FCS Analysis of Protein Mobility on Lipid Monolayers

Abstract

In vitro membrane model systems are used to dissect complex biological phenomena under controlled unadulterated conditions. In this context, lipid monolayers are a powerful tool to particularly study the influence of lipid packing on the behavior of membrane proteins. Here, monolayers deposited in miniaturized fixed area-chambers, which require only minute amounts of protein, were used and shown to faithfully reproduce the characteristics of Langmuir monolayers. This assay is ideally suited to be combined with single-molecule sensitive fluorescence correlation spectroscopy (FCS) to characterize diffusion dynamics. Our results confirm the influence of lipid packing on lipid mobility and validate the use of FCS as an alternative to conventional surface pressure measurements for characterizing the monolayer. Furthermore, we demonstrate the effect of lipid density on the diffusional behavior of membrane-bound components. We exploit the sensitivity of FCS to characterize protein interactions with the lipid monolayer in a regime in which the monolayer physical properties are not altered. To demonstrate the potential of our approach, we analyzed the diffusion behavior of objects of different nature, ranging from a small peptide to a large DNA-based nanostructure. Moreover, in this work we quantify the surface viscosity of lipid monolayers. We present a detailed strategy for the conduction of point FCS experiments on lipid monolayers, which is the first step toward extensive studies of protein-monolayer interactions.

Introduction

Biological membranes have been a predominant focus of biophysical research in the last few decades. Highly complex in their organization, as an interplay between numerous lipid and protein partners, biological membranes are not only a physical barrier between cellular compartments but also directly or indirectly play a fundamental role in several key cellular mechanisms. To facilitate the study of complex membrane-associated phenomena under defined and controlled conditions, a variety of minimal model membrane systems have been developed (1).

From the available in vitro membrane model systems, support-free model membranes are especially attractive as additional interactions with the support can strongly influence the studied behaviors (2, 3). Lipid vesicles, black lipid membranes, suspended lipid bilayers and lipid monolayers are some of the most common free-standing model membranes, with each of them bearing particular limitations. Small unilamellar vesicles and large unilamellar vesicles, with diameters smaller than 1 μm, are mainly used to study protein-membrane interactions in which curvature plays a significant role in the binding (e.g., (4, 5)). In contrast, giant unilamellar vesicles (GUVs) with diameters larger than 10 μm are quasi-planar, can be produced at high yields under several salt conditions and are stable over long periods of time (6). Black lipid membranes also do not have an intrinsic curvature but often contain an undefined amount of residual solvent trapped within the lipid leaflets, which is linked to the preparation protocol (7). Solvent-free suspended lipid bilayers, on the other hand, can be formed from the rupture of lipid vesicles but have very limited sizes, up to only a few μm (e.g., (8)).

To study lipid-protein interactions, it is generally desirable to vary a plethora of membrane conditions, including lipid packing density and mobility. In cells, both are strongly related to the local lipid composition and lateral organization (9, 10, 11). In vitro, lipid mobility can potentially be controlled through external factors such as membrane composition, ambient bulk viscosity, temperature, and ionic strength (12, 13, 14). However, all these factors generally also affect the protein behavior itself, which renders the interpretation of experiments challenging. More importantly, the effect of lipid packing is not accessible by typical model membranes. An elegant system to study its effect is the Langmuir lipid monolayer, in which a single layer of lipids, deposited on an air-water interface, is mechanically compressed to tune both lipid packing and lipid mobility (15). Typically, experiments on lipid monolayers are performed in Langmuir-Blodgett troughs that require large quantities of precious sample volume (of order of tens of milliliters per sample), which usually precludes them from studies on proteins and peptides. With the recent introduction of miniaturized chambers (16), lipid monolayers are of easy handling and preparation, and require considerably smaller amounts of protein, making them compatible with recombinant proteins, which are typically produced on small lab scales. Notably, although lipid monolayers lack essential features of biological membranes, complex biological networks (such as a minimal actin cortex (17, 18), the dynamic behavior of pattern-forming Min proteins, and the assembly of FtsZ filaments (19)) have been successfully reconstituted using this model system.

Changes in lipid packing were proposed to influence the diffusion behavior of protein components in the membrane (15). However, measuring the lateral diffusion of molecules in lipid monolayers has been to date restricted to lipids, which were studied to a minor degree only (15, 16, 20, 21). Here, to quantify mobilities in lipid monolayers, we used fluorescence correlation spectroscopy (FCS) (22). In this method, a fluorescence time trace is recorded from a small detection volume, typically a confocal volume (23, 24). Signal fluctuations originating from, e.g., diffusion of fluorescent species through the detection volume are analyzed by means of autocorrelation. The characteristic decay time of the autocorrelation curve is directly linked to the hydrodynamic properties of the fluorescent molecules, whereas the fluorescence intensity is proportional to the average number of molecules in the detection volume. Thus, knowing the size and shape of the confocal volume, estimated through a calibration procedure, and applying an appropriate model function, one gains direct access to the diffusion coefficient D of a given fluorescent species. The amplitude G₀ of the autocorrelation curve scales with the inverse number N of particles in the confocal volume.

In this study, we demonstrate the use of confocal point FCS to study protein mobilities in lipid monolayers. We used miniaturized chambers to measure hitherto unknown diffusion coefficients of proteins on lipid monolayers and correlated the results with the lipid packing and mobility. Furthermore, we characterized the compatibility of several membrane-binding molecules with the lipid monolayer system.

Materials and Methods

Chemicals

The lipids 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (DOGS-NTA(Ni)), ovine brain ganglioside G_M1, Escherichia coli polar lipid extract, were purchased from Avanti Polar Lipids (Alabaster, AL). ATTO655 and ATTO488 head labeled 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) were purchased from ATTO-TEC (Siegen, Germany). Lipid mixtures were prepared in high purity chloroform (Merck KGaA, Darmstadt, Germany) and their concentration was determined by gravimetry.

Bovine serum albumin was purchased from Sigma-Aldrich (Taufkirchen, Germany). Labeled cholera toxin β (Alexa Fluor 488) was purchased from Invitrogen (Carlsbad, CA). The membrane proximal external region (MPER) of the envelope glycoprotein gp41 of HIV-1, namely the peptide Atto488CELDKWASLWNWF (underscored sequence corresponds to aa 662–673 by HXBc2 numbering), which presumably dimerizes through a disulfide bond, was purified by the Biochemistry Core Facility of the Max Planck Institute of Biochemistry with degree of purity >90%. The Biochemistry Core Facility of the Max Planck Institute of Biochemistry also purified the MinD, MinE (25), and eGFP-MinD (26) proteins according to the reported protocols. Ramm et al. developed the construct and purification protocol for the chimeric fluorescent protein mCherry carrying the membrane targeting sequence (Mts) of the protein MinD from Bacillus subtilis (mCherry-Mts) (B. Ramm, P. Glock, J.M., P. Blumhardt, M. Heymann, and P.S., unpublished data). Purified mNeonGreen was kindly provided by Magnus-Carsten Huppertz, Max Planck Institute of Biochemistry (Martinsried, Germany).

Q buffer (10 mM HEPES, 150 mM NaCl, pH 7.4) was used for most described measurements. M buffer (25 mM Tris-HCl, 150 KCl, 5 mM MgCl₂, pH 7.5) was used when working with Min proteins or mCherry-Mts constructs (Table S1). For DNA origami, D buffer (5 mM Tris-HCl, 1 mM EDTA, 5 mM MgCl₂, 300 mM NaCl, pH 8.0) was used.

DNA origami folding and purification

The elongated DNA origami structure described in (27) was used. Two variations were produced: unmodified (N) and cholesterol (Chol)-modified (X5) DNA nanostructure. For X5, the oligonucleotides in the bottom positions A0, A4, B2, C0, and C4 (Fig. S1) were extended with an 18 nucleotide sequence complementary to the 5′ TEG-Chol modified oligonucleotide AACCAGACCACCCATAGC (Sigma-Aldrich). For detection by fluorescence microscopy and spectroscopy, both N and X5 were functionalized by 3 × 5′ ATTO488-modified oligonucleotides GGGTTTGGTGTTTTTT (Eurofins, Planegg, Germany), positioned on the top facet close to the center of the structure. Folding, purification, and quantification of DNA nanostructures was performed as previously reported (27).

Monolayer preparation in miniaturized chambers

Customized miniaturized chambers inspired by (16) were manufactured by laser cutting a 15 mm diameter hole into a 5 mm-high polytetrafluoroethylene (PTFE) sheet (Fig. S2). Before every use, the PTFE spacers were cleaned in a series of sonication steps (30 min each) in acetone, chloroform, isopropanol, and ethanol. A chamber was completed by gluing a #1.5 cover glass (Menzel Gläser, Braunschweig, Germany) to the bottom of the PTFE spacer using picodent twinsil 22 two component glue (picodent, Wipperfuerth, Germany). Directly before use, the miniaturized chamber was thoroughly rinsed with distilled milliQ water and 99% ethanol, dried under airflow and plasma cleaned (MiniFlecto-PC-MFC; plasma technology, Herrenberg-Gültstein, Germany) for 10 min to make the glass hydrophilic. The cleaned chambers were filled with 200 μL of aqueous buffer. The lipid mixture (0.1 mg/mL, containing 0.01 mol% of ATTO655-DOPE or ATTO488-DOPE) was deposited drop-by-drop on the air-buffer interface (Fig. 1 A; see also (16)) to reach the desired lipid density. If required, 20–40 μL of the aqueous phase were pipetted out after complete evaporation of chloroform to adjust the interface position. The miniaturized chamber was covered with a cover slip and sealed with grease. For addition of biomolecules to the system, biomolecule solution was pipetted into the aqueous phase of known volume to reach the desired concentration.

Figure 1.

Stabilization of the lipid monolayer positioning by using an active temperature control. (A) Schematic representation of lipid monolayer deposition on an air-water interface is shown. A known amount of lipids dissolved in chloroform was deposited drop-by-drop on the air-water interface. If necessary, a small volume (20–40 μL) of the aqueous subphase was pipetted out to bring the monolayer within the working distance of the used objective. (B and C) Representative intensity traces (upper panels) and corresponding autocorrelation curves (lower panels) of 0.01 mol% Atto655-DOPE in DMPC obtained by point FCS without and with active temperature (T) control, respectively, are shown. The effect of the axial drift on the autocorrelation curves is highlighted in (B), whereas the autocorrelation curves obtained with T control (C) are indistinguishable. (D) Monolayer positions over time without and with active T control are shown. Circles and crosses correspond to independent time series. (E) Surface pressure (Π) measured for DMPC monolayers deposited in miniaturized chambers (MC) with a fixed area at 21°C (circles) and 30°C (squares) is shown. The average of typically four independent samples and respective standard deviations are shown. As a reference, the DMPC Langmuir isotherm (LB) at 21°C is shown (black line). To see this figure in color, go online.

FCS and confocal imaging

Confocal imaging was performed on a laser-scanning microscope (LSM780; Zeiss AG, Oberkochen, Germany), equipped with gallium arsenide phosphide detectors and a water immersion objective with a long working distance of 620 μm (LD C-Apochromat 40X, NA 1.1; Zeiss AG). The monolayer interface was located by imaging the back-reflection of the excitation laser. For presentation purposes, the brightness and contrast of images was adjusted by using ImageJ software (28).

FCS measurements were performed on the same microscope stand using avalanche photodiode detectors (ConfoCor3; Zeiss AG). The internal pinholes were set to 35 and 45 μm for 488 and 633 nm excitation wavelength, respectively. To circumvent detector afterpulsing, we performed pseudo-cross correlation, i.e., we split the collected fluorescence, projected it on two independent avalanche photodiode detectors, and cross correlated their signals. The optical system was calibrated on a daily basis by using Alexa Fluor 488 (Alexa488; Thermo Fischer, Waltham, MA) or ATTO 655 carboxylic acid (Atto655; ATTO-TEC), freely diffusing with known diffusion coefficients (29, 30, 31) in aqueous solution, corrected for the respective temperature through the relation $D\left(T\right)\propto T/\eta \left(T\right)$ (32, 33). The viscosity of water $\eta \left(T\right)$ at any temperature was calculated (34). In brief, the confocal volume was positioned 50 μm above the bottom cover slip, the lateral pinhole position was optimized for maximal fluorescence signal and the objective’s correction collar was positioned for maximal count rate per particle, and finally the FCS measurement was taken.

For FCS measurements on lipid monolayers, the optimal axial focus position was determined by scanning the volume in the axial direction to locate the intensity maximum. This procedure was repeated in between FCS measurements because of sample drift. For both the initial calibration measurement and the FCS measurements on lipid monolayers, the irradiance was chosen sufficiently low to minimize photobleaching and fluorescence saturation (35, 36, 37).

To control the temperature, the miniaturized chambers were placed in a heating system (Ibidi, Martinsried, Germany) compatible with the commercial microscopy stage.

The recorded correlation curves were analyzed by using PyCorrFit 0.9.7 (38). The used fitting function reads as follows:

$$G\left(\tau \right)=N^{-1}\left(1+\frac{T}{1-T}{e}^{-\raisebox{1ex}{$\tau $}\!\left/ \!\raisebox{-1ex}{${\tau }_T$}\right.}\right){\left(1+\tau /{\tau }_D\right)}^{-1}{\left(1+\tau /\left(S^2{\tau }_D\right)\right)}^{-\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}\text{.}$$ (1)

Here, ${\tau }_D=w_0^2/4D$ is the diffusion time, which depends on the lateral e⁻²-value of the confocal detection volume. Moreover, we introduced the structure parameter S, which represents the ratio of axial/lateral extent of the detection volume, the triplet fraction T, and the triplet decay time τ_T. Atto655 does not show triplet blinking and experimental curves were therefore fitted with T=0. Moreover, when measuring on membranes, the axial extent of the confocal volume is irrelevant (S=∞).

Langmuir compression isotherms

Compression isotherms were measured using a Kibron Micro-Trough XL Langmuir-Blodgett trough equipped with a dyne probe and the analytical software FilmWareX 4.0 (Kibron, Helsinki, Finland). Before every measurement, the trough was thoroughly cleaned by three washing steps with Kimtech paper tissues soaked with chloroform and ethanol. Powder-free gloves were used to avoid any contaminations. The dyne probe was cleaned by flaming it with a butane torch. The instrument was calibrated daily and the surface pressure (Π) zeroed in the aqueous subphase before every measurement. To verify the subphase purity, an isotherm was recorded in absence of lipids with a compression rate of 5 cm²/min. Lipids were deposited on the air-water interface from a 1 mg/mL stock solution in high purity chloroform. After complete evaporation of chloroform, the isotherm was recorded with a compression rate of 5 cm²/min until the monolayer collapsed. Isotherms were measured at least in duplicate at room temperature (21°C).

Surface pressure measurements in miniaturized chambers

For Π measurement in the miniaturized chambers, we used the dyne probe and detection system described above. After system calibration (see above), Π was zeroed in a miniaturized chamber filled with 200 μL of Q buffer before every measurement. Lipids were deposited on the interface as described above. After full solvent evaporation (∼5 min), the resulting Π value was recorded. We measured Π at room temperature (21°C) and at 30°C. For measurements at 30°C, the miniaturized chamber was placed on a hot plate together with tissue soaked in water and was covered by a petri dish to achieve a humidity-saturated environment (Fig. S3). A small hole in the petri dish ensured accessibility for the dyne probe. Under these settings, evaporation was negligible.

Determination of the interface area in miniaturized chambers

A monolayer of defined packing was deposited in a miniaturized chamber and imaged with a Zeiss Plan Apo 10X/0.45 objective (Zeiss AG). We acquired several adjacent tile images of the interface to image the entire cross section of the miniaturized chamber (R = 7.5 mm). This procedure was repeated in 19 different z-planes, each of them 100 μm apart. A circle was imaged in each z-plane, corresponding to the section of the meniscus with the confocal plane (Fig. S4 B). The center of mass of each circle was determined and the intensity values were plotted versus their distance to this center. To reduce the noise, the radial distance was binned (bin width five pixels). The resulting radial intensity distribution was baseline corrected and fitted by a Gaussian, which is centered around the radius of the circle. Based on the known nominal focus position and the determined radii, we determined the radial meniscus profile h(r) (Fig. S4 C), which was extrapolated to the physical size of the chamber R = 7.5 mm. This function is well behaved and the corresponding meniscus area A was calculated numerically: $A=2\pi \underset{0}{\overset{R}{\int }}r\sqrt{1+{\left|\partial h\left(r\right)/\partial r\right|}^2}dr$.

Results and Discussion

Temperature control improves FCS performance on lipid monolayers

Here, we used a miniaturized chamber with a fixed area (16) to study biomolecule mobility on lipid monolayers by FCS. The use of such miniaturized chambers has the major advantage that considerably smaller amounts of lipids and proteins are required compared to conventional Langmuir-Blodgett troughs. Because of the fixed-area of the chamber, we controlled the lipid density by the amount of lipid deposited on the air-water interface (Fig. 1 A). To avoid lipid oxidation by carbon-chain exposure at the air-water interface, we chose to use the fully saturated lipid DMPC.

When performing FCS on two-dimensional systems such as lipid membranes, the z position of the confocal volume needs to be accurately adjusted to the membrane position because axial mismatches between both bias the obtained number of particles N and diffusion coefficient D (39). For monolayers, it is particularly challenging to keep the confocal volume centered on the monolayer. Upon evaporation of the subphase, the air-water interface is lowered, leading to a reduction of the autocorrelation amplitude G₀ and thus an increase in N detected (Eq. 1; Fig. 1 B). Thus, when working with lipid monolayers, the evaporation of the subphase, because of the high surface area/volume ratio is a major concern. Moreover, almost all confocal setups feature a temperature above the ambient temperature, mainly because of active elements hosted in the microscope body. Consequently, the monolayer chamber is exposed to a temperature gradient in which the bottom cover slip is warmer than the top lid of the chamber. Therefore, water condensation occurs on the top lid of the chamber, preventing the gas phase above the monolayer from reaching a humidity-saturated state. The resulting permanent axial drift of the monolayer with respect to the confocal volume renders long high-quality FCS measurements almost impossible and wastes valuable measurement time because the operator constantly needs to refocus on the lipid monolayer.

We avoided this major bottleneck by actively heating the sample and its surroundings to a constant temperature (T = 30°C) above room temperature. By this strategy, we stabilized the lipid monolayer position and eliminated the artifacts arising from axial drift (Fig. 1, C and D). The stable positioning of the focus allowed us to measure for considerably longer periods of time and thus to access long correlation times, e.g., due to low diffusion coefficients (40, 41). Gudmand et al. had previously taken advantage of the natural subphase evaporation to apply a modification of z-scan FCS (39) to determine lipid mobility in monolayers (15). In their approach, each series of intensity trace measurements started with the lipid monolayer above and finished below the fixed focus position such that the maximal autocorrelation amplitude and counts per particle could be found. Although this approach is simple and elegantly makes use of the inherent evaporation, it comes at the cost of very long measurement times (30 min). In contrast, our approach of focus stabilization and point FCS analysis maximizes the counts per particle because the lipid monolayer is constantly in focus and thus reduces the total measurement time per sample considerably. Nonetheless, both approaches are expected to yield identical results (42).

Lipid diffusion coefficient is an effective tool to characterize the monolayer state

To confirm the quality of the obtained lipid monolayers, we measured the surface pressure (Π) of monolayers deposited at different mean molecular areas per molecule (MMA, in Å²) in the miniaturized chambers (see Materials and Methods; Fig. S2). Because the area of the chamber is constant, the MMA is controlled by the amount of lipids deposited on the air-water interface. We compared the Π-A dependence from the miniaturized chambers with the corresponding Langmuir Π-A isotherm (Fig. 1 E). Since the air-water interface forms a meniscus, its area is larger than the cross section of the chamber, in particular in the miniaturized chambers. As a result, the actual MMA is larger than predicted from the amount of lipid deposited, which directly reflects in an overall lower Π. It would be desirable to apply a correction factor taking into account the real interface area of the meniscus in the miniaturized chambers. However, whereas the meniscus shape in capillaries (i.e., cavity radius much smaller than the height of the liquid) has been subject to theoretical studies (43, 44), no analytical expression is known for a cylindrical well structure as used in this study. Additionally, the contact angle of aqueous buffer and plasma cleaned PTFE is unknown. In this study, we estimated the actual interface area by imaging the entire monolayer at different lipid packings (Fig. S4). Interestingly, the lipid packing of the monolayer had only a very minor impact on the meniscus shape. Relative to the miniaturized chamber cross section πR², the monolayer area was (4 ± 1)% larger at 90 Å²/molecule. As at 50 Å²/molecule, the meniscus shape was similar, we corrected all MMAs by a factor of 1.05.

The compression isotherm of DMPC behaved as previously described (15). The monolayers deposited in the fixed-area chambers follow the general trend of the Langmuir isotherm and confirm the reproducibility of the deposition protocol of the lipid monolayers in fixed-area chambers. We attribute the small discrepancies at low MMA to the difference between the two assays in the physical process of increasing lipid packing. In a Langmuir monolayer, the lipids rearrange upon slow physical compression, whereas in the fixed-area chambers, lipid molecules need to incorporate and find their arrangement during the much faster process of lipid spreading on the interface upon organic solvent evaporation. Consequently, when depositing low MMA lipid monolayers, a fraction of the lipid molecules may not insert into the monolayer, resulting in an effective increase of the MMA.

The measurement of Π in our miniaturized chambers although conceptually simple (Fig. S3), is rather impractical when combined with confocal microscopy at temperature-controlled conditions. Instead, we used confocal microscopy and FCS to monitor the lipid monolayer quality. DMPC lipid monolayers deposited at 30°C were homogeneous between 50 and 100 Å²/molecule (Fig. S5). The FCS curves obtained for DMPC monolayers with different MMAs showed an increased N with higher lipid packing (Fig. 2, A and B). Moreover, the measured N is inversely proportional to the estimated MMA, as highlighted in Fig. 2 B by a fit of the inverse proportionality. However, the determined number of particles is consistently lower than the theoretically predicted number of particles.

Figure 2.

Characterization of DMPC lipid monolayers deposited in fixed-area chambers at 30°C by point FCS. (A) Representative autocorrelation curves, single-component diffusion fit, and the respective residuals obtained by point FCS for DMPC lipid monolayers deposited at different MMAs in fixed area-chambers are shown. The monolayers were doped with 0.01 mol% Atto655-DOPE. The increase of autocorrelation amplitude (G₀) with increasing MMA is highlighted and corresponds to the decrease in the number of particles N in the confocal volume. (B) Number of particles N obtained for lipid monolayers at different MMA by fitting a single-component diffusion model (Eq. 1) is shown. Measurements were performed on pure DMPC, and 2 mol% and 6 mol% content of G_M1. Overall N follows the trend of an expected 1/MMA-dependence (black line). The theoretically predicted N is shown by the gray line. (C) Normalized autocorrelation curves for DMPC monolayers deposited at different MMA are shown. The shift of the autocorrelation curves to smaller diffusion times (τ_D) and correspondingly larger diffusion coefficient (D) with increasing MMA is highlighted. (D) Lipid diffusion coefficient varies linearly with lipid monolayer MMA. Measurements were performed on pure DMPC, 2 and 6 mol% content of G_M1. The lines are linear fits (Table S2) for each data set. The obtained critical areas (a_c) can be found in Table S2. The average of typically four independent samples and respective standard deviations are shown in (B) and (D). To see this figure in color, go online.

Another observed feature was the shift of the correlation curves toward larger diffusion times with increasing lipid packing, resulting in slower lipid mobility, as previously shown (15, 16) (Fig. 2, C and D). As described by the free-area model, D varied linearly with the lipid density (15, 16, 45). As surface pressure Π is a monotonic function of the MMA, there is a monotonic relationship between D and Π (15). Consequently, the cumbersome measurement of Π in fixed-area chambers can be replaced by an FCS measurement of lipid diffusion to characterize the current state of the monolayer.

When analyzing the linear dependence of D on the MMA (Table S2), we obtained the intersection of the linear extrapolation to D = 0 μm²/s, which yields an estimate for the critical area a_c, below which the translational diffusion of lipids in a fluid monolayer can theoretically no longer occur. The obtained a_c of (33.0 ± 8.0) Ų/molecule for DMPC is in good agreement with previously reported values (15, 16, 20).

Protein mobility is modulated by lipid monolayer packing

Lipid Langmuir monolayer studies generally suggest that protein/peptide adsorption and potential insertion into lipid membranes changes lateral pressure and lipid packing (e.g., (21, 46, 47)). Therefore, when studying the effect of lipid packing, careful controls need to be performed to ensure that the addition of biomolecules at a certain concentration does not change the macroscopic properties of the monolayer. Moreover, it has been proposed that the mobility of protein components in membranes is influenced by lipid packing (15). To further investigate this hypothesis, we studied the influence of lipid packing, and consequently of lipid mobility, on the mobility of monolayer-bound biomolecules using point FCS.

First, the binding of molecules to lipid monolayers through head-group specific interaction or the insertion of a hydrophobic moiety was tested (Table S1). Independently of the nature of the interaction with the lipid monolayer, the studied molecules can be divided into two groups: homogeneously distributed or aggregated at the lipid interface. Importantly, the presence of protein clusters precludes quantitative FCS measurements. The aggregates have statistically ill-defined size and brightness distributions, distorting both amplitude and shape of the autocorrelation curve (Eq. 1; Fig. S6). Under special circumstances, the effect of aggregates can be corrected by postprocessing of the photon-arrival times (48, 49). However, even these approaches may only deal with the effect of very bright particles passing through the center of the confocal volume. Although passivation strategies to reduce unspecific interactions with the interface and consequent protein aggregation could be conceived, these are of low relevance as the effective MMA of the monolayer is modified. Generally, it is thus advisable to spend considerable efforts to prevent the formation of or to remove aggregates.

We found that the model protein CtxB falls into the category of proteins that do not aggregate at the lipid monolayer interface and is thus suitable for analysis by point FCS. Previously, binding of CtxB to lipid monolayers has been qualitatively assessed in phase-separated lipid mixtures (16). However, to date, the diffusion behavior of CtxB at the lipid monolayer has not been studied. Here, we varied the density of the lipids at the interface and studied its effect on the mobility of CtxB (Fig. 3). To eliminate the influence of protein binding on lipid diffusion, low protein concentrations (≤10 nM) were used. Indeed, the lipid diffusion in the monolayer does not change upon addition of CtxB, as highlighted by the perfectly superimposed autocorrelation curves (Fig. 3 A; Fig. S7). Consequently, protein binding in these conditions did not change the surface pressure Π. Furthermore, the use of a low protein concentration allowed us to use a simple single-component diffusion model (with a triplet component, in accordance with the used dye molecule) (Eq. 1) to fit the obtained autocorrelation curves for the protein channel because there was virtually no contribution from protein in solution to the autocorrelation curve.

Figure 3.

Cholera toxin β (CtxB) interaction with DMPC lipid monolayers in absence or presence of the specific ligand G_M1 analyzed by point FCS. (A) Autocorrelation curves obtained by point FCS for a DMPC monolayer at 70 Å²/molecule with 0.01 mol% Atto655-DOPE, before and after addition of 10 nM CtxB-Alexa488, and for the protein CtxB-Alexa488 are shown. Fits with single diffusion component and respective residuals are shown. (B) Autocorrelation curves obtained for CtxB (10 nM) bound to DMPC monolayers at 90 and 50 Å²/molecule. Fits with single diffusion component and respective residuals are shown. (C) Shown is the relationship between lipid diffusion and CtxB diffusion on a pure DMPC monolayer (magenta) and in the presence of 2 mol% or 6 mol% of the specific ligand G_M1. The diffusion coefficients of CtxB and lipids show a linear relation (Table S3) as highlighted by the linear fit for a pure DMPC monolayer. (D) Shown is the relationship between the detected CtxB number of particles N bound to the lipid monolayer and the lipid MMA in presence and absence of the ligand G_M1. The average of typically three independent samples and respective standard deviations are shown. To see this figure in color, go online.

As expected, the autocorrelation curve obtained for CtxB was shifted to larger decay times compared to the lipid monolayer curves, indicating a slower protein diffusion (Fig. 3 A). The decrease in lipid MMA, and consequent reduction of lipid mobility, again resulted in a shift of the autocorrelation curves of CtxB to larger diffusion times (Fig. 3 B). The CtxB diffusion coefficient (D_CtxB) scaled linearly with the lipid diffusion in the monolayer; less dense lipid monolayers allowed for a faster diffusion of the protein (Fig. 3 C). A similar linear dependence on D_DOPE was observed for another membrane-targeted molecule, the MPER of the envelope glycoprotein gp41 of HIV-1 (Fig. S8). Relative to D_DOPE, D_CtxB was approximately 60% lower, whereas D_MPER was only 40% lower (Fig. 4).

Figure 4.

Comparison of the mobility of biomolecules of different dimensions on lipid monolayers at 70 Å²/molecule by point FCS. (A) Shown are autocorrelation curves obtained by point FCS for 0.01 mol% Atto655-DOPE, 10 nM of the membrane proximal external region (MPER) of the envelope glycoprotein gp41 of HIV-1 and G_M1-bound CtxB, both labeled with Alexa488, and 40 pM DNA structure X5 three-fold labeled with Atto488 in a DMPC monolayer at 70 Å²/molecule. Fits with single diffusional component and respective residuals are shown. (B) Diffusion coefficients obtained for each analyzed biomolecule in a DMPC monolayer at 70 Å²/molecule (filled bars). The studied biomolecules cover a range of sizes and number of membrane insertion points. The average of typically three independent samples and respective standard deviations are shown. Diffusion coefficients determined by others for Atto655- DOPE (23.5°C) (42), G_M1-bound CtxB (23.5°C) (77), and DNA structure X5 (27.5°C; A.K., J.M., Henri G. Franquelim, and P.S., unpublished data) in 1,2-dioleoyl-sn-glycero-3-phosphocholine free-standing lipid bilayers are shown for comparison (shadowed bars). (C) Surface viscosity η_s of DMPC monolayers based on FCS experiments on CtxB (open symbols, Fig. 3C) and ATTO655-DOPE (gray line with 95% confidence intervals, Fig. 2D; Table S2) are shown. η_s was obtained numerically by finding the zero between predicted and measured diffusion coefficients using Newton’s method. As expected, η_s increases with increasing lipid packing. Empirically, the η_s roughly follows a bi-exponential: ${\eta }_S=a_1\exp \left(-a_2\cdot \text{MMA}\right)+a_3\exp \left(-a_4\cdot \text{MMA}\right)$ with $a_1=1.2\cdot {10}^{-8}\text{Pa}\text{s}\text{m}$, $a_2=0.104{\text{Å}}^{-2}$, $a_3=1.7\cdot {10}^{-10}\text{Pa}\text{s}\text{m}$, $a_4=0.025{\text{Å}}^{-2}$. All values correspond to 30°C. To see this figure in color, go online.

Next, we analyzed the influence of a specific ligand, the ganglioside G_M1, on the diffusion behavior of CtxB. Supposedly, the pentameric CtxB binds up to five G_M1 as each monomer exhibits a binding site with a reported dissociation constant K_D = 0.1–1 nM (50, 51, 52). As all experiments were conducted above K_D, CtxB should prevalently bind to its ligand when G_M1 is present in the membrane. Interestingly, the addition of G_M1 does not significantly influence the diffusion coefficient D_CtxB. In theory, the two-dimensional diffusion of a molecule correlates with the size of the molecule’s inclusion in the membrane (12). One can thus hypothesize that the similar diffusion coefficients obtained in presence and absence of the ligand G_M1 indicate that the insertion size of the pentameric CxtB nonspecifically bound to the lipid monolayer is similar to the effective insertion size of the lipid group codiffusing upon binding of the pentameric CtxB to five G_M1 molecules.

Although the diffusion coefficients are similar, the analysis of the amplitude of CtxB correlation curves unravels a significant difference in CtxB binding to the lipid monolayer in presence and absence of G_M1 (Fig. 3 D). In absence of G_M1, CtxB binding to DMPC monolayers is dependent on the lipid packing, with higher binding observed at low lipid density. This can be explained by the affinity of CtxB to the air-water interface that is shielded at higher lipid densities (Fig. S9). In presence of G_M1, on the other hand, CtxB binds stronger to the lipid monolayer at low MMA, as the total amount of G_M1 present in the lipid monolayer is inversely proportional to the lipid packing. For 100 Å²/molecule, a larger scatter of N was obtained in both conditions, as G₀ is particularly sensitive to small density variations and occasional protein clusters were encountered at the monolayer.

Although these are the first steps toward the analysis of protein-lipid interactions on monolayers, the role of specific ligands in the modulation of protein behavior should be subject to future studies. Additionally, as the lipid density has a strong impact on the diffusion of membrane interacting proteins/peptides, the rate of chemical reactions in the membrane may also be sensitive to lipid packing. Furthermore, the single molecule sensitivity of FCS enables the detection and quantification of protein/peptide interaction with lipid monolayers in a regime in which no diffusion coefficient variation and, consequently, no surface pressure variation is detected.

Hydrodynamic length scale is different in lipid monolayers and bilayers

When inspecting the diffusion coefficients discussed above, it becomes clear that the surface viscosity η_s of the lipid monolayer is considerably lower than the surface viscosity of a bilayer, in good agreement with previous studies (53, 54, 55). Specifically, the diffusion coefficients of lipids and CtxB are at least threefold larger in the lipid monolayer (Fig. 4) depending on the MMA. Accordingly, the characteristic hydrodynamic length of the system $l={\eta }_s/\left({\mu }_1+{\mu }_2\right)$ (56, 57) is shorter in lipid monolayers than bilayers. Here, we introduced the bulk viscosities μ₁ and μ₂ below and above the membrane, respectively. Assuming that the monolayer has half the thickness of a bilayer and that the viscosity of air is negligible compared to the viscosity of water, a lower viscosity of the lipid monolayer compared to the lipid bilayer is reasonable. Moreover, in lipid monolayers, no interleaflet coupling occurs and the packing density is lower, especially at high MMA.

For a more quantitative approach, we estimated the surface viscosity η_s of DMPC lipid monolayers at different MMAs based on the measured diffusion coefficients. Notably, the major models used to describe diffusion in lipid membranes, more specifically the Saffman-Delbrück model (12) and the Hughes-Pailthorpe-White model (HPW model) (56), are weakly dependent on the size of the membrane insertion. Thus, having in hand a reasonable estimation of the membrane inclusion radius r should yield a good approximation of η_s. As the Saffman-Delbrück model becomes inapplicable at high MMA, the values have been estimated numerically by using an empirical expression for the HPW model (57). The estimation was based on the determined D_CtxB assuming the previously reported radius r = 3.1 nm of pentameric CtxB (58). Overall, the η_s of DMPC lipid monolayers decreases with increasing MMA as expected (Fig. 4 C). In detail, the η_s of DMPC ranges from 1 × 10⁻¹⁰ Pa s m at 50 Å²/molecule to 2 × 10⁻¹¹ Pa s m at 100 Å²/molecule. The estimated values are relatively low compared to previously reported data (53, 54, 59, 60, 61). However, these reported values scatter considerably and it has been previously discussed that the surface viscosity was frequently overestimated (53). Interestingly, the surface viscosity at 62 Å²/molecule, which is around the expected MMA for DMPC bilayers (62), compared to 70 Å²/molecule changes only by around 31%. Consequently, all larger differences in diffusion coefficients between lipid monolayers and bilayers can be mainly attributed to interleaflet coupling and the surrounding environments. Gudmand et al. made a similar observation when they found that a DMPC monolayer needs to have an MMA as small as 50 Å² to yield the same diffusion coefficients in bilayers and monolayer (15). Moreover, the estimation of ηs based on D_CtxB is surprisingly in good agreement with the estimation based on D_DOPE (Fig. 2 D). With an assumed r = 0.36 nm (15, 20, 63, 64, 65, 66), the size of the lipid probe is not much larger than the lipids themselves and thus violates a key assumption of the HPW model. The corresponding hydrodynamic length scale l of DMPC lipid monolayers will range from 120 nm at 50 Å²/molecule to 24 nm at 100 Å²/molecule, in comparison to estimated 250 nm for lipid bilayers assuming η_s = 5 × 10⁻¹⁰ Pa s m (67, 68, 69). The smaller hydrodynamic length scale of the lipid monolayer implies a slightly larger sensitivity of the lipid monolayer system to size variations on relevant length scales (12, 57).

To test this hypothesis, we studied the interaction of a relatively large synthetic DNA origami-based nanoparticle (70). For simplicity, we chose a flat three-dimensional DNA origami structure (Fig. S1), which was previously studied on free-standing lipid bilayers (27). The bare nanostructure (denoted N), which has no affinity to GUVs in the used buffer conditions, showed no enrichment at the lipid monolayer interface and therefore did not alter the lipid diffusion (Fig. S10, A and B). The highly charged nature of such DNA nanostructures is likely to be the cause of this low affinity to the air-water interface. Interestingly, we also did not observe significant structure clustering.

Next, we functionalized the DNA origami structure with five cholesterol (Chol)-modified oligonucleotides (structure X5) to directly compare its diffusion behavior to that of pentameric CtxB. As a result, the DNA nanostructure did bind to the lipid monolayer (Fig. S10 C), which is in line with the previously shown binding to GUVs (27). As in the case of CtxB and MPER, although of considerably larger dimensions, structure X5 did not influence the lipid mobility upon binding to the lipid monolayer (Fig. S10 D). The autocorrelation curve obtained for the structure X5 decayed at larger diffusion times than CtxB, corresponding to a smaller diffusion coefficient (Fig. 4). As no significant binding has been observed with structure N, we did not expect a full insertion of X5 into the lipid monolayer. X5 is rather likely binding to the lipid monolayer by the insertion of the five Chol-modified oligonucleotides into the lipid monolayer, gliding on the interface as proposed for lipid bilayers (71).

Both the X5 structure and pentameric CtxB bound to G_M1 have five membrane anchors of comparable insertion sizes. In free-standing bilayers of similar nature, their diffusion coefficients differ by a factor of 2.2. Interestingly, their diffusion coefficients measured in DMPC monolayers of 70 Å²/molecule differ by a factor of 3. This small but significant difference between lipid bilayers and monolayers is consistent with the discussed difference in hydrodynamic length scale. Moreover, having determined the η_s of the lipid monolayer (3.8 × 10⁻¹¹ Pa s m at 70 Å²/molecule), we can now estimate the inclusion size for both MPER and X5. For MPER, r = 0.8 nm, in good agreement with estimations based on the alignment of MPER in lipid bilayers (72). For X5, r = 28.3 nm, which is larger than the combined inclusion of five cholesterol anchors. For a more quantitative understanding, the compatibility of current membrane diffusion models with lipid monolayers needs to be addressed in future studies.

We note that the fit to the autocorrelation curve of the structure X5 shows systematic residuals up to 4% of the amplitude, which is considerably larger than for the other studied biomolecules. However, we do not expect any contribution from rotational diffusion to the autocorrelation curve, as the fluorescent labels are located close to the center of the rod (71, 73). On the other hand, when recording autocorrelation curves at different concentrations of nanostructure X5, the autocorrelation function shifted toward larger diffusion times at higher X5 concentration (Fig. S10 E). A similar behavior was previously described on GUVs for long DNA nanoneedles and has been attributed to crowding effects on particle diffusion at high surface densities (73). Thus, lipid monolayers appear to also support the study of more complex diffusional behaviors by point FCS.

Conclusions

In this work, we have established conditions under which reproducible, long, high-quality point FCS measurements can be performed on lipid monolayer systems. Most importantly, we discussed the necessity to reach an equilibrated system in which the lipid monolayer stays at a constant height to minimize focus drift. To reach this state, it is necessary to heat the closed monolayer chamber to the working temperature or above, in our case 30°C. Provided the biomolecule of interest is exclusively located at the lipid monolayer, the throughput of protein-monolayer studies could even be increased by camera detection schemes similar to, e.g., refs (74, 75, 76), coupled with wide-field illumination. In a set of proof-of-principle experiments, we applied FCS to study the lateral diffusion of selected biomolecules at the lipid monolayer. The use of FCS can not only complement canonical Π measurements to characterize the monolayer state but also allows the quantitative characterization of protein-lipid monolayer interactions in a regime in which the physical properties of the monolayer are not modified. To cover a wide range of sizes and membrane insertion points of diffusing particles, we studied not only the diffusion of the rather small peptide MPER but also the model protein CtxB and a large DNA origami construct. We showed that the viscosity and consequently the hydrodynamic length scale in lipid monolayers are smaller than in lipid bilayers. Furthermore, we investigated the effect of lipid packing on protein diffusion in the lipid monolayer and found a linear dependence between the diffusion coefficient of bound protein and the diffusion coefficients of the lipids themselves. The direct impact of lipid packing on the mobility of monolayer-associated biomolecules may have implications on intermolecular reaction rates. We believe that this study forms a basis for novel research on the effects of lipid packing on protein-monolayer interactions.

Author Contributions

A.K. and J.M. contributed to the project design, conducted experiments, analyzed data, and wrote the manuscript. F.C. conducted experiments and wrote the manuscript. G.C. and P.S. contributed to the project design and wrote the manuscript.

Editor: Claudia Steinem.