Life has evolved a plethora of powerful principles to control adhesion, some of which have been recapitulated in engineered materials. Prominent examples include the superhydrophobic leaves of the sacred lotus1 and the omniphobic surfaces of Nepenthes pitcher plants2. Whereas interfacial phenomena in nature are widely studied, the physical mechanisms underlying the control of bioadhesion—the interfacial accumulation of biopolymers and cells (including bacteria)—are not yet thoroughly understood. We previously explored the omniphobic, antibioadhesive cuticula of Collembola (Fig. 1a) and found that it consists of nanoscopic structures with overhanging cross-sectional profiles (Fig. 1b,c) preventing wetting and bacterial colonization3,4,5,6. Later the lipid-rich envelope of the Collembola cuticula (Fig. 1c)—considered to serve as another ‘line of defence’ against bioadhesion—was shown to contain aliphatic hydrocarbons, in particular steroids, fatty acids and wax esters (Fig. 1d and Extended Data Fig. 1)7. Whereas wax esters can be reasonably assumed to support the non-wetting properties of the cuticula8, the role of steroids and fatty acids remains elusive. Free fatty acids have been reported to kill or inhibit the growth of bacteria and fungi9,10, and steroids have been found to reduce bioadhesion on sponges and sea stars11; however, no mechanistic explanation for this effect of steroids is available. Amphiphilic lipid components of the Collembola cuticula are also contained in the membranes of animal and bacterial cells that play key roles in compartmentalization and the functional alignment of molecular machineries12. Cholesterol, in particular, has been comprehensively studied and its presence is considered crucial for the regulation of functional lipid domains and the interaction between proteins and lipids13. However, the functional relevance of cholesterol at interfaces of living structures other than cellular membranes is underexplored.

Fig. 1: Layers of Collembola cuticular lipids and their bioadhesion properties.
figure 1

a, Image of Tetrodontophora bielanensis, an exemplary Collembola sp. Scale bar, 1 mm. b, Scanning electron microscopy image of a T. bielanensis cuticula. Scale bar, 500 nm. c, Cross-sectional schematic of the cuticula, showing a layered structure consisting of a chitin-rich inner skeleton covered by a protein-rich layer. A thin, lipid-rich envelope covers the protein-rich layer. Scale bar, 200 nm. d, Summary of lipids detected in the outer cuticula layer of T. bielanensis7. e, Layers of Collembola cuticular lipids; SCLs containing cholesterol facilitate orientational adaptation of the topmost lipids to the polarity of the environment. ATR–FTIR (Supplementary Fig. 2) and dynamic contact angle measurements (Fig. 2c and Extended Data Fig. 4a) indicate highly ordered cholesterol molecules, with the hydrocarbon tail of the outer cholesterol layer initially oriented towards the interface and the hydroxyl groups oriented inward. SAMs chemisorbed to gold via thiol groups, with either the polar or non-polar side of cholesterol oriented to the interface, served as references in selected experiments. fi, Adsorbed amount of protein (f,i) and normalized adherent cells (g,h). Adsorbed amount of protein (lysozyme or fetal bovine serum) on monocomponent layers of Collembola cuticular lipids (f) and multicomponent SCLs of stearyl palmitate and cholesterol (i), as determined by quartz crystal microbalance measurements. Normalized adherent cells of S. epidermidis on monocomponent layers of Collembola cuticular lipids (g) and multicomponent SCLs of stearyl palmitate and cholesterol (h). Data normalized to average adherent cell density on a silica (SiO2) substrate. h,i, Pure stearyl palmitate SCLs (100/0) and pure cholesterol SCLs (0/100) served as negative and positive controls, respectively. fi, Mean + s.d. The number of observations (n) is indicated. P values (comparison with cholesterol SCL condition in f,g and with the 0/100 condition in h,i) were determined using one-way analysis of variance. AU, arbitrary units.

Source data

Using lipid layers without the cuticula’s morphological features, we dissect the role of molecular assemblies in the control of bioadhesion from the antiadhesive topography effects of the Collembola cuticula5,14. Cholesterol-containing lipid layers were found to counteract bioadhesion by a previously unknown entropic repulsion mechanism that is—unlike earlier reported interfacial effects15,16,17—caused by orientational fluctuations.

Cholesterol layers exibit low bioadhesion

Lipids of the Collembola cuticula (Fig. 1d) were physiosorbed in multilayers on solid supports by spin-coating (spin-coated lipid multilayers, SCLs; Fig. 1e). For cholesterol SCLs, attenuated total reflection Fourier-transform infrared spectroscopy (ATR–FTIR) showed a multilayer structure with the majority of molecules oriented perpendicular to the interface (Supplementary Fig. 2)18,19. Self-assembled monolayers (SAMs; Fig. 1e) chemisorbed on gold served as a comparative analysis of the investigated lipidic cuticle components in geometrically constrained molecular orientations (Extended Data Figs. 1 and 2a–c).

A hallmark of bioadhesion is the formation of a molecular conditioning layer of proteins, nucleic acids, lipids and other biomolecules20. This process was mimicked by adsorption experiments with lysozyme and albumin, the results showing significantly lower adsorbed amounts of proteins on cholesterol SCLs than on all other studied SCLs and SAMs (Fig. 1f and Extended Data Fig. 2f,g). Likewise, adhesion of Gram-positive (Staphylococcus epidermidis, strain PCI 1200) and Gram-negative (Escherichia coli, strain W3110) bacteria was found to be very low on cholesterol SCLs (Fig. 1g and Extended Data Fig. 2d,e). By analysis of cholesterol SCLs of varying layer thickness (Supplementary Fig. 3a), any influence of multilayer thickness on antiadhesive properties could be excluded. Notably, SAMs of thiocholesterol and N-(2-mercaptoethyl)-3-hydroxy-5-cholenic acid amide (thiocholenic acid)—that is, surrogates of chemically fixed cholesterol monolayers of opposing orientational molecular alignment; Fig.1e)—accumulated higher amounts of adsorbed protein and adherent bacteria than cholesterol SCLs (Fig. 1f,g and Extended Data Fig. 2d–g). This observation suggests that restriction of molecular mobility in layered cholesterol impedes antiadhesive properties. Partial dissolution of cholesterol SCLs was excluded as a cause of low bioadhesion by total organic carbon content analysis (Supplementary Fig. 4).

Because cholesterol occurs in the Collembola cuticula together with several other lipid components (Fig. 1d), we further investigated its relevance regarding the adhesive characteristics of multicomponent SCLs. Multicomponent SCLs containing stearyl palmitate—a lipid that causes strong bioadhesion to its single-component SCLs (Fig. 1f,g and Extended Data Fig. 2d,g)—and cholesterol, with cholesterol content ranging from 0 to 100 wt%, were investigated using similar bioadhesion assays (Extended Data Fig. 3a–c). Multicomponent SCLs containing at least 10 wt% cholesterol showed low adherent cell densities similar to pure cholesterol SCLs (Fig. 1h and Extended Data Fig. 3d). Likewise, protein adsorption on multicomponent SCLs was significantly reduced when a minimum of only 1 wt% cholesterol was present (Extended Data Fig. 3e,f). The results of single-protein adsorption experiments were confirmed when applying physiologically relevant protein mixtures (that is, 10 vol% fetal bovine serum; Fig. 1i).

Cholesterol layers adapt to ambient polarity

The results of the bioadhesion assays suggest that mobility of interfacial cholesterol is key to the antiadhesive characteristics of single and multicomponent SCLs. Atomic force microscopy (AFM)-based colloid probe force spectroscopy, AFM-based single-cell force spectroscopy and dynamic contact angle measurements provided further insight into the underlying interfacial dynamics.

AFM-based force spectroscopy (Supplementary Fig. 5a,b and Methods) showed the dynamic interfacial adaptation of immersed cholesterol SCLs. Interaction forces between hydrophobic colloidal probes (Supplementary Fig. 5c) or single E. coli cells and cholesterol SCLs were quantified (Supplementary Fig. 5d) and were found to increase continuously with contact time (Fig. 2a,b). Further analysis showed that third-order kinetics controlled the underlying polarity adaptation process of the SCL interface to maximize hydrophobic interactions (Fig. 2a,b and Supplementary Note 1). Control experiments with thiocholesterol and thiocholenic acid SAMs confirmed that cholesterol mobility within SCLs is key to the observed adaptation (Fig. 2a,b).

Fig. 2: Cholesterol SCLs dynamically adapt to the polarity of their environment.
figure 2

a,b, Contact time-dependent interaction forces (determined by AFM-based force spectroscopy) between a hydrophobic colloidal probe (a) (10 µm silica bead modified with a hydrophobic silane) or individual E. coli cells (b) (attached to a 10 µm silica bead) and cholesterol SCLs or thiocholesterol and thiocholenic acid SAMs. Experiments were performed at 37 °C. a,b, Mean ± s.e.m. Data were obtained from at least three independent experiments. Regressions are the best-fit solutions for a third-order process indicating formation of bonds at the probe cell–SCL interface following \(\frac{{\rm{d}}\sigma }{{\rm{d}}t}=-\,k{\sigma }^{3}\) (for details see Supplementary Note 1). The interaction force was assumed to be directly correlated to the surface concentration of formed bonds, σ. Coefficients of determination (R2) for fits are shown. c, Representative images showing the dynamic contact angles of water on the surface of a cholesterol SCL. The high advancing contact angle (θadv, blue box) reflects a hydrophobic surface. Following immediate withdrawal of the droplet (no resting period, green box), a slightly reduced but still high receding contact angle (θrec) was observed. No receding of the three-phase contact line was observed when the water droplet was withdrawn after a 20 s resting period (red box). d, During the 20 s resting period, oscillations in the three-phase contact line between the initial advancing angle (93°) and an approximately 10° lower angle were observed.

Source data

Dynamic wetting experiments on cholesterol SCLs in air showed high advancing and receding water contact angles when the droplet was immediately withdrawn after deposition (Fig. 2c, Extended Data Fig. 4a and Supplementary Video 1), indicating that the hydrocarbon chains of the outer cholesterol layers conditioned in air are preferentially oriented towards the interface. If the water droplet was kept on the surface for 20 s before receding, oscillations of the three-phase contact line occurred for 15–20 s and subsequently the static contact angle was lower by approximately 10° (Fig. 2d and Supplementary Video 2). Pinning of the three-phase contact line was observed when the droplet was withdrawn (Fig. 2c, Extended Data Fig. 4a and Supplementary Video 2). The contact angle drop was found to be fully reversible following drying in inert gas (Extended Data Fig. 4d). The resting-period-dependent decrease in the contact angle, the oscillating three-phase contact line during the resting period and the increased contact angle hysteresis confirm a dynamic polarity adaptation of cholesterol at the interface of the SCLs in response to changes in the polarity of the environment.

The time dependence of the increase in interaction force between cholesterol SCLs and AFM probes (Fig. 2a) and single bacterial cells (Fig. 2b), as well as the decrease in receding water contact angle on cholesterol SCLs (Fig. 2c), indicate that the interfacial polarity adaptation process of cholesterol SCLs is relatively slow (that is, occurring over a period of a few seconds).

Also underpinning the results of the bioadhesion assays (Fig. 1h,i and Extended Data Fig. 3d–f), low quantities of cholesterol in multicomponent SCLs of cholesterol and stearyl palmitate sufficed to decrease the receding water contact angle (Extended Data Fig. 4c) and to increase the interaction force with hydrophobic AFM probes (Extended Data Fig. 3h).

Entropic repulsion by surface fluctuations

The antibioadhesive properties of cholesterol-containing SCLs were shown above to correlate with a dynamic adaptation of the interfacial orientation of cholesterol in response to changes in environmental polarity. We hypothesized that entropically driven orientational fluctuations of interfacial cholesterol molecules mechanistically connect these features: any adsorption of biomolecules or attachment of (bacterial) cells requires an orientational (polarity) adaptation of the SCL interface that constrains the orientational states of interfacial cholesterol and thereby reduces the entropy of the system.

To validate this hypothesis we first examined the system for the characteristic temperature dependence of entropic effects on protein adsorption (according to the definition of the Gibbs free energy of adsorption, \(\Delta G=\Delta H\,\mbox{--}\,T\Delta S\)) using three selected proteins of increasing size and complexity: lysozyme, albumin and fibrinogen. Indeed, protein adsorption to cholesterol SCLs was observed to decrease when the temperature was increased from 15 to 40 °C (that is, in a temperature range in which significant temperature-induced changes in protein conformation can be excluded21,22,23), in contrast to minor changes in similar experiments with thiocholesterol and thiocholenic acid SAMs (orientationally fixed controls of cholesterol SCLs underneath the adsorbed protein) (Fig. 3a and Extended Data Fig. 5a,b). At thermal adsorption equilibrium, the temperature dependence of the Gibbs free energy of adsorption can be used to quantitatively estimate the repulsive entropic barrier (ΔSchol) to protein adsorption on cholesterol SCLs. A quantitative analysis of adsorption (based on the slopes of ΔG(T) on cholesterol SCLs and the SAM controls) was performed for lysozyme, because adsorption-induced structural changes are negligible for the small, compact protein and provided \(\Delta {S}_{{\rm{chol}}}=-\,200\,\pm \,60\,{\rm{J}}{{\rm{mol}}}^{-1}{{\rm{K}}}^{-1}\) (Extended Data Fig. 5c and Supplementary Note 2)14,24. This value of ΔSchol shows a significant entropic repulsion counteracting bioadhesion to cholesterol SCLs. The plot of ΔG(T) for cholesterol SCLs exhibits a positive slope (that is, a negative ΔS of adsorption), indicating that entropic repulsion overcompensates any entropic gain of protein adsorption for that system. Whereas desorption-determined processes might show a similar temperature dependence, protein adsorption is well known to be strongly determined by adsorption24.

Fig. 3: Quantitative evidence of entropic repulsion by cholesterol SCLs and verification of the molecular mechanism.
figure 3

a, Temperature-dependent adsorbed amounts of lysozyme as determined by quartz crystal microbalance measurements. For cholesterol SCLs, a pronounced temperature dependence on lysozyme adsorption was observed. Data were obtained from at least three independent experiments. Mean ± s.d. b, Schematic illustration of the transition of interfacial cholesterol from the orientationally fluctuating (unbound) to orientationally constrained (protein-bound) state. Three aligned cholesterol molecules are shown to highlight the cooperative reorientation of neighbouring molecules within SCLs derived from geometric considerations and evidenced by the experimentally observed third-order kinetics of interfacial adaptation and orientational correlation length in MD simulations (Supplementary Notes 1 and 4). The entropic penalty of bioadhesion can be estimated from the ratio of constraint and fluctuating interfacial areas, ΔS = R × ln(Aconstr/Afluct), of cholesterol molecules, with Afluct roughly equal to the molecular dimension of cholesterol due to the cooperative orientational fluctuation and Aconstr related to the cross-sectional area of the cholesterol facing the interface in the protein-bound state (Supplementary Note 3). c, MD simulations of cholesterol multilayers in contact with water, with equilibrium state (left) and intentionally reverted molecular orientation of 10% of molecules in interface layer (right) (for details see Supplementary Note 4). Oxygen atoms of hydroxyl groups are shown as orange spheres. d, Angles between the vector connecting atoms C3 and C17 on cholesterol and stigmasterol (Extended Data Fig. 6c and Supplementary Note 4) and the z axis for interfacial molecules as a function of simulation time for one representative simulation trajectory. In cholesterol multilayer systems with 10% of reverted molecules, spontaneous back-reorientation within 1 µs of simulation time was observed. No back-orientation was observed for stigmasterol controls. e, Correlation function (Cr) as a function of the distance to a reference molecule, plotted for three simulation trajectories of cholesterol multilayer systems (with no changed orientation of molecules at the interface). Error bars represent s.d. at that point over time intervals. f, Root mean squared deviation (RMSD) (for cholesterol multilayers and stigmasterol controls) over three simulation trajectories of the z-component of the position of all O atoms at the interface layer as a function of time for frames every 0.01 µs, with error bars denoting s.e.m. g, Normalized interaction forces between hydrophobic AFM tips and cholesterol SCL or thiocholesterol SAM surfaces. Each curve contains 16 data points recorded sequentially at the same location on the substrate. Individual datasets were acquired at different locations on the substrate. Curves are normalized to the average of each dataset. Data were obtained from at least three independent experiments.

Source data

Entropic repulsion effects were previously reported to occur either due to the pinned conformational flexibility of grafted oligo(ethylene glycol)15 and poly(ethylene glycol) brushes16 or enforced smoothening of height fluctuations of lipid bilayers17. Herein, entropic repulsion of cholesterol SCLs is suggested to be mechanistically very different, primarily governed by polarity-controlled orientational fluctuations: The entropic penalty for the suppression of orientational fluctuations of cholesterol molecules at the SCL interface on protein adsorption can be estimated by relating the cross-sectional area of protein-facing cholesterol at the SCL interface to the area of orientationally free cholesterol at the SCL–solution interface (Fig. 3b and Supplementary Note 3)18,19. The obtained value for this entropic penalty (–160 J mol–1 K–1) is close to the experimentally determined entropic repulsion barrier derived from temperature-dependent protein adsorption experiments (–200 J mol–1 K–1). This rather large entropy penalty per adsorbing lysozyme molecule, corresponding to about three hydrogen bonds, emphasizes the relevance of the uncovered repulsive effect. In this estimation of entropic penalty, three associated interfacial cholesterol molecules were assumed to be cooperatively involved in orientational fluctuations, which was reasoned at first by geometrical considerations of cholesterol molecular packing in SCLs. Time-dependent interaction force measurements (Fig. 2a,b) showing a third-order kinetics support this assumption because the data indicate cooperative interactions involving three cholesterol molecules (Fig. 3b).

Analysis of adsorption experiments with the larger proteins albumin and fibrinogen yielded lower values for the estimated repulsive entropic barrier (probably due to entropy effects of conformational changes; Supplementary Note 2). However, it similarly showed a negative (repulsive) entropy contribution of cholesterol SCLs indicating the general relevance of the newly identified antiadhesion mechanism.

In an entirely independent approach, cholesterol multilayers in contact with water were explored by molecular dynamics (MD) simulations and similarly showed the relevance of molecular orientational fluctuations (Supplementary Note 4). Using equilibrium multilayer systems and steered MD simulations with reverted orientation of a variable percentage of interfacial cholesterol molecules, we observed spontaneous reorientation of these molecules within 1 µs simulation time such that their hydroxyl groups aligned again toward the water layer (Fig. 3c,d and Extended Data Fig. 6b,d). Importantly, control simulations using the cholesterol analogue stigmasterol (Fig. 4) did not show this spontaneous molecular reorientation (Fig. 3c,d and Extended Data Fig. 6b,d). The orientation correlation function of cholesterol molecules in the interfacial layer indicated a rapid decay over a distance of 1 nm (Fig. 3e), confirming orientational cooperativity of a few (two or three) cholesterol molecules as discussed above. Together with the observation of high vertical fluctuation of the interfacial layer of cholesterol multilayers (Fig. 3f), the MD simulations support the view that cholesterol SCLs represent a highly dynamic and fluctuating molecular interface.

Fig. 4: SCLs of cholesterol analogue molecules show graded bioadhesion.
figure 4

Bioadhesion to SCLs of cholesterol analogues (polar hydroxyl groups shown in red). To facilitate the comparison, a bioadhesion index was introduced by aggregating the results of bacterial adhesion experiments with E. coli and S. epidermidis cells and protein adsorption experiments with bovine serum albumin, lysozyme and fetal bovine serum (for the full set of experimental results see Extended Data Fig. 10), all normalized to the corresponding results for cholesterol SCLs. The reference value was calculated using the respective controls from bacterial adhesion (SiO2) and protein adsorption tests (gold). AU, arbitrary units.

Source data

To obtain direct experimental evidence of orientational fluctuations of cholesterol at SCL interfaces, we applied AFM-based force mapping to detect the simultaneous exposure of polar and non-polar residues of immersed cholesterol SCLs (Extended Data Fig. 7). The low and high levels of interaction forces between a hydrophobic AFM tip and thiocholenic acid or thiocholesterol SAMs span the range of interaction forces detected for cholesterol SCLs, confirming that both polar and non-polar residues are exposed at the interface of immersed cholesterol SCLs (Extended Data Fig. 7c). Repeated force–distance measurements at a constant position showed substantially higher variations for cholesterol SCLs than for thiocholesterol SAMs, providing proof of orientational fluctuations of cholesterol at the SCL interface (Fig. 3g and Extended Data Fig. 8).

Bioadhesion to layers of cholesterol analogues

The discovered entropic barrier to bioadhesion resulting from orientational fluctuations at the interface of cholesterol-containing SCLs raises the question of what key structural features of cholesterol underlie this effect.

Therefore, we compared the bioadhesion characteristics of SCLs prepared from structurally related cholesterol analogues (Fig. 4 and Extended Data Fig. 9) in terms of bacterial adhesion and protein adsorption (Extended Data Fig. 10). The results confirmed that molecular amphiphilicity is essential—but not sufficient—for low bioadhesion (Fig. 4 shows the results of different assays aggregated into a bioadhesion index) and showed that even small deviations in the molecular structure of cholesterol can strongly diminish its antiadhesive performance. Only SCLs of dehydrocholesterol, the closest chemical relative of the tested cholesterol analogues, approached the bioadhesion index of cholesterol SCLs.

Our results show that cholesterol assembles in molecular arrangements that are capable of entropically restricting bioadhesion. The combination of the spatially decoupled amphiphilicity of cholesterol and its effective alignment in multilayers, generating a slowly adaptive, cooperative interfacial orientational mobility of the assemblies, was identified to be the prerequisite of pronounced entropic repulsion.

The resting-time-dependent decrease in the receding water contact angle indicates that, among the set of cholesterol analogues compared, only cholesterol and dehydrocholesterol SCLs meet these criteria (Extended Data Fig. 9d). Fluorescence recovery after photobleaching (FRAP) measurements on cholesterol SCLs providing lateral diffusion coefficients of 0.3–0.4 µm2 s–1 further support the ordered, but slowly adaptive, characteristics of the system, comparable to smectic liquid crystals and high-cholesterol-content lipid membranes (Supplementary Fig. 6)25,26.

In agreement with the experimental results, MD simulations show that SCLs of analogues with slight structural modifications with respect to cholesterol (Extended Data Fig. 9d) behave significantly differently, with no spontaneous back-rotations of reverted molecules within the simulation time (Fig. 3d and Extended Data Fig. 6b,d). The simulations also indicate that stigmasterol SCLs are subject to significantly lower interfacial fluctuations than cholesterol SCLs, suggesting that the specific intermolecular interactions of cholesterol provide high interfacial mobility and allow for spontaneous molecular fluctuations (Fig. 3f and Extended Data Fig. 6f). These findings strongly support the experimentally deduced criteria for amphiphilic molecules capable of generating entropically repulsive assemblies.

Exploring the compositional line of defence of the antiadhesive Collembola cuticula, we identified physicochemical features of cholesterol-containing layers that result in effective entropic repulsion. Given the widespread occurrence of cholesterol at tissue boundaries, our findings may shed new light on important and ubiquitous interfacial processes of living matter. The development of cholesterol-inspired synthetic and scalable materials to control bioadhesion is highly attractive but demands further analyses of the molecular requirements that govern this newly identified mechanism.



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.

TOC measurements

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.

QCM measurements

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.

Topographic imaging

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.

MD simulations

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).