invitrogen, ThermoFisher Scientific, Waltham, MA) and 2 µg/mL propidium iodide (#CDX-P0023, AdipoGen Life Sciences, San Diego, CA). Micrographs of microorganisms were taken via Zeiss Axio Imager M2 fluorescence microscopy. Three biological replicates were performed.

GPMV formation

GPMVs were generated from 16HBE cells as described previously (46). Briefly, these cells were first gently rinsed with vesiculation buffer (150 mM NaCl, 2 mM CaCl₂, and 20 mM HEPES in water, pH 7.4) twice, incubated with active vesiculation buffer [1.9 mM DTT and 27.6 mM formaldehyde (HCHO) in vesiculation buffer] for 4 h, and followed by centrifugation at 500 × g for 5 min to remove cellular debris.

To concentrate GPMVs, these vesicles were stained with 5 µg/mL CellMask deep red plasma membrane stain and later centrifuged at 13,300 rpm for 15 min. The supernatant was aspirated, while the pellet was resuspended in PBS and ready for fluorescence imaging.

Blood agar culture

Blood agar (TSA with sheep blood) medium (#R01200, Thermo Scientific, ThermoFisher Scientific, Waltham, MA) was utilized here. Aqueous samples, like micelles, were dropped onto the surface (8 µL each droplet). PBS here was used for dilution and also dropped on the surface. After droplets dried up, the plate was transferred to 37°C and incubated for 8 h. Three biological replicates were performed.

Proteomics study and rhamnolipid study

Partially-purified rhamnolipid micelles (floc) were prepared as described above and utilized for LC-MS analysis. All LC-MS analysis was carried out on an Agilent 6545XT qToF mass spectrometer that was interfaced with an Agilent 1290 Infinity ii chromatography system through a Jet Stream Electrospray Ionization source.

For proteomic analysis, samples were diluted to a final concentration of ~1 µg/µL in 100 mM ammonium bicarbonate buffer (pH ~7.8). Samples were reduced with tris(2-carboxyethyl)phosphine at 55°C for 1 h and then alkylated with iodoacetamide at room temperature for 1 h, protected from light. Samples were then digested with Trypsin/Lys-C (#V5071, Promega, Madison, WI) overnight for ~18 h and subjected to LC-MS/MS analysis. The LC injection volume was 20 µL corresponding to ~20 µg of total digest loaded on column. The LC separations were carried out on a HALO 160 Å ES-C18, 2 µm, 2.1 × 150 mm column operated at 35°C and a flow rate of 0.4 mL/min. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. The gradient was run from 2% B to 95% B over 86 min as follows: initial conditions 2% B held at 2% B from 0 to 2.5 min, 2.5–5 min 8% B, 5–33 min 15% B, 33–73 min 35% B, 73–79 min 65% B, 79–82 min 95% B, held at 95% B until 86 min. At the end of the run, the column was re-equilibrated to initial conditions (2% B) for 4 min. MS data were collected in the positive ionization mode using AutoMS2. MS data were collected over a range of 300–1,500 m/z at a scan rate of 8 spectra/s. MS/MS data were collected over a range of 100–1,700 m/z at a scan rate of 4 spectra/s and an isolation width of 4 amu. Collision energy was selected by the MassHunter acquisition software based on z and m/z values. Data analysis and protein database searching were carried out using MASCOT v. 2.7 (Matrix Science, London, UK). Searches were run against the UniRef100 database restricted by taxonomy to Bacteria (Eubacteria).

For rhamnolipid analysis, samples were diluted to a final concentration of ~0.5 µg/µL in 10 mM ammonium formate buffer (pH ~3.5). The LC injection volume was 2 µL corresponding to ~1 µg of total sample on column. The LC separations were carried out on an Agilent Eclipse Plus C18 RRHD, 1.8 µm, 2.1 × 50 mm column operated at 40°C and a flow rate of 0.4 mL/min. Mobile phase A was 10 mM ammonium formate in water (pH ~3.5), and mobile phase B was methanol. The gradient was run from 25% B to 95% B over 25 min as follows: initial conditions 25% B, 0–20 min 95% B, 20–25 min hold 95% B. At the end of the run, the column was re-equilibrated to initial conditions (25% B) for 3 min. MS data were collected in the negative ionization mode. MS data were collected over a range of 100–1,100 m/z at a scan rate of 6 spectra/s. MS/MS data were collected over a range of 75–1,100 m/z at a scan rate of 2 spectra/s and an isolation width of 4 amu. Collision energies of 10, 20, and 30 were used for MS/MS data acquisition. Data analysis and database searching were carried out using MassHunter Qualitative Analysis v.10. Initial rhamnolipids were identified by searching the Agilent METLIN PCD Lipids Database v.8. Additional rhamnolipids were identified through manual data inspection.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/iai.00407-23.

Supplemental figures and tables. iai.00407-23-s0001.pdf.

Figures S1 to S6, Tables S1 to S3, and legends for Movies S1 and S2.

Movie S1. iai.00407-23-s0002.wmv.

Interactions between RAW264.7 cells and material from JJ692pvdL.

Movie S2. iai.00407-23-s0003.wmv.

Interactions between RAW264.7 cells and purified lipid-rich material from PA14pvdF.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.