Biofilms Formed by Gram-Negative Bacteria Undergo Increased Lipid A Palmitoylation, Enhancing In Vivo Survival

To identify potential modifications of LPS in E. coli biofilms, we compared patterns of rough LPS (Ra, with no O antigen) produced in biofilm and planktonic E. coli K-12 BW25113 bacteria. Tricine SDS-PAGE analysis revealed that LPS extracted from 96-h mature biofilm grown in microfermentors with constant medium renewal had a higher molecular weight than LPS species extracted from 15- to 24-h overnight stationary-phase planktonic culture (Fig. 1A; see Fig. S1 in the supplemental material). An additional band was progressively detected along with aging biofilms and disappeared in LPS extracted from 96-h biofilm bacteria reinoculated under planktonic conditions (Fig. 1A). We also observed an additional LPS band associated with biofilms formed by various pathogenic E. coli strains (Fig. 1B). To further investigate the correlation between biofilm formation and appearance of the additional LPS band, we compared the E. coli BW25113 strain with its closely related derivative, the BW25113 F strain, forming more biofilm than BW25113 due to the presence of the biofilm-promoting F conjugative plasmid (Fig. 1C) (16). Analysis of LPS extracted from biofilms formed by E. coli BW25113 and BW25113 F strains at different time points showed that the additional LPS band appeared more rapidly in the BW25113 F strain (48 h versus 72 h) (Fig. 1D). Whereas this suggested a link between biofilm capacity and LPS modification, monitoring LPS profiles in planktonic cultures (without medium renewal) also revealed the emergence of the additional LPS band in aging planktonic culture (48 h and beyond) (see Fig. S2A in the supplemental material). In contrast, the regular renewal of the growth medium in these very late stationary-phase cultures, performed to limit nutrient exhaustion and approximate microfermentor conditions, significantly reduced the appearance of this band (Fig. S2B). These results therefore suggested that the studied LPS modification occurs in mature biofilms and under conditions created in unusually long planktonic cultures.

The presence of the additional band in LPS extracted from rough E. coli K-12, and enteropathogenic E. coli strains 55989 and O42 and uropathogenic 536 and CFT073 smooth E. coli strains (with O antigen) suggested that the modification occurs in the smallest common LPS part in all tested strains, corresponding to lipid A and the inner core. Moreover, detection of the additional band in LPS extracted from 96-h biofilms formed by E. coli K-12 BW25113 waaC deep rough mutant suggested that the biofilm-associated modification could take place on lipid A bound to a 3-deoxy-d-manno-octulosonic acid (Kdo) disaccharide part of the LPS (lipid A-Kdo₂) (Fig. 2A and B). Of the E. coli enzymes potentially involved in chemical modifications of the lipid A-Kdo₂ portion, four are common to E. coli 55989, O42, 536, CFT073, and K-12 strains: ArnT (amino-arabinose addition to lipid A), EptA (phosphoethanolamine addition to lipid A), EptB (phosphoethanolamine addition to Kdo) and PagP (palmitate addition to lipid A) (17). We analyzed LPS extracted from planktonic (24-h) and biofilm (96-h) bacteria corresponding to arnT, eptA, eptB, and pagP E. coli mutants and detected the biofilm-associated LPS band in biofilm formed by all mutants except the pagP mutant (data not shown and Fig. 2B). Complementation of the E. coli pagP mutant by a plasmid expressing pagP restored LPS modification in 96-h biofilm bacteria (Fig. 2B). These results suggested lipid A palmitoylation in biofilm bacteria (18). Matrix-assisted laser desorption ionization−time of flight mass spectrometry (MALDI-TOF MS) analysis of lipid A in 24-h planktonic bacteria showed one major peak at m/z = 1,797.4, corresponding to a hexa-acylated nonpalmitoylated lipid A species and only a minor species at m/z = 2,035.6, corresponding to a hepta-acylated palmitoylated lipid A species (Fig. 3A and B). In contrast, the spectrum obtained for 96-h biofilm showed two major peaks, at m/z = 1,797.4 and m/z = 2,035.6, thereby indicating significant lipid A palmitoylation under biofilm conditions. We did not detect palmitoylated lipid A in E. coli ΔpagP biofilm bacteria (Fig. 3D), while production of palmitoylated lipid A was restored upon complementation by plasmid pPagP under planktonic and biofilm conditions (Fig. 3E and F). MALDI-TOF MS analysis of LPS extracted from E. coli waaC deep rough (Re) mutant grown under planktonic condition already showed significant lipid A palmitoylation, potentially due to increased membrane stress-dependent pagP expression (19). However, comparison between planktonic and biofilm bacteria showed that the only detected modification occurring in biofilm is a difference at m/z = 2,035 and m/z = 2,475 peaks, corresponding to palmitoylated lipid A and palmitoylated lipid A-Kdo₂, respectively (see Fig. S3 in the supplemental material). Finally, biofilm-associated lipid A palmitoylation is a general feature, since MALDI-TOF mass spectrometry analysis of lipid A extracted from biofilms formed by several pathogenic E. coli strains as well as various Gram-negative bacteria, including Serratia marcescens, Pseudomonas aeruginosa, Klebsiella pneumoniae, and Citrobacter koseri species, consistently showed increased levels of lipid A palmitoylation compared to corresponding planktonic cultures (Table 1 and Fig. S4A to S4D).

To investigate regulation of lipid A palmitoylation in biofilm, we introduced a lacZ transcriptional fusion downstream of the pagP stop codon in E. coli BW25113 (transcriptional operon fusion) and observed a 3-fold increase in β-galactosidase activity in 96-h biofilm compared to planktonic culture (Fig. 4A). Moreover, constitutive expression of pagP in E. coli BW25113 PcL-pagP led to production of the additional LPS band under both planktonic and biofilm conditions (Fig. 4B), suggesting that lipid A palmitoylation is regulated at the transcriptional level. Previous studies involved PhoP and EvgA as positive regulators of pagP expression in response to magnesium limitation (19, 20); however, deletion in these 2 genes had no effect on pagP expression or on the LPS profile, demonstrating that biofilm-associated palmitoylation is PhoP and EvgA independent (Fig. 4A and C). Interestingly, the emergence of the additional LPS band under very late stationary-phase conditions (96 h of planktonic culture) is also PhoP and EvgA independent and is not inhibited upon supplementation with excess magnesium in the culture medium (see Fig. S5A in the supplemental material). Moreover, inactivation of various stress response regulators, including cpxR, rcsB, pspF, oxyR, soxR, arcA, rpoS, relA, and luxS, had no impact on lipid A palmitoylation (Fig. S5B). To identify regulators of pagP expression in E. coli, we took advantage of the white color displayed by E. coli BW25113 pagP-lacZ colonies on 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) agar plates, and we used TnSC189 mariner-based transposon mutagenesis to screen for blue BW25113 pagP-lacZ mutants derepressed for pagP expression. We identified 7 dark blue colonies corresponding either to trivial TnSC189 insertion upstream of the pagP-lacZ fusion or to insertion into the hns gene, which encodes the global silencing protein repressor H-NS (21). To confirm the role of H-NS in pagP expression, we monitored pagP expression in E. coli BW25113 Δhns pagP-lacZ planktonic cultures and observed 18-fold-increased pagP expression in the hns deletion mutant compared to the parental strain (Fig. 4A). Consistent with this result, the previously biofilm-associated LPS band could be detected in 24-h planktonic bacteria in LPS extracted from E. coli Δhns pagP-lacZ mutant (Fig. 4C). Since the ability of this mutant to form a biofilm was strongly affected, we could not meaningfully monitor pagP expression or LPS palmitoylation in biofilms. Complementation of the Δhns pagP-lacZ mutant with plasmid pBAD33-hns decreased expression of pagP (Fig. S6A), therefore demonstrating that H-NS represses lipid A palmitoylation in E. coli planktonic bacteria.

Increased pagP expression under mature biofilm conditions suggested potential alleviation of H-NS repression by another E. coli regulator. Analysis of the pagP promoter region actually revealed three sequences closely matching the proposed consensus binding site for SlyA, an anti-H-NS factor that antagonizes H-NS binding on a number of cell envelope E. coli genes (see Fig. S6B in the supplemental material) (22–25). We therefore tested the contribution of SlyA to pagP regulation and compared pagP expression in E. coli BW25113 pagP-lacZ and BW25113 ΔslyA pagP-lacZ both under planktonic and biofilm conditions. We observed that a slyA deletion very significantly reduced pagP induction in biofilm, with only a 1.3-fold induction of pagP expression in the ΔslyA background, compared to 2.9-fold induction observed in the wild-type (WT) background (Fig. 4A). Consistent with this result, analysis of LPS extracted from the ΔslyA mutant showed only a minor modification of the LPS profile between planktonic and biofilm conditions (Fig. 4C). However, complementation of the slyA mutation in E. coli ΔslyA pagP-lacZ mutant with plasmid pCA24N-slyA partially restored biofilm-associated induction of pagP expression (Fig. S6C). Finally, we compared slyA expression under planktonic and biofilm conditions, and we observed a 1.7-fold induction of slyA expression in biofilm, therefore confirming the role of SlyA in biofilm-associated pagP expression (Fig. 4D).

To investigate the role of lipid A palmitoylation in biofilm bacteria, we first compared E. coli wild-type and ΔpagP and PcL-pagP mutant capacities to form in vitro biofilm on microtiter plates or continuous-flow microfermentors. We observed that neither the lack of, nor constitutive PagP-dependent palmitoylation, affected commensal E. coli BW25113 or pathogenic E. coli 55989 in vitro biofilm formation (see Fig. S7A and S7B in the supplemental material). We then used our previously described in vivo rat model of biofilm-associated infection in a totally implanted venous access port (TIVAP) (26) to compare the extent of in vivo biofilm development of bioluminescent wild-type, ΔpagP, or PcL-pagP E. coli 55989 derivatives in an inoculated TIVAP (Fig. 5A). Monitoring of bioluminescent biofilm biomass formed in an implanted TIVAP during the 7 days of the experiment showed a decrease in luminescence in the TIVAP colonized by E. coli 55989 ΔpagP compared to wild-type and PcL-pagP strains (Fig. 5B and C). However, inoculated TIVAP showed no initial difference in CFU recovered 3 h after inoculation, indicating that the lack of lipid A palmitoylation has no effect on initial adhesion on TIVAP in vivo (Fig. S7C). Nevertheless, we observed a significant decrease in 55989 ΔpagP bacterial number at day 7 compared to TIVAP colonized with wild-type 55989 bacteria (Fig. 5D). Since in vivo formation of E. coli 55989 biofilm in the chamber of the implanted device also led to PagP-dependent modifications of the LPS profile (Fig. 5E), we investigated the potential contribution of lipid A palmitoylation to control of biofilm infection dynamics by the host.

Lipid A palmitoylation in E. coli was previously shown to decrease the inflammatory activity of lipid A (27). To test in a controlled manner whether nonpalmitoylated E. coli 55989 ΔpagP lipid A triggers a higher inflammatory response than palmitoylated biofilm bacteria, we injected E. coli 55989 wild-type ΔpagP and PcL-pagP biofilm bacteria intravenously into rats. Two hours after injection, we used rat serum enzyme-linked immunosorbent assay (ELISA) to measure the amount of interleukin-6 (IL-6) proinflammatory cytokine and observed a significant increase in the IL-6 level induced by E. coli 55989 ΔpagP biofilm bacteria compared to wild-type and PcL-pagP bacteria (Fig. 6A). We consistently observed a 2-fold reduction in the release of IL-6 by the macrophage when brought into contact with bacteria producing palmitoylated lipid A (Fig. 6B). Moreover, palmitoylated E. coli biofilm bacteria also displayed resistance to the antimicrobial peptide protegrine-1 (PG-1) compared to unpalmitoylated bacteria (see Fig. S8 in the supplemental material). Taken together, these results showed that lipid A palmitoylation increased biofilm bacteria survival in vivo, potentially by damping the inflammatory response to and host immune defenses against biofilm bacteria.

Animals were housed in the Institut Pasteur animal facilities, accredited by the French Ministry of Agriculture for performing experiments on live rodents (permit A-75-1061). Work on animals was performed in compliance with French and European regulations on care and protection of laboratory animals (European Commission directive 2010/63; French law 2013-118, 6 February 2013). The protocols used in this study for the animal model, catheter placement, in vivo biofilm formation, and in vivo study of inflammatory responses were approved by the ethics committee of “Paris Centre et Sud no. 59” under reference no. 2012-0045.

Bacterial strains used in this study are described in Table S1 in the supplemental material. Antibiotics were used at the following concentrations: kanamycin, 50 µg/ml; chloramphenicol, 25 µg/ml; ampicillin, 100 µg/ml; and zeocin, 50 µg/ml. Biofilm formation in continuous-flow microfermentors was performed as previously described in reference 16. Briefly, continuous-flow microfermentors containing a removable glass spatula were used as described (http://www.pasteur.fr/recherche/unites/Ggb/matmet.html) to maximize biofilm development and minimize planktonic growth. Inoculation was performed by dipping the glass spatula for 2 min in a culture adjusted to an optical density at 600 nm (OD₆₀₀) of 1 from overnight bacterial cultures grown in M63B1 minimal medium [KH₂PO₄ 100 mM, (NH₄)₂SO₄ 15 mM, MgSO₄ 0.8 mM, vitamin B1 3 µM, pH 7.4] supplemented with 0.4% glucose and the appropriate antibiotics. The spatula was then reintroduced into the microfermentor, and biofilm culture was performed at 37°C in M63B1 with glucose. Biofilm biomass produced at different time points was rapidly resuspended in 15 ml of microfermentor medium (OD₆₀₀ < 0.01), and biofilm bacterial LPS was analyzed after centrifugation and elimination of the resuspension supernatant.

Planktonic and biofilm cultures were performed in M63B1 medium supplemented with 0.4% glucose at 37°C.

TnSC189 mariner-based transposon of E. coli pagP-lacZ was performed as described in reference 44. Transposon insertion sites were determined as described in reference 45. Constitutive expression of the pagP gene was carried out by insertion of the previously described ampPcL genetic element in front of pagP at its native chromosomal location, leading to the constitutive expression of the pagP gene from the phage lambda constitutive promoter (λPr) (23). Insertion of the ampPcL cassette and deletion of the pagP gene were performed using the λ-Red recombinase gene replacement system and a three-step PCR procedure described previously (46, 47). The primers used are listed in Table S2 in the supplemental material. When necessary, the antibiotic resistance marker of the inserted cassette was removed using the flippase-encoding pCP20 plasmid. For construction of pagP-lacZ and slyA-lacZ fusions, the same principle was used. The lacZ-zeo cassette was inserted downstream of the stop codon of the target gene. To construct pagP-lacZ derivatives, mutations were transferred by P1vir transduction from Keio collection E. coli JW1116 (ΔphoP), JW2366 (ΔevgA), JW1225 (Δhns), and JW5267 (ΔslyA) (48) mutants into the E. coli BW25113 pagP-lacZ strain. All constructs were confirmed by PCR and sequencing.

Bacteria (10⁸) were pelleted and resuspended in 100 µl of lysing buffer (Bio-Rad) containing 1% SDS, 20% glycerol, 100 mM Tris (pH 6.8), and Coomassie blue G-250. Lysates were heated at 100°C for 10 min; then, proteinase K was added at 1 mg/ml and incubated at 37°C for 1 h. These samples (4 µl) were electrophoresed in a tricine SDS-PAGE system, which improves resolution of the low-molecular-weight LPS band, using a 4% stacking gel and a 20% separating gel (49). LPSs were then visualized by the periodate-silver staining method adapted from reference 50. Gels were immersed in fixing solution (30% ethanol, 10% acetic acid) for 1 h, washed three times for 5 min each time in water, oxidized in 0.7% periodic acid for 10 min, and washed three times for 5 min each time in water. Gels were then immersed for 1 min in 0.02% thiosulfate pentahydrate, rinsed quickly in water, and stained in 25 mM silver nitrate for 10 min. After a 15-s wash in water, gels were developed in 3.5% potassium carbonate and 0.01% formaldehyde. Development was stopped in 4% Tris base and 2% acetic acid for 30 min.

Lipid A was isolated directly by hydrolysis of bacterial cells as described in references 51 and 52. Briefly, 5 mg of lyophilized bacterial cells was suspended in 100 µl of a mixture of isobutyric acid−1 M ammonium hydroxide (5:3 [vol/vol]) and kept for 1.5 h at 100°C in a screw-cap test tube in a Thermomixer system. The suspension was cooled in ice water and centrifuged (2,000 ×  g, 5 min). The recovered supernatant was diluted with 2 volumes of water and lyophilized. The sample was then washed once with 100 µl of methanol (by centrifugation at 2,000 × g for 5 min). Finally, lipid A was extracted from the pellet in 50 µl of a mixture of chloroform, methanol, and water (3:1.5:0.25 [vol/vol/vol]).

LPS samples were dispersed in water at 1 µg/µl. Lipid A extracts in chloroform-methanol-water were used directly in this mixture of solvents. In both cases, a few microliters of sample solution was desalted with a few grains of ion-exchange resin Dowex 50W-X8 (H⁺). Aliquots of 0.5 to 1 µl of the solution were deposited on the target, and the spot was then overlaid with matrix solution and left to dry. Dihydroxybenzoic acid (DHB) (Sigma-Aldrich) was used as the matrix. It was dissolved at 10 mg/ml in 0.1 M citric acid solution in the same solvents as those used for the analytes (53). Different analyte/matrix ratios (1:2, 1:1, and 2:1 [vol/vol]) were tested to obtain the best spectra. Negative- and positive-ion mass spectra were recorded on a PerSeptive Voyager-DE STR time of flight mass spectrometer (Applied Biosystems) in the linear mode with delayed extraction. The ion-accelerating voltage was set at [minus]20 kV, and the extraction delay time was adjusted to obtain the best resolution and signal-to-noise ratio.

To determine the level of β-galactosidase enzyme activity, bacteria were grown in M63B1 minimal medium supplemented with 0.4% glucose for 24 h under planktonic conditions or for 96 h under continuous-flow biofilm conditions. β-Galactosidase activity was assayed in triplicate as described previously (54) and expressed in Miller units.

Totally implanted venous access ports (TIVAPs) were surgically implanted in CD/SD (IGS:Crl) rats (Charles River) as described previously (26). Briefly, the port was implanted at dorsal midline toward the lower end of the thoracic vertebrae, and the catheter was inserted into the jugular vein. Prior to inoculation of clinical strains, all rats were checked for the absence of infection by plating 100 µl blood and by monitoring rats for the absence of luminescence signals.

The inoculum dose of 10⁴ cells for overnight grown cultures of E. coli 55989 pAT881 wild type (WT), ΔpagP, or PcL-pagP were injected into the port in a 50-µl volume. Planktonic bacteria were removed after 3 h of injection. Progression of colonization was monitored using an IVIS100 imaging system. Rats were sacrificed either 3 h or 7 days postinjection, and TIVAPs were harvested. Serial dilutions from samples were plated on LB agar medium for enumerating CFU/ml.

To estimate the host inflammatory response due to LPS palmitoylation, E. coli 55989 WT, ΔpagP, or PcL-pagP grown in a microfermentor for 4 days were adjusted to 10⁹ cells per 500 µl and injected into rats intravenously. Rats were sacrificed 2 h after injection, and blood was harvested aseptically and analyzed for IL-6 cytokine release in serum using ELISA.

Two-tailed unpaired Student t test analyses were performed using Prism 5.0 for Mac OS X (GraphPad Software, Inc.). Each experiment was performed at least 3 times. Statistical significance was indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.