The impact of plasma membrane lipid composition on flagella-mediated adhesion of enterohemorrhagic Escherichia coli

Enterohemorrhagic Escherichia coli (EHEC) O157:H7 is a major cause of foodborne gastrointestinal illness. The adhesion of EHEC on host tissues is the first step enabling bacterial colonization. Adhesins like fimbriae and flagella mediate this mechanism. Here, we studied the interaction of the bacterial flagellum with the host cell’s plasma membrane using Giant Unilamellar Vesicles (GUVs) as a biologically relevant model. Cultured cell lines contain many different molecular components including proteins and glycoproteins. In contrast, with GUVs we can characterize the bacterial mode of interaction solely with a defined lipid part of the cell membrane. Bacterial adhesion on GUVs was dependent on the presence of the flagellar filament and its motility. By testing different phospholipid head groups, the nature of the fatty acid chains or the liposome curvature, we found that lipid packing is a key parameter to enable bacterial adhesion. Using HT-29 cells grown in the presence of polyunsaturated fatty acid (α-linolenic acid) or saturated fatty acid (palmitic acid), we found that α-linolenic acid reduced adhesion of wild type EHEC but not of a non-flagellated mutant. Finally, our results reveal that the presence of flagella is advantageous for the bacteria to bind to lipid rafts. We speculate that polyunsaturated fatty acids prevent flagellar adhesion on membrane bilayers and play a clear role for optimal host colonization. Flagella-mediated adhesion to plasma membranes has broad implications to host-pathogen interactions. Importance Bacterial adhesion is a crucial step to allow bacteria to colonize their hosts, invade tissues and form biofilm. Enterohemorrhagic E. coli O157:H7 is a human pathogen and the causative agent of diarrhea and hemorrhagic colitis. Here, we use biomimetic membrane models and cell lines to decipher the impact of lipid content of the plasma membrane on enterohemorrhagic E. coli flagella-mediated adhesion. Our findings provide evidence that polyunsaturated fatty acid (α-linolenic acid) inhibits E. coli flagella adhesion to the plasma membrane in a mechanism separate from its antimicrobial and anti-inflammatory functions. In addition, we confirm that cholesterol-enriched lipid microdomains, often called lipid rafts are important in bacterial adhesion. These findings significantly strengthen plasma membrane adhesion via bacterial flagella in an important human pathogen. This mechanism represents a promising target for the development of novel anti-adhesion therapies.

Giant Unilamellar Vesicles (GUVs) as a biologically relevant model. Cultured cell lines contain 23 many different molecular components including proteins and glycoproteins. In contrast, with 24 GUVs we can characterize the bacterial mode of interaction solely with a defined lipid part of 25 the cell membrane. Bacterial adhesion on GUVs was dependent on the presence of the 26 flagellar filament and its motility. By testing different phospholipid head groups, the nature of 27 the fatty acid chains or the liposome curvature, we found that lipid packing is a key parameter 28 to enable bacterial adhesion. Using HT-29 cells grown in the presence of polyunsaturated fatty 29 acid (α-linolenic acid) or saturated fatty acid (palmitic acid), we found that α-linolenic acid 30 reduced adhesion of wild type EHEC but not of a non-flagellated mutant. Finally, our results 31 reveal that the presence of flagella is advantageous for the bacteria to bind to lipid rafts. We 32 speculate that polyunsaturated fatty acids prevent flagellar adhesion on membrane bilayers 33 Enterohemorrhagic Escherichia coli (EHEC) serotype O157:H7 are Shiga-toxin producing 19 strains characterized by peritrichous flagella and is responsible for major food-borne diseases 20 and for serious infections (1). When ingested, infection with EHEC is characterized by 21 symptoms ranging from hemorrhagic colitis to life-threatening complications (2). These 22 bacteria have the capacity to infect and to multiply in a wide variety of host species including 23 human, animals and even plants (3). The persistence in their hosts, including humans, occurs 24 through adhesion onto tissues (4). EHEC adhere to the intestinal mucosa in a manner termed 25 the attaching and effacing effect (5) but other mechanisms, involving flagella and pili, have 26 been described but not fully characterized (6). The pili are the most described adhesins present 27 at the bacterial surface (7,8). More recently, bacterial flagella have also been identified in 28 bacterial adhesion on different host tissues (9,10). The bacterial flagellum is a multiprotein 29 complex, best known as a filament responsible for bacterial movement toward preferred 30 environmental niches (11). The presence of flagella can be seen as a characteristic marker of 31 early-stage colonization. 32 The flagellum is mainly composed of a globular protein, the flagellin, which is organized in four 33 connected domains named D0, D1, D2 and D3. Flagellin peptides fold back on themselves 34 and the D0-D1 domains are interacting through a coiled-coil interface and hydrophobic 35 contacts, which are essential in the flagellin polymerization process. These N-and C-terminal 1 regions are well conserved across all bacterial flagellins. Conversely, the D2-D3 domains 2 generate antigenic diversity and are exposed on the filament exterior (12, 13). These 3 monomers form a helix made of 11 protofilaments of flagellin (14) with lengths up to 15-20 µm 4 and a diameter around 20 nm. 5 Recent evidence has suggested that flagella bind to the plasma membrane phospholipids 6 mainly through hydrophobic effects (15,16). However, little is known about the parameters 7 that govern the direct interaction between the lipid bilayers of plasma membranes and the 8 flagella of enteropathogenic bacteria. Previously, membrane rafts, which are membrane 9 domains enriched in cholesterol and sphingolipids (17, 18), have been documented as targets

20
Interactions of bacterial flagella with phosphatidylcholine vesicles. 21 To expand our knowledge about bacterial flagellum adhesion on plasma membranes (15,16), 22 we initiated this study by visualizing the interaction of bacteria with biomimetic GUVs 23 composed mainly of phosphatidylcholine (PC) from eggs (egg-PC) at room temperature 24 (23°C). This was performed by combining fluorescence and phase contrast microscopy. The 25 lipid bilayer was doped with 2% molar of a green-emitting fluorescent phospholipid, namely 26 (NBD-PE) ( Fig. 1A and 1B). The GUV average diameter was 6.27 µm with diameters ranging 28 from 2 to 41 µm (calculated from more than 450 GUVs) (Fig. S1). Such sizes are relevant with 29 that of EHEC host cells (5-30 µm in diameter). Following incubation with both non-flagellated 30 and flagellated EHEC, the former bacteria were not easily found around the GUVs (Fig. 1A) 31 whereas the latter were mainly observed around the GUVs (Fig. 1B).

11
The Brownian motion of GUVs and the motility of bacteria make the interaction between the 12 two species difficult to quantify by microscopy. Therefore, we used a quantitative binding assay 13 on GUVs as described previously (16) and illustrated in Fig. S2. 14 The lipid quantity immobilized on the GUVs' surface was dose-dependent (Fig. 1C). Lipid GUV 15 suspension containing from 0 to 10 µg of PC were deposited onto the gold surface and 16 subsequently incubated with EHEC. An increase of adherent bacteria was observed up to 5 17 µg. This plateau means that saturation of bacterial adhesion was achieved under the 18 experimental conditions. As a result, all the other experiments were performed with 5 µg of 19

GUVs. 20
Both the presence of the flagella and their motility was assessed to understand the role of 21 flagella in bacterial adhesion. The comparison between EHEC wild type (WT) and the flagella-22 free EHEC mutant (ΔfliC) confirmed that the lack of flagella is responsible for less adherent 23 bacteria (Fig. 1D). A non-motile but flagellated EHEC mutant (ΔmotA) was also studied. EHEC 1 ΔmotA was more adherent than ΔfliC (Fig. 1D), suggesting that the flagellar movement is 2 essential to bacterial adhesion. Collectively, these results confirmed the suitability of the GUV 3 adhesion assay to investigate the adhesive properties of EHEC flagella and reveal a key role 4 of bacterial flagella and active motility in the adhesion process on plasma membrane lipid 5 bilayers.

16
Significance levels are defined as follows: ***: p ≤ 0.001; NS: non-significant. The bar graphs represent 17 the mean of the reported data.

19
The role of vesicle curvature in flagellar adhesion. 20 To tackle the impact of curvature on flagellar adhesion, Large Unilamellar Vesicles (LUVs) 21 were produced with a diameter of 400 ± 30 nm. Surprisingly, EHEC were less adherent on 22 LUVs than on GUVs ( Fig. 2A), suggesting an impact of the size of the vesicles, thus membrane 23 curvature, on flagellar adhesion. To further investigate vesicle curvature-dependence, 24 generalized polarization (GP) measurements were achieved with Laurdan (6-dodecanoyl-2-1 dimethylaminonaphthalene) an amphiphilic fluorescent dye sensitive to local packing. Because 2 Laurdan GP sensed changes in the phospholipid order (23, 24), if the bacterial flagella can 3 penetrate the plasma membrane, a lower GP order will be observed (Fig. 2B). Vesicles with 4 different sizes were incubated with purified H7 flagella to investigate the impact of four different 5 bilayer curvatures. The GP measurement could not be performed on the whole bacteria 6 because the probe detects mainly the bacterial motility in the surrounding environment of the 7 vesicles (data not shown). As the initial packing of the lipid bilayers depends on the curvature, 8 the initial GP value of the vesicles with different diameters were normalized by subtracting the 9 GP value in the absence of flagella for each size of liposome (GP) (Fig. 2C). The presence 10 of H7 with vesicles allowed slight changes in the lipid order only for vesicle diameters greater 11 than 2 µm. This confirms the existence of a threshold diameter above which the interaction 12 becomes significant. 13 The role of phospholipid head groups in flagellar adhesion. 14 We next evaluated more thoroughly how different lipid species may affect EHEC flagellar 15 adhesion. Focus was first given to the influence of the lipid head groups as they could modulate 16 bilayer curvature and lipid packing (25). Since our first results were obtained with vesicles 17 composed of egg-PC, we characterized precisely the fatty acid profile and the lipid species of 18 the egg-PC that we used by gas chromatography-flame ionization detection (GC-FID) and 19 liquid chromatography-mass spectrometry (LC-MS) respectively. The main fatty acids were 20 C16:0 (palmitic acid), C18:0 (stearic acid), C18:1 (oleic acid) and C18:2 (linoleic acid). PC 21 34:1, constituted of palmitic acid and oleic acid, named 1-palmitoyl-2-22 oleoylphosphatidylcholine or POPC was the most abundant lipid (Fig. S3). 23 In the following experiments, membranes made of POPC alone were compared to membranes 24 made of POPC associated to other two phospholipids found in host plasma membrane with 25 the same fatty acid composition: 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine 26 (POPE) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol) (POPG). To prevent 27 from destabilization of the bilayer, POPG and POPE were mixed with POPC at 60% mol (26). 28 This allowed tackling the influence of the lipid head groups, which is known to impact on 29 various bilayer's properties including curvature or lipid packing (27). As POPC, POPE is 30 zwitterionic but has a negative curvature. Conversely, POPG carries a negative charge at 31 physiological pH and exhibits a zero spontaneous curvature as POPC (28).

20
A clear reduction of adhering bacteria was observed when POPC was mixed with POPE or 21 POPG (Fig. 3A). To unravel the role of membrane fluidity, membrane thicknesses and area 22 per lipids were calculated from MD simulations. The presence of POPE dramatically affected 23 these membrane properties by decreasing the area per lipid and increasing membrane 24 thickness. In other words, POPE increased ordering, as also seen by the order parameters of 25 lipids in the POPE/POPC binary mixture with respect to the pure POPC bilayer (Fig. 3B). 1 Conversely, the presence of POPG only slightly affected the area per lipid and increased 2 membrane thickness. To further investigate membrane ordering, steady-state fluorescence 3 anisotropy measurements were performed on GUVs (Fig. 3B) using two probes, 4 diphenylhexatriene (DPH) and Laurdan (28). Their partitioning allows monitoring lipid 5 dynamics in different membrane regions. The amphiphilic structure of Laurdan is localized at 6 the hydrophobic-hydrophilic interface region whereas DPH locates in the hydrophobic core of 7 the lipid bilayer (29). The three lipids, DOPE, DOPG and DOTAP, were mixed with POPC at 40:60 molar ratio. Less 25 bacteria were recovered on DOPC (~2.5 × 10 6 CFU/ml) alone than with POPC (~3.0 × 10 6 26 CFU/ml) (Fig. 4B). When mixing POPC with DOPC or DOPE, this number was significantly 27 lower than DOPC alone. Interestingly, we observed no significant differences neither with 28 DOPG nor with DOTAP mixed with POPC with respect to pure DOPC (Fig. 4B). The bacterial 29 adhesion has a low negative correlation except for the area per lipid parameter, which is 30 positive (Fig. 4C). An impact of the head group charge was observed on EHEC flagellum 31 adhesion. Zwitterionic phospholipids for the DO-bilayer series decreased bacterial adhesion 32 and conversely negative and positive head group had no impact. We thus reasoned that lipid 33 charge and membrane properties could differentially affect flagella adhesion on plasma 34 membrane lipids driven in part by the acyl chain content.

19
The role of phospholipid acyl chains in flagellar adhesion. 20 To investigate further the impact of phospholipid unsaturation, directly related to the fluidity, on 21 bacterial adhesion, the membrane parameters were assessed in the subsequent series: pure 22 POPC, DOPC 9-cis (DOPC cis), DOPC 9-trans (DOPC trans) and a poly-unsaturated lipid, 23 PC 18:3/18:3 (PC 18:3). All lipids were tested at room temperature as before (23 °C) because 24 as long as we work above their phase transition temperature (Tm), no changes are introduced 1 to the ordered gel phase. DOPC trans was also tested at 4 °C as it has a phase Tm of 12 °C. 2 Namely at 23 °C, all lipids were in fluid (disordered) phase (Fig. 5A), whereas at 4 °C, the 3 bilayer made of DOPC trans is ordered. At 23 °C, GUVs made of 100% DOPC exhibited less 4 bacteria adhesion regardless of whether the double bond was in trans or cis compared to the 5 POPC membrane (Fig. 5B). This is in line with their anisotropy under the same temperature  To provide further evidence about the impact of lipid composition with physiological 2 membranes, colon epithelial cells (HT-29) enriched in saturated (palmitic acid) or unsaturated 3 fatty acids (α-linolenic acid) were used. The lipid content of treated HT-29 cells was evaluated 4 by GC-FID (Fig. 6A), which confirmed enrichment in both palmitic and α-linolenic acids. A LC-5 MS analysis confirmed these results and showed an accumulation of palmitic acid and linolenic 6 acid when treated with the corresponding fatty acid (Fig. 6B). Palmitic acid was more 7 incorporated into intracytoplasmic lipid droplets through triglycerides (TAG) than  to POPC only, due to the gel phase state of the bilayer. The coexistence of lo and ld phases 20 improved adhesion of the WT. In the ld phase, the ternary lipid system exhibited no significant 21 differences with respect to POPC only. However, except for ld phase, non-flagellated bacteria 22 were clearly less adherent. The fact that no significant difference was observed for ld when 23 comparing the WT to the ΔfliC mutant, can be attributed to the presence of other adhesins 24 capable of binding to chol or PSM (Fig. 7B). Correlations between WT adhesion and the 25 parameters determined by MD and anisotropies, give the best result with the fluidity state with 26 a moderate negative correlation ( Fig. 7C and 7D). These results substantiate the previous 27 hypothesis concerning an ideal plasma membrane fluidity to increase bacterial adhesion on 28 lipid vesicles. However, with the ternary lipid system, the membrane is not homogeneous, 29 which provides different sizes of lipid rafts depending on the lipid proportion. lo does not have 30 rafts, lo/ld contains large rafts (>75-100nm) and ld forms small rafts (<20nm) (38).   adhere. This interaction was originally identified to occur with negatively charged phospholipids 3 by using immobilized lipids and purified flagella on a thin layer chromatography, which are not 4 physiologically organized (15). Likewise, S. enterica serorvar Typhimurium (Salmonella 5 Typhimurium) flagella were described to interact with pure cholesterol coated on surfaces but 6 not organized in the complex structure of a plasma membrane (39). However, recently, we 7 found that methylated flagella of Salmonella Typhimurium facilitate bacterial adhesion to PC 8 GUV and negatively affect adhesion on pure POPG GUV (16). 9 The present work illustrates both the importance of flagellar motility, and the size and resulting 10 curvature of the lipid vesicles, as important factors for optimal bacterial adhesion. Until now, 11 the fatty acid composition was largely ignored but this work reveals its key role. Fatty acid 12 saturation strongly impacts membrane thickness and area per lipids, both parameters thus 13 appear as important biomarkers of flagellar adhesion. In the presence of saturated fatty acid, 14 flagellar adhesion is optimal and modulated by the head group moiety of the membrane lipids. such as a human cell line (Fig. 6). These results reflect a clear role of lipid packing on flagellar 20

adhesion. 21
The computed bilayer parameters, obtained by MD simulations, such as membrane thickness 22 and area per molecule, have been related to the membrane ordering and fluidity (40). A more 23 fluid membrane exhibits less order and has higher value of area per lipid-molecule and lower 24 value of membrane thickness (when comparing lipid chains of same size). A global correlation 25 between these two structural parameters and the fluidity state of the GUV bilayer showed a 26 very high level of correlation with most of the bilayer composition used in this study. In turn, 27 these two parameters correlated with bacterial adhesion. Although not perfect because many 28 other parameters are at stake, these two parameters provide a very good trend about the 29 optimal conditions that increase, or not, bacterial adhesion. In other words, these two 30 parameters, which can be computed at a relatively low computational cost, are easily obtained 31 descriptors of the bacterial adhesion process. By adding a few other descriptors, we believe 32 that we could establish a quantitative structure activity relationship (QSAR) that could predict 33 flagellar adhesion with high robustness. 34 Ectothermic organisms incorporate fatty acids in phospholipids via a mechanism termed 1 "homeoviscous adaptation" to have constant viscosities at the temperature of cell growth (41). 2 It is only within homeoviscous adaptation limits that the diet of organisms can influence 3 membrane lipid profile. In endothermic animals, the membrane composition has been 4 thoroughly documented to be influenced by dietary fats in the erythrocyte plasma membrane 5 and later in the liver, the brain and other organs (42-45). However, the influence of dietary 6 intake on plasma membrane composition depends on the type of fatty acids ingested. Omega-7 3 and omega-6 fatty acids are not produced in mammals due to the lack of ad hoc desaturases, 8 therefore humans must consume such fatty acids (46). In contrast, palmitic acid can be 9 synthesized endogenously and its quantity is controlled under normal physiological conditions 10 in order to not affect the membrane properties (47). It is noteworthy that the incorporation of 11 diet fatty acids into membranes is influenced by the omega-3/omega-6 balance (48). To Recently, bacterial flagellar motility was described to help bacteria to reach preferred sites in 24 host plasma membrane containing sphingolipid-rich domains (56). Our findings show that not 25 only motility but also bacterial flagella per se, acting as an adhesin on lipids, help bacteria to 26 reach their host cell targets. As lipid rafts are the most documented plasma membrane lipids 27 involved in bacterial adhesion (19, 22), we investigated three different types of lipid rafts: lo, 28 lo+ld and ld, reflecting no, large and small size rafts respectively (38). No and small rafts did 29 not substantially facilitate flagellar adhesion whereas large size rafts promoted adhesion (Fig.  30   7). These results suggest that an optimal size of lipid rafts exists that allows flagellar adhesion, 31 which is consistent with previous works describing lipid rafts and bacterial adhesion (19, 22). properties associated to lipids in flagellated bacterial species will be needed. 8

9
Bacteria growth and preparation. EHEC O157:H7 TUV93-0 derived from strain EDL933 was 10 used (60).All The isogenic mutants ∆fliC and ∆motA were obtained as described (61) were spread manually onto a cover glass with a needle and dried on a heating plate at 50°C 21 for 30 min. 5 µl of lipid (Avanti Polar Lipids) solution in chloroform at 3 mg/ml were subsequently 22 deposited four times and spread until solvent evaporation. The residual solvent that could 23 remain on the lipid-coated cover glass is evaporated under vacuum at least 1 h. To form a well 24 on the cover glass slide, a ring was glued onto it and 500 µl of HEPES buffered saline solution 25 was added for hydration. After swelling of 1h at room temperature, the giant unilamellar 26 vesicles (GUVs) formed were either directly stored in a fridge or reengage in hydration of 27 another preparation to obtain a higher lipid concentration. Most of the GUVs size were 28 evaluated manually under microscope (Fig. S1).

29
Another method of formation can be applied in order to form vesicles with smaller diameter: 30 from Large Unilamellar Vesicles (LUV) to small GUVs. The lipid solution in chloroform was 31 dried under a nitrogen stream, and then under vacuum for 2 h to remove remaining solvent. 32 The film was hydrated to the desired lipid concentration in HEPES buffered saline solution. 33 After vortexing, the multilamellar vesicle suspension was extruded 25 times using a syringe 34 type extruder with polycarbonate filter having a pore size of 400 nm for 400-nm liposome, 2 35 µm for 1 µm-liposome and 5 µm for 2 µm-liposome (Liposofast, Avestin Inc). Prior extrusion 36 for 400-nm liposome, the solution was sonicated using a tip sonicator (Ultrasonic Processor 37 Vibra Cell, Sonic Materials). Liposome size was determined either manually with ImageJ by 38 epifluorescence microscopy or by dynamic light scattering (Zetasizer Nano ZS, Malvern 39 Instruments) for liposomes with a diameter smaller than 1 µm.

41
Labelling and epifluorescence microscopy. Liposomes were first immobilized on gold-42 coating glass. Prior to liposome preparation, DSPE-PEG-PDP (Avanti Polar Lipids) was added 43 to the lipid mixture in chloroform at 3%w and was as well doped with 2% mol NBD-PE for 44 liposome labelling (Avanti Polar Lipids). At the same time, a cover slip was coated with 1 nm 45 of chromium and 10 nm of gold by thermal evaporation (Evaporator Edwards model Auto 306).

46
The incubation between liposomes and bacteria was done in a separated container for 1 h and 47 transferred onto a microscope chamber. This latter was composed of the gold-coated glass 48 surface at the bottom, spaced from a common microscope slide with lateral spacers of molten 1 Parafilm. Observation was carried out on Leica DMI6000 B epifluorescence microscope. 2 Bacterial adhesion assays on liposomes. Bacterial adhesion assays were done in 6-well 3 plates (9.6 cm 2 per well). Gold-coated slides of 1.9x2.5 cm were placed onto a 3 mm high 4 pedestal in each well. 1 ml of 5 µg/ml liposome solution containing DSPE-PEG-PDP was 5 homogeneously deposited on gold-coated surface and incubated for 1 h. After adding 10 ml of 6 HEPES buffered saline solution, the pedestal was carefully removed with a bended pipette tip. 7 A volume of 8 ml was then removed to eliminate non-immobilized liposomes with a minimal 8 volume, to keep immerged the gold-coated surface. The immobilized liposomes were 9 incubated for 1 h with 5 ml of bacterial suspension at 10 8 UFC/ml in HEPES buffered saline.

10
Prior to discard all medium, the excess of bacteria was removed by taking away 5 ml and the 11 non-adherent bacteria were removed from the wells with 8 ml of fresh HEPES buffered saline. 12 With 1 ml of PBS, adherent bacteria were detached by pipetting vigorously several times 13 directly onto gold-coated surface. Samples were serially diluted and plated on LB agar for 14 viable bacterial counts. The bacterial count for 0 µg of lipid corresponds to the nonspecific 15 bacterial adhesion onto the gold surface without lipids. The non-specific adhesion of a well 16 alone (without glass and lipids) from the 6-well cell-culture plate use to perform the assay was 17 systematically subtracted. All adherence assays were performed at 23 ± 2 °C and at 4°C for 18 DOPC (Δ9-trans). 19 Bacterial adhesion assays on HT-29. The human colonic cell line HT-29 was obtained from 20 the American Tissue Culture Collection (ATCC). The cell line was maintained in modified 21 McCoy medium supplemented with 10% (v/v) heat-in-activated fetal calf serum, 2mM L-22 glutamine, 100 unit/ml penicillin and 100 unit/ml streptomycin at 37 °C in 5% CO2. Cultures 23 were used between passages 15 to 20. The cells were seeded in 24-well culture plates (1. where I0 is the fluorescence intensity measured with polarizer in parallel orientation (0º) and I90 7 the intensity in perpendicular orientation (excitation 0º and emission 90º). G is the correction 8 factor derived from the ratio of emission intensity at 0 and 90º with the excitation polarizer at 9 90º and is taking into account the different sensitivity of the detection system for vertically and 10 horizontally experiments were performed leading to a total acquisition time of 6 hours. After data 20 processing, phase and baseline correction, the area of the peaks of interest was determined 21 by integration and the molar ratio of lipids was calculated with relative integrations.  allowed to control the parameters of the machine, acquired and processed the data. The mass 5 spectra were acquired in positive and negative-ion mode. 6 7 Data processing and annotation 8 Agilent generated files (*.d) were converted to *.mzXML format using MSConvert (66). Files 9 *.mzXML datasets were processed using MZmine 2 v2.37 (67). The noise level was 2.0E3 for 10 MS1 and 0E00 for MS2 in centroid. The chromatogram builder was used using a minimum 11 time span of 0.10 (min), a minimum of height of 1.0E3 and m/z tolerance of 5ppm. The 12 chromatogram deconvolution was done with the local minimum search algorithm. The 13 chromatographic threshold was 30.0%; the search minimum in RT range was 0.05 min with a 14 minimum relative height of 5% and a minimum ratio of peak top/edge of 2. Peak duration range 15 0.05 -3 min. MS2 scans were paired using a m/z tolerance range of 0.05 Da and RT tolerance 16 range of 0.1 min. Isotopologues were grouped using the isotopic peaks grouper algorithm with 17 a m/z tolerance of 0.008 and a RT tolerance of 0.3 min. A peak alignment step was performed 18 using the join aligner module (m/z tolerance = 0.008, weight for m/z = 50, weight for RT = 50, 19 absolute RT tolerance 2 min). Peak finder module was used with intensity tolerance of 10%, 20 m/z tolerance of 0.008 and retention time tolerance of 1.0 min. The resulting peak list was then 21 filtered using the peak list row filter module with a minimum peak in a row of 2, a minimum 22 peak in an isotope pattern of 2 and by keeping only peaks with MS2 scan (GNPS). The peak 23 list was then exported to *.csv using the module "Export to CSV file". Moreover, a *mgf was 24 exported using the module: "export for/submit to GNPS". The peak list was annotated using a 25 combination of four databases, GNPS (68) lipid blast (69), lipid match (70)  were considered as significant for a p value≤0.05. Secondary data analysis and correlation 5 matrices were calculated using the open-source software GNU-R. Bacterial growth counts 6 across all the tested scenarios as well as metrics related to each lipid composition (viz. 7 Membrane thickness, Area per molecule, DPH and Laudan anisotropy) were loaded as csv 8 files into the R environment. Correlations were computed using the "corr" function of the R 9 library "Hmisc", which calculates the significance level (p-value) using Pearson correlation 10 coefficient on raw values. The correlation matrix plots were generated using the R packages 11 "corrplot" and "PerformanceAnalytics". To interpret the size of the correlation 12 coefficient it was as follow: 1.00-0.9, very high; 0.9-0.7, high; 0.7-0.5, moderate; 0.5-0.3, low; 13 0.3-0, negligible correlation as described (77). 14 15 Acknowledgement. 16 We are grateful to David Gally and Eliza Wolfson for Stx-negative derivative strain of TUV 93-17 0 and its isogenic mutants (ΔfliC and ΔmotA). We thank the European Regional Development