A technique to probe the lipid raft partitioning of fluorescent proteins at the plasma membrane of living cells is described. It takes advantage of the disparity in diffusion times of proteins located inside or outside of lipid rafts. Acquisition can be performed dynamically in control conditions or after drug addition.
In the past fifteen years the notion that cell membranes are not homogenous and rely on microdomains to exert their functions has become widely accepted. Lipid rafts are membrane microdomains enriched in cholesterol and sphingolipids. They play a role in cellular physiological processes such as signalling, and trafficking1,2 but are also thought to be key players in several diseases including viral or bacterial infections and neurodegenerative diseases3.
Yet their existence is still a matter of controversy4,5. Indeed, lipid raft size has been estimated to be around 20 nm6, far under the resolution limit of conventional microscopy (around 200 nm), thus precluding their direct imaging. Up to now, the main techniques used to assess the partition of proteins of interest inside lipid rafts were Detergent Resistant Membranes (DRMs) isolation and co-patching with antibodies. Though widely used because of their rather easy implementation, these techniques were prone to artefacts and thus criticized7,8. Technical improvements were therefore necessary to overcome these artefacts and to be able to probe lipid rafts partition in living cells.
Here we present a method for the sensitive analysis of lipid rafts partition of fluorescently-tagged proteins or lipids in the plasma membrane of living cells. This method, termed Fluorescence Correlation Spectroscopy (FCS), relies on the disparity in diffusion times of fluorescent probes located inside or outside of lipid rafts. In fact, as evidenced in both artificial membranes and cell cultures, probes would diffuse much faster outside than inside dense lipid rafts9,10. To determine diffusion times, minute fluorescence fluctuations are measured as a function of time in a focal volume (approximately 1 femtoliter), located at the plasma membrane of cells with a confocal microscope (Fig. 1). The auto-correlation curves can then be drawn from these fluctuations and fitted with appropriate mathematical diffusion models11.
FCS can be used to determine the lipid raft partitioning of various probes, as long as they are fluorescently tagged. Fluorescent tagging can be achieved by expression of fluorescent fusion proteins or by binding of fluorescent ligands. Moreover, FCS can be used not only in artificial membranes and cell lines but also in primary cultures, as described recently12. It can also be used to follow the dynamics of lipid raft partitioning after drug addition or membrane lipid composition change12.
1. Calibration of the FCS Setup
2. Staining of Living Cells with Lipid Rafts Marker
3. FCS Data Acquisition on Living Cells
4. FCS Data Analysis
5. Representative Results
An example of an FCS calibration with a cholera toxin-Alexa488 solution is shown in Figure 3. After checking that individual measures of fluorescence as a function of time did not show any photobleaching (Figure 3A), individual and mean FCS curves were calculated. Mean FCS curves were fitted with equations corresponding to various diffusion models (examples in Figure 2). The parameter classically considered to determine the quality of a fit is the coefficient of determination R2. The closer R2 is to 1, the better the fit. In this case, the most accurate model to fit the mean FCS curve is the one describing a population of fluorescent molecules diffusing freely in three dimensions (equation 1 in Figure 2 and Figure 3B). The diffusion time derived from the fit is 0.32 ms. Residuals from curve-fitting (Figure 3C) and R2 factor (0.99906) give an estimate of the quality of the fit.
An example of FCS analysis for cholera toxin-Alexa488 stained HEK293 cells is shown in Figure 5. The multiphasic mean FCS curve shape reveals the existence of populations of fluorescent molecules with different diffusion times. The best fit for this curve corresponds to a model with three populations of fluorescent probes: two with an hindered diffusion (diffusion in two dimensions as in the membrane plane) and one freely diffusing in three dimensions (equation 2 in Figure 2 and Figure 5). This latter population corresponds to fluorescent molecules moving outside of the membrane plane, i.e., either binding or unbinding to their membrane targets, reaching the membrane through the secretion or recycling pathway, or leaving the membrane by endocytosis. The two diffusion times at the membrane, corresponding to cholera toxin bound to GM1, were 2 ms (25% of molecules), corresponding to diffusion outside of lipid rafts, and 75 ms (50% of molecules), corresponding to diffusion in lipid rafts. Please note that any photobleaching during acquisition will lead to artificial longer diffusion times thus possibly creating a bias towards localization of GM1 in lipid raft domains.
Figure 1. Schematic representation of the FCS set-up (picture modified from Marquer et al.12)
Figure 2. Example of diffusion models and corresponding equations used to fit autocorrelation curves. The structure parameter S can be written as S = z0/w0 with z0 the effective focal radius along the optical axis at 1/e2 intensity and w0 the effective lateral focal radius at 1/e2 intensity. These values can be extracted from a classical point spread function (PSF) measurement.
Figure 3. Assessment of diffusion time of cholera toxin-Alexa488 in solution for FCS calibration. A) Fluorescence fluctuation as a function of time for a representative example of a 30 seconds acquisition. B) Mean autocorrelation curve obtained from 10 samples of 30 seconds acquisition fitted with equation 1 (see Figure 2).C) Residuals from curve fitting.
Figure 4. HEK-293 cell stained with cholera toxin-Alexa488. Cells were imaged on an SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany) with the internal 488nm laser line. Fluorescence was collected with a x60 plan apochromat oil immersion objective between 500 and 650 nm.
Figure 5. Assessment of diffusion time of cholera toxin-Alexa488 at the plasma membrane of HEK293 cells. A) Mean autocorrelation curve fitted with equation 2 (see Figure 2). B) Residuals from curve fitting. (modified from Marquer et al.12)
The FCS method presented here enables a sensitive and rapid analysis of the lipid raft partitioning of fluorescent probes of interest in living cells. FCS combines the accuracy of localization of confocal microscopy with the sensitivity of single photon counting. The main difference between FCS and standard biochemical techniques is that FCS enables the absolute determination of the lipid rafts partition of the target and not the relative partition as is the case for DRMs isolation or co-patching.
Acquisition of FCS data takes approximately 5 minutes (10 acquisitions of 30 seconds) for each sample and is thus quite rapid as compared to usual biochemical techniques. This rapidity makes it possible to follow in time the perturbation in lipid raft partitioning that may result from drug addition. However, analysis of autocorrelation curves can be a bit tricky as the different models for fitting have to be browsed through to determine the most accurate one. Moreover, photobleaching during acquisition has to be avoided as this may lead to artifacts.
Here we describe an example where we can delineate the lipid raft partitioning of a lipid: ganglioside GM1 (Fig. 5). Such a study would not have been possible with standard biochemical procedures which can only be applied to proteins. FCS is not limited to lipids though and can also be used for proteins, either tagged with a fluorescent ligand (e.g. transferrin-Alexa555 binds to the transferrin receptor, known to be located outside of rafts9) or expressed as a fusion with a fluorescent protein (e.g. APP-YFP or Bace1-GFP9, two key protein players in Alzheimer’s disease). It can thus be used for a wide range of targets. Further, FCS can be implemented not only in cell lines, as described here, but also in primary cultures9.
Concerning cholera toxin, you should keep in mind that it has a tendency to aggregate GM1 in pentamers, thus creating membrane micro-domains that can be different from native lipid rafts. That is why you cannot use colocalisation with cholera toxin to determine if your protein of interest is inside or outside of rafts. Nevertheless, membrane proteins, such as APP-YFP and Bace1-GFP, diffuse in lipid rafts with diffusion times of 60-80 ms9, very similar to what was observed with cholera toxin (75 ms). Thus cholera toxin can be used to calibrate your set-up and check that you can differentiate between fast and slow diffusion times.
FCS is also suited for many other applications13,14 such as determination of the oligomerisation state of a protein15 or ligand/receptor pharmacological studies16. It can also be coupled to other microscopy techniques such as Total Internal Reflection Fluorescence (TIRF)17,18 or STimulated Emission Depletion (STED)19. Indeed, STED-FCS allows an even more accurate determination of lipid raft partitioning19 but up to now, STED microscopy is still costly and thus not widespread.
The main limits of FCS are the need for low concentrations of fluorophores (in the nanomolar range) and fast diffusion times so that fluorophores will not photobleach before leaving the excitation volume. Complementary techniques to assess the diffusion of proteins include FRAP (Fluorescence Recovery After Photobleaching) and ICS (Image Correlation Spectroscopy) techniques (reviewed in 20).
In FRAP, a high intensity laser beam is used to photobleach a region of a cell and the recovery of fluorescence after photobleaching is then monitored. This recovery comes from the diffusion of fluorophores from non-bleached areas to the bleached area. Thus the diffusion times can be extracted from the recovery kinetics. FRAP can be used with high concentrations of fluorophores and can access large range of diffusion times, thus making it complementary to FCS. FRAP has been used to assess whether a fluorophore was majoritarily inside or outside of rafts21,22 but it is still a rather bulk method to quantify the proportion of fluorophores diffusing inside or outside rafts in the area of interest.
ICS is an imaging analog of FCS, in which spatial correlation is calculated pixel by pixel for a given image23. It has the advantage to be amenable to the study of slowly diffusing fluorescent probes but is limited to 2D analysis. This limit was overcome by the development of ICS termed Spatio-Temporal ICS (STICS)24. STICS enables spatio-temporal correlation of fluorescent probes on a stack of images and is thus a powerful technique though it requires a lot of computational implementation and has a lower time resolution than FCS. In fact, both ICS and STICS are limited to process much slower than the acquisition time of one image frame. Another extension of ICS called raster ICS (RICS)25,26 enables the analysis of fast diffusion process by taking advantage of the raster scan mode available on most commercial confocal microscopes. RICS is versatile as it can be used for a large diffusion range (from μs to ms) but its implementation (scanning parameters, data processing) can be cumbersome. We did not find many papers in the literature reporting the use of ICS and its derived techniques to study lipid rafts27-29.
The authors have nothing to disclose.
This work was supported by a grant from Agence Nationale de la Recherche (ChoAD). We are also grateful to the Fondation ICM (Institut du Cerveau et de la Moelle) for their financial support.
Name of the reagent | Company | Catalogue number | Comments |
Cholera toxin subunit B-Alexa 488 | Invitrogen | C-34775 | MW (pentamer) = 57 kg/mol |
Confocal microscope | Leica | SP5 | |
Incubator for temperature and CO2 control | Life imaging services | The Cube and the Box | |
SPAD (Single Photon Avalanche Diode) | MPD (Micro Photon Devices) | PDM serie (100 μm sensitive area) | |
High pass 488 nm filter | Semrock | 488 nm blocking edge BrightLine long-pass filter Part # FF01-488/LP-25 |
|
FCS detection unit | Picoquant | Picoharp 300 module | |
Acquisition and auto-correlation software | Picoquant | SymPhoTime | |
Fitting software | OriginLab | OriginPro8 |