All chemicals (solvents, lipids, media, proteins and so on) used in this article were purchased from Sigma-Aldrich without further purification, except where stated otherwise.
Lipid layer preparation
SCLs were prepared on silicon wafers (10 × 15 mm2). The substrates were cleaned by immersion in a solution of deionized water, ammonia and hydrogen peroxide (volume fraction 5/1/1) for 15 min at 70 °C, rinsed repeatedly in Milli-Q water and then dried in a nitrogen stream. The cleaned substrates were immediately used for the preparation of SCLs by spin-coating. Compounds were dissolved in chloroform at a concentration of 2 wt% (unless stated otherwise), and spin-coating (LabSpin6, SÜSS MicroTec) was performed at 3,000 rpm s–1 for 30 s.
Silicon wafers were cleaned as described for SCL preparation. Clean substrates were first coated with a 3 nm chromium adhesion layer and then a 50 nm gold layer by chemical vapour deposition performed at 5 × 10–5 mbar (Univex 300, Leybold). Gold substrates were cleaned by immersion in a solution of deionized water, ammonia and hydrogen peroxide (volume fraction 5/1/1) for 15 min at 70 °C, rinsed repeatedly in Milli-Q water and then dried in a nitrogen stream. To minimize defects, the gold substrates were used for SAM formation immediately after cleaning. Thiol compounds were dissolved in ethanol (analytically pure, p.a.) to a concentration of 1 mM. Before use, thiol solutions were sonicated for 5–10 min. Gold-coated substrates were additionally cleaned in ethanol (p.a.) for 30 min using an ultrasonic bath. Subsequently, samples were immersed in thiol solutions and incubated for 24 h to ensure complete assembly. Containers housing the solution and samples were filled with dry nitrogen and sealed to minimize oxygen exposure. After incubation, sample surfaces were rinsed for 15–20 s with ethanol (p.a.), dried under nitrogen and used directly for further experiments.
Preparation of thiocholenic acid
For the reaction scheme see Supplementary Fig. 1. One equivalent of 3-acetoxy-5-cholenic acid (A) was dissolved in anhydrous dimethyl sulfoxide (DMSO) under an argon atmosphere. The solution was stirred on ice, and six equivalents of carbonyldiimidazole (B), dissolved in anhydrous DMSO, were added slowly to the solution. The reaction mixture was stored at 4 °C overnight for activation. A solution of 2.5 eq. cystamine (D) was added dropwise to the stirred, ice-cooled imidazole intermediate (C) and left to react overnight at room temperature. The product (E) was purified, freeze-dried, dissolved in ethanol and diluted in water before the addition of 1 M NaOH to obtain ester cleavage. After purification and lyophilization, the product (F) was again dissolved in ethanol and diluted in water. The disulfide bond was cleaved by the addition of a sixfold excess of tris(2-carboxyethyl)phosphine hydrochloride to cholenic acid in aqueous solution at neutral pH. The cleavage product (G) was purified and lyophilized (purity around 80%). For reaction/purity control and purification, reverse-phase high-performance liquid chromatography (RP–HPLC) with a linear gradient of acetonitrile in water with additive formic or trifluoroacetic acid was used. The analytical HPLC instrument (1260 Infinity II, Agilent) was equipped with a diode array detector (210 and 278 nm) and an electron spray ionization–time-of-flight (ESI–TOF) detector. The preparative HPLC instrument (1200 series, Agilent) used a diode array detector and a fraction collector in manual collection mode.
The total organic carbon (TOC) content of solutions was analysed using a Sievers 5310C laboratory TOC analyser (GE Analytical Instruments) in accordance with the manufacturer’s specifications. Information on the preparation of samples is provided in Supplementary Fig. 4.
Bacterial adhesion assays
S. epidermidis (strain PCI 1200, ATCC) and E. coli (strain W3110) were grown overnight from single colonies in lysogeny broth (LB) at 37 °C and 200 rpm. Overnight cultures were centrifuged at 4,000g for 5 min. The supernatant was removed and the remaining pellet resuspended in LB. This washing step was repeated three times. Cell densities were adjusted to an optical density (OD600) of 0.2 in fresh LB, and sample substrates were incubated in the bacterial solution for 1 h at 37 °C (no shaking). After incubation, adherent bacteria were fixed with 4% paraformaldehyde in PBS for 10 min, washed in fresh PBS and Milli-Q water and dried under nitrogen. Samples were sputter-coated with a 15 nm gold layer (SCD 050, Balzers) and imaged by scanning electron microscopy (SEM; XL30 ESEM-FEG, Philips/FEI) in high-vacuum mode at an acceleration voltage of 5 kV. For each sample, at least six images were acquired at random positions and cells were counted using the counting tool in Fiji27. The scale bars of SEM images were used to calibrate pixel width. The Fiji ‘cell counter’ plugin was used to count the number of cells in the calculated area (approximately 103 µm2). The numbers counted were either normalized against the median value of the silicon reference for relative comparisons or scaled up to cells per square millimetre for absolute values.
Quartz crystal microbalance (QCM) measurements were performed using a QCM-D model E4 (Biolin Scientific) equipped with a peristaltic pump system (IPC, Ismatec). Gold-coated quartz crystals (QSX301, Quantum Design) with a resonance frequency of 5 MHz were used for QCM measurements. SCLs and SAMs were prepared on QCM crystals as described above. All measurements were performed at a flow rate of 100 µl min−1. Protein solutions (lysozyme, bovine serum albumin and fibrinogen (100 µg protein ml–1 PBS)) or 10 vol% fetal bovine serum (Merck) in PBS were adsorbed on the layered samples for 1 h and subsequently subjected to a desorption regime for 30 min with PBS. Frequency and dissipation shifts induced by the adsorbed proteins were recorded in real time at the third, fifth, seventh, ninth, 11th and 13th overtones (15, 25, 35, 45, 55 and 65 MHz, respectively). The mass of adsorbed protein was calculated using the Sauerbrey equation28 with Q-Sense DFind software (Biolin Scientific).
Dynamic contact angle measurements
Contact angle measurements were performed using an OCA 30 optical contact angle measuring and contour analysis system equipped with a TPC 160 temperature-controlled chamber (DataPhysics Instruments). Droplets of degassed deionized water were dispensed and redispensed at varying velocities of 0.3–2.0 µl s–1 to monitor advancing and receding contact angles and their time-dependent behaviour.
Ellipsometry measurements were performed with an M-2000 ellipsometer (J. A. Woollam) equipped with a 50 W QTH lamp operating at wavelengths from 371 to 1,000 nm and an angle of incidence of 75°. The oxide layer thickness of the silicon wafer was determined by ellipsometry before layer assembly. The thickness of the assembled layers was calculated by an optical model that included three layers: Si, SiO2 and a Cauchy layer.
In situ ATR–FTIR
For in situ ATR–FTIR, 500 µl of a 2% cholesterol-chloroform solution was spin-coated on a germanium ATR crystal. Characterization of the deposited cholesterol SCLs by in situ ATR–FTIR was performed as described previously29. In situ ATR–FTIR spectroscopy was conducted using the single-beam sample reference technique to obtain compensated ATR–FTIR spectra in dry and aqueous environments30,31. Dichroic measurements were performed according to a previously described method32. Infrared light was polarized by a wire grid polarizer (SPECAC). The ATR–FTIR attachment was operated on an IFS 55 Equinox spectrometer (BRUKER Optics) equipped with a Globar source and mercury-cadmium-telluride detector. P- and s-polarized spectra were recorded from dry cholesterol SCLs. The high dichroic ratios of the n(C-O) band of the C-O-H headgroup of cholesterol can be verified in the line of the ATR–FTIR dichroism measurements of lipid bilayers. The dichroic ratio R = AP/AS, with absorbance Ap measured in p-polarization and AS measured in s-polarization, of infrared bands with a transition dipole moment (M) located perpendicular to the surface plane also showed high values, with R > 4. Briefly, under conditions of ATR at the interface of the dense medium (Si) and rare medium (air), an evanescent wave was established with an electrical field split into the three electrical field components, Ex, Ey and Ez, which interact with, for example, adjacent organic layers. Parallel polarized infrared light (EP) forms Ex and Ez whereas vertically polarized light (ES) forms Ey. High values of either R or Ap are obtained when the M of a functional group within the organic layer lies parallel to Ez (out of plane), whereas low R or high AS values are obtained when M lies parallel to Ey (in plane), which is due to the scalar product \(A=E\times M=E\times M\times \cos (E,M)\) of the vectors E and M.
Time-of-flight secondary ion mass spectrometry
Time-of-flight secondary ion mass spectrometry (ToF–SIMS) was conducted with a ToF–SIMS 5-100 instrument equipped with a 30 kV Bi liquid metal ion gun (IONTOF). Data were acquired in Bi3++ mode and calibrated against a list of reference peaks (SurfaceLab7, IONTOF). The area of analysis was 300 × 300 µm2, which was scanned over 128 × 128 pixels. The sampling depth of this technique is as low as a few nanometres—that is, only the uppermost molecular layers of the sample contribute to the analysis. Characteristic signals of cholesterol (mass to charge ratio, m/z = 369.3) and stearyl palmitate (m/z = 257.2) were selected for semiquantitative characterization of SCLs.
We applied a previously described FRAP protocol to analyse diffusion coefficients and mobile fractions with a fluorescence confocal laser scanning microscope using the FRAP tool (SP5, Leica)33,34. For FRAP measurements, fluorescent cholesterol SCLs were prepared by the addition of 1/100 or 1/20 NBD cholesterol (ThermoFisher) to pure cholesterol solutions (2 wt%) and spin-coating to clean no. 1.5 glass coverslips (Corning). Cholesterol SCLs were subsequently submerged in deionized water or PBS, and FRAP was performed by photobleaching a defined spot with a diameter of 10 ± 1 µm using the following protocol: ten images before bleaching followed by a high-power laser beam with subsequent bleaching for 4 s to achieve a completely bleached area. Recovery was recorded with a 40×/ 1.4 numerical aperture oil immersion objective at an image acquisition speed of 1 s per frame at 256 × 256 pixels, for a total of 300 s (SP5, Leica). The resulting time-lapse was analysed using the MATLAB (MathWorks) programme frap_analysis35.
All AFM measurements were performed with a NanoWizard IV AFM (JPK Instruments). The cantilevers used were calibrated before measurements. Measurements were conducted in PBS at room temperature (25 °C), except where stated otherwise.
Surface topography of the layered surfaces was recorded using the Quantitative Imaging mode of the AFM instrument using qp-BioAC cantilevers (Nanosensors). The acquisition parameters employed were as follows: 300 nm ramp, 10 ms pixel time and force trigger of 100 pN. Images of 30 × 30 µm2 with a resolution of 256 × 256 pixels were recorded. The data-processing software provided by the AFM manufacturer (JPK Instruments) was used to extract the surface roughness (Ra) from topography images.
Colloidal probe force spectroscopy
For colloidal probe measurements, individual silica beads (Kisker Biotech, Ø10 µm) were attached to a tipless cantilever (PNP-TR-TL-Au, Nanoworld, nominal force constant 0.08 N m–1) as described previously36. Colloidal probe-modified AFM cantilevers were cleaned in isopropanol, and adsorbed water was removed by heating at 120 °C for 10 min. The colloidal probe was hydrophobized by incubation in hexamethyldisilazane vapour for 12 h and subsequent heating at 120 °C for 1 h. The force spectroscopy parameters employed were as follows: 3 nN setpoint force, 5 µm s–1 approach/retract velocity and 5 µm pulling distance. Interaction forces were extracted from the retraction force–distance curves using data from the processing software provided by the AFM manufacturer.
Single-cell force spectroscopy
The colloidal probe-modified AFM cantilever (see section ‘Colloidal probe force spectroscopy’) was made cell adhesive by application of a polydopamine coating, and individual E. coli cells with cytoplasmic green fluorescent protein (strain MG1655 eGFP) were attached as described previously37. Measurements were conducted at 37 °C using a PetriDishHeater (JPK Instruments). The same force spectroscopy parameters and data-processing routines were used for colloidal probe spectroscopy measurements. Only datasets in which the position and orientation of the bacterial cell on the AFM cantilever was unchanged before and after measurements (that is, the contact conditions/geometry were constant during the measurement) were analysed.
Force spectroscopy measurements to quantify the spatial heterogeneity of cholesterol SCLs
Measurements were performed using the Quantitative Imaging mode of the AFM instrument using qp-BioAC cantilevers (CB-2, Nanosensors). The cantilevers were either hydrophobized by incubation in hexamethyldisilazane vapour as described above or hydrophilized by plasma cleaning (Harrick Plasma) for 10 min. The acquisition parameters used were as follows: 100 nm ramp, 20 ms pixel time and force trigger of 500 pN. Images of 5 × 5 µm2 with a resolution of 50 × 50 pixels were recorded at different locations on the sample surface. Interaction forces were extracted from retraction force–distance curves using the data-processing software provided by the AFM manufacturer.
Force spectroscopy measurements to quantify the temporal heterogeneity of cholesterol SCLs
Time-dependent force spectroscopy measurements were performed with hydrophobized (see previous section) qp-BioAC cantilevers (CB-2, Nanosensors). The force spectroscopy parameters employed were as follows: 1 nN setpoint force, 1 µm s–1 approach/retract velocity and 200 nm pulling distance. A waiting period of 20 s was maintained between the 16 consecutive measurements made at a single location on the sample surface, to minimize the influence of measurements on the dynamics of cholesterol molecules. Measurements were repeated at different locations of the sample. Interaction forces were extracted from retraction force–distance curves using the data-processing software provided by the AFM manufacturer.
Cholesterol or stigmasterol SCLs with four molecular layers—that is, two double layers (Extended Data Fig. 6a)—were modelled. At the interface of two double layers the hydrophilic hydroxyl groups of cholesterol or stigmasterol face each other; at the interface, hydroxyl groups face the water layer. The double layer–water interface normal was along the z axis in all simulations. To model a solid substrate at the bottom of the SCL, the motion of molecules was restrained in the xy plane for the lowermost lipid molecules. The simulation box dimensions were 7.1554 × 7.1554 × 50 nm3. Hydrophobic walls, modelled as direct 12-6 LJ potential at z = 0 and z = 50 nm, were included at the bottom and top of the simulation box along the z axis. Vacuum layers above and below the multilayer prevent water–wall and lipid–wall short-range non-bonded interactions.
Cholesterol and stigmasterol molecules were modelled with the CHARMM36 force field38,39. The TIP3P water model in CHARM40,41,42 was used and included KCl salt in the simulations. First, double layers of cholesterol or stigmasterol were constructed on CHARMM-GUI, where the hydrophobic tails face each other43,44. The CHARMM36 force field parameters for cholesterol and stigmasterol, TIP3P water parameters and parameters for ions were obtained from CHARMM-GUI. Each layer in the double layer of cholesterol or stigmasterol contained 128 molecules. The double layer obtained from CHARMM-GUI was translated along the z axis by 4 nm using the Visual Molecular Dynamics visualization programme45, constructing multilayers of four layers containing 512 lipid molecules in total (Extended Data Fig. 6a). Water and ions were added on top of the multilayers. The cholesterol multilayer system contained 12,284 water molecules with 72 K+ and Cl− ions and the stigmasterol multilayer system contained 12,363 water molecules with 74 K+ and Cl− ions. The system was energy minimized and equilibration simulations for both cholesterol- and stigmasterol-containing systems were conducted using Gromacs 2019.4 (for details see Supplementary Note 4)46,47.
To construct the systems with a reversed molecular orientation at the interface (top layer), 10% (13 molecules), 30% (39 molecules) and 50% (64 molecules) of the molecules were reversed in orientation compared with the equilibrium system. The Alchembed tool48 was used to remove any overlaps between coordinates that might have appeared on reverting the orientation of molecules. Minimization and short equilibration runs were then conducted as described above (for details see Supplementary Note 4).